|Home | About | Journals | Submit | Contact Us | Français|
The goal of this study was to determine whether subclinical thyroid dysfunction was associated with incident heart failure (HF) and echocardiogram abnormalities.
Subclinical hypothyroidism and hyperthyroidism have been associated with cardiac dysfunction. However, long-term data on the risk of HF are limited.
We studied 3044 adults ≥65 years initially free of HF in the Cardiovascular Health Study (CHS). We compared adjudicated HF events over a mean 12-year follow-up and changes in cardiac function over 5 years among euthyroid participants, those with subclinical hypothyroidism (subdivided by TSH levels: 4.5–9.9, ≥10.0 mU/L), and those with subclinical hyperthyroidism.
Over 12 years, 736 participants developed HF events. Participants with TSH≥10.0 mU/L had a greater incidence of HF compared to euthyroid participants (41.7 vs. 22.9/1000 person years, p=0.01, adjusted hazard ratio=1.88, 95% confidence interval 1.05–3.34). Baseline peak E velocity, an echocardiographic measurement of diastolic function associated with incident HF in the CHS cohort, was higher in those with TSH≥10.0 compared to euthyroid participants (0.80 vs. 0.72 m/sec, p=0.002). Over 5 years, left ventricular mass increased among those with TSH≥10.0, but other echocardiographic measurements were unchanged. Those with TSH 4.5–9.9 or with subclinical hyperthyroidism had no increase in risk of HF.
Compared to euthyroid older adults, those with a TSH≥10.0 mU/L have a moderately elevated risk of HF and alterations in cardiac function, but not older adults with TSH<10. Clinical trials should assess whether the risk of HF might be ameliorated by thyroxine replacement in individuals with TSH≥10.0 mU/L.
Subclinical thyroid dysfunction is present in patients who have an abnormal thyrotropin (thyroid-stimulating hormone, TSH) level and a normal free thyroxine (FT4) level(1). Subclinical thyroid dysfunction is common, particularly in older individuals, with a prevalence of subclinical hypothyroidism of 10% and subclinical hyperthyroidism of 1.5%(2,3). Whereas it is generally accepted to treat the abnormal free thyroxine levels of overt thyroid dysfunction, the indications and threshold TSH for treatment of subclinical hypothyroidism and subclinical hyperthyroidism are areas of clinical controversy(1,3,4), as current evidence about the risks is limited(1,3).
Based on the known effects of thyroid hormone on the heart, it is reasonable to expect adverse cardiac effects in subclinical thyroid dysfunction(5). Subclinical thyroid disease has been associated with systolic and diastolic cardiac dysfunction, and small studies have shown that thyroxine replacement improved measurements of cardiac function in subjects with subclinical hypothyroidism(6,7). However, the clinical importance of these effects is unclear(3,8). Data on cardiovascular risks are conflicting(9–11), and no randomized clinical trials (RCT) have assessed the impact of thyroxine replacement on clinical cardiac endpoints(3). Only one study has examined the relationship between subclinical thyroid dysfunction and HF events. In a population-based study of adults aged 70–79, participants with TSH≥7.0 mU/L had a more than 2-fold higher risk of HF events, compared to euthyroid subjects(10), but echocardiography was not performed. No study has directly addressed the relationship between subclinical hyperthyroidism and HF events.
As the number of hospitalizations for HF has greatly increased(12), examining a common and easily treatable potential risk factor for HF is warranted. To determine whether subclinical thyroid dysfunction was associated with HF and cardiac dysfunction, we examined a large cohort study of community-dwelling older adults.
Participants were part of the Cardiovascular Health Study (CHS), a population-based, longitudinal study of risk factors for the development of cardiovascular disease (CVD) in 5888 adults ≥65 years(13). Enrolment of an original cohort of 5201 adults occurred in 1989–1990, and an additional cohort of 687, predominantly African Americans, was enrolled in 1992–1993. Eligible individuals were identified from an age- and sex-stratified random sample of the Medicare-eligible adults in 4 US communities. Details of the eligibility criteria have been previously described(9). All participants gave written informed consent; the Institutional Review Boards at study sites approved the protocol.
