Credible Mendelian randomization studies: approaches for evaluating the instrumental variable assumptions

Area of Science:

• Epidemiology
• Genetics
• Biostatistics

Background:

• Mendelian randomization (MR) studies, a type of instrumental variable (IV) analysis, rely on strong assumptions.
• These assumptions are often not systematically evaluated in MR applications, potentially impacting study credibility.

Purpose of the Study:

• To present methods for systematically evaluating the validity of MR studies.
• To apply these methods to an MR study investigating the effect of obesity on mental disorder using FTO genotype as an IV.
• To describe the assumptions underlying these IV assessment methods.

Main Methods:

• The authors detail several methods for assessing the validity of instrumental variable assumptions in MR studies.
• These methods were applied to a specific MR study examining the relationship between obesity and mental disorder.
• The assessment techniques are applicable to any IV analysis, including those using genetic or other sources of exogenous variation.

Main Results:

• The presented methods can help evaluate the validity of MR studies, though they are not infallible.
• The application to an FTO-based MR study demonstrated the utility of these assessment techniques.
• The evaluation approaches may sometimes fail to detect biased IVs or incorrectly flag valid IVs as biased.

Conclusions:

• Routinely applying methods to assess MR assumptions is crucial for improving the scientific rigor and contributions of MR studies.
• While not conclusive, these assessment methods enhance the credibility of IV analyses.
• The described techniques are broadly relevant for any instrumental variable analysis.