Abstract / Summary
Cardiometabolic multimorbidity (CMM) is an important health burden in middle-aged and older adults. CTI-FI combines the C-reactive protein–triglyceride glucose index (CTI) and frailty index (FI) to summarize inflammatory–metabolic burden and accumulated health deficits. Its association with incident CMM and the contribution of repeated measurements remain uncertain. To examine the associations of Wave 3 CTI-FI, the time-weighted two-point mean CTI-FI, and CTI-FI two-point longitudinal exposure patterns with incident CMM, and to assess their performance relative to the individual components and alternative combined representations. We included 5,586 participants from the China Health and Retirement Longitudinal Study (CHARLS) with CTI-FI measurements at Wave 1 (2011–2012) and Wave 3 (2015) who remained free of CMM through Wave 3. Follow-up began at Wave 3 and continued through 2020. CMM was defined as the coexistence of at least two of three conditions: heart disease, stroke, and diabetes. CTI-FI was calculated as CTI multiplied by FI, and two-point longitudinal exposure patterns were derived using K-means clustering. Associations were assessed using Cox models and restricted cubic splines. Sensitivity analyses addressed death as a competing event, interval-based outcome ascertainment, extreme exposure values, and selection related to exposure availability. Exploratory comparisons assessed model fit and predictive performance, including 3- and 5-year prediction measures and bootstrap internal validation. During a median follow-up of 5.00 years, 494 participants developed CMM. In fully adjusted Cox models stratified by baseline cardiometabolic disease status, hazard ratios (HRs) per 1-standard-deviation (SD) increment were 1.342 (95% confidence interval [CI], 1.246–1.445) for Wave 3 CTI-FI and 1.370 (95% CI, 1.273–1.476) for the time-weighted two-point mean CTI-FI. Corresponding HRs for the highest versus lowest quartile were 3.263 (95% CI, 2.316–4.599) and 4.217 (95% CI, 2.938–6.053). The high-level pattern was associated with a higher CMM hazard than the low-level pattern (HR, 2.072; 95% CI, 1.707–2.515). Both continuous CTI-FI measures showed nonlinear associations with incident CMM. Associations were consistent across sensitivity analyses. CTI contributed additional longitudinal information beyond FI in model-fit comparisons, but gains in discrimination were small. Neither CTI-FI nor its repeated-measurement summaries showed a consistent predictive advantage over corresponding FI-based models. Higher CTI-FI across current and repeated-measurement dimensions was associated with incident CMM. These findings support examining inflammatory–metabolic burden alongside accumulated health deficits in studies of cardiometabolic disease progression. Clinical use of CTI-FI requires further validation, particularly given the absence of a consistent predictive advantage over FI.