The absorption of regular human insulin from subcutaneous injection sites is delayed due to the self-association of insulin to multimeric forms. The insulin analogue insulin lispro has a weak self-association and a fast absorption rate. We examined the safety and efficacy of insulin lispro in the premeal treatment of patients with diabetes mellitus. A 12-month study was performed in 336 patients with insulin-dependent diabetes mellitus (IDDM) and 295 patients with non-insulin-dependent diabetes mellitus (NIDDM). The patients were randomized to inject either regular human insulin 30 to 45 minutes before eating, or insulin lispro immediately before each meal, in addition to basal insulin. The postprandial rise in serum glucose was lower in patients receiving insulin lispro than in those receiving regular human insulin therapy. At end point the increment was significantly lower at 1 hour (35%) and at 2 hours (64%) after the meal in IDDM patients; in NIDDM patients, the increment was nonsignificantly lower at 1 hour (19%) and significantly lower at 2 hours (48%). IDDM patients receiving insulin lispro achieved significantly lower glycated hemoglobin (HbA1c) levels in patients receiving regular human insulin (8.1% vs 8.3%). In NIDDM patients, HbA1c levels decreased equally in both treatment groups. Due to its fast absorption rate, insulin lispro improves postprandial control in diabetes. Insulin lispro can be considered one step toward optimal insulin therapy and improved patient convenience.