Context: Thyroid-stimulating hormone (TSH), as the most sensitive and specific marker of thyroid status, is associated with multiple health outcomes, including mortality. However, whether TSH levels are causally associated with the risk of mortality remains unclear.
Objective: This study aims to investigate the causal association between TSH levels and all-cause mortality using Mendelian randomization (MR) analyses.
Methods: MR analyses using single-nucleotide polymorphisms (SNPs) associated with TSH levels (P < 5 × 10-8) as instruments. Mortality data were obtained from the UK Biobank, including 384 344 participants who were recruited from 22 assessment centers across the UK taken between 2006 and 2010. Cox proportional hazards regression was used to estimate the association of the TSH genetic risk score (GRS) with all-cause and cause-specific mortality.
Results: 15 557 individuals died during a median of 9.00 years of follow-up in the UK Biobank. A total of 70 SNPs were included in the MR analysis. The main MR analyses showed that 1 SD increase in TSH was associated with a decreased risk of all-cause mortality (OR 0.972, 95% CI 0.948-0.996), which may be largely attributed to respiratory disease mortality (OR 0.881, 95% CI 0.805-0.963). The multivariable hazard ratios (HRs) (95% CI) of all-cause mortality across 3 TSH GRS categories were 1.00 (reference), 0.976 (0.940-1.014), and 0.947 (0.911-0.985), respectively (P for trend < .01). Moreover, except digestive diseases mortality, genetically predicted TSH levels were negatively associated with mortality from CVD, cancer, noncancer diseases causes, and dementia, although not statistically significant.
Conclusion: Higher TSH levels were causally associated with lower risk of all-cause mortality, which may be largely attributed to respiratory disease mortality.
Keywords: Mendelian randomization; genetic variants; mortality; thyroid-stimulating hormone.
© The Author(s) 2023. Published by Oxford University Press on behalf of the Endocrine Society. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.