In the Early Treatment Diabetic Retinopathy Study, a randomized clinical trial sponsored by the National Eye Institute, one eye of each patient was assigned to early photocoagulation and the other to deferral of photocoagulation (i.e., careful follow-up and initiation of photocoagulation only if high-risk proliferative retinopathy developed). This design allowed observation of the natural course of diabetic retinopathy in the initially untreated eye. Gradings of baseline stereoscopic fundus photographs of eyes with nonproliferative retinopathy assigned to deferral of photocoagulation were used to examine the power of various abnormalities and combinations of abnormalities to predict progression to proliferative retinopathy in photographs taken at the 1-, 3-, and 5-year follow-up visits. Severity of intraretinal microvascular abnormalities, hemorrhages and/or microaneurysms, and venous beading were found to be the most important factors in predicting progression. On the basis of these analyses and other considerations, a retinopathy severity scale was developed. This scale, which divides diabetic retinopathy into 13 levels ranging from absence of retinopathy to severe vitreous hemorrhage, can be used to describe overall retinopathy severity and change in severity over time.