Response to Gómez-Ulla et al. Free

We strongly disagree with the assertion of Gómez-Ulla et al. (1) that European protocols did not define any standards by which diabetic retinopathy screening programs should be assessed. Recently, Bachmann and Nelson (2) pooled the data from many available screening studies, dated up to 1996, and determined that retinal photography was the screening test of choice. In fact, the U.K. National Screening Committee was considering this topic as a possible national priority. In addition, a cross-sectional epidemiologic study on 258 randomly selected diabetic patients in Spain was published by Hernáez-Ortega et al. (3) in 1998. The authors compared the efficiency of a 45° nonmydriatic retinal camera with a Polaroid instant film versus the standard method, and they concluded that fundus photography would be advisable for its use in medical settings that usually control for potential ophthalmologic complications resulting from diabetes (3).

Gómez-Ulla et al. (1) concluded in their article that sending digital retinal images via the Internet was a suitable method for detecting and grading diabetic retinopathy in a nonselected diabetic population. This conclusion was reached after comparison between grading by direct examination and grading by inspecting digital images. However, we would like to point out some methodological issues in their article that could invalidate the generalization of their results. What do we know about their patients? The authors stated that 70 consecutive diabetic patients were recruited either from an endocrinology and/or an ophthalmology unit, but they did not describe their study population (i.e., age, sex, diabetes type, diabetes duration, etc.). Otherwise, it seems that this diagnostic method will work equally well despite the population. It is well-known that older populations will have a lower degree of collaboration, higher degree of media opacities, and smaller pupil size. We also suggest that the authors pull together those patients excluded from their analysis (i.e., seven patients with opaque cataract and seven with poor image quality) to provide the percentage of eyes that did not allow detection of retinopathy with their proposed method (i.e., 10%). Hernáez-Ortega et al. (3) reported 6.6% of ungradable fundus photography in a much larger study population of diabetic patients in the North of Spain.

Kappa (κ), the measure of agreement corrected for chance, may be interpreted as the proportion of the best possible improvement in agreement beyond chance that was actually obtained by the observers. The authors calculated the intraclass correlation coefficient (ICC) to determine agreement on the stage of diabetic retinopathy. The authors should have noted that the ICC can be interpreted as a special form of weighted κ. In fact, when the variable is ordinal, calculating a weighted κ is preferable (4).

Therefore, we recommend that the authors provide the missing data on their study population to show how much it influenced their results. We also suggest the calculation of a weighted κ to determine agreement on degree of diabetic retinopathy.