Genetic liability to cannabis use disorder and migraine susceptibility: evidence from Mendelian randomization

Background Migraine is a highly burdensome neurological disorder. Cannabis-based medications have gained attention for migraine symptom management, but their relationship remains uncertain. We investigated the bidirectional associations between genetic liability to cannabis-related traits and migraine susceptibility. Given their frequent psychiatric comorbidity, we also explored the contribution of psychiatric liability to these relationships. Methods We conducted bidirectional Mendelian randomization (MR) analyses using genome-wide association study summary statistics from individuals of European ancestry. Genetic liability to cannabis use disorder (CUD, inability to cease cannabis consumption despite harm) and lifetime cannabis use (LCU, self-reported ever using cannabis) was evaluated in migraine discovery and replication datasets, with subgroup analyses of migraine with aura (MA) and migraine without aura (MO). Multivariable MR (MVMR) assessed the contribution of psychiatric liability, including major depressive disorder (MDD), anxiety disorder, schizophrenia, and bipolar disorder. Effect estimates represent changes in outcome odds per one-unit increase in the log-odds of genetic liability, with false discovery rate (FDR) correction for multiple testing. Results Forward MR showed a positive IVW point estimate for genetic liability to CUD and migraine susceptibility in the discovery dataset, but this association did not remain significant after FDR correction (OR IVW = 1.078; 95%CI, 1.004–1.158; p = 0.039; p FDR = 0.078). In the replication dataset, a small positive association was observed and remained significant after FDR correction (OR IVW = 1.097; 95%CI, 1.027–1.172; p = 0.006; p FDR = 0.024). No clear association was observed for LCU with migraine or for CUD with MO or MA. Reverse MR showed that the association between migraine liability and LCU did not survive FDR correction in the discovery dataset and was not replicated. Standard MVMR estimate for the CUD–migraine association was attenuated after adjustment for MDD, with similar patterns in sensitivity analyses. Potential sample overlap and low conditional F-statistics in MVMR may limit inference. Conclusions Our findings support a small association between genetic liability to CUD and migraine susceptibility in the replication dataset, although the corresponding discovery association did not remain significant after FDR correction. Exploratory MVMR findings suggest that MDD-related liability may contribute to this relationship, but should be interpreted cautiously.