Several GWAS have targeted smoking initiation (Vink et al. 2009; The Tobacco and Genetics Consortium 2010; Thorgeirsson et al. 2010; Siedlinski et al. 2011; Argos et al. 2014) or cessation (The Tobacco and Genetics Consortium 2010; Siedlinski et al. 2011; Argos et al. 2014) with phenotypes dichotomized into ever versus never or used as quantitative age of onset phenotypes. Only the large GWAS meta‐analysis of the Tobacco and Genetics Consortium yielded signals in tyrosine kinase and dopamine signaling pathway genes that genome‐wide significantly associated with smoking initiation (never vs. ever smokers) and smoking cessation (former vs. current smokers), respectively (The Tobacco and Genetics Consortium 2010). Time‐to‐event analysis is more powerful than analysis of binary traits or transformed quantitative phenotypes because it incorporates information of follow‐up time span and allows for censoring. There is a huge gap in understanding the contribution of an associating variant to a specific trait. Causal mediation analysis has been used to improve the understanding of the mechanisms underlying detected associations (Jiang et al. 2013; Liu et al. 2013). The estimation of mediation effects in the context
