Among a cohort of 348 women aged 50 on entering a 12-year prospective study, the incidence of diabetes was increased (17.1%) during follow-up in those who initially had fasting glucose concentration above the upper quintile, a fasting serum insulin concentration above the upper quintile (14.9%), a disappearance rate of glucose below the lowest quintile in an IV glucose tolerance test (12.7%), or early insulin response below the lowest quintile (17.1%). The incidence in all women was 4.9%. By multivariate analysis, the highest risk was for high fasting serum insulin concentration. Obesity and treatment with antihypertensive drugs further increased the risk. An initial low early insulin response was not however a prerequisite for the development of manifest diabetes. Determination of fasting insulin concentration, especially in overweight hypertensive subjects, is of value in order to find out which subjects are at high risk of developing diabetes.