Dietary antioxidants and comorbidity of cardiovascular disease and osteoarthritis: A machine learning analysis with SHAP interpretation

Comorbidity of cardiovascular disease (CVD) and osteoarthritis (OA) poses a significant global health burden, leading to increased disability and mortality. The influence of dietary antioxidants on the risk of CVD-OA comorbidity remains unclear, and new predictive tools are needed for early identification. We utilized nationally representative data from the US National Health and Nutrition Examination Survey (2007-2010) in a cross-sectional study design to systematically assess the association between dietary antioxidant intake and CVD-OA comorbidity risk. A total of 7682 participants with complete dietary and clinical information were included. Intake data for 44 dietary antioxidants were derived from repeated 24-hour dietary recalls. Comorbidity status was ascertained by self-report of physician-diagnosed CVD and OA. After preprocessing and addressing multicollinearity, 9 interpretable machine learning (ML) models - including eXtreme Gradient Boosting (XGBoost), Random Forest, LightGBM, and others - were developed and benchmarked for comorbidity prediction. Model performance was evaluated using receiver operating characteristic-area under the curve and other standard metrics. Feature importance was further interpreted using SHapley Additive exPlanations (SHAP). XGBoost outperformed all other ML models, achieving the highest area under the curve (0.908, 95% CI: 0.873-0.942) and demonstrating robust accuracy in identifying CVD-OA comorbidity. SHAP analysis revealed that dietary catechins, vitamin A, epigallocatechin gallate, and zinc were the strongest protective factors for comorbidity risk. Participants with comorbidity exhibited significantly lower intakes of these antioxidants compared to controls. The model, based on accessible dietary and demographic variables, showed strong potential as a noninvasive clinical screening tool. Among 9 ML models, XGBoost showed the best predictive accuracy for CVD-OA comorbidity. SHAP analysis identified catechins, vitamin A, epigallocatechin, and zinc as the top protective dietary antioxidants.