A total of 166 patients (102 type 1, 64 type 2) were randomised to insulin lispro (LP) combined with insulin lispro protamine suspension (NPL), an intermediate-acting formulation of LP, or to regular human insulin (HR) combined with human NPH insulin (NPH) in this open-label, parallel study. Insulin doses were similar at endpoint. Blood glucose (BG) measurements (before and two hours after meals, bedtime, 3 a.m.)(mmol/l) were lower with LP/NPL two hours after breakfast (8.84 +/- 0.32 vs 10.29 +/- 0.41, p < 0.001), before lunch (6.21 +/- 0.20 vs 7.10 +/- 0.31, p = 0.016), two hours after the evening meal (10.18 +/- 0.36 vs 7.86 +/- 0.28, p < 0.0.001), and at bedtime (7.85 +/- 0.28 vs 9.43 +/- 0.40, p = 001). HbA1c was lower for LP/NPL at endpoint (7.54 +/- 0.11% vs 7.92 +/- 0.10%, p = 0.019). There was no difference in hypoglycaemia or insulin antibody levels. LP/NPL resulted in better glycaemic control than HR/NPH without increasing the risk of hypoglycaemia.