to test whether the detected association represented an independent signal. We list top five SNPs, regardless of their P‐values, to allow for comparison with results from the follow‐up studies adjusting for the relevant covariates. For SNPs identified as significantly associated with the transitions, we further performed mediation analyses to investigate whether the association is through plausible mediators. Details of the mediation analyses are presented in the Appendix S1. We used GWAVA (Ritchie et al. 2014) for predicting the potential functional effects of the associating noncoding region SNPs.
