Association of an exploratory triglyceride glucose–visceral lipoprotein composite with incident stroke in CKM-staged Chinese middle-aged and older adults: a prospective CHARLS study

Cardiovascular-kidney-metabolic (CKM) syndrome links adiposity, insulin resistance, kidney dysfunction, and vascular disease. Whether a composite combining the triglyceride-glucose index (TyG), visceral adiposity, and LDL-C burden is associated with incident stroke beyond simpler component indices remains uncertain. We conducted a prospective landmark analysis of the China Health and Retirement Longitudinal Study. Baseline TyG-VLR used 2011 biomarkers, and cumulative TyG-VLR summarized 2011 and 2015 measurements. Participants free of stroke at the 2015 landmark were followed through 2020. Cox models used a covariate set that excluded the modifiable biochemical and adiposity components embedded in TyG-VLR while retaining age as an essential stroke confounder. Fine–Gray models accounted for death as a competing event. TyG, CVAI, VLR, and TyG-VLR were compared in mutually exclusive models using Harrell’s C statistic. Restricted cubic splines, medication analyses, 20-dataset multiple imputation, and selection-weighted and outcome-definition sensitivity analyses were performed. Among 5,226 participants, 452 incident strokes were reported over a median follow-up of 5.0 years. The principal fully adjusted model included 3,347 participants and 253 events. Each 1-SD increment in baseline TyG-VLR was associated with incident stroke (HR 1.15, 95% CI 1.02–1.30; P = 0.022), and Q4 versus Q1 had an HR of 1.62 (95% CI 1.10–2.38; P = 0.015). Cumulative TyG-VLR was not statistically significant per SD (HR 1.13, 95% CI 0.99–1.28; P = 0.070), although Q4 versus Q1 had an HR of 1.66 (95% CI 1.09–2.55; P = 0.019). Fine–Gray estimates were similar. The basic-model C statistic was 0.7028; it was 0.7034 after adding baseline TyG-VLR, compared with 0.7045 after adding TyG and 0.7032 after adding VLR. Higher TyG-VLR was associated with incident stroke, particularly at the upper end of the exposure distribution. However, the association was attenuated after comprehensive adjustment, cumulative continuous estimates were imprecise, and TyG-VLR did not improve model concordance or outperform simpler indices. TyG-VLR should therefore be regarded as an exploratory epidemiologic composite rather than a validated clinical prediction tool.