A total of 6695 diabetic men and women, aged 35 to 54 years, from 14 centres and representing 13 national groups, participated in a vascular disease prevalence survey. A random sample was drawn after stratification of each centre's diabetic base population by sex, duration of diabetes and age. A common agreed protocol, standardized examination procedures, and centralized laboratory methods were used in the investigation. Within the age range examined there was considerable variation between centres in a number of variables, including degree of obesity (measured as Body Mass Index (BMI)), proportion treated with insulin and proportion of cigarette smokers. The latter also showed considerable sex differences within centres. Subjects with age at onset below 25 years were notably few in Hong Kong, Tokyo and Oklahoma. There was also considerable variation in the apparent prevalence of both large- and small-vessel (macrovascular and microvascular) disease between centres. In pooled data, measures of large-vessel disease were significantly and independently associated with age, blood pressure and BMI in both sexes, and with diabetes duration and plasma cholesterol in men only. Within-centre analyses showed blood pressure to be the most consistently associated variable in both sexes. In pooled data, small-vessel disease of the eye was significantly and independently associated with diabetes duration, blood pressure, BMI and type of treatment in both sexes. In within-centre analyses, diabetes duration was the most consistently associated variable, followed by blood pressure. Proteinuria as an index of small-vessel disease of the kidney was, in pooled data, significantly and independently correlated with diabetes duration, blood pressure and plasma cholesterol in both sexes. In within-centre analyses, blood pressure was the most consistently associated variable, with diabetes duration and plasma cholesterol equal second - significant in 12 of the 28 centre/sex groups. Heterogeneity of large-vessel disease prevalence in diabetic subjects is confirmed by this study, and the possibility of heterogeneity in small-vessel disease prevalence and severity is suggested.