Fig. 3

Illustration of Egger regression approach to Mendelian randomisation. Each point in the scatter plot represents an instrumental SNP. The vertical axis represents the estimated slope of the regression of Y (serum total triglycerides) on the SNP. The horizontal axis represents the estimated slope of the regression of X on the SNP. The black and red lines correspond to the Egger regression and weighted median estimate of the causal effect of X on Y, respectively. In this particular example, the intercept of the Egger regression is not significantly different from zero, which we interpret to indicate little evidence of directional pleiotropy (p=0.574). The estimates of the causal effect produced by Egger regression and by the weighted median estimator are both significant and in reasonable concordance. Grey dashed lines represent the 95 percent confidence intervals for the slopes in the regressions