搜索结果：sets

This study aims to evaluate multiple feature sets composed of sensor-based biomarkers acquired during walking for the automated estimation of post-stroke motor impairment levels using Fugl-Meyer Lower Extremity Assessment (FMA-LE)-derived classes. Sensor-based walking data from the open-source ARRA dataset were combined with data collected at the Hospital of Braga. Data from 32 post-stroke individuals (FMA-LE motor score: 24 ± 3) were included. A decision tree classifier was evaluated using stratified six-fold cross-validation across different feature sets, including: correlated with motor impairment levels versus full feature sets; spatiotemporal versus surface electromyographic (sEMG) features; inclusion of demographic variables; and the use of data augmentation. The best performance was achieved using correlated sEMG features combined with age, paretic side, and body mass, along with noise-based data augmentation, yielding a validation Matthews Correlation Coefficient (MCC) of 0.85 ± 0.16 and a test MCC of 0.70. sEMG features provided improved classification performance compared to spatiotemporal features, and comparable results were obtained using a reduced subset of muscles. These results demonstrate the feasibility of using sEMG-based features acquired during walking to classify post-stroke motor impairment levels. Feature reduction and inclusion of demographic variables may support efficient model design, while data augmentation may enhance generalization. Further validation in larger and more diverse datasets is required to assess robustness and clinical applicability.

This study aimed to compare the effects of pre-exhaust training and traditional training on muscular hypertrophy, strength, body composition, and muscular endurance in resistance-trained participants over an 8-week study period. We randomly assigned 48 young, resistance trained individuals (male = 32, female = 9; height = 173.0 &#xb1; 10.3 cm; weight = 81.7 &#xb1; 15.9 kg; age = 22.5 &#xb1; 4.1 yrs) to 1 of 2 experimental groups after initial testing: a pre-exhaust resistance training (RT) group (PreEx: n = 24) or a traditional RT group (TRAD: n = 24). The RT protocol consisted of the following lower body exercises: leg extension, Smith squat, seated hamstring curl, barbell Romanian deadlift. Participants in the TRAD group completed all sets for one exercise before performing a different exercise with ~2 minutes of rest between sets. The PreEx group performed a set of a single-joint exercise immediately prior (<10 seconds) to a set of the corresponding multi-joint exercise for that muscle group followed by ~2 minutes of rest. Both groups performed 4 sets of each exercise twice weekly with loads corresponding to 8-12 repetition maximum (RM). Assessments included pre-post measures of muscle thickness of the quadriceps and hamstrings, body composition, 1RM squat strength, local muscle endurance of the quadriceps, and the countermovement jump. Results showed slightly greater improvements in muscle size and body composition favoring TRAD, with statistical uncertainty for a true between-group difference. Changes in strength, local muscular endurance and countermovement jump height were similar between groups. In conclusion, TRAD would seem to be a better option than PreEx for those seeking to optimize muscle hypertrophy. Measures of strength, power and local muscular endurance were relatively similar between conditions, suggesting that PreEx is a viable alternative to TRAD for these outcomes.

Factor Xa (FXa) remains a clinically validated and chemically tractable anticoagulant target despite the therapeutic role of direct oral FXa inhibitors. Contemporary FXa inhibitor literature, however, is heterogeneous in scaffold design, endpoint reporting, assay consistency, translational depth, and suitability for computer-aided drug design (CADD). This review evaluates published series of small-molecule FXa inhibitors through a framework that combines translational structure-activity relationships (SARs), assay-aware data quality, and QSAR-readiness. A structured narrative synthesis focused mainly on post-2014 studies reporting discrete small-molecule or semisynthetic FXa inhibitors. Eligible series were classified as fully synthetic or natural-product-derived/semisynthetic chemotypes, and extraction covered scaffold architecture, potency endpoints, assay context, selectivity, clotting or antithrombotic readouts, PK/ADME, structural clarity, translational context, and extraction confidence. QSAR-readiness was assessed using analog density, congenericity, endpoint quality, assay comparability, activity range, structural interpretability, and curation burden. Fully synthetic chemotypes, particularly anthranilamide-derived and related scaffolds, provided the most coherent and modellable FXa datasets, whereas natural-product-derived and semisynthetic series expanded structural diversity. Many exploratory series, however, were limited by small analog sets, heterogeneous endpoints, incomplete translational characterization, narrow activity ranges, or higher curation burden. The practical value of published FXa inhibitor series, therefore, depends not only on potency but also on whether chemical and biological information can be reconstructed with confidence for reproducible SAR interpretation, local QSAR modeling, AI/ML-enabled CADD reuse, and clinical benchmark-aware prioritization. The QSAR-readiness framework is a critical triage tool, not a substitute for formal validation, distinguishing datasets suitable for curated local modeling from those better suited to qualitative SAR, scaffold inspiration, or translational hypotheses.

