The observed bi-directional relationship between depression and neuroticism could be confounded by non-independent instrumental variants across both tests, i.e. a region containing variants associated with depression was used to test for a causal effect on neuroticism and then variants in that same region were also used to test for a causal effect of neuroticism on depression. To account for any overlap, we identified and removed 15 instrumental variants from each bi-directional test where there was evidence of linkage disequilibrium (LD r2 > 0.1). The effect size for the IVW regression analysis MR test for depression on neuroticism was attenuated from 0.146 (s.e. = 0.039) to 0.112 (s.e. = 0.042) and the P-value was no longer significant after false discovery rate correction (PFDR = 0.037). The effect size for the IVW regression analysis MR test for neuroticism on depression was attenuated from 0.366 (s.e. = 0.037) to 0.289 (s.e. = 0.042) and the P-value remained significant after false discovery rate correction (PFDR = 1.34 × 10-10).
