Reassessing Instrument Strength in Two-Sample Mendelian Randomization Analysis

Area of Science:

• Genetic Epidemiology
• Statistical Genetics
• Causal Inference

Background:

• Mendelian randomization (MR) is a key method for estimating causal effects using genetic variants.
• Two-sample MR, utilizing Genome-Wide Association Study (GWAS) summary statistics, is increasingly popular.
• Instrumental variable (IV) selection is crucial, with a trade-off between statistical power and potential bias.

Purpose of the Study:

• To investigate the impact of including weak genetic variants (IVs) in two-sample MR analyses.
• To evaluate the influence of pleiotropy and IV strength on causal effect estimates.
• To assess the role of exposure GWAS sample size in the reliability of MR findings with weak IVs.

Main Methods:

• Simulations were conducted to assess the effects of pleiotropy and weak IVs on MR estimates.
• Real data analyses utilized two versions of FinnGen GWAS summary statistics with varying sample sizes.
• The impact of including IVs with weaker association p-values on causal estimates was examined.

Main Results:

• Pleiotropy significantly increased the variability of causal effect estimates, even at modest levels.
• Inclusion of weak IVs generally did not substantially alter the direction or variability of causal estimates in simulations.
• In real data, weak IVs attenuated effect sizes, particularly in smaller exposure GWAS, risking false null findings.

Conclusions:

• The inclusion of weak IVs in MR analyses is context-dependent, influenced by exposure GWAS sample size.
• Weak IVs can be cautiously incorporated when exposure GWAS sample sizes are large.
• Caution is advised when using weak IVs with small exposure GWAS sample sizes due to the risk of biased, null results.