Werner & Ingbar’s The Thyroid: A Fundamental & Clinical Text
9th Edition

13
Measuring Serum Thyrotropin and Thyroid Hormone and Assessing Thyroid Hormone Transport
Anthony D. Toft
Geoffrey J. Beckett
This chapter reviews the various serum tests of thyroid function; outlines how these tests should be selected, performed, reported, and interpreted; and discusses how the pitfalls of the tests can be recognized or avoided. The United States National Academy of Clinical Biochemistry has published extensive guidelines concerning the Laboratory Support for the Diagnosis and Monitoring of Thyroid Disease, which are of relevance to both clinicians and laboratories (1). In general, measurements of serum thyrotropin (TSH) and the thyroid hormones thyroxine (T4) and triiodothyronine (T3) should be performed to determine thyroid status before embarking upon the tests necessary to determine the cause of the thyroid dysfunction or any treatment for it. Diagnosis of thyroid disorders relies heavily on the results of measurements of serum TSH and thyroid hormones, and it is essential that reliable assay systems be used for these measurements.
SERUM THYROTROPIN
Measurement of TSH in a basal serum sample by a sensitive immunometric assay is the single most sensitive, specific, and reliable test of thyroid status. In patients with primary (thyroidal) hypothyroidism, serum TSH concentrations are high, whereas in almost all patients with thyrotoxicosis, serum TSH concentrations are low, usually below 0.02 mU/L. However, measurement of serum TSH alone is not a reliable test for detecting thyroid dysfunction arising from hypothalamic-pituitary disease, or other specific instances, such as assay interference and end-organ resistance (2).
Evolution of Serum Thyrotropin Measurements
The introduction in the 1980s of sensitive and specific immunoassays for TSH revolutionized the way in which thyroid function testing is approached (3,4). Before then, serum TSH was measured by relatively insensitive and nonspecific radioimmunoassays (5). These assays could only detect serum TSH concentrations as low as about 1 mU/L, and therefore serum TSH could not be detected in all normal subjects. The assays could detect the high serum TSH concentrations found in patients with primary hypothyroidism, but because of their limited sensitivity they were unable to distinguish reliably between normal subjects and patients with thyrotoxicosis or other conditions associated with decreased TSH secretion without recourse to thyrotropin-releasing hormone (TRH) stimulation.
Serum TSH is now measured by immunometric assays that use noncompetitive labeled-antibody methods. These assays use two highly specific monoclonal antibodies raised to epitopes on the alpha and beta subunits, respectively, of the TSH molecule. Nonisotopic labels are now preferred over radioactive labels, because they improve assay sensitivity and readily lend themselves to automation (Fig. 13.1). With immunometric assay methodologies it is now possible to detect serum TSH concentrations < 0.02 mU/L with a high degree of specificity (1,6). As a result of the improvements, serum TSH assays have become the most sensitive and specific method for assessing thyroid status (7).
FIGURE 13.1. Configurations of immunoradiometric assays (IRMAs). An IRMA usually consists of a capture antibody and a signal antibody. Both antibodies may be added together with the sample to form a sandwich, or the capture antibody and analyte may be added first, followed by a wash step, before the addition of signal antibody. The capture antibody is usually immobilized on an insoluble matrix. The signal antibody may be labeled with 125I, but is more often labeled with a fluorescent or chemiluminescent chemical or an enzyme, usually horseradish peroxidase or alkaline phosphatase. These enzymes can be linked to substrates that give a colorimetric end point. Greater sensitivity can be achieved if the enzyme is linked to a substrate that produces fluorescence or chemiluminescence.
Functional Sensitivity
The introduction of immunometric assays for serum TSH led to a problem regarding the nomenclature describing the assays. Manufacturers described their products with a variety of qualifying adjectives such as “sensitive,” “supersensitive,” and “ultrasensitive,” terms that did not necessarily equate with the relative degree of sensitivity of the assays. Also, analytical sensitivity was frequently quoted. This is the lowest serum concentration of TSH that can be distinguished by statistical analysis from the zero standard within a single assay when multiple replicates are analyzed. This is an inappropriate and misleading way of estimating sensitivity because laboratories do not analyze multiple
P.330

replicates of a sample routinely (1,6,8,9). Also, the sensitivity of a single assay does not represent the day-to-day variability that occurs in routine use. The key to identifying the sensitivity of an assay lies in constructing a precision profile. This is a plot of the coefficient of variation (CV) against serum TSH concentration, using results obtained over a long period of time with multiple batches of reagents (Fig. 13.2). The concentration of TSH at which the response in the assay has a CV of 20% is the “functional sensitivity” of the assay, and is the value that most accurately represents the performance of an assay when used routinely (1,6,8,9,10). The use of functional sensitivity gave rise to the “generation” nomenclature for TSH assays. The early assays with a functional sensitivity of 1.0 to 2.0 mU/L were termed first-generation assays, and those with a functional sensitivity of 0.1 to 0.2 mU/L and 0.01 to 0.02 mU/L were termed second- and third-generation assays, respectively (6). Fourth-generation assays, which have a sensitivity of 0.001 to 0.002 mU/L, have been developed for research purposes (11). These are arbitrary terms and should not be used; instead, serum TSH assays should simply be described in terms of their functional sensitivity in milliunits per liter.
FIGURE 13.2. The precision profile and functional sensitivity of a serum thyrotropin (TSH) assay. The precision profile of an assay is derived from multiple assays using different batches of reagents. The quantity of TSH that can be measured with a coefficient of variation of 20% is the functional sensitivity of the assay. In this example the functional sensitivity is 0.0125 mU/L.
It is essential that serum TSH assays be sufficiently sensitive for clinical diagnostic purposes. To this end, the laboratories should use a method to measure serum TSH that has a functional sensitivity of ≤0.02 mU/L, and that functional sensitivity should be defined as briefly noted earlier as the 20% between-run CV determined by a protocol that allows for changes in operator and reagent batch, and is assessed in at least 10 different assay runs. Laboratories should also confirm the functional sensitivity quoted by the manufacturer and have quality-assurance procedures in place to ensure that the functional sensitivity of the assay is regularly monitored.
MEASUREMENTS OF SERUM THYROID HORMONES
Thyroid hormones are transported in serum almost entirely reversibly bound to protein (see Chapter 6). Most (70% to 80%) of the T4 and T3 in serum is bound to thyroxine-binding globulin (TBG), and the remainder is bound to transthyretin (also known as thyroxine-binding prealbumin) and albumin. Approximately 0.05% of the T4 and 0.2% of the T3 in serum are free (i.e., not bound to any protein). Most available evidence indicates that only the free hormone can cross the cell membrane and affect intracellular metabolism (12,13). Thus, patients with marked TBG deficiency have low serum total T4 and T3
P.331

