AIMS To evaluate a multiple daily injections (MDI) regimen combining lispro with multiple NPH insulin injections in order to replace basal insulin optimally. METHODS Twenty-five C-peptide negative Type 1 patients already trained to MDI were randomized to lispro (lispro + NPH 5 min before breakfast and lunch, lispro before dinner, NPH at bedtime) or soluble insulin (20-30 min before each meal and NPH at bed-time) for 3 months before crossing over to the other regimen for another 3 months. The mean initial HbA1c level was 8.32+/-1.5%. RESULTS The variability of capillary blood glucose values was significantly lower with lispro (MAGE 0.75+/-0.36 g/l vs. 0.99+/-0.50, P<0.01; MODD 0.64+/-0.26 g/l vs. 0.80+/-0.40, P<0.05). There was a nonsignificant reduction in HbA1c with lispro: -0.40+/-0.86 vs. -0.08+/-0.71. Mean daily blood glucose levels were significantly lower with lispro (1.53+/-0.48 g/l vs. 1.82+/-0.57 g/l, P<0.05). The frequency of all hypoglycaemic episodes was the same with both regimens but the number of severe hypoglycaemic events was reduced with lispro, P = 0.048. At the end of the study, 75% of the patients chose the lispro associated with multiple NPH regimen for their own treatment. The total insulin doses was the same with both regimens but the proportion of NPH was higher with lispro (53% vs. 34%). CONCLUSIONS An MDI regimen using lispro combined with multiple NPH compared to a standard MDI regimen using soluble insulin reduced day-to-day blood glucose fluctuations, was generally preferred by patients and was associated with a reduced incidence of severe hypoglycaemia with no loss of overall control.