Although our data included an extraordinarily detailed smoking history, there are still some limitations in our study. Smoking behavior encompasses psychiatric and social behaviors in which both complex genetic and environmental factors are involved; these were not accounted for in our analyses. Also other plausibly relevant covariates, such as socio‐economic status, and working environment, were not considered. Further, our phenotype data were collected retrospectively in subjects with a mean age of 56 years at the time of the interview; thus the accuracy of self‐reported ages of onsets may be influenced by recall bias. Although the interviews contained detailed measures of ND, the age of onset of ND was not assessed, and thus we were not able to perform time‐to‐event analyses of ND. However, we analyzed tolerance, which is the key dimension of ND. The efficient long‐term survival model accounting for family structure, which to the best of our knowledge is not available, would be more appropriate when analyzing the transitions to daily smoking and quitting, and it should be considered in the future. Future studies should also attempt to investigate low‐frequency variants (MAF < 0.05) in the context of transitions in smoking behavior.
