This study, in 3 phases, compared the long-term acceptability and efficacy of insulin administered by nasal spray with an intensified subcutaneous regimen in nine type I (insulin-dependent) diabetic subjects on baseline therapy with ultralente insulin. In phase 1, patients were begun and stabilized on a regimen of ultralente daily and Actrapid insulin three times daily. Phase 2 consisted of 4 mo of this intensified subcutaneous regimen. In phase 3, intranasal administration of insulin, with 1% (wt/vol) sodium glycocholate, replaced Actrapid insulin for 4 mo. Glycemic control was compared in each of the three phases. It was possible to maintain the dose of ultralente insulin relatively constant in only six of the nine subjects during the intranasal phase of the study. The six subjects showed a significant rise in glycosylated hemoglobin during the intranasal phase (10.4 ± 0.6% intranasal vs. 9.1 ± 0.3% subcutaneous, P < .05) but not in plasma or urinary glucose levels. There was no significant change in the incidence of hypoglycemic episodes during intranasal insulin therapy in this group. The other three subjects were considered treatment failures. Six of the nine original subjects expressed a preference for intranasal insulin, and one subject complained of mild nasal irritation insufficient to cease treatment. The intranasal route of administration of insulin has the potential to replace short-acting insulin as an adjunct to longer-acting insulin in some insulin-treated diabetic patients.