Fig. 1: Study design and Mendelian Randomisation (MR) assumptions.

Study design: Solid paths are hypothesised to exist, whereas dotted paths are hypothesised not to exist according to MR assumptions; β is the causal relationship of interest to be estimated, where β = α/γ. γ and α are the estimated direct effects of a SNP on the exposure and the outcome, respectively. MR assumptions: MR relies on three core assumptions for valid instrumental variables. These include: Relevance (IV1) – the instrument is associated with the risk factor of interest; Exchangeability (IV2) – the instrument is not associated with any potentially confounding variable; and Exclusion Restriction (IV3) – the instrumental variable can only influence the outcome via the risk factor (Fig. 1). In light of the first assumption, the genetic instruments were constructed using top SNPs associated with the exposure variables. The second and third assumptions are violated if instrument SNPs show horizontal pleiotropy, influencing the outcome through other causal pathways than the exposure, or correlated pleiotropy, where genetic variants for the exposure are also associated with a confounder. Therefore, several sensitivity analyses were conducted to detect and remove possible pleiotropic genetic variants, as detailed in the Methods and Results. SNP single nucleotide polymorphism.