Nine triglyceride-glucose-frailty composite indices and incident cardiovascular disease across cardiovascular-kidney-metabolic syndrome stages 0–3: a prospective cohort study

Cardiovascular-kidney-metabolic (CKM) syndrome links metabolic, renal and cardiovascular disease (CVD). Evidence comparing composite indices that combine the triglyceride-glucose (TyG) index, frailty and adiposity or inflammation across CKM Stages 0–3 is limited. We evaluated nine TyG-frailty (TFI) composite indices in relation to incident CVD across CKM Stages 0–3. In this prospective cohort study of 6,083 participants from the China Health and Retirement Longitudinal Study (CHARLS, 2011–2020), nine TFI-related indices were assessed. TFI-WC was designated the primary index, and the remaining eight were analysed as exploratory exposures. Multivariable Cox regression, restricted cubic splines, subgroup analyses, internal validation of prediction performance, and exploratory machine-learning and clustering analyses were performed. Over a median 9.0-year follow-up, 1,644 incident CVD events occurred (27.03%), with incidence rising across CKM Stages 0–3 from 15.01% to 31.82%. TFI-WC showed the strongest association among the nine indices (per 1-SD HR 1.258, 95% CI: 1.204–1.316; p < 0.0001), with a nonlinear association, and the remaining eight indices were also significant (all p < 0.0001). Associations were consistent across CKM stages, with no significant interaction (all p > 0.05). Adding TFI-WC to the base model improved time-dependent area under the receiver operating characteristic curve (AUC) from 0.623 to 0.641 at 3 years, from 0.652 to 0.669 at 5 years and from 0.639 to 0.665 at 9 years, and yielded the highest 5-year integrated discrimination improvement (IDI) (0.0075, 95% CI: 0.0059–0.0091). The remaining eight indices also added incremental predictive value (all p < 0.0020). Logistic regression showed the highest machine-learning discrimination (AUC = 0.639), and SHAP analysis identified each composite index as a leading contributor within its own model. Clustering identified three burden strata, the highest carrying a 1.797-fold increased risk (95% CI: 1.530–2.111). TFI-related indices were positively and nonlinearly associated with incident CVD across CKM Stages 0–3. TFI-WC, the primary index, showed the strongest association among the nine indices and the highest time-dependent discrimination at 3, 5 and 9 years. Clustering identified three burden strata with progressively increasing risk. These findings suggest that TFI-WC may be a practical instrument for risk stratification in middle-aged and older adults, pending external validation.