Five hundred and seven elderly hypertensive patients were followed for 1 year, 371 for 2 years and 270 for 3 years in a double-blind, randomized, controlled trial in which they received either placebo or 25-50 mg hydrochlorothiazide and 50-100 mg of triamterene daily. One third of the active treatment group also received 250 mg to 2 g methyldopa daily. After 1 year the active treatment group had an average increase in fasting blood sugar of 2.5 mg/dl compared with an average fall of 1.4 mg/dl in the placebo group (P = 0.01). The increase in blood sugar 1 hour and 2 hours after 50 g oral glucose tended to be greater in the actively treated group but these increases did not achieve statistical significance. The effects of diuretic treatment were established after one year and did not increase further over the next 2 years. Overall there was an increase in fasting blood sugar of 5 mg/dl in the active treatment group which occurred mainly in the first year. The hyperglycaemic effect of diuretics appeared to be partly or wholly related to potassium loss since, in both groups, impairment of glucose tolerance was most marked in those in whom serum potassium decreased. The measures of blood sugar were also positively related to systolic pressure before and after treatment.