concentrations, but they are euthyroid and have normal serum concentrations of TSH, free T4, and free T3. This does not necessarily mean that the complexes of T4 or T3 and binding protein have no biological role; binding proteins may be involved in targeting T4 and T3 to specific tissues. The reversible binding of T4 and T3 to these binding proteins also ensures that a relatively constant concentration of free hormone is supplied to tissues. The ability of T4 and T3 to dissociate from the binding proteins and maintain a constant free hormone concentration forms the basis of the methods used to measure free T4 and free T3 in serum.
MEASUREMENTS OF SERUM FREE THYROXINE AND FREE TRIIODOTHYRONINE
Theoretical Considerations
Serum free hormone measurements theoretically provide a more reliable means of diagnosing thyroid dysfunction than measurements of serum total hormone concentration, because the latter can be altered not only by thyroid dysfunction, but also by changes in the production of the binding proteins. Because the serum concentrations of free T4 and free T3 are extremely low, measuring them in the presence of the high concentrations of T4 and T3 that are protein-bound has proved challenging. Indeed, poor assay design has resulted in a mass of conflicting literature regarding the results of measurements of serum free T4 and free T3 in patients with nonthyroidal illness and those taking certain drugs (14,15). The theoretical basis on which reliable measurements of serum free T4 and free T3 should be designed has been reviewed recently in detail (16). In humans, the serum free T4 (and free T3) fraction is mostly affected by changes in the affinity and concentration of TBG, whereas changes in the affinity and concentration of albumin or transthyretin have little effect (16).
In essence, the serum free T4 (and free T3) concentration results from the equilibrium between the bound and free hormone.
The law of mass action dictates the proportion of T4 that binds to TBG, i.e., the affinity of the binding protein (Keq) multiplied by its concentration. This is the “relative binding capacity.” Thus, if the affinity or concentration of TBG decreases, the proportion of T4 bound will also diminish, and the serum free T4 concentration will increase. This illustrates the mechanism for the low serum total T4 and high free T4 concentrations found in patients with nonthyroidal illness who have a decrease in serum TBG-binding capacity.
Measurements of serum free T4 (and T3) involve sampling the fraction of T4 in serum that is free (unbound). This can be done using physical separation of free T4 from bound T4 [through equilibrium dialysis (ED) or ultrafiltration], or, alternatively, by adding an antibody that “captures” a proportion of the pool of free T4. The removal of free T4 from the original equilibrium will result in dissociation of T4 from the binding protein to create a new equi librium. A crucial requirement of any valid method for measurement of serum free T4 is that there is minimal disruption of the original equilibrium during the assay process. If this principle is adhered to, then the measured serum free T4 value will be a good estimate of the serum free T4 concentration in vivo. The disruption of the equilibrium will be minimal if there is little dilution of the sample and only a small proportion of the T4 pool is captured. Other factors that influence the affinity of binding proteins for T4 and thus modify serum free T4 concentrations include temperature, pH, and the ionic composition of any buffer used in the assay. These factors all need to be controlled in any assay for serum free T4 (16). The inadequacies of many commercial assays have led to the suggestion that the term “free hormone estimate” be used for all serum free T4 measurements used in clinical practice (1). However, this term fails to indicate that some assays are better than others, and it may also cause the clinician to have inappropriate lack of confidence in the result.
Methods for Measuring Serum Free Thyroxine and Triiodothyronine
Equilibrium Dialysis and Ultrafiltration
Equilibrium dialysis and ultrafiltration are widely regarded as reference methods for assays of serum free T4 and free T3, although neither is completely satisfactory (1,16,17,18,19,20,21). Both methods involve the initial separation of free hormone through a semipermeable membrane followed by quantification of the separated hormone by sensitive and specific immunoassay.
A typical dialysis apparatus comprises two compartments separated by a semipermeable membrane. Serum is added to one compartment and buffer to the other. The sample is then incubated with agitation, during which time small molecules like T4 pass across the membrane into the opposite compartment. This process will proceed until equilibrium has been reached such that the concentration of the permeable molecules is equal on both sides of the membrane. The T4 concentration in the buffer compartment at this new equilibrium should provide a good estimate of the free T4 concentration in the original sample. Equilibrium is reached typically in 16 to 24 hours. This long incubation period can cause problems in certain samples (e.g., patients given heparin), because nonesterified fatty acids (NEFAs) may be released by the hydrolysis of endogenous triglyceride through the action of lipoprotein lipase, which is activated by heparin (19). When present in high concentrations, these fatty acids diminish the binding capacity of
P.332

TBG for T4, thus increasing the free T4 concentration during the dialysis period. A similar increase in free T4 will occur if serum samples from patients given heparin are stored unfrozen before analysis. Ions from the buffer compartment can also move into the sample during dialysis and modify the pH and composition of the sample, thereby affecting binding of T4 to TBG or other protein. It is also essential to minimize disruption of the original equilibrium by keeping both the sample dilution and ratio of the volume of buffer to volume of sample as low as possible. It can be predicted that, for serum samples with normal or high binding capacity, the free T4 concentration will not change significantly, even when the sample is diluted more than 100-fold. Figure 13.3 illustrates this in serum samples from a normal subject, and in patients with thyrotoxicosis and hypothyroidism. In contrast, serum free T4 values decline markedly when serum from a patient with low serum-binding capacity (e.g., renal failure or other severe illness) or serum rich in NEFAs is diluted (Fig. 13.3). In these sick patients, the low serum-binding capacity may result from the presence of drugs, fatty acids, or various compounds that accumulate when hepatic or renal function is impaired, or by the synthesis of modified TBG that has a low affinity for T4 (19,22,23,24,25). The effect of dilution on serum free T4 concentrations has been advocated as a useful test to validate commercial assays for serum free T4 (16,26,27,28).
FIGURE 13.3. The effect of sample dilution on serum free thyroxine (T4) concentrations measured by equilibrium dialysis. Serum samples from a normal subject (euthyroid) and patients with thyrotoxicosis (hyperthyroid), a high serum nonesterified fatty acid concentration (high NEFA), renal failure, and hypothyroidism were analyzed in dilutions, ranging from 1:2 to 1:80. The results of the measurements in the serum samples from the normal subject and the patients with thyroid dysfunction did not vary when the samples were diluted up to 80-fold. In contrast, the serum free T4 concentrations decreased with dilution in the samples from a patient with renal failure or a patient who had a high serum nonesterified fatty acid concentration after heparin administration (i.e., low serum-binding capacity). The dashed horizontal lines denote the normal reference range. To convert serum free T4 values to ng/dL, multiply by 0.078.
Validity of Commercial Methods for Measurements of Serum Free Thyroxine and Triiodothyronine
Routine assays for serum free T4 and free T3 have been given a variety of names, depending on the methodology. The first widely available assays were the so-called analogue assays. These assays are based on a T4 analogue that has similar immunoreactivity to T4 but does not bind to the serum-binding proteins. If this is achieved, the analogue should equilibrate only with the free T4 fraction, and therefore the results should accurately reflect the free T4 content of the serum sample. However, while many of the early analogues did not bind to TBG, they did bind to albumin or transthyretin. As a consequence, the free serum T4 results were dependent on the binding capacity of those proteins in the sample (29,30,31,32).
All assays for serum free T4, now in wide use have the following steps in common: The serum sample is diluted, an anti-T4 antibody is added to “capture” a proportion of the free T4 pool, and the unoccupied T4-binding sites on the antibody are then estimated by adding T4 that has been labeled in some way. Sample dilution, temperature, and buffer composition can be controlled to some extent, but these assays have additional problems over those of equilibrium dialysis (16). T4 binding to the antibody tends to disrupt the original equilibrium, an effect that is dependent on the T4-binding capacity of the serum sample, and the affinity and concentration of the anti-T4 antibody. Also, the binding capacity of the serum sample may be quite different from that used in the assay calibrators. The effect of this is that serum samples with low binding capacity (low serum TBG, e.g., nonthyroidal illness) tend to yield serum free T4 values that are lower than the true value, whereas serum samples with high binding capacity (high serum TBG, e.g., pregnancy) tend to yield results that are higher than the true value. Many assays include albumin, added in an attempt to diminish nonspecific binding and nullify the effects of fatty acids on serum free T4 concentrations (33,34). Unfortunately, albumin binds T4 and, if present in sufficient concentrations, will disrupt the equilibrium (35). Serum samples with low binding capacity are more prone to this effect than are samples with normal or high binding capacity. These various effects explain why serum free T4 concentrations may be low, normal, or high in a patient with nonthyroidal illness, depending on which assay was used. Figure 13.4 shows the correlation between the results of measurements of serum free T4 by ED and two serum free T4 assays currently on the market in several groups of patients. The labeled-antibody assay (left panel) contains no added bovine serum albumin, whereas the
P.333

