Abstract
Purpose :
Glaucomatous visual field (VF) damage manifests primarily in a localized manner due to the patterns of retinal nerve fiber layer (RNFL) loss. Therefore, using abnormal OCT RNFL sectors to guide the VF location to follow over time may improve detection of glaucoma worsening compared to a location agnostic approach. The purpose of this study is to compare trend-based, event-based, and OCT-guided methods to identify visual field worsening among a large cohort of glaucoma patients.
Methods :
We conducted a retrospective longitudinal study of glaucoma and glaucoma suspect patients with seven or more reliable VFs over time, and a reliable OCT scan within one year of the initial VF to implement our OCT guided method. Our OCT guided method followed the VF sector with the lowest percentile RNFL thickness on OCT Garway-Heath zone along with the immediately adjacent sector within the same hemifield, with other sectors disregarded. We assessed for progression using the following methods across the first five VFs: mean deviation (MD) slope, pointwise linear regression (PLR), guided progression analysis (GPA), OCT-guided mean slope, OCT-guided PLR, and OCT-guided GPA. AUC was computed by comparing each methods progression decision for each eye over all but the last two VFs, to that methods progression decision over all VFs (reference standard).
Results :
44,193 VFs from 5,191 eyes were included in the study. Demographics and characteristics of included eyes are summarized in Table 1. MD, PLR, and GPA methods had an AUC of 0.79 (0.77,0.82), 0.77 (0.74,0.79), and 0.82 (0.79,0.84), respectively. The OCT-guided methods: mean slope, PLR, and GPA, had AUCs of 0.78 (0.74,0.81), 0.75 (0.72,0.79), and 0.89 (0.85,0.92), respectively.
Conclusions :
The performance of OCT-guided GPA at detecting VF worsening was better than non-OCT GPA (p < .01). However, OCT guided mean slope and OCT guided PLR performed no better than non-OCT guided methods. OCT-guided event-based methods, but not trend-based methods, may perform better at identifying visual field progression compared to algorithms performed without using OCT guidance.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.