The comparative performance of insulin resistance and visceral adiposity composites over traditional indices for cardiometabolic multimorbidity risk stratification: a multi-dimensional analysis of 16 markers and 5 pathophysiological pathways in the CHARLS cohort

To compare the associations and incremental discriminatory ability of 16 alternative metabolic markers with incident cardiometabolic multimorbidity (CMM) and explore potential mediating pathways. Using the CHARLS cohort ( n = 7,646), we evaluated 16 markers. Primary analyses comprised Cox regression, competing-risk models, and ROC/IDI/NRI assessments of incremental discrimination. XGBoost, qgcomp, and SEM served as secondary/exploratory analyses of feature importance, mixture effects, and mediation pathways. During follow-up, 633 participants developed CMM. All markers were significantly associated with CMM in quartile analyses; CVAI showed the strongest association (Q4 vs. Q1 HR: 5.15, 95% CI: 3.52–7.54). TyG-WC provided the greatest incremental improvement in discrimination (IDI = 0.026, NRI = 0.255, P < 0.001). Composite insulin resistance (IR) and atherogenic lipid indices accounted for 87.8% of the joint risk weight. SEM revealed a suppression pattern: biochemical IR statistically accounted for the association between structural (anthropometric) adiposity and CMM (indirect effect: 0.34, P < 0.001; mediation proportion: 121.9%, reflecting inconsistent mediation rather than complete mediation), while the direct effect of structural adiposity was negative and non-significant ( P > 0.05). Integrating biochemical and anthropometric markers (particularly TyG-WC) improves CMM risk stratification, although the incremental gains were modest and require external validation before clinical implementation. The association between structural adiposity and CMM was statistically accounted for by biochemical IR; in exploratory SEM analyses, structural obesity without metabolic dysfunction was not independently associated with CMM in this cohort. Whether this reflects a metabolically healthy phenotype warrants further study. Not applicable.