Abstract
Purpose :
To compare methods of combined structure-function information to common structure-only and function-only metrics for detecting progression in early glaucomatous eyes.
Methods :
151 eyes/individuals had repeat OCT scans and 24-2&10-2 visual fields (VF) within 6 months (mean: 3.2 visits, range 2-5). This formed a variability (V) group, with 49 healthy (H), and 102 glaucoma (G) or glaucoma suspect (GS) eyes. OCT and VF tests >1year after the baseline tests were obtained from 100 (74 G/GS, 26 H) of the 151 eyes to comprise the study group. Summary metrics of the thickness measures from the circumpapillary retinal nerve fiber layer (cRNFL) and a ±8° retinal ganglion cell (RGC) region around the fovea, as well as mean and pattern standard deviation (MD; PSD) metrics from the 24-2/10-2 VF, were obtained (Table). Topographical agreement of abnormal structure-abnormal function (aS-aF) locations between RNFL and RGC probability and 24-2&10-2 PD probability maps from each session was determined with an objective, automated method.[1] Quantile regression was applied to each metric to derive 95% limits and define “statistical progressors”.[2] Four glaucoma experts evaluated all OCT and 24-2&10-2 VF tests to form a progression reference standard (P-RS).
Results :
The P-RS identified 27 progressing G eyes. For the OCT alone, the RGC metrics in general, and the I-Mac in particular, detected the largest number of the 27 P-RS eyes (Hits in Table)(I-Mac: 19 hits). However, the VF metrics alone detected the fewest; although the PSD 10-2 identified almost double the number of P-RS eyes than PSD 24-2 (n= 9 vs 5). A combination of global cRNFL or RGC (G OR G Mac) metrics identified more P-RS eyes, (n=23), but also produced another 19 false positives (FPs). Fig shows a progressing eye that was not detected by G or the VFs. For structure-function, a combined summary metric [(G or G Mac) AND (PSD 24-2 or 10-2)] was more specific [4 (5.5%) FPs] but detected only a third of the P-RS eyes (n=9). An automated aS-aF agreement method slightly improved performance with 3 FPs and 13 P-RS eyes.
Conclusions :
Summary metrics perform poorly in identifying progressing eyes, due to a high number of FPs. A structure-function summary metric significantly reduces the number of FPs, while a topographic agreement method slightly improves sensitivity, while maintaining high specificity. [1]Hood, Tsamis et al. IOVS 2019 [2]Wall, Doyle et al. IOVS 2013
This is a 2020 ARVO Annual Meeting abstract.