To examine the association between diabetes and prediabetes at baseline, and disability, mortality over a 22-year period among middle-aged Japanese adults. Participants consisted of 1,788 adults aged 45-64 years at baseline from the cohort study National Integrated Project for Prospective Observation of Non-communicable Disease and its Trends in the Aged 1990 (NIPPON DATA90). Disability, defined as having a decline in activities of daily living (ADL), was assessed by a modified Katz questionnaire at four time points. Disability and death without disability for 22-year follow up were used as outcomes to test the association with a diagnosis of diabetes or prediabetes at baseline, using multinomial logistic regression. Adjusted odds ratios (ORs) were obtained from four models that contained appropriate adjustment factors, such as age, sex, smoking status, drinking status, body mass index and cardiovascular risk factors (hypertension, hypercholesterolemia, triglycerides, low serum high-density lipoprotein), at baseline. In the present study, 334 participants (18.7%) reported at least one disability, and 350 (19.6%) were reported dead without observation of disability during follow up. Adjusting sex and other risk factors, participants with diabetes and prediabetes had a higher risk for disability (OR 1.43, 95% confidence interval [CI] 1.07-1.91 and OR 1.66, 95% CI 1.10-2.50, respectively) and for mortality (OR 1.56, 95% CI 1.16-2.08 and OR 1.77, 95% CI 1.18-2.65, respectively) than individuals with normal glucose tolerance. In middle-aged Japanese adults, individuals with diabetes and prediabetes were more likely to be associated with disability and mortality. Our findings suggest that prediabetes and diabetes in middle-aged adults should be paid more attention, and requires more intervention to prevent disability and mortality in later life.