We explored potential causal relationships between the mood disorders and other traits using Mendelian randomisation. The interpretation of these analyses is challenging, especially for complex traits, when the ascertainment of cases varies, and when there are relatively few (< 20) variants used as instruments (for example, in the PGC BD and down-sampled analyses presented) (41, 67, 68). Major depressive disorder and bipolar disorder demonstrate considerable heterogeneity (as our subtype analyses show for bipolar disorder types 1 and 2), potentially confounding the results of Mendelian randomisation. That said, our analyses are consistent with a bidirectional influence of educational attainment on risk for mood disorders (and vice versa), with different directions of effect in the two mood disorders. We found no significant relationship between IQ and either mood disorder. We also find results consistent with major depressive disorder increasing the risk for coronary artery disease in a relatively well powered analysis. This mirrors epidemiological findings, although the mechanism remains unclear (69).
