We performed 4 sensitivity analyses that rely on distinct assumptions regarding instrument validity, to determine the robustness of these findings. The sensitivity analyses showed a consistent pattern supporting weak evidence for a causal effect of cannabis use on schizophrenia and strong evidence for a causal effect of schizophrenia on cannabis use (Table 3). As an exception, the evidence provided by MR-Egger SIMEX for a causal relation from schizophrenia risk to cannabis use, was very weak. However, since the Egger intercept was not significantly different from 0 (Supplementary Table S10), indicating no pleiotropic effects for the SNPs included in the genetic instruments27, it is likely that this method simply lacked power to be able to reject the null hypothesis of no causal effect28.
