Tehran University Medical Journal (Jan 2023)
Mendelian randomization, a method for inferring causal relationships in observational studies and an alternative to clinical trial studies: a brief report
Abstract
Background: Mendelian randomization (MR) is a new generation in the statistical method that uses genetic variants as instrumental variables in data from non-experimental studies to evaluate and estimate the causal effects of risk factors. Methods: The weakness of observational studies to detect causality, the difficulties of conducting clinical trials, the dramatic advancement of Genome-Wide Association Studies (GWAS) have led to the emergence of a new type of study called MR. It is increasingly being used to determine causality MR is an approach based on meta-analysis methods. The main idea of the MR is based on using the instrument variable (IV) to find the causality between exposure and outcome. This variable does not need to adjust the confounding effects found in observational studies. Results: Data for this study were collected from the beginning of January 2003 to October 2020 in PubMed. Our results showed that MR has an increasing trend. The data used in MR includes summarized statistical data, individual-level data, and meta-analysis. Choosing the suitable IV is essential to successfully conduct an MR. For an unbiased estimate, three main hypotheses should be considered: 1) The IV has a strong relationship with the desired exposure (i.e., potential risk factor), 2) The IV is not related to the confounding variable, and 3) The IV is not directly related to the outcome and should only relate to the outcome through exposure. If these conditions are not met, one solution is to use robust methods. Besides, this research introduced the study designs, estimation methods, limitations, software packages, and some applications of MR in medical research. Conclusion: When we seek to find a causal relationship, but it is not possible to use a clinical trial as a standard method, the MR design can be used in observational studies. Therefore, it is possible to obtain causal relationships between exposure and outcome using the MR.
