| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
REVIEW |
Institute of Epidemiology and Social Medicine (H.V., C.S.) and Departments of Gastroenterology, Endocrinology, and Nutrition (H.W.) and Cardiology (M.D.), University of Greifswald, D-17487 Greifswald, Germany
Address all correspondence and requests for reprints to: Henry Völzke, M.D., Institute of Epidemiology and Social Medicine, Ernst Moritz Arndt University, Walther Rathenau Str. 48, D-17487 Greifswald, Germany. E-mail: voelzke{at}uni-greifswald.de.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Two investigators independently searched the MEDLINE database. All case-control and cohort studies published in peer-reviewed journals were selected. Studies on nonthyroidal illness or low-T3 syndrome and reports from highly selected populations were not considered. Risk estimates from studies with appropriate adjustment for confounders were metaanalyzed.
Results: Four among eight studies performed to investigate the association between hyperthyroidism and mortality revealed an increased risk of either all-cause or circulatory mortality. Only the minority of studies, however, adjusted analyses for relevant confounders besides age, sex, and race. Studies after radioiodine therapy were generally biased by indication. Findings from 11 studies that investigated the relation between hypothyroidism and mortality were highly discrepant and partly even mutually exclusive. Some of these discrepancies are explained by confounding and selection.
Conclusions: The currently available evidence for a causal relation of both hyperthyroidism and hypothyroidism with mortality is weak and should particularly not be used to decide whether patients with subclinical thyroid conditions should be treated. Very old individuals might represent an exception from this rule and may benefit from mildly reduced thyroid function, but this has to be substantiated by further research.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Here we systematically reviewed studies on the relation between thyroid dysfunction and mortality questioning whether thyroid dysfunction is a causal factor for mortality in general. In particular, such causality would have impact on the ongoing controversy on current guidelines regarding treatment of subclinical thyroid disorders (5, 6).
A highly relevant methodological factor in judging whether thyroid dysfunction plays a causal role for mortality is to consider the concept of confounding. Confounders may represent the real causal factor for an association of interest, and nonconsideration of confounders may thus overestimate the strength of such an association. On the other hand, confounders may also suppress an association of interest. Considering suppressors in data analysis prevents underestimating the strength of such an association.
For assessing the evidence for possible causation, we used evidence-based medicine criteria. For this we evaluated the quality of previous studies according to methodological criteria and specifically addressed the question whether relevant confounders for the association between thyroid disorders and mortality had been considered.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In all metaanalyses a fixed-effects model was assumed. The P value of the homogeneity test was used only to describe the extent of heterogeneity inherent in a metaanalysis. The Mantel-Haenszel method was used to pool the risk estimates for studies with appropriate adjustment for confounding factors. Overall survival was calculated as hazard ratios, and values are given with 95% confidence intervals.
The grades of recommendation were assessed according to the Oxford Centre for Evidence-Based Medicine Levels of Evidence (7). The possible causation between thyroid dysfunction and mortality was graded by assessing the quality of considered studies and finally categorized into four levels with level A being the highest (consistent results of high quality cohort studies) and level D being the lowest level of evidence (based on physiology, first principles, troublingly inconsistent findings, or inconclusive studies) (7).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We identified 14 cohort studies (8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21) and two case-control studies (22, 23) that investigated the relation between thyroid dysfunction and mortality. There was an overlap of 542 subjects between two of these studies (11, 15). Two investigations (24, 25) were excluded being based on data of other publications (8, 11), and one report was not considered because no quantitative data had been given (26). Other studies (27, 28, 29, 30, 31, 32) were excluded because highly selected populations such as patients taking amiodarone, patients with various malignancies, dialysis patients, or otherwise severely ill patients were investigated. In addition, this review does not include one study (33) that analyzed mortality among 29 middle-aged women taking levothyroxine because of insufficient information on the reasons underlying levothyroxine treatment.
Most of the studies (9, 10, 11, 12, 14, 15, 16, 17, 21, 22, 23) were conducted in Europe, three in the United States (8, 19, 20), one in Australia (18), and one in Japan (13). All studies were carried out in areas with sufficient iodine supply. Study participants were recruited from the general population (10, 13, 14, 17, 18, 19, 20), registries (11, 15, 21), a primary care practice (12), or hospitals (8, 9, 16, 22, 23). The size of the study populations investigated varied from 298 (16) to 524,152 subjects (21). Likewise, the mean duration of the follow-up was heterogeneous ranging from 3.7 (14) to 20 yr (18). Six studies (8, 9, 11, 12, 15, 21) compared the outcome of subjects having thyroid dysfunction with the mortality of background populations using age- and sex- (and race) specific standardized mortality ratios. The other studies (10, 13, 14, 16, 17, 18, 19, 20, 22, 23) and subgroup analyses of a further study (15) compared the outcome of exposed subjects with an internal reference group of euthyroid subjects.
