MR Dictionary

Multivariable MR

MR analyses including multiple exposures in a single estimation. Multivariable MR can be used to estimate mediating effects of an independent variable, to adjust for possible pleiotropy bias due to horizontal pleiotropy of a specific effect, or to adjust for potential confounding. The estimate obtained from a multivariable MR analysis can be interpreted as the direct effect of an exposure of interest (adjusting for the other exposures included in the estimation) on the outcome.

All of the genetic IVs should fulfil the MR assumptions. In addition, each exposure should be strongly predicted by the set of genetic IVs conditional on the other exposures included. When using multivariable MR to adjust for confounding or horizontal pleiotropy, it is important to make sure that an exposure on the causal path between the primary exposure of interest and the outcome is not being adjusted for. In the context of mediation, multivarible MR can be coupled with univariable MR results and formally through two-step MR to estimate the total, direct and indirect effects of an exposure on an outcome of interest.

Multivariable Mendelian randomization. Adapted from Zheng et al. Multivariable MR uses multiple instruments (Z1, ..., Zn) associated with multiple, potentially correlated exposures (e.g., X1, X2 and X3) to jointly estimate the independent causal effect of each of the exposures on a particular outcome (Y). It can also be formally used to explore mediation following two-step MR.
Figure 2.3 - Multivariable Mendelian randomization. Adapted from Zheng et al. Multivariable MR uses multiple instruments (Z1, ..., Zn) associated with multiple, potentially correlated exposures (e.g., X1, X2 and X3) to jointly estimate the independent causal effect of each of the exposures on a particular outcome (Y). It can also be formally used to explore mediation following two-step MR.

References

Other terms in 'Definition of MR and study designs':