Causal Association Between Thyroid Function and Myeloproliferative Disease: A Two-Sample Mendelian Randomization Study.

BACKGROUND: Thyroid dysfunction has been linked to hematologic abnormalities, but whether thyroid function has a causal role in myeloproliferative diseases (MDs), including chronic myeloproliferative disease (CMD), remains unclear. We used Mendelian randomization (MR) to assess the potential causal effects of thyroid-related traits on MD risk.

METHODS: A two-sample MR analysis was performed using genome-wide association study (GWAS) summary statistics. Outcome data for CMD and MD, excluding chronic myelogenous leukemia (CML), were obtained from the Finnish R12 cohort (European ancestry). Genetic instruments for thyroid traits-including thyrotropin (TSH), free thyroxine (FT4), free/total triiodothyronine (FT3/TT3), FT3/FT4 and TT3/FT4 ratios, high/low TSH, and genetic liability to hyperthyroidism and hypothyroidism-were selected at genome-wide significance (p < 5 × 10-8), clumped for linkage disequilibrium (R 2 < 0.001, 10,000 kb), and required F-statistics > 10. Primary causal estimates were obtained using inverse-variance weighted (IVW) MR, with sensitivity analyses using MR-Egger, weighted median, weighted mode, MR-PRESSO, heterogeneity tests, and leave-one-out analyses. Steiger tests were applied to assess directionality, and false discovery rate (FDR) correction was used for multiple comparisons.

RESULTS: Most thyroid function measures (TSH, FT4, FT3, TT3, hormone ratios, and hypothyroidism) showed no evidence of a causal association with CMD or MD excluding CML (IVW p > 0.05 p > 0.05 p > 0.05). A nominal association was observed between self-reported hyperthyroidism/thyrotoxicosis and CMD, but the effect estimate was implausibly extreme (IVW OR = 5.53 × 10-14, 95% CI: 7.78 × 10-27-0.3935, p = 0.0432) and did not remain significant after FDR correction. After outlier removal, genetically predicted hyperthyroidism was nominally associated with MD excluding CML (OR = 0.8806, 95% CI: 0.7954-0.9750; p = 0.0144), but this also did not survive FDR correction (FDR = 0.2015). Sensitivity analyses and pleiotropy tests did not indicate strong directional pleiotropy for the nominal signals, and Steiger tests supported the hypothesized exposure-to-outcome direction.

CONCLUSIONS: In this MR study, we found no robust evidence for a causal association between most genetically predicted thyroid function traits and MD after multiple-testing correction. The nominal signals observed for hyperthyroidism were phenotype-dependent and statistically unstable, particularly for self-reported CMD. These results emphasize that previously observed associations may be driven by nongenetic factors or clinical confounding.Jingjing Xiang and Jun Yan were co-first authors.