Nonlinear association between remnant cholesterol inflammation index and incident stroke among middle-aged and older adults with hypertension: a CHARLS cohort study

The remnant cholesterol inflammation index (RCII) combines remnant cholesterol (RC) and high-sensitivity C-reactive protein (hs-CRP), two markers that may reflect residual lipid-related and inflammatory vascular burden. We examined whether baseline RCII was associated with incident stroke among hypertensive middle-aged and older adults and whether this association was nonlinear.

We analyzed 3,823 stroke-free hypertensive participants aged 45 years or older from the China Health and Retirement Longitudinal Study (CHARLS). Participants were excluded if they lacked follow-up information, were younger than 45 years, had missing or biologically implausible RCII components, had hs-CRP > 10 mg/L, or had missing covariates. RCII was analyzed per 1-standard deviation (SD) increase and by quartiles. Cox proportional hazards models, restricted cubic splines, an exploratory two-piecewise Cox model, subgroup analyses, and sensitivity analyses were performed.

During a median follow-up of 9.08 years (interquartile range, 9.08-9.17), 526 incident strokes occurred (13.8%). In the fully adjusted model, the original-scale linear estimate was positive but not statistically significant (HR per 1-SD increase, 1.06; 95% CI, 0.99-1.14; P = 0.110). Compared with Q1, adjusted HRs were 1.38 (95% CI, 1.06-1.81) for Q2, 1.66 (95% CI, 1.28-2.15) for Q3, and 1.60 (95% CI, 1.23-2.08) for Q4 (P for trend <0.001). Spline and threshold analyses suggested a nonlinear association, with a data-driven inflection point near RCII 3.80 and apparent flattening at higher values. Subgroup findings were broadly consistent, although current drinking status showed an exploratory interaction. Sensitivity analyses supported the robustness of the quartile-based associations.

Among hypertensive CHARLS participants, higher RCII categories were associated with incident stroke. The association was better characterized by categorical and nonlinear analyses than by a single original-scale linear estimate. RCII should be interpreted as an association marker, not as a validated predictive or clinical decision tool.