Purchase this article with an account.
Antonio Mattia Modarelli, Giovanni Montesano, Paolo Fogagnolo, Francesco Oddone, Paolo Lanzetta, Andrea Perdicchi, Chris A Johnson, David F Garway-Heath, David P Crabb, Luca Mario Rossetti; Test-retest variability in anatomically defined visual field clusters with fundus perimetry. Invest. Ophthalmol. Vis. Sci. 2019;60(9):2479. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To investigate the effect of fundus tracking on test-retest variability in visual field clusters aimed at capturing perimetric damage in glaucoma.
44 glaucoma patients and 54 healthy subjects were tested four times with a 24-2 grid, twice using the Compass Fundus Perimeter (CMP, CenterVue, ZEST Strategy) and twice with the Humphrey Field Analyzer (HFA, Zeiss Meditec, SITA-Standard Strategy), in random order. All visual field locations were divided into 6 clusters according to Garway-Heath et al., following the path of the nerve fibre bundles. Mean sensitivity per cluster was calculated for each test. Bland-Altman plots and 95% limits or repeatability (LoR) were calculated for each cluster. Percentage change in variability of CMP compared to HFA were calculated as (CMP-LoR - HFA-LoR)/HFA-LoR, so that negative values indicated lower variability with CMP and positive values lower variability with HFA. A paired bootstrap procedure was used to assess the variability of the percentage change and its confidence intervals (CIs).
LoR in CMP were generally reduced except for cluster 3 (Figure 1) in healthy subjects. For glaucoma patients, they were reduced in all clusters except for cluster 2. However, bootstrap distributions and CIs showed that the estimates of such reductions had a large variability, especially in glaucoma subjects (Figure 2). A significant reduction (p < 0.05) could only be obtained for Clusters 1, 4 and 6 in healthy subjects and Cluster 5 for glaucoma patients.
Fundus tracking is a promising technology to achieve reduction in cluster test-retest variability, with possible important implications for structure-function analyses. However, CIs on these estimates of such a reduction are still too large to be conclusive. Learning effect and small sample size might have contributed to such variability.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
Figure 1. Bland Altman plots for each cluster with HFA (red) and CMP (blue). The schematic highlights the corresponding sector. Shaded regions indicate the 95% LoR.
Figure 2. Bootstrap distributions of percentage improvement estimates in the LoR of CMP over HFA for each cluster
This PDF is available to Subscribers Only