We used an independent sample of 8689 women from the EPIC-InterAct study35 to follow-up our menarche signals. To test associations between each identified SNP and age at menarche with correction for cryptic relatedness, we ran a linear mixed model association test implemented in GCTA34 (--mlma-loco option), adjusting for birth year, disease status and research centre. Given the relatively small sample size compared to our discovery set, directional consistency with results from the discovery-meta analysis was assessed using a binomial sign test. Variance explained by menarche loci was estimated using restricted maximum likelihood analysis in GCTA34. In addition to the 123 confirmed menarche loci, variance explained in subsets of menarche loci below the genome-wide significance thresholds was also assessed.
