Exploration of errors in variance caused by using the first-order approximation in Mendelian randomization.

Area of Science:

• Genetics
• Epidemiology
• Statistical Genetics

Background:

• Mendelian randomization (MR) leverages genetic variants to infer causal relationships between exposures and outcomes.
• Two-sample MR (2SMR) enhances statistical power by using summary statistics from large genome-wide association studies (GWAS).
• Current 2SMR practices often rely on a first-order approximation for standard error calculations.

Purpose of the Study:

• To quantify the underestimation of variance caused by the first-order approximation in 2SMR.
• To evaluate the accuracy and robustness of the second-order approximation for standard error in 2SMR.

Main Methods:

• Simulated Mendelian randomization (MR) analyses were conducted.
• The variance of causal effects was compared between first-order and second-order standard error approximations.
• The accuracy of each approximation was assessed against simulated true variance.

Main Results:

• The first-order approximation was shown to significantly underestimate variance, in some cases by nearly 50%.
• This underestimation can lead to an elevated false-positive rate in causal inference studies.
• The second-order approximation demonstrated robustness and accuracy in correcting for this underestimation.

Conclusions:

• The standard first-order approximation in 2SMR is prone to underestimating variance, potentially compromising study validity.
• The second-order approximation provides a more reliable method for estimating standard errors in 2SMR.
• Adopting the second-order approximation is recommended for more accurate causal effect estimation in MR studies.