Testing and correcting for weak and pleiotropic instruments in two‐sample multivariable Mendelian randomization

Abstract

Multivariable Mendelian randomization (MVMR) is a form of instrumental variable analysis which estimates the direct effect of multiple exposures on an outcome using genetic variants as instruments. Mendelian randomization and MVMR are frequently conducted using two‐sample summary data where the association of the genetic variants with the exposures and outcome are obtained from separate samples. If the genetic variants are only weakly associated with the exposures either individually or conditionally, given the other exposures in the model, then standard inverse variance weighting will yield biased estimates for the effect of each exposure. Here, we develop a two‐sample conditional F ‐statistic to test whether the genetic variants strongly predict each exposure conditional on the other exposures included in a MVMR model. We show formally that this test is equivalent to the individual level data conditional F ‐statistic, indicating that conventional rule‐of‐thumb critical values of 10, can be used to test for weak instruments. We then demonstrate how reliable estimates of the causal effect of each exposure on the outcome can be obtained in the presence of weak instruments and pleiotropy, by repurposing a commonly used heterogeneity Q ‐statistic as an estimating equation. Furthermore, the minimized value of this Q ‐statistic yields an exact test for heterogeneity due to pleiotropy. We illustrate our methods with an application to estimate the causal effect of blood lipid fractions on age‐related macular degeneration.

Education Attainment, Intelligence and COVID-19: A Mendelian Randomization Study

“…Whether our findings and policy implications could be generalized to other ethnicities will require further investigation. Fourth, in multivariable MR analysis, we were unable to provide the conditional F-statistics as illustrated by a very recent publication implemented by the MVMR package in R [ 48 ]. As we lacked individual-level data, and we did not estimate the gene-exposure association in separate non-overlapping samples, we could not calculate the pairwise covariance between each genetic instrument with any two exposures for all genetic instruments and all exposures.…”

Noise-augmented directional clustering of genetic association data identifies distinct mechanisms underlying obesity

“…If the traits are estimated in the same sample, then the off-diagonal entries of will be non-zero. Although the correlation between and is not easily estimated, provided the j th genetic variant explains only a small proportion of the variance in the k th and l th traits, then , where X k and X l are the k th and l th traits, respectively [ 50 ]. We can therefore compute the ( k , l )th entry of , i ≠ j , by where is an estimate of the correlation between X k and X l .…”

Non-Causal Effects of Asthma on COVID-19 Susceptibility and Severity

“…Meanwhile, in order to quantitatively verify instrument strength, F-statistics for each instrumental variables individually and cumulatively were calculated via the formula F-statistic = R 2 × (SampleSize-2)/(1-R 2 ) ( Palmer et al, 2012 ). If instrumental variables with a F-statistic much than 10, the association was regarded as strong enough to avoid the weak instrument bias ( Sanderson et al, 2021 ).…”