A review of instrumental variable estimators for Mendelian randomization

Area of Science:

• Epidemiology
• Biostatistics
• Genetics

Background:

• Observational data often contains confounding factors, making causal inference challenging.
• Instrumental variable (IV) analysis offers a method to estimate causal effects by leveraging specific assumptions.
• Mendelian randomization (MR), utilizing genetic variants as IVs, has emerged as a powerful tool in recent years.

Purpose of the Study:

• To provide a comprehensive overview of instrumental variable analysis methods.
• To discuss techniques for statistical inference and confidence interval construction.
• To compare the statistical properties of various IV estimation methods, with a focus on Mendelian randomization.

Main Methods:

• The paper reviews several IV estimation techniques, including the ratio method, two-stage methods, likelihood-based methods, and semi-parametric methods.
• It details methods for obtaining statistical inferences and confidence intervals.
• Special attention is given to bias and coverage properties, particularly concerning weak instruments.

Main Results:

• The study compares the statistical properties of different IV estimators.
• It provides practical guidance on selecting appropriate analysis methods based on study design and data characteristics.
• The paper prioritizes settings relevant to Mendelian randomization, such as continuous exposures with continuous or binary outcomes.

Conclusions:

• Instrumental variable analysis, especially Mendelian randomization, is a valuable approach for causal inference in observational studies.
• Understanding the properties and limitations of different IV methods is crucial for reliable causal effect estimation.
• The paper equips researchers with the knowledge to choose suitable methods and interpret results effectively, particularly in genetic epidemiology.