Another limitation of this study involved the limited availability of drug, smoking, and alcohol histories for use in the discovery GWAS analyses. Drug abuse cases were defined by reported injection and use of illicit drugs 10 or more times in the past 30 days; lifetime drug abuse history was not known, and DSM-IV or other diagnostic criteria were not assessed. However, the ascertainment of cases from UHS (a street-recruited cohort of PWID) and the high correlation between the study threshold and clinical levels of any abuse/dependence, a positive predictive value of 87% in analysis of NSDUH data (see Methods), suggests that the cases very likely meet diagnostic criteria for abuse/dependence. Furthermore, sensitivity analyses in our replication cohort (GAIN) showed that increasing the symptom threshold from 1 to 3 maintained significant replication in spite of the reduced number of cases and statistical power (P=0.0019 to 0.015, Table S12) and very similar magnitudes of association (OR=0.53 to 0.55, Table S12). These results support the generalizability of this study’s genome-wide significant and replicated association to other drug abuse phenotypes, including DSM criteria. Other
