After a 2 weeks hospital treatment with a strong dietary regime 100 patients not sufficiently controlled (blood glucose fasting 120-180, pp 180-250 mg/100 ml), were randomly allocated to a purely dietary therapy or to a diet-metformin-combination for a duration of 2 years. After 1 and 2 years the patients were rehospitalized for an accurate evaluation. Additionally, HbA1 was controlled every 3 months, in case of problems after shorter periods. If HbA1 exceeded 10% for more than 4 weeks the patients were hospitalized for 5 days in order to decide whether the deterioration of metabolic control was due to the patients' non-compliance to treatment or to treatment failure. During 2 years of observation 30 patients were withdrawn from the treatment for external reasons. 13 patients (6 diet, 7 metformin + diet) were non-compliant with diet, 3 further patients were not metformin-compliant. Therapeutic failures were confirmed in 4 diet patients, none of the metformin + diet patients failed to respond to treatment. After excluding all these patients there was a group of 29 diet and 25 metformin + diet patients with a similar development of criteria of metabolic control during 2 years. Lipid levels deteriorated in the dietary group more noticeably than during additional metformin treatment although without reaching a statistically significant level, though. The stimulated C-peptide was significantly reduced by metformin in comparison to diet only, which supports former findings of an insulin-lowering effect of the drug.