Caution against examining the role of reverse causality in Mendelian Randomization

Abstract

Recently, Mendelian Randomization (MR) has gained in popularity as a concept to assess the causal relationship between phenotypes in genetic association studies. An extension of standard MR methodology, the MR Steiger approach, has recently been developed to infer the causal direction between two phenotypes in prospective studies. Through simulation studies, we examined and quantified the ability of the MR Steiger approach to determine the causal direction between two phenotypes (i.e., effect direction). Through simulation studies, our results show that the MR Steiger approach may fail to correctly identify the direction of causality. This is true, especially in the presence of pleiotropy. We also applied the MR Steiger method to the COPDGene study, a case-control study of chronic obstructive pulmonary disease (COPD) in current and former smokers, to examine the role of smoking on lung function. We have created an R package on Github called reverseDirection which runs simulations for user-specified scenarios to examine when the MR Steiger approach can correctly determine the causal direction between two phenotypes in any user specified scenario. In summary, our results emphasize the importance of caution when the MR Steiger approach is used in to infer the direction of causality.

Keywords: Mendelian Randomization; causal direction; reverse causality.

Figure 1:

Directed acyclic graphs are shown for how the 8 sub scenarios were generated for the SNPs G, phenotype 1 X, phenotype 2 Y, unmeasured confounder U, and measurement error represented by ε: (A) no pleiotropy, no measurement error, and no unmeasured confounding of the two phenotypes (i.e. X,Y), (B) pleiotropy, no measurement error, and no unmeasured confounding, (C) no pleiotropy, measurement error, and no unmeasured confounding, (D) pleiotropy, measurement error, and no unmeasured confounding, (E) no pleiotropy, no measurement error, and unmeasured confounding, (F) pleiotropy, no measurement error, and unmeasured confounding, (G) no pleiotropy, measurement error, and unmeasured confounding, (H) pleiotropy, measurement error, and unmeasured confounding.

Figure 2:

For scenario 1 in the plots below in Figure 2A (sub scenarios A-D) and Figure 2B (sub scenarios E-H), the x-axis represents the values of β_X, the association between the two phenotypes in equation 2 if there was no pleiotropy generated and equation 4 if there was pleiotropy generated. Column 1 shows the proportion of simulations case 1–3 were concluded. Column 2 shows the proportion of simulations where pSteiger < α, pMR < α, and the test statistic Z > 0. As seen in the plots below, when there is no measurement error for phenotype 1, no unmeasured confounding, or no pleiotropy was generated (row 1 of Figure 2A), the power of the MR Steiger approach initially increases and then decreases as the association between the two phenotypes increases. When pleiotropy was generated, the MR Steiger approach concludes the wrong direction between the two phenotypes (i.e. case 2 is concluded when the data is generated under case 1)

Figure 2:

For scenario 1 in the plots below in Figure 2A (sub scenarios A-D) and Figure 2B (sub scenarios E-H), the x-axis represents the values of β_X, the association between the two phenotypes in equation 2 if there was no pleiotropy generated and equation 4 if there was pleiotropy generated. Column 1 shows the proportion of simulations case 1–3 were concluded. Column 2 shows the proportion of simulations where pSteiger < α, pMR < α, and the test statistic Z > 0. As seen in the plots below, when there is no measurement error for phenotype 1, no unmeasured confounding, or no pleiotropy was generated (row 1 of Figure 2A), the power of the MR Steiger approach initially increases and then decreases as the association between the two phenotypes increases. When pleiotropy was generated, the MR Steiger approach concludes the wrong direction between the two phenotypes (i.e. case 2 is concluded when the data is generated under case 1)