| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
CONTROVERSY IN CLINICAL ENDOCRINOLOGY |
Department of Medicine, Washington Hospital Center, Washington, D.C. 20010; Uniformed Services University of the Health Sciences, Bethesda, Maryland 20814; and Georgetown University School of Medicine, Washington, D.C. 20006
Address all correspondence and requests for reprints to: Dr. Leonard Wartofsky, Department of Medicine, Washington Hospital Center, 110 Irving Street NW, Washington, D.C. 20010-2975. E-mail: leonard.wartofsky{at}medstar.net.
 |
Abstract |
|---|
 Top Abstract References |
|---|
ADVANCES OVER THE past two decades in both the sensitivity and precision of assays for TSH and in our understanding and definition of mild thyroid disease (subclinical hypothyroidism and subclinical hyperthyroidism) have helped to fuel a controversy regarding what constitutes the normal range for TSH. The issue is important because it relates to whether to screen for thyroid disease and what to do when a patient is found to have mild abnormalities in TSH, whether by screening or otherwise. Thus, a more precisely defined reference range will allow the detection of patients with mild thyroid dysfunction who could benefit from therapy or at least closer follow-up. Indeed, the universe of patients with mild thyroid dysfunction is sufficiently large, and the benefits of therapy sufficiently clear (see below) to justify screening. These issues were the subject of deliberations by an expert panel convened by The Endocrine Society, American Thyroid Association, and American Association of Clinical Endocrinologists and subsequently published (1). Remarkably, reviewers representing all three professional organizations disagreed substantively with the conclusions of the panel (2), which have, in turn, been the subject of rebuttal arguments (3) and comment (4, 5). This present commentary summarizes our opinion based on the best available data from current literature and our own clinical judgment, both of which lead us to recommend general acceptance of a proposed revision of the TSH reference range now generally extant into what should be considered the true biologically normal TSH range under most circumstances.
Clinicians should appreciate that a presumed normal range given on a laboratory report is actually only a reference range for an ostensibly normal population. Such reference ranges were derived from cross-sectional studies of populations uncorrected for any underlying or occult disease, and for TSH have ranged from 0.5 up to 7.0 mU/liter. Indeed, the earlier first generation TSH RIAs often described the upper limit of normal at 10 mU/liter. With minor refinements in the past decade, the reference range has dropped to 0.5–5.5 mU/liter, reflecting the mean of all samples ± 2 SD. A symmetrical Gaussian distribution would place the midpoint or mean value of such a population at 3.0 mU/liter. This would appear to be inconsistent with recently published data indicating a population mean value of 1.5 mU/liter for an iodine-sufficient population (6, 7, 8, 9, 10). The discrepancy arises because the raw value reference interval for TSH is a skewed curve with a long tail toward the higher TSH values and is not a bell-shaped curve typical of a true Gaussian distribution curve. Thus, to create a normally distributed curve of the values, the reference interval for TSH is calculated by log transformation of the arithmetic TSH values.
We have learned that there are multiple possible reasons for this skew at the upper tail end of the TSH curve, principally related to the inclusion of individuals with underlying factors that confound the results. Principal among these is occult autoimmune thyroid disease (e.g. Hashimoto thyroiditis), and the population of individuals at the upper end of the reference range has a high prevalence of antithyroperoxidase (anti-TPO) antibodies (anti-TPOAb). Other confounders or possible explanations for slightly elevated TSH levels are listed in Table 1
.