Single-cell RNA sequencing (scRNA-seq) studies increasingly combine expression-based and mutation-derived signals, but small-cohort designs with repeated runs from the same biological unit can make standard cross-validation overly optimistic. We reanalyzed PRJNA736095 (14 SRR runs from 7 GSM/donor proxies) using a GATK-centered RNA-seq variant-calling and gene-burden workflow, then evaluated mutation-derived, expression-only, and combined feature sets with leakage-safe preprocessing inside each validation fold. Run-level repeated stratified cross-validation showed high within-dataset separability for GATK gene-burden features (balanced accuracy 0.973 +/- 0.113), but GSM-grouped leave-one-GSM-out validation reduced balanced accuracy to 0.708 and exact GSM-level permutation testing was not significant (p = 0.257). Expression-only and combined feature sets did not improve GSM-grouped balanced accuracy over the variant-only branch. Expression-mutation marker overlap was not significant after FDR correction, and public external datasets were used only to define feasibility or processed biological context rather than as strong external classifier validation. These findings position the workflow as an auditable, hypothesis-generating framework and highlight pitfalls of standard cross-validation in small-cohort scRNA-seq machine-learning analyses.

Initial data analysis (IDA) is essential for valid and reproducible statistical analyses, but existing IDA frameworks were primarily developed for single-study datasets. Cancer registries (CRs) are extensive data systems characterized by continuous updates, repeated data extraction, multiple analytical uses, and evolving classification systems, which create requirements not addressed by existing IDA frameworks. This study aims to develop a structured IDA framework adapted to CRs. We conceptualized IDA in CRs as a process spanning three data states: operational registry data, the extracted dataset and the analysis-ready dataset. The framework was developed by adapting existing IDA principles to the CR setting and organizing them into four stages: metadata, cleaning, screening, and reporting. The approach is based on predefined and versioned rule sets, structured recording of data processing, and metadata-based linkage between dataset definitions, data cleaning rules, screening outputs, the final report and dataset. A demonstrative survival dataset from the Slovenian Cancer Registry was used to illustrate implementation. The framework is represented by the item set defining the activities and expected outputs of the structured IDA process in CRs. In the use case, metadata specified the dataset scope, intended use, variables, coding context, and applicable rules. Execution of the selected rules produced an analysis-ready dataset with traceable data cleaning steps, documented eligibility decisions, screening outputs describing the general and analysis-specific data properties, and the IDA report intended for external researchers to be delivered alongside the data. The proposed framework extends existing IDA approaches to meet the specific requirements of CRs. It supports consistent dataset preparation and with that improves transparent and reproducible data use.

The present study aimed to compare the effects of high-intensity resistance training (HIRT) with different training volumes on physical performance and blood-derived markers adaptations in recreationally trained subjects. Thirty subjects of both genders (21.7 &#xb1; 1.8 years old; body mass 62.4 &#xb1; 6.5 kg; height 1.7 &#xb1; 0.03 m) were randomly allocated to perform lower (G4; women = 7; men = 8) or higher (G6; women = 8; men = 7) training volume, which consisted of 4 and 6 weekly sets per muscle group, respectively. The HIRT method consisted of 2 sets of the following: 6 repetitions maximum (RM) at 85% of the 1RM followed by 20-seconds of rest, 3RM, another 20-seconds of rest, followed by 3RM, and finally, 2:30 minutes of rest. After 2:30 minutes of rest, the participants repeated the entire sequence a second time (i.e., the second series was performed). Muscle performance and blood-derived markers were assessed before and after 4 weeks of training. No significant differences in physical performance were observed between groups. However, there were significant changes in creatine kinase (p = 0.03), creatinine (p = 0.01), and lactate dehydrogenase (p = 0.01), with higher values for G6 in comparison to G4. HIRT improved muscle performance regardless of training volume. Nevertheless, the greater volume resulted in more pronounced increases in blood-derived markers of muscle damage when compared to the lower volume group. Therefore, considering the similar benefits on muscle performance, it would be recommended to perform lower volumes to prevent unnecessary physiological stress.

