A controlled trial of a new microprocessor device for insulin-dosage adjustment was undertaken in two matched groups of a priori well-controlled diabetic children. A prospective study design with three equal 8-wk periods was used. In the first period, both groups used manual methods for insulin-dosage adjustment after manual criteria. In the second period, one group of children adjusted insulin dosage by computer algorithms, whereas the other continued to use manual methods. In the third period, both groups again adjusted insulin by traditional methods. Mean premeal glycemia and glycosylated hemoglobin levels did not change in either group throughout the study. During the second period, episodes of hypoglycemia were more frequent in children without the computer than in those who used the device. In keeping with the latter outcome, the group that used the microprocessor device was given less insulin in the second period than the first (0.88 +/- 0.02 vs. 0.94 +/- 0.02 U.kg-1.day-1, P less than 0.0001) and in comparison to the control group of patients who concurrently were given an increased insulin dose in the second period compared with the first. This study showed that insulin treatment through specific computer-mediated dosage-adjusting algorithms was safe and minimized hypoglycemia by effectively accommodating seasonally changing insulin requirements. We recommend the device to help diabetic children and their families in the care of insulin-dependent diabetes.