Comparative diagnostic performance of six imaging modalities for detecting macular atrophy in neovascular age-related macular degeneration

Purpose: To evaluate the diagnostic accuracy of six imaging modalities - colour fundus photography (CFP), multicolour imaging (MC), blue autofluorescence (BAF), green autofluorescence (GAF), near-infrared reflectance (NIR) and structural optical coherence tomography (OCT) - for detecting macular atrophy (MA) in patients with previously treated neovascular age-related macular degeneration (AMD). Methods: This retrospective cohort study included 113 eyes from patients treated with anti-vascular endothelial growth factor for at least 1 year. All eyes underwent six imaging modalities during the same visit. MA was identified by a senior retinal specialist using multimodal criteria and served as the diagnostic reference (ie, gold standard). Sensitivity and specificity of each imaging modality were calculated relative to this reference. Two masked graders independently assessed images using modality-specific definitions of MA. Discrepancies were resolved by consensus. Sensitivity, specificity, positive predictive value, negative predictive value and inter-reader agreement were calculated for each modality. Results: MA was present in 46.9% of eyes based on multimodal imaging. Structural OCT detected MA in 43.4% of cases, followed by GAF (36.3%), BAF and NIR (35.4% each), MC (33.6%) and CFP (30.1%). Structural OCT demonstrated the highest diagnostic accuracy, with 92.4% sensitivity and 96.7% specificity. Other modalities showed moderate sensitivity (64.1-77.4%) and specificity (75.0-88.3%). Inter-reader agreement was moderate for BAF (κ=0.46) and lower for the remaining modalities, lowest for CFP (κ=0.18). Conclusions: Structural OCT offers the highest accuracy and reliability for detecting MA in neovascular AMD, supporting its use in both clinical practice and research. Other modalities may supplement OCT when needed, but CFP alone is limited.