Fasting TSH was measured at baseline in a subsample (n=3678) of CHS participants, selected according to availability of stored serum for analysis. Because our primary study question pertained to unrecognized thyroid function abnormalities, participants taking thyroid hormone preparations at baseline (n=339) or other medications that could affect thyroid testing, including antithyroid drugs (n=1), corticosteroids (n=77) and amiodarone (n=2), were excluded. We also excluded participants with known HF at baseline (n=144) to examine incident HF.
Serum TSH and FT4 concentrations were assayed, as previously described(9), with FT4 (normal range: 0.7–1.7 ng/dL [9–22 pmol/L]) measured in individuals with TSH <0.10 or >4.50 mU/L for the 95% of samples with sufficient serum for this additional test. Compared to those without thyroid function testing (n=1523), participants with thyroid function testing were less likely to be men (38 vs. 55%, p<0.001) and to have prevalent CVD (22 vs. 26%, p=0.002), but mean ages, race proportions, thyroxine use, and prevalent HF were similar.
Study participants were classified into 3 groups based on their thyroid function tests(9):
Participants whose testing suggested nonthyroidal illness (low TSH and FT4, n=2), ≥TSH 4.50 and a missing FT4 (n=21), and overt thyrotoxicosis (TSH<0.10 with an elevated FT4, n=1) or overt hypothyroidism (TSH≥20 mU/L or >4.50 with a FT4<0.70 ng/dL, n=47) were excluded from analyses. The sample for our analyses was 3044.
During the 15-year follow-up, we assessed incident HF events among participants free of HF at baseline. Clinical outcomes were ascertained every six months. Diagnoses have been adjudicated until 6/30/2004 based on interview, review of medical records, and other support documents without knowledge of thyroid status. HF events were defined on the basis of diagnosis from a physician and consideration of symptoms, signs, chest radiographs, and treatment of heart failure (current prescription for a diuretic agent and digitalis or a vasodilator)(14,15).
As previously described(16), echocardiographic images were obtained in 1989–1990 and 1994–1995, using a standardized protocol with core reading centers. The following measurements were examined without knowledge of thyroid status; left ventricular (LV) mass and fractional shortening at the midwall, a quantitative measurement of global chamber function, were calculated, as reported(17). Qualitative assessment of LV ejection fraction (LVEF) was classified as normal, borderline, or abnormal, approximately corresponding to values ≥55%, ≥45% and <45%, respectively(17). Agreement between baseline and 5-year follow-up echocardiography core centers has been previously reported(17). For diastolic function, left atrial size, Doppler peak E and A velocities were measured(18,19). Doppler filling velocities and their ratios, consistent with impaired relaxation and restrictive filling, were previously found in CHS to be predictive of incident HF(14). 896 participants had missing 1994–1995 echocardiograms, because of death (36%), missed visit (17%) or follow-up visit outside the clinics with no echocardiogram (47%). The 576 alive participants without follow-up echocardiograms were older, more likely to be nonwhite, smoker, have diabetes, atrial fibrillation and subclinical hyperthyroidism (2.9 vs. 1.4%); prevalence of hypertension, CVD, subclinical hypothyroidism and gender proportions did not differ significantly. LVEF at the time of incident HF was based on data abstracted from reports of echocardiograms performed during hospitalization for HF, and was classified as described above.
We collected self-reported race and smoking status, classified as never, current, or former. Thyroid medication use was assessed annually via medication bottle examination. Physical examination included blood pressure, heart rate, and body mass index (BMI). Hypertension was defined as self-report and use of anti-hypertensive medications, or blood pressure ≥140/90 mmHg. Diabetes was defined as a fasting glucose ≥126 mg/dL (7.0 mmol/L) or use of hypoglycemic medication. Atrial fibrillation at baseline was self-reported or determined on ECG or Holter monitor(9). For prevalent HF at baseline, self-reports were confirmed by physical examination or, if necessary, by a validation protocol that included surveys of treating physicians or review of medical records(20). Prevalent CVD (defined as coronary heart disease, stroke, or transient ischemic attack) was based on self-report, medical record verification and use of selected drugs.
