Clinical and Clustering-Based Subtyping of Extensive Macular Atrophy With Pseudodrusen-Like Appearance (EMAP)

To explore the usefulness of hierarchical clustering to classify extensive macular atrophy with pseudodrusen-like appearance (EMAP) based on macular and peripheral atrophy extension.

In this cross-sectional study, all participants underwent best-corrected visual acuity (BCVA) testing, optical coherence tomography (OCT), short-wavelength fundus autofluorescence (SW-AF), and ultra-widefield color fundus photography (UWF-CFP). Peripheral atrophy was quantified in degrees using OptosAdvance software, and macular atrophy area was manually measured. Based on the relative proportion of macular and peripheral atrophy, eyes were clinically classified as having predominantly central, mixed, or predominantly peripheral disease. Hierarchical clustering was performed using macular atrophy area and peripheral angular extension.

Eighty-five eyes from 45 EMAP patients (mean age, 64.0 ± 8.7 years; 26 females, 57.8%) were included. Peripheral pavingstone-like degeneration was observed in 57% of eyes. Four clusters were identified: Cluster 1, minimally atrophic (20%); Cluster 2, predominantly central (20%); Cluster 3, mixed (27%); and Cluster 4, predominantly peripheral (33%). Cluster 2 showed significantly worse BCVA (P = 0.013 vs. Cluster 1; P = 0.007 vs. Cluster 4) and higher prevalence of foveal atrophy (P = 0.004) and fibrosis (P < 0.0001). No significant differences in age or refractive error were found among the groups.

Clinical and clustering-based classifications converged on consistent EMAP subtypes defined by macular and peripheral atrophy distribution. These findings suggest that combining clinical expertise with data-driven methods can help describe disease variability and guide future research.

Integrating clinical expertise with unsupervised clustering may improve the identification of EMAP subtypes, aiding patient selection and outcome stratification in future trials and longitudinal studies.