two-step assay (right panel) has albumin included. The results from both methods agree with the results of equilibrium dialysis in serum samples from pregnant women (high serum TBG) and euthyroid ambulatory patients with normal serum binding capacity. However, the results of the labeled antibody assay (no added albumin)—not the results of the two-step assay (albumin added)—agree with those of dialysis in serum samples with low binding capacity from patients with nonthyroidal illness.
FIGURE 13.4. Relationship between measurements of serum free thyroxine (T4) by equilibrium dialysis (ED) and by two commercial methods in ambulatory patients, patients with nonthyroidal illness, and pregnant women. Left panel: The results of measurements by equilibrium dialysis and the Vitros ECi–labeled antibody serum free T4 assay (Ortho Clinical Diagnostics, High Wycombe, Bucks, UK), which has no bovine serum albumin included in the reagents, are similar. Right panel: The results of ED and the AxSYM two-step free T4 assay (Abbott Diagnostics, North Chicago, IL, USA), which includes albumin, disagree markedly in patients with nonthyroidal illness. The ED was performed using the Nichols method (Nichols Institute Diagnostics, San Clemente, CA, USA). To convert serum free T4 values to ng/dL, multiply by 0.078. (From Christofides ND, Wilkinson E, Stoddart M, et al. Serum thyroxine binding capacity-dependent bias in an auomated free thyroxine assay. J Immunoassay 1999;20:201; and Christofides ND, Wilkinson E, Stoddart M, et al. Assessment of serum thyroxine binding capacity-dependent biases in free thyroxine assays. Clin Chem 1999;45:520, with permission.)
The validity of an assay can be assessed in two ways (16,27,28). One is by comparison with a reference method. For any assay, the results should be similar to those obtained using a validated equilibrium dialysis or ultrafiltration method. This applies not only to patients with thyroid dysfunction, but also to patients with a wide range of serum-binding capacities. These should include pregnant women, women taking estrogen, patients with nonthyroidal illness, and patients taking drugs such as furosemide or salicylate that inhibit T4 binding to TBG or other binding proteins. The second validation method is to document that the results are similar when serum samples, particularly those with normal or high binding capacity, are diluted. Because dilution itself may modify nonspecific effects, dilutions of no more than eightfold should be tested. Assays in which the measured serum free T4 values fall substantially at these dilutions are likely to underestimate the true free T4 values in patients with low serum-binding capacity. An example of this dilution test performed using two current assays is shown in Fig. 13.5. Many current assays for serum free T4 fail these validation tests (27,36,37).
FIGURE 13.5. Dilution test used to assess the validity of serum free thyroxine (T4) assay methods. This plot shows the effects of dilution of serum on the measured free T4 concentration in a pool of serum samples from normal pregnant women, who have a high serum thyroxine-binding globulin concentration, as determined by equilibrium dialysis and the two commercial methods shown in Fig. 13.4. Dilution of serum did not alter the results as measured by equilibrium dialysis and the Vitro ECi method (Ortho Clinical Diagnostics, Hugh Wycombe, Bucks, UK), but resulted in underestimation of the serum free T4 concentration (negative free T4 bias) in the Abbott AxSYM two-step serum free T4 assay (Abbot Diagnostics, North Chicago, IL, USA). (From Christofides ND, Wilkinson E, Stoddart M, et al. Serum thyroxine binding capacity-dependent bias in an automated free thyroxine assay. J Immunoassay 1999;20:201, and Christofides ND, Wilkinson E, Stoddart M, et al. Assessment of serum thyroxine binding capacity-dependent biases in free thyroxine assays. Clin Chem 1999;45:520, with permission.)
Nomenclature of Assays for Serum Free Thyroxine and Free Triiodothyronine
Serum free T4 assays have been given a wide range of names, including one-step, analogue, two-step, back-titration, and labeled antibody. The principles of each of these assay have been reviewed in the literature (1,12,16,19,26). The names do not guarantee validity or performance. For example, clearly different serum free T4 concentrations may be found in the same samples assayed using two different assays of the same type (Fig. 13.6).
FIGURE 13.6. Serum free thyroxine (T4) concentrations in patients with chronic renal failure measured by two commercial two-step methods (TS1, TS2). The reference ranges of the two assays were similar, but the results in the patients, who had decreased serum binding of T4, were consistently lower in the TS1 assay. To convert serum free T4 values to ng/dL, multiply by 0.078.
Serum Free Triiodothyronine Assays
Serum free T3 concentrations are lower than those of free T4, and their measurement is much more problematic. Under normal circumstances, approximately 80% of the T3 in serum is derived from the extrathyroidal deiodination of T4 (see Chapter 7). In patients with nonthyroidal illness, tissue uptake and deiodination of T4 are decreased, and therefore serum total T3 concentrations fall (see section on nonthyroidal illness in Chapter 11). However, for
P.334

serum free T3 other factors come into play, including modification of serum T3-binding capacity, which tends to increase the free T3 fraction (14,15). Although results from most commercial methods indicate that serum free T3 concentrations are low in these patients, the concentrations as measured by ultrafiltration or some labeled antibody methods may be normal or occasionally even slightly high (38,39).
Serum Free Thyroxine and Free Triiodothyronine Indexes
Serum free T4 and free T3 can be estimated indirectly as the serum free T4 index and the serum free T3 index. These index methods require measurement of serum total T4 or total T3 (see later in the chapter) together with measurement of the thyroid hormone–binding ratio (THBR, also known as the T3-resin uptake) (1). The THBR is an estimate of the number of unoccupied thyroid hormone–binding sites in a serum sample, as compared with the number in normal subjects. The THBR is usually measured by adding labeled T3 (T4 can be used instead) and an insoluble hormone-binding resin to the serum sample. The proportion of labeled T3 that binds to the resin (resin uptake) is then measured. The ratio of the resin uptake in the presence of the patient’s serum, as compared with that in a pool of normal serum, is the THBR. In patients with thyrotoxicosis there are few unoccupied serum-binding sites, and proportionately more labeled T3 binds to the resin, whereas in pregnant women, in whom there are many unoccupied binding sites, proportionately less labeled T3 binds to the resin. The serum free hormone index is then calculated by multiplying the serum total hormone concentration by the THBR (1). In thyrotoxicosis, both the serum THBR and total T4 values are high, giving an appropriately high serum free T4 index. In pregnancy, although the serum total T4 value is high, the THBR is low, and the serum free T4 index is normal. Sometimes, the results are expressed as the percentage of labeled T3 bound to the resin, and the serum free T4 index is calculated as the product of the serum total T4 concentration and the percentage T3-resin uptake.
In a similar test, serum total T4 and TBG are measured separately by immunoassays, and a serum T4/TBG ratio is calculated. This ratio fails to take into account changes in
P.335

the other thyroid hormone–binding proteins, although, as discussed earlier, even large changes in the concentration or affinity of transthyretin TTR or albumin have relatively little effect on the free T4 (or free T3) concentration in serum. More importantly, modifications in the binding capacity or affinity of TBG that may occur in patients taking certain drugs or those with nonthyroidal illness are not accounted for by the serum T4/TBG ratio (23,40).
Furthermore, the ratio does not accurately describe the serum free T4 concentration, because the calculation uses the serum total TBG concentration rather the concentration of unoccupied binding sites; the relationship between the serum T4/TBG ratio and serum total T4 concentrations is linear, whereas the relationship between serum free T4 and total T4 concentrations is curvilinear. At high serum total T4 concentrations, there are few unoccupied T4 binding sites on TBG, and thus the T4/TBG ratio underestimates the serum free T4 concentration. The converse is true when serum total T4 concentrations are low (16). These indirect tests have largely been replaced by direct measurements of serum free T4 and free T3.
Introduction of a Method for Assay of Serum Free Thyroxine and Free Triiodothyronine
Clinicians should inform themselves on how well the serum free T4 and free T3 assays used by the laboratory perform in patients with thyroid disorders, pregnant women, patients with nonthyroidal illness, patients taking drugs that alter serum protein binding of T4 or patients who have abnormalities in the production of one of the binding proteins (see section on effects of drugs and other substances on thyroid hormone synthesis and metabolism in Chapter 11). Laboratories should take appropriate action to minimize sample deterioration during storage, for example, by freezing samples that cannot be assayed within 24 hours of collection.
MEASUREMENT OF SERUM TOTAL THYROXINE AND TOTAL TRIIODOTHYRONINE
It is much easier to measure serum total T4 and T3 than free T4 and T3. To do this, T4 or T3 is displaced from the binding proteins using various chemical agents, and the released hormone is measured using a competitive immunoassay. The technical problems associated with measurements of serum total T4 and total T3 are limited to assay interference and calibration issues. The purity of the standard, the type of matrix (serum) used to prepare the calibrators, and the lack of an international reference method are important problems (1).
In clinical practice, most problems associated with measurements of serum total T4 (and total T3) arise in patients with abnormalities in serum hormone binding that may increase or decrease the serum total hormone concentration (41). Among these abnormalities, the most common is the increase in serum TBG concentrations, and therefore the increases in serum total T4 and total T3 concentrations, that occur in pregnant women and women taking any form of estrogen. The estrogen analogue tamoxifen has a small effect, and raloxifene has very little effect, on serum TBG concentrations. Genetic variants of TBG, trans thyretin, and albumin have been described to have altered T4-binding characteristics, which in turn alter serum total T4 and total T3 concentrations (see Chapter 6) (41). In patients with nonthyroidal illness, the TBG that is produced may have lower affinity for T4 and T3 than normal TBG. Many drugs may also lower the T4-binding capacity of serum. Some patients produce autoantibodies to T4 or T3, which increase the hormone-binding capacity of serum and interfere with the assays for the respective hormone (42).
The use of serum total T4 and total T3 measurements is becoming less popular with the advent of more reliable methods for measurements of serum free T4 and free T3. However, the results of measurements of serum total T4 are still considered to have a role in assessing thyroid status in some patients with nonthyroidal illness in whom serum free T4 and TSH values do not concur (1).
EFFECTS OF DRUG TREATMENT
Drugs may modify secretion of TSH, or the production, secretion, transport and metabolism of T4 and T3 (Table 13.1)(43,44). Some drugs alter thyroid secretion, whereas others result in abnormal thyroid function test results in otherwise euthyroid subjects in other ways (see section on the effects of drugs and other substances on thyroid hormone synthesis and metabolism in Chapter 11).
TABLE 13.1. DRUGS THAT ALTER THYROID FUNCTION AND THE RESULTS OF THYROID FUNCTION TESTS

Decreased TSH Secretion Decreased Thyroid Hormone Secretion Increased Thyroid Hormone Secretiona Decreased Thyroid Hormone Synthesisa

Dopamine Iodide Iodide Methimazole
Glucocorticoids Lithium Lithium Carbimazole
Octreotide Amiodarone Amiodarone Propylthiouracil
  Thalidomide   Lithium
High Serum TBG, Total T4, Total T3b Low Serum TBG, Total T4, Total T3b Displacement of T4 from Serum Proteins Increased Hepatic Metabolism of T4b

