Each cohort performed two SNV association analyses using an additive model implemented with the R package SeqMeta, http://cran.r-project.org/web/packages/seqMeta/index.html. Analyses were stratified by ancestry. One SNV association analysis used an untransformed model with RR-interval as the outcome, adjusted for age, sex, BMI and cohort-specific adjustments. The other SNV association analysis was a model based on the rank-based inverse-normal transformed residuals (RR-INVN), with residuals taken from a linear regression RR-interval adjusted for age, sex and BMI covariates. The RR-INVN analysis was performed to check for potential sensitivity to deviations from normality within the analysis of rare variants. Additional cohort-specific covariate adjustments were also applied, which included for example PCs or family structure.
