Article
Testing for the absence of causal effects – Mendelian randomization turned around
Search Medline for
Authors
-
Maren Vens
- Institut für Medizinische Biometrie und Statistik, Universität zu Lübeck, Universitätsklinikum Schleswig-Holstein, Campus Lübeck, Lübeck, Germany
-
Michelle Kretschmer
- Institut für Medizinische Biometrie und Statistik, Universität zu Lübeck, Universitätsklinikum Schleswig-Holstein, Campus Lübeck, Lübeck, Germany
-
Inke R. König
- Institut für Medizinische Biometrie und Statistik, Universität zu Lübeck, Universitätsklinikum Schleswig-Holstein, Campus Lübeck, Lübeck, Germany
| Published: | February 26, 2021 |
|---|
Outline
Text
Mendelian randomization is a robust approach for assessing causal relationships using data from observational studies. Genetic variants are used as instrumental variables for causal inference about the effect of a putative causal risk factor on an outcome. Critically, the genetic variants have to fulfill three fundamental assumptions that are strong, whether for estimation or for testing, but at least two of these cannot be tested directly. Since the number of applications of Mendelian randomization is expanding quickly, more attention needs to be paid to these assumptions, and positive and negative results from Mendelian randomization studies are potentially subject to biases from violations.
As a solution, it has been suggested that the strength of Mendelian randomization might not be in the proof but the exclusion of causality, arguing that only in very specific settings would the evidence for no association lack robustness. We therefore propose a statistical procedure following equivalence testing for exclusion of causality using Mendelian randomization. Extensive simulation studies show in which settings this approach is robust and can be recommended for practical application.
The authors declare that they have no competing interests.
The authors declare that an ethics committee vote is not required.
