Insulin Resistance Indices Predict Mortality in Cardiovascular Disease: A Large‐Scale NHANES Study With Machine Learning Validation

Insulin resistance (IR) is a key driver of cardiovascular disease (CVD), the leading cause of global mortality. This study evaluated the prognostic value of two surrogate IR indices—the McAuley index and the Metabolic Score for Insulin Resistance (METS‐IR)—for predicting all‐cause and CVD mortality. Data from 22,308 NHANES participants with established CVD (1999–2018) was analyzed. Outcomes were all‐cause and CVD mortality. Cox proportional hazards models and restricted cubic splines assessed associations, while machine learning methods (random forest, XGBoost, CoxBoost, DeepHit) evaluated predictive performance. Model interpretability was assessed using SHapley Additive exPlanations (SHAP). Over a median 9.2‐year follow‐up, 3484 deaths occurred, including 1093 from CVD. A higher McAuley Index was inversely associated with risk, with each 1‐unit increase predicting a 9.2% reduction in all‐cause and 11.3% reduction in CVD mortality. Higher METS‐IR values were associated with increased mortality. Restricted cubic spline analysis confirmed significant U‐shaped relationships. Across nine models, the Cox model demonstrated the best performance (C‐index: 0.87 for all‐cause and 0.85 for CVD mortality), with time‐dependent AUCs consistently above 0.8. SHAP analysis highlighted the McAuley Index and METS‐IR as leading predictors. The McAuley Index and METS‐IR are robust, independent predictors of all‐cause and CVD mortality. Their integration with interpretable machine learning enhances risk stratification, underscoring the role of metabolic dysfunction and central adiposity in long‐term outcomes. These indices may help identify high‐risk patients who could benefit from targeted interventions.