Datasets:

agentic-health

/

s2orc-PubMed_processed

like 1

Follow

Agentic Health, Inc. 3

https://www.semanticscholar.org/paper/6d563db62d7cf50dc560bc15b904e002987bf7a2

Exploiting probability density function of deep convolutional autoencoders’ latent space for reliable COVID-19 detection on CT scans

Journal of Supercomputing

26ed29a9-64ce-4d6c-9024-8b022fd2fe22

Exploiting probability density function of deep convolutional autoencoders’ latent space for reliable COVID-19 detection on CT scans We present a probabilistic method for classifying chest computed tomography (CT) scans into COVID-19 and non-COVID-19. To this end, we design and train, in an unsupervised manner, a deep convolutional autoencoder (DCAE) on a selected training data set, which is composed only of COVID-19 CT scans. Once the model is trained, the encoder can generate the compact hidden representation (the hidden feature vectors) of the training data set. Afterwards, we exploit the obtained hidden representation to build up the target probability density function (PDF) of the training data set by means of kernel density estimation (KDE). Subsequently, in the test phase, we feed a test CT into the trained encoder to produce the corresponding hidden feature vector, and then, we utilise the target PDF to compute the corresponding PDF value of the test image. Finally, this obtained value is compared to a threshold to assign the COVID-19 label or non-COVID-19 to the test image. We numerically check our approach's performance (i.e. test accuracy and training times) by comparing it with those of some state-of-the-art methods. # Introduction There is general agreement on chest computed tomography (CT) scans having a potential role in diagnosing COVID-19 and being particularly effective when used as a complement to polymerase chain reaction (PCR) testing [bib_ref] The role of computed tomography scan in the diagnosis of COVID-19 pneumonia, Axiaq [/bib_ref] [bib_ref] Thoracic imaging tests for the diagnosis of COVID-19, Islam [/bib_ref] [bib_ref] Jicheng Xie et al (2020) Sensitivity of chest CT for COVID-19: comparison..., Fang [/bib_ref] [bib_ref] Hongyan Hou et al (2020) Correlation of chest CT and RT-PCR testing..., Ai [/bib_ref] [bib_ref] Chest CT for typical coronavirus disease 2019 (COVID-19) pneumonia: relationship to negative..., Xie [/bib_ref]. Clearly, cheaper testing methods exist, like lateral flow test. Moreover, the radiation dose inflicted to the patient by a scan may be harmful. However, the method can be useful when a different test is unavailable or too expensive or when a quick diagnosis is needed. Due to its noteworthy detection capabilities, deep learning (DL) is often employed to assist the evaluation of CT scans for the diagnosis of COVID-19. Indeed, classification of chest CT scans is an important and active research area. In recent contributions, classification is carried out by exploiting supervised DL, using only a few classes of chest CTs (for instance, COVID-19 class and normal one). However, the symptoms of chest diseases manifest themselves in a broad spectrum of visual characteristics. Hence, supervised-trained models could run into trouble when they are fed with chest CTs that do not belong to any of the classes used in the training phase. This issue can be addressed by taking advantage of unsupervised DL, where the neural network models are trained only on data sets of the COVID-19 class. In this manner, the trained model can differentiate the COVID-19 class (the target class) from any other types of chest images (anomalies). This paper develops an unsupervised classification approach based on autoencoders (AEs). An AE is composed of an encoder and a decoder. The encoder takes an image and transforms it into a hidden feature vector, a compressed input representation. The decoder is used to reconstruct the input image from the hidden feature vector. Since deep convolutional neural networks (CNNs) have proved their effectiveness in image processing and feature extraction, in this paper, we design a deep convolutional autoencoder (DCAE) as the neural network model [bib_ref] Stacked convolutional auto-encoders for hierarchical feature extraction, Masci [/bib_ref]. We train the proposed DCAE, in an unsupervised manner, on a data set of chest CT scans obtained from COVID-19 patients. After training, the encoder is used to obtain the hidden feature vectors of the training set CT scans. These hidden feature vectors are, in turn, exploited to estimate the probability density function (PDF) of COVID-19 hidden feature vectors, by means of the multivariate kernel density estimation (KDE) method. The classification of a test image is performed by feeding it into the trained encoder to produce its hidden feature vector. Afterwards, the PDF is used to compute the PDF value of the test image. Finally, the obtained PDF value is compared to a suitably tuned threshold to classify the test image as either COVID-19 or non-COVID- [bib_ref] A histogram-based low-complexity approach for the effective detection of covid-19 disease from..., Scarpiniti [/bib_ref]. For comparison, we also consider a second method for classification based on the DCAE reconstruction error. The resulting error is noticeably higher than the average error corresponding to the training set instances when the trained DCAE is fed with a CT scan of a non-COVID-19 case and attempts to recover it. If this error is below a suitably tuned threshold, the image is classified as COVID-19, otherwise as non-COVID- [bib_ref] A histogram-based low-complexity approach for the effective detection of covid-19 disease from..., Scarpiniti [/bib_ref]. The rest of the paper is organised as follows: In Sect. 2, we review the related work and present the contributions of this paper. In Sect. 3, we provide details of 1 3 the used data sets and pre-processing tasks. Sections 4 and 5 describe the proposed DCAE architecture and classification approach, respectively. We discuss the obtained numerical results in Sect. 6 and compare the performance of our approach to the ones of some state-of-the-art approaches that rely on the supervised training method. Finally, in Sect. 7, we point out the summarised observations and provide some hints for future research. ## Related work and paper contributions The literature on the application of DL-based algorithms to the detection of COVID-19 is vast. The small volume of available data on COVID-19 patients has motivated the researchers to take this deficiency into account. For instance, the transfer learning approach is adopted in [bib_ref] A deep learning algorithm using CT images to screen for Corona Virus..., Wang [/bib_ref] [bib_ref] Classification of positive COVID-19 CT scans using deep learning, Muhammad Attique Khan [/bib_ref] [bib_ref] Seyed Mohammad Sakhaei (2021) A fully automated deep learning-based network for detecting..., Rahimzadeh [/bib_ref] [bib_ref] Classification of COVID-19 by compressed chest CT image through deep learning on..., Zhu [/bib_ref] [bib_ref] The ensemble deep learning model for novel COVID-19 on CT images, Zhou [/bib_ref] [bib_ref] Corodet: a deep learning based classification for COVID-19 detection using chest X-ray..., Hussain [/bib_ref] [bib_ref] Diagnosing COVID-19 pneumonia from X-ray and CT images using deep learning and..., Halgurd S Maghdid [/bib_ref] to deal with the lack of large-size data sets. In [bib_ref] Application of deep learning technique to manage COVID-19 in routine clinical practice..., Ali Abbasian Ardakani [/bib_ref] , the authors utilise GoogleNet and ResNet for supervised COVID-19 classification. The authors of [bib_ref] A histogram-based low-complexity approach for the effective detection of covid-19 disease from..., Scarpiniti [/bib_ref] take a statistical method to address issues like huge computational complexity and large datasets required by deep networks. The adopted approach is based on the evaluating and comparing the statistical representation of medical images. The authors of [bib_ref] Deep supervised learning using self-adaptive auxiliary loss for COVID-19 diagnosis from imbalanced..., Hu [/bib_ref] consider an unbalanced-data supervised algorithm and obtained good results comparable with benchmark architectures. A large number of research papers adopt supervised learning approaches [bib_ref] Classification of COVID-19 pneumonia from chest CT images based on reconstructed super-resolution..., Tan [/bib_ref] [bib_ref] COVID-Net: a tailored deep convolutional neural network design for detection of COVID-19..., Wang [/bib_ref] [bib_ref] Detection of coronavirus disease (COVID-19) based on deep features and support vector..., Prabira Kumar Sethy [/bib_ref] [bib_ref] CovXNet: a multi-dilation convolutional neural network for automatic COVID-19 and other pneumonia..., Mahmud [/bib_ref] [bib_ref] An accuracy vs. complexity comparison of deep learning architectures for the detection..., Sarv Ahrabi [/bib_ref] [bib_ref] Deep learning enables accurate diagnosis of novel coronavirus (COVID-19) with CT images, Song [/bib_ref] [bib_ref] Convolutional capsnet: a novel artificial neural network approach to detect COVID-19 disease..., Suat Toraman [/bib_ref] [bib_ref] Deep learning for detecting corona virus disease 2019 (COVID-19) on high-resolution computed..., Yang [/bib_ref]. In [bib_ref] Classification of COVID-19 pneumonia from chest CT images based on reconstructed super-resolution..., Tan [/bib_ref] , the authors consider a binary classification problem and apply the off-theshelf VGG-16. In [bib_ref] CovXNet: a multi-dilation convolutional neural network for automatic COVID-19 and other pneumonia..., Mahmud [/bib_ref] , the authors use depth-wise convolutions with varying dilation rates to extract more diversified features of chest images. The authors use a pretrained model and reach the overall 90.2% accuracy. In, the authors design a neural network model as a combination of convolutional and capsule layers, called COVID-FACT. Despite their great effort, the considered model achieves the 90.82% accuracy. The authors of [bib_ref] Deep learning enables accurate diagnosis of novel coronavirus (COVID-19) with CT images, Song [/bib_ref] propose a model based on the pre-trained ResNet50 and achieve the accuracy, rather similar to the original ResNet50. The DenseNetbased approach, considered in [bib_ref] Deep learning for detecting corona virus disease 2019 (COVID-19) on high-resolution computed..., Yang [/bib_ref] , achieves 92% accuracy. The work by [bib_ref] Deep learning for covid-19 diagnosis from ct images, Loddo [/bib_ref] compares different classification architectures on a specific dataset to discover the most suitable real-world scenarios. The approaches developed in all these papers typically deal with only a limited class of chest images. Hence, a naturally posed question is: to which category does a CT fit when it does not belong to any of the classes learned by the supervisedtrained models? In any case, the aforementioned supervised methods must be trained on both COVID-19 and non-COVID-19 CTs. Lastly, we provide an overview of AE-based approaches. The authors of [bib_ref] Stacked-autoencoder-based model for COVID-19 diagnosis on CT images, Li [/bib_ref] consider a stacked AE (SAE), composed of four CAEs, followed by a dense layer, and a final softmax classifier. Each layer is equipped with regularisation to improve the local optimum. The binary classification task, considered by the authors, occurs at the last stage, where the softmax classifier obtains the probability of the two types of labels and performs the classification task with an average accuracy of 94.7% . This method is different from our approach because we perform the classification by directly using the PDF of the hidden feature vectors, without inserting any classifier at the top of our model. The deep convolutional denoising AE, proposed in [bib_ref] A chest X-ray image retrieval system for COVID-19 detection using deep transfer..., Layode [/bib_ref] , is trained on COVID-19, pneumonia, and a few other types of chest X-rays. Then, the hidden feature vector of a test image is compared to the features of the selected training data sets. The considered AE exhibits good performance. However, unlike our work, this approach relies on training the considered model over each selected class and therefore cannot detect chest CTs except those that belong to the classes of training data sets. The work by [bib_ref] CoroNet: a deep network architecture for enhanced identification of COVID-19 from chest..., Agarwal [/bib_ref] focuses on a two-stage learning method and a triple classification task. The authors train their considered model on classes of COVID-19, pneumonia, and normal cases separately. Once the hidden feature vectors of all classes are independently obtained, a feature classifier is employed and trained-in a supervised manner-to detect each decision class. The considered approach reaches a quite good accuracy of 93.50% . In contrast to this work, we train our DCAE model on only one class, i.e. the COVID-19. The paper by [bib_ref] Unsupervised deep learning based variational autoencoder model for COVID-19 diagnosis and classification, Mansour Romany [/bib_ref] is based on a variational autoencoder (VAE) model for COVID-19 classification. The VAE model involved adaptive Wiener filtering (AWF)-based pre-processing technique to enhance the image quality. Besides, Inception v4 with Adagrad technique is employed as a feature extractor and unsupervised VAE model is applied for the classification process. As the last research paper, the method introduced by [37] builds a robust statistical generating target histogram of the deep denoising convolutional autoencoder's (DDCAE) latent vector. It then estimates the statistical distance between unknown and target histograms to classify the images according to proper thresholds. A brief overview of the given literature is provided in [fig_ref] Table 1: A synoptic overview of main related work on COVID-19 detection/classification [/fig_ref]. Finally, the main contributions of this paper are listed below: - we base the classification task on exploiting an unsupervised deep neural network model, which is trained only on COVID-19 images and, in so doing, we strengthen the robustness of the proposed model concerning the presence of CT scans of other diseases which are not seen during training phase; - we propose an ad hoc DCAE with an optimised number of layers for the best classifying test performance; - as an additional novelty, we base the classification task on the estimation of the probability density of the training hidden feature vectors by adopting the KDE method; and finally, - we carry out numerical tests under benchmark data sets available in the literature and compare the performance of our approach to the ones of some supervised/ unsupervised deep neural networks, both in terms of test accuracy and processing times. ## Utilised data sets and pre-processing The training data set used in this paper is composed of 4000 CTs of COVID-19 cases collected from over 500 patients. These CT scans have been selected from the 'COVIDx CT-2A' data set; the 'A' variant of the '2nd' version of the open-source 'COVIDx-CT' data set [bib_ref] COVIDNet-CT: a tailored deep convolutional neural network design for detection of COVID-19..., Gunraj [/bib_ref]. The 'COVIDx CT-2A' data has been validated by medically qualified personnel. The data are split into training and validation sets containing 80% and 20% of instances, respectively. - Transfer learning Ref. [bib_ref] Seyed Mohammad Sakhaei (2021) A fully automated deep learning-based network for detecting..., Rahimzadeh [/bib_ref] - ResNet50V2/Xception + Softmax 0.9849 - BC (Normal, COVID) - Transfer learning Ref. [bib_ref] The ensemble deep learning model for novel COVID-19 on CT images, Zhou [/bib_ref] - AlexNet/GoogleNet/ResNet + Softmax 0.9905 - MC (Normal, COVID, Tumour) - Transfer learning Ref. [bib_ref] Diagnosing COVID-19 pneumonia from X-ray and CT images using deep learning and..., Halgurd S Maghdid [/bib_ref] - Modified AlexNet 0.9410 - BC (Normal, COVID) - Transfer learning 0.9410 Ref.- DensNet-121 0.8711 - MC (Normal, COVID, Cancer) - Transfer learning Ref. [bib_ref] Application of deep learning technique to manage COVID-19 in routine clinical practice..., Ali Abbasian Ardakani [/bib_ref] - AlexNet/GoogleNet/VGG-16 0.9951 - MC (COVID, Atypical/Viral Pneumonia) - Transfer learning Ref. [bib_ref] Deep supervised learning using self-adaptive auxiliary loss for COVID-19 diagnosis from imbalanced..., Hu [/bib_ref] - Deep CNN 0.9943 - MC (Normal, COVID, Bacterial Pneumonia) Ref. [bib_ref] Classification of COVID-19 pneumonia from chest CT images based on reconstructed super-resolution..., Tan [/bib_ref] - SRGAN/VGG-16 0.9800 - BC (Normal, COVID) Ref. [bib_ref] Detection of coronavirus disease (COVID-19) based on deep features and support vector..., Prabira Kumar Sethy [/bib_ref] - Deep CNN + SVM 0.9866 - MC (Normal, COVID, Bacterial Pneumonia) Ref. [bib_ref] Deep learning enables accurate diagnosis of novel coronavirus (COVID-19) with CT images, Song [/bib_ref] - ResNet50 + FPN 0.9300 - MC (Normal, COVID, Bacterial Pneumonia) Ref. [bib_ref] Deep learning for detecting corona virus disease 2019 (COVID-19) on high-resolution computed..., Yang [/bib_ref] - DenseNet 0.9500 - BC (Normal, COVID) Ref. [bib_ref] Stacked-autoencoder-based model for COVID-19 diagnosis on CT images, Li [/bib_ref] - SAE + Softmax Classifier 0.9470 - BC (Normal + COVID) Ref. [bib_ref] A chest X-ray image retrieval system for COVID-19 detection using deep transfer..., Layode [/bib_ref] - Denoising AE + Hidden Features -- MLC (COVID, Pneumonia, and other classes) Ref. [bib_ref] CoroNet: a deep network architecture for enhanced identification of COVID-19 from chest..., Agarwal [/bib_ref] - AE + FPN + Classifier - [formula] - MC (Normal, COVID, Pneumonia) [/formula] Ref. [bib_ref] Unsupervised deep learning based variational autoencoder model for COVID-19 diagnosis and classification, Mansour Romany [/bib_ref] - VAE + AWF 0.9920 Ref. [bib_ref] Momenzadeh Alireza (2022) A novel unsupervised approach based on the hidden features..., Michele [/bib_ref] - DDCAE + Latent Space 1.0000 [formula] - (Pneumonia & Normal) 1 3 [/formula] ## Exploiting probability density function of deep convolutional… The test data set comprises 4000 CT slices of COVID-19, normal, pneumonia, and three types of lung cancers, namely adenocarcinoma, large cell carcinoma and squamous cell carcinoma. The test images have been randomly selected from two separate data sets [bib_ref] COVIDNet-CT: a tailored deep convolutional neural network design for detection of COVID-19..., Gunraj [/bib_ref] , composed of CTs of over 500 patients different from those involved in the training set. The data sets include both male and female patients and cover a wide age range. The information about the employed training/test data sets is summarised in [fig_ref] Table 2: Train/validation/test set compositions and related web links [/fig_ref] , where we also give the links to the corresponding web pages. For illustrative purpose, in , we present some instances of various chest CT categories, which are drawn from training and test data sets. As can be seen from the figure, the COVID-19 image is characterised by the presence of several opacities. Indeed, commonly reported imaging features specific of COVID-19 pneumonia are peripheral, bilateral, ground-glass opacities with or without visible intralobular lines: see [bib_ref] Chest CT in COVID-19: what the radiologist needs to know, Thomas [/bib_ref] for a wider discussion. As a pre-processing task, the margin of every CT is cropped before performing the training phase. Moreover, all the images are resized to 300 pixels in width and 200 pixels in height: this size was selected as a compromise between computational complexity and resolution. Two samples of the COVID-19 dataset in the original, cropped and resized versions are demonstrated in ## The proposed dcae architecture A DCAE is composed of an encoder and a decoder as depicted in [fig_ref] Figure 3: The autoencoder architecture training parameters have been selected because they yielded the... [/fig_ref]. The encoder takes an image as input and transforms it into a hidden feature vector called latent space. The decoder takes the hidden feature vector and attempts to recover the input image. The difference between the input and the output images results in the reconstruction error, which is the cost function to minimise during the training phase. We point out that the hidden feature vectors have fewer dimensions than those of the input images since the DCAE produces compressed versions. The DCAE, exploited in this work, has been designed to take account of the tradeoff between test