Remnant cholesterol inflammatory index and new-onset Fried-phenotype frailty in middle-aged and older Chinese adults: a prospective CHARLS cohort study

Abstract Background Composite indices combining remnant cholesterol with high-sensitivity C-reactive protein (hs-CRP) may capture frailty-relevant inflammatory–lipid risk. The remnant cholesterol inflammatory index (RCII) has been linked to cardiometabolic disease and Frailty-Index progression, but not to new-onset Fried-phenotype frailty. Framed as an etiologic question, natural-log-transformed RCII (lnRCII) was evaluated within the American Heart Association 2023 Cardiovascular–Kidney–Metabolic (CKM) framework, with the C-reactive protein–triglyceride–glucose index (CTI) as a pre-defined comparator. Methods Among 17,708 China Health and Retirement Longitudinal Study (CHARLS) Wave 1 (2011–2012) national-baseline respondents, 3,224 baseline non-frail adults aged ≥ 45 years (CKM Stages 0–3) with complete fasting biomarkers were followed to Wave 3 (2015). Because lnRCII and CTI share hs-CRP and are strongly correlated, they were modelled in parallel single-exposure Cox regressions, with a joint model as a collinearity check. Nomogram, decision-curve and machine-learning analyses were exploratory. Robustness to interval censoring, competing risk and selection was examined. Results Over 12,845 person-years, 212 participants (6.6%) developed frailty. Each 1-SD increment in lnRCII carried a 25% higher risk in the primary fully adjusted model (hazard ratio 1.25, 95% confidence interval 1.09–1.45, P = 0.002; top-quartile 1.94, 1.29–2.92; P -trend 0.004); CTI yielded 1.23 (1.06–1.43, P = 0.007). Associations were dose-dependent; no subgroup interaction was significant. Incremental discrimination over hs-CRP was negligible (ΔC ≈ 0.000) and the hs-CRP-free triglyceride–glucose index was null (1.03, P = 0.68), indicating that the association appears to be largely related to the inflammatory component. Two analyses were weaker: the competing-risk estimate (subdistribution hazard ratio 1.13, 0.98–1.31, P = 0.10) and lnRCII in the joint model with CTI (1.22, P = 0.12), as expected at r = 0.84. The exploratory nomogram reached an optimism-corrected concordance index of 0.753 (calibration slope 0.909), but its decision-curve net benefit over the covariate-only model was approximately zero. Conclusions In community-dwelling middle-aged and older Chinese adults, higher lnRCII was associated with new-onset Fried-phenotype frailty after covariate adjustment (dose-dependent), paralleling CTI; the association was not evident in crude analyses and was sensitive to modelling choices, and is best interpreted as a covariate-adjusted rather than an established independent association. Because the association appears to be largely related to the shared hs-CRP component and lnRCII adds no discrimination beyond hs-CRP, lnRCII is best regarded as an interpretable integrated etiologic marker rather than a validated clinical prediction tool. The exploratory prediction analyses require external and temporal validation before any clinical application.