Purpose: To investigate and compare the diagnostic accuracy of the Heidelberg Retinal Tomograph 3 (HRT3) diagnostic algorithms and establish whether they are affected by optic disc size and glaucoma severity.
Design: Multicenter cross-sectional evaluation of diagnostic tests.
Participants: Two hundred forty-two eyes from 139 normal subjects and 103 glaucomatous patients classified by the presence of a repeatable visual field (VF) defect.
Testing: Eyes were imaged by the HRT3. The diagnostic accuracies of Moorfields regression analysis (MRA) and the glaucoma probability score (GPS) was explored by sensitivity and specificity and area under the receiver operating characteristics curves (AUC). The analysis was performed globally and by optic disc size quartiles and by 3 VF severity groups.
Main outcome measures: Sensitivity, specificity, and AUC.
Results: The GPS showed a sensitivity (80% vs. 77%) similar to and a specificity (57% vs. 67%) lower than that of MRA Result. It showed a higher specificity in small discs than MRA Result (77% vs. 68%) but a low to very low specificity in medium to very large discs (medium, 61%; large, 50%; very large, 26%). Moorfields regression analysis Global showed the highest sensitivity and specificity (68% and 78%) in very large discs. R. Burke linear discriminant function (RB-LDF) and cup shape measure (CSM) showed the best and least variable AUC across optic nerve head sizes and glaucoma stages. The sensitivity of both MRA and the GPS decreased at the earlier glaucoma stages. The MRA and GPS agreement was moderate throughout the entire population and in small discs and early stage, whereas it was weaker among the other disc size and glaucoma stage subgroups.
Conclusions: HRT3 diagnostic algorithms' accuracy is moderate. The GPS is less specific and more influenced by disc size than MRA. Cup shape measure and the RB-LDF offer the best and less variable performances across different disc sizes and glaucoma stages.