Horizontal Pleiotropy

The core defining feature of horizontal pleiotropy is the direct violation of the exclusion restriction (IV3) assumption. This occurs when a genetic instrument (SNP) influences the outcome through a biological pathway that does not pass through the exposure of interest. For example, a variant in the FTO gene used as an instrument for BMI may also independently affect smoking behavior, creating a spurious association between BMI and lung cancer that is not causal. This renders the instrumental variable estimate inconsistent and biased.

It is crucial to distinguish horizontal pleiotropy from vertical pleiotropy (also known as mediated pleiotropy). Vertical pleiotropy is a natural part of the causal chain and does not violate MR assumptions. In vertical pleiotropy, the genetic variant affects the exposure, which in turn affects a downstream outcome (e.g., SNP → LDL Cholesterol → Myocardial Infarction). Horizontal pleiotropy represents an alternative, independent path (e.g., SNP → Blood Pressure → Myocardial Infarction, bypassing LDL).

Horizontal pleiotropy can be categorized by its net effect on causal estimates:

• Directional Pleiotropy: The pleiotropic effects systematically bias the causal estimate in a specific direction (away from the null). This is detected by methods like MR-Egger regression where the intercept term is non-zero.
• Balanced Pleiotropy: The pleiotropic effects are independent of the instrument-exposure association and cancel each other out across multiple instruments. In this case, the Inverse-Variance Weighted (IVW) estimate remains consistent, though heterogeneity will be inflated.

The presence of horizontal pleiotropy induces heterogeneity among the causal estimates derived from individual genetic variants. This is formally tested using:

• Cochran's Q statistic: A significant Q value suggests that the variability between variant-specific causal estimates is greater than expected by chance alone, a hallmark of pleiotropy.
• I² statistic: Quantifies the percentage of total variation across instruments attributable to heterogeneity rather than sampling error. High I² values signal potential pleiotropic bias.

Specific methods identify and remove individual pleiotropic outliers to salvage a valid causal estimate:

• MR-PRESSO (Pleiotropy RESidual Sum and Outlier): This method systematically tests each SNP for significant deviation from the expected causal effect, removes detected outliers, and provides a corrected causal estimate.
• Radial MR: A visualization and formal statistical framework that plots the contribution of each variant to the overall estimate, making it easy to identify variants with large residuals that drive heterogeneity.

Several statistical approaches have been developed to provide valid causal inference even in the presence of horizontal pleiotropy:

• Weighted Median Estimator: Provides a consistent estimate if at least 50% of the weight in the analysis comes from valid instruments, making it robust to a minority of pleiotropic variants.
• MR-Egger Regression: Allows for directional pleiotropy by fitting an unconstrained intercept, though it requires the InSIDE (Instrument Strength Independent of Direct Effect) assumption to hold.
• Contamination Mixture Method: Models the genetic variants as a mixture of valid and invalid instruments, using robust likelihood-based methods to estimate the causal effect.

An instrumental variable analysis method that uses genetic variants as proxies for modifiable exposures to estimate causal effects on health outcomes. MR relies on three core assumptions: the genetic instrument must be associated with the exposure (relevance), must not be associated with confounders (independence), and must affect the outcome only through the exposure (exclusion restriction). Horizontal pleiotropy represents a direct violation of this third assumption.

A pleiotropy-robust Mendelian randomization method that allows for directional horizontal pleiotropy by fitting a weighted linear regression with an unconstrained intercept term. The intercept provides a formal test for directional pleiotropy: a non-zero intercept indicates systematic bias. The slope yields a causal estimate corrected for this bias, though at the cost of reduced statistical power compared to IVW.

The Mendelian Randomization Pleiotropy RESidual Sum and Outlier test. This method detects and corrects for horizontal pleiotropy by:

• Identifying outlier genetic variants whose causal estimates deviate significantly from the consensus
• Removing these variants and recalculating the causal effect
• Providing a global pleiotropy test statistic It is particularly effective when a minority of instruments exhibit pleiotropic effects.

An extension of MR that estimates the direct causal effect of multiple correlated exposures on an outcome simultaneously. MVMR accounts for shared genetic architecture by including genetic variants associated with any of the exposures. This method can mitigate horizontal pleiotropy when the pleiotropic pathway operates through a measured, included covariate, effectively conditioning it out of the primary exposure estimate.

A statistical method to assess whether two potentially related traits share the same causal genetic variant at a given genomic locus. In the context of pleiotropy, colocalization helps distinguish between:

• Horizontal pleiotropy: one variant independently affects two traits
• Vertical pleiotropy: one variant affects an exposure which causally affects the outcome
• Linkage disequilibrium: two distinct variants are correlated

A bias in instrumental variable analysis that occurs when genetic variants are only weakly associated with the exposure (typically F-statistic < 10). Weak instruments amplify the impact of even minor violations of the exclusion restriction, including horizontal pleiotropy, leading to inflated and unreliable causal estimates that are biased toward the confounded observational association in two-sample MR designs.