classification accuracy and model complexity. Its "ad hoc" designed architecture is detailed in: the encoder is composed of three convolutional [formula] Training 3200 −−− −−− −−− WP1 1 Validation 800 −−− −−− −−− WP1 Test 2500 300 900 300 WP1 & WP2 2 [/formula] layers, two batch normalisation layers, two max-pool layers, followed by flatten and dense layers with rectifier activation functions (ReLUs). The DCAE is trained to recover its input, i.e. COVID-19 CT scans. As training cost function, we use the mean squared error (MSE), and in order to minimise the considered cost function over the training set, we employ the Adam (Adaptive moment estimation) solver, which is a gradient-based stochastic optimisation algorithm that takes account of the first and second moments of underlying gradients [bib_ref] Adam: a method for stochastic optimization, Diederik [/bib_ref]. The training is carried out over 100 epochs. After the training phase, we select the DCAE weights, which give rise to the minimum squared error on the validation data set.recap details on the adopted model setting. For illustrative purpose, [fig_ref] Figure 4: Numerically evaluated plots of accuracy-vs [/fig_ref] presents the graphs of MSE and accuracy versus the number of epochs, over which the training phase is carried out. Note that we have considered other options for both the architecture and the training parameters, using several alternatives proposed in the literature. The various options have been compared by means of unreported tests: the final architecture and The DCAE is implemented in Python language, using TensorFlow and Keras API. All the numerical trials have been carried out on a PC equipped with an AMD Ryzen 9 5900X 12-Core 3.7 GHz processor, two GeForce RTX 3070 graphics cards, and 128 GB RAM. ## Image classification In this section, we describe the classification approach that is based on the PDF estimation of the DCAE hidden feature vectors. ## Pdf estimation of the training hidden feature vectors In order to estimate the PDF of the hidden feature vectors, a first choice that needs to be done is whether to use a parametric or a nonparametric method. Since we have no clues on the shape of the pdf and we do not want to polarise the estimation by guessing, we decided to use a nonparametric estimate. Among the nonparametric methods, we selected the KDE approach, 1 which is well knownand outperforms the simpler histogram approach. [bib_ref] The role of computed tomography scan in the diagnosis of COVID-19 pneumonia, Axiaq [/bib_ref] The KDE is a traditional density estimation method. Such methods may become unmanageable when the number of variables is too high. Luckily, the computational power of modern computers makes this method suitable for the problem at hand. ## 3 Exploiting probability density function of deep convolutional… The KDE method is based on a univariate kernel function denoted by K(x). Although a wide variety of different kernels can be used, according to, we consider the Gaussian one, i.e. K(x) = e −x 2 (see [fig_ref] a: Normal [/fig_ref]. The kernel is used as an interpolating function to build up the PDF estimate. In order to describe the KDE approach, we first illustrate it for the simple case of a univariate PDF. Let us consider a set of n real numbers: x i for i = 1, ..., n , drawn from a (hidden) random variable (RV) X, which exhibits an unknown PDF, f X (x) , that we want to estimate. The KDE estimate of the PDF is: where the constant is a normalisation factor, used to set the integral of f X (x) to one, while the parameter h is the kernel bandwidth which is used to set the width of the kernel. The estimation is illustrated in [fig_ref] Figure 5: An example of KDE estimation over n = 4 data points [/fig_ref] , where we see that the density is obtained as the superposition of n scaled and shifted copies of the kernel. To describe the multivariate case, let us assume that we have a p dimensional RV X, with a multivariate density f X ( ) where ∈ ℝ p is a p-dimensional vector. Moreover, we have a set of vectors i ∈ ℝ p for i = 1, ..., n which are samples drawn from the RV X. The KDE estimate generalises to the multivariate case as in: where ||.|| denotes the Euclidean norm. In order to estimate the PDF of the DCAE hidden feature vector when the input is a COVID-19 image, we use the KDE estimate of (2), where the vectors i are the hidden feature vectors obtained from the images of the training set, with n = 3200 and p = 128 in our setting. For the computation of the KDE estimate, we used Scikit-learn [bib_ref] Scikit-learn: Machine Learning in Python, Pedregosa [/bib_ref] , an open-source machine learning library developed for the Python environment. One parameter that needs to be carefully tuned is the bandwidth h. Indeed, the choice of bandwidth controls the bias-variance trade-off in the estimation of the PDF. This means that a too narrow bandwidth leads to a high-variance estimate (over-fitting), while a too wide bandwidth results in a high-bias estimate (under-fitting). For selecting the bandwidth, we employed the Scikit-learn built-in method of grid search cross-validation. This algorithm selects the best option from a grid of parameter values provided by the user, automating the 'trial-and-error' method. A second option could be to use the maximum likelihood cross-validation (MLCV) approach, introduced in [bib_ref] Van den Broeck K (1974) A stepwise discriminant analysis program using density..., Habbema [/bib_ref] [bib_ref] On the choice of smoothing parameters for Parzen estimators of probability density..., Duin R P W [/bib_ref]. We point out that the computational complexity of the evaluated KDE-based estimation approach depends on the length p of the hidden feature vectors and the number n, of employed training images. specifically, since n Euclidean distances among p-dimensional vectors must be computed, the resulting computational complexity scales as: [formula] (1) f X (x) = 1 n ∑ i=1 K x − x i h (2) f X ( ) = 1 n ∑ i=1 K || − i || h , [/formula] ## Classification based on the estimated pdf In our approach, the estimated PDF of the DCAE latent space is used to classify the test images. To this end, we feed the image to the DCAE encoder to produce the corresponding hidden feature vector. Next, the obtained vector is used as the argument of the estimated PDF, in order to compute its corresponding PDF value. In practice, since the obtained values of the PDF are minimal, it is more robust to work with log probability densities. If the obtained value of log density is above a (suitably tuned) threshold, the image is classified as a COVID-19 case; otherwise, it is labelled as a non-COVID-19 one. In order to suitably set the decision threshold, we evaluate the real-valued log probability densities of n images in the training set, denoted by l i , for i = 1, 2, ⋯ , n . Then, we compute the mean and the standard deviation of the evaluated l i s, denoted by l and l , respectively. Thus, the threshold is set to: [formula] (3) O(n × p). (4) TH = l + l , [/formula] where the constant is evaluated by using the validation set. In particular, we select so that the threshold equals the minimum log probability of the validation set, namely −233.5 : in this way the whole validation set is classified as COVID-19. The overall proposed classification procedure is summarised in [fig_ref] Figure 6: Estimation of the PDF of the latent space [/fig_ref] and Algorithm 1. ## A benchmark classifier For comparison purposes, we consider a benchmark classification method. To this end, we recall that the DCAE is trained to produce an output similar to the input as much as possible. The difference between the input and the output is the reconstruction error, which minimises the cost function during the training phase. However, since the training set contains COVID-19 scans only, the DCAE is effective at recovering COVID-19 images, but it is ineffective for recovering non-COVID-19 images. Therefore, a high reconstruction MSE can be an index that the image does not belong to the COVID-19 class. By considering this fact, we build up the following benchmark classification procedure. Given an image, we feed it into the DCAE and compute the Euclidean norm of its reconstruction error. If the obtained error norm is below a (suitably) set threshold, the image is classified as COVID-19. The threshold is evaluated by using the instances of validation set. In particular, we select the threshold equal to the maximum error of the validation set, namely 0.08: in this way the whole validation set is classified as COVID-19. presents the flow diagram of the classification process based on reconstruction error. ## Numerical results and performance comparisons In order to evaluate the performance of our classification method, we carried out several tests. We split the presentation into three parts that are: (i) the results that are obtained from the KDE probabilistic approach; (ii) the results that are achieved through reconstruction error evaluation; and (iii) performance comparisons with some state-of-the-art solutions. In the next subsection, we describe the employed performance metrics as a preliminary step. ## Performance metrics Given a binary classifier, the considered performance metrics are the rates of true-positive (TP), true-negative (TN), false-positive (FP) and false-negative (FN) assignments. These metrics are summarised in . These basic metrics can be represented in a compact form as the four elements of the resulting confusion matrix. From these basic metrics, a number of affiliated performance indexes can be derived. This paper will consider accuracy, recall, precision and F1-score, as performance indexes. The formal definitions of these indexes are given in. ## Performance of the proposed approach As a first experiment, we carried out a classification of the test data set by using the proposed approach. To this end, each test image is fed to the trained DCAE, and the corresponding hidden feature vector is obtained. Afterwards, the obtained [fig_ref] Table 1: A synoptic overview of main related work on COVID-19 detection/classification [/fig_ref] False positive (FP) Non-COVID-19 image classified as COVID-19 False negative (FN) COVID-19 image classified as non-COVID-19 hidden feature vector is used as an argument of the multivariate PDF estimated by the KDE and then the corresponding log density value is computed. The soobtained log densities are plotted in [fig_ref] a: Normal [/fig_ref] for all the test images and in [fig_ref] Figure 8: Results under the proposed approach [/fig_ref] for validation set. From [fig_ref] Figure 8: Results under the proposed approach [/fig_ref] , we see that the log densities of the COVID-19 images are almost separated from those of the non-COVID-19 ones. In [fig_ref] a: Normal [/fig_ref] and c, the vertical dashed line denotes the threshold, which is laid between the two classes. As a result, the proposed method achieves a 97.12% test accuracy. The corresponding confusion matrix is presented in [fig_ref] Figure 8: Results under the proposed approach [/fig_ref]. The last row ofreports the evaluation performance metrics. ## Performance of the benchmark classifier As a second experiment, we evaluate the performance of the benchmark classification approach of Sect. 5.3, which is based on the reconstruction error. To this end, in [fig_ref] a: Normal [/fig_ref] , we plot the reconstruction errors obtained by feeding the whole test set to the DCAE, using two different colours for the COVID-19 and non-COVID-19 scans. From the figure, we see that, as expected, the error is lower for the COVID-19 images. However, the two classes are not disjoint. The threshold is shown in [fig_ref] a: Normal [/fig_ref] by the vertical dashed line. The resulting accuracy is equal to 86.35% . The corresponding confusion matrix is presented in [fig_ref] Figure 9: Performance results of the reconstruction error approach [/fig_ref] , and the related performance indexes are reported in. From, we conclude that the obtained performance of the classification approach based on the reconstruction error is about 11% inferior performance than the corresponding one of the KDE-based proposed approach. ## Performance comparison against state-of-the-art approaches and robustness test In this subsection, we study the classification accuracy of several other approaches and compare their results with ours. For computational complexity reasons, we do not use the whole test set but a subset of it, comprising 1000 images (500 COVID, 120 pneumonia, 120 normal and 260 cancer). On this reduced set, our approach reaches a 100% accuracy, while the reconstruction error approach reaches 97% accuracy (see [fig_ref] Table 8: Comparison with supervised classificationThe training set is composed of 4000 images [/fig_ref]. test CTs. In order to compare our results with those obtained by supervised learning, we have considered three state-of-the-art supervised models, namely GoogLeNet [bib_ref] Going deeper with convolutions, Szegedy [/bib_ref] , AlexNet [bib_ref] Imagenet classification with deep convolutional neural networks, Krizhevsky [/bib_ref] , and ResNet18 [bib_ref] Deep residual learning for image recognition, He [/bib_ref] , which are typically used for image classification and able to successfully classify out-of-sample examples. As a first experiment, we have trained-in a supervised way-the aforementioned models, using a training set composed of 2000 COVID-19 CTs and 2000 non-COVID-19 CTs. The non-COVID-19 set is composed of five classes, namely normal CTs, pneumonia and three types of lung cancers CTs. Afterwards, each model has been evaluated on the reduced test set. The obtained performance indexes are [fig_ref] Table 8: Comparison with supervised classificationThe training set is composed of 4000 images [/fig_ref]. These results are comparable to those obtained from the KDEbased approach. While the supervised DL models have a performance similar to our approach, we expect that their performance is more sensitive (i.e. less robust) to unseen test images. To corroborate this statement, we carry out a final experiment, where we retrain all the supervised models using a modified data set: we eliminate pneumonia CTs from the non-COVID-19 images, replace them with the normal CTs, apply all the procedures from scratch. Once the best weights are achieved, we perform the test phase on the reduced test set, including pneumonia CTs. The corresponding performance indexes are presented in, while the confusion matrices are shown in [fig_ref] Figure 10: Confusion matrices [/fig_ref]. It is observed that the supervised models are able to distinguish COVID-19 perfectly, but since the pneumonia images have not been included in the training phase, the models are in trouble with these images. In other words, if some classes of images are not present in the training set, the supervised-trained models are not capable of correctly classifying them in the test phase. We conclude that our approach is more robust in the presence of outliers in the test set. [formula] (a) AlexNet (b) GoogleNet (c) ResNet18 [/formula] As a second experiment, we compare our approach with several other methods presented in [bib_ref] Momenzadeh Alireza (2022) A novel unsupervised approach based on the hidden features..., Michele [/bib_ref]. In particular, we consider the Histogram-Based DDCAE (HB-DDCAE) method proposed in that work, together with several shallow methods. The shallow methods are: support vector machine (SVM) with 2-degree (SVM-2D) and 3-degree (SVM-3D) polynomial, and radial basis function (SVM-RBF) kernels; multilayer perceptron (MLP) equipped with single hidden layers composed by 50 ## 3 Exploiting probability density function of deep convolutional… (MLP-50), 100 (MLP100), and 200 (MLP-200) neurons; random forest (RF) composed by 100 (RF-100), 500 (RF-500), and 1000 (RF-1000) binary trees. See [bib_ref] Momenzadeh Alireza (2022) A novel unsupervised approach based on the hidden features..., Michele [/bib_ref] for details about the training. The results are presented in [fig_ref] Table 1: A synoptic overview of main related work on COVID-19 detection/classification [/fig_ref]. From [fig_ref] Table 1: A synoptic overview of main related work on COVID-19 detection/classification [/fig_ref] , we note that the shallow algorithms have an inferior performance. On the other hand, the HB-DDCAE method has the same performance of the proposed approach. [fig_ref] Table 1: A synoptic overview of main related work on COVID-19 detection/classification [/fig_ref] reports the (numerically evaluated) average times required by the implemented methods for classifying a batch of 10 images in the test phase. A comparison of the entries of this table leads to the conclusion that the average test time of our method is over 25% less than the corresponding ones of the implemented benchmark models. We have numerically ascertained that this is due to the fact that our proposed method works on the reduced-size (i.e. compressed) hidden feature vectors, while all the benchmark models directly process the full-size input test images. This conclusion provides further support about the actual effectiveness of the proposed KDE-based classifying approach. ## Test-time comparisons ## Conclusion and hints for future research We propose a method for classifying lung CTs as COVID-19 or non-COVID-19. The method exploits a DCAE trained on COVID-19 CTs only and a KDE estimation of the PDF of the DCAE hidden feature vectors. The DCAE is used to produce the corresponding hidden feature vector to classify an image. Afterwards, we use the soobtained PDF evaluation to compute, in the test phase, the PDF value of the hidden feature vector, corresponding to a test image: if the PDF value is above a suitable threshold, that image is classified as COVID-19, otherwise as non-COVID-19. We compare our KDE-based approach to the benchmark method that is based on the reconstruction error. In addition, we also check the accuracy performance of three widely known supervised models and the results of some recent papers. The carried out tests support the conclusion that the proposed approach is highly effective in terms of both achieved test accuracy and the needed test times. The presented results could be extended, at least, along two main research lines. A first research line could concern the utilisation of generative adversarial networks (GANs) for the generation of additional training examples in the case of new COVID-19 mutations (as, for example, the Omicron one), in order to quickly provide reliable automatic detection of these mutations without pausing for the acquisition of sufficiently large new datasets. A second research line could regard the implementation of the proposed algorithmic framework atop distributed Fog/Cloud networked technological platforms [bib_ref] Alireza Momenzadeh et al (2021) Learning-in-the-Fog (LiFo): deep learning meets Fog computing..., Baccarelli [/bib_ref] [bib_ref] *: Energy and delay-efficient dynamic queue management in tcp/ip virtualized data centers, Baccarelli [/bib_ref] , in order to be capable to quickly generate reliable clinical diagnosis by leveraging the low-delay and (possibly, adaptive [bib_ref] Recursive Kalman-type optimal estimation and detection of hidden Markov chains, Baccarelli [/bib_ref] and /or multiantenna empowered [bib_ref] Optimized power allocation and signal shaping for interference-limited multi-antenna "ad hoc" networks, Baccarelli [/bib_ref] [bib_ref] Optimized power-allocation for multiantenna systems impaired by multiple access interference and imperfect..., Baccarelli [/bib_ref] capability of emerging Fog computing platforms for allowing ubiquitous wireless access to computing-demanding medical diagnostic services. [fig] a: Normal. (b) (c) Cancer. (d) Pneumonia. [/fig] [fig] Figure 3: The autoencoder architecture training parameters have been selected because they yielded the highest classification accuracy. [/fig] [fig] Figure 4: Numerically evaluated plots of accuracy-vs.-epochs and MSE-vs.-epochs under the training phase [/fig] [fig] Figure 5: An example of KDE estimation over n = 4 data points: a copy of the kernel is placed on each data point and the copies are summed to produce the final PDF estimate [/fig] [fig] Figure 6: Estimation of the PDF of the latent space (stage 1) and classification of the test images (stage 2) Fig. 7 Evaluation of reconstruction errors of training set and the threshold (stage 1), and classification of the test images (stage 2) [/fig] [fig] Figure 8: Results under the proposed approach [/fig] [fig] Figure 9: Performance results of the reconstruction error approach [/fig] [fig] Figure 10: Confusion matrices: pneumonia CTs are present in test data set, but not in the train data set reported in [/fig]

