To establish whether lispro may be a suitable short-acting insulin preparation for meals in intensive treatment of Type 1 diabetes mellitus (DM) in patients already in chronic good glycaemic control with conventional insulins, 69 patients on intensive therapy (4 daily s.c. insulin injections, soluble at each meal, NPH at bedtime, HbA1c <7.5%) were studied with an open, cross-over design for two periods of 3 months each (lispro or soluble). The % HbA1c and frequency of hypoglycaemia were assessed under four different conditions (Groups I-IV). Lispro was always injected at mealtime, soluble 10-40 min prior to meals (with the exception of Group IV). Bedtime NPH was continued with both treatments. When lispro replaced soluble with no increase in number of daily NPH injections (Group I, n = 15), HbA1c was no different (p = NS), but frequency of hypoglycaemia was greater (p < 0.05). When NPH was given 3-4 times daily, lispro (Group II, n = 18), but not soluble (Group III, n = 12) decreased HbA1c by 0.35 +/- 0.25% with no increase in hypoglycaemia. When soluble was injected at mealtimes, HbA1c increased by 0.18 +/- 0.15% and hypoglycaemia was more frequent than when soluble was injected 10-40 min prior to meals (Group IV, n = 24) (p < 0.05). It is concluded that in intensive management of Type 1 DM, lispro is superior to soluble in terms of reduction of % HbA1c and frequency of hypoglycaemia, especially for those patients who do not use a time interval between insulin injection and meal. However, these goals cannot be achieved without optimization of basal insulin.