New horizons in Mendelian randomization

When considering whether an exposure is a causal risk factor for an outcome, evidence from randomized trials is reliable but typically slow or impractical to gather, whereas evidence from conventional observational studies is often unreliable, as it is subject to bias from confounding and reverse causation.

Mendelian randomization (MR) is an example of a quasi-experimental approach: it is analogous to a randomized trial, but relies on nature doing the randomization for us.

MR can be implemented rapidly for a range of exposures to provide insights about causal relationships that can prioritize or deprioritize exposures for further investigation.

The aim of our research is to develop methods that enable detailed MR analyses to inform policymakers and drug developers about the nature of causal effects: enabling trial interventions to target the right mechanism in the right population group at the right time.

We will develop methodology for MR that informs us on how causal effects occur, when interventions are most effective, and which population groups they are strongest in.

We will also apply our methods to address questions of key biomedical importance, and provide resources and training so that others can apply our methods to their research questions.