Background-Proteinuria often changes dynamically, showing either regression or progression. The impact of changes in proteinuria on future stroke risk remains largely unknown. We hypothesized that changes in proteinuria would be associated with stroke risk in patients with diabetes mellitus and prediabetes mellitus. Methods and Results-The study population included 17 380 participants with diabetes mellitus or prediabetes mellitus enrolled in a prospective Chinese cohort. From the baseline and 2-year dipstick screening results, participants were classified as having no proteinuria or remittent, incident, or persistent proteinuria. Reduction in proteinuria was calculated as the baseline minus 2-year proteinuria. Stroke outcomes were assessed in subsequent follow-ups. Data were analyzed using Cox proportional-hazards models. During a median follow-up of 6.9 years, we identified 751 patients with stroke. Stroke risk was increased for participants with persistent (hazard ratio [HR], 1.64; 95% confidence interval [CI], 1.18-2.30), incident (HR, 1.52; 95% CI, 1.22-1.89), and remittent (HR, 1.42; 95% CI, 1.01-2.02) proteinuria compared with those with no proteinuria. Persistent proteinuria was associated with a higher risk of stroke for participants with prediabetes mellitus (HR, 2.58; 95% CI, 1.58-4.22) compared with those with diabetes mellitus (HR, 1.35; 95% CI, 0.86-2.12 [P for interaction=0.0083]). Proteinuria reduction contributed to a decrease in stroke incidence (HR, 0.88; 95% CI, 0.81-0.95). The results were confirmed by sensitivity analyses. Conclusions-Persistent, incident, and remittent proteinuria are independent indicators of stroke risk in both diabetic and prediabetic populations.