Testing for a Causal Role of Thyroid Hormone Measurements Within the Normal Range on Human Metabolism and Diseases: a Systematic Mendelian Randomization

Background Variation in thyroid function parameters within the normal range has been observationally associated with adverse health outcomes. Whether those associations reflect fl ect causal effects is largely unknown. Methods We systematically tested associations between genetic differences in thyrotropin (TSH) and free thyroxine (FT4) within the normal range and more than 1100 diseases and more than 6000 molecular traits (metabolites and proteins) in three large population-based cohorts. This was performed by combining individual and summary level genetic data and using polygenic scores and Mendelian randomization (MR) methods. We performed a phenomewide MR study in the OpenGWAS database covering thousands of complex phenotypes and diseases. Findings Genetically predicted TSH or FT4 levels within the normal range were predominately associated with thyroid-related outcomes, like goitre. The few extra-thyroidal outcomes that were found to be associated with genetic liability towards high but normal TSH levels included atrial fi brillation (odds ratio = 0.92, p-value = 2.13 x 10-3), - 3 ), thyroid cancer (odds ratio = 0.57, p-value = 2.97 x 10-4), - 4 ), and specific fi c biomarkers, such as sex hormone binding globulin ((3 (3 = - 0.046, p-value = 1.33 x 10-6) - 6 ) and total cholesterol ((3 (3 = 0.027, p-value = 5.80 x 10-3). - 3 ). Interpretation In contrast to previous studies that have described the association with thyroid hormone levels and disease outcomes, our genetic approach fi nds little evidence of an association between genetic differences in thyroid function within the normal range and non-thyroidal phenotypes. The association described in previous studies may be explained by reverse causation and confounding.