Table 5.

Assumptions of the Study Designs and Action Taken to Satisfy Them

Observational Study		MR study
Assumption	Action to satisfy assumption and minimise bias	Assumption	Action to satisfy assumption and minimise bias
Absence of unmeasured or residual confounding	Inclusion of potential confounders in analysis models	Absence of pleiotropic effects that influence outcome	Use of genetic instruments associated with exposure at genome-wide significance level; assessment of pleiotropy in the IVW estimate; comparison of IVW estimate with those of sensitivity analyses more robust to pleiotropic effects; assessment of consistency of instrument effects
Absence of reverse causation	Assessment of anxiety exposures prior to most common period of AN onset	Absence of association between genetic instrument and confounders of exposure-outcome association	Use of genetic instruments associated with exposure at genome-wide significance level; assessment of consistency of instrument effects; use of GWAS completed in European samples only
Missing data does not depend on unobserved data	Imputation of missing data; inclusion of predictors of missingness in imputation and analysis models; comparison of models with imputed, maximum available and complete case data	Robust association between genetic instrument and exposure	Use of genetic instruments associated with exposure at genome-wide significance level