Abstract
PURPOSE: The purpose of this study was to investigate the change of ellipsoid zone (EZ)/retinal pigment epithelium (RPE) loss ratio on optical coherence tomography (OCT) in geographic atrophy (GA) secondary to age-related macular degeneration using deep learning-based analysis. METHODS: OCT volumes of patients from the OAKS (NCT03525613) and DERBY (NCT03525600) trials, two phase III prospective randomized trials of GA treated with sham, intravitreal pegcetacoplan monthly (PM) and every other month (PEOM) were included. RPE and EZ loss were measured on OCT using validated deep learning-based algorithms. Pooled study eyes were divided into 4 quartiles determined by EZ/RPE loss ratios on OCT at baseline, month 12, and month 24. The change in ratio with treatment and its association with further disease progression and therapeutic response were analyzed. RESULTS: Eight hundred eighty-nine OCT volumes were included. Overall, quartiles shifted downward at 12 and 24 months, representing a moderate decrease in disease activity, particularly in treated eyes. The EZ/RPE loss quartiles at month 12 were prognostic for further disease progression in the sham eyes for RPE and EZ loss. There was a higher likelihood of a quartile decrease in the PM and PEOM groups versus the sham at month 24 (P value quartile shift PM versus sham = 0.14, PEOM versus sham = 0.0053). CONCLUSIONS: The change in EZ/RPE loss ratios at month 12 remained to be predictive for disease activity and therapeutic response, providing insights into disease mechanisms and relevant guidance for GA management in clinical practice. The EZ/RPE loss ratio decreased either due to advanced GA or induced by therapy in treated patients versus sham.