Hyperthyroidism and mortality
Eight studies (8, 12, 14, 17, 18, 20, 21, 22) observed the association between hyperthyroidism and mortality (Tables 1
and 2), whereas three (17, 18, 20) specifically addressed subclinical hyperthyroidism (Table 2). Four studies (8, 12, 14, 22) investigated relations between hyperthyroidism and all-cause mortality, yielding an increased risk of fatal events among hyperthyroid subjects. This relation, however, has not been confirmed by three other studies (17, 20, 21). Likewise, an increased circulatory mortality was described in three studies (8, 12, 14) but has not been found in four other investigations (18, 20, 21, 22).
|
|
].
|
). These differences were not present among men when sex-stratified analyses were performed in two of the reports (11, 15). With regard to circulatory mortality, one study (9) demonstrated an increased standardized mortality ratio (1.6, 95% confidence interval 1.6–1.7) among hyperthyroid patients treated with radioiodine in the whole study population and two studies (11, 15) in the female subpopulation (standardized mortality ratio 1.3, 95% confidence interval 1.2–1.4 and 1.3, 95% confidence interval 1.1–1.4). One other study (23) comparing the outcome in patients treated for hyperthyroidism with a population-based control sample also demonstrated an increased circulatory mortality in the patient cohort. Such findings were not confirmed in the female population of another study (8). There was also no relation between the exposure and circulatory mortality in the male subpopulations of two studies (11, 15). One study (8) observed that the increased all-cause mortality among treated females persisted in the subpopulation of patients who used thyroid hormone replacement therapy in the aftermath of radioiodine treatment. In contrast, two other studies (11, 15) found that differences in all-cause and circulatory mortality were no longer present in females receiving T4 substitution. Notably, in one study (16) no confounding was considered at all, and most other studies listed in Table 3 (8, 9, 11, 15) were controlled only for age, sex, and race. One study (23) further considered diabetes and baseline coronary artery disease, but none of the studies (8, 9, 11, 15, 16, 23) included smoking.
|
There were 11 studies (8, 10, 12, 13, 14, 16, 17, 18, 19, 20, 21) exploring the association of hypothyroidism with all-cause and circulatory mortality (Tables 4 and 5); five of them (13, 17, 18, 19, 20) provided specific information on the association between subclinical hypothyroidism and mortality (Table 5). Compared with the reference populations, all-cause mortality in hypothyroid subjects was increased in one study (8) and in the male subpopulation of another study (13) but was similar in the female subpopulation of the latter (13) as well as in the whole population of three other studies (10, 17, 20, 21). In contrast, one study (14) revealed decreased all-cause mortality in hypothyroid subjects. Furthermore, among patients of a general care practice, lowest mortality was found in subjects with hypothyroidism (12), albeit comparisons with the background population did not attain statistical significance. Likewise, findings for circulatory mortality were inconsistent. Compared with reference groups, circulatory mortality in subjects with hypothyroidism was increased in the total population of one study (18) and in a subpopulation of subjects free of baseline coronary artery disease having overt hypothyroidism (19) and was similar in six studies (8, 10, 12, 13, 19, 20, 21) and decreased in a further study (14).
|
|
). |
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Several mechanisms explain a causal relation between overt hyperthyroidism and mortality including atrial fibrillation (34), cardiovascular hypertrophy (35), and inflammation (36). Similar mechanisms also apply for subclinical hyperthyroidism (20, 36, 37, 38, 39). On the other hand, lone atrial fibrillation without underlying cardiac disease does not increase the mortality risk (40), and hyperthyroid conditions might protect from atherosclerosis (41). Hence, no or even inverse associations between hyperthyroidism and mortality may also be interpreted as biologically plausible.
