Conventional treatment of obese noninsulin dependent diabetes mellitus (NIDDM) patients is often unsatisfactory. In this study the efficacy of Modifast, a commercial very low calorie diet (VLCD), was evaluated in a population of obese poorly controlled NIDDM patients. The mechanisms of action of VLCD in these patients were also studied by comparing: (i) Plasma insulin and glucose profiles after a VLCD and an isocaloric mixed meal and (ii) plasma amino acid levels, both at baseline and after four weeks of VLCD treatment. A total of 14 obese NIDDM patients (M/F 7/7. median body mass index (BMI) 38.7 kg-2, interquartile range (IQ) 34.7-46.5 kg-2, waist circumference 116 cm, IQ 106-139 cm, insulin treated 7/14) with poor diabetic control (HbA1c 8.6%, IQ 7.8-10%) were studied. Patients were given a VLCD (425 kcal/day) for 12 weeks. At baseline, VLCD and isocaloric meal tests were performed on consecutive mornings. Fasting plasma amino acid levels were also determined at baseline and after 4 weeks of VLCD treatment. Weight, waist circumference, HbA1c, blood pressure, fasting plasma insulin, total cholesterol and triglyceride levels all fell significantly following VLCD treatment. Insulin therapy was able to be ceased in the seven insulin treated patients. Oral hypoglycaemic agent dosage fell from a median of eight (IQ 6-12) to two (IQ 0-8) tablets per day (P = 0.03) in patients initially on this form of therapy. Insulin secretion was higher after VLCD than isocaloric meal (P = 0.04). Fasting plasma alanine level fell from 512.0 (IQ 412.0-563.0) to 374.0 (IQ 342-472.0) mumol/l (P = 0.04) following VLCD treatment. In conclusion, the short term use of a VLCD is very effective in rapidly improving glycaemic control and promoting substantial weight loss in obese NIDDM patients. Moreover, a VLCD diet increases insulin secretion and reduces substrate for gluconeogenesis. Thus, VLCD treatment may improve glycaemic control by factors more than caloric restriction alone.