Optic disk and nerve fiber layer imaging to detect glaucoma

Abstract

Purpose: To compare optic disk and retinal nerve fiber layer (RNFL) imaging methods to discriminate eyes with early glaucoma from normal eyes.

Design: Retrospective, cross-sectional study.

Methods: In a tertiary care academic glaucoma center, 92 eyes of 92 subjects (46 with early perimetric open-angle glaucoma and 46 controls) were studied. Diagnostic performance of optical coherence tomography (StratusOCT; Carl Zeiss Meditec, Dublin, California, USA), scanning laser polarimetry (GDx VCC; Laser Diagnostic Technologies, San Diego, California, USA), confocal laser ophthalmoscopy (Heidelberg Retinal Tomograph [HRT] III; Heidelberg Engineering GmbH, Heidelberg, Germany), and qualitative assessment of stereoscopic optic disk photographs were compared. Outcome measures were areas under receiver operator characteristic curves (AUCs) and sensitivities at fixed specificities. Classification and regression tree (CART) analysis was used to evaluate combinations of quantitative parameters.

Results: The average (+/- standard deviation) visual field mean deviation for glaucomatous eyes was -4.0 +/- 2.5 dB (decibels). Parameters with largest AUCs (+/- standard error) were: average RNFL thickness for StratusOCT (0.96 +/- 0.02), nerve fiber indicator for GDx VCC (0.92 +/- 0.03), Frederick S. Mikelberg (FSM) discriminant function for HRT III (0.91 +/- 0.03), and 0.97 +/- 0.02 for disk photograph evaluation. At 95% specificity, sensitivity of disk photograph evaluation (90%) was greater than GDx VCC (P = .05) and HRT III (P = .002) results, but not significantly different than those of StratusOCT (P > .05). The combination of StratusOCT average RNFL thickness and HRT III cup-to-disk area with CART produced a sensitivity of 91% and specificity of 96%.

Conclusions: StratusOCT, GDx VCC, and HRT III performed as well as, but not better than, qualitative evaluation of optic disk stereophotographs for detection of early perimetric glaucoma. The combination of StratusOCT average RNFL thickness and HRT III cup-to-disk area ratio provided a high diagnostic precision.

Figure 1

Comparison of the areas under receiver operator characteristic curves (±SE) of the best parameters from StratusOCT (average retinal nerve fiber layer thickness: 0.96 ± 0.02), GDx-VCC (Nerve Fiber Indicator: 0.92 ± 0.03), HRT III (FSM discriminant function: 0.91 ± 0.03), and the cumulative score of the three observers for disc photograph evaluation (0.97 ± 0.02).

Figure 2

Results of Classification and Regression Trees (CART) analysis for combination of the best parameters from optical coherence tomography (StratusOCT), scanning laser polarimeter (GDx-VCC), scanning laser ophthalmoscopy (HRT III). The three best parameters from each instrument were entered into the final CART analysis. With appropriate cutoff points for the StratusOCT’s average RNFL thickness and HRT III’s cup/disc area ratio, their combination provides a sensitivity of 91% (42 out of 46 glaucomatous eyes identified as such) and a specificity of 96% (44 out of 46 normal eyes classified as glaucomatous).