|
An argument given against screening and finding mild TSH elevations and then treating the patient has been the incidence of reversion of elevated TSH to normal. This is the rationale for repeating the measurement. Normalization may reflect the presence of one of the causes of transient elevation of TSH listed in Table 1
, such as previous systemic illness or subacute thyroiditis. A previously unrecognized cause of elevated TSH levels, mutations in the TSH receptor (14), would appear to be relatively rare (15, 16). Estimates of subsequent reversion to a normal level have varied from 5% when apparently due to nonthyroidal causes (11) to as high as 40–50% in some European studies (12, 17, 18). There are several possible reasons for a greater rate of transient elevations or normalization of TSH levels in Europe. First, Hashimoto’s disease or prevalence of anti-TPOAb is much lower; thus, the frequency of true early thyroid failure on an autoimmune basis would be lower. Secondly, iodine deficiency is still ubiquitous in Europe, and transient higher iodine exposures could induce temporary TSH elevations in susceptible individuals. Finally, in the articles cited, TSH normalization was typically described as lowering of TSH to less than 5 mU/liter (19, 20), whereas we would consider levels between 3 and 5 mU as probably still elevated and warranting incremental dosage titration and continued follow-up. Other arguments have been presented both for (21) and against (22, 23) treatment of mild thyroid failure. It is of interest that risks associated with subclinical hypothyroidism are not restricted to the adult population. Moreover, some ostensibly transient elevations may be the harbinger of true thyroid failure to come; this is the case in some newborns who exhibit reversion of an initially high TSH level back into the reference range, but are found to subsequently develop mild thyroid failure with evidence of either TSH receptor mutations or positive TPOAb (24). In a diabetic population followed longitudinally, Warren et al. (25) found that the baseline TSH level was a better predictor of future thyroid dysfunction than thyroid autoantibodies. Consistent with recommendations for a lower reference range for TSH, they found that baseline TSH levels more than 1.53 mU/liter predicted subsequent thyroid dysfunction, whereas no thyroid dysfunction developed in the 293 patients with TSH levels less than 1.53 mU/liter.
As implied by the reports of Demers and Spencer (9) and Hollowell et al. (7), a compelling case may be made for revision of the method used to arrive at normal or reference interval values for TSH. The recommendation would be to draw upon a cohort of individuals with no personal or family history of thyroid dysfunction, no visible or palpable goiter, who are taking no medication, whose blood samples are drawn fasting in the morning hours (0600–1000 h), and who are seronegative for TPOAb using one of the newer, more sensitive immunometric assays (10). This would assure a reference interval that is more truly normal and should have a normal Gaussian distribution. When data for subjects with positive antithyroid antibodies or a family history of autoimmune thyroid disease are excluded from a so-called normal cohort, the normal reference range becomes 0.4–2.5 mU/liter (9, 10). It would be even better if it were feasible to exclude from the selected cohort those with diffuse hypoechogenicity of the thyroid on ultrasound in view of the recent finding that this precedes TPOAb positivity in autoimmune thyroid disease (26). Cost constraints preclude routine screening echosonography, but a sonogram might be considered in the setting of a marginally elevated TSH and negative TPOAb. Recently revised thyroid disease guidelines of the American Association of Clinical Endocrinologists propose a reference TSH range of 0.3–3.0 mU/liter (27).
Once having established the appropriate reference range, excluding confounding factors such as those listed in Table 1, it is also important to recognize the limitations of a population-based reference range from an individual patient-based reference range. Fraser and Harris (28, 29) pointed out that when an individual’s variation for a given test is narrow (i.e. their personal reference range), the value of the population-based reference range will be limited. In a study of normal men, Andersen et al. (8) demonstrated remarkably narrow individual reference ranges within a relatively small segment of the population reference range, i.e. confined to only 25% of a range of 0.3–5.0 mU/liter. This would suggest that a shift in the TSH value of the individual outside of his or her individual reference range, but still within the population reference range, would not be normal for that individual. For example, an individual (as in Anderson’s series) with a personal range of 0.5–1.0 mU/liter would be at subphysiological thyroid hormone levels at the population mean TSH of 1.5 mU/liter. This is consistent with studies of twins indicating that each of us has a genetically determined free T4 (FT4)-TSH set point or relationship (9, 30). In this context, Baloch et al. (10) estimated that it would require a measured TSH difference of 0.75 mU/liter to be significant in a given patient, a difference consistent with the narrow individual range observed by Anderson et al. (8). Thus, levels of TSH in an individual patient respond to fluctuations in serum FT4, but remain in a very narrow individual range and change very little unless the patient becomes hypothyroid or hyperthyroid (8).