The major drawback of most studies on the association between hyperthyroidism and mortality is that major confounders were not considered. Thus, smoking promotes the development of toxic goiter (42, 43) as well as the induction of hyperthyroidism in Graves’ disease (44). Because tobacco consumption is also an important predictor of both all-cause and circulatory mortality (45), nonconsideration of smoking may have given rise to an overestimation of hyperthyroidism-related effects on mortality (8, 9, 11, 12, 14, 15, 21, 22, 23). In addition, preexisting heart disease may represent a common cause for thyroid dysfunction and mortality. In euthyroid patients, the use of iodine-containing contrast agents for cardiac catheterizations leads to a 20% decrease of serum TSH levels (46), and cardiac arrhythmias may require antiarrhythmic therapy with the iodine-containing amiodarone. Both coronary artery disease and arrhythmias thus indirectly affect serum TSH levels and are at the same time highly relevant predictors of all-cause and circulatory mortality (47, 48).
Data pooling was possible only combining two studies (18, 20) with appropriate adjustment for relevant confounders. Although the aggregated study population comprised 5341 subjects, only 68 of them had subclinical hyperthyroidism. Both investigations (18, 20) were performed in regions with sufficient iodine supply where hyperthyroidism is a rare condition, thereby limiting the statistical power of these studies. Although iodine-deficient areas would instead constitute optimal study regions for investigating relations between subclinical hyperthyroidism and mortality, such studies have not yet been performed.
Studies investigating the association between hyperthyroidism and mortality in patient cohorts after radioiodine treatment (8, 9, 11, 15, 16, 23) were performed in large study populations. However, with respect to the question of whether hyperthyroidism is a causal factor for mortality, the interpretation of studies that investigated patients undergoing radioiodine treatment for hyperthyroidism is generally hampered because no clear differentiation is possible between the effects of the underlying thyroid disease, the effects after the change in thyroid function by treatment, and the side effects of treatment. Important confounders such as smoking, however, might have also influenced these studies (8, 9, 11, 15, 16, 23). In the absence of randomization, selection bias through indication may have produced spurious associations between the exposure of interest and mortality in these studies (8, 9, 11, 15, 16, 23). Particularly in Europe, radioiodine is commonly administered in patients who have contraindications against surgery (49, 50, 51). These contraindications include comorbidities such as poor general condition and severe cardiac or pulmonary diseases, all of which limited the general prognosis. On the other hand, young and healthy patients and women of child-bearing age are less likely to receive radioiodine treatment, thereby affecting selection and consequently the association of interest (23).
The high mortality of radioiodine-treated patients reported during the first year after radioiodine treatment, compared with the general population, may further mirror such selection (11), as does the finding that mortality from all causes increased with increasing cumulative doses of radioiodine (9, 11, 23). The latter may have produced by the wish of therapists to avoid persistent thyroid disease in particularly ill patients. Similar selection bias might be present in studies for which subjects have been recruited from hospitals (8, 9, 11, 15, 16, 23). In comparison with ambulatory diagnosed and treated patients, these subjects may have reduced prognosis due to poor general health and comorbidities. This problem might be especially evident in studies comparing the outcome of clinically recruited patients with a reference group selected from the general population (8, 9, 11, 15, 23).
Previously, new statistical methods such as propensity scores and instrumental variables have been developed to tackle the problem of selection bias and confounding in nonrandomized observational studies in treated patients (52). These methods are also promising for studies investigating the association between hyperthyroidism and mortality in treated patients.
Hypothyroidism and mortality
Experimental evidence suggests that hypothyroidism prolongs the life span of different animal models (53, 54). Reduced metabolic rate in hypothyroidism, which is related to increased survival in several species, provides the best explanation for this phenomenon (55). On the other hand, overt hypothyroidism is strongly associated with major cardiovascular risk factors such as hypertension (56), dyslipidemia (57), systemic inflammation (58), and insulin resistance (59). Although not always confirmed (60, 61), similar effects on cardiometabolic factors have also been reported for subclinical hypothyroidism (59, 62, 63, 64). Hence, from the perspective of biological plausibility, both increased and decreased mortality might a priori be expected to be present in hypothyroid individuals.
The inconsistency of findings from studies investigating the association between hypothyroidism and mortality can partly be explained by confounding and selection bias. Because there is an inverse relation between smoking and hypothyroidism (65, 66), nonconsideration of smoking may have given rise to an underestimation of the association between hypothyroidism and mortality.
