Determinants of Intergrader Agreement for Key Retinal Photography and OCT Biomarkers in AMD

To quantify determinants of intergrader agreement for key retinal photography and optical coherence tomography (OCT) biomarkers in early to intermediate age-related macular degeneration (AMD) and to inform future consensus definitions and grading protocol.

Single eyes from 269 participants with early to intermediate AMD were consecutively sampled. Retinal photographs were evaluated for the presence of large drusen and pigmentary abnormalities. OCT B-scans were graded for reticular pseudodrusen (RPD), outer retinal band abnormality (external limiting membrane [ELM], ellipsoid zone [EZ], and retinal pigment epithelium [RPE]), nascent geographic atrophy (nGA), intraretinal hyperreflective foci (IHRFs), hyporeflective drusen cores (hDCs), shallow irregular RPE elevations (SIREs), and drusenoid pigment epithelial detachment (DPED). Biomarkers were classified using a trinary certainty system: "definitely present" (≥90%), "questionably present" (50%-90%), or "absent" (<50%). The main outcome was odds of intergrader agreement from generalized estimating equations adjusting for intraparticipant correlations, biomarker type, and eye- and participant-level factors.

Odds of agreement were comparable for large drusen, nGA, RPE abnormality, IHRF, and ELM abnormality; higher for RPD (odds ratio [95% confidence interval], 3.17 [1.56-6.45], P < 0.01); and lower for SIRE, EZ abnormality, pigmentary abnormalities, hDC, and DPED (up to 0.5 [0.32-0.78], P < 0.01). Agreement improved with higher grading certainty (1.4 [1.22-1.6], P < 0.0001) but declined with age (per decade, 0.8 [0.7-0.93], P < 0.01).

Biomarker type, grading certainty, and participant age influence OCT agreement. Consensus definitions and grading protocols, including the use of high certainty thresholds (≥90%), may improve consistency.

Implementing consensus definitions and protocols can improve agreement and strengthen the clinical utility of OCT biomarkers in AMD.