A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization

Area of Science:

• Epidemiological research
• Statistical genetics
• Biostatistics

Background:

• Mendelian randomization (MR) employs genetic data as instrumental variables (IVs) to investigate causal relationships in epidemiology.
• Two-sample summary data MR is increasingly common but susceptible to pleiotropy, where genetic variants violate IV assumptions.
• Detecting and correcting for pleiotropy is crucial for maintaining the validity of MR analyses.

Purpose of the Study:

• To clarify the adaptation of meta-analysis techniques for detecting and correcting pleiotropy in MR.
• To compare the Inverse Variance Weighted (IVW) method with MR-Egger regression in handling pleiotropy.
• To propose statistics for quantifying the goodness-of-fit between IVW and MR-Egger approaches.

Main Methods:

• Adaptation of meta-regression and random effects modeling from mainstream meta-analysis.
• Focus on two contrasting MR approaches: Inverse Variance Weighted (IVW) and MR-Egger regression.
• Investigation of random effects models for robustness to pleiotropy under the IVW approach.

Main Results:

• The study clarifies how meta-analysis methods are adapted to address pleiotropy in MR.
• It contrasts the assumptions and performance of IVW and MR-Egger regression.
• Statistics are proposed to assess the relative goodness-of-fit of these methods.

Conclusions:

• Established meta-analysis techniques can be adapted to detect and correct for pleiotropy in MR.
• The IVW and MR-Egger regression methods offer different ways to handle violations of IV assumptions.
• Quantifying goodness-of-fit is important for selecting appropriate methods in pleiotropic MR studies.