Few studies (13, 18, 19, 20) investigating the role of hypothyroidism for mortality considered smoking. Two of the studies (13, 14) comprised selected populations, thereby limiting generalizability of their findings. One study (13) was performed in atomic bomb survivors, who have a restricted comparability with the general population, given the history of radiation exposure and intense surveillance afterward. The other study (14) comprised 85-yr-old individuals. In very old individuals, subclinical hypothyroidism may reflect a physiologically low metabolic rate rather than thyroid dysfunction (67). This hypothesis has been partly confirmed by one study (17) that was conducted in men who were 73 yr old or older, demonstrating low serum-free T4 levels to be associated with decreased mortality risk. The results obtained from elderly subjects, however, must not be extrapolated to younger individuals who might experience potential adverse outcomes if not treated (68). Subgroup analyses of one study (19) indicated that a relation between hypothyroidism and circulatory mortality was present only in subjects with both overt hypothyroidism and an a priori low cardiovascular risk. Such results, however, have not always been confirmed (20).
Studies examining mortality outcomes might have less statistical power and probably lower validity than studies of similar size that examine fatal and nonfatal cardiovascular events combined. A recent metaanalysis (69) pooled the data of 14 observational studies (10, 12, 13, 70, 71, 72, 73, 74, 75, 76) investigating the association between subclinical hypothyroidism and coronary artery disease (nonfatal and fatal). The summary odds ratio was 1.65 (95% confidence interval 1.28, 2.12) over all studies. For the five prospective cohort studies (10, 12, 13, 70, 71), a similar but nonsignificant trend was observed (odds ratio 1.42; 95% confidence interval 0.91, 2.21). Similar to the studies presented herein, only three of these 14 studies (13, 71) considered major cardiovascular risk factors as confounders for the association investigated. In contrast to the present analyses, pooling of the latter three studies (13, 71) yielded a clearly increased odds ratio of 2.38 (95% confidence interval 1.53, 3.69). One might hypothesize from the discrepancies between these and our results that patients with hypothyroidism may have an increased risk of cardiovascular disease, but a higher level of care after the detection of hypothyroidism may prevent fatal cardiovascular disease. This notion is supported by one study (21), demonstrating higher risks of nonfatal cardiovascular events but similar mortality in patients treated for hypothyroidism relative to the background population.
Further methodological issues
Besides confounding and selection bias, misclassification might also be relevant for cohort studies investigating the effects of thyroid dysfunction. Due to intraindividual variation of thyroid function, a single test result within the reference range does not exclude thyroid dysfunction (77). Furthermore, the awareness of thyroid disease during follow-up usually leads to fundamental changes in thyroid status after diagnosis. For example, overt hypothyroidism at diagnosis will be replaced by euthyroidism, subclinical hypothyroidism, or hyperthyroidism after treatment with thyroid hormones. In particular, studies with long follow-up periods should ascertain whether the exposure status has been changed after baseline examinations. Currently the minority of studies for the association between thyroid function and mortality performed repeated evaluations of the thyroid function status (10, 14, 23).
A further weakness of some studies (8, 9, 10, 11, 12, 15, 16, 21) is that subclinical thyroid dysfunction was not distinguished from yet undiagnosed overt dysfunction. Hypothetically, higher proportions of subjects with overt thyroid dysfunction in one study may yield different results, compared with another study with lower proportions of those subjects among the exposed. If subjects with overt thyroid dysfunction had been left untreated, effects on mortality might have been stronger with a higher probability to be detected in the former, compared with the latter study. If, however, subjects with overt thyroid dysfunction had been treated shortly after baseline examinations and subjects with subclinical conditions had not, misclassification bias would have been stronger in the former study and thus would attenuate measures of the real effect, compared with the latter study. Against this background, randomized, controlled trials would have been useful to specifically address the question of whether treatment of subclinical thyroid disorders influence mortality.
Finally, another explanation for differing results among studies is the different methods used for analysis. For example, two studies (12, 13) placed much emphasis on mortality outcomes year by year, leading to the finding that mortality was increased in certain years of the studies but not overall, findings that are hard to interpret. An international consensus on how to measure, define, and use variables in clinical and epidemiological research would be helpful to improve comparability among independently performed studies.
Conclusions
Most studies on the relation of both subclinical and overt hyperthyroidism with mortality are limited by nonconsideration of major confounders and selection bias. High-quality studies revealed inconsistent results. The recommendation regarding (overt and subclinical) hyperthyroidism playing a causal role with respect to mortality is therefore assessed as grade D. In very old subjects, decreased serum TSH values might indicate an increased mortality risk, but this should be confirmed by further studies.
