Minimal model (MINMOD) analysis of the frequently sampled intravenous glucose tolerance test (FSIVGTT) is dependent on an adequate insulin response to the glucose load. As this is characteristically deficient in subjects with non-insulin-dependent diabetes mellitus (NIDDM), the technique has been modified by the use of an intravenous bolus of insulin. Previous validation of this modification in humans has relied on agreement between insulin sensitivity indexes (SI) estimated from tolbutamide- and insulin-modified tests and not on direct comparison with estimates derived from the isoglycemic glucose clamp. We have compared estimates of insulin sensitivity derived from minimal modeling of a 4-h insulin-modified FSIVGTT and the glucose clamp in subjects with NIDDM. Twelve subjects underwent an insulin-modified FSIVGTT and an isoglycemic hyperinsulinemic clamp in random order 2–4 weeks apart. Fasting plasma glucose (8.4 vs. 9.0 mmol/l) and immunoreactive insulin (IRI) concentrations (104.5 vs. 101.5 pmol/l) were not different between the 2 study days. SI(clamp) was derived from the steady-state glucose infusion rate during the 3rd h of the clamp, corrected for the ambient insulin and glucose concentrations. SI(ivgtt) was derived using MINMOD. SI(ivgtt) was 1.06 ± 0.18 min−1 · mU−1 · ml × 104, and mean SI(clamp) was 4.97 ± 0.69 l · min−1/pmol · 1−1 × 104 (mean ± SE). SI(ivgtt) was positively correlated with SI(clamp) (r = 0.73, P = 0.004) and negatively correlated with body mass index (r = −0.7, P = 0.005) and fasting IRI(ivgtt) (r = −0.64, P = 0.008). In summary, MINMOD analysis of the insulin-modified FSIVGTT provides a valid measure of insulin sensitivity in subjects with NIDDM.