AMD Genetics: Methods and Analyses for Association, Progression, and Prediction
Age-related macular degeneration (AMD) is a leading cause of blindness in the elderly population of Western countries. In the past few years, over one dozen AMD risk loci have been identified through genome-wide association studies (GWAS), either by individual studies or through meta-analyses of multiple studies from the National Eye Institute (NEI) supported AMD Gene Consortium. However, the genetic causes for disease progression have not been well studied yet. In this talk, using the data from two large multi-center randomized clinical trials -- AREDS (Age-Related Eye Disease Study) and AREDS2, we propose a novel approach to evaluate the effect of the known AMD risk SNPs on disease progression and establish prediction models for AMD progression. Specifically, we calculate the time-to-advanced AMD (either choroidal neovascularization or geographic atrophy) from the baseline visit time for each eye of each patient, and model this eye-level progression time using a bivariate approach, which appropriately accounts for between-eye correlations. We show that our method is computationally efficient for genome-wide studies and is more powerful than existing methods in many scenarios. In addition, we develop prediction models on AMD progression based on the genetic score, demographic factors, and eye-level macular information. We find that the model with genetic score and demographic factors (but without macular information) predicts the progression well, suggesting an effective and efficient screening process which uses only genetic and demographic factors to identify patients at high risk of progression for early interventions.