https://www.semanticscholar.org/paper/a70d64565d3e430a1ef95e9677b07cbe85b781bb

The Emesis Trial: Depressive Glioma Patients Are More Affected by Chemotherapy-Induced Nausea and Vomiting

Frontiers in Neurology

3513ce18-3757-4358-b15f-7060af98e6d0

The Emesis Trial: Depressive Glioma Patients Are More Affected by Chemotherapy-Induced Nausea and Vomiting Purpose: Glioma patients face a limited life expectancy and at the same time, they suffer from afflicting symptoms and undesired effects of tumor treatment. Apart from bone marrow suppression, standard chemotherapy with temozolomide causes nausea, emesis and loss of appetite. In this pilot study, we investigated how chemotherapy-induced nausea and vomiting (CINV) affects the patients' levels of depression and their quality of life.Methods: In this prospective observational multicentre study (n = 87), nausea, emesis and loss of appetite were evaluated with an expanded MASCC questionnaire, covering 10 days during the first and the second cycle of chemotherapy. Quality of life was assessed with the EORTC QLQ-C30 and BN 20 questionnaire and levels of depression with the PHQ-9 inventory before and after the first and second cycle of chemotherapy.Results: CINV affected a minor part of patients. If present, it reached its maximum at day 3 and decreased to baseline level not before day 8. Levels of depression increased significantly after the first cycle of chemotherapy, but decreased during the further course of treatment. Patients with higher levels of depression were more severely affected by CINV and showed a lower quality of life through all time-points.Conclusion:We conclude that symptoms of depression should be perceived in advance and treated in order to avoid more severe side effects of tumor treatment. Additionally, in affected patients, delayed nausea was most prominent, pointing toward an activation of the NK 1 receptor. We conclude that long acting antiemetics are necessary to treat temozolomide-induced nausea. # Introduction Brain tumors are among the most aggressive neoplasms. Glioblastoma, the malignant glioma with the worst prognosis, is associated with a median survival time of 16-18 months and a 5 year survival rate of 6 % for male and 9 % for female patients (1). Standard treatment includes bulk surgery, if possible, followed by radiotherapy combined with concomitant and adjuvant chemotherapy with temozolomide (TMZ). TMZ is an orally available alkylating agent administered concomitantly during radiotherapy at 75 mg/m²/d followed by six adjuvant cycles at 150-200 mg/m² of body surface on day 1-5 of a 28 day cycle. Common side effects are bone marrow suppression and, in rare cases, liver toxicity with elevated transaminases [bib_ref] Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma, Stupp [/bib_ref] , skin erythema, alopecia and others. Close monitoring of neutrophils, lymphocyte and thrombocyte count and transaminases on a weekly basis and dose reduction, if required, is crucial. The most common non-hematological side-effects are nausea, emesis and loss of appetite. At the standard dose of 150-200 mg/m 2 , TMZ is considered to be moderately emetogenic, which means that 30-90 % of patients would experience nausea, emesis and loss of appetite during treatment without appropriate emetogenic prophylaxis. Chemotherapy-induced nausea and vomiting (CINV) can occur as an acute or delayed reaction. Acute nausea and vomiting occur within 24 h after application of chemotherapy, usually with a peak at 5-6 h. Nausea is induced via the peripheral 5-hydroxytryptophan receptor 3 (5-HT 3 ) (3). Delayed nausea occurs from 24 to 120 h and is activated through a central pathway, mainly activated through the neurokinin-1 (NK 1 ) receptor. Anticipatory nausea is a conditioned response starting already before application of chemotherapy in expectancy of nausea, i.e., when the chemotherapy infusion comes in sight. The most important breakthrough in antiemetic treatment took place in 1992 when ondansetron was launched as the first 5-HT 3 antagonist in the market. A second important member of this class of agents is granisetron. With a median half-life of approximately 4 h (ondansetron) and 10 h (granisetron), both substances are useful to treat acute, but not delayed nausea. Prophylactic antiemetic treatment with steroids is usually not applied in brain tumor patients since patients are often heavily pretreated with corticosteroids to reduce peritumoral edema and rapid tapering is desired. In addition, several publications suggest tumor-promoting effects of corticosteroids [bib_ref] Dexamethasone protected human glioblastoma U87MG cells from temozolomide induced apoptosis by maintaining..., Das [/bib_ref]. The usual antiemetic treatment in patients with glioma receiving TMZ consists of a 5-HT 3 antagonist like ondansetron or granisetron, approximately 1 h before chemotherapy. However, clinical experience shows that about one third of patients suffer from severe nausea and emesis despite antiemetic treatment, affecting the patients' health-related quality of life (HRQoL). A recent randomized phase-II trial showed that combination of aprepitant plus ondansetron may increase acute anti-emetic response on day 1 and may have benefits regarding CINV's effect on HRQoL [bib_ref] Randomized open-label phase II trial of 5-day aprepitant plus ondansetron compared to..., Patel [/bib_ref]. In addition to treatment burden, patients with gliomas develop depression during the first six months after diagnosis in about 15-20 % of cases [bib_ref] Frequency, clinical associations, and longitudinal course of major depressive disorder in adults..., Rooney [/bib_ref] and up to 30 % of brain tumor patients suffer from clinically relevant depression (assessed at any time during the course of disease) [bib_ref] Single-institution cross-sectional study to evaluate need for information and need for referral..., Reinert [/bib_ref]. Depression is associated with reduced physical function, cognitive impairment and HRQoL reduction [bib_ref] Depression in cerebral glioma patients: a systematic review of observational studies, Rooney [/bib_ref]. HRQoL is impaired in patients with high grade gliomas as compared to healthy controls, and similar results were found in patients with other types of solid cancer, e.g., NSCLC [bib_ref] Neurobehavioral status and health-related quality of life in newly diagnosed high-grade glioma..., Klein [/bib_ref]. Patients treated with TMZ experience no worsening but rather a slight improvement of HRQoL as compared to their baseline pretreatment assessment [bib_ref] Health-related quality of life in patients treated with temozolomide versus procarbazine for..., Osoba [/bib_ref]. Adding TMZ after radiotherapy has no negative implications on HRQoL [bib_ref] Health-related quality of life in patients with glioblastoma: a randomised controlled trial, Taphoorn [/bib_ref]. Nonetheless, treatment associated side-effects like CINV may seriously affect patients' HRQoL. Accordingly, one of the most common fears of patients from chemotherapy is nausea [bib_ref] Rankings and symptom assessments of side effects from chemotherapy: insights from experienced..., Sun [/bib_ref]. In the study presented here, we investigated the level and time course of nausea, emesis and loss of appetite in patients with malignant brain tumors during their first two cycles of chemotherapy with TMZ. In addition, we asked for the patients' HRQoL and levels of depression prior to chemotherapy and after the first and second cycle of chemotherapy. Our aim was to determine whether there is an interaction between CINV and patients' levels of depression and HRQoL at any of the given time-points. # Methods ## Study population In this prospective, observational, multicentre study, we investigated patients suffering from primary or recurrent malignant glioma receiving chemotherapy in six hospitals in Germany specialized in treatment of glioma patients (University Hospitals Marburg, Münster, Regensburg, Würzburg as well as DIAKOVERE Henriettenstift Hannover and Hospital Barmherzige Brüder Regensburg) in between 2012 and 2016. All 87 patients were included consecutively. Permission of the local ethics committee was obtained (08/13, 26.02.2013), and all patients gave informed consent to participate. Main inclusion criteria were age older than 18 years, qualification for legal acts and a primary or recurrent glioma requiring chemotherapy during the adjuvant phase of the treatment. HRQoL and levels of depression were assessed at least 1 week prior to chemotherapy (t0) and at least 1 week after the first (t1) and second (t2) cycle of chemotherapy. The level and time course of nausea, emesis and loss of appetite were asked for during the first two cycles of chemotherapy with TMZ (c0, c1). This study was conducted following the STROBE guidelines for observational studies. ## Questionnaires Patients' baseline characteristics (sex, age, Karnofsky Performance Status (KPS), WHO-grade (low: WHO grade I+II, high: WHO grade III+IV), chemotherapeutic agent and dosage and concomitant antiemetic therapy) were assessed by a questionnaire designed for this study's purpose. The validated MASCC questionnaire was used to evaluate nausea, emesis and loss of appetite. It scales nausea from 0 to 10 with 0 meaning no nausea at all, frequency of emesis and loss of appetite (on a dichotome scale with yes/no) on a daily basis [bib_ref] MASCC and ESMO guideline update for the prevention of chemotherapy-and radiotherapy-induced nausea..., Roila [/bib_ref]. We expanded the original MASCC questionnaire from 5 to 10 days in order to additionally cover the five days after the last application of TMZ, which is given day 1-5 in cycles of 28 days (Supplement 2). Timepoints of evaluation were 1 day prior to chemotherapy as baseline, on the first day of chemotherapy (before and after application) and day 2-10 during c1 and c2. Patients were asked to indicate their level of nausea on a numeric rating scale to visualize the extent of nausea. The PHQ-9 is an established tool to evaluate depression by patient self-report [bib_ref] Screening for depression in medical settings with the Patient Health Questionnaire (PHQ):..., Gilbody [/bib_ref] and is validated for glioma patients [bib_ref] Screening for major depressive disorder in adults with cerebral glioma: an initial..., Rooney [/bib_ref]. PHQ-9 is sensitive for intra-patient changes [bib_ref] The PHQ-9: validity of a brief depression severity measure, Kroenke [/bib_ref] and consists of nine questions, ranging on a scale from 0 to 3 with a maximum of 27 points. Results can be subclassified in five groups (no symptoms: 0-4 points, minimal symptoms: 5-9 points, minor depression: 10-14 points, moderate major depression: 15-19 points, severe major depression: 20-27 points). In this study, levels of depression were evaluated prior to the first cycle of chemotherapy (t0), after completion of the first cycle of therapy (t1) and after completion of the second cycle of therapy (t2). In order to identify changes in patients' HRQoL, we asked patients to fill in the EORTC QLQ-C30 and Modul QLQ-BN20 questionnaires at t0, t1 and t2. The EORTC QLQ-C30 consists of 30 questions, which can be subclassified in 15 categories (global health, physical functioning, role functioning, emotional functioning, cognitive functioning, social functioning, fatigue, nausea, pain, dyspnea, insomnia, appetite loss, constipation, diarrhea, financial difficulties) [bib_ref] Testing the EORTC Quality of Life Questionnaire on cancer patients with heterogeneous..., Ringdal [/bib_ref] [bib_ref] Reference data for the quality of life questionnaire EORTC QLQ-C30 in the..., Schwarz [/bib_ref]. Answers are ranging on a scale from 0 to 4 (except global health item: 0-7). The EORTC QLQ-BN20 was designed to measure HRQoL particularly in glioma patients [bib_ref] An international validation study of the EORTC brain cancer module (EORTC QLQ-BN20)..., Taphoorn [/bib_ref]. Answers range on a scale from 0 to 4 which are subclassified in 11 brain tumor specific categories (future uncertainty, visual disorder, motor dysfunction, communication deficit, headache, seizures, fatigue, rash, alopecia, weakness of legs, and loss of bladder control). ## Statistical analyses Statistical analyses were performed using IBM SPSS Statistics 25 (SPSS Worldwide, Chicago, IL, USA). For patients' characteristics, descriptive statistics were performed. For EORTC QLQ-C30 and QLQ-BN20, scores for each subcategory and overall scores were calculated via linear transformation using the official EORTC QLQ-C30 Scoring Manual [bib_ref] The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life..., Aaronson [/bib_ref]. Patients with missing data were included if more than 50 % of questions per item were completed. Missing single items, items with <50 % of given information and missing questionnaires were not taken into account. For PHQ-9, overall points achieved were summed up and summarized into the five given subcategories described above. Mean values for nausea, emesis and loss of appetite (MASCC) were calculated for each time point during the first two cycles of chemotherapy. Data was examined for Gaussian distribution by Kolmogorov-Smirnov testing. We performed the student's t-test in equally distributed data and the Wilcoxon test in non-equally distributed data to evaluate , a minor part of the patients received a combination of Lomustine (CCNU) and TMZ [n = 6 (6.9 %) in c1]; n = 5 (5.7 % in c2). Serotonine receptor antagonists were the most prevalent antiemetic prophylaxis during c1 (ondansetrone n = 46, 52.8 %; granisetrone n = 13, 14.9 %; palonosetrone n = 6, 6.9 %) and c2 (ondansetrone n = 39, 44.8 %; granisetrone n = 9, 10.3 %; palonosetrone n = 13, 14.9 %) [fig_ref] TABLE 1 |: Patients'characteristics, n = 87, chemotherapy and concomitant antiemetic therapy in cycle 1 [/fig_ref]. ## Gastrointestinal symptoms During c1, we spotted an increase of nausea directly after the application of the chemotherapeutical agent using the MASCC questionnaire [fig_ref] FIGURE 1 |: Mean of nausea [/fig_ref]. Symptoms remained constantly high until day 7. The CINV associated symptoms lasted ∼2 days longer than chemotherapy was applied. Similarly, emesis increased directly after application and took 5 days to return to baseline levels [fig_ref] FIGURE 1 |: Mean of nausea [/fig_ref]. During c1, patients gradually lost their appetite with a minimum of appetite at day 5 and did not completely recover until day 10 [fig_ref] FIGURE 1 |: Mean of nausea [/fig_ref]. During c2, nausea slowly increased with a maximum at day 6 [fig_ref] FIGURE 1 |: Mean of nausea [/fig_ref]. In contrast to c1, emesis most often developed not before day 2 of chemotherapy and was back to baseline levels by day 4 [fig_ref] FIGURE 1 |: Mean of nausea [/fig_ref]. Appetite, on the contrary, hit its minimum at day 4 during c2 and was not back to former levels at day 10 [fig_ref] FIGURE 1 |: Mean of nausea [/fig_ref]. Exact frequencies of nausea, emesis and loss of appetite at the respective days of chemotherapy during c1 and c2 are provided in [fig_ref] TABLE 2 |: Frequencies of symptoms of nausea, emesis and loss of appetite during c1... [/fig_ref]. In order to investigate if the choice of chemotherapeutic regimen had any impact on nausea, emesis or loss of appetite, we performed a subanalysis in patients who received TMZ only (c1: n = 81, c2: n = 70) or TMZ + CCNU (c1: n = 6, c2: n = 5). The chemotherapeutic regimen had no significant effect on nausea (c1: p = 0.607, c2: p = 0.514), emesis (c1: p = 0.471, c2: p = 0.412) or loss of appetite (c1: p = 0.471, c2: p = 0.207). The extent of CINV (nausea c1: p = 0.969, c2: p = 0.614; emesis c1: p = 0.260, c2: p = 0.863; loss of appetite c1: 0.368, c2: 0.716) was not significantly significantly different in patients with low (n = 13) or high grade (n = 74) tumors during c1 nor c2. A poorer general condition as assessed with the KPS (≤70) was not significantly associated with nausea (c1: p = 0.969, c2: p = 0.614), emesis (c1: p = 0.260, c2: p = 0.863) or loss of appetite (c1: p = 0.368, c2: p = 0.716), as compared with patients with a KPS > 70 at c1 or c2. ## Depression Prior to chemotherapy, the mean baseline PHQ-9 score was 6.79 (0-22). At t1, it increased to 8.25 (0-25), but dropped to 7.13 (0-27) at t2 . In total, mean PHQ-9 scores indicated minimal depressive symptoms. However, single patients with moderate or severe major depression could be identified after chemotherapy [fig_ref] TABLE 3 |: Classification of PHQ-9 symptoms, the absolute and relative number of patients and... [/fig_ref]. The mean PHQ-9 was significantly higher at t1 as compared to the level prior to chemotherapy, with an effect size r of 0.35 (p = 0.003). By contrast, at t2, levels of depression were not significantly different from the scores at t0 (p = 0.341) . Patient drop-out is summarized in Supplement 1. We performed a subanalysis to investigate if the chemotherapeutic regimen (TMZ or CCNU + TMZ) would have any impact on depression in c1 (TMZ: n = 81, TMZ + CCNU: n = 6) or c2 (TMZ: n = 70, TMZ + CCNU: n = 5). No significant effect on the PHQ-9 score was found at t0 (c1: p = 0.648, c2: p = 0.503), t1 (c1: p = 0.158, c2: p = 0.308) or t2 (c1: p = 0.629, c2: p = 0.629). Patients with low grade gliomas (n = 13) had a significant higher likelihood of a higher PHQ-9 score at t1 (p = 0.010) and t2 (p = 0.041) as compared with patients with high grade glioma (n = 74). There was no significant difference to the baseline values at t0 (p = 0.133). Patients with a lower KPS (≤ 70) had a significantly higher PHQ-9 score at t1 (p = 0.010) and t2 (p = 0.041) as compared to patients with a KPS of >70. At baseline assessment at t0, however, no significant difference of PHQ-9 was found (p = 0.133). Patients with higher levels of depression at t0 showed a significantly higher likelihood of developing nausea (p = 0.00) and emesis (p = 0.023) during c1. Similarly, patients with higher levels of depression at t1 also had a significantly higher incidence of emesis (p = 0.00) and loss of appetite (p = 0.03) during c2. Vice versa, patients experiencing nausea (p = 0.00) or emesis (p = 0.002) during c1 showed significantly elevated levels of depression at t1. This was also found to be true for patients' levels of depression at t2, if they experienced nausea (p = 0.027) and emesis (p = 0.00) during c2. ## Quality of life Patients' HRQoL assessment with the QLQ-C30 questionnaire showed a significant drop in the mean of the global health item with an effect size r of 0.22 (p = 0.044) and physical function with an effect size r of 0.22 (p = 0.044) at t1. Fatigue (p = 0.002) and nausea (p = 0.009) increased at t1 with effect sizes r of 0.34 and 0.29, respectively. Global health was also reduced at t2 with an effect size r of 0.24 (p =0.029), as well as nausea with an effect size r of 0.28 (p = 0.01). The other items of the QLQ-C30 questionnaire showed no significant changes in t1 or t2. The QLQ-BN20 questionnaire showed a significant increase of the weakness of legs item at t1 with an effect size r of 0.027 (p = 0.014). At t2 loss of hair worsened significantly with an D-1 means day prior to chemotherapy application. D1 * day 1 prior to application of chemotherapy and D1# day 1 after application of chemotherapy. FIGURE 2 | PHQ-9 prior to (t0) and after the first (t1) and second (t2) cycle of chemotherapy: The mean PHQ-9 at t1 is significantly (p = 0.003) higher than mean PHQ-9 at t0 indicating a higher burden of depression at t1. No significant difference was found in PHQ-9 at t1 and t2. Patients whose PHQ-9 levels reached a score above 15 were defined as moderately or severely depressed and analyzed in a separate HRQoL subanalysis. In contrast to patients with a PHQ-9 score lower than 15 during all time-points of observation (t0, t1, t2), patients with signs of major depression showed a significant impairment in their HRQoL concerning global health, physical function, role function, social function, future uncertainty and fatigue during all time points of measurement . Chemotherapy-induced nausea was not significantly different between the two groups, whereas loss of appetite was significantly more frequent in patients with higher levels of depression at t1 and t2 . In order to analyze the impact of general condition, our patient series was divided in a group with a lower (≤70, n = 20) and higher (>70, n = 62) KPS. We performed a HRQoL subanalysis comparing these two groups. Patients with a lower KPS showed a significant impairment in HRQoL concerning global health, physical functioning, role functioning, social functioning, future uncertainty, motor dysfunction and weakness of legs compared to patients with a KPS > 70 at all time-points of observation (t0, t1, t2). Neither nausea nor loss of appetite were significantly different in the two groups [fig_ref] TABLE 5 |: Comparison of the mean of the items of the EORTC QLQ-C30 and... [/fig_ref]. # Discussion To our knowledge, this is the first prospective multicenter study assessing glioma patients under the following conditions: a defined 10-day period before, during and after application of chemotherapy and its effects on HRQoL and levels of depression. In order to measure nausea, emesis and loss of appetite, we applied the expanded MASCC questionnaire, modified with a numeric rating scale and assessed nausea, emesis and loss of appetite for 10 consecutive days, during the c1 and c2 of chemotherapy. Overall, the burden of CINV symptoms was moderate. Interestingly, the application of TMZ during day 1-5 in both c1 and c2 appeared to cause delayed and prolonged nausea, emesis and loss of appetite. In view of the significant delay of nausea and emesis observed in this study, we speculate that a relevant activation of the NK1 pathway takes place, supported by several clinical trials reducing nausea by combining a NK1 receptor antagonist with a 5 HT 3 antagonist setron [bib_ref] Profile analysis of chemotherapy-induced nausea and vomiting in patients treated with concomitant..., Matsuda [/bib_ref] [bib_ref] Combination of palonosetron, aprepitant, and dexamethasone effectively controls chemotherapy-induced nausea and vomiting..., Matsuda [/bib_ref] [bib_ref] Antiemetic prophylaxis with temozolomide: an audit from a tertiary care center, Patil [/bib_ref]. Shorter acting antiemetics should therefore be substituted with longer acting substances like palonosetron, or through the addition of a NK1 receptor antagonist like aprepitant, rolaprepitant or the fix combination of netupitant and palonosetron [bib_ref] European Society for Medical Oncology (ESMO) position paper on supportive and palliative..., Jordan [/bib_ref] [bib_ref] Review of NEPA, a novel fixed antiemetic combination with the potential for..., Hesketh [/bib_ref]. We also observed a tendential decrease of emesis in c2, possibly as a consequence of an adjustment in antiemetic prophylaxis after c1, e.g., increase in palonosetron intake [fig_ref] TABLE 1 |: Patients'characteristics, n = 87, chemotherapy and concomitant antiemetic therapy in cycle 1 [/fig_ref]. As higher levels of nausea and emesis exhibit significant intercorrelations with depressive 4 | Comparison of the mean of the items of the EORTC QLQ-C30 and QLQ-BN20 questionnaire at t0 prior to chemotherapy and after the first (t1) and second cycle of chemotherapy (t2) in patients with a PHQ-9 score of <15 and ≥15. symptoms and HRQoL, constant monitoring and treatment of gastrointestinal side effects would be crucial. While the PHQ-9 score prior to chemotherapy indicated only minimal symptoms of depression in most patients, PHQ-9 scores of 15 or higher in single patients pointed toward moderate to severe pre-existing symptoms of depression in a specific subpopulation. After completion of c1, levels of depression increased significantly. Chemotherapy effects such as nausea and emesis or myelosuppression and infections, but also the fear of these symptoms may enhance the psychosocial burden of patients and lead to a higher level of psychological stress [bib_ref] Role of psychosocial variables on chemotherapy-induced nausea and vomiting and health-related quality..., Grassi [/bib_ref] [bib_ref] The impact of chemotherapy-related nausea on patients' nutritional status, psychological distress and..., Farrell [/bib_ref]. After completion of c2, however, levels of depression decreased. This may point toward a reduced level of stress once the treatments have become routine. Interestingly, we observed that not only was the extent of gastrointestinal symptoms associated with a significantly higher level of depression after the respective cycle of chemotherapy, but also vice versa-patients with higher baseline levels of depression experienced significantly more severe nausea, emesis or loss of appetite. We presume that treatment-resistant or anticipatory nausea during chemotherapy may be psychosomatic to a relevant extent [bib_ref] Evaluation of risk factors predicting chemotherapy-related nausea and vomiting: results from a..., Molassiotis [/bib_ref] [bib_ref] Treatment of breakthrough and refractory chemotherapyinduced nausea and vomiting, Navari [/bib_ref]. The QLQ-C30 and QLQ-BN20 questionnaire assessed prior to and after c1 and c2 indicated fatigue and loss of hair, which may not necessarily have been caused by chemotherapy alone, but possibly resulted also from previous radiotherapy [bib_ref] Hair disorders in patients with cancer, Freites-Martinez [/bib_ref] [bib_ref] Relationship between fatigue and quality of life in patients with glioblastoma multiformae, Lovely [/bib_ref] [bib_ref] A European Organisation for research and treatment of cancer phase III trial..., Soffietti [/bib_ref]. Interestingly, the QLQ-C30 questionnaire showed a significant increase of nausea at t1 and t2, respectively, thus supporting results from the MASCC questionnaire. Global health dropped significantly at t1 and t2. Patients with signs of depressive mood, as indicated by a PHQ-9 score of 15 or higher, showed more severe effects through decreased HRQoL than non-depressed patients. Global health, physical function, role function, social function, future uncertainty and fatigue were already significantly impaired prior to chemotherapy in depressed patients. In the further course of disease, these executing aspects of the patients' lives deteriorated more markedly than in non-depressed patients. By contrast, emotional functioning, dyspnea, appetite loss, headaches and financial difficulties were significantly impaired only during chemotherapy at either t1 or t2. This subanalysis should be interpreted with care as there were less patients represented in the group of a PHQ-9 score of 15 or higher (at t0 n = 7, at t1 n = 13, at t2 n = 8) compared to the group with lower depression scores (at t0 n = 66, at t1 n = 64, at t2 n = 59) and the two subgroup are not equally distributed. Due to its design, the results obtained in this pilot study should be interpreted with some caution. At first, the study is not adequately powered for the quantity of HRQoL parameters assessed with the EORTC QLQ-C30 and QLQ-BN20. Second, we investigated a series of primary and recurrent glioma of different WHO grading treated at different hospitals with inhomogeneous chemotherapy and antiemetic medication representing the daily practice of outpatient care. While most patients received TMZ alone, some patients were treated additionally with lomustine. Third, we neither assessed the general toxicity nor tolerability of chemotherapy. General side-effects of therapy might have had interactions with depression, CINV and HRQol. Even though we documented baseline depression, CINV and HRQoL scores, we did not interview the patients about preexisting psychiatric disorders. In addition, we cannot provide information on the consecutive development of depression, CINV or HRQoL beyond the first two courses of chemotherapy. Although these factors may have influenced the severity of nausea, emesis and loss of appetite, the mode of evaluation established in this study appears to be adequate and the observations on duration of gastrointestinal side effects, intercorrelation with depressive symptoms and effect on HRQoL seems to be robust enough to draw initial conclusions. Taken together, we observed a relevant interaction between gastrointestinal side effects of chemotherapy and depressive symptoms. Neither KPS, WHO grading nor chemotherapeutical regimen did influence CINV symptoms significantly. CINV may be underestimated in glioma patients, may last longer than anticipated, and appears to be aggravated by pre-existing depressive symptoms, severely affecting the HRQoL of the affected patients. During treatment, CINV should be asked for thoroughly and treated with effective, long-lasting antiemetics not only to reduce gastrointestinal symptoms, but also to prevent depressive mood and impairment of HRQoL. Moreover, HRQoL was impaired after initiation of chemotherapy, especially in patients suffering from pre-existing depressive mood. According to the standard within German certified oncological centers, we consider it important to introduce regular screening of the extent of psychosocial burden and depressive symptoms during the course of disease. Early detection and treatment of depression may probably not only stabilize the patient's mood, but also prevent deterioration of gastrointestinal symptoms and HRQoL. # Data availability statement The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. # Ethics statement The studies involving human participants were reviewed and approved by the respective Ethik-Kommission der Medizinischen Fakultät. The patients/participants provided their written informed consent to participate in this study. # Author contributions Study conception and design were prepared by HS, LJ, and KJ, and consented with all co-authors. Material preparation was performed by AK, PH, HP, OG, BW, ML, KJ, and HS. Data collection was coordinated by HS and LJ. Statistical analysis was performed by VD. The manuscript was prepared by VD, HS, and AK, and commented by all authors. All authors read and approved the final manuscript. # Funding This publication was supported by the Open Access Publication Fund of the University of Wuerzburg. [fig] FIGURE 1 |: Mean of nausea (A), emesis (B) and loss of appetite (C) during the first 10 days of the c1 (black rhombus) and c2 (gray square) of chemotherapy. Respective days during the course of chemotherapy are displayed on the x-axis. The median MASCC is shown on the y-axis (nausea: 0-10; emesis: frequency per day; loss of appetite: 0: not at all, 1: loss of appetite). Frontiers in Neurology | www.frontiersin.org 5 February 2022 | Volume 13 | Article 773265 [/fig] [table] TABLE 1 |: Patients'characteristics, n = 87, chemotherapy and concomitant antiemetic therapy in cycle 1 (c1) and cycle 2 (c2), TMZ, Temozolomide; CCNU, Lomustine. [/table] [table] TABLE 2 |: Frequencies of symptoms of nausea, emesis and loss of appetite during c1 and c2 in %. [/table] [table] TABLE 3 |: Classification of PHQ-9 symptoms, the absolute and relative number of patients and the severity of their symptoms respectively at t0, t1 and t2. [/table] [table] TABLE 5 |: Comparison of the mean of the items of the EORTC QLQ-C30 and QLQ-BN20 questionnaire at t0 prior to chemotherapy and after the first (t1) and second cycle of chemotherapy (t2) in patients with a KPS of ≤70 and >70.P-values are provided for each time-point and each item; significant p-values are highlighted. [/table]

