Chapter 9:

In this chapter, we consider extensions to the basic Mendelian randomization paradigm. These include approaches with multiple exposure variables and those which aim to estimate a parameter other than that estimated in a standard Mendelian randomization investigation. We consider in turn multivariable Mendelian randomization, network Mendelian randomization, non-linear Mendelian randomization, factorial Mendelian randomization, bidirectional Mendelian randomization, and meta-analysis in Mendelian randomization.

Mendelian randomization provides a framework for using genetic variants as proxy measurements for an exposure to assess the impact on an outcome of an intervention in the exposure. We demonstrate how this framework can be exploited in various ways beyond simply considering the causal effect of a single exposure variable