An insulin-modified frequently sampled intravenous glucose tolerance test (FSIGTT) with minimal model analysis was compared with the glucose clamp in 11 subjects with normal glucose tolerance (NGT), 20 with impaired glucose tolerance (IGT), and 24 with non-insulin-dependent diabetes mellitus (NIDDM). The insulin sensitivity index (S_I) was calculated from FSIGTT using 22- and 12-sample protocols (S_I(22) and S_I(12), respectively). Insulin sensitivity from the clamp was expressed as S_I(clamp) and S_IP(clamp). Minimal model parameters were similar when calculated with S_I(22) and S_I(12). S_I could not be distinguished from 0 in ∼ 50% of diabetic patients with either protocol. S_I(22) correlated significantly with S_I(clamp) in the whole group (r = 0.62), and in the NGT (r = 0.53), IGT (r = 0.48), and NIDDM (r = 0.41) groups (P < 0.05 for each). S_I(12) correlated significantly with S_I(clamp) in the whole group (r = 0.55, P < 0.001) and in the NGT (r = 0.53, P = 0.046) and IGT (r = 0.58, P = 0.008) but not NIDDM (r = 0.30, P = 0.085) groups. When S_I(22), S_I(clamp), and S^IP(clamp) were expressed in the same units, S_I(22) was 66 ± 5% (mean ± SE) and 50 ± 8% lower than S_I(clamp) and S_IP(clamp), respectively. Thus, minimal model analysis of the insulin-modified FSIGTT provides estimates of insulin sensitivity that correlate significantly with those from the glucose clamp. The correlation was weaker, however, in NIDDM. The insulin-modified FSIGTT can be used as a simple test for assessment of insulin sensitivity in population studies involving nondiabetic subjects. Additional studies are needed before using this test routinely in patients with NIDDM.