The effect of metformin on Type 1 (insulin-dependent) diabetes has been assessed with the artificial pancreas. Fourteen Type 1 diabetic patients of normal body weight received in addition to their usual insulin therapy 850 mg metformin or placebo three times a day for 4-6 weeks. The sequence was placebo-metformin in eight patients and metformin-placebo in the other six. On the last day of metformin or placebo treatment, an artificial pancreas was used for about 36 h to assess insulin requirement. There was a 25.8% reduction in insulin requirement during metformin management despite slightly lower blood glucose levels (5.25 +/- 0.20 versus 5.98 +/- 0.18 mmol/l, p less than 0.01). Maximum reduction (about 50%) occurred 2 h after both lunch and dinner. There was no nocturnal effect. A marked decrease in specific insulin binding before metformin was found (0.56 +/- 0.27% to 10(7) monocytes versus 2.82 +/- 0.75 of control subjects) and significant increase after metformin (1.36 +/- 0.36%, p less than 0.05). There were no significant changes in blood lactate, total and HDL-cholesterol, triglycerides and C-peptide levels. These results show that insulin receptor binding is diminished in Type 1 diabetes, perhaps as a consequence of higher peripheral blood insulin levels and that metformin can improve binding, and so reduce the amount of insulin needed to reach euglycaemia. The insulin sparing effect is greatest after meals, and interference with intestinal absorption of sugars may also be important. It follows that metformin could be usefully administered to Type 1 diabetic patients with unimpaired liver and renal function to reduce their insulin requirement.