Abstract
Purpose :
Over 100 optical coherence tomography (OCT) prognostic biomarkers for late age-related macular degeneration (AMD) have been described, but validation of an empirically-derived, OCT-based model for AMD is lacking. This project evaluated the prognostic discrimination of an OCT-based model.
Methods :
Seventy-eight eyes from 78 individuals (13 converters and 65 non-converters to late AMD) were propensity-score matched by age and sex from consecutive patients with AMD and longitudinal follow-up from the Centre for Eye Health, Sydney, Australia. Fifteen biomarkers (empirically-derived from recent meta-analysis and the simplified AREDS model) were identified by majority grading from three independent graders, including:
*OCT-derived reticular pseudodrusen, hypo-reflective drusen cores, intra-retinal hyper-reflective foci, shallow irregular retinal pigment epithelium elevation (SIRE), nascent geographic atrophy (nGA), external limiting membrane/ellipsoid zone/interdigitation zone/retinal pigment epithelium (RPE) reflective abnormality, and
*Colour fundus photography (CFP)-derived large drusen, drusenoid pigment epithelium detachment (DPED), pigmentary abnormality, fellow eye large drusen, fellow eye pigmentary abnormality, and bilateral intermediate drusen.
Prognostic discrimination (area under the receiver operating characteristic curve [AUC]) was calculated for: an optimised model (using logistic regression with independent, backward and forward selection), an all-biomarkers model, and the simplified AREDS [3yr] model.
Results :
Participants were 71.86 years old, 56% female, and follow-up time was 2.74 years. The optimised model (using either selection method) included only four OCT biomarkers (DPED, RPE abnormality, SIRE, nGA) demonstrating excellent prognostic discrimination (AUC, 0.9 [0.82, 0.99], P < 0.0001). This was similar to the all-biomarkers model (0.94 [0.88, 1], P < 0.0001) and greater than the simplified AREDS [3yr] model (0.78 [0.66, 0.9], P < 0.01; Figure).
Conclusions :
Assessing four OCT biomarkers – DPED, RPE abnormality, SIRE, and nGA – facilitates excellent prognostic discrimination of late AMD. Further validation in larger cohorts across longer periods is needed.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.