We used an inverse-variance meta-analysis to test the effects of each genetic variant on age at menarche across the 32 studies. Fixed effects models were used, although in the absence of significant heterogeneity choice of model has little impact on the results. In order to correct for potential relatedness between two Icelandic cohorts (AGES and deCODE), the corrected association results for these cohorts were first meta-analysed and the GC method was re-applied to the results of the combined sample. These results were then meta-analysed with the remaining 30 studies. We also display further results following a second correction for GC using the overall genomic inflation factor calculated from the meta-analysis of all 32 studies. All meta-analyses were conducted using the METAL software package. We considered P-values <5×10−8 to indicate genome-wide significance.
