Table 2.

Mendelian randomization results.

Parameter	X	Y	Causal estimate	se	p-value	# SNPs
α	SIT	TC	0.419	0.052	7.6 × 10−16	6
β	TC	CAD	0.347	0.059	4.8 × 10−9	36
γ	SIT	CAD	0.308	0.065	1.9 × 10−6	6
Indirect effect (α * β) mediated by TC	SIT	CAD	0.145	0.031	2.2 × 10−6
Direct effect (γ − α * β)	SIT	CAD	0.163	0.072	2.3 × 10−2

We performed Mendelian randomization analyses of total sitosterol, total cholesterol and CAD. A schematic figure of the investigated causal relationships is displayed in Supplementary Fig. S8. We provide single causal effect estimates based on the given number of SNPs used as instruments (method: inverse-variance weighting). From these estimates, the direct effect of total sitosterol on CAD and the indirect effect mediated by total cholesterol are calculated as described in the ‘Methods’ section. Both causal effects were positive and significant while the direct effect is slightly larger than the indirect effect. Considering more restricted sets of instrumental variables provided similar results (see Supplementary Data S14) as well as applying different methods of Mendelian randomization analysis (Supplementary Fig. S9).