https://www.semanticscholar.org/paper/1c3417fab5c74216e0084ce390f86eee6787a912

Retrospective evaluation of pregnant women with celiac disease

Journal of the Turkish-German Gynecological Association

d1f286a5-06ff-433a-9156-a73979475dcb

Retrospective evaluation of pregnant women with celiac disease Objective: To show celiac disease (CD) and its poor pregnancy outcome relationship, and to demonstrate the importance of a gluten-free diet together with low-dose low-molecular-weight heparin (LMWH) and low-dose corticosteroid (LDC) in the management of pregnancies with CD.Material and Methods:This study consisted of 2 groups of patients. Six patients with CD (control group) on a gluten-free diet were monitored during their first pregnancies within the framework of antenatal care program and their pregnancy outcomes were compared with eight poorlytreated pregnant patients with CD (study group) who were referred from other medical institutions. LMWH (enoxaparine 1x2000 Anti-XA IU/0.2 mL/day), and LDC (methylprednisolone 1x4 mg p.o/day) were used in the control group. Their obstetric histories and outcomes of their last pregnancies were compared. The patients' obstetric risk levels were evaluated using the "Beksac Obstetrics Index" (BOI).Results: There were miscarriages in 50% of the study group. There were also 50% and 75% preterm deliveries in the control and study groups, respectively. The BOI of the study group was significantly worse than the control group (1.31 vs. 0.31±0.21, p<0.01). There were no statistically significant differences between age (24±4.7 vs 31.7±6 years, p=0.448), gestational day of birth (259.3±8.5 vs 246.6±24.3), birthweight (2691±698 vs 2262±359 g, p=0.394), and cesarean section rates (p=0.118).Conclusion: CD is a risk factor for adverse pregnancy outcome. Miscarriage and preterm labor are critical complications in pregnancies complicated by CD. A gluten-free diet is important in the treatment. LMWH and LDC seem to be helpful in the management of pregnant women with CD. (J Turk Ger Gynecol Assoc 2017; 18: 56-9) # Introduction Celiac disease (CD) is a small intestinal enteropathy, which is activated by gluten ingestion in patients with a genetic background [bib_ref] Potential new mechanisms of placental damage in celiac disease: anti-transglutaminase antibodies impair..., Simone [/bib_ref]. The prevalence of the disease is around 1% [bib_ref] Screening of tissue transglutaminase antibody in healthy blood donors for celiac disease..., Tatar [/bib_ref]. "Gluten, gluten-specific T-cells, the major histocompatibility complex antigen HLA-DQ, and transglutaminase type 2 (TG2)" are the main actors of this disorder. It has been reported that repeated miscarriages and obstetric complications are more frequent in patients with CD [bib_ref] Celiac disease and pregnancy outcome, Ciacci [/bib_ref]. CD is an autoimmune disorder characterized by circulating anti-TG2 autoantibodies [bib_ref] Anti-tissue transglutaminase antibodies from celiac patients are responsible for trophoblast damage via..., Simone [/bib_ref]. It has been reported that anti-TG2 antibodies act negatively on endometrial receptivity and impair decidual angiogenesis together with interstitial trophoblast migration along with various mechanisms [bib_ref] Coeliac disease-specific autoantibodies targeted against transglutaminase 2 disturb angiogenesis, Myrsky [/bib_ref]. It has also been reported that the chorionic villus (materno-fetal interface) is one of the main targets of anti-TG2 autoantibodies and these antibodies directly attack syncytiotrophoblasts [bib_ref] Maternal celiac disease autoantibodies bind directly to syncytiotrophoblast and inhibit placental tissue..., Anjum [/bib_ref]. In other words, impaired endometrial receptivity and disturbed syncytiotrophoblastic apoptosis might be the main causes of impaired fetal perfusion (intrauterine hypoxia) and poor obstetric outcome. In this report, we compared the pregnancy outcomes of 6 primigravida patients at complete remission on a gluten-free diet and 8 referred patients with CD who had various gestational symptoms in terms of obstetric outcome. # Material and methods This retrospective study was conducted at the Department of Obstetrics and Gynecology, Hacettepe University, between March 2014 and March 2016. There were two groups of patients. The control group comprised six patients with CD on glutenfree diet who were monitored during their first pregnancies. Their pregnancy outcomes were compared with eight pregnant patients (study group) with CD from other medical institutions who had been referred because they had not adhered to their recommended gluten-free diet. All patients were diagnosed with CD before their pregnancies. We used the Beksac Obstetrics Index (BOI), which is an obstetrics index for the assessment of risk levels of high-risk pregnancy groups [(number of alive children + π/10)/Gravida], in order to compare these two groups (π=3.14) [bib_ref] An Obstetrics Index for the Assessment of Risk Levels of "High Risk..., Beksac [/bib_ref]. Patients in the control group were followed up under the "autoimmune disorders in pregnancy" protocol within the special antenatal care program of the Division of Perinatal Medicine. Laboratory tests were performed (complete blood count, liver function enzymes, antithrombin-III and activated protein-C activities, complement 3 and 4, blood glucose level, hereditary thrombophilia-related polymorphisms, antibodies such as antinuclear antibodies, antiphospholipid antibodies, anti-smooth muscle antibodies, anti-double stranded DNA, and others according to individual differences), and necessary precautions were undertaken. Low-dose low-molecularweight heparin (LMWH) (enoxaparine 1x2000 Anti-XA IU/0.2 mL/day), and low-dose-corticosteroid (methylprednisolone 1x4 mg p.o/day) were used together with CD-specific treatment in the control group. In one patient, low-dose corticosteroid (LDC) was used alone without LMWH. Study group patients who did not have an abortion received standard CD treatment at the outpatient clinic. The Statistical Package for the Social Sciences version 17 (IBM SPSS Statistics, Chicago, IL, USA) was used for data analysis. Pearson's Chi-square and Fisher's exact test were used for categorical variables and the t-test was used for continuous variables. This study was performed in compliance with the ethics principles of the university board and those of the national committee. All patients were informed about the study and signed informed consent. The non-interventional clinical research ethics board approval number is GO 16/100 (2016). # Results The demographics of the patients are given in Four of six patients in the control group and three of four patients in non-treatment group gave birth via cesarean section. There were no statistically significant difference regarding their delivery routes (p=0.118). The patients' obstetrics risk levels were evaluated using the BOI. The BOIs of the entire control group was 1.31 because they were all primigravida patients. The mean BOI value of the non-treatment group patients was 0.31±0.21. This difference was statistically significant (p=0.001). # Discussion CD is an autoimmune small intestinal enteropathy, which is activated by dietary gluten (cereal prolamins) and its incidence is about 1% (1, 3). Increased risk of pregnancy failure and obstetric complications has been reported in patients with CD [bib_ref] The impact of maternal celiac disease on birthweight and preterm birth: a..., Khashan [/bib_ref] [bib_ref] Celiac disease and pregnancy outcome, Ciacci [/bib_ref]. It has been reported that up to 50% of women with untreated CD have a history of miscarriage and other unfavorable pregnancy outcomes, which is similar to our findings [bib_ref] Celiac disease and obstetrical-gynecological contribution, Casella [/bib_ref]. Untreated patients with CD also have a higher risk of developing intrauterine growth retardation, low birthweight, stillbirth, pre-term birth, and small-for-gestational-age babies compared with pregnancies with treated CD [bib_ref] Celiac disease and obstetrical-gynecological contribution, Casella [/bib_ref] [bib_ref] Celiac disease and obstetric complications: a systematic review and metaanalysis, Saccone [/bib_ref]. In our study, there were 50% and 75% preterm deliveries in the control and study groups, respectively. Anti-TG2 autoantibodies were reported to be the main source of placenta-specific inflammatory process in patients with CD, which resulted in intrauterine hypoxia and impaired fetal perfusion [bib_ref] Anti-tissue transglutaminase antibodies from celiac patients are responsible for trophoblast damage via..., Simone [/bib_ref]. Impaired apoptosis of syncytiotrophoblasts and disturbed endometrial receptivity by circulating anti-TG2 antibodies seems to be the reason for these implantation and placentation disorders [bib_ref] Coeliac disease-specific autoantibodies targeted against transglutaminase 2 disturb angiogenesis, Myrsky [/bib_ref] [bib_ref] Maternal celiac disease autoantibodies bind directly to syncytiotrophoblast and inhibit placental tissue..., Anjum [/bib_ref]. In our small series, we demonstrated that patients with active CD who were not on a proper gluten-free diet experienced poor pregnancy outcomes, and their BOIs were statistically significantly lower than patients with CD on a gluten-free diet. It has been reported that anti-TG2 plasma levels decreased when CD went into complete remission with a gluten-free diet [bib_ref] Celiac disease, Green [/bib_ref]. Control of circulating anti-TG2 antibodies should be the goal during perinatal surveillance. This may be a rationale in for prophylactic use of LDCs in certain cases, especially when the complement system is activated. On the other hand, the rising prevalence of venous thromboembolism among patients with inflammatory bowel disease and autoimmune diseases should be our concern during the management of these pathologies [bib_ref] Rising prevalence of venous thromboembolism and its impact on mortality among hospitalized..., Nguyen [/bib_ref]. The importance of prophylactic low-dose low-molecular-weight heparin use in diseases such as CD in necessary cases lies in the following: anti-TG2 antibodies bind to human endometrial endothelial cells and impair endometrial angiogenesis by inhibiting the activation of matrix metalloprotease-2 (MMP-2) activity [bib_ref] Coeliac disease-specific autoantibodies targeted against transglutaminase 2 disturb angiogenesis, Myrsky [/bib_ref]. Thus, these biologic changes may be responsible for the induction of venous thromboembolic events. The other possible mechanism for the endothelial injury of vascular structures around the materno-fetal interface (chorionic villae) is the direct attack of anti-TG2 antibodies on endothelial cells of spiral veins, together with endovascular trophoblasts covering and occluding the tip of spiral arteries, which are the opening to the intervillous space, and the syncytiotrophoblasts that cover the outer surface of chorionic villae. Autoimmune antibody positivity should be taken into consideration as a risk factor for poor pregnancy outcome [bib_ref] Does the presence of autoantibodies without autoimmune diseases and hereditary thrombophilia have..., Mumusoglu [/bib_ref]. The final goal should be the elimination of anti-TG2 antibodies through dietary precautions and/or suppression of antibodies by the preventive use of LDCs in necessary cases, especially when the complement system is activated. Endothelial injury of vascular structures should be eliminated. Low-dose LMWH might be critical in cases with antithrombin III and activated protein-C activity changes. Elimination of thrombus formation is also critical to prevent secondary activation of the complement system, which itself may also give harm to surrounding tissues. In our series, we used low-dose LMWH in 5 of the 6 CD patients in the control group (patients with CD under long-term followup) due to active protein-C and antithrombin III activity changes. All of these patients delivered successfully without perinatal mortality and severe morbidity. Three of these patients had preterm deliveries without important neonatal complications. We must also remember that the destruction of chorionic villae by these toxic materials (anti-TG2 antibodies, cell degregades of endothelial cells of spiral veins and complement system proteins) will result in the release of fetal cell degregades (syncytiotrophoblasts) into the maternal circulation and cause a graft-versus-host-like inflammatory process in the placenta. All these patho-biologic events are most probably the reason of impaired fetal perfusion and hypoxia, and these might be the reason of increased obstetric complications such as miscarriage, intrauterine growth retardations, preterm deliveries, and possibly preeclampsia in patients with CD, as observed in the uncontrolled patients with CD of our clinical series. We believe that patient-specific individualized management is essential in pregnancies with CD and dietary control is necessary to provide better pregnancy outcomes. LMWH and LDC seem to be helpful in the management of pregnancies with CD. Further studies are necessary in this field to understand CD in pregnancy. This study is limited by the low number of patients. Ethics Committee Approval: Ethical approval was obtained and given in detail. [table] Table 2: Demographic findings of patients in their last pregnancies (without any treatment) [/table]

https://www.semanticscholar.org/paper/4a92e648e9c906b68f899ce00b257d57660330f5

Genome-wide characterization of WRKY gene family in Helianthus annuus L. and their expression profiles under biotic and abiotic stresses