Background/Objectives: Glaucoma is a leading cause of irreversible blindness, and it is hard to catch early be-cause it rarely causes symptoms until real damage has already occurred. Existing au-tomated detection methods still miss the subtle structural changes that show up before vision loss begins. This study presents an automated framework based on optic nerve structure and retinal biomarkers, aimed at enabling earlier and more consistent glau-coma screening. The framework combines segmentation and classification in a single deep learning pipeline, trained and tested on the ORIGA and REFUGE2 fundus image datasets. Each image is resized, denoised with median filtering, and contrast-enhanced through histogram equalisation and normalisation, then cropped down to the optic nerve region using central cropping and intensity-based localisation. Methods: A custom encod-er-decoder CNN segments the optic disc and cup, and from that segmentation, we ex-tract biomarkers ophthalmologists already rely on, including the vertical cup-to-disc ratio and disc/cup area measurements. A Lightweight Vision Transformer separately learns broader structural patterns across the retina, and the two feature sets are fused and passed through a SoftMax classifier. Results: Segmentation accuracy was strong: Dice scores of 0.9082 for the optic disc and 0.9994 for the optic cup, IoU scores of 0.8351 and 0.9988, and an overall mean Dice of 0.9538 and mean IoU of 0.9169. Classification performance held up well, too, with high F1-score, recall, accuracy, and precision across normal and glaucomatous cases. Conclusions: The combination of interpretable, clinically established biomarkers with transformer-based global feature learning gives the framework the ability to support automated glaucoma risk assessment and demon-strates promising performance for glaucoma detection using retinal fundus images.

Cementitious sand and gravel (CSG) is commonly used in construction engineering; however, its mix proportion design is complex, and traditional physical experiments face limitations such as long cycles, high costs, and susceptibility to external factors when obtaining high-quality sample data. In this study, a foundational dataset was first acquired through physical experiments: 100 sets of CSG specimens with different mix proportions (cement content 40, 50, 60, 70 kg/m3; water-to-binder ratio 1.0, 1.2, 1.4; sand ratio 0.1, 0.2, 0.3, 0.4; fly ash content 20, 30, 40, 50 kg/m3) were prepared. After 28 days of standard curing, compressive strength and splitting tensile strength tests were conducted using a WAW-1000 electro-hydraulic servo universal testing machine, yielding 100 sets of real mechanical property data. The coefficients of variation for all test groups were below 10%, confirming the reliability and repeatability of the experimental data. On this basis, a data augmentation method based on Conditional Tabular Generative Adversarial Networks (CTGAN) is proposed. Through adversarial training between the generator and the discriminator, the model learns the multi-dimensional distribution characteristics of the original CSG data and generates 100 synthetic samples, which are then merged with the original data to expand the dataset to 200 samples. The quality of the synthetic data is evaluated using Wasserstein distance and correlation matrix heatmaps. Furthermore, a hybrid XGBoost-LSTM prediction model is proposed-XGBoost is used for feature construction to capture nonlinear interactions among mix proportion variables, and the constructed features are then fed into an LSTM network for sequential learning and regression prediction. The results show that the CTGAN-generated data are highly consistent with the original data in terms of kernel density distributions and variable correlations, with Wasserstein distance significantly superior to four comparative methods: Bootstrap, SMOTE, GaussianCopula, and TVAE. After augmentation, the XGBoost-LSTM model achieves a coefficient of determination (R2) of 0.9897 for compressive strength prediction (vs. 0.9793 before augmentation) and 0.9801 for splitting tensile strength (vs. 0.9882 before augmentation, a slight decrease). The mean absolute percentage errors (MAPE) are 4.49% and 4.11%, and the root mean square errors (RMSE) are 0.201 and 0.049, respectively; both error metrics are reduced compared with those before augmentation. Compared with baseline models including XGBoost, LSTM, Random Forest (RF), and Support Vector Regression (SVR), the XGBoost-LSTM model exhibits the best performance across all evaluation metrics, and Wilcoxon signed-rank tests confirm that the performance differences are statistically significant (p < 0.05). The proposed method of CTGAN-based data augmentation combined with the XGBoost-LSTM hybrid model provides an effective solution to the problem of insufficient CSG sample data and offers a reference for data enhancement and performance prediction of other small-sample materials.

