Although regular screening for diabetic retinopathy with ophthalmoscopy or retinal photography is widely recommended in the United States and Europe, few reports of its use in developing countries are available. We compared the performance of screening by retinal photography with that of indirect ophthalmoscopy by using data from a population-based survey of diabetes and its complications in Egypt. During that project, 427 persons with diabetes underwent an eye examination and fundus photography with a non-mydriatic camera through a dilated pupil. Data from the examinations of the right eye of each patient are presented. Ninety-two (22%) of the 427 retinal photographs were ungradable; in 58 eyes (63%), this was due to media opacity (42 eyes with cataract, 3 with corneal opacity, and 13 with both). Agreement between retinal photography and indirect ophthalmoscopy was poor (kappa = 0.33; 95% CI = 0.27-0.39) and primarily due to the large number of eyes (n = 79) with ungradable photographs that could be graded by ophthalmoscopy. None of these eyes was judged by ophthalmoscopy to have sight-threatening retinopathy. Fifty-four photographs were diagnosed with greater retinopathy than found on ophthalmoscopy. Retinal photography with the nonmydriatic camera through a dilated pupil is a useful method to screen for diabetic retinopathy in most adults in Egypt. However, such screening strategies have limited use in older persons and in persons with corneal disease or cataract.