The efficacy and safety of metformin in the treatment of obese, non-insulin-dependent, diabetic subjects poorly controlled by insulin after secondary failure to respond to sulphonylureas has been investigated. Fifty insulin-treated, obese diabetics participated in this prospective, randomised double-blind six-month trial. After a four-week run-in period, during which all patients were given placebo (single-blind), patients were randomly assigned to continue to receive placebo or to active treatment with metformin. At six months, there was a relevant and significant improvement in glycaemic control in diabetics receiving the combined insulin-metformin treatment (decrease in glucose -4.1 mmol.l-1; glycosylated haemoglobin A1 decrease -1.84%). No significant changes were seen in diabetics receiving insulin and placebo. There was a significant decrease in blood lipids (trygliceride and cholesterol), an increase in HDL-cholesterol and a reduction in blood pressure in diabetics taking metformin. These positive findings were most marked in the 14 diabetics who experienced a good response to metformin (glucose profile < 10 mmol.l-1), and were less marked but still significant in the remaining 13 diabetics, whose response to therapy was not so good (glucose profile > 10 mmol.l-1). The fasting insulin level was significantly lower after six months of combined insulin-metformin treatment as shown by a 25% reduction in the daily dose of insulin (-21.6 U/day). Metformin was well tolerated by all diabetics.(ABSTRACT TRUNCATED AT 250 WORDS)