Table S1. SNP selection process for lifetime cannabis use ® bipolar disorder-association. MAF = mean allele frequency. SNPs = single nucleotide polymorphisms.

Table S2. SNP selection process for bipolar disorder ® lifetime cannabis use-association. MAF = mean allele frequency. SNPs = single nucleotide polymorphisms.

Table S4. Post-hoc power calculations performed using CNSGenomics mRnd app (https://shiny.cnsgenomics.com/mRnd/) for Binary outcomes. The calculator used for this analysis was developed for one-sample Mendelian Randomization, therefore and the present these plots represent approximations. α = Type-1 error rate. K = proportion of cases in the study. OR = odds ratio (point estimates from Mendelian randomization analyses). ?!” # = Proportion of variance explained for the association between the SNP or allele score (z) and the exposure variable (z).

Figure S1. Power plot for the lifetime cannabis use (P < 5e-8) ® bipolar disorder analysis. Calculated using CNSGenomics mRnd app (https://shiny.cnsgenomics.com/mRnd/) for Binary outcomes. The calculator used for this analysis was developed for one-sample Mendelian Randomization, therefore and the present these plots represent approximations. Sample size = 51,710. α = 0.05. K = 0.39. ?!” # = 0.0017. OR = odds ratio, representing the true effect of the association.

Figure S2. Power plot for the lifetime cannabis use (P < 1e-6) ® bipolar disorder analysis. Calculated using CNSGenomics mRnd app (https://shiny.cnsgenomics.com/mRnd/) for Binary outcomes. The calculator used for this analysis was developed for one-sample Mendelian Randomization, therefore these plots represent approximations. Sample size = 51,710. α = 0.05. K = 0.39. ?!” # = 0.0043. OR = odds ratio, representing the true effect of the association.

Figure S3. Power plot for the lifetime cannabis use (P < 5e-8) ® bipolar disorder analysis. Calculated using CNSGenomics mRnd app (https://shiny.cnsgenomics.com/mRnd/) for Binary outcomes. The calculator used for this analysis was developed for one-sample Mendelian Randomization, therefore these plots represent approximations. Sample size = 68,082. α = 0.05. K = 0.28. ?!”# = 0.0029. OR = odds ratio, representing the true effect of the association.

Figure S4. Results from leave-one-out analysis. When testing whether bipolar disorder is causal for lifetime cannabis use, we repeated IVW regression 17 times. For each repetition, we excluded one of the 17 instrumental variables (SNPs). The red lines show the slope (solid line) and 95% confidence intervals (dashed lines) from these regressions, while the black lines show the slope and confidence interval from the original analysis (that included all 17 SNPs).