OBJECTIVE Whether brachial-ankle pulse wave velocity (baPWV), a noninvasive marker for arterial stiffness, is a useful predictive maker for cardiovascular events in subjects with diabetes is not established. In the present cohort study, we evaluated the benefit of baPWV for the prediction of cardiovascular morbidity and mortality in subjects with diabetes. RESEARCH DESIGN AND METHODS A total of 4,272 outpatients with diabetes were enrolled in the Kyushu Prevention Study of Atherosclerosis. Of these, 3,628 subjects, excluding those with an ankle-brachial index of <0.9, were prospectively followed for 3.2 ± 2.2 years. The baPWV at baseline was classified by recursive partitioning (RP) for each end point. We plotted the Kaplan-Meier curves for high- and low-baPWV groups, which were designated based on the cutoff points, and calculated Cox proportional hazards models. RESULTS The elevation of baPWV quartiles was significantly correlated to the incidence of coronary artery events, cerebrovascular events, and all-cause mortality. RP revealed baPWVs of 14 and 24 m/s as statistically adequate cutoff points for cardiovascular events and mortality, respectively. High-baPWV classes showed significantly low event-free ratios in Kaplan-Meier curves for all end points and remained independent risks for all-cause mortality and cerebrovascular events, but not for coronary artery events after adjustments for age, sex, BMI, hypertension, hyperlipidemia, smoking, and hemoglobin A1c by Cox proportional hazards models. CONCLUSIONS This large-scale cohort study provided evidence that high baPWV is a useful independent predictor of mortality and cardiovascular morbidity in subjects with diabetes.