Estrogen Androgen Furosemide Phenytoin
Tamoxifen Anabolic steroids Salicylates Carbamazepine
Heroin Glucocorticoids Fenclofenac Phenobarbital
Methadone   Mefenamic acid Rifampin
Impaired T4 to T3 Conversion Impaired Absorption of T4c    

β-Adrenergic antagonists Calcium carbonate    
  Cholestyramine    
Glucocorticoids Colestipol    
Amiodarone Aluminium hydroxide    
Propylthiouracil Ferrous sulfate    
Iopanoic acid Sucralfate    

aModify thyroid hormone synthesis and thus may alter thyroid status.
bModify the requirement for T4 in patients with hypothyroidism.
cInterfere with T4 absorption from the gastrointestinal tract. Patients taking T4 should be advised against taking these drugs and T4 at the same time.
T3, triiodothyronine; T4, thyroxine; TBG, thyroxine-binding globulin; TSH, thyrotropin.
In general, serum TSH concentrations are less affected by drugs than are serum T4 and T3 concentrations. High doses of glucocorticoids and dopaminergic drugs inhibit TSH release acutely, but do not cause hypothyroidism (45,46). High doses of propranolol and other β-adrenergic antagonist drugs (but not all of them), and glucocorticoids decrease serum T3 concentrations slightly by inhibiting the extrathyroidal conversion of T4 to T3 (47).
Phenytoin, carbamazepine, and phenobarbital accelerate the clearance of T4 and T3, and may lower serum free T4 and free T3 concentrations slightly. High doses of furosemide, salicylates, and some other nonsteroidal antiinflammatory drugs competitively inhibit T4 and T3 binding to serum TBG and may transiently raise serum free T4 and free T3 concentrations (22,48,49,50,51). In vivo administration
P.336

of heparin liberates NEFAs, which may then displace T4 and T3 from the binding proteins, and therefore increase serum free T4 and free T3 concentrations; the effect of these drugs on measured serum free T4 (and free T3) concentrations is very method dependent (52,53,54).
Drugs that alter the serum concentration or binding capacity of the thyroid hormone–binding proteins (particularly TBG) will alter serum total T4 and total T3 concentrations, but not serum free T4 and free T3 concentrations if the hypothalamic-pituitary-thyroid axis is intact.
REFERENCE RANGES
By convention, a reference range usually comprises 95% of a reference population. Thus, whatever is being measured, 2.5% of “normal” subjects will have values above the reference range, and 2.5% will have values below the range. The distribution of serum total and free T4 and total and free T3 concentrations is normal in reference populations. In contrast, for serum TSH the distribution is log normal, even when subjects who have high serum antithyroid peroxidase antibody concentrations are excluded from the reference population (55). Serum TSH, but not serum free T4 and free T3, concentrations vary twofold at different times of the day, peak values occurring during the night (between 2300 and 0200 hours) and nadir values during the day (between 1000 and 1600 hours) (56). TSH secretion is also pulsatile, with about 8 to 12 small pulses per 24 hours. For serum TSH, reference ranges should be established using specimens collected between 0800 and 1800 hours and using 95% confidence limits from log-transformed data. The reference population should have no personal or family history of thyroid disease, have normal serum antithyroid peroxidase antibody concentrations, and should not be taking medications known to alter TSH or thyroid secretion or serum binding proteins (1). Clearly, method-related reference ranges need to be used for all thyroid function tests. Manufacturers’ reference ranges should be confirmed locally in at least 120 normal subjects (1). Typical adult reference ranges are serum TSH, 0.3 to 4.0 mU/L; serum total T4, 5 to 12 μg/dL (65 to 155 nmol/L); serum total T3, 65 to 170 ng/dL (1.0 to 2.6 nmol/L); serum free T4, 0.7 to 2.0 ng/dL (9 to 25 pmol/L); and serum free T3, 0.2 to 0.5 ng/dL (3 to 8 pmol/L).
The results of all these measurements are higher in neonates, infants, and children, and therefore age-related reference ranges are needed (see Chapter 74) (1,57). Also, serum TSH and free and total T4 and T3 concentrations all change during pregnancy, and therefore trimester-related reference ranges should be available for each of these analytes (see Chapter 80) (58,59).
P.337

CHANGES IN THYROID FUNCTION TESTS
Knowledge of the analytical variability and estimates of the intra-individual and inter-individual variability allows estimation of the differences in thyroid tests that can be considered as clinically important when repeated measurements are made during thyroid hormone or antithyroid treatment (60,61). For serum TSH, total T4, and total T3, these are approximately 0.8 mU/L, 1.2 μg/dL (15 nmol/L), and 40 ng/dL (0.6 nmol/L), respectively. For serum free T4 and free T3, these values are probably about 0.5 ng/dL (6 pmol/L) for free T4 and 0.1 ng/dL (1.5 pmol/L) for free T3, but many of these results were determined using older assays, and these data may not be applicable to the newer assays.
ASSAY INTERFERENCE
All assays are prone to interference from a range of substances in serum, including human anti-animal (heterophil) antibodies, antithyroid antibodies, and rheumatoid factor. For example, if mouse monoclonal antibodies are used in an immunometric assay for measurements of serum TSH, then antimouse antibodies in a serum sample might bind to both the capture antibody and signal antibody and form a bridge between the antibodies in a similar manner to TSH, resulting in a spuriously high serum TSH value. Assay manufacturers try to minimize such interference by adding mouse immunoglobulins to the reagents in the hope that they will neutralize any antimouse antibodies and prevent bridging. Unfortunately, in some serum samples the concentration of these antimouse antibodies may be so high that interference still occurs (42,62,63,64,65). Laboratories should have protocols available to determine if results are analytically valid. Such protocols could include determining if the result shows linearity on dilution (not valid for free hormone measurements), testing for the presence of heterophil antibodies, antimouse antibodies, and antithyroid hormone antibodies, and confirmation by an alternative assay method (66).
THYROID HORMONE–BINDING PROTEINS
The thyroid hormone–binding proteins TBG, trans thy retin, and albumin transport the hydrophobic thyroid hormones in serum and provide a buffer to stabilize free T4 and free T3 for cell uptake (41). TBG is a member of the serine protease inhibitor (serpin) family of proteins, and it may facilitate the transport of maternal T4 to the fetus. Most structural variants of TBG have either normal or diminished affinity for T4 and T3 (see Chapter 6.) Genetic variants of albumin may have a markedly increased affinity for T4, but not T3, giving rise to a disorder known as familial dysalbuminemic hyperthyroxinemia. Patients with this disorder have high serum total T4 concentrations, but normal serum total T3, free T3, and TSH concentrations. Some analogue-based methods for measuring serum free T4 give high values due to assay artefact, whereas serum free T4 concentrations measured by equilibrium dialysis or two-step assays are normal (67). A much less common mutation in albumin has been described that results in an increased affinity for T3 but not T4 (68). Several mutations in transthyretin (TTR) have been described which result in high serum total T4 concentrations (24,69,70).
Electrophoresis of serum after the addition of radiolabeled T3 or T4 allows the identification of abnormal proteins that have high affinity for either hormone. Incubation of serum with radiolabeled T3 or T4 and precipitation of the gammaglobulin fraction using 20% polyethylene glycol is a simple way to detect serum anti-T3 and anti-T4 antibodies (71).
INTERPRETATION OF THE RESULTS OF THYROID FUNCTION TESTS
Thyrotoxicosis
With the exception of the rare cases of thyrotoxicosis caused by a TSH-secreting pituitary tumor, serum TSH concentrations are low, usually < 0.02 mU/L, in patients with thyrotoxicosis. Both serum total and free T3 and T4 concentrations are high in over 95% of these patients. Those patients with thyrotoxicosis caused by Graves’ disease are usually more clinically thyrotoxic than those with nodular goiter, and this is reflected in higher serum total and free T4 and T3 concentrations. A typical patient with Graves’ thyrotoxicosis has a serum free T4 concentration of 3 to 4 ng/dL (40 to 52 pmol/L), a serum total T4 concentration of 15 to 17 μg/dL (190 to 220 nmol/L), a serum free T3 concentration of 0.8 to 1.0 ng/dL (12 to 15 pmol/L), and a serum total T3 concentration of 350 to 450 ng/dL (5.4 to 7.0 nmol/L). The corresponding values in a patient with thyrotoxicosis caused by a nodular goiter are likely to be about 25% to 30% lower. In younger patients, there is a reasonably good correlation between serum free T4 and free T3 concentrations and symptoms and signs of thyrotoxicosis, but the correlation is weak in older patients.
Triiodothyronine Thyrotoxicosis
About 2% to 4% of patients with thyrotoxicosis have high serum T3 concentrations, but their serum T4 concentrations are within the reference range (usually the upper part of that range); they are said to have T3-thyrotoxicosis. The incidence of T3-thyrotoxicosis is higher when serum total rather than free T4 is measured (72). It is most likely to be
P.338

found in patients who live in areas of iodine deficiency and in patients with nodular goiter, those receiving antithyroid drug therapy, and those with recurrent thyrotoxicosis after subtotal thyroidectomy or antithyroid drug therapy.
Subclinical Thyrotoxicosis
The combination of a low serum TSH concentration and normal serum free T4 and free T3 concentrations is known as subclinical thyrotoxicosis. This condition reflects the sensitivity of the pituitary thyrotrophs to the inhibitory action of small increases in serum T4 and T3 concentrations (see section on regulation of thyrotropin secretion in Chapter 10) (73). An absence of symptoms was once part of the definition of subclinical thyrotoxicosis, but, in fact, subtle clinical features of thyrotoxicosis may be present (see Chapter 79). Also, subclinical thyrotoxicosis is a risk factor for atrial fibrillation (74,75) and low bone mineral density (76). These biochemical findings may also be present in patients with nonthyroidal illness or central hypothyroidism (Table 13.2)(77).
TABLE 13.2. CAUSES OF LOW SERUM THYROID-STIMULATING HORMONE CONCENTRATIONS

