Several limitations are worth noting. First, despite interesting findings and the partially high risk sample design, our sample size is underpowered to detect the modest effect sizes typically associated with substance use disorders12. For instance, in the AA subsample, 80% power to detect a common variant (MAF ≥ 35%) is only expected for genotype relative risks≥1.2878, which is fairly high for psychiatric disorders. Second, due to the high degree of relatedness in our data, ascertainment, and the relatively small number of “cases”, methods such as GCTA79 or LD Score regression80 that are typically used to assess SNP-heritability were not appropriate. Instead, we report heritability using familial relatedness. Third, we elected to derive diagnoses based on the DSM-IV nomenclature for dependence instead of DSM-5 definitions for substance use disorders. Even though the DSM-5 definition of substance use disorders is more contemporary, it relies on a lower symptom burden (e.g., ≥2 of 11 vs ≥3 of 7 criteria for DSM-IV dependence), which may dilute identification of genetic effects on more severe forms of the disorder. However, when we examined the association between
