Risk assessment model for development of advanced age-related macular degeneration.

Publication Type
Journal Article
Year of Publication
Klein, Michael L; Francis, Peter J; Ferris, Frederick L; Hamon, Sara C; Clemons, Traci E
Arch Ophthalmol
Date Published
2011 Dec
Aged; Apolipoproteins E; Complement C2; Complement C3; Complement Factor H; environment; Female; Genetic Predisposition to Disease; Genotype; Humans; Macular Degeneration; Male; Models, Theoretical; Phenotype; Polymorphism, Single Nucleotide; Proportional Hazards Models; Proteins; Risk Assessment; Risk Factors; ROC Curve; Severity of Illness Index; visual acuity

OBJECTIVE: To design a risk assessment model for development of advanced age-related macular degeneration (AMD) incorporating phenotypic, demographic, environmental, and genetic risk factors.

METHODS: We evaluated longitudinal data from 2846 participants in the Age-Related Eye Disease Study. At baseline, these individuals had all levels of AMD, ranging from none to unilateral advanced AMD (neovascular or geographic atrophy). Follow-up averaged 9.3 years. We performed a Cox proportional hazards analysis with demographic, environmental, phenotypic, and genetic covariates and constructed a risk assessment model for development of advanced AMD. Performance of the model was evaluated using the C statistic and the Brier score and externally validated in participants in the Complications of Age-Related Macular Degeneration Prevention Trial.

RESULTS: The final model included the following independent variables: age, smoking history, family history of AMD (first-degree member), phenotype based on a modified Age-Related Eye Disease Study simple scale score, and genetic variants CFH Y402H and ARMS2 A69S. The model did well on performance measures, with very good discrimination (C statistic = 0.872) and excellent calibration and overall performance (Brier score at 5 years = 0.08). Successful external validation was performed, and a risk assessment tool was designed for use with or without the genetic component.

CONCLUSIONS: We constructed a risk assessment model for development of advanced AMD. The model performed well on measures of discrimination, calibration, and overall performance and was successfully externally validated. This risk assessment tool is available for online use.