OBJECTIVE--To evaluate the influence of antihypertensive treatment and metabolic characteristics on the development of diabetes mellitus in middle aged men. DESIGN--Prospective study over an average of nine years. SETTING--Community based health survey of middle aged men carried out at the University of Uppsala. SUBJECTS--Seventy three hypertensive men aged 49-54 and 65 normotensive controls matched for body mass index, glucose disappearance rate (k value) at an intravenous glucose tolerance test, and serum triglyceride and cholesterol concentrations. INTERVENTIONS--Hypertensive group was treated with beta blockers, thiazides, hydralazine, or combinations of these drugs. Treatment was not randomised. MEASUREMENTS and MAIN RESULTS--Intravenous glucose tolerance, fasting blood glucose and serum lipid and insulin concentrations, body weight and height, three skinfold measurements, and blood pressure were recorded both during an initial health screening survey in 1970-3 and at a follow up survey in 1980-3. In the period between the two surveys 12 hypertensive men and two controls developed diabetes. Review of values obtained at the initial survey showed that the hypertensive men who developed diabetes or impaired glucose tolerance could be distinguished from those hypertensive men who did not by virtue of a higher fasting serum insulin concentration (26.1 v 15.2 mU/l (confidence interval of difference -15.2 to -6.2)), a lower peak serum insulin concentration (78.9 v 94.3 mU/l (confidence interval of difference -1.1 to 41.1)), and a lower k value (1.29 v 1.68 (confidence interval of difference -0.02 to 0.68)). The insulin index (peak insulin concentration divided by fasting insulin concentration), however, decreased significantly in the hypertensive men over time irrespective of whether they developed diabetes but did not change in the controls. Furthermore, the serum triglyceride concentration increased in the treated group and decreased in the controls. CONCLUSION--A severalfold difference in the incidence of diabetes between treated hypertensive and non-treated, normotensive men may be a consequence of the treatment, which may be particularly deleterious in men predisposed to diabetes.