We calculated HF incidence per 1000 person-years of follow-up, and used log-rank tests to compare Kaplan-Meier estimates of HF incidence across the four thyroid groups. We used Cox proportional hazard models to examine associations between the four groups and HF. The multivariate models adjusted for known clinical risk factors of HF in older adults(12,15,21,22) that might be potential confounders in the relationship between subclinical thyroid disease and HF. To check the sensitivity of our results to the selection of covariates, we assessed models that excluded potential confounders with p-values >0.2 after adjustment, and obtained similar results. We hypothesized some interactions a priori: the relationship between subclinical thyroid disease and HF might differ by gender, prevalent CVD or thyroxine use during follow-up. As previously(9), we first examined models in which follow-up was censored at the time of first thyroxine use, as the main analysis. Because of interaction with thyroxine use, we also used models in which baseline hazard was stratified by thyroxine use during follow-up as a time-dependent covariate, and participants changed stratum when thyroxine use was initiated.
We graphically examined smooth estimates of hazard function for HF against TSH to examine cut-points at which HF risk might be increased, as well as tabulation of event rates within smaller TSH intervals, and used Schoenfeld residuals to check proportional hazards assumption for thyroid status and for covariates included in the multivariate models. We used models with baseline hazard jointly stratified by four variables (prevalent atrial fibrillation, CVD, diabetes and hypertension) that did not meet this assumption(23). We also categorized age, the only continuous covariate that did not meet the assumption of log-linearity. Results were reported as hazard ratios (HR), with 95% confidence intervals (CI).
Among participants with echocardiographic data, multiple linear regression was used to assess baseline differences and changes over time in echocardiographic measurements across thyroid groups, after adjustment for age and gender, similar to previous analyses of echocardiograms in CHS(14). We conducted analyses using Stata 9.2 (Stata Corporation, Texas).
The mean age was 72.6 years; 60% were women (Table 1). Mild subclinical hypothyroidism (TSH 4.5–9.9 mU/L) was present in 428 (14.5%) participants, severe subclinical hypothryoidism (TSH 10.0–19.9 mU/L) in 46 (1.5%), and subclinical hyperthyroidism in 44 (1.4%). Subclinical hypothyroidism was more common in women, and associated with lower systolic blood pressure. Subclinical hyperthyroidism was associated with higher BMI and lower LDL-cholesterol.
Most echocardiographic measurements at baseline did not differ significantly by thyroid status (Table 2). Participants with TSH≥10.0 had a higher peak E velocity (0.80 vs 0.72 m/sec, p=0.002), and this difference persisted after adjustment for age, gender, heart rate and systolic blood pressure. Peak E velocity was associated with incident HF in the overall study sample (HR 1.14 for each 0.1 m/sec increment, 95%CI:1.09–1.18, p<0.001) and in those with TSH≥10.0 (HR 1.45, 95%CI:1.20–1.76, p<0.001) after adjustment for age, gender and systolic blood pressure. Compared to euthyroidism, subclinical hyperthyroidism was associated with larger left atrial size, higher proportions with E/A ratio <0.7, and increased heart rate (Table 2), differences that persisted after excluding those with atrial fibrillation.
Over a median (interquartile range) follow-up of 12 years (7.0–14.4), 736 (24%) participants developed HF events. Among the 474 participants with sublinical hypothyroidism, 109 received thyroxine replacement during follow-up, with 28 reporting intermittent use. Among the 46 participants with TSH≥10.0, 22 received thyroxine during follow-up (7 intermittent use). In the analysis censoring follow-up at first thyroxine use, participants with TSH≥10.0 had a greater incidence of HF compared to euthyroid participants (41.7 vs. 22.9/1000 person-years, p=0.01), but rates were similar for those with subclinical hyperthyroidism or TSH 4.5–9.9 (Figure). In multivariate analysis censoring at first thyroxine use, the risk of HF was increased among those with TSH≥10.0 (HR: 1.88, 95%CI: 1.05–3.34; Table 3, model 1). Multivariate model omitting lipids yielded similar results.
The relationship between subclinical hypothyroidsim and HF differed by thyroxine use during follow-up (p=0.02 for interaction). In a model (model 2) including all follow-up and thyroxine use as a time-dependent covariate, participants with TSH≥10.0 had an increased risk of HF during periods where thyroxine use was not reported, but no increased risk during periods of thyroxine use. Almost identical results were obtained using a model that incorporated history of thyroxine use as a time-dependent covariate (participants classified as user from first use) and allowing for interaction of this measure with thyroid status (data not shown).
