Cardiometabolic vulnerability and incident frailty across three aging cohorts: a multicohort longitudinal study of HRS, ELSA, and CHARLS

Abstract Background Cardiometabolic dysfunction is linked to frailty, but whether a fixed-weight biomarker score developed for one functional aging outcome retains an association with incident frailty without re-estimating its weights is unknown. We evaluated the systolic blood pressure–HDL–A1c–CRP (SHAC) index, originally developed for incident mobility limitation, across three aging cohorts. Methods We analyzed adults aged ≥ 65 years from HRS, ELSA, and CHARLS. Participant-specific baseline was the first eligible assessment with complete SHAC biomarkers and a valid frailty index (FI); participants frail at baseline were excluded. Incident frailty was defined as FI ≥ 0.25. Cohort-specific Cox models estimated hazard ratios (HRs) per 1-SD higher SHAC using the original fixed weights and cohort-internal standardization, and fully adjusted estimates were pooled by random-effects meta-analysis. Sensitivity analyses addressed exposure scaling, baseline FI, outcome timing, missingness and selection, competing mortality, and longitudinally updated SHAC. Results Fully adjusted analyses included 6,965 HRS participants with 3,334 events, 4,150 ELSA participants with 1,192 events, and 1,473 CHARLS participants with 476 events. HRs (95% CIs) were 1.12 (1.08–1.16), 1.21 (1.14–1.29), and 1.19 (1.08–1.30), respectively. The pooled HR was 1.16 (1.10–1.23; I²=65.9%). Adjustment for baseline FI attenuated the associations to 1.05 (1.01–1.09), 1.13 (1.06–1.20), and 1.12 (1.02–1.23). Fixed ELSA development-reference scaling also retained positive associations. Findings were broadly similar across other sensitivity analyses. HbA1c was the individual biomarker most consistently associated with incident frailty. Adding SHAC produced small increases in apparent discrimination, but SHAC did not clearly outperform HbA1c alone. Conclusions Fixed-weight SHAC was positively associated with subsequent observed frailty across HRS, ELSA, and CHARLS. These findings support further study of the previously specified SHAC weight structure as a marker of cardiometabolic vulnerability in functional aging. Cohort-dependent scaling, baseline-FI attenuation, selection, heterogeneity, and modest discrimination limit stronger claims regarding universal score transportability or clinical prediction.