To evaluate metabolic control and safety parameters (hypoglycaemia frequency and nasal mucosa physiology), 31 insulin-dependent diabetic patients were treated with intranasal insulin at mealtimes for 1 month and with subcutaneous fast-acting insulin at meals for another month in an open, crossover randomized trial. During both treatment periods the patients were treated with intermediate-acting insulin at bedtime. Six of the patients were withdrawn from the study during intranasal insulin therapy due to metabolic dysregulation. Serum insulin concentrations increased more rapidly and decreased more quickly during intranasal as compared with subcutaneous insulin administration. Metabolic control deteriorated, as assessed by haemoglobin A1c concentrations, slightly but significantly after intranasal as compared with subcutaneous insulin therapy. The bioavailability of intranasally applied insulin was low, since intranasal insulin doses were approximately 20 times higher than subcutaneous doses. The frequency of hypoglycaemia was similar during intranasal and subcutaneous insulin therapy, and nasal mucosa physiology was unaffected after intranasal insulin. We conclude that due to low bioavailability and to a high rate of therapeutic failure, intranasal insulin treatment is not a realistic alternative to subcutaneous insulin injections at the present time.