Event

Mendelian randomisation (MR) allows estimation of the casual effect of a modifable phenotype on a disease outcome using genes as instrumental variables. If the phenotype causally affects the disease, genes altering the phenotype level will be associated with disease, and the effect of the phenotype can be estimated from the gene-phenotype and gene-disease associations. MR estimates are unconfounded provided that some assumptions are met. The main assumption is the absence of pleiotropy, that is the gene in uences the outcome only through the given phenotype. Excluding pleiotropy may be difficult even for well-studied genes, and the use of multiple instruments can indirectly address the issue: if all genes represent valid instruments, their MR estimates will vary only by chance. Formal testing of pleiotropy can be performed by the Hausman test, but the test requires individual data on both gene-phenotype and gene-disease associations for each gene. An alternative approach to test for the possible presence of pleiotropy, based on the use of between-instrument heterogeneity in a meta-analysis of MR estimates from multiple instruments, is presented. This approach can be used with aggregate data, that is estimates and standard errors of both gene-phenotype and gene-disease associations for each instrument.