Background/objectivesHMG-CoA reductase inhibitors (statins) reduce ischemic heart disease (IHD) in middle-aged diabetic individuals, and LDL-cholesterol (LDL-C) is a risk factor. However, their preventive effects on cerebrovascular attack (CVA) have not been identified in elderly, especially in elderly ≥75years (late elderly), who account for approximately 30% of diabetic individuals in Japan. Randomized controlled studies of statins for late elderly are difficult to carry out, because many co-morbidities in elderly disrupt randomized controlled conditions. MethodsWe performed a prospective cohort study (Japan Cholesterol and Diabetes Mellitus Study) with 5.5years of follow-up since 2004. A total of 4014 type 2 diabetic patients without previous IHD or CVA (n=1936 women; age=67.4±9.5years; ≥75years: n=1016) were enrolled, while 405 patients were registered as sub-cohort patients. We recorded detailed information on medications and laboratory data after every change in medication in patients of sub-cohort and suffered from IHD or CVA. We subdivided statin-users into prevalent, new and non-users. ResultsA total of 104 CVAs occurred during 5.5-years. Plasma HDL-C level was inversely correlated with CVA in patients ≥65years. In case–control study, among patients who were not prescribed statins, CVA increased in age-dependent manner. CVA incidence was lower in prevalent and new statin-users than in non-users (hazard ratio [HR]:0.46, 0.523), especially in late elderly (HR: 0.51, 0.21). Statins reduced CVAs mainly due to a direct effect and partially due to the effects of HDL-C and glucose metabolism. No significant differences were observed between statins. ConclusionStatins prevented CVA in middle-aged, elderly and late elderly diabetic patients via a direct effect. This study is the first to demonstrate the usefulness of observational studies for statistically analyzing agents' effects on late elderly.