A second important distinction between CATS and SAGE relates to the distribution of cannabis dependence symptoms. Due to study design (i.e. opioid dependent cases and controls from high risk neighborhoods), the rate of lifetime cannabis use in CATS was 96.1%, and a majority of the participants reported experiencing cannabis dependence symptoms. Despite the oversampling for alcohol dependence in SAGE, rates of cannabis use and endorsement of dependence symptoms were lower, presumably due to the control group, which was not environmentally matched to the alcoholic cases. This necessitated the inclusion of never-users of cannabis in the replication analyses; however, restricting analyses to ever users produced a similar, albeit non-significant, pattern of results (Figure S7). The distinction between the phenotypes used in CATS and SAGE is nontrivial. By including never-users in SAGE, the effects of genotype and environment and their interplay on later, more problematic stages of cannabis involvement (i.e. dependence symptoms) cannot be disentangled from their effects on cannabis use. As a corollary, the high rate of cannabis use in CATS precluded any study of contributors to its variance. Therefore, we