PLoS ONE

0aed7a40-85f3-4c66-9e1b-c1556c57001b

Genome-wide characterization of WRKY gene family in Helianthus annuus L. and their expression profiles under biotic and abiotic stresses WRKY transcription factors play important roles in various physiological processes and stress responses in flowering plants. Sunflower (Helianthus annuus L.) is one of the important vegetable oil supplies in the world. However, the information about WRKY genes in sunflower is limited. In this study, ninety HaWRKY genes were identified and renamed according to their locations on chromosomes. Further phylogenetic analyses classified them into four main groups including a species-specific WKKY group. Besides, HaWRKY genes within the same group or subgroup generally showed similar exon-intron structures and motif compositions. The gene duplication analysis showed that five pairs of HaWRKY genes (HaWRKY8/9, HaWRKY53/54, HaWRKY65/66, HaWRKY66/67 and HaWRKY71/ 72) are tandem duplicated and four HaWRKY gene pairs (HaWRKY15/82, HaWRKY25/65, HaWRKY28/55 and HaWRKY50/53) are also identified as segmental duplication events, indicating that these duplication genes were contribute to the diversity and expansion of HaWRKY gene families. The dN/dS ratio of these duplicated gene pairs were also calculated to understand the evolutionary constraints. In addition, synteny analyses of sunflower WRKY genes provided deep insight to the evolution of HaWRKY genes. Transcriptomic and qRT-PCR analyses of HaWRKY genes displayed distinct expression patterns in different plant tissues, as well as under various abiotic and biotic stresses, which provide a foundation for further functional analyses of these genes. Those functional genes related to stress tolerance and quality improvement could be applied in marker assisted breeding of the crop. OPEN ACCESS Citation: Li J, Islam F, Huang Q, Wang J, Zhou W, Xu L, et al. (2020) Genome-wide characterization of WRKY gene family in Helianthus annuus L. and their expression profiles under biotic and abiotic stresses. PLoS ONE 15(12): e0241965. https://doi. # Introduction The WRKY gene family is considered as one of the largest transcription factor (TF) family in higher plants, which basically contain an approximate 60-residue DNA-binding domain, named as WRKY domain, with a highly conserved heptapeptide motif WRKYGQK and a C 2 H 2 -or C 2 HC-type of zinc-finger motif included. Both the heptapeptide motif and zinc-finger motif are needed for binding of WRKY TFs to the cis-acting element W-box (C/T)TGAC (C/T). WRKY gene family can be classified into three main groups (I-III), based on the number of WRKY domains and the structure of their zinc-finger motifs. The group I WRKY proteins consist of two WRKY domains, whereas groups II and III contain only one. The group II and III WRKY proteins are distinguished by the type of zinc-finger motif, with a C-X 4-5 -C-X 22-23 -H-X 1 -H type of motif in group II and a C-X 7 -C-X 23 -H-X 1 -C type in group III. The first WRKY gene was cloned and identified from sweet potato, encoding a 549 amino acid protein called SPF1 (SWEET POTATO FACTOR1). Since then, a large number of WRKY genes have been discovered from different plants. Functional analyses showed that WRKY genes are associated with various aspects of physiological processes, including seed dormancy and germination, root development, leaf senescence, modulation of flowering time, plant nutrient utilization etc. The knockout mutant of AtWRKY2 resulted in hypersensitivity of Arabidopsis to ABA during seed germination and post-germination, early growth, suggesting that AtWRKY2 mediates seed germination and post-germination development. Overexpression of OsWRKY31 in rice inhibited plant lateral root formation and elongation, and also affected the transport process of auxin. AtWRKY12, AtWRKY13 and AtWRKY71 are three main genes regulating Arabidopsis flowering time, with AtWRKY12 and AtWRKY13 working antagonistically under short daylight conditions, and AtWRKY71 accelerating flowering. It has been also documented that 12 WRKY genes are involved in leaf senescence in Arabidopsis and rice, as their mutants inhibited or promoted leaf senescence to different extents. In addition to plant growth and development, WRKY genes also participate in modulation of plant tolerance to abiotic and biotic stress. Qiu and Yureported that overexpression of OsWRKY45 in Arabidopsis significantly increased the expression level of PR genes and ABA/ stress regulated genes, thus contributed to the enhancement of disease resistance and salt and drought tolerance of the plant. GmWRKY54 from soybean, which was confirmed in a DNA binding assay that could interact with the W-box, conferred salt and drought tolerance to transgenic Arabidopsis, possibly through the regulation of DREB2A and STZ/Zat10. In tobacco, overexpression of grape VvWRKY2 reduced the susceptibility to fungal pathogens like Botrytis cinerea, Pythium spp. and Alternaria tenuis. The WRKY1 in tobacco could be phosphorylated by a salicylic acid-induced protein kinase (SIPK), resulting in enhanced DNAbinding activity to a W-box sequence from the tobacco chitinase gene CHN50, and subsequently formation of hypersensitive response-like cell death. Thus, WRKY genes may be involved in mitigating the damage caused by stresses, through interacting with the cis-element W-box and activating downstream plant defense signaling. Common sunflower (Helianthus annuus L.) is grown throughout the world as an industrial crop for edible oil. It is the fourth important oilseed crop which contributes to 12% of the edible oil produced globally. However, sunflower production has been threatened by different stresses, among which drought and salinity are two major abiotic constraints. Moreover, parasitic weed Orobanche cumana is a new emerged biotic issue worldwide. WRKY transcription factors are involved in regulation of plant tolerance to both abiotic and biotic stresses. Thus, it is of great interest to characterize a WRKY gene family in sunflower and identify their functions under different stresses. The WRKY gene family has been well studied in sunflower. Giacomelli et al.have identified a total number of 97 WRKY genes in the Asteraceae family, while only 26 of them belong to H. annuus, and this identification was all based on EST database. The publication of reference genome will provide an opportunity to reveal the organization, expression and evolutionary traits of common sunflower WRKY gene family at the genome-wide level. Badouin et al.reported a high-quality reference for the sunflower genome (3.6 gigabases), with 17 chromosomes and 52,232 protein-coding genes on them. Guo et al.identify 112 sunflower WRKY genes from this reference genome and Liu et al.have extended this family to 119 members. In the current study, another sunflower database from a different sunflower genotype was used to search WRKY genes as support and addition to the previous works. A total of 90 HaWRKY genes were identified, among which 89 had corresponding genes in the updated sunflower WRKY gene family, whereas the rest one was newfound (S3 . The 90 WRKY genes could be classified into four main groups, including an extra WKKY group. Analyses on exon-intron organization, motif composition, gene duplication, chromosome distribution, phylogenetic relationship and gene synteny were further conducted to systemically characterize these common sunflower WRKY genes. Additionally, the expression patterns of HaWRKY genes in different plant tissues and in responses to different abiotic and biotic stresses were also recorded, to identify the implication of specific WRKY genes in different biological processes. The present findings provide a foundation for future research on functional characterization of WRKY genes in common sunflower. # Materials and methods ## Gene identification The genome of H. annuus (HA412.v1.1.bronze) was downloaded from Sunflower Genome Database (https://www.sunflowergenome.org/). The protein sequences of the WRKY family of A. thaliana were obtained from Plant Transcription Factor Database (http://planttfdb.cbi.pku. edu.cn/index.php), which were used to search the WRKY genes from H. annuus genome via BlastP and tBlastN (E-value � 1e-20). Then Pfam database (http://pfam.xfam.org/) and SMART database (http://smart.embl-heidelberg.de/) were used for verification of the WRKY domains. These potential sequences were further queried in the NCBI Conserved Domains Database (https://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml) and InterProScan Database (http://www.ebi.ac.uk/interpro/search/sequence-search) to validate the conserved domain. The molecular weight (Mw) and isoelectric point (pI) of the full-length proteins were predicted using the pI/Mw tool (https://web.expasy.org/compute_pi/) in ExPASy. ## Phylogenetic analysis and gene structure Multiple sequence alignment based on WRKY domain sequences were conducted by clustal W analysis with default parameters. A neighbor-joining (NJ) tree was constructed in MEGA 5.2 with the following criteria: Poisson model, pairwise deletion, and 1000 bootstrap replications. Further maximum likelihood (ML) analysis of WRKY gene family from sunflower and Arabidopsis was conducted, to confirm the reliability of the result. The intron-exon structures of sunflower WRKY genes were analyzed by comparing predicted coding sequences with their corresponding full-length sequences using the online tool Gene Structure Display Sever (GSDS, http://gsds.cbi.pku.edu.cn/). The MEME online program (Multiple Expectation Maximization for Motif Elicitation) version 4.11.1 (http://meme-suite.org/index.html) was used to identify conserved motifs in the sunflower WRKY proteins. ## Chromosomal distribution and gene duplication Multiple Collinearity Scan toolkit (MCScanX) was adopted to analyze the gene duplication events with default parameters. PAL2NAL v14 was subsequently used to calculate dN and dS, with a dN/dS ratio of 1 indicative of neutral selection. ## Plant materials, growth conditions and treatments Seeds of both O. cumana and two sunflower cultivars JY207 and TK0409 were provided by the Institute of Plant Protection, Inner Mongolia Academy of Agricultural and Animal Husbandry Sciences, Hohhot, China. For abiotic stresses, common sunflower cultivar TK0409 was used. The seeds were germinated and grown in peat moss according to our previous study. At the four-leaf stage, seedlings with uniform size were selected to expose to NaCl (0, 150, and 300 mM) for salt stress and polyethylene glycol-6000 (PEG-6000, 0, 10% and 20% w/v) for simulated drought stress, respectively. The treatment concentrations were selected based on our preliminary experiments. All the plants were placed in a growth chamber with light intensity ranging from 250 to 350 μmol m -2 s -1 , temperature at 16-20˚C, and relative humidity at approximately 55-60%. After another one week, roots and leaves of the sunflower seedlings were sampled for RNA isolation. For biotic stress, root parasitic weed Orobanche cumana were applied to common sunflower cultivars TK0409 (susceptible) and JY207 (resistant). 200 mg of O. cumana seeds were homogeneously mixed with 0.5 kg of the peat and vermiculite (1:1, v/v) substrate. Sunflowers were grown in the substrate containing O. cumana seeds as mentioned above. All the plants were placed in a growth chamber with 20˚C at the daytime and 14˚C at night, photoperiod for 14 h, and an irradiance of 300 μmol m −2 s −1 . Three weeks after inoculation, sunflower roots were collected for experiments. In order to avoid tissue contamination of O. cumana, sunflower roots that were over 1 cm adjacent to the interaction site with O. cumana were harvested. Each treatment was replicated three times. # Gene expression analysis To determine the expression profiles of the HaWRKY genes under natural conditions, the transcriptomic data from 10 sunflower tissues, including bract, corolla, leaf, ligule, ovary, pollen, seed, stamen, stem, and style, were downloaded from the "Gene Expression Browser" of Sunflower Genome Database. Data were transformed by log 2 (FPKM+1). For biotic stress, the transcriptomic data were from our previous work. Genes with false discovery rate (FDR) less than 0.01 and fold change more than 2 were considered differentially expressed between infected and non-infected sunflower. The expression profile heat-maps were generated using HemI 1.0 software. For abiotic stress, TaKaRa MiniBEST Plant RNA Extraction Kit (Takara Bio, Kyoto, Japan) was used to extract total RNA from sunflower leaves and roots. The quantity and quality of RNA samples were assessed by agarose gel electrophoresis and using a Nanodrop 2000 Spectrophotometer for A260/A280 ratio. 200 ng of total RNA was reverse transcribed by using TaKaRa PrimeScript TM RT reagent Kit with gDNA Eraser. The gene-specific primers of sunflower for qRT-PCR amplification were designed by using Primer Premier 5.0 software and provided in S1 [fig_ref] Table: Primers for qRT-PCR [/fig_ref] SYBR Premix Ex Taq II (Tli RNaseH Plus, TaKaRa) in CFX96TM Real-Time PCR detection System (Bio-Rad, Hercules, CA, USA) was used to conduct the qRT-PCR experiments. The PCR conditions consisted of pre-denaturation at 95˚C for 30 s, 40 cycles of denaturation at 95˚C for 5 s, and annealing and extension at 58˚C for 30 s. The default setting was used for the melting curve stage. The 2 −ΔΔCt method with three replications was performed for analysis and the ACT2 was selected as the reference gene. Genes with relative expression fold change (stress/control) � 2 and � 0.5 were considered significantly (Tukey's HSD test, P < 0.05) up and down-regulated, respectively. Ethical approval. This article does not contain any studies with human participants or animals performed by the authors. ## Ethics approval and consent to participate # Results ## Identification of the wrky genes A total of 104 candidate WRKY genes were predicted from H. annuus, among which 14 were removed after domain check and the rest 90 were named as HaWRKY1-HaWRKY90 (S2 . In comparison to the results of Guo et al.and Liu et al., one gene (HaWRKY51) in our study was found not included in the previous works, whereas the rest had their corresponding genes. The length of these genes ranged from 388 bp (HaWRKY5) to 8445 bp (HaWRKY49), with molecular weight (MW) from 10.48 to 74.25 kDa. The isoelectric point (pI) of these proteins ranged from 4.81 (HaWRKY86) to 10.44 (HaWRKY53) (S2 . ## Phylogenetic analysis of wrky family members Phylogenetic analysis of WRKY family members of H. annuus and A. thaliana was conducted based on WRKY domain. The common sunflower WRKY domains were divided into four large groups [fig_ref] Fig 1: Phylogenetic relationships of WRKY genes from common sunflower and Arabidopsis [/fig_ref] , corresponding to group I, II and III in Arabidopsisand an extra WKKY group. In contrast, Guo et al.and Liu et al.just distributed WRKY genes into group I, II and III. Among 90 HaWRKY family members, group II accounts for the largest part with 48 HaWRKY proteins, followed by group I with 18 proteins and group III with 17 proteins [fig_ref] Fig 1: Phylogenetic relationships of WRKY genes from common sunflower and Arabidopsis [/fig_ref]. There were 7 HaWRKY proteins in group WKKY, with C-X 5 -C-X 23 -H-X-H type zinc-finger motifs [fig_ref] Fig 2: Alignment of multiple HaWRKY protein sequences [/fig_ref] , which is not found in Arabidopsis [fig_ref] Fig 1: Phylogenetic relationships of WRKY genes from common sunflower and Arabidopsis [/fig_ref]. In addition, each group could be divided into several subgroups. Proteins with two WRKY domains were assigned as the N-terminal and the C-terminal WRKY domains according to their locations on protein. The proteins grouped either in N-terminal or C-terminal WRKY domains, usually followed by C 2 H 2 -type zinc-finger motifs (C-X 4 -C-X 22-23 -H-X-H), were classified as group I, with 16 identified as N-terminal WRKYs (I N) and 15 C-terminal (I C), and among them 13 members contained two WRKY domains [fig_ref] Fig 2: Alignment of multiple HaWRKY protein sequences [/fig_ref]. Group II of HaWRKY family could be clustered into five subgroups, with 4 in IIa, 10 in IIb, 13 in IIc, 11 in IId and 10 in IIe [fig_ref] Fig 1: Phylogenetic relationships of WRKY genes from common sunflower and Arabidopsis [/fig_ref]. 