  Serum Triiodothyronine
Serum Thyroxine
Condition or Factor Free Total Free Total

Endogenous subclinical thyrotoxicosis High normal High normal High normal High normal
Exogenous subclinical thyrotoxicosis (T4 therapy) Normal Normal High normal or high High normal or high
Nonthyroidal illness Normal, low, or higha Normal or low Normal, low, or highb Normal, low, or highc
Central hypothyroidism Normal or low Normal or low Low normal or low Low normal or low

aIn patients with nonthyroidal illness, serum free T3 concentrations may be high if a particular commercial assay system (e.g., Vitros ECi, Ortho Clinical Diagnostics, High Wycombe, Bucks, UK) is used.
bThe values depend on both the severity of the illness and the method of measurement used.
cThe values depend on the type and severity of the illness.
Adapted from Toft AD. Subclinical hyperthyroidism. N Engl J Med 2001;345:512, with permission. Copyright 2001 Massachusetts Medical Society.
In the absence of clinical manifestations of thyroid disease, and even after additional studies such as measurements of thyroid radioiodine uptake or TSH-receptor antibodies, it may be difficult to decide whether patients with low serum TSH concentrations and normal serum free T4 and T3 concentrations have subclinical thyrotoxicosis or nonthyroidal illness. In such circumstances, thyroid function tests should be repeated several weeks later; a normal or high serum TSH concentration at this time suggests recovery from nonthyroidal illness or the hypothyroid phase of thyroiditis. If the pattern persists, the choices are a trial of antithyroid drug therapy or close clinical follow-up.
Factitious Thyrotoxicosis
In this condition, excessive quantities of thyroid hormone are taken without the knowledge of the physician, usually by women wishing to lose weight. In all cases, serum TSH concentrations are low. Serum concentrations of T3 and T4 vary, depending upon the thyroid hormone preparation being taken. Ingestion of T4 alone results in a greater rise in serum total and free T4 than in total and free T3 concentrations, so that the serum T4:T3 ratio is higher in these patients than in those with spontaneously occurring thyrotoxicosis; for example, a patient might have a serum free T4 concentration of 4.6 ng/dL (60 pmol/L) but a serum free T3 concentration of 0.9 ng/dL (1.4 pmol/L). If T3 is taken, serum total and free T4 concentrations are low, whereas serum total and free T3 concentrations are high (how high will depend on when the patient last took T3). If thyroid extract is taken, the serum T3 and T4 concentrations will depend upon the relative proportions of the two thyroid hormones in the tablets, which may vary from batch to batch. If factitious thyrotoxicosis is suspected, serum thyroglobulin should be undetectable, in contrast to all other forms of thyrotoxicosis (see Chapters 14 and 29) (78).
Changes in Serum Thyroxine and Triiodothyronine Concentrations during Treatment of Thyrotoxicosis
In patients with thyrotoxicosis, serum T4 and T3 concentrations are usually normal within 6 to 8 weeks after iodine-131 therapy and in 3 to 4 weeks after initiation of antithyroid drug therapy (79). There is a delay, however, in the recovery of TSH secretion; as a result, serum TSH concentrations may remain low for several weeks or even a few months after serum T4 and T3 concentrations have fallen to within, or even below, their respective normal ranges (see Chapter 45) (80). It follows that measurements of serum free T4 are a better indicator of the patient’s clinical status than measurements of serum TSH during the first few months of any type of antithyroid therapy. After recovery
P.339

of TSH secretion, measurements of serum TSH are the best way to gauge the effect of treatment for thyrotoxicosis.
Predicting Relapse of Thyrotoxicosis
No single test of thyroid function is reliable in predicting relapse of thyrotoxicosis in patients with Graves’ disease (see Chapter 45). On a group basis, antithyroid drug treatment can be expected to induce long-term remission in many patients with mild thyrotoxicosis [serum total T3 concentration less than approximately 300 ng/dL (4.5 nmol/L)] and a small goiter, but not in patients with a large goiter and a higher serum T3 concentration (81). Patients with Graves’ thyrotoxicosis who have a serum T3:T4 ratio of > 20, calculated in nanograms per micrograms (> 0.024 calculated on a molar basis) that persists during antithyroid drug treatment are unlikely to have a remission (82).
Pregnancy
In normal pregnant women, serum TSH concentrations fall slightly during the first trimester; up to 20% of these women may have values below the normal reference range at that time Chapter 80 (83). Therefore, trimester-related reference values should be used (1). This decrease in serum TSH concentrations is due to the weak thyroid-stimulating activity of chorionic gonadotropin, which has structural homology with TSH (see section on chemistry and biosynthesis of thyrotropin in Chapter 10). In a few pregnant women, the resulting increase in serum free T4 concentrations may be sufficient to cause clinical mani festations of thyrotoxicosis (gestational thyrotoxicosis). Gestational thyrotoxicosis is frequently associated with hyperemesis, in which serum chorionic gonadotropin concentrations tend to be high (see Chapter 26) (84,85). Serum chorionic gonadotropin concentrations also are higher in women with multiple pregnancies, as compared with those who have singleton pregnancies, and they are therefore more likely to have high serum free T4 concentrations (and low serum TSH concentrations) during the first trimester.
While serum free T4 and free T3 concentrations rise slightly in response to the thyroid-stimulating action of chorionic gonadotropin during the first trimester, they then fall. As for serum TSH, it is wise, as noted earlier, to use trimester-related reference ranges in evaluating the results of these measurements, particularly in pregnant women with thyrotoxicosis (1,58,59). In the past, good control of thyrotoxicosis consisted of maintaining a normal serum TSH concentration in the mother; however, but there is now evidence that mothers with no thyroid disease and who have a serum free T4 concentration in the lowest tenth percentile in the first trimester have children with lower scores on neurodevelopment tests, as compared with children of mothers who have higher serum free T4 concentrations (see Chapter 74) (86). Pregnant women with thyrotoxicosis, therefore, should be given the lowest dose of antithyroid drug which results in a low normal serum TSH concentration and a serum free T4 concentration in the upper part of the reference range throughout pregnancy.
Hypothyroidism
Serum TSH concentrations are usually in excess of 20 mU/L in patients with overt primary hypothyroidism, defined as high serum TSH and low T4 concentrations, whereas most patients with serum TSH concentrations between 5 and 20 mU/L have normal serum T4 concentrations, albeit in the lower part of reference range (subclinical hypothyroidism) (see Chapter 78). Markedly high serum TSH values of 500 mU/L and higher are usually found only in young patients who have severe, long-standing hypothyroidism. The increase in TSH secretion in these patients is accompanied by hypertrophy and hyperplasia of the thyrotrophs, which occasionally is sufficiently intense to cause enlargement of the pituitary (see Chapter 3). Otherwise, there is little relationship between serum TSH concentrations and clinical findings in patients with hypothyroidism (87). Patients with central hypothyroidism have normal or low serum TSH concentrations (see Chapter 51). Measurements of serum T3 are not indicated in evaluating patients with hypothyroidism, because the values are normal in a substantial proportion of the patients.
Subclinical Hypothyroidism
Subclinical hypothyroidism is defined as high serum concentrations of TSH and normal serum concentrations of T4 and T3. The serum TSH concentrations are usually between 5 and 20 mU/L. Most patients are asymptomatic, but some have nonspecific symptoms compatible with hypothyroidism, such as tiredness and lack of energy (see Chapter 78). Subclinical hypothyroidism represents the mildest form of thyroid failure; whether these patients should be treated is debated. Those favoring therapy are most influenced by the knowledge that 25% to 50% feel better when taking T4 and by the evidence that approximately 3 to 5% have progression to overt hypothyroidism each year. Progression is most likely in patients who have a high serum antithyroid peroxidase antibody concentration (88).
Changes in Serum Thyroxine and TSH Concentrations during Treatment of Hypothyroidism
After initiation of T4 therapy in patients with hypothyroidism, serum TSH concentrations fall slowly as serum T4 concentrations rise. For example, in hypothyroid patients with serum TSH concentrations of 100 to 200 mU/L, treatment with T4 in a dose of 100 μg daily results in a fall
P.340