There are also ethnic differences to be considered when establishing so-called normal ranges. The report by Hollowell et al (7) that analyzed the NHANES data implied that the Caucasian population among the 13,344 individuals screened may have skewed the upper end of the TSH curve due to the greater frequency of autoimmune thyroid disease in whites (12.3%) than in blacks (4.3%). Indeed, the mean TSH level in African-Americans was 1.18 mU/liter, in contrast to a mean of 1.40 mU/liter in Caucasians. Thus, interpretation of the reference range of TSH levels in African-Americans as being as high as 4 or 5 mU/liter would seem even more problematic. However, the NHANES survey concluded that age and ethnic differences per se do not have a significant enough effect on TSH levels to require adjustments in the proposed reference ranges (9, 25). Nevertheless, when data for subjects with positive TPOAb or a family history of autoimmune thyroid disease are excluded, the normal reference interval becomes much tighter, i.e. 0.4–2.0 mU/liter (7, 9), and this tighter reference range may be more applicable to African-Americans. Recognition of the need to tighten up the reference range for TSH has led to the National Academy of Clinical Biochemistry (NACB) reducing the upper limit of the reference range from 5.5 to 4.1 mU/liter (10).
The log-linear relationship between TSH and FT4 implies that a minor change in FT4 results in an amplified change in TSH to outside of the usual population-based reference range, although the FT4 is still within its own population-based reference range. This pattern, previously defined as subclinical thyroid disease, is perhaps best described as mild or minimally symptomatic disease (31, 32). In the case of subclinical hypothyroidism, for example, a slight drop in FT4 results in an amplified and inverse response in TSH secretion. Thus, although there is little doubt in our minds that a TSH level between 5 and 10 is abnormal, we would go further to state that a TSH level that rises in a given individual from a set point of 1.0 to a value of 3.5 is likely to be abnormally elevated and imply early thyroid failure. The NACB guideline 8 states that "analytical variability together with between-person and within-person estimates of biological variability suggest that the magnitude of difference in ... TSH... values that would be clinically significant when monitoring a patient’s response to therapy... is 0.75 mU/liter" (10). This concept that small differences within an individual’s normal range are significant is suggested by the study by Michalopoulu et al. (33), who treated TPOAb-positive hypercholesterolemic patients with TSH levels between 2 and 4 mU/liter with low dose levothyroxine and demonstrated normalization of TSH levels and improvement in their lipid profile.
Indeed, the new NACB guidelines (10) state that "greater than 95% of healthy, euthyroid subjects have a serum TSH concentration between 0.4 and 2.5 mU/liter." They go on to state that "ambulatory patients with a serum TSH >2.5 mU/liter, when confirmed by repeat TSH measurement made after 3–4 wk, may be in the early stages of thyroid failure, especially if TPOAb is detected." Logically it would seem that the exclusion of TPOAb-positive individuals from populations employed to establish a normal TSH reference range constitutes recognition that this TPOAb-positive population has Hashimoto’s disease and is vulnerable to progression to overt hypothyroidism.
The issue of the true normal range for TSH is far from simply an academic argument. It relates to the issues of screening for thyroid disease as well as treatment of the early or mild disease implied by slight increases or decreases in serum TSH. The frequency of abnormal TSH values found in the studies by Canaris et al. (34) and the NHANES III surveys (25) would seem to justify screening for thyroid disease in the general population. A case has been made for screening to detect mild thyroid dysfunction (2, 21, 35, 36), largely predicated on the potential benefits of treatment. The rationale for treatment is based upon a growing body of literature that describes adverse effects, although often mild, of the state of so-called subclinical thyroid disease. It is in this context that we disagree with the conclusions of the consensus panel (1). Although the panel concluded that there was good evidence that patients with slight elevations of TSH above 4.5 will progress to overt hypothyroidism, and that levothyroxine therapy would prevent symptoms, they did not hold that early treatment provided any benefit. This conclusion of the consensus panel notwithstanding, there have been extensive clinical studies and reviews indicating both the abnormalities present in mild thyroid failure and the benefits of T4 treatment (20, 21, 31, 32, 33, 36, 37, 38, 39, 40, 41, 42, 43) as well as more recent studies since the consensus panel was convened (44, 45, 46). The often cited U.S. Preventive Services Task Force (47) concurred that there was evidence of mild thyroid failure associated with slight elevations of serum TSH, but "could not determine the balance of benefits and harms of screening asymptomatic adults for thyroid disease." Like the nonendocrinologists on the U.S. Preventive Services Task Force, Helfand, using a rigorous evidence-based medicine approach (48), also concluded that it was "uncertain whether treatment will improve quality of life" in patients with mild thyroid failure. Several of the inconsistencies in these recommendations and those of the consensus panel (1) have been cited (2, 4). Some of these relate to the panel’s acknowledgment of the value of T4 therapy in a number of clinical situations, such as mild thyroid failure complicating pregnancy, while denying the value of TSH screening in pregnancy. We are also befuddled by the practice of supporters of the recommendations of the consensus panel who promote a target TSH range of 1.0–1.5 mU/liter in patients already receiving T4 therapy, whereas they refuse to accept TSH levels of 3–10 mU/liter as abnormal in patients not receiving T4 therapy.
