Fig. 1.

Schematic modelling of assumptions: (A) Mendelian randomisation framework, where an instrumental variable ($Z$) influences only the exposure ($X$), allowing inference on causal influence of exposure on outcome ($Y$) even in the presence of confounds ($U$); the causal effect of interest is indicated by the dashed arrow and $\beta$ denotes the causal effect estimate. (B) Instrumental variable assumptions: (1) The relevance assumption, that the IV is associated with the exposure; (2) the independence assumption, that there are no unmeasured confounds of IV and outcome; and (3) the exclusion restriction, that the IV is only associated with the outcome via the exposure.