Abstract
Purpose :
Geographic atrophy (GA) in age-related macular degeneration may significantly impair central vision, particularly when the fovea is involved. Identifying prognostic biomarkers for foveal involvement is vital for categorizing patients at increased risk of vision loss. We investigated the incidence and risk factors for foveal involvement in patients with initial foveal-sparing GA, using machine learning to assess the importance of each risk factor.
Methods :
Retrospective, longitudinal study conducted at San Raffaele Scientific Institute (Milan, Italy) including 167 eyes from 115 patients. Patients had a minimum GA of 0.049 mm2 within 800 microns from the fovea and mean follow-up of 50±29 months (range 6-137). We collected clinical and imaging data. We employed mixed-model Cox regression analysis and Random Survival Forests (RSF) to identify and rank risk factors for foveal involvement. Higher Variable Importance (VIMP) indicated greater predictive importance.
Results :
Median survival time for foveal involvement was 46 months(95% CI 38-55). Incidence rates were 26% at 24 months and 67% at 60 months. Risk factors included proximity of GA to the fovea (HR=0.97 per 10-mm increase[95% CI 0.96-0.98]), worse baseline vision (HR=1.37 per 0.1-LogMAR increase[95% CI 1.21-1.53]), and thinner outer nuclear layer (HR=0.59 per 10-micron increase[95% CI 0.46-0.74]). RSF analysis identified these as the most crucial predictors (VIMP=17, p=0.002; VIMP=6.2, p=0.003; VIMP=3.4, p=0.01). Lesser GA area at baseline (HR = 1.09[95% CI 1.01-1.16]) for 1-mm2 increase) and presence of a double layer sign (HR=0.42[95% CI 0.20-0.88]) were protective but had lesser importance (VIMP=0.9, p=0.08; VIMP=-0.03, p=0.7).
Conclusions :
Our study harnesses RSF analysis to pinpoint key factors influencing foveal involvement in GA, highlighting the crucial role of anatomical and functional parameters in its progression. These insights can aid clinicians in identifying at-risk patients early and tailoring preventive strategies, thereby improving GA management and patient outcomes.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.