Although many of the earlier studies indicating little benefit of T4 therapy were of patients who achieved TSH reductions down to only the range of 3–3.5 mU/liter, it is remarkable that some did show benefit with minimal TSH reductions in this range (20). Other studies employing appropriate dosage titration to TSH levels under 3.0 are more uniformly associated with improvement in symptoms, lipid abnormalities, and cardiovascular function (33, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46).
We will only comment briefly on the lower end of the normal or reference range for TSH because there is general agreement that it lies between 0.2 and 0.4 mU/liter (10), as indicated by a number of clinical studies (11, 25, 34, 49, 50, 51). However, although the consensus panel group (1) concluded that there was good evidence for an association of low TSH levels with atrial fibrillation and progression to overt thyrotoxicosis, they concluded that there were insufficient data to support intervention with therapy to prevent these outcomes! The panel was less sanguine about the association of low TSH and low bone mineral density and weakly implied that intervention might be indicated. As in patients with slightly elevated TSH levels, we are equally concerned about appropriate diagnosis and treatment of patients with TSH levels that are slightly below the reference interval (subclinical hyperthyroidism) because of risks to both heart (51, 52) and bone (53).
Surks et al. (1, 3) argue that initiation of levothyroxine therapy for mild thyroid failure would be inappropriate because it could result in overtreatment with attendant risks of subclinical hyperthyroidism. Although this risk might apply to a fraction of the population to be treated, an equivalent risk of undertreatment of such individuals might apply as well. Both results could be minimized by education of our primary care physicians about the desirable TSH target in their patients and the appropriate use of the available multiple dosage strengths of levothyroxine to achieve those targets. To us, individual failure on the part of physicians to appropriately monitor levothyroxine therapy and adjust dosage is not a rationale to withhold the indicated therapy. We find the reluctance of the consensus panel to consider treatment for mild TSH elevations puzzling when it is most likely that they would not argue with the wisdom and rationale for early therapeutic intervention for mild diabetes mellitus with slight, but definite, elevations in blood glucose, mild elevations in low-density lipoprotein cholesterol, or mild elevations in blood pressure. After all, few endocrine disease states appear suddenly in an "on or off" or "black or white" manner. Rather, the disordered physiology must start at a subintense level and then will have the potential to progress from mild to moderate to overt or severe. Just as we have revised downward our concept of normal range blood pressure and cholesterol, so we now should consider the evidence for doing so with TSH. Indeed, in the context of possible subclinical diabetes, recent data indicate that the cutpoint for hemoglobin A1C of 7.0% may be too high; 72% of excess cardiovascular risk occurred in patients with hemoglobin A1C levels between 5.0 and 6.9% (54). Thus, subclinical or mild disease states may apply to most disorders, and the recently described possibility of subclinical hyperaldosteronism as the cause a significant fraction of the hypertension that we see is another case in point (55, 56). Ultimately, the optimal control of our patients’ thyroid status will require inexpensive and readily available access to precise TSH measurements, and this may become feasible in the near future with home or point of care TSH testing. In the final analysis, we aim to employ our best clinical judgment and do what is optimal for our patients. Given the wealth of data on the abnormalities present in untreated subclinical hypothyroidism or hyperthyroidism and the demonstrated benefits of therapy to date, we are not disposed to have our hands tied by the deficiencies inherent in analyses of this issue by evidence-based medicine and allow our patients to continue to be at risk as a consequence. Clearly, one thing that all parties to this controversy can agree upon is the need for large scale, carefully constructed and performed studies (4). Until those data become available, a more precisely determined reference range for TSH of 0.3–2.5 (Fig. 1) will permit detection of individuals at risk of overt thyroid disease and should prompt their additional follow-up to confirm progression into thyroid dysfunction and thereby justify initiation of therapy. We will probably never have an absolute cutoff value for TSH distinguishing normal from abnormal, but recognition that the mean of normal TSH values is only between 1.18 and 1.40 mU/liter (7) and that more than 95% of the normal population will have a TSH level less than 2.5 mU/liter (10) clearly imply that anyone with a higher value should be carefully assessed for early thyroid failure. Thus, we believe that a TSH level between 5 and 10 mU/liter deserves confirmation and, if confirmed, warrants treatment. More judgment is required until more definitive data are available for the management of those patients with TSH values between 2.5 and 5.0. Assessment could include a review of their personal and family medical history and serum cholesterol and TPOAb levels, and the decision as to whether to initiate a trial of levothyroxine therapy is based more upon the "art of medicine" at this time than the science.
