Association analysis was performed separately in AA and EA families using a generalized estimating equation (GEE) framework to account for family relatedness by considering each family as a cluster. The GEE employs a logistic regression model (i.e., binomial distribution) to account for relatedness in the R package GWAF46. Gender, birth cohort, GWAS array indicator, and the first four principal components (as in23) were included as covariates in the model. A trans-ancestral (EA+AA) GWAS was performed by meta-analyzing summary statistics from the EA and AA GWAS using inverse-variance weighting in METAL47, with genomic control.