Since in normal persons the hypoglycemic effect of low-dose intramuscular exceeds that of subcutaneous insulin we studied the effect of routes of insulin therapy in diabetic ketoacidosis. Forty-five patients with diabetic ketoacidosis entered a randomized prospective protocol with insulin administered either intravenously, subcutaneously or intramuscularly. Initial priming dose of insulin had to be repeated in two of 15, three of 15 and six of 15 of the intravenous, subcutaneous and intramuscular groups respectively. The intravenous group had a more rapid fall in plasma glucose (P less than 0.01) and ketone bodies (P less than 0.05) during the first two hours. Thereafter, there were no significant differences in the rate of decline of plasma glucose or ketones nor in the time required for glucose to reach 250 mg per deciliter or for complete recovery from diabetic ketoacidosis. The data confirm the efficacy of low-dose insulin therapy for diabetic ketoacidosis and indicate that the optimal route of insulin administration is by initial intravenous combined with subcutaneous or intramuscular.