17 members of HaWRKYs in group III contained the C-X 7 -C-X 23 -H-X-C type zincfinger motifs [fig_ref] Fig 2: Alignment of multiple HaWRKY protein sequences [/fig_ref] , and were classified as subgroup IIIa. There were no HaWRKY proteins found in subgroup IIIb, not as that in Arabidopsis [fig_ref] Fig 1: Phylogenetic relationships of WRKY genes from common sunflower and Arabidopsis [/fig_ref]. ## Gene structure and motif composition of wrky family members The distributions of exons and introns on HaWRKY genes were investigated via GSDS program, to gain further insight into the structure diversity of the WRKY family in sunflower. As shown in [fig_ref] Fig 3: Phylogenetic relationships, gene structures and motif compositions of WRKY genes in common... [/fig_ref] almost half number (43) of the HaWRKY genes had three exons, followed by 17 with two exons, 17 with four exons, 12 with five exons and 1 with seven exons. Genes within same groups generally shared similar structures, such as group IIIa, in which all HaWRKY genes possessed three exons and two introns. Most of the WRKY domains spanned an exonexon junction, whereas HaWRKY genes with two WRKY domains in the group I at least had one complete domain within one exon, except HaWRKY1. Further analyses on introns indicated that HaWRKY genes only with phase-0 introns (between two consecutive codons) were clustered into group IIa and IIb, and only with phase-2 introns (between the second and third nucleotide of a codon) into the group IId, IIe and IIIa. The phase-1 introns (between the first and second nucleotide of a codon) were widely distributed among these groups, except group IIa. Motif structures on HaWRKY proteins were constructed via MEME program. As exhibited in [fig_ref] Fig 3: Phylogenetic relationships, gene structures and motif compositions of WRKY genes in common... [/fig_ref] ## Evolution of group iii hawrky genes In order to understand the evolution of common sunflower group III WRKY genes, a phylogenetic tree of group III WRKY proteins from two monocots (rice and maize) and three dicots (sunflower, Arabidopsis and grape) was constructed. All the group III WRKY family members were divided into 10 clades as shown in [fig_ref] Fig 4: Phylogenetic relationships and motif compositions of group III WRKY proteins from five... [/fig_ref] WRKY proteins from closer species were clustered into same clades. Most proteins from dicots gathered in clade 1 and 3, whereas monocots MEME analysis was also conducted to search the conserved motifs of group III WRKY proteins from five species. Proteins within same clades usually displayed similar motif structures, indicating potential functional similarities among WRKY proteins. Motifs 1 and 7 were WRKY domains. Interestingly, motif 1 was found in all clades, whereas motif 7 was unique to clades 9 and 10, two clades only containing rice proteins, implying that motif 1 might have common function among different species, while motif 7 might play specific roles in rice and contribute to the divergence of group III WRKY genes. Motifs 1, 10 and 18 were specific to dicots. In contrast, motifs 12 and 19 were only observed in monocots. These motifs might be also important to the divergence of WRKY genes. ## Chromosomal location and synteny of hawrky genes HaWRKY genes are distributed unevenly on 17 chromosomes (S2 [fig_ref] Fig 5: Genome localization and synteny analyses of WRKY genes within common sunflower, and... [/fig_ref]. Chromosomes Ha10 and Ha15 both have 13 HaWRKY genes as the largest groups, whereas there was no HaWRKY gene observed on chromosomes Ha2. No correlation between chromosome length and HaWRKY gene number could be determined. Two or more genes located within 200 kb on same chromosome is defined as a tandem duplication eventHaWRKY50/53) are also identified [fig_ref] Fig 4: Phylogenetic relationships and motif compositions of group III WRKY proteins from five... [/fig_ref]. These results indicated that tandem and segmental duplication possibly contributes to the diversity and expansion of HaWRKY gene families. The dN/dS ratio of these duplicated gene pairs were calculated to understand the evolutionary constraints. The synonymous substitution rates (dS) of all segmental and tandem duplicated HaWRKY gene pairs were higher than non-synonymous substitution rate (dN) as shown in [fig_ref] Table 1: dN/dS analyses for the duplicated WRKY gene pairs of sunflower [/fig_ref] , indicating that HaWRKY gene family probably went through strong purifying selection during evolution. Dual syntenies of common sunflower with Arabidopsis and rice were also conducted. A total of eight HaWRKY genes showed syntenic relationship with those in Arabidopsis, composing 9 orthologous pairs, whereas only one HaWRKY gene was collinear with one in rice [fig_ref] Fig 5: Genome localization and synteny analyses of WRKY genes within common sunflower, and... [/fig_ref]. Similarly, more collinear gene pairs were observed between sunflower and Arabidopsis than rice, as sunflower is phylogenetically closer to Arabidopsis. HaWRKY16 was associated with two Arabidopsis genes and HaWRKY25 and HaWRKY65 are syntenic with a same Arabidopsis gene. HaWRKY25 is also found to be syntenic with a rice gene, indicating that these orthologous pairs might occur before the divergence of monocots and dicots. ## Transcriptomic pattern of hawrky genes from different tissues The transcriptome data of HaWRKY genes of different sunflower tissues were downloaded from Sunflower Genome Database. 20 of the 90 identified HaWRKY genes weren't expressed in all ten tissues [fig_ref] Fig 6: Expression profile of WRKY genes in different tissues of common sunflower [/fig_ref] , which might be pseudogenes, have special temporal and spatial expression patterns or express in other tissues. The expression patterns of HaWRKY genes in sunflower were organ-specific, as bract, corolla, ligule, ovary, seed and stamen, which are related to flower, were clustered into a big group, and leaves and stem were in another group [fig_ref] Fig 6: Expression profile of WRKY genes in different tissues of common sunflower [/fig_ref]. Most of the HaWRKY genes didn't express in pollen [fig_ref] Fig 6: Expression profile of WRKY genes in different tissues of common sunflower [/fig_ref]. In general, the expression levels of HaWRKY genes in bract, ligule, leaves and stem were higher than that in other tissues [fig_ref] Fig 6: Expression profile of WRKY genes in different tissues of common sunflower [/fig_ref]. HaWRKY17/22/79/81 displayed highest transcript abundances across all tissues except pollen and were clustered into a group, whereas the expression levels of HaWRKY23/31/37/40/68/84 were extremely low in all tested tissues [fig_ref] Fig 6: Expression profile of WRKY genes in different tissues of common sunflower [/fig_ref]. The expression patterns of some genes were tissue-specific, for example, HaWRKY73 was abruptly induced only in leaves, HaWRKY3 in stem, HaWRKY11 in style, etc. ## Profiles of hawrky genes under abiotic and biotic stress Twenty-three HaWRKY genes which were highly induced in different tissues of common sunflower (except in pollen) were selected to test the reactions of different WRKY genes to different abiotic stresses. Generally, HaWRKY genes were inhibited in sunflower leaves after treatment of PEG and NaCl with different concentrations, whereas HaWRKY29/30 at 150 mM NaCl and HaWRKY48/89 at 20% PEG were significantly (P < 0.05) up-regulated by 46%/ 140% and 70%/51% compared with control, respectively. But no any significant (Tukey, P<0.05) changes of HaWRKY55/57 were observed in response to abiotic stresses in all comparisons [fig_ref] Fig 7: Expression profile of 23 selected HaWRKY genes in responses to treatments of... [/fig_ref]. Similarly, in roots as shown in [fig_ref] Fig 7: Expression profile of 23 selected HaWRKY genes in responses to treatments of... [/fig_ref] HaWRKY52 and HaWRKY89 were up-regulated by PEG, and 10% PEG significantly (P < 0.05) increased by 152% and 179% compared with control. In contrast, Most of the HaWRKY genes were significantly (P < 0.05) up-regulated after treatment with NaCl in roots, as compared to the control. Among them, the transcript levels ofIn order to understand the role of sunflower WRKY gene family against biotic stress, transcription levels of two contrasting common sunflower cultivars (TK0409, susceptible; JY207, [fig_ref] Fig 8: Expression profile of HaWRKY genes in response to infection of parasitic weed... [/fig_ref] , suggesting these genes might partly contribute to the resistance of sunflower against O. cumana. # Discussion The WRKY transcription factor family is considered to be involved in diverse stress responses, developmental and physiological processes in plants. Systematical characterization of WRKY genes in several species has been studied, including Arabidopsis, rice, tomato, maize etc. Sunflower WRKY genes have been well characterized benefiting from the release of its reference genome in previous studies. However, there are two sunflower genome database available ("HanXRQr1.0" and "HA412.v1.1.bronze") assembled from different sunflower genotypes. Guo et al.and Liu et al.have used the "HanXRQr1.0" database for retrieval of WRKY genes. In our study, we used "HA412.v1.1.bronze" database to search WRKY genes as support and addition to the previous works. Indeed, we found 89 WRKY genes which have corresponding genes in the results of Guo et al.and Liu et al., and we discovered another WRKY gene (HaWRKY51) which was neglected in their works. Multiple protein sequence alignments revealed domain variations in common sunflower WRKY family. In comparison to the results of Guo et al.and Liu et al.that classify sunflower WRKY proteins into three normal groups, an extra WKKY group with 7 proteins was identified in our study based on the phylogenetic analysis, including HaWRKY 60 with WKKYGQK, HaWRKY33/50/53/54 with WKKYGEK, and HaWRKY51 with WKKYGKK [fig_ref] Fig 2: Alignment of multiple HaWRKY protein sequences [/fig_ref]. Interestingly, this group has been also found in Helianthus exilis, Helianthus petiolaris and Helianthus tuberosus, indicating that the WKKY variation is common in the Asteraceae. In addition, WKKYGQK, WKKYGKK and WKKYGEK are also observed in different legumes, but with low frequencies. However, there are no more reports about WKKY group in other plant species. Although the WRKYGQK is highly conserved in most WRKY domains, variation in the core sequence has been documented. In our studies, mutations happened to R and Q sites, while the others were conserved. WRKYGKK and WRKYGEK are the most frequently occurring variants of the core sequence in most plant species. As the WRKYGQK core sequence can interact with the W-box to activate downstream genes, the variations in this motif might influence the function of downstream target genes. Thus, further investigations on functions and binding specificities of these sunflower proteins with mutated WRKY motifs might provide deep insight into this transcription factor family. A domain loss is common in the WRKY gene family in plants, which is recognized as a divergent force for expansion of this gene family. In the current study, 5 domain loss events were found in group I, suggesting a potential cause of the diversity of WRKY genes in this group. Tandem and segmental duplication events also played a pivotal role in the expansion of WRKY gene family. Five pairs of tandem and four pairs of segmental duplicated genes were identified in the present study, with five pairs in group IIIa, two pairs in group IId, one pair in group IIb and one pair in group IIc. This result indicated that tandem and segmental duplication events might contribute to the amplification of sunflower WRKY genes in these groups, as compared to those of Arabidopsis. The origin of WRKY genes from group III appears to have occurred prior to the divergence of monocots and dicots, and then numerous duplications and diversifications happened after that event. In order to explore how the WRKY group III gene family evolved, a phylogenetic tree of WRKY group III proteins from sunflower with two dicots (Arabidopsis, grape) and two monocots (rice, maize) was constructed, which divided the 17 group III HaWRKYs into three clades. WRKY proteins from closer species appeared to be clustered together. Both monocots and dicots proteins occurred in many clades, suggesting group III WRKY genes diversified before the monocot-eudicot split. In addition, clade 7 contained group III WRKY proteins from all 5 species, which tended to form monocot-and dicot-specific subclades, implying that group III WRKY genes evolved separately after the divergence of monocots and dicots. It is well known that WRKY genes play essential roles in plant growth and development. Li et al.reported that AtWRKY13 functioned in stem development, as a weaker stem phenotype was observed and lignin-synthesis-related genes were repressed in Arabidopsis wrky13 mutants. In the current study, the orthologous of AtWRKY13 in sunflower, HaWRKY28 displayed high expression levels in stem, indicating that these genes might also act in stem development in sunflower. Overexpression of WRKY15 exhibited an increased leaf area of Arabidopsis, which implied that AtWRKY15 seemed to be involved in leaf growth. HaWRKY79, which is the orthologous of AtWRKY15, were highly induced not just in leaves, but across all tissues, indicating that this gene might be constitutive in sunflower plant growth and development. In contrast, HaWRKY7 was specifically expressed in leaves, suggesting its role in sunflower leaf growth. Among all WRKY genes in sunflower, only HaWRKY30 displayed a high level of expression in pollen, with other WRKY genes extremely low expressed. Interestingly, two pollen-specific regulators in Arabidopsis, AtWRKY34 and AtWRKY2, have phylogenetically close relationship with this sunflower WRKY gene, indicating that HaWRKY30 might be associated with pollen developmental modulation. In addition to their role in plant growth and development, WRKY TFs also play pivotal roles in various stress responses, providing an important basis for genetic improvement of crops. Drought and salinity, both of which can cause plant cellular dehydration, are two major constraints to sunflower production. Responses of plants to drought and salinity usually result in accumulation of reactive oxygen species (ROS) and abscisic acid (ABA), which activate downstream WRKY genes. Overexpression of a membrane-localized cysteine-rich receptor-like protein kinase, CRK5 in Arabidopsis, led to increase of ABA sensitivity and promotion of stomatal closure, and subsequent enhancement of plant drought tolerance. Knockout of AtWRKY18, AtWRKY40 and AtWRKY60 significantly increased the expression of CRK5, suggesting negative regulation of these three genes on CRK5. In our study, the relative expression levels of two orthologous of AtWRKY40, HaWRKY74 and HaWRKY81, were recorded in sunflower roots and leaves. Expression levels of both two genes decreased as the concentrations of PEG increased in sunflower leaves, while in sunflower roots, two genes were induced under low concentration of PEG and inhibited under high concentration. It has also been reported that AtWRKY46, AtWRKY54, and AtWRKY70 are implicated in promotion of BR-regulated plant growth and inhibition of drought response, as reduced BR-regulated growth and higher survival rates under drought stress was observed in their triple mutant. HaWRKY9 and HaWRKY22, which were phylogenetically close to AtWRKY46, were both repressed under PEG treatments in sunflower roots and leaves. These results are suggesting that sunflower probably enhanced drought tolerance via down-regulating specific WRKY genes and subsequently activating downstream signal pathways. The increase of ABA level caused by drought usually induces high expression of AtWRKY57, which binds to W-box in the promoter region of the downstream response genes. HaWRKY57, the orthologous of AtWRKY57, displayed a high expression level under PEG treatment in sunflower roots. Interestingly, the increase of ROS level caused by salinity also activates AtWRKY57, and consistently, HaWRKY57 was highly expressed under treatment of 300 mM NaCl in sunflower roots. These results indicated that HaWRKY57 might share similar functions with AtWRKY57 in sunflower under drought and salinity. AtWRKY15 is another WRKY gene induced by ROS, but will make Arabidopsis more susceptible to osmotic stress and oxidative stress. In our study, HaWRKY79, the orthologous of AtWRKY15, was significantly suppressed in sunflower roots under treatment of NaCl, implying their similar roles in conferring salt tolerance. The parasitic weed Orobanche cumana is a new emerged threat to sunflower production worldwide. Previous studies proposed that O. cumana deployed effectors in sunflower to suppress host defense responses and resistant sunflower cultivars recognized effectors with the help of R proteins to activate effector-triggered immunity. WRKY family has been found to be involved in the microbe-associated molecular pattern-triggered immunity, PAMP-triggered immunity or effector-triggered immunity # Conclusion In this study, we identified 90 WRKY genes from Helianthus annuus L. and characterized their structure, duplication, chromosomal distribution, phylogenetic tree, followed by tissue-differential gene expression and differential expression in response to biotic and abiotic stress. Supporting information S1 [fig] Fig 1: Phylogenetic relationships of WRKY genes from common sunflower and Arabidopsis. The different-colored braches indicate different groups (or subgroups). The red solid circles and blue solid squares represent WRKY genes from Arabidopsis and common sunflower, respectively. https://doi.org/10.1371/journal.pone.0241965.g001 in clade 2, 4, 5, 6, 8, 9 and 10. Clade 7 contained proteins from all 5 species, indicating these proteins might be orthologues from a single ancestral gene. [/fig] [fig] Fig 2: Alignment of multiple HaWRKY protein sequences. "N" and "C" indicate the N-and C-terminal of WRKY domains. https://doi.org/10.1371/journal.pone.0241965.g002 [/fig] [fig] Fig 3: Phylogenetic relationships, gene structures and motif compositions of WRKY genes in common sunflower. (A) Phylogenetic tree of WRKY genes. The different-colored braches indicate different groups (or subgroups). (B) Exon-intron structures of WRKY genes. Blue boxes indicate 5' and 3' UTRs. Yellow boxes indicate exons. Black lines indicate introns. Red boxes indicate WRKY domains. The numbers indicate the phases of introns. (C) Motif compositions of WRKY proteins. Different motifs are displayed with different colored boxes. https://doi.org/10.1371/journal.pone.0241965.g003 [/fig] [fig] Fig 4: Phylogenetic relationships and motif compositions of group III WRKY proteins from five monocot and dicot plants. On the left side, proteins are clustered into 10 clades, marked with different colors. On the right side, different motifs are displayed with different colored boxes. https://doi.org/10.1371/journal.pone.0241965.g004 [/fig] [fig] Fig 5: Genome localization and synteny analyses of WRKY genes within common sunflower, and between common sunflower and two representative plant species. (A) Chromosomal distribution and interchromosomal relationships of common sunflower WRKY genes. Gray lines indicate all sytenic gene pairs in common sunflower genome and red lines indicate duplicated WRKY gene pairs. (B-C) Synteny analyses of common sunflower WRKY genes with Arabidopsis and rice, respectively. Gray lines indicate collinear gene pairs between common sunflower and other plant genomes and red lines indicate syntenic WRKY gene pairs. https://doi.org/10.1371/journal.pone.0241965.g005 [/fig] [fig] Fig 6: Expression profile of WRKY genes in different tissues of common sunflower. (A) Hierachical clustering of expression profile of WRKY genes from different tissues. Data were transformed with a log 2 (FPKM+1) transformation. (B) Boxplot of expression levels of WRKY genes in different tissues. https://doi.org/10.1371/journal.pone.0241965.g006 [/fig] [fig] Fig 7: Expression profile of 23 selected HaWRKY genes in responses to treatments of PEG and NaCl. (A) Expression profile of WRKY genes in common sunflower leaves. (B) Expression profile of WRKY genes in common sunflower roots. https://doi.org/10.1371/journal.pone.0241965.g007 [/fig] [fig] Fig 8: Expression profile of HaWRKY genes in response to infection of parasitic weed Orobanche cunama. JY represents resistant sunflower cultivar JY207; JO represents JY207 with infection of O. cumana; TK represents susceptible sunflower cultivar TK0409; TO representsTK0409 with infection of O. cumana. https://doi.org/10.1371/journal.pone.0241965.g008 [/fig] [table] PLOS ONE: | https://doi.org/10.1371/journal.pone. [/table] [table] Table 1: dN/dS analyses for the duplicated WRKY gene pairs of sunflower. [/table] [table] Table: Primers for qRT-PCR. (DOCX) S2Table. Basic information regarding the presence of WRKY genes in sunflower. (DOCX) S3 Table. Corresponding gene names. (DOCX) Project administration: Weijun Zhou, Ling Xu. Software: Juanjuan Li, Qian Huang, Jian Wang, Chong Yang. Validation: Juanjuan Li, Faisal Islam. Visualization: Juanjuan Li, Qian Huang, Jian Wang. Writing -original draft: Juanjuan Li. Writing -review & editing: Faisal Islam, Weijun Zhou, Ling Xu, Chong Yang. [/table]

