Methodological challenges in mendelian randomization

Abstract

We give critical attention to the assumptions underlying Mendelian randomization analysis and their biological plausibility. Several scenarios violating the Mendelian randomization assumptions are described, including settings with inadequate phenotype definition, the setting of time-varying exposures, the presence of gene-environment interaction, the existence of measurement error, the possibility of reverse causation, and the presence of linkage disequilibrium. Data analysis examples are given, illustrating that the inappropriate use of instrumental variable techniques when the Mendelian randomization assumptions are violated can lead to biases of enormous magnitude. To help address some of the strong assumptions being made, three possible approaches are suggested. First, the original proposal of Katan (Lancet. 1986;1:507-508) for Mendelian randomization was not to use instrumental variable techniques to obtain estimates but merely to examine genotype-outcome associations to test for the presence of an effect of the exposure on the outcome. We show that this more modest goal and approach can circumvent many, though not all, the potential biases described. Second, we discuss the use of sensitivity analysis in evaluating the consequences of violations in the assumptions and in attempting to correct for those violations. Third, we suggest that a focus on negative, rather than positive, Mendelian randomization results may turn out to be more reliable.

Figure 1

Diagrams illustrating a genetic variant (G) which is an instrument for the effect of exposure (X) on outcome (Y) when the exposure-outcome relationship is confounded by unmeasured factors (U)

Figure 2

Diagrams illustrating violations of the exclusion restriction with the genetic variant (G) not independent of outcome (Y) conditional on exposure (X) and confounders (U) because of a direct effect of G on Y not through X

Figure 3

Setting in which the variant (G) affects an intermediate (W) on the pathway to exposure (X), violating the exclusion restriction

Figure 4

Setting in which the exclusion restriction holds when (X₁, X₂) are taken together as the exposure but does not hold for X₁ alone.

Figure 5

Setting in which the exposure is time-varying and in which the exclusion restriction holds when (X₁, X₂) are taken together as the exposure but does not hold for X₁ alone.

Figure 6

Diagram illustrating that when the exposure (X) is measured with error (X*) the exclusion restriction will not in general hold for the mismeasured exposure X*

Figure 7

Setting in which prior outcome (Y₀) can affect subsequent exposure (X₁) so that the variant (G) is not independent of final outcome (Y₁) conditional on exposures (X₀,X₁) and confounders (U), violating MR assumptions, due to reverse causation

Figure 8

Setting in which variant (G) is associated with outcome (Y) conditional on exposure (X) and confounders (U), violating the MR assumptions, because of using data from a case-control study of another outcome (D)

Figure 9

Setting in which variant (G) is associated with outcome (Y) conditional on exposure (X) and confounders (U), violating MR assumptions, because of linkage disequilibrium with another variant (G_U) that affects the outcome.

Figure 10

Setting in which variant (G) is associated with outcome (Y) conditional on exposure (X) and confounders (U), violating MR assumptions, because of linkage disequilibrium with another variant (G_U) that affects the outcome, and in which the exclusion restriction is violated.