MR-SPLIT: a novel method to address selection and weak instrument bias in one-sample Mendelian randomization studies

Ruxin Shi,Ling Wang,Stephen Burgess,Yuehua Cui

DOI: https://doi.org/10.1101/2024.02.11.579683

2024-02-12

Abstract:Mendelian Randomization (MR) is a widely embraced approach to assess causality in epidemiological studies. Two-stage least squares (2SLS) method is a predominant technique in MR analysis. However, it can lead to biased estimates when instrumental variables (IVs) are weak. Moreover, the issue of the winner’s curse could emerge when utilizing the same dataset for both IV selection and causal effect estimation, leading to biased estimates of causal effects and high false positives. Focusing on one-sample MR analysis, this paper introduces a novel method termed as Mendelian Randomization with adaptive Sample-sPLitting with cross-fitting InstrumenTs (MR-SPLIT), designed to address bias issues due to IV selection and weak IVs, under the 2SLS IV regression framework. We show that the MR-SPLIT estimator is more efficient than its counterpart cross-fitting MR (CFMR) estimator. Additionally, we introduce a multiple sample-splitting technique to enhance the robustness of the method. We conduct extensive simulation studies to compare the performance of our method with its counterparts. The results underscored its superiority in bias reduction, effective type I error control, and increased power. We further demonstrate its utility through the application of a real-world dataset. Our study underscores the importance of addressing bias issues due to IV selection in high dimensions and weak IVs in one-sample MR analyses and provides a robust solution to the challenge.

Genetics

What problem does this paper attempt to address?

### The Problem the Paper Attempts to Solve This paper aims to address the bias issues caused by instrument variable selection and weak instrument variables in single-sample Mendelian Randomization (MR) studies. Specifically: 1. **Instrument Variable Selection Bias**: When the same dataset is used for both instrument variable selection and causal effect estimation, "winner’s curse" may occur, leading to increased bias in causal effect estimation and a higher false positive rate. 2. **Weak Instrument Variable Bias**: When instrument variables are weak, traditional two-stage least squares (2SLS) may produce biased estimates, especially in the presence of multiple instrument variables. To tackle these issues, the paper proposes a new method—Mendelian Randomization with adaptive Sample-sPLitting with cross-fitting InstrumenTs (MR-SPLIT). This method improves existing MR analysis methods in the following ways: - **Adaptive Selection of Primary and Weak Instrument Variables**: Instrument variables are divided into primary and weak instrument variables based on their association strength with the exposure variable, and weak instrument variables are combined into a composite instrument variable. - **Multiple Sample Splitting Strategy**: By repeatedly randomly splitting the dataset, the robustness of the estimates is improved. - **Cross-Fitting**: By selecting instrument variables and estimating causal effects in different subsamples, the bias from using the same dataset for selection and estimation is avoided. The paper demonstrates the superior performance of the MR-SPLIT method in reducing bias, effectively controlling the type 1 error rate, and improving statistical power through extensive simulation studies and real data applications.

MR-SPLIT: a novel method to address selection and weak instrument bias in one-sample Mendelian randomization studies

Approximation of Bias and Mean-Squared Error in Two-Sample Mendelian Randomization Analyses.

On Mendelian Randomization Mixed-Scale Treatment Effect Robust Identification (MR MiSTERI) and Estimation for Causal Inference

Selecting invalid instruments to improve Mendelian randomization with two-sample summary data

Robust instrumental variable methods using multiple candidate instruments with application to Mendelian randomization

A Novel One‐Sample Mendelian Randomization Approach for Count‐Type Outcomes That Is Robust to Correlated and Uncorrelated Pleiotropic Effects

Breaking the Winner's Curse in Mendelian Randomization: Rerandomized Inverse Variance Weighted Estimator

MR-BOIL: Causal Inference in One-Sample Mendelian Randomization for Binary Outcome with Integrated Likelihood Method.

Inference After Selecting Plausibly Valid Instruments with Application to Mendelian Randomization

Leveraging Local Distributions in Mendelian Randomization: Uncertain Opinions are Invalid

A Versatile and Efficient Novel Approach for Mendelian Randomization Analysis with Application to Assess the Causal Effect of Fetal Hemoglobin on Anemia in Sickle Cell Anemia

A new robust and accurate two-sample Mendelian randomization method with a large number of genetic variants

Overlapping-sample Mendelian randomisation with multiple exposures: a Bayesian approach

Debiased Estimating Equation Method for Versatile and Efficient Mendelian Randomization Using Large Numbers of Correlated Weak and Invalid Instruments

Mendelian Randomization With Refined Instrumental Variables From Genetic Score Improves Accuracy and Reduces Bias

LASSO-type instrumental variable selection methods with an application to Mendelian randomization

Winner's Curse Free Robust Mendelian Randomization with Summary Data

A latent outcome variable approach for Mendelian randomization using the expectation maximization algorithm

A Random Effects Model-based Method of Moments Estimation of Causal Effect in Mendelian Randomization Studies

Unbiased estimation and asymptotically valid inference in multivariable Mendelian randomization with many weak instrumental variables

TwoStepCisMR: A Novel Method and R Package for Attenuating Bias in cis-Mendelian Randomization Analyses