Polygenic contribution to cannabis use disorder shared with and independent of ADHD: a GWAS-by-subtraction approach

Abstract Cannabis use disorder (CUD) is a heterogeneous condition characterized by different patterns of comorbidity and clinical presentation, yet the genetic contributors to this heterogeneity remain poorly understood. CUD frequently co-occurs with attention-deficit/hyperactivity disorder (ADHD), reflecting, in part, their shared genetic liability. Distinguishing the genetic component of CUD that operates through ADHD from the component independent of ADHD may help clarify whether these two components differ in their association with substance-use phenotypes, psychiatric traits, and broader health outcomes. We applied genomic structural equation modeling–based GWAS-by-subtraction to large-scale ADHD and CUD genome-wide association studies to characterize these two CUD components and investigate their genetic and phenotypic association profiles. The ADHD-independent component accounted for approximately 35% of CUD SNP-based heritability and included three genome-wide significant loci. Although both components showed genetic correlations with substance dependence and risk-taking behaviors, they exhibited distinct patterns of genetic correlation with substance use phenotypes, other mental disorders, and health outcomes. In a phenome-wide association study of 114,410 Estonian Biobank participants, both genetic components were associated with multiple health conditions, but the ADHD-related genetic component generally showed larger effects. Across two clinical samples, polygenic score generated with both components further revealed distinct associations with the comorbidity between ADHD and substance use disorders (SUD). Both scores were associated with SUD among individuals without ADHD. In contrast, only the polygenic score for the ADHD-related component (PGSadhd) was associated with ADHD and comorbid ADHD and SUD, with the strongest association observed in the comorbid group. Mediation analysis further indicated that ADHD diagnosis accounted for approximately 20% of the association between PGSadhd and SUD. Together, these findings demonstrate that ADHD-related and ADHD-independent components of CUD have distinct genetic and phenotypic association profiles, suggesting that genetic heterogeneity may contribute to the clinical heterogeneity of SUD. This framework may help guide further research into the etiology and clinical presentation of the disorder.