Purpose: The aim of this study was to evaluate and compare 4 different discriminant analysis formulas and the new Glaucoma Probability Score (GPS) for the detection of morphometric optic nerve head changes in chronic open-angle glaucoma.
Methods: This is a prospectively planned cross-sectional study. Two hundred and fourteen consecutive eyes were recruited into this study. For each patient, the eyes were evaluated by a slit lamp examination, and the visual fields were assessed by a Humphrey Field Analyzer 750 (HFA, Humphrey Inc, San Leandro, CA), using the standard full threshold 24-2 (Swedish Interactive Threshold Algorithm) program. The optic nerve heads were morphometrically evaluated using the Heidelberg Retina Tomograph 3 (HRT 3, Heidelberg Engineering, Heidelberg, Germany; software version 3.0). From the HRT data, 4 discriminant analysis formulas and the GPS were considered. All data were analyzed by Student t test and Pearson r coefficient. A linear regression model was also used to determine the independent contribution of variables included in the model. Sensitivity, specificity, diagnostic precision, and receiver operating characteristic curve areas were calculated for all the 5 methods examined. kappa statistic was used to study the agreement among, and between, the 5 different methods.
Results: One hundred and nineteen normal eyes and 95 eyes with primary open-angle glaucoma were included in the study. No significant difference was found between the 2 study subgroups in both age and refractive error. Significant (P<0.001) correlations were found between visual field indices and the HRT parameters. Sensitivity, specificity, and diagnostic precision of the 4 formulas ranged between 50% and 99.16%. Bathija et al's formula had the highest diagnostic precision, followed by Mikelberg's formula. Using kappa statistics, kappa ranged from 0.177 to 0.528 when comparing each single discriminant formula with the GPS.
Conclusions: The GPS showed similar sensitivity and specificity to the Mikelberg and Bathija formulas; this method is a promising one for differentiating between healthy and glaucomatous eyes, requiring no subjective user input.