Troglitazone and metformin lower glucose levels in diabetic patients without increasing plasma insulin levels. We compared the insulin sparing actions of these two agents and their effects on insulin sensitivity and insulin secretion in 20 type 2 diabetic patients. To avoid the confounding effect of improved glycemic control on insulin action and secretion, patients were first rendered euglycemic with 4 weeks of continuous subcutaneous insulin infusion (CSII) before randomization to CSII plus troglitazone (n = 10) or CSII plus metformin (n = 10); euglycemia was maintained for another 6-7 weeks. Insulin sensitivity was assessed by a hyperinsulinemic-euglycemic clamp 1) at baseline, 2) after 4 weeks of CSII, and 3) after CSII plus either troglitazone or metformin. The 24-h glucose, insulin, and C-peptide profiles were performed on the day before the second and third glucose clamps. Good glycemic control was achieved with CSII alone and was maintained with CSII plus an oral agent (mean 24-h glucose: troglitazone, 6.2+/-0.6 mmol/l; metformin, 6.2 +/-0.3 mmol/l). Insulin requirements decreased 53% with troglitazone compared with CSII alone (48+/-4 vs. 102+/-13 U/day, P < 0.001), but only 31% with metformin (76+/-13 vs. 110+/-18 U/day, P < 0.005). The 24-h C-peptide profiles were similar. Normal fasting hepatic glucose output was maintained with both agents despite lower insulin levels than on CSII alone. Insulin sensitivity did not change significantly with CSII alone or with CSII plus metformin, but improved 29% with CSII plus troglitazone (P < 0.005 vs. CSII alone) and was then 45% higher than in the CSII plus metformin patients (P < 0.005). In conclusion, metformin has no effect on insulin-stimulated glucose disposal independent of glycemic control in type 2 diabetes. Troglitazone (600 mg/day) has greater insulin-sparing effects than metformin (1,700 mg/day) in CSII-treated euglycemic patients. This is probably explained by the peripheral tissue insulin-sensitizing effects of troglitazone.