Preliminary data from our group indicated that cyclosporin A induced frequent remissions of insulin dependency in a group of 40 insulin-dependent (type I) diabetic children if given at the onset of clinical manifestations of diabetes. We report a 2-yr analysis of the response to cyclosporin A in the group of 81 patients included in the initial study. As observed before, a remission could be obtained in most of the patients (65%) in association with a shorter duration of symptoms, less severe hyperglycemia, lower incidence of ketoacidosis, and higher plasma C-peptide concentrations. All remissions ended during the follow-up period after a mean +/- SE duration of 316 +/- 21 days (range 31-850 days). Two parameters were linked to the duration of remissions: the mean circulating level of cyclosporin during the first 3 mo and the duration of prediagnostic polyuria. We were unable to relate the end of a remission to variations in the cyclosporin regimen, titer of autoantibodies, or progression of beta-cell failure. The euglycemic clamp technique revealed that insulin sensitivity decreases with time in patients not taking insulin. At 24 mo, the patients who had a remission of insulin dependency had better glycemic control, lower insulin dosages, and C-peptide levels two- to threefold higher than the nonremission patients and four- to sixfold higher than the historical control subjects. The cyclosporin regimen was well tolerated over the observed period: more specifically, serum creatinine remained unchanged, and kidney biopsies performed at 18-24 mo of treatment were within normal limits.(ABSTRACT TRUNCATED AT 250 WORDS)