The incidence of and risk factors for renal failure were determined in 912 Oklahoma Indians with non-insulin-dependent diabetes mellitus in a follow-up study conducted between 1987 and 1990. The incidence rate was 15.7/1,000 person-years after an average follow-up time of 10.2 years. Among those who had no qualitatively positive proteinuria at baseline, the incidence of renal failure was 10.3/1,000 person-years compared with 19.3- and 56.2/1,000 person-years, respectively, in those with slight and heavy proteinuria at baseline. Fasting plasma glucose (FPG) > or = 11.1 mM (200 mg/dl) increased the risk of renal failure to 2.9-fold (95% confidence interval [CI] = 1.9-4.6) higher than a level < 7.8 mM (140 mg/dl), and twofold (95% CI = 1.4-3.1) higher than a level between 7.8 (140 mg/dl) and 11.1 mM (200 mg/dl). The hypertensive patient had twice the incidence of renal failure than the normotensive subject (rate ratio = 2.1, 95% CI = 1.4-3.0). Patients with a lower blood pressure under antihypertensive medication had a lower incidence of renal failure than those whose hypertension remained uncontrolled with or without use of medication. Significant independent risk factors for renal failure, identified from Cox's proportional hazards model, were duration of diabetes, FPG, age, hypertension, and insulin use (P < 0.05). In patients without proteinuria at baseline, FPG and hypertension were significant predictors of renal failure as identified by multivariate analyses, whereas in patients who had proteinuria at baseline, insulin use was significant. Thus, hyperglycemic and hypertension control are suggested strongly for diabetic Oklahoma Indians as potential strategies to prevent the development of renal failure.