To establish whether the short-acting insulin analog lispro can be successfully implemented in long-term intensive insulin therapy in type 1 diabetes, and if so, what its effects are on glycemic control and frequency and awareness of hypoglycemia. We randomized 56 type 1 diabetic patients to treatment with either lispro (n = 28) or human regular insulin (Hum-R; n = 28) as mealtime insulin for 1 year (open design, parallel groups). Lispro was injected at mealtime and Hum-R was given 10-40 min before meals (bedtime NPH was continued on both occasions). With lispro, NPH was added at breakfast (approximately 70/30), lunch (approximately 60/40), and supper (approximately 80/20) (mixing percentage of lispro/NPH) to optimize premeal and bedtime blood glucose. Total daily insulin units were no different in the two treatment groups, but with lispro approximately 30% less short-acting insulin at meals and approximately 30% more NPH was needed versus Hum-R (P < 0.05). The bedtime NPH dosage was no different. With lispro + NPH, the mean daily blood glucose was lower than with Hum-R (8.0 +/- 0.1 vs. 8.8 +/- 0.1 mmol/l; P < 0.05), HbA1c was lower (6.34 +/- 0.10 vs. 6.71 +/- 0.11%, mean value over 1 year; P < 0.002), and hypoglycemia (blood glucose < or = 3.8 mmol/l) was less frequent (7.4 +/- 0.5 vs. 11.5 +/- 0.7 episodes/patient-month) and tended to occur more within 90 min after meals than in the postabsorptive state (P < 0.05 vs. Hum-R). After 1 year, plasma adrenaline and symptom responses to experimental, stepped hypoglycemia improved with lispro and were closer to the responses of 12 nondiabetic control subjects versus Hum-R both in terms of thresholds and magnitude (P < 0.05). We concluded that mealtime injection of lispro + NPH improves the 24-h blood glucose and the percentage HbA1c as compared with Hum-R. The improvement can be maintained long term. Intensive therapy with lispro + NPH results in less frequent hypoglycemia and better awareness and counterregulation of hypoglycemia.