To maximize discovery, we leveraged genomic structural equation modeling (gSEM)30 to combine new and existing GWAS data with varied, but closely related, phenotypes for OA to enable the largest GWAS of OA to date (23,367 cases, 384,629 controls: effective sample size 88,114). We brought together novel results from the Genetics of Opioid Addiction Consortium (GENOA) with publicly available summary statistics from the MVP-SAGE-YP27, the Psychiatric Genetics Consortium-Substance Use Disorder Group (PGC-SUD)26, and the Partners Health Group (PH)28. We examined SNP-based heritability and genetic correlation among the varied phenotypic definitions of OA across the contributing cohorts, including diagnostic and frequency of use-based cases and different types of controls: opioid exposed, unexposed, and population-based. We conducted a variant level gSEM analysis in the full complement of cohorts and a gene-based association test based on those results. Whereas gSEM accounts for the sample overlap among the GENOA, PGC-SUD, and MVP-SAGE-YP analyses, we were able to combine the samples and increase substantially available sample size compared to standard meta-analysis. Follow-up analyses included: (1) evaluation of genetic correlation with brain-related phenotypes; (2) estimation of predicted
