Association of atopic dermatitis with autoimmune diseases: A bidirectional and multivariable two-sample mendelian randomization study

BACKGROUND: Observational studies have suggested the association between atopic dermatitis (AD) and the risks of autoimmune diseases. It is still unclear, however, whether or in which direction causal relationships exist, because these associations could be confounded. OBJECTIVES: Our study seeks to assess the possibility of AD as a cause of autoimmune diseases, and to estimate the magnitude of the causal effect. METHODS: Two-sample mendelian randomization (MR) analyses were performed using genome-wide association study (GWAS) summary-level statistics. Specifically, bidirectional MR analyses were conducted to examine the direction of association of AD with autoimmune diseases; multivariable MR analyses (MVMR1) were used to test the independence of causal association of AD with autoimmune diseases after controlling other atopic disorders (asthma and allergic rhinitis), while MVMR2 analyses were conducted to account for potential confounding factors such as smoking, drinking, and obesity. Genetic instruments for AD (Ncases=22 474) were from the latest GWAS meta-analysis. The GWAS summary data for asthma and allergic rhinitis were obtained from UK Biobank. The GWAS summary data for smoking, alcohol consumption, obesity and autoimmune diseases (alopecia areata, vitiligo, systemic lupus erythematosus, ankylosing spondylitis, rheumatoid arthritis, and type 1 diabetes) were selected from the largest GWASs available. Causal estimates were derived by the inverse-variance weighted method and verified through a series of sensitivity analyses. RESULTS: Genetically predicted AD linked to higher risks of rheumatoid arthritis (OR, 1.28; P=0.0068) (OR(MVMR1), 1.65; P=0.0020) (OR(MVMR2), 1.36; P<0.001), type 1 diabetes (OR, 1.37; P=0.0084) (OR(MVMR1), 1.42; P=0.0155) (OR(MVMR2), 1.45; P=0.002), and alopecia areata (OR, 1.98; P=0.0059) (OR(MVMR1), 2.55; P<0.001) (OR(MVMR2), 1.99; P=0.003) in both univariable and multivariable MR. These causal relationships were supported by sensitivity analyses. No causal effect of AD was identified in relation to systemic lupus erythematosus, vitiligo, and ankylosing spondylitis. Concerning the reverse directions, no significant association was noted. CONCLUSION: The results of this MR study provide evidence to support the idea that AD causes a greater risk of rheumatoid arthritis, type 1 diabetes and alopecia areata. Further replication in larger samples is needed to validate our findings, and experimental studies are needed to explore the underlying mechanisms of these causal effects.