In 20 patients with non-insulin-dependent diabetes mellitus (NIDDM) and secondary failure to sulfonylurea, a double-blind randomized study was performed comparing two regimes: insulin plus placebo (IP) and insulin plus glyburide (IG). The protocol included two hospitalization periods (days 1-18 and 78-85) and follow-up at the outpatient clinic for 325 days. The metabolic control was kept as tight as possible. The subjects underwent normoglycemic clamp studies and meal tests with determination of insulin, C-peptide, glucagon, somatostatin, and gastric inhibitory polypeptide in plasma. On IG, they demonstrated marked and long-lasting improvement of metabolic control: HbA1c decreased from 11.1 +/- 0.3% on day 3 to 8.3 +/- 0.4% (P less than .001) on day 78 and 9.1 +/- 0.5% (P less than .001) on day 325. In subjects on IP, the corresponding values were 10.3 +/- 0.5, 8.4 +/- 0.4 (P less than .001), and 8.9 +/- 0.5% (P less than .05). Body weight increased by 6.0 +/- 1.5 kg (P less than .005) on IG and 2.9 +/- 2.1 kg (NS) on IP. The daily insulin requirement decreased on IG from 62.5 +/- 12.9 U/day on day 7 to 33.5 +/- 8.8 U/day on day 83 and 34.6 +/- 8.9 U/day on day 325. On IP the insulin requirement was almost constant: 62.0 +/- 10.7 U/day on day 7, 55.5 +/- 7.7 U/day on day 83, and 54.7 +/- 7.9 U/day on day 325. Insulin sensitivity measured with the hyperinsulinemic clamp (plasma insulin approximately equal to 130 microU/ml) was similar on IP and IG at the initiation of the study and was unchanged on days 18 and 85. A key observation of this study, although the mechanism is unclear, is that isoglycemic-meal-related insulin requirement was diminished by insulin treatment, indicating improvement of meal-related insulin sensitivity. Glyburide increased basal and meal-but not glucagon-stimulated insulin and C-peptide levels, and also augmented the effect of meals on somatostatin release. We conclude that in NIDDM, IG regime promptly and continuously decreased insulin requirement and improved metabolic control. This effect is, at least during the first 3 mo, mainly due to enhanced insulin secretion. IG and IP treatment had no effect on insulin sensitivity during hyperinsulinemic-normoglycemic clamp, whereas meal-related insulin sensitivity was augmented.