Inter-device and inter-scan comparability of drusen volume assessments in age-related macular degeneration: A MACUSTAR report

Abstract

Purpose : Optical coherence tomography (OCT)-based drusen volume quantification is a relevant biomarker in age-related macular degeneration (AMD). An FDA-approved automated software for quantifying drusen volume using Cirrus (Zeiss Meditec, Dublin, CA) is available, however the comparability with similar approaches on other devices is largely unknown. Hence, we compared drusen volume measured by a newly developed, automated software running on Spectralis OCT data (Heidelberg Engineering, Heidelberg, Germany) against the above-mentioned software on Cirrus.

Methods : We included 306 eyes from 159 individuals with bilateral intermediate (iAMD, n=133) or early (eAMD, n=26) AMD, both reading center confirmed using multi-modal imaging. The Spectralis imaging protocol included a 20°×20° (25 B-scans) and a 30°×25° macular volume scan (241 B-scans), while the Cirrus protocol included a 200×200 macular cube. We assessed inter-device and inter-scan differences in 3- and 5mm diameter fovea-centered grid with intra-class correlation coefficients (ICCs) and root mean squared error (RMSE). Furthermore, we calculated the agreement based on exceeding a drusen volume cut-off indicating higher AMD progression risk (previously shown at >0.03mm³ for Cirrus, set at >0.09mm³ for Spectralis by adding the mean difference between the two devices to 0.03).

Results : Mean age of participants was 71.3±7.4 years. In the 5-mm diameter grid, no drusen were detected in 2 eyes (0.6%) in the Spectralis 241 B-scans, in 13 eyes (4%) in the 25 B-scans and 75 eyes (23.6%) in Cirrus scans. Agreement on drusen volume measurements was high between the two Spectralis scans (ICC 0.993 [0.991-0.994]). Drusen volume measurements between the denser Spectralis scan and Cirrus showed good reliability (ICC 0.803 [0.76-0.84]); the RMSE was 0.007 in eAMD and 0.106 in iAMD. Considering the cut-off mentioned above, agreement (kappa) between the two Spectralis scans was 97.4 (0.94); while with Cirrus it was 89.3 (0.77) for 241 B-scans and 89.4 (0.77) for 25 B-scans, respectively.

Conclusions : Drusen volume measures across the two softwares are not interchangeable. Comparability improves using a binary cut-off. Less dense scan patterns decrease sensitivity of drusen detection in Spectralis OCT scans.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

 

Figure 1. Bland-Altman plot of Spectralis 241 B-scans and Cirrus drusen volume measurements.