Abstract
Purpose :
Many novel OCT biomarkers have been proposed for diagnosis, risk stratification and management of AMD, but clinical translation capacity is limited by poor understanding of inter-grader reliability. This study compared the inter-grader agreement of 10 empirically-derived prognostic biomarkers for AMD to the 2 clinical gold standard.
Methods :
Single eyes from 78 individuals with intermediate AMD were independently graded by four optometrists with 2-10 years of experience in retinal imaging. OCT B-scans of each eye was graded for the absence (<90% certainty) or presence (≥90% certainty) of empirically-derived prognostic biomarkers: reticular pseudodrusen (RPD), hyporeflective drusen cores (hDC), hyperreflective foci (HRF), shallow irregular RPE elevation (SIRE), nascent geographic atrophy (nGA), and hyperreflective outer-retinal band reflective abnormalities of the external limiting membrane (ELM) / ellipsoid zone (EZ) / interdigitation zone (IZ) / retinal pigment epithelium (RPE). Colour fundus photography of each eye was graded for drusenoid pigment epithelial detachment (DPED) and two gold-standard biomarkers: large drusen, and pigmentary abnormalities. If presence of a biomarker was questionable, an auxiliary retinal imaging modality (OCT en face, near-infrared reflectance and/or fundus autofluorescence) was used for re-grading. Free-marginal kappa (κ) and percentage overall agreement were assessed.
Results :
Gold-standard AMD biomarkers, large drusen and pigmentary abnormalities, had 83% (κ = 0.67) and 68% (κ = 0.37) overall agreement, respectively. In comparison, the overall agreement for the ten empirically-derived prognostic biomarkers ranged from 46 – 90%. Only one OCT biomarker, nGA, outperformed both large drusen and pigmentary abnormalities (90% agreement; κ = 0.81). Aside from outer-retinal band reflective abnormality, all other OCT biomarkers were non-inferior to pigmentary abnormalities (Figure 1).
Conclusions :
Most OCT biomarkers are more reliably graded that the gold standard AMD biomarker, pigmentary abnormalities, but not large drusen. These results indicate utilising a multi-modal approach can improve inter-grader agreement relative to current clinical gold standard alone.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.