in serum TSH concentrations to near normal only after 6 to 8 weeks, whereas serum T4 concentrations increase to within the normal range in about 2 weeks (89). The appropriate dose of T4 is that which restores the patient to the euthyroid state with a serum TSH concentration at the lower part of the reference range; when this is done, serum free T4 concentrations are usually in the upper part of its reference range or slightly high, e.g., serum TSH 0.5 mU/L and serum free T4 1.6 ng/dL (20 pmol/L) (see Chapter 67). Once the maintenance dose of T4 is established, it remains relatively constant. However, some medications reduce the absorption of T4 (Table 13.1). Patients should be advised not to take these medications within 4 hours after ingestion of T4. Drugs such as phenytoin and carbamazepine, which increase hepatic metabolism of T4, may result in the need for a higher dose of T4 as well. Sertraline, by an unknown mechanism, may also increase the requirement for T4 (90,91). The rise in serum TBG concentrations that occurs in pregnant women and those given estrogen results in the need for more T4; conversely, the fall in serum TBG concentrations that occurs in patients given androgen results in the need for less T4. A high serum TSH concentration in a patient taking T4 who had been previously well controlled does not necessarily imply poor compliance; it should raise the suspicion of whether the patient is taking a new medication.
One biochemical combination that is very suggestive of poor compliance with T4 therapy is a high serum free T4 concentration and a high serum TSH concentration. These results indicate that the patient, having been erratic in tablet-taking, consumed high doses of T4 in the few days before the clinic visit.
Many patients taking T4 have a low serum TSH concentration and feel better than when the concentration is normal. There is little evidence that this exogenous form of subclinical thyrotoxicosis is detrimental, as long as the patient’s serum T3 concentration is normal (92).
Nonthyroidal Illness
Patients who have any of a wide range of acute or chronic nonthyroidal illnesses may have abnormalities in thyroid function tests, even though they are clinically euthyroid. The majority of these patients have normal serum TSH concentrations, and therefore this measurement provides the best guide to thyroid status in sick people, as it does in more healthy people. However, some acutely ill patients have low serum TSH concentrations, and then during recovery their concentrations may be slightly high (5 to 20 mU/L) for a few days (93,94). Serum total and free T3 concentrations usually fall as a result of impaired tissue uptake of T4 and conversion of T4 to T3 (15). However, serum free T3 concentrations measured by ultrafiltration of undiluted serum (38) or some commercial assays may be normal or high in these patients (39). The serum concentrations and binding capacity (or affinity) of thyroid hormone-binding proteins tend to be reduced in patients with nonthyroidal illness (95), which in turn tends to reduce serum total T4 and T3 concentrations and raise the free fraction of serum T4 and T3. The contribution of each of these mechanisms varies with the type, severity, and stage of the illness, and therefore the results of thyroid function tests may be extremely variable; they in fact may mimic the changes that occur in patients with thyroid dysfunction. In hospitalized patients, a low serum TSH value is at least twice as likely to be caused by nonthyroidal illness as thyrotoxicosis, whereas a high serum TSH concentration is as likely to be associated with recovery from illness as hypothyroidism (93). The poor predictive value of thyroid function tests in hospitalized patients indicates that these tests should be requested only in patients who have some clinical manifestations of thyroid dysfunction, such as unexplained atrial fibrillation or hypothermia. In such patients, low serum total or free T3 concentrations strongly suggest nonthyroidal illness, and repeated measurements after the patients have recovered from the illness will determine whether the abnormal results were caused by the illness or by thyroid dysfunction (see section on nonthyroidal illness in Chapter 11) (96).
PHILOSOPHY OF THYROID FUNCTION TESTING
Measurement of serum TSH is the best single test of thyroid function (4). The sensitivity of this measurement is such that some organizations have recommended that all adults be screened for thyroid disease using TSH, beginning at age 35 years and every 5 years thereafter (97). However, the yield will be low (see Chapter 19). Furthermore, it is important to note that measurements of serum TSH as a first-line test may yield misleading information, for example, missing patients with central hypothyroidism in whom serum TSH concentrations may be normal or even high (98), and those admittedly rare patients with thyrotoxicosis caused by a TSH-secreting pituitary tumor or thyroid hormone resistance (see Chapter 24 and see Chapter 81) (Table 13.3). On the other hand, patients with low serum TSH concentrations as a result of nonthyroidal illness may be mistakenly diagnosed as having central hypothyroidism. Measurements of serum TSH are widely used for screening newborn infants for hypothyroidism (see section on neonatal screening in Chapter 75).
TABLE 13.3. CLINICAL SITUATIONS IN WHICH MEASUREMENTS OF SERUM THYROTROPIN ALONE MAY YIELD MISLEADING RESULTS

Condition Serum TSH Consequences of Clinical Action Based on Serum TSH Value Alone Serum Free T4 (If Measured)

Heterophil antibodies Normal Failure to diagnose thyrotoxicosis High
Central hypothyroidism Normala Failure to diagnose hypothyroidism and investigate hypothalamic-pituitary structure and function Low
TSH-secreting pituitary adenoma Normala Failure to diagnose thyrotoxicosis and investigate pituitary structure and function High
Thyroid hormone resistance Normala Failure to recognize the condition High
Poor compliance with T4 therapy High Inappropriate increase in dose of T4 High
Delayed recovery of TSH secretion Normal or low Failure to diagnose impending hypothyroidism Low

aSerum TSH concentrations may be high in these conditions, which should prompt measurements of serum free T4 and further investigation. T4, thyroxine; TSH, thyrotropin.
Isolated TSH deficiency is rare; impaired thyrotroph function is usually accompanied by growth hormone and gonadotropin deficiencies, although corticotropin secretion is less frequently affected. Clinical suspicion of pituitary
P.341

or hypothalamic disease should override any misleading biochemical findings (99). However, such a policy not only gives undue weight to the clinical skills of a profession that is more reliant than ever before on biochemical tests, but also fails to appreciate that the majority of requests for thyroid function tests originate in primary care. Few primary care physicians have much experience in the diagnosis and treatment of central hypothyroidism, the symptoms of which are often nonspecific. In this respect, recent reports cast considerable doubt on the wisdom of adopting measurement of serum TSH alone as the first-line test of thyroid function (100,101). Together, these reports describe a series of 21 patients with pituitary or hypothalamic disease in whom hypothyroidism was manifested by low serum T4 but normal serum TSH concentrations. Although some of the patients had clinical signs of pituitary disease, and almost 50% had a pituitary tumor, the diagnosis of hypopituitarism was unsuspected in the majority. The study from Liverpool, United Kingdom (101), concluded that the incidence of partial or complete hypopituitarism was 55 cases per million per year, severalfold higher than that recognized previously (102). If translated to the United Kingdom as a whole, there would be some 2500 new cases of central hypothyroidism each year, of which some 60% would not be detected by either clinical examination or measurement of serum TSH.
The main reason to measure serum TSH alone is to control laboratory expenditure. However, the issue of cost of thyroid function tests is complex and depends upon factors such as total laboratory workload, the costs of reagents for particular tests, discounts given by diagnostic laboratories, and degree of laboratory automation. Restricting assessment to measurements of serum TSH alone, rather than measurements of serum TSH and free T4, is unlikely to result in the savings anticipated. Indeed, any savings realized by not measuring serum free T4 at the initial stage are minor when set against the potential costs of misdiagnosis. If both serum TSH and free T4 are measured at the outset and interpreted correctly, it should be possible to detect all causes of thyroid dysfunction or prompt measurement of serum T3 to identify nonthyroidal illness or assay interference (103).
REFERENCES
1. Baloch Z, Carayon P, Conte-Devolx B, et al. Laboratory medicine practice guidelines. Laboratory support for the diagnosis and monitoring of thyroid disease. Thyroid 2003;13:3.
2. Dayan CM. Interpretation of thyroid function tests. Lancet 2001;357:619.
3. Seth J, Kellett HA, Caldwell G, et al. A sensitive immunoradiometric assay for serum thyroid stimulating hormone: a replacement for the thyrotrophin releasing hormone test? BMJ 1984; 289:1334.
4. Caldwell G, Kellett HA, Gow SM, et al. A new strategy for thyroid function testing. Lancet 1985;i:1117.
5. Hall R. The need for more sensitive assays for the measurement of thyroid-stimulating hormone (TSH) and the long-acting thyroid stimulator (LATS). Clin Endocrinol (Oxf) 1974;3:361.
6. Spencer CA, Takeuchi M, Kazarosyan M. Current status and performance goals for serum thyrotropin (TSH) assays. Clin Chem 1996;42:140.
7. Nicoloff JT, Spencer CA. The use and misuse of the sensitive thyrotropin assays. J Clin Endocrinol Metab 1990;71:553.
8. Spencer CA, LoPresti JS, Patel A, et al. Applications of a new chemiluminometric thyrotropin assay to subnormal measurement. J Clin Endocrinol Metab 1990;70:453.
9. Spencer CA, Takeuchi M, Kazarosyan M, et al. Interlaboratory/intermethod differences in functional sensitivity of immunometric assays of thyrotropin (TSH) and impact on reliability of measurement of subnormal concentrations of TSH. Clin Chem 1995;41:367.
10. McConway MG, Chapman RS, Beastall GH, et al. How sensitive are immunometric assays for thyrotropin? Clin Chem 1989; 35:289.
11. Spencer CA, Schwarzbein D, Guttler RB, et al. Thyrotropin (TSH)-releasing hormone stimulation test responses employing third and fourth generation TSH assays. J Clin Endocrinol Metab 1993;76:494.
12. Ekins R. The free hormone hypothesis and measurement of free hormones. Clin Chem 1992;38:1289.
P.342