Stratifying sublinical hypothyroidism into those with TSH 4.5–6.9 and 7.0–9.9 mU/L, we similarly found no significantly increased risk of HF; graphical examination of hazard function showed no increased risk up to around 10 mU/L. A cubic spline analysis also showed the non-linearity of this relationship. None of the 6 participants with TSH<0.10 had HF events. The relationship between subclinical thyroid disease and HF did not differ by prevalent CVD or gender (each interaction p value >0.20). Excluding 595 participants with prevalent CVD yielded similar results, including in those with TSH≥10.0 (HR: 2.17, 95%CI: 1.15–4.07) in multivariate analysis censoring at first thyroxine use. Based on baseline echocardiographic data, peak E velocity might be a potential mediator of this relationship; further adjusting Model 1 for peak E velocity yielded an HR of 1.69 (95%CI: 0.94–3.02) for those with TSH≥10, showing that this relationship was not fully explained by increased peak E velocity and that other factors, including unmeasured echocardiographic parameters, might play a role in this relationship.
Repeat echocardiograms were obtained on 2148 participants after 5 years; those with TSH≥10.0 had a larger increase in LV mass (+21 vs. +4 g, p=0.04, Table 4) and a greater proportion of HF events associated with low EF than in euthyroid participants (80 vs. 45%, p=0.08). Peak E velocity decreased more in subjects with TSH≥10 than in euthyroid participants (−0.10 vs. −0.01m/sec, p=0.005), which might be related to increase in LV mass over time, increasing impairment of LV relaxation, a plateau effect and/or regression to the mean due to higher baseline values.
Compared to euthyroidism, subclinical hyperthyroidism was associated with a smaller increase in left atrial size and peak A velocity over time, but cross-sectional year-5 left atrial sizes and peak A velocities did not differ (40.5 vs. 40.0 mm, p=0.65; 0.81 vs. 0.81 m/sec, p=0.90). Excluding participants with atrial fibrillation at baseline or those who developed HF before year 5 examination (n=237) yielded similar results. For the 499 participants who developed HF after year 5 examination, the time interval±SD between the 5-year echocardiograms and HF was 4.4±2.6 years (3.5±1.9 for those with TSH 10.0).
In this large, population-based study of older adults, subclinical hypothyroidism was associated with a moderately elevated risk of HF among older adults with TSH≥10.0 mU/L, consistent with another large cohort study of older individuals(10). The risk of CHF was not increased among the high proportions of older adults with TSH levels between 4.5 and 9.9 mU/L or among those with TSH<0.45. Adverse alterations in two echocardiographic measurements, peak E velocity at baseline - an echocardiographic measurement of diastolic function associated with HF in this cohort sample and previous CHS analysis(14) - and increase in left ventricular mass over 5 years, were also found exclusively in the subgroup of subclinical hypothyroidism with TSH≥10.0 mU/L.
To our knowledge, only one previous population-based prospective study directly examined the relationship between subclinical hypothyroidism and HF events, and found that adults aged 70–79 with TSH≥7.0 mU/L had a higher risk of HF (HR: 2.88, 95%CI 1.39–5.99) in multivariate analysis compared to euthyroid participants, with a more pronounced risk in those with TSH≥10 (HR 3.10, 95%CI: 1.30–7.39), and no increased risk in those with TSH 4.5–6.9(10). Possible explanations for the weaker point estimate found in CHS as compared to the Health, Aging, and Body Composition (Health ABC) Study include differences in the study population that was mainly white in CHS (40% were black in Health ABC), younger mean age (72.6 vs. 74.7), different length of follow-up (12 vs. 4 years), no upper TSH cut-off in Health ABC analysis (1% with TSH≥20) and less precision due to lower power in the Health ABC. In the present study, the fact that HF risk was limited to participants with TSH≥10 who were not taking thyroxine, with a HF risk similar to the euthyroid group under thyroxine replacement, strengthens the case for a potential causal relationship, which could only be definitively proven by RCTs. Potential mechanisms for the impact of thyroid dysfunction on HF may be related to the effects of thyroid hormones on the heart(5,7), notably by the regulation of genes coding for cardiac proteins(5). Overt hypothyroidism also alters cardiovascular function(24).
