Except for TOPMed studies, in which the genetic data were derived from deep whole-genome sequencing, participants in all studies were genotyped on genome-wide arrays. The majority of studies imputed their genotypes to the Haplotype Reference Consortium41 (for EUR ancestries) or 1000 Genomes42 (Supplementary Table 1). GWAS summary statistics were generated in each study sample typically using RVTESTS43, BOLT-LMM44 or SAIGE45 with covariates of sex, age, age squared and genetic principal components according to an analysis plan detailed in the Supplementary Note. Studies composed primarily of closely related individuals (for example, family studies) first regressed out covariates, inverse-normalized the residuals as necessary and then tested additive variant effects under a linear mixed model with a genetic kinship matrix for all phenotypes. Some studies of unrelated individuals followed the same analysis for quasi-continuous phenotypes (AgeSmk, CigDay and DrnkWk), but estimated additive genetic effects under a logistic model for binary phenotypes (SmkInit and SmkCes).
