Which insulin resistance indices best predict future frailty progression in a cardiovascular-kidney-metabolic syndrome stage 0–3 population: a national prospective cohort and machine learning study

Abstract

Background: Insulin resistance (IR) is implicated in frailty progression within Cardiovascular-Kidney-Metabolic (CKM) syndrome populations. While multiple non-insulin-based IR indices have been proposed, their comparative utility in predicting frailty across early CKM stages (0–3) remains unclear. Identifying reliable, non-insulin-based IR indices for predicting frailty index (FI) across early CKM stages (0–3) remains challenging. Methods: This prospective cohort study analyzed 4,354 adults (≥ 45 years) from the China Health and Retirement Longitudinal Study (2011–2015). We evaluated and compared 12 IR indices for predicting frailty progression. Associations were assessed using multivariable logistic regression. Machine learning (RFE, Boruta, and LASSO) identified optimal predictors, and a Random Forest (RF) model incorporating key covariates was developed and validated. Results: After full adjustment, CTI (OR = 1.19), TyG-WHtR (OR = 1.12), TyG-WC (OR = 1.00), and eGDR (OR = 0.87) significantly predicted FI (all p < 0.05). Among all indices, TyG-WHtR demonstrated the most stable discriminative performance across CKM stages 0–3 (AUCs: 0.52–0.59, all p < 0.001). Machine learning consistently selected TyG-WHtR as the top predictor. The final 13-variable RF model achieved an AUC of 0.74. SHAP analysis confirmed TyG-WHtR, age, depressive symptoms, and renal biomarkers as key predictors. Conclusion: TyG-WHtR is a robust, stable predictor of frailty progression in individuals with CKM syndrome stages 0–3. Its integration into clinical practice, potentially via the developed web tool, could enhance early frailty risk stratification.