13. Ekins R. Measurement of free hormones in blood. Endocr Rev 1990;11:5.
14. Kaptein E. Nonthyroidal illness. In: Henneman G KE, ed. Thyroid international. Darmstadt: Merck KG, 1998.
15. Beckett GJ, Wilkinson E. Thyroid hormone metabolism in non-thyroidal illness. CPD Bulletin of Clinical Biochemistry 1998;1:9.
16. Christofides N. Free analyte immunoassay. In: Wild D, ed. The immunoassay handbook. 2nd ed. London: Nature Publishing Group, 2001:61.
17. Faber J, Waetjen I, Siersbaek-Nielsen K. Free thyroxine measured in undiluted serum by dialysis and ultrafiltration: effects of non-thyroidal illness, and an acute load of salicylate or heparin. Clin Chim Acta 1993;223:159.
18. Tikanoja S. Ultrafiltration devices tested for use in a free thyroxine assay validated by comparison with equilibrium dialysis. Scand J Clin Lab Invest 1990;50:663.
19. Stockigt J. Free thyroid hormone measurement. A critical appraisal. Endocrinol Metab Clin North Am 2001;30:265.
20. Nelson JC, Wilcox RB. Analytical performance of free and total thyroxine assays. Clin Chem 1996;42:146.
21. Holm SS, Andreasen L, Hansen SH, et al. Influence of adsorption and deproteination on potential free thyroxine reference methods. Clin Chem 2002;48:108.
22. Wang R, Nelson JC, Wilcox RB. Salsalate and salicylate binding to and their displacement of thyroxine from thyroxine-binding globulin, transthyrin, and albumin. Thyroid 1999;9: 359.
23. Hennemann G, Everts ME, de Jong M, et al. The significance of plasma membrane transport in the bioavailability of thyroid hormone. Clin Endocrinol (Oxf) 1998;48:1.
24. Lim CF, Stockigt JR, Curtis AJ, et al. A naturally occurring furan fatty acid enhances drug inhibition of thyroxine binding in serum. Metabolism 1993;42:1468.
25. Lim CF, Stockigt JR. Influence of uremic toxins and nonesterified fatty acids on drug and thyroid hormone binding in serum. Clin Chem 1998;44:2380.
26. Ekins R. Analytic measurements of free thyroxine. Clin Lab Med 1993;13:599.
27. Christofides ND, Wilkinson E, Stoddart M, et al. Serum thyroxine binding capacity-dependent bias in an automated free thyroxine assay. J Immunoassay 1999;20:201.
28. Christofides ND, Wilkinson E, Stoddart M, et al. Assessment of serum thyroxine binding capacity-dependent biases in free thyroxine assays. Clin Chem 1999;45:520.
29. Bayer MF. Free thyroxine results are affected by albumin concentration and nonthyroidal illness. Clin Chim Acta 1983;130:391.
30. Csako G, Zweig MH, Benson C, et al. On the albumin dependence of measurements of free thyroxin. I. Technical performance of seven methods. Clin Chem 1986;32:108.
31. Csako G, Zweig MH, Benson C, et al. On the albumin-dependence of measurements of free thyroxin. II. Patients with non-thyroidal illness. Clin Chem 1987;33:87.
32. Csako G, Zweig MH, Glickman J, et al. Direct and indirect techniques for free thyroxin compared in patients with nonthyroidal illness. I. Effect of free fatty acids. Clin Chem 1989;35: 102.
33. Wilkins TA. Albumin in analog FT4 assay reagents: the facts. Clin Chem 1987;33:1293.
34. Wilkins TA, Midgley JE. Albumin-dependence of free thyroxin in nonthyroidal illness. Clin Chem 1987;33:1494.
35. Ekins R. One-step, labeled-antibody assay for measuring free thyroxine. I. Assay development and validation. Clin Chem 1992;38:2355.
36. Sapin R. Serum thyroxine binding capacity-dependent bias in five free thyroxine immunoassays: assessment with serum dilution experiments and impact on diagnostic performance. Clin Biochem 2001;34:367.
37. Sapin R, d’Herbomez M. Free thyroxine measured by equilibrium dialysis and nine immunoassays in sera with various serum thyroxine-binding capacities. Clin Chem 2003;49: 1531.
38. Faber J, Siersbaek-Nielsen K. Serum free 3,5,3′-triiodothyronine (T3) in non-thyroidal somatic illness, as measured by ultrafiltration and immunoextraction. Clin Chim Acta 1996;256: 115.
39. Sapin R, Schlienger JL, Kaltenbach G, et al. Determination of free triiodothyronine by six different methods in patients with non-thyroidal illness and in patients treated with amiodarone. Ann Clin Biochem 1995;32:314.
40. Wiersinga WM. Thyroid hormone metabolism in non-thyroidal illness. Curr Opin Endocrinol Diabetes 1996;3:422.
41. Schussler GC. The thyroxine binding proteins. Thyroid 2000; 10:141.
42. Despres N, Grant AM. Antibody interference in thyroid assays: a potential for clinical misinformation. Clin Chem 1998;44: 440.
43. Surks MI, Sievert R. Drugs and thyroid function. N Engl J Med 1995;333:1688.
44. Kailajarvi M, Takala T, Gronroos P, et al. Reminders of drug effects on laboratory test results. Clin Chem 2000;46:1395.
45. Brabant A, Brabant G, Schuermeyer T, et al. The role of glucocorticoids in the regulation of thyrotropin. Acta Endocrinol (Copenh) 1989;121:95.
46. Kaptein EM, Spencer CA, Kamiel MB, et al. Prolonged dopamine administration and thyroid hormone economy in normal and critically ill subjects. J Clin Endocrinol Metab 1980;51:387.
47. Geffner DL, Hershman JM. Beta-adrenergic blockade for the treatment of hyperthyroidism. Am J Med 1992;93:61.
48. Hawkins RC. Furosemide interference in newer free thyroxine assays. Clin Chem 1998;44:2550.
49. Lim CF, Munro SL, Wynne KN, et al. Influence of nonesterified fatty acids and lysolecithins on thyroxine binding to thyroxine-binding globulin and transthyretin. Thyroid 1995;5:319.
50. Munro SL, Lim CF, Hall JG, et al. Drug competition for thyroxine binding to transthyretin (prealbumin): comparison with effects on thyroxine-binding globulin. J Clin Endocrinol Metab 1989;68:1141.
51. Surks MI, DeFesi CR. Normal serum free thyroid hormone concentrations in patients treated with phenytoin or carbamazepine. A paradox resolved. JAMA 1996;275:1495.
52. Beckett GJ, Ratcliffe WA, Chapman B, et al. Non-isotopic, two-step free thyroxine immunoassay: a better measure of free thyroxine than analogue radioimmunoassay. Ann Clin Biochem 1990;27:581.
53. Beckett GJ, Wilkinson E, Rae PW, et al. The clinical utility of a non-isotopic two-step assay (DELFIA) and an analogue radioimmunoassay (SimulTRAC) for free thyroxine compared. Ann Clin Biochem 1991;28:335.
54. Mendel CM, Frost PH, Kunitake ST, et al. Mechanism of the heparin-induced increase in the concentration of free thyroxine in plasma. J Clin Endocrinol Metab 1987;65:1259.
55. Hollowell JG, Staehling NW, Flanders WD, et al. Serum TSH, T4, and thyroid antibodies in the United States population (1988 to 1994): National Health and Nutrition Examination Survey (NHANES III). J Clin Endocrinol Metab 2002;87: 489.
56. Brabant G, Prank K, Hoang-Vu C, et al. Hypothalamic regulation of pulsatile thyrotopin secretion. J Clin Endocrinol Metab 1991;72:145.
P.343

