Diagnostic accuracy of Heidelberg Retina Tomograph III classifications in a Turkish primary open-angle glaucoma population

Abstract

This study aimed to evaluate the diagnostic accuracy of Moorfields regression analysis (MRA) and the glaucoma probability score (GPS) in primary open-angle glaucoma (POAG) and to measure the level of agreement between the two algorithms in classifying eyes as normal or abnormal in a Turkish population. We prospectively selected 184 healthy subjects and 158 subjects with POAG, who underwent an ophthalmological examination, visual field analysis and imaging with the Heidelberg Retina Tomograph II, using HRT III software, Version 3.0. The diagnostic accuracies of the two classifications were measured when the borderline was taken as either normal (highest specificity criteria) or abnormal (highest sensitivity criteria). The agreement between them was calculated using the unweighted kappa (kappa) coefficient. Optic nerve head topographic parameters showed statistically significant differences between the control and POAG groups (p < 0.001). The parameters with the highest area under the receiver operating characteristic curves were global GPS (0.86), cup : disc area (0.85), rim : disc area (0.85) and vertical cup : disc (0.85). According to the highest specificity criteria, MRA had a sensitivity of 67.7% and a specificity of 95.1%, whereas the GPS had a sensitivity of 70.9% and a specificity of 88.0%. According to the highest sensitivity criteria, MRA had a sensitivity of 81.0% and a specificity of 75.0%, whereas the GPS had a sensitivity of 89.2% and a specificity of 57.6%. A moderate agreement of 68% (233 eyes) with a kappa coefficient of 0.51 was found between MRA and the GPS. The GPS automated classification showed similar sensitivity to MRA, but considerably lower specificity, when applied in a Turkish population.