UGI External Seminar - Professor George Davey Smith - University of Bristol

Abstract : “Natural selection is through phenotype, thus the issue of phenotypic causality has been central to evolutionary and genetic thought in their modern (i.e. post-1850s) formulations. The rediscovery of Mendel’s work at the turn of the 20th Century provided a potential mechanism integrating the two, as reflected in the work of George Udny Yule, John Brownlee, Edward East and others.

This culminated in RA Fisher’s (not generously referenced/ footnoted) 1918 paper. Sewall Wright introduced an analogous framework simultaneously with Fisher, together with the entirely original formulation of path analyses as a causal analytical strategy when reliable causal anchors were available. Wright also contributed to his father’s introduction of the instrumental variable framework in 1928, though did not link this to his work on genetics and phenotypic causality. In the mid-20th Century Fisher related thinking about randomization in agricultural studies to the factorial randomization of Mendelian genes.

Occasional informal uses of genetics to strengthen inference on phenotypic causation in humans was seen from the 1960s onwards. The term Mendelian randomization was introduced by Richard Gray and Keith Wheatley in 1991 when proposing a method using family relatedness to provide evidence of effectiveness of bone marrow transplant in hematopoietic malignancies. The term was used for studies of the kind today recognized as MR from the early 2000s on. Echoing the development of Wright’s work, neither the term, nor the concept, was influenced in any way by instrumental variable analysis. For making inferences on phenotypic causality – both in evolutionary and epidemiological settings – the key assumption is of gene-environmental equivalence, which should be the starting point for any serious consideration of the approach.