57. Elmlinger MW, Kuhnel W, Lambrecht HG, et al. Reference intervals from birth to adulthood for serum thyroxine (T4), triiodothyronine (T3), free T3, free T4, thyroxine binding globulin (TBG) and thyrotropin (TSH). Clin Chem Lab Med 2001; 39:973.
58. Gow SM, Kellett HA, Seth J, et al. Limitations of new thyroid function tests in pregnancy. Clin Chim Acta 1985;152:325.
59. Price A, Obel O, Cresswell J, et al. Comparison of thyroid function in pregnant and non-pregnant Asian and western Caucasian women. Clin Chim Acta 2001;308:91.
60. Andersen S, Pedersen KM, Bruun NH, et al. Narrow individual variations in serum T4 and T3 in normal subjects: a clue to the understanding of subclinical thyroid disease. J Clin Endocrinol Metab 2002;87:1068.
61. Browning MC, Ford RP, Callaghan SJ, et al. Intra- and interindividual biological variation of five analytes used in assessing thyroid function: implications for necessary standards of performance and the interpretation of results. Clin Chem 1986;32:962.
62. Boscato LM, Stuart MC. Heterophilic antibodies: a problem for all immunoassays. Clin Chem 1988;34:27.
63. John R, Henley R. Antibody interference in free thyroxine assays. Ann Clin Biochem 1992;29:472.
64. John R, Othman S, Parkes AB, et al. Interference in thyroid-function tests in postpartum thyroiditis. Clin Chem 1991;37: 1397.
65. Kricka LJ. Interferences in immunoassay—still a threat. Clin Chem 2000;46:1037.
66. Ismail AA, Walker PL, Cawood ML, et al. Interference in immunoassay is an underestimated problem. Ann Clin Biochem 2002;39:366.
67. Stockigt JR, DeGaris M, Barlow JW. “Unbound-analogue” methods for free T4: a note of caution. N Engl J Med 1982; 307:126.
68. Sunthornthepvarakul T, Likitmaskul S, Ngowngarmratana S, et al. Familial dysalbuminemic hypertriiodothyroninemia: a new, dominantly inherited albumin defect. J Clin Endocrinol Metab 1998;83:1448.
69. Curtis AJ, Scrimshaw BJ, Topliss DJ, et al. Thyroxine binding by human transthyretin variants: mutations at position 119, but not position 54, increase thyroxine binding affinity. J Clin Endocrinol Metab 1994;78:459.
70. Stockigt JR, Stevens V, White EL, et al. “Unbound analog” radioimmunoassays for free thyroxin measure the albumin-bound hormone fraction. Clin Chem 1983;29:1408.
71. Stockigt JR, Dyer SA, Mohr VS, et al. Specific methods to identify plasma binding abnormalities in euthyroid hyperthyroxinemia. J Clin Endocrinol Metab 1986;62:230.
72. Seth J, Beckett G. Diagnosis of hyperthyroidism: the newer biochemical tests. Clin Endocrinol Metab 1985;14:373.
73. Snyder PJ, Utiger RD. Inhibition of thyrotropin response to thyrotropin-releasing hormone by small quantities of thyroid hormones. J Clin Invest 1972;51:2077.
74. Forfar JC, Feek CM, Miller HC, et al. Atrial fibrillation and isolated suppression of the pituitary-thyroid axis: response to specific antithyroid therapy. Int J Cardiol 1981;1:43.
75. Sawin CT. Subclinical hyperthyroidism and atrial fibrillation. Thyroid 2002;12:501.
76. Faber J, Jensen IW, Petersen L, et al. Normalization of serum thyrotrophin by means of radioiodine treatment in subclinical hyperthyroidism: effect on bone loss in postmenopausal women. Clin Endocrinol (Oxf) 1998;48:285.
77. Toft AD. Subclinical hyperthyroidism. N Engl J Med 2001; 345:512.
78. Cohen JH, Ingbar SH, Braverman LE. Thyrotoxicosis due to ingestion of excess thyroid hormone. Endocr Rev 1989;10:113.
79. Page SR, Sheard CE, Herbert M, et al. A comparison of 20 or 40 mg per day of carbimazole in the initial treatment of hyperthyroidism. Clin Endocrinol (Oxf) 1996;45:511.
80. Toft AD, Irvine WJ, Hunter WM, et al. Anomalous plasma TSH levels in patients developing hypothyroidism in the early months after 131I therapy for thyrotoxicosis. J Clin Endocrinol Metab 1974;39:607.
81. Young ET, Steel NR, Taylor JJ, et al. Prediction of remission after antithyroid drug treatment in Graves’ disease. QJM 1988; 66:175.
82. Takamatsu J, Kuma K, Mozai T. Serum triiodothyronine to thyroxine ratio: a newly recognized predictor of the outcome of hyperthyroidism due to Graves’ disease. J Clin Endocrinol Metab 1986;62:980.
83. Panesar NS, Li CY, Rogers MS. Reference intervals for thyroid hormones in pregnant Chinese women. Ann Clin Biochem 2001;38:329.
84. Talbot JA, Lambert A, Anobile CJ, et al. The nature of human chorionic gonadotrophin glycoforms in gestational thyrotoxicosis. Clin Endocrinol (Oxf) 2001;55:33.
85. Hershman JM. Human chorionic gonadotropin and the thyroid: hyperemesis gravidarum and trophoblastic tumors. Thyroid 1999;9:653.
86. Pop VJ, Brouwers EP, Vader HL, et al. Maternal hypothyroxinaemia during early pregnancy and subsequent child development: a 3-year follow-up study. Clin Endocrinol (Oxf) 2003; 59:282.
87. Meier C, Trittibach P, Guglielmetti M, et al. Serum thyroid stimulating hormone in assessment of severity of tissue hypothyroidism in patients with overt primary thyroid failure: cross sectional survey. BMJ 2003;326:311.
88. Cooper DS. Subclinical hypothyroidism. N Engl J Med 2001; 345:260.
89. Cotton GE, Gorman CA, Mayberry WE. Suppression of thyrotropin (h-TSH) in serums of patients with myxedema of varying etiology treated with thyroid hormones. N Engl J Med 1971;285:529.
90. Sagud M, Pivac N, Muck-Seler D, et al. Effects of sertraline treatment on plasma cortisol, prolactin and thyroid hormones in female depressed patients. Neuropsychobiology 2002;45: 139.
91. Harel Z, Biro FM, Tedford WL. Effects of long term treatment with sertraline (Zoloft) simulating hypothyroidism in an adolescent. J Adolesc Health 1995;16:232.
92. Toft AD, Beckett GJ.Thyroid function tests and hypothyroidism. BMJ 2003;326:295.
93. Spencer C, Eigen A, Shen D, et al. Specificity of sensitive assays of thyrotropin (TSH) used to screen for thyroid disease in hospitalized patients. Clin Chem 1987;33:1391.
94. Hamblin PS, Dyer SA, Mohr VS, et al. Relationship between thyrotropin and thyroxine changes during recovery from severe hypothyroxinemia of critical illness. J Clin Endocrinol Metab 1986;62:717.
95. Wilcox RB, Nelson JC, Tomei RT. Heterogeneity in affinities of serum proteins for thyroxine among patients with non-thyroidal illness as indicated by the serum free thyroxine response to serum dilution. Eur J Endocrinol 1994;131:9.
96. Stockigt JR. Guidelines for diagnosis and monitoring of thyroid disease: nonthyroidal illness. Clin Chem 1996;42:188.
97. Ladenson PW, Singer PA, Ain KB, et al. American Thyroid Association guidelines for detection of thyroid dysfunction. Arch Intern Med 2000;160:1573.
98. Oliveira JH, Persani L, Beck-Peccoz P, et al. Investigating the paradox of hypothyroidism and increased serum thyrotropin (TSH) levels in Sheehan’s syndrome: characterization of TSH carbohydrate content and bioactivity. J Clin Endocrinol Metab 2001;86:1694.
99. Price A, Weetman AP. Screening for central hypothyroidism is unjustified. BMJ 2001;322:798.
P.344

100. Waise A, Belchetz PE. Lesson of the week: unsuspected central hypothyroidism. BMJ 2000;321:1275.
101. Wardle CA, Fraser WD, Squire CR. Pitfalls in the use of thyrotropin concentration as a first-line thyroid-function test. Lancet 2001;357:1013.
102. Lamberts SW, de Herder WW, van der Lely AJ. Pituitary insufficiency. Lancet 1998;352:127.
103. Mitchell DR, Parvin CA, Gronowski AM. Rules-based detection of discrepancies between TSH and free T4 results. Clin Chim Acta 2003;332:89.