Retest variability of quantitative fundus autofluorescence and its determinants in age-related macular degeneration: A MACUSTAR study report

Abstract

Purpose : To determine the test-retest reliability of quantitative fundus autofluorescence (QAF) as well as associated patient factors (i.e. its determinants) in participants with and without age-related macular degeneration (AMD) in the MACUSTAR study, a European multicenter study.

Methods : A total of 132 eyes of 68 subjects with and without AMD (22 no macular disease; 2 early, 28 intermediate, 16 late AMD) from 6 clinical study sites underwent duplicate QAF imaging at baseline and two weeks later. QAF values were measured at the inner, middle and outer “Delori ring”. Test-retest reliability was assessed using intra-class correlation (ICC) analysis. QAF image quality was graded by two researchers using mean opinion scores (MOS, criteria graded: focus, illumination, symmetry, zoom, centering). The impact of age, AMD disease status, lens status and MOS on retest reliability was evaluated in a mixed-effect linear model. Furthermore, a random forest regressor was trained to evaluate the extent to which manual image quality grading can be replaced by an automated assessment.

Results : ICC (mean [95% confidence interval]) was 0.95 [0.93 – 0.96] in the intra-day and 0.79 [0.72 – 0.86] in the inter-day analysis for all eyes, with slightly lower values in AMD affected eyes. For QAF segments within the QAF8 ring, the ICC was 0.77 [0.70 – 0.84] and 0.74 [0.65 – 0.81 for peripheral segments]. Image quality was predicted with a mean absolute error of 0.27. Finally, from all evaluated reliability indices, MOS and inferred MOS proved most important for predicting retest reliability.

Conclusions : QAF is highly reliable when implemented in a standardized fashion at the posterior pole, over time and between study centers. Automated image quality assessment may be helpful to assist in reliable QAF analysis. Correcting for patient factors may further enhance retest reliability.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.