A combination of OCT metrics can detect glaucoma progression sooner than a VF benchmark criterion

Abstract

Purpose : To evaluate the performance of optical coherence tomography (OCT) and visual field (VF) summary metrics in detecting progression of early glaucomatous damage.

Methods : OCT scans and 24-2 and 10-2 VFs were obtained from 107 eyes [37 healthy controls (HC), 70 patients with early glaucoma, at least 4 tests (mean: 10.3±1.5 tests) and at least 2 years between baseline (first) and the last follow-up test (range: 2.0–6.8yrs). Based on confirmation of structural and functional evidence of progression, two graders independently judged 21 of the 70 patient eyes as ‘Progressors’ (P). In addition, 42 eyes were judged ‘stable’ (S), based upon a lack of evidence of change on OCT and VF. Metrics of the circumpapillary retinal nerve fiber layer (cpRNFL) and the macular ganglion cell layer (GCL) thickness were extracted from the OCT scans. From the 24-2 and 10-2 VFs, we recorded the total deviation values from each test location and the mean deviation (MD) global index for both patterns. A benchmark criterion (BC) of progression of 7dB loss in at least 5 locations was used as a ‘benchmark’. Linear regression over time was performed for all metrics. A negative slope less than a criterion value based on the HC eyes and a 95% specificity was deemed progressing. A Kaplan-Meier method identified the time to detect progression.

Results : The 24-2 and 10-2 BC labelled 6 S eyes as ‘progressing’, for a specificity of 86% (Table 1). The sensitivity was remarkably low, 38% (24-2) and 52% (10-2). Only one metric, the 24-2 MD Superior, had similar specificity (88%) and sensitivity (52%). All other OCT and VF metrics had worse specificity. Notably, all OCT metrics had the same or better sensitivity than the BC, ranging from 52% to 86%. Composite metrics of Structure-Structure (S-S) and Structure-Function (S-F) showed higher specificity and sensitivity than the BC. The best performing metric, an S-S metric (Table, red), which evaluates cpRNFL and GCL topographic agreement, had a specificity of 93% and a sensitivity of 86%, detecting nearly half of P eyes within 2 years. (Fig.1, green)

Conclusions : A strict benchmark criterion based upon pointwise VF loss was not as highly specific as anticipated. OCT and VF metrics, in isolation, had similar or worse performance. A S-S metric achieved the highest specificity and sensitivity and detected progression in half the eyes within 2 years.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.