It is important to detect the damage in the rollers and repair them since the damage to the rollers has a negative impact on the quality of the rolled products. Identifying the types of damage helps determine the repair process and normal production work. Ultrasonic testing technology has the advantages of large detection depth, accurate defect localization, low cost, convenient use, fast speed, and harmlessness to the human body. In order to improve the intelligence of ultrasonic detection for identifying damages in rollers, this article proposes a deep learning classification method of damages based on Rayleigh wave signals and power spectrum images with specific sampling rate, automatic identification of four common types of damages (void, hole, crack, and adhesion) is achieved by establishing end-to-end learning models for one-dimensional (1D) and two-dimensional (2D) data. Firstly, an organic glass inclined block and a clamping device were designed. In the experiment, time-domain signals were received on the right side of the damaged sample, and signal data sets were established for signals with different sampling rates. Then, the power spectrum image data sets were established after a short-time Fourier transform was performed. Next, a damage detection model is established based on a deep learning framework, which includes ResNet, GoogLeNet, DenseNet, and AlexNet with 1D and 2D convolutional channels to extract signal features for classifying damage. Finally, the performances of DenseNet models with different structures and depths are compared based on key indicators such as accuracy and training time. The experiment demonstrates that under high sampling rate conditions, using the power spectrum image of Rayleigh waves as data input yields better results than directly using Rayleigh wave signals. Moreover, for the power spectrum images of 0.5 MS/s Rayleigh waves, using ResNet-18 to establish a deep learning model can achieve high accuracy and shorter training time.

Outlier detection serves as an effective technique for identifying anomalous samples in complex data. Existing methods are often disturbed by noise and boundary samples, which degrade the quality of sample relationships. Moreover, traditional random walk approaches are vulnerable to weak and spurious connections that can mislead the walking process. To address these issues, this paper proposes a robust random walk based on natural neighbors for outlier detection (RWNOD) method. First, an adaptive smoothing mechanism is proposed to leverage natural neighbors to actively adjust sample positions, reducing local noise while preserving structural information. Then, a robust random walk strategy is developed to incorporate shadowed sets into the transition matrix, preserving reliable connections while suppressing unreliable ones. At the same time, a corresponding outlier detection algorithm is proposed. Experiments on datasets are conducted to compare the proposed algorithm with seven other algorithms. The experimental results demonstrate that the proposed algorithm achieves superior performance and strong robustness.

Bisphenol A (BPA) is a common industrial chemical primarily used in the manufacture of plastics, and it has been found in more than 90% of people worldwide. As an endocrine disruptor, BPA can impair reproduction, development, immunity, metabolism, and cognition; it also disturbs immune balance and thus fosters chronic inflammation. A number of population-based studies have indicated a link between environmental BPA exposure and atopic dermatitis (AD). Nevertheless, the detailed molecular pathways connecting BPA to AD remain poorly understood. AD is the leading chronic recurrent inflammatory skin disorder, characterized by severe itching and repeated eczema-like lesions. Its prevalence is roughly 13% among children and 5% among adults, and its global incidence continues to rise, imposing heavy health and economic burdens on societies. To clarify whether and how BPA may promote or worsen AD, we carried out a comprehensive computational study that integrated network toxicology, transcriptomic data, machine learning, molecular docking, and molecular dynamics simulations. From the CTD, ChEMBL, and SwissTargetPrediction databases, we collected 5701 potential BPA targets; from GeneCards and OMIM, we obtained 3270 genes linked to AD. The overlap between these two gene sets gave a group of common candidate genes. Enrichment analyses using GO and KEGG showed that these common genes were significantly overrepresented in the PI3K-Akt signaling pathway, Th17 cell differentiation, and the JAK-STAT signaling pathway-all central to immune and inflammatory regulation. We then built a protein-protein interaction (PPI) network by submitting the common genes to the STRING database and employed Cytoscape to extract hub genes from that network. By integrating human AD transcriptomic profiles with the hub genes and applying two machine learning techniques (LASSO and SVM), we identified six core toxic targets of BPA in AD: TIGIT, JAK3, IL22, S100A8, CCL2, and FCER1G. These six targets fall into two main functional categories: immune dysregulation and inflammatory cell infiltration. Subsequent molecular docking and molecular dynamics simulation experiments confirmed that BPA binds well to all six targets and can form stable complexes with them. Collectively, our findings offer a preliminary experimental foundation for future investigations into the pathogenesis of BPA-induced AD and provide important molecular evidence for understanding how environment-gene interactions contribute to complex inflammatory skin diseases such as AD.

