Premium
Is Mendelian randomization ‘lost in translation?’: Comments on ‘Mendelian randomization equals instrumental variable analysis with genetic instruments’ by Wehby et al.
Author(s) -
Lawlor Debbie A.,
Windmeijer Frank,
Davey Smith George
Publication year - 2008
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.3308
Subject(s) - mendelian randomization , terminology , term (time) , context (archaeology) , instrumental variable , computer science , psychology , epistemology , linguistics , biology , genetics , genetic variants , gene , genotype , paleontology , philosophy , physics , quantum mechanics , machine learning
Wehby et al. suggest that researchers use the term ‘instrumental variable analysis with genetic instruments’ rather than ‘Mendelian randomization’ for studies that use genetic variants that proxy for modifiable risk factors to investigate the causal effects of these risk factors [1]. They state that ‘Using common language between disciplines applying IV analysis with genetic variants is essential to increasing collaborations and fostering the application of this method’ [1]. However, we would argue that truly translational research requires direct collaboration between individuals from different disciplines, clear definitions of discipline-specific terminology and the willingness to learn from each other. Any attempts to develop a common language must reflect this reality and may be less important than being able to translate between disciplines. In the limited context of ‘Mendelian randomization’ developing the necessary multidisciplinary approach clearly goes well beyond the term used to describe these studies. What is required is capitalizing on the insights that all disciplines can bring and not arguing over which discipline should be the one with terminological hegemony. It is notable, for example, that throughout their commentary Wehby et al. refer to ‘endogenous’ variables, without any explanation of this term in language that would be understood by clinicians, epidemiologists, geneticists, basic scientists or indeed many biomedical statisticians for whom this term will be unknown. Similarly, they use the term ‘direct’ where most epidemiologists and clinicians would specify ‘causal’ (the latter also having a clearer understanding for lay readers of research). Others have also suggested alternative terms for ‘Mendelian randomization’ including ‘Mendelian deconfounding’ [2] and ‘Mendelian triangulation’ [3], so which name should be given prominence? We are not particularly committed to the name ‘Mendelian randomization’ and have discussed the origin of this term in previous publications [4, 5]. However, it is now recognized