*Geek Box: Mendelian Randomisation
Mendelian randomisation [MR] is a principle of using genetics to mimic a long-term randomised controlled trial, particularly where a long-term intervention study may be unethical or practically infeasible. Because an individuals’ genes are ‘assigned’ when they are conceived, this in effect it is the purest form of randomisation, i.e., the genetic lottery from Mom and Pops.
Well-conducted MR can provide an unconfounded estimate of the relationship between an exposure and an outcome. It is unconfounded because the genetic variant results in a certain physiological response that is independent of other considerations. Thus, to be properly conducted, a MR study has to satisfy three criteria:
1. The genetic variant must be associated with the specific mediating exposure’, e.g., LDL-C;
2. The genetic variant must not be associated with any potential confounders that could influence the outcome, and;
3. The genetic variant must only influence the disease outcome through the specific exposure pathway, not through other mechanisms.
An IV is only valid where the 3 assumptions above hold. This is crucial, because it means that claims of “causality” can only be made where these assumptions are met. Where there may be factors that undermine these assumptions, then an MR study should be considered genetic associations, not necessarily a cause-effect relationship. Several potential issues may arise for nutrition research.
In the first instance, examples of where a genetic variant provides a strong IV for mimicking dietary intake are rare. And importantly, a genetic variant IV may only be associated with a specific, isolated tissue compartment, e.g., plasma, which may not be the sole pathway through which a given nutrient may be associated with an outcome. Where an IV mimics nutrient levels in a specific tissue compartment, conclusions should be confined to that tissue compartment, not stated as the effect of a nutrient broadly in relation to the outcome of interest.
When long-term randomised studies are not possible, Mendelian randomisation is a powerful tool to examine potential cause-effect relationships. But we should temper our enthusiasm for thinking anything genetic solves all methodological challenges in our field, as MR of nutritional exposures faces several methodological challenges itself that need to be considered.