Investigating the Causal Relationship Between Sleep-Related Traits and Self-Reported Diabetes: A Mendelian Randomization Study

Abstract

Objective: Self reported data can be a valuable resource for understanding health outcomes, behaviors, disease prevalence, and risk factors, yet underutilized in epidemiological research. While observational studies have linked sleep traits with diabetes, evidence using self reported diabetes data for causal connection is lacking. Methods: We performed a two sample Mendelian randomization (MR) analysis using Inverse Variance Weighting (IVW), IVW with multiplicative random effects (IVW MRE), Maximum Likelihood (ML), MR-Egger regression, and Weighted Median models, with genetic variants linked to five sleep traits (sleep duration, insomnia, snoring, daytime dozing, and chronotype) and self reported diabetes from the UK Biobank dataset. The study utilized MR Egger and MR PRESSO regression to evaluate pleiotropy and outliers, IVW Q statistics to detect heterogeneity, the MR Steiger test to assess directionality, and leave one out sensitivity analysis to ensure the reliability. Results: ML provided positive causal associations between genetically predicted insomnia (p = 0.002, OR = 1.021, 95% CI: 1.008 to 1.035) and daytime dozing (p = 0.014, OR = 1.029, 95% CI: 1.006 to 1.052) with diabetes, while IVW and IVW-MRE analysis showed a trend towards significance. Snoring showed mixed evidence, while genetically predicted sleep duration was marginally associated with diabetes (p = 0.053, OR = 0.992, 95% CI: 0.984 to 1.000) with the weighted median method, indicating a potential small protective effect. No causal association was found between chronotype and diabetes. Conclusion: This exploratory MR study provides evidence for the effect of insomnia, daytime dozing, sleep duration and snoring on diabetes risk. These findings underscore the importance of considering self reported health outcomes in epidemiological research.