In 13 non-obese patients with Type 2 diabetes mellitus who failed to achieve adequate blood glucose control on dietary treatment (fasting blood glucose 13.4 +/- 2.7 (+/- SD) mmol l-1, glycosylated haemoglobin 13.0 +/- 1.7%), the effects of 6 months insulin or sulphonylurea therapy on blood glucose control and lipid metabolism were compared in a randomized crossover study. Three patients, who showed a clear improvement on insulin (median glycosylated haemoglobin fell from 14.7 to 8.6%), withdrew from the study prematurely because of subjective and objective signs of hyperglycaemia after crossover from insulin to sulphonylurea. Daily dose after 6 months was 2000 mg tolbutamide (n = 3), 18 +/- 1 mg glibenclamide (n = 7), or 34 +/- 3 U insulin. On insulin, fasting (8.0 +/- 1.9 mmol l-1) and postprandial blood glucose (10.4 +/- 2.7 mmol l-1), and glycosylated haemoglobin (9.5 +/- 1.1%) were lower than on sulphonylurea (11.0 +/- 3.4 mmol l-1, 14.4 +/- 4.8 mmol l-1 and 11.0 +/- 2.5%, respectively, p less than 0.05 in each case). Median increase in body weight was greater on insulin (4.2 vs 1.1 kg, p less than 0.05). Six patients experienced improved well-being on insulin compared with sulphonylurea. Median plasma non-esterified fatty acids decreased from 825 mumol l-1 to 476 mumol l-1 (sulphonylurea) and 642 mumol l-1 (insulin, both p less than 0.05).(ABSTRACT TRUNCATED AT 250 WORDS)