Subclinical Hypothyroidism and Cardiovascular Risk: How to End the Controversy

The indications for screening and TSH threshold levels for treatment of subclinical hypothyroidism have remained a clinical controversy for over 20 years. Subclinical thyroid dysfunction is a common finding in the growing population of older adults, occurring in 10–15% among those age 65 and older, and may contribute to multiple common problems of older age, including cardiovascular disease, muscular impairment, mood problems, and cognitive dysfunction (1). In 2004, both the U.S. Preventive Services Task Force (2) and a clinical consensus group of experts (3) concluded that the existing evidence about the association between subclinical hypothyroidism and cardiovascular risks, primarily cross-sectional or case-control studies (4), was insufficient. For example, a frequently cited analysis from the Rotterdam study found a cross-sectional association between subclinical hypothyroidism and atherosclerosis, as measured by abdominal aortic calcification (odds ratio, 1.7; 95% confidence interval [CI], 1.1–2.6) and prevalent myocardial infarction (MI) (odds ratio, 2.3; 95% CI, 1.3–4.0) (5). Conversely, the prospective part of this study included only 16 incident MIs; the hazard ratio (HR) for subclinical hypothyroidism was 2.50, with broad 95% CIs (0.70–9.10). Potential mechanisms for the associations with cardiovascular diseases among adults with subclinical hypothyroidism include elevated cholesterol levels, inflammatory markers, raised homocysteine, increased oxidative stress, insulin resistance, increased systemic vascular resistance, arterial stiffness, altered endothelial function, and activation of thrombosis and hypercoagulability that have all been reported to be associated with subclinical hypothyroidism (1, 6).

Since 2004, several large prospective cohorts investigated this issue, leading to conflicting data on the associations between subclinical hypothyroidism and cardiovascular risk (7–9). Study-level meta-analyses tried to clarify these conflicting data (9, 10), and found modestly increased risks for coronary heart disease (CHD) and CHD mortality, but with heterogeneity among individual studies that used different TSH cutoffs, different confounding factors for adjustment, and varying CHD definitions (9). To help clarify this issue, we formed the Thyroid Studies Collaboration, which successfully collected individual participant data from 55 287 participants with 542 494 person-years of follow-up from 11 international prospective cohorts. In a series of individual participant data analyses, generally considered the highest level of nonrandomized evidence (11), we found that subclinical hypothyroidism was associated with an increased risk of CHD events and CHD mortality in adults with higher TSH levels (12) and with an increased risk of heart failure events (13), particularly when TSH exceeds 10 mU/L.

A particular shortcoming of the available evidence to date is the lack of randomized clinical trials on relevant clinical outcomes (2, 3). Small, short-term trials among individuals with subclinical hypothyroidism have demonstrated that low-density lipoprotein cholesterol (3) and several surrogate measurements for CHD, such as carotid intima-media thickness (6) and brachial artery endothelial function (14), improved with thyroid hormone replacement, but no randomized clinical trial with clinical cardiovascular outcomes has been published.

In this context, a nested case-cohort study in this issue of the JCEM by LeGrys et al (15) from the Women's Health Initiative (WHI) Observational Cohort is of interest. They examined the association between subclinical hypothyroidism and incident MI among postmenopausal women by measuring TSH and free T₄ on archived serum from 736 cases with incident MI and 2927 randomly selected subcohort members. After multivariate adjustment and exclusion of those with overt thyroid disease, the HR for MI was 0.99 (95% CI, 0.67–1.46) for TSH 4.69–6.99 mU/L, and 1.19 (95% CI, 0.72–1.96) for TSH ≥ 7.0 mU/L. The presence of positive thyroid peroxidase antibodies did not lead to increased risk among those with TSH ≥ 7.0 mU/L. The study has several strengths. Cases of MI, clearly an important clinical endpoint, were rigorously collected and adjudicated. The case-cohort sampling was performed within a well-executed cohort study with excellent follow-up. The study also has some limitations. Separate analyses among those with TSH ≥ 10.0 mU/L, perhaps the group of greatest interest, were not reported because of the small number with such TSH levels (n = 36). Although participants were followed for 5–10 years, information about the use of thyroid medication was not available beyond 3 years of follow-up and was missing for about 23% of the study population. Previous studies have found that accounting for initiation of replacement therapy during follow-up may increase the risk estimates, particularly for elevated TSH levels (12, 13). In addition, thyroid function testing was performed only at baseline, and we have no information on how many participants normalized their TSH over time, which is a limitation of most published large cohorts (7, 8, 12) and might lead to underestimating the true risks of persistent subclinical hypothyroidism. Surprisingly, a recent study found no increased cardiovascular risk for persistent subclinical hypothyroidism among the elderly (16).

