The U.K. National Screening Committee recommended digital fundus photography as the screening method of choice for diabetic retinopathy (DR). However, concerns have been expressed about replacing ophthalmoscopy with slit-lamp biomicroscopy by digital photography. These concerns included the possibility of missing peripheral and stereoscopic visible retinal lesions; increased chance of a technical failure compared with ophthalmoscopy, resulting in more reexaminations; and higher cost (1). New data from our study of 453 patients with diabetes, aged 31–86 years, highlight the need for a careful consideration of the retinal area photographed and the minimal resolution of images to detect DR.
We compared nonstereoscopic two-field 45° digital fundus photography after pharmacological mydriasis with indirect ophthalmoscopy and slit-lamp biomicroscopy performed by an ophthalmologically trained physician for the detection of any DR. The photographs were made with a Canon CR5 nonmydriatic camera, interfaced to a 3-CCD color video camera (resolution 785 × 576 pixels). They were digitized, compressed (10:1 JPEG), and independently analyzed by the ophthalmologically trained physician and an ophthalmologist with a 17-inch monitor using Adobe Photoshop. In cases of disagreement, the judgement of another ophthalmologist was decisive. DR was detected by ophthalmoscopy in 16% of 442 right eyes that had both ophthalmoscopy and any gradable photography, as well as by photography in 8%. The agreement was moderate (κ statistic 0.54 ± 0.06) due to a sensitivity for detection of DR by photography of 44% compared with ophthalmoscopy. One reason for this low sensitivity could be that the resolution of the digital photographs was too low to detect small retinal lesions; 68% of the eyes with retinopathy missed by digital photography had only microaneurysms, small dot hemorrhages, or hard exudates in the area of both photographic fields. This is expected to become less of a problem because digital resolution improves rapidly in new cameras. However, the other reason for the low sensitivity of fundus photography was that retinal lesions were located outside the two photographic fields in 17% of the eyes with DR discovered by ophthalmoscopy.
This study focused on comparison of detecting any DR, since, except for two eyes treated with photocoagulation, only eyes with minimal and moderate nonproliferative retinopathy were detected. The U.K. Prospective Diabetes Study showed that microaneurysms alone are predictive for worsening retinopathy after 3–12 years (2). Therefore, it may be important to also detect small lesions, since these may be the first signs that blood glucose or blood pressure control needs to be reevaluated or possibly intensified. Furthermore, detection of any retinopathy instead of no retinopathy might indicate other reexamination intervals, while less sensitive detection methods could give patients false reassurance that their microvascular status is in good condition.
References
Address correspondence to Hendrik A. van Leiden, MD, VU University Medical Center, Department of Ophthalmology, P.O. BOX 7057, 1007 MB Amsterdam, The Netherlands. E-mail: [email protected].