Given the limitations of previous studies and contrary results demonstrated by well-designed studies, the recommendation for a causal role of (overt and subclinical) hypothyroidism for mortality has also to be graded with D. The finding of lower mortality in subclinically hypothyroid older individuals should be replicated by independent research.
We conclude that there is some biological plausibility to assume a causal relation of thyroid dysfunction with all-cause and circulatory mortality. The currently available data from observational studies are, however, insufficient to draw final conclusions. Evidence for a causal relation of both hyperthyroidism and hypothyroidism with mortality is weak and should particularly not be used to decide whether patients with subclinical thyroid conditions must be treated. Very old patients having subclinical hyperthyroidism might represent an exception from this general rule, but this has to be substantiated by independent research.
High-quality observational studies on the relation between thyroid dysfunction and mortality that particularly attempt to reduce selection as well as misclassification bias and control for important confounders are needed. Studies conducted in previously or currently iodine-deficient areas should clarify the role of subclinical and overt hyperthyroidism for all-cause and circulatory mortality. Randomized, controlled trials should be designed to specifically address the role of subclinical thyroid dysfunction for all-cause and circulatory mortality.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Author Disclosure Summary: H.V., C.S., H.W., and M.D. have nothing to declare. Nobody was previously employed by a company or has equity interests. Nobody consulted for a company. Nobody was previously employed or received lecture fees from a company. Nobody is an inventor on a patent.
First Published Online May 1, 2007
Received January 24, 2007.
Accepted April 23, 2007.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A de Rooij, J P Vandenbroucke, J W A Smit, M P M Stokkel, and O M Dekkers Clinical outcomes after estimated versus calculated activity of radioiodine for the treatment of hyperthyroidism: systematic review and meta-analysis Eur. J. Endocrinol., November 1, 2009; 161(5): 771 - 777. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Goichot, R. Sapin, and J. L. Schlienger Subclinical Hyperthyroidism: Considerations in Defining the Lower Limit of the Thyrotropin Reference Interval Clin. Chem., March 1, 2009; 55(3): 420 - 424. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E Forestier, S Vinzio, R Sapin, J L Schlienger, and B Goichot Increased reverse triiodothyronine is associated with shorter survival in independently-living elderly: the Alsanut study Eur. J. Endocrinol., February 1, 2009; 160(2): 207 - 214. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Haentjens, A. Van Meerhaeghe, K. Poppe, and B. Velkeniers Subclinical thyroid dysfunction and mortality: an estimate of relative and absolute excess all-cause mortality based on time-to-event data from cohort studies Eur. J. Endocrinol., September 1, 2008; 159(3): 329 - 341. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Biondi Should we treat all subjects with subclinical thyroid disease the same way? Eur. J. Endocrinol., September 1, 2008; 159(3): 343 - 345. [Full Text] [PDF]
|
|
|
|
 |
|
 S. Mariotti Mild Hypothyroidism and Ischemic Heart Disease: Is Age the Answer? J. Clin. Endocrinol. Metab., August 1, 2008; 93(8): 2969 - 2971. [Full Text] [PDF]
|
|
|
|
|
|
S. Razvi, A. Shakoor, M. Vanderpump, J. U. Weaver, and S. H. S. Pearce The Influence of Age on the Relationship between Subclinical Hypothyroidism and Ischemic Heart Disease: A Metaanalysis J. Clin. Endocrinol. Metab., August 1, 2008; 93(8): 2998 - 3007. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Ochs, R. Auer, D. C. Bauer, D. Nanchen, J. Gussekloo, J. Cornuz, and N. Rodondi Meta-analysis: Subclinical Thyroid Dysfunction and the Risk for Coronary Heart Disease and Mortality Ann Intern Med, June 3, 2008; 148(11): 832 - 845. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. N. Pearce, P. W. F. Wilson, Q. Yang, R. S. Vasan, and L. E. Braverman Thyroid Function and Lipid Subparticle Sizes in Patients with Short-Term Hypothyroidism and a Population-Based Cohort J. Clin. Endocrinol. Metab., March 1, 2008; 93(3): 888 - 894. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Biondi and D. S. Cooper The Clinical Significance of Subclinical Thyroid Dysfunction Endocr. Rev., February 1, 2008; 29(1): 76 - 131. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. C. Ridgway, Y. Tomer, and S. M. McLachlan Update in Thyroidology J. Clin. Endocrinol. Metab., October 1, 2007; 92(10): 3755 - 3761. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
