Abstract Background Acute Coronary Syndrome (ACS) remains a leading cause of cardiovascular morbidity and mortality worldwide. Lifestyle factors, including diet, hydration, and medication adherence, influence ACS outcomes, and fasting introduces significant metabolic and behavioral modifications. Prior studies report conflicting results, with some suggesting cardioprotective benefits due to improved metabolic control, reduced oxidative stress, and enhanced cardiovascular function, while others highlight risks such as dehydration, electrolyte imbalance, and medication non-adherence. Current risk stratification methods may not fully capture fasting-related ACS risks. This study evaluates the impact of 30 days of continuous 12hour fasting on ACS incidence and mortality while leveraging machine learning (ML) models for improved risk stratification compared to traditional scoring systems. Purpose This study aims to determine whether 30 day continuous 12hour fasting affects ACS incidence and short-term mortality and whether ML-based models provide superior predictive accuracy in fasting-related ACS risk stratification. By integrating ML techniques, we seek to identify key clinical and biochemical predictors of ACS among fasting individuals, enhancing precision in risk assessment, informing clinical decision-making, and refining ACS prevention strategies. Methods A retrospective cohort study analyzed 91,881 ACS patient records (2006–2019) from the Malaysian National Cardiovascular Disease Database (NCVD-ACS). Patients were categorized into fasting (n=6,923) and non-fasting (n=84,958) groups based on admission timing relative to the month of Ramadan. Ramadan is based on the Muslim calendar and is observed for a period of 30days. Data preprocessing included Multiple Imputation by Chained Equations (MICE) and k-Nearest Neighbors (KNN) for missing values, while the Synthetic Minority Over-sampling Technique (SMOTE) balanced class distributions. Feature selection was performed using L1 regularization, recursive feature elimination (RFE), and tree-based importance rankings. ML models—including Logistic Regression, Random Forest, XGBoost, and a Stacking Model—were trained and validated using AUC-ROC, accuracy, precision, recall, and F1-score. SHAP (Shapley Additive Explanations) analysis was conducted for the fasting group to interpret model predictions and identify key predictors of ACS incidence and mortality, enhancing model transparency. Results This is the largest cohort to date, examining the impact of fasting to CV outcome. Continuous 30day fasting was associated with a modest reduction in ACS incidence, but did not significantly impact in-hospital mortality. The most impactful time frame was during the fasting period, where ACS incidence reached its lowest.ACS Distribution by Grp ML Performance Matrix
Kasim et al. (Sat,) studied this question.