[Analysis on prediction and mediating effects of triglyceride-glucose body mass index on stroke incidence based on...

Objective: To investigate the prediction and mediating effects of triglyceride-glucose body mass index (TyG-BMI) on the incidence of stroke. Methods: A total of 18 062 residents free of stroke at baseline surveys of chronic disease cohort study in Pudong New District of Shanghai during 2013-2020 were followed until December 2024. Data were collected through questionnaires, physical examinations, and laboratory tests. Missing data were handled using multiple imputation. A Cox proportional hazards regression model was constructed to predict stroke risk and model performance was evaluated by using the concordance index (C-index), time-dependent receiver operating characteristic (ROC) curves, and the Brier score. Subsequently, the prediction and mediating effects of TyG-BMI were analyzed. Results: In the prediction model, the TyG-BMI index showed a HR of 1.00 (95%CI: 1.00-1.01), with a P=0.017. The model demonstrated a C-index of 0.71. Time-dependent area under ROC values were 0.770 (5-year), 0.717 (8-year), and 0.716 (10-year), while Brier scores were 0.005 3 (5-year), 0.043 7 (8-year), and 0.065 2 (10-year). Mediation analysis revealed hypertension mediated 51.9% of total effect of TyG-BMI on stroke (indirect effect: 0.028; total effect: 0.054), diabetes mediated 28.6% of total effect of TyG-BMI on stroke (indirect effect: 0.012; total effect: 0.042). Conclusions: The prediction model demonstrated reasonable accuracy and robustness, indicating that it could serve as a practical tool for screening the populations at high risk. TyG-BMI has both direct and indirect effects, mediated through hypertension or diabetes, on stroke incidence.