8274751 Considerations for applying mendelian randomisation to shift work research

<h3>Objective</h3> Mendelian Randomisation (MR) is a research method that uses genetic markers, mainly single nucleotide polymorphisms (SNPs), as instrumental variables (IVs) to infer causality. MR requires three assumptions, one of which is that IVs must be independent of potential confounders. Several studies have been explored health effects from shift work as an exposure variable for MR.1-4 Conversely, body mass index and educational level are related to shift work in the MR analysis.5 This study shows the selection process of IVs for shift work and discusses its appropriateness. <h3>Material and Methods</h3> For the ‘Job involves shift work’ question in the UK Biobank, we listed SNPs that are highly (p &lt; 5×10-8) or moderately (p &lt; 5×10-6) associated with the question. To assess potential association with confounders, each SNP was cross-referenced with trait associations (p &lt; 1×10^-⁶) from a wide database. In the created data frame, the data was listed based on specific keywords related to covariates (e.g., body mass index, education) to identify the frequency of occurrence for each keyword. <h3>Results</h3> There were 55 SNPs associated with shift work with p &lt; 5×10-8 and 446 SNPs with p &lt; 5×10-6. After clumping, 2 and 39 SNPs remained, respectively. All 55 highly relevant SNPs for shift work were also associated with education. Of the 39 SNPs, 4 were associated with education and 6 with BMI. 25 SNPs had no association with other traits. <h3>Conclusion</h3> Prior to using MR for occupational and environmental health research, a rigorous review of instrumental variables should be carried out to ensure that the assumptions of MR are not violated. To maintain methodological robustness, rather than attempting to establish a comprehensive causal relationship with shift work itself, we suggest that Mendelian Randomisation may be better suited for exploring specific biological mechanisms that may underlie its health effects.