In a population-based epidemiological study, 991 Pima Indians with non-insulin-dependent (Type 2) diabetes mellitus (NIDDM) and 288 without diabetes aged > or =15 years were examined for retinopathy by fundus photography with a 45 degrees fundus camera after mydriasis. The photographs were graded using a modified Airlie-House classification scheme. The associations of several factors with retinopathy were studied by logistic regression. Non-proliferative retinopathy was present in 11.2 % (19/169) subjects at the time of diagnosis of diabetes and in 8.3% (4/48) in newly diagnosed subjects who had a documented non-diabetic oral glucose tolerance test within 4 years prior to diagnosis of diabetes. The prevalence of retinopathy in subjects with impaired glucose tolerance was 12% (8/68). Retinopathy at the time of diagnosis of diabetes was significantly associated with lower body mass index and higher systolic blood pressure but not glycaemia. Retinopathy was present in 375 (37.8 %) diabetic subjects and 14 (5.2 %) non-diabetic subjects. Among all subjects with diabetes (duration 0-37 years), stepwise multivariate analysis showed non-proliferative retinopathy to be associated with duration of diabetes, mean blood pressure, fasting plasma glucose, treatment with insulin and albuminuria. Proliferative retinopathy was seen in 34 (2.7%) of diabetic and none of the non-diabetic subjects, and was associated with 2 h post-load glucose concentrations, as well as albuminuria, insulin treatment, younger age, and diastolic blood pressure. These data confirm the need for fundus examination at the time of diagnosis of diabetes and during long-term follow-up. Albuminuria and blood pressure are potentially modifiable risk factors and the impact of treating these on incidence and progression of diabetic retinopathy need to be assessed.