https://www.semanticscholar.org/paper/aa2cdd1cea9b4c7c995819e6fad8d2b55df0e820

Intravascular Fasciitis in the Femoral Vein with Hypermetabolic Signals Mimicking a Sarcoma: The Role of Preoperative Imaging Studies with Review of Literature

Vascular Specialist International

3ec9fff7-a664-4723-b15c-3251b0aeaeb7

Intravascular Fasciitis in the Femoral Vein with Hypermetabolic Signals Mimicking a Sarcoma: The Role of Preoperative Imaging Studies with Review of Literature Kim et al. Intravascular Fasciitis in FV Kim et al. Intravascular Fasciitis in FV

https://www.semanticscholar.org/paper/d2b8aa1a8380e9519cb40e397945e95922247ffc

Highly selective determination of dopamine in the presence of ascorbic acid and serotonin at glassy carbon electrodes modified with carbon nanotubes dispersed in polyethylenimine.

Journal of Nanoscience and Nanotechnology

a90bff56-f708-4cd5-a11e-ea7f9bdcf6fa

Highly selective determination of dopamine in the presence of ascorbic acid and serotonin at glassy carbon electrodes modified with carbon nanotubes dispersed in polyethylenimine. We report the highly selective and sensitive voltammetric dopamine quantification in the presence of ascorbic acid and serotonin by using glassy carbon electrodes modified with a dispersion of multiwall carbon nanotubes (MWCNT) in polyethylenimine, PEI (GCE/MWCNT-PEI). The electrocatalytic activity of the MWCNT deposited on the glassy carbon electrode has allowed an important decrease in the overvoltages for the oxidation of ascorbic acid and dopamine, making possible a clear definition of dopamine, serotonin and ascorbic acid oxidation processes. The sensitivities for dopamine in the presence and absence of 1.0 mM ascorbic acid and serotonin were (2 18 ± 0 03 × 10 5 AM −1 (r = 0 9998); and (2 10 ± 0 07 × 10 5 AM −1 (r = 0 9985), respectively, demonstrating the excellent performance of the GCE/MWCNT-PEI. The detection limit for dopamine in the mixture was 9 2 × 10 −7 M. The R. S. D. for the determination of 50 M dopamine using four different electrodes was 3.9% when modified with the same MWCNT/PEI dispersion, and 4.6% when using four different dispersions. The modified electrode has been successfully applied for recovery assays of dopamine in human blood serum. Therefore, the new sensor represents an interesting and promising alternative for the electrochemical quantification of neurotransmitters and other analytes of clinical interest. # Introduction Dopamine (DA) is an important neurotransmitter present in the central nervous system of mammalian. It plays a very important role in the functioning of central nervous, hormonal, cardiovascular and renal systems. Several alterations in dopaminergic transmission have been related to severe problems like Parkinson disease, schizophrenia and drug dependence.Therefore, sensitive and selective strategies to determine DA are highly required. In this sense, electroanalytical techniques have been widely used for the quantification of DA due to their known advantages and the easy oxidation of DA at different electrodes, mainly involving carbon materials.The major drawback of the electrochemical determinations of DA is the interference of easily oxidizable compounds usually present in nervous systems, like ascorbic acid (AA) and serotonin (SER) that are oxidized at * Author to whom correspondence should be addressed. potentials close to those of DA at most of the common electrodes. Several attempts have been made to circumvent this problem. The modification of electrodes with permselective layers has demonstrated to be very successful.The use of electrochemical sensors based on carbon nanotubes (CNT) represents another very advantageous alternative.Zhang et al.have proposed the use of a glassy carbon electrode (GCE) modified with poly(styrene sulfonic acid)sodium salt/single wall carbon nanotubes for the determination of DA in the presence of excess of AA based on the different electrostatic interaction of these compounds with the polyelectrolyte. Shervedani et al.have reported the highly selective quantification of DA at micromolar levels in the presence of 1.0 mM AA using a gold electrode modified with cysteamine. Zhu et al.have proposed the determination of DA in the presence of uric acid (UA) and AA using GCE modified with 1-octyl-3-methylimidazolium hexafluorophosphate containing multi-wall carbon nanotubes (MWCNT). The selective determination of DA in the presence of AA using a GCE modified by assembling of poly(diallyldimethylamonium chloride) and negatively charged-shortened-MWCNT has also been demonstrated.Hu et al.have proposed the use of a GCE modified with a dispersion of MWCNTs in water in the presence of dihexadecyl/hydrogen phosphate for the simultaneous determination of DA and serotonin (SER) without interference of AA. They have reported that the fast adsorption of DA and SER on the modified GCE makes possible a very sensitive determination of DA (D.L. 1 1 × 10 −8 M) and SER (D.L. 5 × 10 −9 M) and the quantification of SER in human blood serum after spiking DA and SER. Xie et al.have proposed the use of gold electrodes modified with overoxidized polypyrrole-multiwalled carbon nanotubes nanocomposite film for the nanomolar detection of DA by Differential Pulse Voltammetry. Chen et al.have reported the advantages of using a conductive composite film containing MWCNT with poly(methylene blue) synthetized on different electrodes for the simultaneous detection of DA and AA, with detection limit of 67 M DA. The successful use of a platinum electrode modified with single wall carbon nanotubes (SWCNT) and phitic acid to determine DA in the presence of AA and UA, with detection limit of DA of 0.08 M, has been proposed.Glassy carbon electrodes modified with a dispersion of MWCNTs in polyacrylic acid have been also exploited for the highly sensitive and selective quantification of DA (20 nM) and UA (110 nM) in the presence of 0.3 mM AA.Shahrokhian et al.have reported the highly sensitive (detection limit 0.08 M) and selective determination of DA in the presence of AA at carbon paste electrode modified with thionine-Nafion supported on MWCNT. Recently, we have proposed the highly efficient dispersion of MWCNT in the polycation polyethylenimine (PEI) and the excellent performance of glassy carbon electrodes modified with this dispersion as amperometric detector not only in batch 20 but also in Flow Injection Analysis and Capillary Electrophoresis.The irreversible PEI adsorption onto the sidewalls of CNT by wrapping around them is responsible for the highly efficient n-doping of SWC-NTs due to the electron donating ability of the large numbers of amine groups present in the polymer.In this work we propose the use of GCE modified with MWCNTs dispersed in PEI as a sensing layer for the highly selective dopamine quantification even in the presence of a large excess of AA and SER. ## Experimental details ## Reagents Hydrogen peroxide (30% v/v aqueous solution) was purchased from Baker. Ascorbic acid (AA) was obtained from Fluka. Dopamine (3,4-dihydroxyphenethylamine, DA), serotonin (5-hydroxy tryptamine, SER), and polyethylenimine (PEI, Average MW 750,000, Catalog number P-3143) were purchased from Sigma. Multi-walled carbon nanotubes powder (MWCNT, 20-50 nm diameter, 5-20 microns length) was obtained from NanoLab (USA). Lyophilized human blood serum (Standatrol) was provided by Wienner. Other chemicals were reagent grade and used without further purification. Ultrapure water ( = 18 M cm) from a Millipore-MilliQ system was used for preparing all the solutions. A 0.050 M phosphate buffer solution pH 7.40 was employed as supporting electrolyte. ## Apparatus The measurements were performed with an EPSILON potentiostat (BAS). The electrodes were inserted into the cell (BAS, Model MF-1084) through holes in its Teflon cover. A platinum wire and Ag/AgCl, 3 M NaCl (BAS, Model RE-5B) were used as counter and reference electrodes, respectively. All potentials are referred to the latter. A magnetic stirrer provided the convective transport during the amperometric measurements. Scanning Electronic Microscopy (SEM) images were obtained with a Hitachi S3000N Microscope equipped with secondary and backscattered electron detectors. IR experiments were performed with a Nicolet 5-SXC FT-IR spectrometer. ## Preparation of gce/(mwcnt/pei) Preparation of MWCNT-PEI dispersion: It was obtained by dispersing 1.0 mg of MWCNTs within 1.0 mL of 1.0 mg/mL PEI solution (prepared in 50:50 v/v ethanol/water) followed by sonication for 15 min. ## Preparation of the glassy carbon electrode modified with the mwcnt-pei dispersion (gce/(mwcnt-pei)): The glassy carbon electrode was previously polished with alumina slurries of 1.0, 0.30 and 0.05 m for 2 min each and sonicated in water for 30 s. Then it was modified with the MWCNT-PEI dispersion (GCE/(MWCNT-PEI)) in the following way: an aliquot of 20 L was dropped on top of a polished GCE and then the electrode was left in the air to allow the solvent to evaporate at room temperature (approximately 90 min in our experimental conditions). Preparation of the glassy carbon electrode modified with PEI solution (GCE/PEI): It was prepared by dropping 20 L of 1.0 mg/mL PEI solution (prepared in 50:50 v/v ethanol/water) on polished and clean GCE. ## Preparation of the samples for ir CNTs dispersions for IR were prepared in ethanol/water 50/50 V/V and sonicated for 15 min. CNT/PEI dispersions were prepared as indicated above. ## Procedure Cyclic voltammetric experiments were performed at 0.100 Vs −1 . Differential pulse voltammograms (DPV) were obtained with a pulse height of 0.040 V and pulse duration of 200 ms. All the experiments were conducted at room temperature. # Results and discussion ## Characterization of mwcnt-pei dispersion and GCE/(MWCNT-PEI) [fig_ref] Figure 1: FT-IR spectra for unmodified MWCNT [/fig_ref] illustrates FT-IR spectra for unmodified MWCNT (A) and MWCNT dispersed in PEI (B). The spectrum for unmodified MWCNT shows a moderate to weak absorption between 1667 and 1640 cm −1 , region assigned to C C stretching of phenyl ring vibrations. These peaks are originated from the underlying CNTs structure and can be observed in all CNTs spectra 23 (The spectrum of was magnified in order to show more clearly the different contributions.). On the contrary, the spectrum corresponding to CNTs dispersed in PEI shows several additional peaks as a consequence of the presence of basic nitrogen-containing groups (e.g., amine) on the CNTs. These peaks are due to the bending vibration of N-H (1648, 1567 and 1475 cm −1 , stretching vibration of C-N (1318, 1100 and 1022 cm −1 . Medium to strong absorption was observed between 909-666 cm −1 , due to NH wagging. These functional groups are hydrophilic and make possible the easy dispersion of MWCNTs in aqueous medium. Absorption peaks at 2946 cm −1 and 2850 cm −1 can be assigned to aliphatic C-H stretching frequencies of the ethylene group of PEI. Therefore, the results confirm that PEI is effectively wrapped on the MWCNTs walls upon ultrasonication. [fig_ref] Figure 2: SEM picture obtained for a GCE modified with the dispersion of CNTs... [/fig_ref] shows a SEM image of a GCE coated with the MWCNT/PEI dispersion obtained at 50,000×. The picture reveals that MWCNTs are efficiently entrapped within the polymeric matrix and distributed in a homogeneous way on the glassy carbon surface. A quick and efficient way to demonstrate the catalytic activity of MWCNTs dispersed in the polymeric film and deposited on the glassy carbon surface is the evaluation of the voltammetric behavior of hydrogen peroxide at the resulting electrode. shows the voltammetric response of 50 mM hydrogen peroxide at different electrodes: bare glassy carbon electrode (GCE) (a), GCE modified with PEI (GCE/PEI) (b), and GCE modified with the MWCNT-PEI dispersion (GCE/MWCNT-PEI) (c). In a way similar to that observed at GCE (a), elevated overvoltages are necessary to oxidize and reduce hydrogen peroxide at GCE/PEI (b). On the contrary, when MWCNTs are present at the electrode (GCE/MWCNT-PEI), an important decrease in the oxidation and reduction overvoltages is obtained as a consequence of the electrocatalytic activity of MWCNT, in agreement with previous results.As it was previously reported, the MWCNT-PEI dispersion is highly stable. Electrodes prepared with the same dispersion even 14 days after the first day, gave the same response as the one obtained the first day of the dispersion.and GCE/MWCNT-PEI (c). A decrease of 90 mV in the peak potential separation and an increase of around 8-fold in the oxidation current is observed at GCE/MWCNT-PEI. A better definition of the dopaminequinone reduction peak and associated redox processes is also evident. depicts cyclic voltammograms for AA at the three different electrodes. A shifting of 140 mV in the negative direction is observed for the oxidation peak potential at GCE-PEI (dashed line), due to the favored electrostatic interaction of ascorbate with the polycation. When MWCNT are present, a shifting of 490 mV in the negative direction for AA oxidation peak potential and an increase of around 50% in the current are observed (compared to GCE). Cyclic voltammograms for SER obtained at the different electrodes are shown in . At bare GCE the voltammogram shows an oxidation peak at 0.366 V, with no reduction peak. In the presence of PEI, the oxidation peak potential remains almost constant while the current decreases (34.3 vs. 24.5 A at GCE and GCE/PEI, respectively) due to some electrostatic repulsion with the positively charged PEI. In the presence of MWCNT the oxidation peak potential decreases 20 mV and the current largely increases (34.3 vs. 351.8 A at GCE and GCE/MWCNT-PEI, respectively). In summary, the modification of GCE with MWCNT-PEI largely improves the electro-oxidation of AA, DA and SER, even when the capacitive currents are higher than those at GCE and GCE/PEI. This behavior can be attributed to the electrocatalytic activity of CNTs efficiently dispersed in PEI and successfully deposited on the surface of GCE and to the significant increase in the surface area of the resulting electrodes. . As expected, a broad peak is obtained at the bare electrode, indicating that is not possible to detect a mixture of these compounds. [fig_ref] Figure 5: A [/fig_ref] displays the voltammetric response for the same mixture (1.0 mM DA and 1.0 mM AA) at GCE/MWCNT-PEI. Two very well-defined DPV peaks at −0.108 V and 0.128 V are observed for the oxidation of AA and DA, respectively. Therefore, the catalytic activity of MWCNT allows a clear definition of the two oxidation processes as well as a significant enhancement in the associated currents, ensuring, in that way, the individual determination of the two compounds. At GCE/PEI the signals are smaller and the resolution is very poor (not shown). ## Simultaneous determination of da and aa Considering the interesting results shown in [fig_ref] Figure 5: A [/fig_ref] , we evaluate the feasibility to quantify DA in the presence of a large excess of AA using GCE/MWCNT-PEI. [fig_ref] Figure 5: A [/fig_ref] , two clearly separated peaks are obtained at −0.120 V and 0.090 V for AA and DA, respectively. A slight shifting in the positive direction is observed for DA oxidation peak potential as the concentration increases, while the peak potentials and currents for the oxidation of AA remain constant. GCE/MWCNT-PEI in the presence (full circles) and in the absence (empty circles) of SER and AA. The sensitivities for DA in the presence and absence of AA and SER are (2 18 ± 0 03 × 10 5 AM −1 r = 0 9998); and (2 10 ± 0 07 ×10 5 AM −1 r = 0 9985), respectively, representing a difference in sensitivities in the absence and presence of AA and SER just of 3.7%. As it is also shown in, the signals for AA (squares) and SER (triangles) remain constant in the presence of the different concentrations of DA, indicating the feasibility of the method to determine DA in such a complex mixture. The detection limit for DA in the mixture, obtained as 3.3 times the ratio between standard deviation of the blank and sensitivity, was 9 2 × 10 −7 M. As it was shown in calibration plots shown in, the analytical methodology was highly reproducible. The R.S.D. for the determination of 50 M DA using four different electrodes modified with the same dispersion was 3.9%. Analogous The usefulness of the proposed methodology for determinations in human blood serum was also evaluated. [fig_ref] Figure 9: Differential pulse voltammograms for undiluted blood serum sample before [/fig_ref] shows DPVs for undiluted human blood serum before (a) and after (b) the addition of 100 M AA + 100 M DA + 100 M SER. Experiments performed by separate additions of AA, DA and SER to the blood serum sample demonstrated that the oxidation of these compounds occurred at −0.09 V, 0.06 V and 0.22 V, respectively (not shown). Therefore, the contributions of AA, DA and SER can be clearly distinguished, making possible the determination of these compounds even in a matrix as complex as blood serum sample. Separate additions of uric acid to the buffer solution or serum sample demonstrated that the peak at 0.13 V can be attributed to the oxidation of uric acid present in the serum sample (not shown). Additions of 50 M and 100 M DA to undiluted serum samples gave recoveries of 104% and 99%, respectively. Adsorptive stripping experiments of DA at CNTmodified-GCE (5 min accumulation at −0.250 V) with medium exchange demonstrated that DA can be preconcentrated at the CNTs-modified GCE, although no improvement in sensitivity was obtained under these conditions compared to the direct DA determination (not shown). # Conclusions In summary, the MWCNT/PEI film-coated GCE exhibits remarkable electrocatalytic effects on the oxidation of DA, AA and SER, improving their oxidation peak currents and lowering their oxidation overpotentials. At variance with GCE, the GCE modified with MWCNT-PEI film permits a very favorable voltammetric resolution of DA, AA and SER oxidation processes. The GCE/MWCNT-PEI demonstrated to be highly efficient in detecting DA in the presence of large excess of AA and SER. These characteristics make GCE/MWCNT-PEI a novel analytical tool for the selective and sensitive quantification of DA in the presence of AA and SER, and open the doors to new challenges in the electroanalytical determination of other neurotransmitters and further practical applications. [fig] Figure 1: FT-IR spectra for unmodified MWCNT (A) and MWCNT-PEI (B) (The amount of MWCNT to get the spectrum is higher than that of MWCNT-PEI.). [/fig] [fig] Figure 2: SEM picture obtained for a GCE modified with the dispersion of CNTs in PEI. Magnification: 50,000×. [/fig] [fig] Figure 3, Figure 4, Figure 4: GCE/MWCNT-PEI also demonstrated to be highly stable in flow experiments performed either through Cyclic voltammogram for 50 mM hydrogen peroxide at different electrodes: GCE (a), GCE-PEI (b), and GCE/MWCNT-PEI (c). Scan rate: 0.100 Vs −1 . Supporting electrolyte: 0.050 M phosphate buffer solution pH 7.40. shows cyclic voltammograms for 1.0 mM DA (A), AA (B) and SER (C) at GCE (a), GCE/PEI (dashed line, b) Cyclic voltammograms for 1.0 mM dopamine (A), ascorbic acid (B) and serotonin (C) at different electrodes: GCE (a), GCE-PEI (b), and GCE/MWCNT-PEI (c). Scan rate: 0.100 Vs −1 . Supporting electrolyte: 0.050 M phosphate buffer solution pH 7.40. [/fig] [fig] Figure 5: A) shows DPVs for a mixture of 1.0 mM DA and 1.0 mM AA at bare GCE (A), and GCE/MWCNT-PEI (B) [/fig] [fig] Figure 6, Figure 7, Figure 8: (B) compares calibration plots for DA obtained in the absence (empty circles) and in the presence (full circles) of AA, as well as the corresponding response for AA in the presence of different concentrations of DA (triangles). It is important to remark that every experiment was obtained with a new electrode and that each point represents the average of the currents obtained with three different electrodes. The signal for AA remains constant even for the solution containing 100 M DA. The sensitivities for DA obtained in the presence and absence of AA are (2 16 ± 0 05 × 105 AM −1 r = 0 998 ; and (2 25 ± 0 03 × 10 5 AM −1 r = 0 998), respectively. This small difference in sensitivity for DA at GCE/MWCNT-PEI in (A) Differential pulse voltammograms for mixtures containing 1.0 mM ascorbic acid and different concentrations of dopamine: 10 (a), 25 (b), 35 (c), 50 (d), 75 (e) and 100 (f) M. (B) Current versus dopamine concentration plot for dopamine in the presence (full circle) and absence (empty circle) of ascorbic acid; as well as for ascorbic acid (triangle) in the presence of the different concentrations of dopamine. Other conditions as in Figure 5. the absence and presence of AA (4.2%), clearly demonstrate the feasibility to determine DA and AA in mixtures of both compounds using the GCE/MWCNT-PEI without any interference, even for 100-fold excess of AA compared to DA and using an AA concentration as high as 1.0 mM. In a similar way, determinations of AA in the presence of 1.0 mM Do are also possible. Figure 7(A) shows DPVs for increasing concentrations of AA from 0.50 to 5.0 mM in the presence of 1.0 mM DA. Two clearly resolved voltammetric peaks can be observed in all cases. Figure 7(B) displays calibration plots for AA at GCE/MWCNT-PEI in the absence (empty circles) and in the presence (full circles) of DA, as well as the response for DA for the solutions containing different concentrations of AA in the mixture (triangles). The peak current for DA remains constant even for the highest concentration of AA in the solution. The sensitivities for AA obtained in the presence and absence of DA are (1 6 ± 0 3 × 10 4 r = 0 9991), and (1 5 ± 0 5 × 10 4 AM −1 r = 0 998), respectively, showing a difference just of 6.2%. The signal for DA remains constant even for (A) Differential pulse voltammograms for mixtures containing 1.0 mM dopamine and different concentrations of ascorbic acid: 0.50 (a), 1.00 (b), 2.00 (c), 3.00 (d), 4.00 (e) and 5.00 (f) mM. (B) Current versus ascorbic acid concentration plot for ascorbic acid in the presence (full circle) and absence (empty circle) of dopamine; as well as for dopamine (triangle) in the presence of the different concentrations of ascorbic acid. Other conditions as in Figure 5. 5.0 mM AA, clearly indicating the feasibility to quantify AA in the presence of DA.3.4. Determination of DA in the Presence ofAA and SER A) shows DPVs for mixtures containing 1.0 mM AA, 1.0 mM SER and increasing concentrations of DA from 10 to 100 M at GCE/MWCNT-PEI. At variance with the voltammograms obtained at GCE, where just two peaks were obtained (not shown), at GCE/MWCNT-PEI the signal is resolved into three well-defined DPV peaks at −0.120 V, 0.090 V and 0.270 V for the oxidation of AA, DA, and SER, respectively. Once more, the presence of MWCNT ensures a highly sensitive response and an adequate shifting in the peak potential for the oxidation of AA, and makes possible the individual determination of the three compounds.The concentration of DA in the mixtures can be determined even in the presence of 1.0 mM AA and 1.0 mM SER.Figure 8(B) depicts calibration plots for DA at [/fig] [fig] Figure 8: (A) Differential pulse voltammograms for mixtures containing 1.0 mM ascorbic acid, 1.0 mM serotonin and different concentrations of dopamine: 10 (a), 25 (b), 35 (c), 50 (d), 75 (e) and 100 (f) M. (B) Current versus dopamine concentration plot for dopamine in the presence (full circle) and absence (empty circle) of ascorbic acid and serotonin; as well as for ascorbic acid (square) and serotonin (triangle) in the presence of the different concentrations of dopamine. Other conditions as in Figure 5. [/fig] [fig] Figure 9: Differential pulse voltammograms for undiluted blood serum sample before (a) and after (b) the addition of 100 M AA + 100 M DA + 100 M SER. Other conditions as in Figure 5. experiments using four electrodes prepared with four different dispersions gave a R.S.D. of 4.6%. [/fig]

https://www.semanticscholar.org/paper/2e14d4080650e9f19d957cbf04317138c6866e19

Listeria monocytogenes DNA Glycosylase AdlP Affects Flagellar Motility, Biofilm Formation, Virulence, and Stress Responses

Applied and Environmental Microbiology

643a31e6-3016-4604-beca-0edfc0e2a834

Listeria monocytogenes DNA Glycosylase AdlP Affects Flagellar Motility, Biofilm Formation, Virulence, and Stress Responses [fig] Supplementary, Figure 2 Supplementary, Figure 3 Supplementary, Figure 4: Role of AdlP in the oxidative stress response. (B) Three strains of bacteria were incubated with 10mM H2O2 or PBS at room temperature for the indicated time and measured survival rates. Survival percentages were calculated by comparing the recovered bacteria (cfu) from H2O2 to the recovered bacteria (cfu) from PBS. Data are mean and SEM of three independent experiments. (N=3; * indicates P<0.05, **P<0.01, ANOVA) Contribution of AdlP to streptomycin resistance. (A, B) Bacteria number changes following exposure to streptomycin. We exposed F2365∆adlP and the parental strain to 25 µg/ml streptomycin for the indicated time (A) or to streptomycin at the indicated concentration for 4 hours (B). Data represent mean of one independent experiment of triplicates. Contribution of AdlP to bactericidal antibiotic resistance. (A) Bacteria number changes following exposure to gentamicin. Three strains of bacteria were exposed to gentamicin at the indicated concentration for 4 hours. Data are mean and SEM of three independent experiments. (N=3; * indicates P<0.05, **P<0.01, ANOVA) (B, C) Bacteria number changes following exposure to ciprofloxacin and ampicillin. Three strains of bacteria were exposed to 1 µg/ml ciprofloxacin for the indicated time (B) or to the indicated concentration of ampicillin for 4 hours (C). Data are mean and SEM of three independent experiments. (N=3) [/fig]

https://www.semanticscholar.org/paper/af9ef6e4494bcf9bb64bb6b04e3e60cddfa111f4

Predictors of Primary Care Practice Among Medical Students at the Michigan State University College of Human Medicine

