Association between the WBC/HDL ratio and incident stroke in middle-aged and older adults: a longitudinal study from CHARLS

Inflammation and lipid dysfunction contribute to cerebrovascular disease. The white blood cell count-to-high-density lipoprotein cholesterol ratio (WBC/HDL) reflects both inflammation and the protective effects of HDL. Evidence relating WBC/HDL to incident stroke in middle-aged and older adults remains limited. Participants aged ≥ 45 years from CHARLS were included in the study. WBC/HDL, WBC, and HDL were evaluated per 1-standard deviation (SD) increment and by quartiles. Cox models, restricted cubic splines (RCS), exploratory subgroup analyses, and comparisons of model fit and discrimination were performed. The proportional hazards assumption was assessed using scaled Schoenfeld residuals, with time-varying effects modeled where appropriate. Sensitivity analyses addressed uncertainty in the estimated timing of stroke onset, death as a competing event, missing covariates, early stroke events, and potential acute inflammation. In the Model 3 complete-case analysis ( N = 6712; 635 stroke events), each 1-SD increase in WBC/HDL was associated with a higher risk of incident stroke (HR 1.109, 95% CI 1.050–1.170; P < 0.001). Stroke risk increased across WBC/HDL quartiles ( P for trend = 0.012). There was no evidence that WBC/HDL violated the proportional hazards assumption ( P = 0.191). The RCS analysis showed an overall association for WBC/HDL ( P overall = 0.006; P nonlinearity = 0.229), whereas the results for WBC were ( P overall = 0.134; P nonlinearity = 0.938) and for HDL were ( P overall = 0.050; P nonlinearity = 0.788). Model performance was comparable between the WBC/HDL model and the model containing WBC and HDL jointly, with no statistically significant differences in model fit or discrimination. The WBC/HDL–stroke association was supported by Fine–Gray competing-risk analysis (sHR per SD 1.102, 95% CI 1.045–1.162; P < 0.001) and multiple imputation (HR 1.106, 95% CI 1.052–1.162; P < 0.001). Higher WBC/HDL was independently associated with incident stroke, and the association remained robust across multiple sensitivity analyses. Compared with WBC or HDL alone, WBC/HDL showed more consistent associations across quartile and RCS analyses. When WBC and HDL were modeled jointly, however, WBC/HDL showed comparable model fit and discrimination, with no clear evidence of incremental predictive value beyond joint modeling of its components.