Purchase this article with an account.
Emmanouil (Manos) Tsamis, Jennifer Leah Grossman, Sol La Bruna, Gustavo De Moraes, Donald C Hood; Evaluation of standard and novel structural and functional methods for detecting progression in glaucoma. Invest. Ophthalmol. Vis. Sci. 2021;62(8):3358.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To assess standard and novel optical coherence tomography (OCT) and visual field (VF) methods for detecting glaucomatous progression
Disc circle and macular cube OCT scans and 24-2 and 10-2 VF data were obtained from 100 eyes [71 patients/suspects and 29 healthy controls (HC)], as part of a prospective, longitudinal study (P-group)(ClinicalTrials.gov Identifier: NCT02547740). All eyes had a minimum of 4 OCT and VF tests (mean 9.2 tests), with the last visit at least 1 year after the baseline visit. We evaluated standard OCT and VF metrics, combinations of OCT-OCT and OCT-VF, and a metric based on a region-of-interest (ROI) approach (Table 1, left-most column). For event-based, test-retest variability was estimated based on 176 eyes (146 patients/suspects, 30 HC). Repeated OCT and VF data collected within 4 months were analyzed with quantile regression to define cut-offs (the 95th percentile defined ‘statistical progression’). Those cut-offs were then applied to the first vs. last test of the P-group. For trend-based, based on least squares regression eyes were categorized as “progression” if the slope was significantly negative (one-tail p<0.05). Finally, 2 OCT experts (E-OCT) reviewed all OCT information and judged whether each eye had progressed on a scale of 0 (definitely did not progress) to 100% (definitely did progress). Eyes with scores ≥95% were labelled “progressors” (P), and those with scores ≤5% “not progressors” (NP). The E-OCT identified 13 P eyes, none of which was a HC
We used the 60 NP and 13 P eyes as proxies for specificity and sensitivity (see Tables -bold columns). For the Event Analysis (Table 1) only one metric, TI (green in Tables), had a percent agreement greater than 90% for both proxies, while for the Trend Analysis (Table 2) none of the metrics had percentages greater than 90% for both. In general, an event-based approach performed better than trend analyses (Tables 1&2)
All objective methods so far explored miss clear progression revealed in some eyes by a careful examination of circumpapillary b-scans, with or without the aid of OCT probability maps, as well as identify “progressors” that are clearly not progressing. We recommend that a qualitative evaluation of OCT images and thickness/probability maps be included at least as part of a post-hoc analysis.[1,2] 1. Hood et al. JoG 2020; 2. Eguia et al. TVST 2020
This is a 2021 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only