Lesion area, perimeter and diameter as prognostic markers for the progression of geographic atrophy (GA) secondary to age-related macular degeneration

Abstract

Purpose : Multiple shape-descriptive factors have previously been demonstrated to be prognostic for geographic atrophy (GA) progression. The aim of this study was to compare the prognostic value of these factors in a longitudinal natural history study.

Methods : Patients participating in the prospective natural history FAM (NCT00393692) and DSGA (NCT02051998) studies were included in the analysis. Serial fundus autofluorescence (FAF) images were graded using the RegionFinder software (Heidelberg Engineering, Germany) to generate annotated FAF images. A custom-built ImageJ plug-in (Bethesda, Maryland, USA) was used to measure automatically the lesion area, circularity, perimeter and caliper diameters based on the annotation. The baseline variables lesion area, perimeter and maximum caliper-diameter were evaluated for prediction of the upcoming square root lesion progression rate using linear mixed-effects models (LMM). The goodness-to-fit (R²) was used to compare the prognostic model fit among the different LMMs.

Results : A total of 296 eyes of 201 patients (female 130; mean age: 72.2±13.08; median follow-up [Q1–Q3]: 2.36 years [1.45–4.2]) were included in the analysis.
In a subset of 100 visits graded independently by two readers, the intraclass-correlation-coefficient (inter-rater reliability) was 0.99 (95%CI 0.996-0.99) for the area, 0.99 (0.98-0.99) for the perimeter, 0.99 (0.98-0.99) for the circularity, 0.97 (0.95-0.98) for the maximum caliper-diameter.
LMM analysis for the individual factors disclosed that the (sqrt) circularity LMM exhibited the best goodness-to-fit (R²=0.22) followed by the (sqrt) perimeter LMM (0.13) and the maximum caliper-diameter LMM (0.08). Leave-one-out cross-validation on the patient level resulted in a prognostic R² of 0.22 for a combined model containing (sqrt) lesion size and (sqrt) circularity.

Conclusions : These findings confirm the importance of shape-descriptive factors as prognostic variables for GA progression rates. However, the substantial remaining variation in GA progression appears to depend on further variables. These might include FAF phenotype, SD-OCT characteristics and genotype. In the ongoing study, these variables will be included in the model to assess their exact prognostic value and to reveal possible interactions.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.