We conducted a randomized placebo-controlled study to determine the effects of the thiazolidinedione compound troglitazone on whole-body insulin sensitivity (S_I), pancreatic β-cell function, and glucose tolerance in 42 Latino women with impaired glucose tolerance (IGT) and a history of gestational diabetes mellitus (GDM), characteristics that carry an 80% risk of developing NIDDM within 5 years. After baseline oral (OGTT) and intravenous (IVGTT) glucose tolerance testing, subjects were assigned to take placebo or 200 or 400 mg troglitazone daily for 12 weeks (14 subjects per treatment group). An OGTT and IVGTT were repeated during the 12th week of treatment. Five subjects failed to complete the trial for personal reasons, and medication compliance averaged 90% in the remaining subjects, none of whom experienced a serious adverse event. S_I, calculated by minimal model analysis of IVGTT results, changed by only 4 ± 14% during 12 weeks of placebo administration, but increased 40 ± 22 and 88 ± 22% above basal during treatment with 200 and 400 mg troglitazone, respectively (P = 0.01 among groups). Troglitazone administration was also associated with a dose-dependent reduction in the total insulin area during IVGTTs, which was highly significant (P < 0.001), and with a reduction during OGTTs, which approached statistical significance (P = 0.09). Glucose tolerance improved slightly in all groups, but the magnitude of change did not differ significantly among groups, whether it was assessed as the number of subjects who continued to manifest IGT at 12 weeks (P = 0.64 among groups), the change in total glucose area during OGTTs (P = 0.58), or the change in fractional glucose disappearance rates during IVGTTs (P = 0.28). Among the women who received troglitazone, the greatest improvement in S_I occurred in the women who had the highest diastolic blood pressures and the best IVGTT insulin responses during baseline testing. Our findings indicate that troglitazone improved whole-body insulin sensitivity and lowered circulating insulin concentrations in women with prior GDM who are at very high risk for NIDDM. The lack of improvement in glucose tolerance despite improved insulin sensitivity may be a manifestation of the β-cell defect that predisposes the women to NIDDM. The overall pattern of response to troglitazone in our high-risk patients indicates that the drug is an ideal agent with which to test whether the amelioration of insulin resistance can delay or prevent diabetes in women with limited β-cell reserve.