Polygenic prediction of cannabis-related outcomes: Evidence from the Avon Longitudinal Study of Parents and Children longitudinal cohort and the European gene-environment interaction case-control study.

Spinazzola E.; Austin-Zimmerman I.; Trotta G.; Li Z.; Leung P.; Chesney E.; Li Q.; Alameda L.; Arango C.; O'Donovan M. et al. · Addiction (Abingdon, England) · 2026

**Background & Methods** This observational study examined whether polygenic risk scores (PRSs) for cannabis use disorder, major depressive disorder, insomnia, and chronic pain predict problematic cannabis use and early initiation. Two cohorts were analyzed: the Avon Longitudinal Study of Parents and Children (ALSPAC; n=8,224 UK birth cohort participants) and the European Gene-Environment Interaction study (EU-GEI; n=776 controls and 491 first-episode psychosis cases across Europe and Brazil), with outcomes measured via self-reported cannabis use and Cannabis Abuse Screening Test scores. **Key Findings** • In ALSPAC, CUD PRS was significantly associated with problematic cannabis use at ages 20 and 24 years (β=0.04-0.11, pFDR≤0.03); MDD PRS showed consistent associations across all ages (17, 20, 24 years; β=0.06-0.1, pFDR≤0.01); chronic pain PRS associated with problematic use at all three timepoints (β=0.07-0.09, pFDR<0.001). • In EU-GEI, CUD PRS (RRR=1.82, pFDR=0.001) and chronic pain PRS (RRR=1.27, pFDR=0.007) were associated with weekly-to-daily cannabis use; CUD PRS demonstrated evidence of additive interaction (RERI=2.4, P=0.009), while MDD and schizophrenia PRS showed no additive interaction effects. • Age at cannabis initiation explained substantially more variance in problematic cannabis use than any single polygenic risk score, suggesting environmental timing factors are more predictive than genetic liability alone. **Dosage & Administration** Not reported. **Safety & Adverse Effects** Not reported. **Evidence Quality** This study provides moderate evidence given its prospective longitudinal design and conceptual replication across two independent cohorts. However, limitations include reliance on self-reported cannabis use (subject to recall and social desirability bias), variable sample sizes across timepoints, attrition in longitudinal follow-up, and recruitment of FEP cases in EU-GEI (limiting generalizability to non-psychotic populations). The modest effect sizes (β<0.15) and explained variance suggest polygenic scores have limited clinical predictive utility independently. Replication in diverse populations and incorporation of environmental factors would strengthen conclusions.

More recent cannabis research