Mendelian randomization (MR) is a technique that can allow investigators to test for causal relationships using observational data. MR studies separate people based on gene variants known to affect an outcome of interest. Since genes are sorted more or less randomly during conception, separating participants based on gene variants allows researchers to form a natural randomized experiment that eliminates confounders that can limit observational studies.

For example, if there are genes that raise LDL levels in the blood, investigators can examine people who have these gene variants and compare them to people who don’t, in terms of heart disease outcomes. Since these genes work on a chemical pathway, they’re unlikely to correlate with confounders that could also affect heart disease risk, like smoking or diet habits. As a result, researchers can test for a cause-and-effect relationship between LDL and heart disease.

MR studies are not infallible, since they can have important sources of bias that need to be accounted for. For one, investigators need to examine whether gene variants might result in exposures unrelated to the exposure of interest, a phenomenon known as horizontal pleiotropy. Investigators also need to watch out for weak instrument bias, which is when gene variants have such a small effect on exposure that their effects are undetectable. Finally, some gene variants are not randomly distributed in a population and are instead overrepresented in geographic regions or ethnic groups. This can introduce the exact type of confounding MR is designed to circumvent.