For most echocardiographic measurements, we found no significant differences for subclinical hypothyroidism or hyperthyroidism compared to euthyroidism, in contrast to several non population-based studies(6,25–27), but consistent with other population-based studies(28,29). Several explanations exist for these discrepant findings, including the age of the underlying study population and the degree of subclinical thyroid dysfunction(8). All of these previous studies had insufficient sample sizes with subclinical hypothyroidism (n=8–66 as compared to 474 in the present study) to investigate different TSH cutoffs(7,28), and none linked abnormalities in the surrogate echocardiographic markers with HF events(8). Indeed, our participants with subclinical hyperthyroidism demonstrated some differences in baseline echocardiographic measurements, but no increase in HF events, suggesting that these echocardiographic abnormalities, while statistically significant, do not result in HF. In our cohort, those with TSH≥10.0 had a higher peak E velocity at baseline, an echocardiographic measurement of diastolic function that was associated with incident HF in this cohort sample and previous CHS analysis(14). The higher early diastolic filling velocity we found in participants with TSH≥10 may reflect increased left atrial pressure and diastolic dysfunction from more marked subclinical hypothyroidism. Some small RCTs have been performed that have shown improvement in echocardiographic measures of cardiac function(30–32), supporting our findings of reduction of risk of HF after thyroxine initiation.
Among limitations, our data may not be generalizable to younger age groups and it has been suggested that there may be age differences in the associations between subclinical thyroid dysfunction and adverse outcomes(33). Thyroid function testing was performed at a single point in time, which is a limitation of all published observational cohorts(10,11,34). Our power was limited in those with TSH≥10, as this group was small. Our echocardiographic findings could be affected by multiple comparisons and missing data for 5-year and incident HF echocardiograms; we could not definitively conclude on the type of cardiac dysfunction (systolic or diastolic) involved in those with TSH≥10, which needs to be explored in future studies.
These data have a number of strengths: the large, population-based cohort of older adults, designed to examine cardiovascular risk factors; the mean 12-year follow-up; the formal adjudication of HF events; the exclusion of individuals taking thyroxine or other medications that could affect thyroid function testing; and the incorporation of thyroxine use over time analytically(9). In addition, few large prospective studies have the availability of baseline and follow-up echocardiography data.
Our data suggest that subclinical hypothyroidism with a TSH≥10.0 mU/L represents a potentially modifiable risk factor for HF in older adults, but not subclinical hypothyroidism with moderate TSH levels (TSH 4.5–9.9 mU/L) and subclinical hyperthyroidism. Our study builds on the prior study demonstrating increased HF risk in marked subclinical hypothyroidism(10), with additional mechanistic support in our study from echocardiographic data and information on reversibility of HF risk with thyroxine replacement. Our findings of a lack of HF risk in the high proportions of older adults with less severe subclinical hypothyroidism are also important, as many patients with TSH levels of 4.5–9.9 mU/L are treated in clinical practice(35), without consistent evidence to support increased risk without thyroxine replacement and improved risk with replacement(1,3). This and previous studies have mostly found no increased cardiovascular risk in subjects with TSH<10(9,10,36), with some conflicting data(11). Moreover, about 20 % of patients are currently overtreated by thyroxine replacement with an increased risk of subclinical hyperthyroidism that has been associated with atrial fibrillation and increased fracture risk(3,37). In aggregate, our findings might help refine a treatment threshold at which clinical benefit would be expected(8) and demonstrate a subpopulation at risk for a life-threatening condition. Clinical trials should examine the efficacy of screening for and treating subclinical thyroid dysfunction, and assess whether the risk of HF might be ameliorated by thyroxine replacement in individuals with TSH levels above 10.0 mU/L.
The research reported in this article was supported by contracts N01-HC-35129, N01-HC-45133, N01-HC-75150, N01-HC-85079 through N01-HC-85086, N01 HC-15103, N01 HC-55222, and U01 HL080295 from the National Heart, Lung, and Blood Institute, with additional contribution from the National Institute of Neurological Disorders and Stroke. A full list of participating CHS investigators and institutions can be found at http://www.chs-nhlbi.org. The TSH measurement and this study were supported by an American Heart Association Grant-in-Aid (to Dr Fried) with funding from July 1991 to June 1993.
Role of the Sponsor
This study was funded through contracts with the NHLBI and included substantial NHLBI involvement in study design and oversight. A member of the NHLBI serves on the executive committee of the study, and NHLBI reviewed the manuscript, and approved its publication.
Conflict of interest
No author has a financial interest in the subject matter or materials discussed in the manuscript.
Dr. Vittinghoff, Professor of Biostatistics in the Department of Epidemiology and Biostatistics University of California, San Francisco, reviewed the statistical analyses of the paper and is included in the authors of this paper.