Abstract
KEY POINTS
This is a large retrospective cohort analysis of adults ages 25-84 years with outcomes of all-cause mortality and cardiovascular mortality.
Overt hypothyroidism was associated with increased all-cause and cardiovascular mortality after multivariable adjustment.
Subclinical hypothyroidism was not associated with increased mortality, but it was associated with all-cause mortality among individuals younger than 50 years in age-stratified analyses.
Interpretation is limited by multiple sources of selection bias, incomplete exposure characterization, and insufficient details about the a priori hypothesis and analysis plan.
SUMMARY
Background
Overt hypothyroidism (OH) is associated with adverse cardiovascular outcomes. 1 Studies on subclinical hypothyroidism (SCH) and these end points have been heterogeneous and vary in the risk reported both by degree of thyrotropin (TSH) elevation and patient age. No primary randomized data directly address these end points. Accordingly, treatment thresholds for SCH and whether they vary by population remain an area of ongoing debate.
The Thyroid Studies Collaborative individual patient data level meta-analysis (n > 55,000) found an increased risk of coronary heart disease events and mortality in patients with SCH who had TSH levels ≥10 mIU/L. 2 Other meta-analyses have found an association between ischemic heart disease prevalence and SCH at lower TSH levels only in cohorts that include adults younger than 65 years. 3 Combined secondary analysis of the TRUST and IEMO80 + trials did not demonstrate a mortality benefit from treatment in older adults. 4 However, observational data suggest that treatment of SCH with levothyroxine may reduce ischemic cardiac events in younger individuals (40-70 years). 5 In this study, 6 Ran et al. examine the association between hypothyroidism and mortality using a very large retrospective cohort from China that includes younger adults.
Methods
This retrospective cohort included adults ages 25-84 years undergoing standardized routine health examinations at a tertiary center in China (2017-2022). Of 116,267 available individuals, 42,051 were excluded for incomplete thyroid testing (missing TSH, free thyroxine [FT4], or free triiodothyronine [FT3]), and 487 were excluded for loss to follow-up before December 31, 2024. Additional exclusions included patient-reported thyroid disease or medication use (n = 502), low TSH (n = 1772), or normal TSH with low FT3 (n = 1179).
Thyroid status was defined using population-based reference ranges for TSH (0.60-4.90 mIU/L) and FT4 (8.50-14.50 pmol/L). Mortality outcomes were obtained from national death registries. The primary outcome was all-cause mortality; cardiovascular mortality was secondary.
Multivariable Cox proportional-hazards models adjusted for age, sex, body mass index, hypertension, dyslipidemia, and diabetes. Analyses were stratified by age (<50, 50-70, >70 years) and diabetes status.
Results
The final analytic cohort included 70,276 participants, of whom 66,987 (95.3%) were euthyroid, 2436 (3.5%) had SCH, and 853 (1.2%) had OH. Median follow-up was 5.1 years (range, 3.6-6.0). Participants with SCH and OH were older (45 and 48.7 vs. 42 years) and more often female (62% and 66.7% vs. 46.6%) as compared to euthyroid participants. There were also significant differences across the groups for the prevalence of hypertension, diabetes, and dyslipidemia. Mean (±SD) TSH levels across the three categories were reported as 2.1 ± 0.9 mIU/L for euthyroid, 6.3 ± 1.3 mIU/L for SCH, and 8.7 ± 6.0 mIU/L for OH.
There were 329 deaths in euthyroid individuals, 16 in SCH individuals, and 14 in OH individuals. After adjustment, OH was associated with increased all-cause mortality (HR, 2.01; 95% CI, 1.17-3.45) and cardiovascular mortality (HR, 2.70; 95% CI, 1.18-6.19). When stratified by disease status, all-cause mortality continued to be associated with OH among participants with diabetes (HR, 5.45; 95% CI, 2.46-12.07) but not among those without diabetes (HR, 1.19; 95% CI, 0.56-2.54).
SCH was not associated with either outcome. However, among individuals < 50 years of age, SCH was associated with higher all-cause mortality (HR, 3.28; 95% CI, 1.31-8.23), while those ages 50-65 years had a lower, nonsignificant risk (HR, 0.32; 95% CI, 0.08-1.31). Finally, in the analysis of SCH with TSH 4.9-10 mIU/L, there was a similar HR for those younger than 50 and no significant risk for those 50 years and older, but the sample size for this analysis was not reported.
Conclusions
In this large retrospective cohort, OH as defined by a single elevated TSH and low FT4 was associated with increased all-cause and cardiovascular mortality. SCH was associated with increased mortality only in the subanalysis of those < 50 years, including for TSH < 10 mIU/L.
COMMENTARY
This large observational study of a relatively young population in China adds a novel cohort to the data on the relationship between hypothyroidism and mortality. These data are largely consistent with other findings. However, methodological limitations, several inherent in the study design, impact the interpretation of the reported associations.
First, as with any retrospective study, depending on clinically defined inclusion criteria, substantial selection bias may be present. More than one third of the potential participants were excluded owing to incomplete thyroid testing. Individuals without incomplete testing may differ from the analytical cohort, particularly if testing was prompted by clinical symptoms or perceived illness rather than per protocol.
Second, exposure was incompletely characterized. Despite verified follow-up, levothyroxine initiation was not captured or modeled as a time-varying exposure. Thus, it is not known whether observed associations reflect untreated disease or treatment effects. In addition, thyroid status was defined with a single measurement, which creates classification bias; up to a third of mild TSH elevations will normalize on repeat testing. This can result in bias toward the null if people who do not have SCH are wrongly characterized or away from the null if transient elevations occur more commonly among participants with other risk factors who are therefore more likely to have poor outcomes. Finally, mean TSH was only 8.7 mIU/L with a large standard deviation in the OH group compared to 6.3 mIU/L in the SCH group, which raises questions about the accuracy of diagnostic categorization.
The interpretation of significance in this study also needs to be evaluated in the context of performing multiple subgroup analyses, particularly as an a priori justification for the diabetes stratification is not provided. Based on prior literature, the other covariables used, particularly lipids, 7 would have been more likely targets for a stratified analysis. And if these were performed and not reported, then the number of analyses is higher than reflected in the article. The risk of false-positive findings increases with the number of analyses performed and is especially relevant when the findings are of marginal significance.
These limitations highlight the challenges inherent in trying to understand causal relationships between complex exposures and broad composite end points like mortality by the proximal study of associations in observational cohorts, even large ones. Although the findings of Ran et al. are broadly consistent with prior literature, incomplete exposure characterization and multiple sources of residual confounding constrain interpretation. Further studies incorporating longitudinal thyroid measurements and treatment exposure will be necessary to clarify these associations in the absence of powered randomized trials.