SNP–phenotype associations were tested in logistic regression models implemented in SAS version 9.1. Genotypic effects were modeled as the additive effect of the number of minor alleles carried, and adjusted for race and gender. Analyses were also performed separately within the two racial groups. To control for multiple comparisons, we computed the false discovery rate (FDR) [34] for each of the 25 observed P-values. Briefly, this procedure uses the ranking of the observed P-values, the number of tests preformed and the hypothesized proportion of true null tests to determine the FDR or Q-value, defined as the expected proportion of tests called significant at a given P-value that would be false positives.
