Primary association analyses were performed between all SNPs in the two chromosome regions and smoking behavior using logistic regression run in PLINK v1.06. Smoking at age 14/31 were dichotomized to increase statistical power and entered as dependent variables. Previous studies(3, 7, 17) have consistently shown an association between genetic variation within 15q25 and heavy-smoking/dependence, but not with light-smoking(3, 17, 48). Therefore, for CHRNA5-CHRNA3-CHRNB4, smoking behavior was dichotomized as: 1) Smoking at 14: regular-smokers Vs. occasional + non-smokers 2) Smoking at 31: heavy-smokers Vs. light-smokers + non-smokers (i.e. in both time points we compared heavier smokers with the rest of the sample). In contrast, for the TTC12-ANKK1-DRD2 gene-cluster, smoking behavior was dichotomized as smokers Vs. non-smokers, because we hypothesized that genetic variation at this site might be relevant for smoking use regardless of smoking severity.
