The primary limitation of this study is combining drug abuse cases from one cohort and population controls drawn from several other cohorts. This design enabled us to construct the largest GWAS to date for a drug abuse phenotype, other than nicotine and alcohol; and there are other examples of GWAS comparing cases to population controls.(Luca et al., 2008; Silverberg et al., 2009; Wellcome Trust Case Control, 2007) Nonetheless, we recognize the caveats of using population controls. Potential phenotype misclassification among controls may reduce statistical power, generating false negative results. However, the relatively low prevalence of our phenotype (injection drug use in the past 30 days plus using illicit drugs 10+ times in the past 30 days) in the U.S. population (lifetime history of injecting drugs of 2.6% (Lansky et al., 2014)) suggests that phenotype misclassification was unlikely to substantial effect the study results.
