To estimate the proportion of variation in age at menarche and age at natural menopause that can be explained by the identified genome-wide significant SNPs, we pooled the original genotype and phenotype data from the NHS and the WGHS for the 23 genome-wide significant SNPs. After first adjusting the age at menarche or age at natural menopause for principal components of population substructure within each study separately, we performed a forward/backward stepwise selection of candidate SNPs with the BIC criteria in regression models that also included a study variable to select a minimal set of non-redundant SNPs for each phenotype. The proportion of variation explained (R2) in the combined phenotype was calculated by comparing the model containing the final selected SNPs and a study variable to the model containing a study variable alone.
