Detection of glaucoma-like optic discs in a diabetes teleretinal program

We would like to provide follow-up commentary to our article titled, “Detection of glaucoma-like optic discs in a diabetes teleretinal program,” that appeared in the December 2007 issue of Optometry. Our stated purpose was to assess the sensitivity and specificity of a diabetes teleretinal program to identify the glaucoma-suspicious optic disc. We established 2 groups of patients drawn from a diabetic teleretinal imaging cohort: (1) subjects with diabetes mellitus regarded by teleretinal readers as having glaucoma-like discs (N=175 subjects) and (2) subjects with diabetes in whom the imagers did not note any concerns regarding the optic disc appearance (N=175 subjects). To calculate sensitivity (the percentage of people whose test results are abnormal [in this instance a teleretinal assessment] who truly have the abnormality) and specificity (the percentage of people whose test results are normal by the same test who truly do not have the abnormality), one needs to determine the number of true-positives, false-positives, true-negatives and false-negatives there are in the study. We accomplished this by applying a definition of the glaucoma-like disc that is clearly described in the Methods section of the report to all 350 subjects. After data analysis there were 103 true-positives, 72 false-positives, 168 true-negatives and 7 false-negatives. Although these data remain accurate, in the article we reported the percentage of true-positives among those felt to have a glaucoma-like disc in group 1 (103 of 175; 59%) and incorrectly labeled this value as sensitivity. We also reported the percentage of true-negatives among those felt not to have a have glaucoma-like disc in group 2 (168 of 175; 96%) and incorrectly labeled this value as specificity. The sensitivity is actually 94% (not 59% as reported in the article), and the specificity is 70% (not 96% as reported in the article). This makes sense because although the teleretinal readers tended to overcall the glaucoma-like disc, they missed very few cases of the glaucoma-like disc. This actually is good news in that teleretinal programs designed to detect diabetic retinopathy actually show even more promise than we originally thought in detecting the glaucoma-like disc. We still feel, however, that improvements in teleretinal programs can be made to achieve the ultimate goal: detection of manifest glaucomatous damage in the asymptomatic patient.