Correction for sample overlap, winner's curse and weak instrument bias in two-sample Mendelian Randomization
Inverse-variance weighted two-sample Mendelian Randomization (IVW-MR) is the most widely used approach that uses genome-wide association studies summary statistics to infer the existence and strength of the causal effect between an exposure and an outcome. Estimates from this approach can be subject to different biases due to: (i) the overlap between the exposure and outcome samples; (ii) the use of weak instruments and winner's curse. We developed a method that aims at tackling all these biases together. Assuming spike-and-slab genomic architecture and leveraging LD-score regression and other techniques, we could analytically derive and reliably estimate the bias of IVW-MR using association summary statistics only. This allowed us to apply a bias correction to IVW-MR estimates, which we tested using simulated data for a wide range of realistic scenarios. In all the explored scenarios, our correction reduced the bias, in some situations by as much as 30 folds. When applied to real data on obesity-related exposures, we observed significant differences between IVW-based and corrected effects, both for non-overlapping and fully overlapping samples. While most studies are extremely careful to avoid any sample overlap when performing two-sample MR analysis, we have demonstrated that the incurred bias is much less substantial than the one due to weak instruments or winner's curse, which are often ignored.