reports in other populations have noted lower prevalence. e.g.,62 A more detailed characterization of the OUIP groups’ opioid use would have been useful, but was not obtained. We examined somewhat divergent ODE’s in our confirmation samples because of ascertainment and assessment differences (Supplementary Table 3). While these methodological differences are a limitation, the observed confirmation in two multi-site U.S.-based studies supports the generalizability of our findings. The highly comorbid composition of the ODE and OUIP groups may raise concerns that intergroup differences are better attributable to another phenotype. Post-hoc association analyses conducted in CATS to address this possibility (Supplementary Table 10) found few associations for comorbid disorders reaching nominal significance. Comorbidity pattern differences across samples (Supplementary Tables 1 and 2) further argue against this possibility. Finally, our exclusion of participants with non-European ancestry and the lack of genomic inflation (λ=1.01) suggest spurious association due to uncorrected admixture is unlikely.
