Mendelian randomization with multiple exposures-the importance of thinking about time

Abstract

Have you ever wondered how Mendelian randomization (MR) studies can estimate a lifetime effect when the exposure is only measured once? This is incredible, considering that other familiar methods2 would require that the exposure (and time-varying covariates) be measured repeatedly and frequently throughout the life course to estimate the same effect. MR avoids this by making important assumptions about time to estimate effects. For example, the assumption that the relationship between the genetic variant(s) and the exposure is constant through time3 allows the estimation of a lifetime effect even when exposure is only measured once. Regardless of the methods used to infer causality, it is not possible to define a causal effect or hypothesis without thinking about time. This may be even more true for MR than for conventional methods because of the potentially long periods of time between when the proposed instruments are set and the exposure and outcome are measured. MR studies increasingly use multiple instruments, multiple outcomes and, as can be seen in recent publications, multiple exposures. The incorporation of multiple exposures in MR adds another wrinkle, however–is it possible to capture longitudinal relationships between variables without longitudinal data? In short, what assumptions do we have to make about time in order to use MR to estimate causal effects which involve the relationship between multiple exposures? And what causal effects are we even estimating?

Publication
International Journal of Epidemiology
Click the Cite button above to demo the feature to enable visitors to import publication metadata into their reference management software.
Click the Slides button above to demo Academic’s Markdown slides feature.

Supplementary notes can be added here, including code and math.

Avatar
Jeremy A. Labrecque
Assistant professor, Epidemiology and causal inference

My research is on how we know what we know.