To improve criteria for entry into future trials of immunosuppression, we enrolled 40 children with recent-onset Type I insulin-dependent diabetes in a pilot trial of cyclosporine. Twenty-seven patients were able to discontinue insulin therapy 48 +/- 5 days after the start of immunosuppression. At four months, their fasting and postprandial blood glucose concentrations averaged 110 and 160 mg per deciliter (6.1 and 8.9 mmol per liter) with a mean hemoglobin A1c level of 6.15 percent. Seventy-five percent of these patients with early remission still did not need insulin at 12 months, and their glycemic control was similar to that at 4 months. The major differences between the 27 patients with remission and the 13 without remission were the duration of symptoms before diagnosis (26.8 vs. 48.0 days, P less than 0.01), the degree of weight loss (3.2 vs. 10 percent of body weight, P less than 0.001), the initial hemoglobin A1c level (10.7 vs. 13.2 percent, P less than 0.001), and the frequency of ketoacidosis (11 vs. 61.5 percent, P less than 0.001). The lesser degree of weight loss was the strongest independent predictor of remission. The response of C-peptide to intravenous glucagon (0.50 vs. 0.17 pmol per milliliter, P less than 0.05) was also an independent predictor. No differences were observed between the two groups of patients in age, sex, HLA phenotype, autoantibodies to insulin or islet-cell antigens, or doses or trough levels of cyclosporine. Only minimal manifestations of toxicity were detected over the period of observation. We conclude that early treatment with cyclosporine in children with recent-onset Type I diabetes can induce remission from insulin dependence, with half the patients not requiring insulin after a full year.