Diagnostic Accuracy of Conventional Sector-based Optic Nerve Head and Retinal Nerve Fibre Layer Thickness Analysis Across Different Levels of Glaucoma Severity

Abstract

 

Purpose

 

There is little evidence that conventional sector-based optic nerve head (ONH) and retinal nerve fibre layer thickness (RNFLT) analysis is optimized for glaucoma detection. Small areas of localized damage could potentially elude detection when sector means are taken. We investigated the merit of sector-based analysis relative to a method with a variable minimum unit of analysis across the entire ONH and RNFLT profile.

 

Methods

 

Data from one eye of 151 glaucoma patients and 33 healthy controls, tested with spectral-domain OCT were analyzed. Bruch’s membrane opening-minimum rim width (BMO-MRW) and peripapillary RNFLT were obtained at each degree and also averaged for each sector. A classification of glaucoma was made with (1) sectoral analysis: when the sectoral parameter value was below the corresponding 1% sectoral normative limit, and (2) total analysis: when a given number of individual measurements was below the corresponding 1% normative limit. Analyses were performed on all subjects, and with a median split based on visual field severity (Mean Deviation, MD) to examine the influence of disease severity on diagnostic accuracy.

 

Results

 

The median MD (interquartile range) of patients and controls was -3.6 (-7.7 - -1.7) and -0.3 (-1.1 - 0.8) dB, respectively. The width and depth of locations outside the 1% limits are shown in Figure 1. With BMO-MRW, sectoral analysis yielded 85% (all patients), 77% (MD ≥ -3.6 dB) and 94% (MD < -3.6 dB) sensitivity; while at the same specificity, total analysis yielded 79%, 64% and 94% sensitivity, respectively (Fig. 2). The sensitivity of sectoral analysis was significantly higher than total analysis (p < 0.01) except in the analysis of patients with MD < -3.6 dB. With RNFLT, sectoral analysis yielded 85% (all patients), 73% (MD ≥ -3.6 dB) and 97% (MD < -3.6 dB) sensitivity; while at the same specificity, total analysis yielded 74%, 56% and 91% sensitivity, respectively (Fig. 2). Regardless of severity, sectoral analysis of RNFLT had significantly higher sensitivity compared to total analysis (p < 0.01).

 

Conclusions

 

The higher diagnostic accuracy of sectoral analysis compared to total analysis indicates that irrespective of the level of visual field damage, BMO-MRW and RNFLT defects are wide enough to be detected by conventional sectoral analysis.