Micro-Unmanned Aerial Vehicles (micro-UAVs) require compact radar arrays under severe Size, Weight, Power, and Cost (SWaP-C) constraints. Spread Spectrum Digital Beamforming (SSDBF) reduces receiver hardware by multiplexing multiple antenna channels before a shared RF chain, but the receiver-side switching operation folds wideband noise and injects switch-related transient noise before digital demultiplexing. This paper develops a hardware-waveform co-design framework that links the transition density of bipolar spreading sequences to high-frequency code energy and switching-event count, while explicitly distinguishing modulation-induced thermal-noise folding from switch-transient injection. Transition density is used as a physically interpretable design surrogate rather than a sufficient statistic for folded noise, and it is constrained jointly with non-zero-shift cross-correlation to preserve spatial isolation under receiver-side timing skew. An &#x3f5;-constrained Greedy Coordinate Space Search (&#x3f5;-GCSS) algorithm is proposed to synthesize low-transition-density SSDBF code sets. Commercial circuit-level transient simulations and system-level MATLAB R2024b simulations show that the proposed code set reduces the PRN-like transition-density level from about 0.50 to about 0.19, lowers the circuit-simulation-derived integrated IF noise power by 9.63 dB relative to the Gold/PRN-like baseline, and improves the normalized maximum detection range by 8.2 percentage points at K=4 without adding analog front-end hardware.

Background/Objective: The management and prevention of impacted maxillary canines is both an art and a science. These important teeth serve both aesthetic and functional purposes, so they demand meticulous evaluation. Deep learning model tools are game-changing technologies, elevating diagnostic capabilities and therapeutic planning. This study comprised a comparative analysis of the prediction of permanent maxillary canine eruption as a preventive measure between experts and a deep learning model. Methods: In this retrospective cross-sectional study, 2230 panoramic radiographs of patients aged 9-14 years were analyzed to assess the patterns of eruption of unerupted maxillary permanent canines (UPMCs). The study images were classified into three sectors according to the modified Ericson and Kurol sectors. Data preprocessing techniques were used to prepare images for the deep learning model by using a DenseNet121-based Convolutional Neural Network (CNN). The data were split into training and testing sets to train the AI to predict sectors. The deep learning model's predictions were evaluated using sector accuracy, precision, recall, and F1 score. Results: The sample included 796 (35.6%) men and 1434 (64.3%) women. The mean age of the participants was 12.2 &#xb1; 1.60 years. For sector 1, the right side of UMPCs, the AI reported an accuracy of 99.55% and perfect precision at 100%, while for the left side, accuracy and precision were 99.24% and 100.00%, respectively. For sector 2, prediction on the right side, the performance accuracy values reached 99.73%, and the precision was 98.15%. For the left side of UMPCs, the prediction accuracy was 98.79%, and precision was 94.10%. Regarding sector 3, the right side saw 99.82% accuracy and 98.83% precision. For the left side, the AI achieved an overall accuracy of 99.46% and precision of 99.72%. Conclusions: The deep learning-based system significantly reduced the time and human resources required for landmark identification and parameter generation, making the diagnostic process more efficient in interceptive/preventive orthodontics.

Dextran sodium sulfate (DSS) was used to induce ulcerative colitis in a porcine model to investigate the transcriptional responses in inflamed colonic tissue. Eleven pigs were divided into two groups, of which five pigs were administered an oral dose of DSS daily for five days, whereas six non-treated pigs served as a control group. Differences in transcript expression between treated and control pigs identified 425 down-regulated and 780 up-regulated mRNAs in response to DSS treatment. Fifty-nine differentially expressed miRNAs were also identified, comprising 20 up-regulated and 39 down-regulated miRNAs. The top enrichment KEGG pathways for the up-regulated genes were breast cancer (ssc05224), gastric cancer (ssc05226), focal adhesion (ssc04510), and the PI3K-Akt signaling pathway (ssc04151). The top gene ontology terms for the up-regulated genes were blood vessel development (GO:0001568), extracellular matrix and external encapsulating structure (GO:0031012). Protein-protein interaction network analysis identified three hub genes, including LOXL1, MFAP2, and FSTL3. Seventeen high-confidence miRNA-mRNA pairs were recognized, and two genes (CD101 and AVL9) have been confirmed as targets for ssc-miR-24-3p by dual-luciferase reporter assay. Our study provides a better understanding of the key roles of mRNAs and miRNAs in regulating DSS-induced colitis in pigs and defines sets of coding and non-coding RNA transcripts, which may serve as intervention targets or biomarkers for ulcerative colitis.

