risk factor. Indeed, the effect of smoking on these outcomes is sufficiently strong that variants associated with heaviness of smoking achieve genome-wide significance even in unstratified GWAS (i.e., where smokers and never-smokers are not considered separately). When stratified, one should see the association only in ever-smokers and not in never-smokers (see Box 3) [16–18], although due to misclassification (i.e., misreporting of smoking status), this is not always the case.
