In this study, we evaluated the effects of high-(55%) and low-(40%) carbohydrate diets on insulin requirements in nine type 1 diabetic subjects treated intensively with ultralente as basal insulin and regular insulin as premeal insulin adjusted to the carbohydrate content of meals. Nine subjects were randomized in a crossover design to follow two diets consecutively for a period of 14 days each. A 3-day food diary was completed for each diet with the amount of carbohydrate in the mixed meals ranging from 21 to 188 g. Preprandial (5.9 vs. 6.1 mmol/l) and postprandial (8 vs. 8.9 mmol/l) capillary glucose and fructosamine (310 vs. 316 mumol/l) were comparable on both the low- and high-carbohydrate diets. The assessment of meal carbohydrate content by the patients was excellent, with > 85% of cases falling within 15% of computer-assisted evaluation. When premeal regular insulin was prescribed in U/10 g of carbohydrate, the postprandial glycemic rise remained constant (2.4 +/- 2.8 mmol/l) over a wide range of carbohydrate ingested (21-188 g) and was not affected by the glycemic index, fiber, and caloric and lipidic content of the meals. This tight control was maintained during the low- and high-carbohydrate diet without any change in insulin requirements (breakfast, 1.5 vs. 1.5 U/10 g of carbohydrate; lunch, 1.0 vs. 1.0; supper, 1.1 vs. 1.2) and in basal ultralente insulin requirements (22.5 vs. 21.4 U/day). These results indicate that in type 1 diabetic subjects 1) increasing the amount of carbohydrate intake does not influence glycemic control if premeal regular insulin is adjusted to the carbohydrate content of the meals; 2) algorithms based on U/10 g of carbohydrate are effective and safe, whatever the amount of carbohydrate in the meal; 3) the glycemic index, fiber, and lipidic and caloric content of the meals do not affect premeal regular insulin requirements; 4) wide variations in carbohydrate intake do not modify basal (ultralente) insulin requirements; and, finally 5) the ultralente-regular insulin regimen allows dissection between basal and prandial insulin requirements, so that each can be adjusted accurately and independently.