Mendelian randomization
Mendelian randomization methods
A major goal of epidemiology is to reduce the burden of disease in populations through interventions that target causal determinants of disease risk. Although observational studies, such as prospective cohort studies and case-control studies, can provide evidence with regard to disease etiology, limitations such as residual confounding, reverse causation bias, and measurement error severely constrain the ability to infer causality.
Mendelian randomization (MR) is a relatively new form of evidence synthesis and causal inference that is of growing importance in observational epidemiology. The approach can be viewed as an application of instrumental variable analysis, a technique originally developed in the field of econometrics, and exploits the principle that genotypes are not generally associated with confounders in the population and should be immune to reverse causation bias.
MR primer
In the following video, Professor George Davey Smith gives us an overview of Mendelian randomization in two minutes. He explains what it is and how it can help us to understand the causal impact of behaviours, such as smoking, on health.
Mendelian randomization in IEU
Within the MRC IEU, we have been developing a series of methods for Mendelian randomization. The approach is proving ever more popular, due in part to the explosion in publicly available data from large international genome-wide association consortia and cohort studies, which has led to a dramatic rise in the number of genetic variants available for MR analyses and an increased power for testing causal hypotheses.
The landscape of methods available for MR studies is also undergoing a rapid expansion, as new and rich data sources are enabling increasingly sophisticated statistical methods to be applied to assess causal hypotheses and to probe the assumptions necessary for valid causal inference. The IEU is at the forefront of this initiative.