Predictive value of cholesterol, high-density lipoprotein and glucose (CHG) index combined with C-reactive protein for stroke risk: a cohort study

Background Systemic inflammation and insulin resistance (IR) are closely linked in cerebrovascular disease. Triglyceride–Glucose (TyG) index and cholesterol–high-density lipoprotein–glucose (CHG) index are proposed markers of IR. Yet no existing biomarker integrates an inflammatory marker with CHG index for stroke risk prediction. This study combined C-reactive protein (CRP) with CHG to create a novel composite index (CCHG) and sought to validate its predictive value for stroke. Methods In this cohort study, 6,804 adults aged ≥ 45 years were recruited from the CHARLS database, with data collected from 2011 to 2020. Stroke events were recorded during a median nine-year follow-up. The TyG, CHG, and CCHG indices were calculated, and their associations with stroke risk were evaluated using univariate and multivariate Cox regression models. Restricted cubic spline (RCS) analysis was employed to investigate potential non-linear dose–response relationships. Receiver operating characteristic (ROC) curve analysis was used to compare the predictive performance of the three indices, and subgroup analyses were conducted to assess their applicability across different populations. Results Among the 6,804 participants, 675 (9.9%) developed stroke during the nine-year follow-up. In unadjusted Cox regression models, the TyG index (HR = 1.37, 95% CI: 1.24–1.52, p < 0.001), the CHG index (HR = 1.64, 95% CI: 1.41–1.90, p < 0.001), and the CCHG index (HR = 1.52, 95% CI: 1.37–1.68, p < 0.001) were all significantly associated with stroke risk. After full multivariable adjustment (Model 3), the HR with 95% CI for incident stroke was 1.30 (1.15–1.47) for CCHG, 1.34 (1.10–1.62) for CHG, and 1.23 (1.09–1.39) for TyG. These associations persisted in the adjusted models. RCS analysis revealed a linear relationship between the TyG index and stroke risk, whereas the CHG and CCHG indices showed positive non-linear associations. ROC curve analysis demonstrated that the CCHG index had a modestly higher discriminative ability than the TyG and CHG indices in predictive performance. Subgroup analyses revealed no significant interactions for CHG or CCHG across subgroups ( p for interaction > 0.05). By contrast, an interaction was observed between the TyG index and hypertension ( p for interaction = 0.046). Conclusions An elevated CCHG index is significantly correlated with an increased risk of stroke, demonstrating a non-linear relationship. Furthermore, the CCHG index showed modestly better discriminative ability for incident stroke compared with CHG and TyG indices. Clinical trial number Not applicable.