Journal of the American Board of Family Medicine

30846d40-3474-415c-bf32-35bd7d821cd6

Predictors of Primary Care Practice Among Medical Students at the Michigan State University College of Human Medicine Introduction: This study examines the evolution of student and physician interest in primary care from medical school matriculation to practice, focusing on student factors that potentiate primary care (PC) practice.Methods: We compiled a dataset of 2,047 Michigan State University College of Human Medicine graduates from 1991 to 2010.PC interest was assessed using the Association of American Medical Colleges (AAMC) matriculating student (MSQ) and graduation (GQ) questionnaires.PC practice was determined using AMA physician Masterfile data.C 2 analyses and logistic regression were used to examine factors that predict PC practice.Results: PC interest at matriculation and at graduation were the factors most likely to predict PC practice.After controlling for URM status, gender, and rural origin, the odds of practicing PC among those with a sustained interest in PC (on both the MSQ and GQ) were 100 times higher than those with no interest in PC, on either survey (P < .01).Among those students who developed an interest in PC by graduation, the odds of practicing PC were 60 times higher than noninterested students (P < .01).Finally, among students who were interested in PC at matriculation, but not graduation, the odds of eventually practicing PC were 3.8 times higher than noninterested students (P < .01).Conclusions: Our study suggests that cultivating PC interest at any point during medical school may predict PC practice.Early and sustained interest in primary care was the most substantial predictor of PC practice in our study, highlighting the need for primary care education even before medical school matriculation.( # Introduction Addressing the nation's primary care shortage and health disparities gaps requires examining the existing educational infrastructure.Institutional practices informed by Flexner [bib_ref] American medical education 100 years after the Flexner report, Cooke [/bib_ref] have cultivated an academic climate that favors specialization, impacting the current primary care workforce.Despite significant evidence that a primary care driven workforce decreases cost and increases quality, [bib_ref] Higher primary care physician continuity is associated with lower costs and hospitalizations, Bazemore [/bib_ref] [bib_ref] More comprehensive care among family physicians is associated with lower costs and..., Bazemore [/bib_ref] the U.S. continues to do a poor job recruiting and training physicians that fulfill the nation's health care needs. [bib_ref] Trends in US medical school contributions to the family physician workforce: 2018..., Phillips [/bib_ref] [bib_ref] Preventive Services Task Force methods to communicate and disseminate clinical preventive services..., Kurth [/bib_ref] Fewer U.S medical school seniors are choosing primary care careers such as family medicine (FM); and in the last decade, less than half of the available FM residency positions have been filled by graduating U.S. medical school seniors. [bib_ref] Trends in US medical school contributions to the family physician workforce: 2018..., Phillips [/bib_ref] ] [bib_ref] General medicine vs subspecialty career plans among internal medicine residents, West [/bib_ref] A similar increasing trend toward subspecialization has seen among pediatricians. [bib_ref] Where have the generalists gone? They became specialists, then subspecialists, Dalen [/bib_ref] Student loan debt, perceived work-life balance, perceived respect, and the income gap between specialists and primary care physicians, have all been examined as potentiators of this trend. [bib_ref] How do medical students view the work life of primary care and..., Phillips [/bib_ref] [bib_ref] Graduating medical student perspectives on factors influencing specialty choice. An AAFP National..., Kost [/bib_ref] here is a large body of literature describing factors influencing medical students' specialty choices.1] [bib_ref] Characteristics of medical students by level of interest in family practice, Bowman [/bib_ref] The temporal relationship between evolving student interest in primary care and eventual practice, however, remains incompletely understood. Our purpose was to examine how student interest in pursuing primary care changes from entry into medical school to graduation and how this changing interest is associated with eventual practice.By using a large retrospective sample examining 19 years of data from our home institution, we were able to assess how factors such as specialty interest, race/ethnicity, gender, and rural origin may affect eventual practice patterns. # Methods This study was a secondary analysis of merged data from several surveys, all using data from students who graduated from a single medical school (Michigan State University College of Human Medicine (MSU-CHM)) from 1991-2010, with the goal of capturing practice location and specialty data for graduates who were post-residency completion.Two national survey questionnaires, the matriculating student questionnaire (MSQ) and the graduation questionnaire (GQ), were administered by the Association of American Medical Colleges (AAMC) annually to all matriculating and graduating medical students.Institutional MSQ and GQ data describing students' career intentions and demographics were obtained from the AAMC.American Medical Association (AMA) Physician Masterfile data were used to create a primary care practice variable, as described below.All nonresponses were treated as missing and removed from the analyses. ## Demographic variables Demographic data were obtained from the MSQ and from internal Michigan State University (MSU) data gathered from medical school admission applications.Female was coded 1 and male as 0. Underrepresented in Medicine (URiM) [bib_ref] Assessing the evolving definition of underrepresented minority and its application in academic..., Page [/bib_ref] was defined using the MSQ underrepresented minority indicator which included: African Americans, Mexican Americans, Native Americans (American Indians, Alaska Natives, and Native Hawaiians), and mainland Puerto Ricans.This variable was coded as 1 if yes, and 0 otherwise.Rural origin was defined using rural-urban commuting area (RUCA) code of childhood residence(medical school application) and coded as 1 if RUCA was greater than or equal to 4, otherwise 0. Thus, rural origin included large rural, small town, and rural areas; on the other hand, nonrural origin included urban and suburban areas. Matriculating Student Questionnaire Career Intentions Table [fig_ref] Table 1: Coding Scheme Used to Indicate Primary Care Intention at Matriculation and Graduation,... [/fig_ref] shows how the primary care intentions were generated using existing variables.On the MSQ, students were asked to select a specialty in which they were most interested.Specifically, the questionnaire states: "What general specialty are you considering?"Responses were coded as "1," designating interest in primary care at matriculation (MSQ PCI), if they responded to this question selecting internal medicine (IM, 200), family medicine (FM, 180), or pediatrics (Peds, 500).All other responses were coded as "0."Graduation Questionnaire Career Intentions Given that the GQ survey questions often vary by year and frequently use different questions to investigate primary care, we used different coding schemes to construct a second binary variable, representing primary care intention at GQ, as seen in Table [fig_ref] Table 1: Coding Scheme Used to Indicate Primary Care Intention at Matriculation and Graduation,... [/fig_ref].Specifically, for 1978 to 1990, if students responded to SPEC_PREF1: "If yes, which specialty are you planning?" by selecting FM (06), IM (07), or Peds (19) and question number 509 (SUB_SPEC_PLAN: "Are you planning to become certified in a subspecialty?"by selecting "No," they were considered interested in primary care at graduation (GQ PCI), and coded 1.For 1991 to 1998, if the students responded "Yes" to SPEC_ PLAN: "Are you planning to become certified in a specialty?")and "No" to (SUB_SPEC_PLAN) with the same responses to SPEC_PREF1, they were coded as 1.For 1999 to 2004, the coding procedure was the same with 1991 to 1998 excluding the SUB_SPEC_ PLAN, which was not used due to an incomplete AAMC dataset for this variable for these years.For 2005 to 2010, if students responded "Yes" to (SPEC_ PLAN: "Are you planning to become certified in a specialty?")and selected family practice (120), IM (140), or Peds (320) for: (SPEC_PLAN: "Are you planning to become certified in a specialty or subspecialty?Choice of specialty/subspecialty"), they were considered interested in primary care at graduation and coded 1. Due to an AAMC error, data from 2008 were incomplete and therefore excluded. ## Primary care practice Using American Medical Association (AMA) Physician Masterfile data, we investigated the current specialty in which physicians were practicing.To confirm accuracy, and verify that Masterfile data captured a clinical practice rather than a residency training site, we enlisted a student research assistant to cross-check each variable using a protocol-driven google search.To minimize bias, we included only offices that could be verified with an active address using google maps, or hospital and medical group web sites.Current specialty and practice location were confirmed, and records from retired, deceased or inactive physicians were removed from the data set, as were records that could not be confirmed.As established in previous methodology, primary care was defined to include physicians practicing FM, IM, Peds, Geriatrics, or Sports Medicine/Family Medicine without subspecialization, unless the specialty is age-specific (eg, geriatrics or adolescent medicine) and the physician describes themselves as practicing primary care. [bib_ref] Predictors of Primary Care Practice Among Medical Students 377 copyright, Wendling [/bib_ref] [bib_ref] The impact of community-based undergraduate medical education on the regional physician workforce, Phillips [/bib_ref] [bib_ref] Trends in subspecialization: A comparative analysis of rural and urban clinical education, Wendling [/bib_ref] ll others were classified as nonprimary care practice. We then examined the specific specialty outcome of physicians, compared with their initial described student-level intention to practice primary care, from MSQ and GQ. # Analysis First, descriptive statistics (proportions and standard deviations) of MSQ PCI, GQ PCI, and PC from key demographic groups (gender, URM, and rural origin) were analyzed, and their statistical associations were assessed using c 2 tests.Next, we examined the influence of change in primary care interest on primary care practice, controlling for the 3 demographic variables using logistic regression analysis.Finally, we explored which (sub) specialties were most often chosen by specific groups of students in detail.All the data management and analyses were conducted using R 4.0.3 and R Studio 1.3.1073for Windows (R Core Team, Vienna, Austria). # Results ## Descriptive statistics of demographics and primary care intentions Of 2,047 responses, 21% (n = 430) were found to be complete and have valid responses that were used for analysis.Complete cases included only those with answers to all questions in the analysis.Complete cases were substantially fewer than original cases due to incomplete responses on the MSQ and GQ as well as graduates having retired or left clinical practice.For complete cases, we assumed data were missing completely at random, and first tested whether there were statistically significant differences between the original and the complete cases using 2-sample test for equality of proportions with continuity correction.Our analytic sample slightly overrepresented the proportion of PC compared with that of the original cases (40% vs 46%, P = .05). Among included graduates, 52% (n = 222) were female, 14% (n = 60) were URM, and 33% (n = 142) were from a rural background, as seen in Table [fig_ref] Table 2: Descriptive Statistics and Response Rates of All MSU-CHM Students Graduating 1971-2010 [/fig_ref].As of 2018, 46% (n = 198) of graduates were practicing primary care.Among those practicing primary care, 57% (n = 247) and 43% (n = 183) expressed their interest in PC at matriculation and graduation, respectively, indicating an overall decrease in PC interest during medical school.Looking at Figure [fig_ref] Figure 1: Figure 1 [/fig_ref] , the 2 largest groups of students either maintained an initial interest (Always PCI, n = 140, 33%) or never had interest (Never PCI, n = 136, 32%).An additional 25% of students had interest in PC at matriculation but no interest at graduation (Initial PCI, n = 107).The smallest group of students developed an interest in primary care during medical school (Developed PCI, n = 47, 11%). Table [fig_ref] Table 3: Frequency and Proportion [/fig_ref] shows how intentions on the MSQ and GQ differed across demographic backgrounds in detail.Specifically, at matriculation, the proportions of students who were interested in primary care were higher than those who were not, regardless of their gender, URM, and rural origin status.However, the pattern was reversed at graduation.At graduation, the proportions of PCI within female (50% vs 63%, P < .01),URM (33% vs 55%, P = .03),and rural origin student groups (41% vs 55%, P = .02)were statistically lower than at matriculation.Looking at their actual primary care practice (PC), female students practiced primary care more than male, but the difference of the proportions of PC for URM and rural origin status were not significant.C 2 test results also support the above patterns (see sixth and seventh columns).At matriculation (MSQ) and graduation (GQ), women were more likely to demonstrate an intention to practice primary care (P < .05),while there were no statistically significant associations demonstrated among URM and rural origin students.Women were more likely to practice primary care at all 3 time points (P < .01). ## Primary care practice We employed a logistic regression model to better understand what factors may predict primary care practice.As seen in Table [fig_ref] Table 4: Logistic Regression Model Predicting Primary Care Practice [/fig_ref] , controlling for other demographic factors and groups of interests, the odds of entering primary care practice among students who were interested in primary care at both matriculation and graduation (Always PCI) were found to be 100 times higher than those who demonstrated no interest in primary care at matriculation and graduation (the reference group: Never PCI).The odds of students entering primary care who had no initial interest, but developed an interest in primary care by graduation (Developed PCI), were 60 times higher when compared with the Never PCI group.Lastly, the odds of students entering primary care with an initial interest at matriculation, but no interest at graduation (Initial PCI), were still 3.9 times higher than those with no interest.Interestingly, after controlling for primary care interest grouping, demographic variables including female, URM, and rural origin status, did not significantly predict primary care practice. ## Specialty choices We then examined which specialties were practiced by students in detail.Looking at Figure , among students who showed an initial interest in primary care at matriculation but not graduation (Initial PCI, n = 107), 25 students (23%) still were practicing a primary care specialty, including Pediatrics (n = 10, 9%), Internal Medicine (n = 8, 7%), Family Medicine (n = 6, 6%), and Internal Medicine/Pediatrics (n = 1, 1%) specialties.The majority of Initial PCI students (n = 82, 77%) did not eventually practice primary care; the most common practice specialties were Obstetrics and Gynecology (n = 21, 20%) followed by Adult Surgical (n = 14, 13%), and Adult Support (n = 13, 12%) specialties.A complete list of specialties within each category can be found in the Appendix.This pattern was very similar to the pattern of the group of students who were never interested in primary care (Never PCI, n = 136).Of these, 126 students (93%) chose a nonprimary care specialty, and mostly practiced We also explored specialty choices of those students who showed an interest in primary care at both matriculation and graduation (Always PCI, n = 140).Consistent with the result of logistic regression model, most (n = 124, 89%) eventually practiced a primary care specialty.The other students mostly entered Adult Medical Subspecialty (n = 6, 4%), Pediatric Subspecialty (n = 3, 2%), and Other (n = 3, 2%).Similarly, 81% (n = 39) of students who did not show an initial interest in primary care but developed it at graduation (Developed PCI, n = 47) chose to be a primary care physician; the remainder mostly entered Adult Medical Subspecialty (n = 2, 4%) and Pediatric Subspecialty (n = 2, 4%) practice. # Discussion Eventual primary care practice may be influenced by several factors.Our study suggests that an initial interest in primary care that is sustained until graduation seems to be one of the most important predictors; highlighting the importance of promoting primary care careers before medical school and selecting students for admission with a primary care interest.While the odds of entering primary care for students who were initially interested in primary care but had lost interest by graduation were still 3 times higher than students demonstrating no interest during their medical school career, we also found that among students with no initial interest, the odds of entering primary care among those who developed an interest at graduation were almost 60 times higher than their never-interested peers.Our data supports that identifying and fostering the interest of students with an initial interest in primary care is an important strategy to support primary care practice, but cultivating an interest among not-yet-interested students is also worthwhile.These results are concordant with prior research.Exposure to family medicine during medical school, for example, potentiates students' attitudes and career choices [bib_ref] Medical student debt and primary care specialty intentions, Phillips [/bib_ref] [bib_ref] Third-and fourth-year medical students' changing views of family medicine, Phillips [/bib_ref] [bib_ref] The effect of required third-year family medicine clerkship on medical students' attitudes:..., Senf [/bib_ref] The curriculum at our institution has historically prioritized primary care, with Michigan State University-College of Human Medicine (MSU-CHM) students during the study years being exposed to robust primary care clinical experiences including approximately 7 weeks of required outpatient family medicine, 1 week of required outpatient internal medicine, and 4 weeks of required outpatient pediatrics during the third and fourth year of medical school.More recently, MSU-CHM has undergone efforts to continue this prioritization, integrating primary care exposure throughout all 4 years of medical school. [bib_ref] Michigan State University College of Human Medicine, Wagner [/bib_ref] nterestingly, our study did not show a significant relationship between demographic factors and eventual primary care practice.Primary care practice in our study seems to be predicted most strongly by primary care interest, not individual demographic characteristics.These results underscore the importance of a holistic approach to primary care recruitment, and may represent an opportunity to improve recruitment strategies.The need for primary care physicians from backgrounds underrepresented in medicine, as well as those interested in rural practice is well documented. [bib_ref] Predictors of Primary Care Practice Among Medical Students 377 copyright, Wendling [/bib_ref] [bib_ref] The racial and ethnic composition and distribution of primary care physicians, Xierali [/bib_ref] [bib_ref] Trends in subspecialization: A comparative analysis of rural and urban clinical education, Wendling [/bib_ref] [bib_ref] Family medicine residents: increasingly diverse, but lagging behind underrepresented minority population trends, Xierali [/bib_ref] Focusing on providing early (prematriculation) primary care career exposure with the goal of developing early interest in PC among students from rural and underrepresented backgrounds may amplify existing PC recruitment strategies by augmenting and diversifying the pool of applicants who demonstrate interest in primary care.Experts have suggested that broad career choices are often chosen as early as middle school, and that premedical recruitment should begin very early in the educational process, and include engaging with communities, high schools, and undergraduate institutions such as community colleges, to find, support, and recruit these students. [bib_ref] Graduating medical student perspectives on factors influencing specialty choice. An AAFP National..., Kost [/bib_ref] [bib_ref] Finding, recruiting, and sustaining the future primary care physician workforce: a new..., Bennett [/bib_ref] [bib_ref] The leaky pipeline: factors associated with early decline in interest in premedical..., Barr [/bib_ref] [bib_ref] Viewpoint: Developing a physician workforce for America's disadvantaged, Freeman [/bib_ref] is well positioned to accomplish this.The College of Human Medicine also has a long and established history of developing pipeline programming dedicated to recruiting students to serve communities where there is a need.Capitalizing on this expertise through the development of primary care-focused programming that includes primary care mentorship and community outreach may enable our institution to bolster primary care recruitment. [bib_ref] Michigan State University College of Human Medicine, Wagner [/bib_ref] ur study endeavored not only to understand factors that would predict primary care outcome, but also what specialties might potentially draw students away from primary care careers.Overwhelmingly, among students with an initial interest in primary care, the most popular non-PC specialty at the time of practice was Obstetrics and Gynecology.This might be reflective of the changing landscape within family medicine.While traditionally family medicine has represented the broadest scope of patient care, spanning care across age groups and including several modalities and care settings including obstetric care and minor procedures, the scope of care provided by family physicians has been shrinking, with fewer family physicians providing obstetric care. [bib_ref] Family physicians practicing high-volume obstetric care have recently dropped by one-half, Barreto [/bib_ref] Obstetrics may represent the most primary care adjacent specialty, providing interested students the opportunity to combine surgical training with longitudinal care of a cohort of patients.Efforts to potentiate the primary care workforce may require rigorous interventions that focus on preserving the breadth and spectrum of care provided by primary care specialties. Our study has important limitations.Approximations using the MSQ and GQ may incompletely characterize primary interest, and our definition of primary care did not include specialties like sports medicine where physicians may be providing some primary care.Moreover, despite being largely representative, our sample overrepresented primary care, and used only a subset of the original data due to incomplete data sets.In addition, throughout different versions of the MSQ and GQ, there was variability in the questions that were used to create outcome variables.Our study is retrospective, linking multiple data sets and using selfreported data from the AMA Masterfile database.While we cross checked all variables with multiple sources, inconsistencies between publicly available specialty practice and actual practice are possible.Lastly, this is a single institution study, limiting generalizability.Future studies will examine additional factors that may affect student primary care interest, such as student socioeconomic status and previous exposure to primary care.Studies will also include data from multiple institutions as an important next step in understanding strategies to cultivate a diverse and robust primary care workforce. To see this article online, please go to: http://jabfm.org/content/35/2/370.full. [fig] Figure 1: Figure 1.Proportions of students distributed among 4 groups based on their intention to practice primary care at matriculation and graduation (n = 430) [/fig] [table] Table 1: Coding Scheme Used to Indicate Primary Care Intention at Matriculation and Graduation, by Survey Year [/table] [table] Table 2: Descriptive Statistics and Response Rates of All MSU-CHM Students Graduating 1971-2010 (Original Cases) and the Subgroup of Those Students With Complete Responses to the Variables Used for This Study (Complete Cases) With Two-Sample Test for Equality of Proportions with Continuity Correction [/table] [table] Table 3: Frequency and Proportion (%) of Primary Care Intention on Matriculating Student Questionnaire (MSQ) and Graduation Questionnaire (GQ) and Primary Care Practice, Across Demographic Variables With Chi-Square Test Results [/table] [table] Table 4: Logistic Regression Model Predicting Primary Care Practice [/table]

https://www.semanticscholar.org/paper/3ad2f0659d823fc04be069afd708f1ab3319b1d1

"Pre-IVF treatment with a GnRH antagonist in women with endometriosis (PREGNANT): study protocol for(...TRUNCATED)

BMJ Open

e6216b43-2428-4aaf-8101-b40e1dfd2810

"Pre-IVF treatment with a GnRH antagonist in women with endometriosis (PREGNANT): study protocol for(...TRUNCATED)

https://www.semanticscholar.org/paper/933719e39da1fa3be9e8473457e5275e35eee318

"Genomic profiling of thousands of candidate polymorphisms predicts risk of relapse in 778 Danish an(...TRUNCATED)

Leukemia

84407eba-2491-4d53-9172-b72cae91c0e3

"Genomic profiling of thousands of candidate polymorphisms predicts risk of relapse in 778 Danish an(...TRUNCATED)