Biological aging enhances long-term cardiovascular risk prediction in middle-aged and older adults with early-stage cardiovascular–kidney–metabolic syndrome

Introduction Cardiovascular–kidney–metabolic (CKM) syndrome defines a population at high long-term cardiovascular risk, particularly in its early stages, where accurate risk prediction remains challenging using conventional staging alone. Although the CKM framework provides a structured approach to cardiovascular prevention, substantial heterogeneity in cardiovascular disease risk exists within the same CKM stage. Biological aging, which reflects cumulative physiological dysregulation beyond chronological age, may capture this latent vulnerability and thereby enhance long-term cardiovascular risk prediction in individuals with early-stage CKM. Purpose To evaluate the association between a biological aging biomarker (Light BioAge) and incident cardiovascular disease (CVD) in individuals with CKM stages 0–3, and to develop an explainable machine learning model to enhance long-term cardiovascular risk prediction and within-stage stratification. Methods This population-based longitudinal study included 10,182 adults aged ≥45 years from two large prospective cohorts in China and the United States with CKM stages 0–3. Light BioAge was derived from routinely measured clinical biomarkers. Incident CVD was assessed using competing risk models. A random survival forest (RSF) model was developed for risk prediction and evaluated in an independent validation cohort using time-dependent discrimination, decision curve analysis, and model explainability to assess stage-specific risk patterns. Results During follow-up, 1,974 incident CVD events occurred. Participants in the highest quartile of Light BioAge had a significantly higher risk of CVD compared with those in the lowest quartile (subdistribution hazard ratio (sHR) 1.54, 95% confidence interval (CI) 1.33-1.79). In the validation cohort, the RSF model achieved strong long-term discrimination, with a 6-year time-dependent area under the curve (AUC) of 0.763 (95% CI 0.736-0.791). At the 6-year prediction horizon, decision curve analysis demonstrated clear clinical utility of the RSF model, yielding net benefits of 0.038 and 0.016 at 10% and 15% risk thresholds, respectively, exceeding those of the treat-all strategy. Model explainability analyses consistently identified Light BioAge as the most influential predictor and demonstrated heterogeneity in predicted risk across CKM stages, with pronounced risk escalation at stage 3. Conclusions Light BioAge is a clinically meaningful biomarker associated with incident CVD in early-stage CKM. An explainable machine learning–based RSF model incorporating biological aging improves long-term cardiovascular risk prediction and refines CKM stage–based stratification, supporting more precise and stage-informed prevention strategies.