To systematically compare the prognostic value of multiple shape-descriptive factors in the natural course of the disease.
A total of 296 eyes of 201 patients (female patients 130; mean age: 72.2 ± 13.08 years) with a median follow-up of 2.38 years from 2 prospective, noninterventional natural history studies (Fundus-Autofluorescence-in-Age-related-Macular-Degeneration [clinicaltrials.gov identifier NCT00393692], Directional-Spread-in-Geographic-Atrophy [NCT02051998]) were included in the analysis. Serial fundus autofluorescence images were annotated using semiautomated image analysis software to determine the lesion area, circularity, perimeter, and caliper diameters. These variables and the fundus autofluorescence phenotype were evaluated for prediction of the future square root progression rates using linear mixed-effects models.
For the combined model, leave-one-out cross validation on patient level (Scenario 1: previously unknown patient) resulted in a goodness-to-fit (R2 value) of 0.244 and leave-one-out cross validation on visit level (Scenario 2: previous observation of the patient) in a R2 value of 0.391. This indicated that shape-descriptive factors could explain 24.4% of the variance in geographic atrophy progression in previously unknown patients and 39.1% in patients with previous observation.
These findings confirm the relevance of shape-descriptive factors and previous progression as prognostic variables for geographic atrophy progression. However, a substantial part of the remaining variation in geographic atrophy progression seems to depend on other variables, some of which are visible in optical coherence tomography.
Fundus autofluorescence imaging allows for fully automated prediction of geographic atrophy progression rates based on annotations that are already used in clinical routine and that are part of the eligibility assessment of currently ongoing trials.
*Department of Ophthalmology, University of Bonn, Bonn, Germany;
†The Nuffield Laboratory of Ophthalmology, Nuffield Department of Clinical Neurosciences, Sleep and Circadian Neuroscience Institute, University of Oxford, Oxford, United Kingdom;
‡Institute for Medical Biometry, Informatics and Epidemiology, University of Bonn, Germany; and
§Doheny Eye Institute, University of California Los Angeles (UCLA), Los Angeles, California.
Reprint requests: Monika Fleckenstein, MD, Department of Ophthalmology, University of Bonn, Ernst-Abbe-Street 2, 53127 Bonn, Germany; e-mail: firstname.lastname@example.org
Supported by BONFOR GEROK Program, Faculty of Medicine, University of Bonn, Grant No. O-137.0022, O-137.0025 (to M.P.) and Grant No. O-137.0020 (to M.L.), DFG Grant FL 658/4-1 and FL 658/4-2 DFG Grant Ho1926/3-1.
None of the authors have any financial/conflicting interests to disclose.
M. Pfau and M. Lindner contributed equally to this work.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Web site (www.retinajournal.com).
Heidelberg Engineering (Heidelberg, Germany) provided research material (Heidelberg Spectralis HRA + OCT). Heidelberg Engineering had no role in the design or conduct of this study.