We report our first experience with a new microprocessor device for assisting individuals with diabetes in the adjustment of insulin therapy. The computer weighs 200 g and can receive, store, and analyze patient-entered capillary blood glucose (CBG) data on an ongoing basis. Changes in the injected mixtures of short- and intermediate-acting insulins are recommended according to algorithms designed to bring the premeal CBG levels to any desired target value set by the physician. Throughout the present study, the premeal target glucose level was set to 110 mg/dl. Seven (type I) insulin-dependent diabetic individuals ranging in age from 11 to 43 yr were selected to participate in the first use of a BCMC (Better Control Medical Computers, Inc., Toronto, Ontario, Canada) computer on an outpatient basis. All subjects were concerned about diabetes control and were fully informed about as well as thoroughly practiced in the use of manual insulin dosage adjustment schemes, based on ≈ 4 times daily CBG estimation, as currently taught in our diabetes clinics. During the last 7 days of the control period of self-adjustment, their mean ± SEM CBG levels (measured before breakfast, lunch, dinner, and bedtime snack) were, respectively, 178 ± 20, 187 ± 35, 208 ± 22, and 207 ± 13 mg/dl. Immediately after the control period they were given the device and were instructed in the procedure for entering glycemic data and following manufacturer's recommendations in regard to insulin dosages. This experimental period lasted 8 wk and the outcome was assessed as before. Thus, 8 wk after starting daily use of the instrument, all glycemic values measured as before had fallen significantly (P < 0.005–0.05) closer to normal: 116 ± 9, 110 ± 6, 148 ± 15, and 135 ± 9 mg/dl, respectively. Concurrently there was also a significant (P < 0.01) reduction in the variability of glycemia measured before the main meals. During this 2-mo interval, glycosylated hemoglobin A1 fractions also fell in each subject and for the group the mean fell from 11.7 ± 0.5% to 9.6 ± 0.3% (P < 0.005). (Upper limit of normal was 9.0%.) Changes in life-style were neither necessary nor encouraged. Individual-injected insulin dosages in the twicedaily mixtures were modified for each subject, but overall for the group, only the mean intermediateacting insulin at breakfast was increased significantly from 28 ± 3 to 39 ± 5 U (P < 0.05). Immediately after the experimental period, the device was returned by 6 of 7 patients, who thereafter reverted to manual-dosage adjustment similar to the control period. During this follow-up period, metabolic control deteriorated, as evidenced by increases in glycosylated hemoglobin in all except the one subject who kept the device and in whom this index of diabetes control has remained unchanged at a level in the normal range for > 18 mo. We conclude from these studies that a microprocessor-based device can receive, store, and interpret patient-entered CBG measurements and in reference to these data recommend day-to-day changes in insulin therapy. Over a period of 8 wk, these dosage changes clearly improved glycemic control. In this regard, reductions occurred in both the mean premeal glycemia and its variability, as well as in the glycosylated hemoglobin fractions, all key indices of metabolic control. In light of these findings, it appears that the use of a computer may well provide both the concerned individual with diabetes mellitus and his physician with a new and practical method for significantly improving metabolic control by directly optimizing conventional insulin therapy.