Previous studies have indicated that beta-cell dysfunction predicts the development of diabetes, although it is unknown whether the use of combinations of insulin secretory measures further improves prediction. The Insulin Resistance Atherosclerosis Study is a prospective, multicenter, epidemiological study of the relationship between insulin sensitivity and the risk of diabetes and cardiovascular disease. At baseline, fasting concentrations of insulin, intact proinsulin (PI), and split PI were measured, and acute insulin response (AIR) was determined during a frequently sampled intravenous glucose tolerance test (FSIGTT). Subjects who were nondiabetic at baseline (n = 903) were reexamined after 5 years of follow-up; 148 had developed diabetes. In separate logistic regression models adjusted for age, sex, clinic, and ethnicity, 1 SD differences in measures of beta-cell dysfunction were associated with diabetes incidence (AIR: odds ratio [OR] 0.37, 95% CI 0.27-0.52; intact PI: OR 1.90, 95% CI 1.57-2.30; split PI: OR 1.94, 95% CI 1.63-2.31). After additional adjustment for BMI, impaired glucose tolerance, and insulin sensitivity, these measures continued to be significantly associated with risk of diabetes (all P < 0.0001). Furthermore, in models that included both PI and AIR, each was an independent predictor, and individuals who had combined low AIR and high PI experienced the highest diabetes risk. In conclusion, both low AIR and high PI independently predicted diabetes in a well-characterized multiethnic population. Although fasting PI is simpler to assess, determining AIR from an FSIGTT may further improve prediction. If pharmacological agents to prevent diabetes are proved to be efficacious in ongoing clinical trials, then it may be beneficial to perform FSIGTTs to identify better (for intensive intervention) prediabetic subjects who would ultimately require lifelong pharmacological therapy.