|
|
Footnotes |
|---|
Received March 1, 2005.
Accepted June 29, 2005.
|
References |
|---|
|
TopAbstractReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  T. E. Hamilton, S. Davis, L. Onstad, and K. J. Kopecky Thyrotropin Levels in a Population with No Clinical, Autoantibody, or Ultrasonographic Evidence of Thyroid Disease: Implications for the Diagnosis of Subclinical Hypothyroidism J. Clin. Endocrinol. Metab., April 1, 2008; 93(4): 1224 - 1230. [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]
|
|
|
|
|
|
M. I. Surks and J. G. Hollowell Age-Specific Distribution of Serum Thyrotropin and Antithyroid Antibodies in the U.S. Population: Implications for the Prevalence of Subclinical Hypothyroidism J. Clin. Endocrinol. Metab., December 1, 2007; 92(12): 4575 - 4582. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. A. Spencer, J. G. Hollowell, M. Kazarosyan, and L. E. Braverman National Health and Nutrition Examination Survey III Thyroid-Stimulating Hormone (TSH)-Thyroperoxidase Antibody Relationships Demonstrate That TSH Upper Reference Limits May Be Skewed by Occult Thyroid Dysfunction J. Clin. Endocrinol. Metab., November 1, 2007; 92(11): 4236 - 4240. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. M. Casey and K. J. Leveno Thyroid disease in pregnancy. Obstet. Gynecol., November 1, 2006; 108(5): 1283 - 1292. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. P. Walsh, L. C. Ward, V. Burke, C. I. Bhagat, L. Shiels, D. Henley, M. J. Gillett, R. Gilbert, M. Tanner, and B. G. A. Stuckey Small Changes in Thyroxine Dosage Do Not Produce Measurable Changes in Hypothyroid Symptoms, Well-Being, or Quality of Life: Results of a Double-Blind, Randomized Clinical Trial J. Clin. Endocrinol. Metab., July 1, 2006; 91(7): 2624 - 2630. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. Teng, Z. Shan, X. Teng, H. Guan, Y. Li, D. Teng, Y. Jin, X. Yu, C. Fan, W. Chong, et al. Effect of iodine intake on thyroid diseases in China. N. Engl. J. Med., June 29, 2006; 354(26): 2783 - 2793. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G Brabant, P Beck-Peccoz, B Jarzab, P Laurberg, J Orgiazzi, I Szabolcs, A P Weetman, and W M Wiersinga Is there a need to redefine the upper normal limit of TSH? Eur. J. Endocrinol., May 1, 2006; 154(5): 633 - 637. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Giavarina, R. M. Dorizzi, and G. Soffiati Indirect Methods for Reference Intervals Based on Current Data Clin. Chem., February 1, 2006; 52(2): 335 - 337. [Full Text] [PDF]
|
|
|
|
|
|
E. Grossi, R. Colombo, S. Cavuto, and C. Franzini Indirect Methods for Reference Intervals Based on Current Data: The authors of the article cited above respond: Clin. Chem., February 1, 2006; 52(2): 337 - 338. [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 |
