We developed a program for a pocket computer (Sharp PC-1500) enabling conversion to and optimizing of meal-related insulin injection regimens. The dosage of regular insulin injected at meal times by a penshaped device (Novo Pen) was calculated in dependence of the carbohydrate content of the ingested meal, daytime, difference of actual and target blood glucose, 4 daily glucose levels of the last 2 or 4 days and physical activity. Long-acting insulin was proposed in dependence of morning blood levels of the preceding days. The response to elevated levels was additionally determined by night values. All quantifications were based on equivalent factors reflecting relations between insulin need and blood glucose changes and carbohydrate intake, respectively. This permits a so-called "quantified adaptation", where all these factors are individually defined and steadily actualized by the computer according to the data processed during adaptation. In a controlled cross-over study 12 type 1 diabetics were treated with conventional therapy (CT = 2-3 injections daily) and computer-assisted meal-related insulin therapy (CAMIT = 3-4 injections daily), each for a 6 weeks period. Although the number of meals was reduced from 6-7 to 3-5 and carbohydrate intake was allowed to vary from day to day, parameters of metabolic control were significantly improved by CAMIT compared to CT: mean blood glucose decreased from 9.10 +/- 2.96 to 6.22 +/- 0.65 mmol/l with CAMIT and only from 8.86 +/- 1.83 to 6.91 +/- 0.90 mmol/l with CT (p less than 0.05) and HbA1 from 10.2 +/- 1.5 to 8.6 +/- 0.8% with CAMIT and only from 9.8 +/- 1.3 to 9.1 +/- 1.0% with CT.(ABSTRACT TRUNCATED AT 250 WORDS)