PurposeThe debate over the causal and longitudinal association between cystatin C and stroke in older adults persists. Our aim was to assess the link between cystatin C levels, both measured and genetically predicted, and stroke risk.MethodsThis study employed a retrospective cohort design using samples of the China Health and Retirement Longitudinal Study (CHARLS), which is a nationally representative cohort recruiting individuals aged 45 years or above. A multivariate logistic model and the two-sample Mendelian randomization framework were used to investigate the longitudinal and genetically predicted effect of serum cystatin C on stroke.ResultsThe study population had a mean age of 59.6 (SD ±9.5), with 2,996 (46.1%) women. After adjusting for confounding factors, compared to those in the first quartile of cystatin C, those in the last quartile had the greatest risk of stroke incidence [odds ratio (OR), 1.380; 95% confidence interval (CI), 1.046–1.825]. The Mendelian randomization analysis showed that a genetically predicted cystatin C level was positively associated with total stroke (OR by inverse variance-weighted method, 1.114; 95% CI, 1.041–1.192).ConclusionsThis national cohort study suggests that higher serum cystatin C is associated with an increased risk of total stroke, which is further supported by Mendelian randomization.