How can we explain these inconsistent data? Risks might differ according to gender (only women in the current study), race (85% white race), and pre-existing cardiovascular disease (excluded), although such interactions were not detected in the individual participant data analysis that included 55 000 participants (12). Some have hypothesized that increasing age may attenuate the risk of cardiovascular disease with subclinical hypothyroidism (1), which is consistent with some (9, 10), but not all studies (12). However, the WHI results may not be as inconsistent as they first appear. When looking at CIs and not at statistical significance, results of the current study and the published individual participant data analysis (12) are both consistent, with no clear increased cardiovascular risk for elevated TSH < 7.0 mU/L and a moderate increased risk with higher TSH levels, particularly among those with TSH ≥ 10.0 mU/L (not separately analyzed in the current study). In the current study, adjusted HRs for risk of MI were 1.19 (95% CI, 0.72–1.96) for TSH ≥ 7.0 mU/L, whereas our Thyroid Studies Collaboration found HRs for risk of CHD events of 1.17 for TSH 7.0–9.9 mU/L (95% CI, 0.96–1.43) and 1.89 for TSH ≥ 10 mU/L (95% CI, 1.28–2.80; p for trend across TSH levels < .001) (12). When there is no clear biological rationale for specific cut-points, such as TSH values above 7 or 10 mU/L, tests of trend may provide a clearer picture of the underlying relationships but were not reported in the WHI analyses.

What are the clinical and research implications of current data? The finding of no significantly increased risk of CHD among the high proportions of adults with minimal TSH elevations is important because many patients with minimal TSH elevations are treated in clinical practice (17). The threshold TSH to define and treat subclinical hypothyroidism remains controversial (2, 3, 18). One argument for therapy at low TSH thresholds is to prevent progression to overt hypothyroidism. Counterarguments are the lifelong daily therapy in asymptomatic individuals, the potential risk of overtreatment, and most importantly, the lack of data from randomized trials showing benefits on clinical outcomes. Given the current lack of such clinical trial data, it might be argued that treatment thresholds should rather be defined according to potential clinical risks. Given current available data, threshold for cardiovascular risk seems to be a TSH ≥ 7.0 or ≥ 10.0 mU/L, whereas future studies should clarify the TSH thresholds discussed for other potential risks (mood, cognition, muscular strength…).

How will this long-standing controversy be solved? Additional observational data are unlikely to adequately address the uncertainties surrounding the risks associated with subclinical hypothyroidism or the potential benefits and harm of replacement therapy. Such concerns are best addressed in an appropriately powered randomized controlled trial with clinical outcomes. The European Commission has now funded such a trial, the Thyroid Hormone Replacement for Untreated Older Adults with Subclinical Hypothyroidism (TRUST). The TRUST study is a multicenter, double-blinded, placebo-controlled randomized trial designed to assess the multimodal effects of thyroid hormone replacement among 3000 adults age 65 and older with persistent subclinical hypothyroidism (European Commission FP7-Health-2011, Specific Programme “Cooperation” – Theme “Health”, Proposal No. 278148-2; clinicaltrials.gov, NCT01660126) (19, 20). The trial is currently recruiting older adults in four countries (Scotland, Ireland, The Netherlands, and Switzerland; http://www.trustthyroidtrial.com). In addition to cardiovascular events (CHD, stroke, heart failure…), the trial will assess disease-specific and health-related quality of life, muscle function, and cognitive function. This trial should definitively clarify the population who benefits from T₄ replacement by assessing multiple clinical outcomes. It will also assess whether any benefits are offset by potential adverse effects of replacement therapy, such as atrial fibrillation and osteoporosis. Until results of this trial and current clinical recommendations are updated (2, 3), clinicians must rely on existing observational data and clinical trials with shorter follow-up that seem to favor a TSH treatment threshold ≥ 10 mU/L for a potential benefit of T₄ replacement on clinical outcomes.