The simulation study has highlighted the benefits of using multivariable MR‐Egger over its univariable counterpart. This is particularly true when the associations of the genetic variants with the risk factor of interest are associated with genetic associations with at least one of the risk factors (measured pleiotropy). Under this scenario, the InSIDE assumption for univariable MR‐Egger is likely to be violated, leading to biased causal estimates. Multivariable MR‐Egger will, however, produce consistent causal estimates if the InSIDE assumption for multivariable MR‐Egger is satisfied. Although the estimates from univariable and multivariable MR‐Egger are asymptotically the same when genetic associations with each risk factor are all independent, multivariable MR‐Egger should also have greater power to detect a causal effect when the InSIDE assumption is satisfied. Given these advantages, and the sensitivity of the multivariable IVW method to directional pleiotropy, we believe that multivariable MR‐Egger should be considered as an important sensitivity analysis for a MR study.
