A period of early, intensive insulin treatment is thought to improve subsequent beta-cell function in insulin-dependent diabetes mellitus (IDDM). To study this hypothesis, we randomly assigned adolescents with newly diagnosed IDDM to receive either conventional treatment (n = 14) (NPH insulin, 1 U per kilogram of body weight per day, in two divided doses) or an experimental treatment (n = 12) (a two-week hospitalization with maintenance of blood glucose levels between 3.3 and 4.4 mmol per liter by continuous insulin infusion delivered by an external artificial pancreas [Biostator]). During the two-week intervention, the experimental-therapy group received four times more insulin than the conventionally treated group, and their endogenous insulin secretion was more completely suppressed, as evidenced by a urinary C-peptide excretion rate one seventh that of the conventionally treated group. After the first two weeks, both groups were treated similarly and received similar amounts of insulin. At one year, the mean (+/- SEM) plasma level of C peptide was significantly higher after mixed-meal stimulation in the experimental-therapy group than in the conventionally treated group (0.51 +/- 0.07 vs. 0.27 +/- 0.06; P less than 0.01). The experimental-therapy group also had better metabolic control, as evidenced by lower glycohemoglobin values (7.2 +/- 0.7 vs. 10.8 +/- 1.2 percent; P less than 0.01). We conclude that suppression of endogenous insulin by intensive, continuous insulin treatment during the first two weeks after the diagnosis of IDDM may improve beta-cell function during the subsequent year.