We studied the incidence of blindness and visual impairment in patients who were enrolled in a photographic control- and screening program for diabetic retinopathy. The study cohort consisted of 2133 patients examined between January 1990 and December 1992 and followed until October 1st 1995. The occurrence of blindness (visual acuity < or = 0.1) and moderate visual impairment (visual acuity 0.2-0.4) was assessed. The Wisconsin scale was used to grade retinopathy. The mean HbA1c value for the last 8 years was used to represent long-term glycaemic control. Average follow-up time was 2.9 years. Seven patients were blind and 24 had visual impairment caused by retinopathy at the entry of the study. Six patients went blind due to retinopathy during the study period, corresponding to an incidence of 1.0 per 1000 person-years (95% confidence interval 0.4-2.1), and 28 became visually impaired, corresponding to an incidence of 4.6 per 1000 person-years (95% confidence interval 3.0-6.6). Multivariate analysis showed a statistically significant association between blindness/visual impairment and old age, long duration of diabetes, and poor glycaemic control. HbA1c values in the highest quartile, i.e. > or = 8.5%, were associated with a 65% increase in risk of blindness/visual impairment (95% confidence interval 14-130%). Retinopathy was the major cause of blindness and visual impairment in patients with diabetes. The study revealed a low incidence of blindness, which is in line with recent reports. Control of hyperglycaemia may be of value for the prevention of visual loss.