The mechanism of action of metformin was studied by comparing glucose turnover before and after a 75-g oral glucose load in 10 nonobese men with non-insulin-dependent diabetes mellitus (NIDDM) during metformin and placebo therapy by the combined application of the forearm and double-isotope techniques. During the study, 9 of the 10 patients were regularly receiving glibenclamide therapy. In 5 of the men, the first study was performed during metformin therapy, and the second study was done during placebo administration; in the other 5 subjects, the order was reversed. The interval between the studies was at least 3 mo. The metformin dosage was 1 g twice daily in 9 of the patients and 850 mg thrice daily in the 10th subject. In the basal state, metformin administration reduced plasma glucose levels from 172 +/- 14 to 103 +/- 9 mg/dl (P less than .005), hepatic glucose output (HGO) from 2.67 +/- 0.15 to 2.20 +/- 0.20 mg X kg-1 X min-1 (P less than .02), and forearm glucose uptake (FGU) from 0.106 +/- 0.18 to 0.039 +/- 0.016 mg X 100 ml-1 forearm X min-1 (P less than .005), whereas insulin (23 +/- 6 microU/ml) and lactate (1.56 +/- 0.18 mM) levels were unchanged. Although the oral glucose tolerance curve (OGTC) was significantly lowered by metformin, the incremental area under the curve and the insulin response were unchanged. The systemic appearance of ingested glucose was unaffected by metformin; 64 +/- 2% of the load was recovered peripherally in 3 h.(ABSTRACT TRUNCATED AT 250 WORDS)