Fasting hyperinsulinemia is a widely used surrogate measure of insulin resistance and predicts type 2 diabetes in various populations. Whether fasting hyperinsulinemia predicts diabetes independent of insulin resistance is unknown. In 319 Pima Indians with normal glucose tolerance, fasting plasma insulin concentration and insulin-stimulated glucose disposal (M) (hyperinsulinemic clamp) were inversely related, but at any given M, there was substantial variation, with some subjects being hyperinsulinemic and others being hypoinsulinemic relative to their degree of insulin sensitivity. In 262 of the 319 subjects followed prospectively over 6.4 +/- 3.9 years, a high fasting plasma insulin concentration was a significant independent predictor of diabetes, in addition to low M and low acute insulin response (AIR) (intravenous glucose challenge). In 161 of the 319 subjects with follow-up measurements of M and AIR (5.1 +/- 3.9 years), a high relative fasting plasma insulin concentration predicted a decline in AIR but not in M before the onset of diabetes. The adjusted fasting plasma insulin concentration was a familial trait (heritability of 0.52) and in a genome-wide scan, there was suggestive evidence of linkage (logarithm of odds score 1.77) to a region on chromosome 3q, which harbors the gene encoding GLUT2. These results provide the first prospective evidence in humans that fasting hyperinsulinemia itself has a primary role in the pathogenesis of diabetes, independent of insulin resistance. Whether amelioration of basal insulin hypersecretion will prevent diabetes remains to be elucidated.