Insulin Resistance Indices and Cardiovascular Disease Risk in Middle-Aged and Older Adults with Prediabetes: A Machine Learning-Assisted Predictive Modeling Study

Background: Insulin resistance (IR) indices—C-reactive protein–triglyceride–glucose index (CTI), triglyceride–glucose (TyG) index, and weight-adjusted waist index (WWI)—are linked to cardiovascular disease (CVD), but their comparative utility in prediabetic populations remains unclear. To evaluate the associations and exploratory mediating roles of CTI, TyG and WWI with prevalent CVD in middle-aged and elderly patients with prediabetes. Methods: The study included 3121 NHANES participants aged ≥45 with prediabetes. Four machine learning algorithms identified key variables. Multivariable logistic regression and restricted cubic spline analyses examined the associations between IR and CVD. Model performance was assessed via receiver operating characteristic curves, net reclassification improvement, integrated discrimination improvement, calibration, and decision curve analysis. Subgroup analyses tested robustness. Mediation analysis assessed the mediating roles of IR in the WWI–CVD relationship. External validation used CHARLS. Results: Individuals in the highest tertiles of TyG, TyG–WWI, and CTI had approximately twofold higher CVD risk (ORs: 1.80–1.97). RCS showed a significant linear increase in CVD risk with higher IR index levels. In fully adjusted models, AUCs improved markedly over baseline: TyG–WWI 0.771 vs. 0.579. Mediation analyses indicated that CTI and TyG mediated 22.13–26.30% of the WWI–CVD association. Subgroup analyses and external validation in CHARLS supported the robustness. Conclusions: IR, as captured by TyG, TyG–WWI, and CTI, independently elevates CVD risk among middle-aged and older adults with prediabetes. WWI is associated with greater odds of prevalent CVD, potentially driven by adiposity accumulation, with CTI and TyG partially mediating this effect.