The prior GWAS of marijuana use disorder were conducted in the SAGE cohort.(Agrawal et al., 2014) Since a subset of the population controls used for our GWAS were drawn from SAGE, we repeated our analyses after excluding controls from SAGE and the similarly ascertained COGEND cohort. Despite the reduced sample size ─ nearly 3,600 fewer controls, rs9891146 was associated with drug abuse at meta-analysis P=0.026, thus providing independent nominally significant evidence for association: OR (95% CI)=1.13 (1.00–1.27) in EAs and 1.08 (0.94–1.24) in AAs. Rs8071463 and rs9897982 both had meta-analysis P values of 0.20 in the reduced sample but consistent directions of association with the full sample.