This study aims to develop a lightweight, edge-deployable artificial intelligence (AI) system for real-time, non-contact recognition of cattle eating, standing, and lying behaviors in farm environments. Automated monitoring of these behaviors in cattle provides fundamental behavioral data for the future development of systems that analyze feeding duration, lying duration, and behavioral rhythms. Nevertheless, practical deployment on farms is hindered by data imbalance, dense animal groupings, scale variation, occlusion, and the need for low-cost edge computing. To address these challenges, we propose BoviFusionNet, a lightweight, edge-deployable AI system. A box balanced augmentation strategy rebalances training instances at the object level without altering the validation or test sets. Built upon YOLO11n, the model integrates three targeted enhancements: information-preserving downsampling (ADown), adaptive bidirectional feature fusion (BiFPN), and local window attention (C2CGA) to improve multi-scale representation and fine-grained behavior discrimination. Experimental results show that BoviFusionNet achieves 0.7851 recall, 0.7763 F1-score, 0.7976 mAP@0.50, and 0.6305 mAP@0.50:0.95, with only 5.4 GFLOPs and a 3.4 MB model size. Compared with the YOLO11n baseline, it improves mAP@0.50:0.95 by 9.92% and reduces the parameter count by 39.8%. After INT8 quantization and deployment on an RK3588S edge device, real-time inference reaches 28.08 frames per second (FPS). Therefore, BoviFusionNet offers an effective accuracy-complexity trade-off for on-farm edge AI applications. By enabling continuous, non-invasive monitoring of health-relevant behaviors, it provides fundamental behavioral data for the future development of veterinary health assessment tools without relying on cloud services or wearable sensors.

Small objects in video streams occupy a small proportion in the image; the texture and shape information they carry is limited, making it difficult to continuously track and identify. To solve this problem, a method for continuous tracking and recognition of small objects in the video stream based on YOLO and spatio-temporal context memory network is proposed. A backbone network based on the improved YOLOv8 model is introduced, and the multi-scale visual features of small objects are extracted at different levels of the video stream using the wavelet pooling module. A mixed attention module enhances the feature response in the spatially significant pixel regions, generating weighted multi-scale visual features. The neck network processes these weighted features through a spatio-temporal context memory network to extract multi-scale spatio-temporal features. Then, a bidirectional feature pyramid module fuses these multi-scale spatio-temporal features. The head network processes the fused features to output continuous recognition results for small objects. Experiments show that the proposed method successfully extracts the spatiotemporal features of small objects from video stream data sets dominated by small objects. Under different conditions of small object occlusion rates, this method achieves a success rate of continuous tracking and recognition of small objects over 0.93.

Recent meta-analyses of randomized controlled trials have raised concerns that treatment with omega-3 fatty acids may increase the risk of atrial fibrillation (AF). However, these meta-analyses included at most 8 trials. The aim of this current meta-analysis was to expand the search by including other eligible omega-3 randomized controlled trials with AF incidence data, incorporating both published and unpublished data. Eligible studies were randomized controlled trials investigating daily doses of &#x2265;500 mg/d of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA). Additional inclusion criteria included &#x2265;12 months of treatment with EPA/DHA, participants &#x2265;50 years of age, and, where possible, the absence of known AF/atrial flutter at baseline. The primary outcome was the occurrence of new-onset AF. Our primary hypothesis was that risk for AF would simultaneously depend on both omega-3 dose (above or below 1500 mg/d) and background cardiovascular disease risk status, and that their combined impact on AF risk would be synergistic. A total of 35 randomized controlled trials (37 data sets; n=114&#x2009;592) were included in this meta-analysis. Only studies including patients at high-risk for cardiovascular disease who were treated with high-doses of EPA/DHA (>1500 mg/d) showed a statistically significant increase in AF risk with a pooled odds ratio (OR) of 1.43 (95% CI, 1.14-1.79) and an absolute risk difference of 0.8% (0.40%-1.1%). None of the other 3 groups showed statistically significant levels of AF risk (odds ratios, 1.07 [high risk-low dose], 1.06 [low risk-low dose], and 1.03 [low risk-high dose]). This meta-analysis suggests that high-dose EPA/DHA treatment is associated with an increased risk of AF in patients at high cardiovascular disease risk, whereas low-dose EPA/DHA does not appear to increase AF risk, even in high-risk populations. Further prospective studies are needed to evaluate any potential increased risk of higher doses balanced against potential benefits.