3.1.4 Thyroid function tests and the effects of drugs

The assessment of thyroid function by laboratory testing began in about 1934 with the measurement of oxygen consumption or basal metabolic rate. Twenty years later measurement of protein-bound iodine became the standard technique and after a further 20 years this assay was superseded by radioimmunoassays of thyroxine (T₄) and triiodothyronine (T₃). Radioimmunoassays for thyroid-stimulating hormone (TSH) were reported from 1965, but early techniques could not distinguish normal values from the suppressed levels found in thyrotoxicosis. Until about 1990 this distinction was made by the administration of intravenous thyrotropin-releasing hormone (TRH), which fails to increase TSH to measurable levels in thyrotoxicosis, while producing a clear 5- to 15-fold increase in serum TSH in euthyroid subjects with normal pituitary function. Immunometric TSH assays now allow the suppressed serum TSH levels of thyrotoxicosis to be clearly distinguished from normal. This fundamental advance has coincided with the development of ingenious techniques to estimate the minute fraction of total serum T₄ that circulates in the unbound state, but even the best free T₄ methods offer only a marginal diagnostic advantage over the measurement of total T₄, e.g. when the concentration of thyroxine-binding globulin (TBG) is abnormal. Current enthusiasm for free T₄ and T₃ estimation needs to be tempered by an understanding of the method-dependent limitations of these techniques, particularly in situations where assessment of thyroid function is most difficult (see below).

All current methods of measuring TSH, T₄, and T₃ in serum, whether by radioimmunoassay or immunometric techniques, are comparative, i.e. they depend on the assumption that the unknown sample and the assay standards are identical in all measured characteristics other than the concentration of analyte. When this condition is not fulfilled, e.g. when the sample shows anomalous binding of tracer to serum proteins or antibodies, the assay result will be spurious and potentially misleading.

While there is little doubt that circulating TSH and T₄ should both be measured when an abnormality of thyroid function is suspected, recent recommendations suggest that it may be appropriate to apply testing more widely in a wide range of patient groups with an increased risk of thyroid dysfunction (Box 3.1.4.1). For example, neonatal screening for congenital hypothyroidism is firmly established. Routine testing of thyroid function with a single measurement of serum TSH in women over 50, the group most likely to have significant thyroid dysfunction (2), first advocated about 2000, (3) has become widely recommended (see also Chapter 3.1.7). Because current TSH assays are very sensitive in detecting either thyrotoxicosis or primary hypothyroidism, there is a trend for T₄ to be estimated in primary care only if TSH is abnormal (see below).

The recognition that an adequate level of maternal thyroxine in the first trimester of pregnancy is a crucial determinant of fetal brain development, has led to increased testing of thyroid function in preparation for pregnancy, especially in women who have impaired fertility or any risk factors for thyroid dysfunction (4) (see Chapter 3.4.5). The frequency of postpartum thyroid dysfunction places a high priority on the assessment of thyroid function for any suggestive clinical features in the first year after childbirth (5) (see Chapter 3.4.6).

The value of routine testing needs to be compared with the sensitivity and accuracy of clinical assessment. Studies of unselected patients assessed by primary care physicians show that clinical acumen alone lacks sensitivity and specificity in detecting previously undiagnosed thyroid dysfunction (6). In up to one-third of patients evaluated for suspected thyroid disease by specialists, laboratory results lead to revision of the clinical assessment (7).

Regardless of the strategy that is used for first-line testing, serum TSH and a valid serum T₄ estimate are both necessary for definitive assessment of thyroid status. As shown in Fig. 3.1.4.1, the common types of thyroid dysfunction can be identified by diagonal deviations from the normal T₄–TSH relationship, which depends on the negative feedback interaction between target gland secretion and trophic hormone. The figure shows primary hypothyroidism due to target gland failure (high serum TSH with low free T₄: A), failure of TSH secretion (both low: B), autonomous or abnormally stimulated target gland function (high serum free T₄ with low TSH: C), and primary excess of TSH or thyroid hormone resistance (both high: D). Abnormal results that fall outside these areas suggest that some other factor has disturbed this relationship, or that the sample has been collected under non-steady state conditions (see below). The figure shows serum free T₄ rather than T₃ because T₄ is the major circulating determinant of TSH secretion, although circulating T₃ also has an important direct inhibitory effect on TSH secretion.

The relationship shown in Fig. 3.1.4.1 allows precise diagnosis of thyroid dysfunction from a single serum sample, subject to the assumptions and limiting conditions summarized in Box 3.1.4.2. The first of these assumptions (steady-state conditions) should always be questioned when associated illness or medications perturb the pituitary–thyroid axis. The large difference between the half-lives of TSH (1 h) and T₄ (1 week) accounts for many transient nondiagnostic abnormalities in the T₄–TSH relationship. Of the six assumptions detailed in Box 3.1.4.2, only the last three can be validated in the laboratory; the first three must be verified clinically. It should be emphasized that optimal assessment of thyroid function depends on collaborative communication across the laboratory–clinical interface. Critical aspects of this approach have been summarized by Stockigt (see section 7 of Chapter 6b on this website) (1).

A general algorithm for the assessment of thyroid function based on initial measurement of TSH is shown in Fig. 3.1.4.2. The application of this strategy will vary depending on the circumstances in which testing is initiated. Several distinct clinical situations can be identified: (1) testing of untreated subjects in screening or case-finding studies with low prediagnostic probability, (2) when clinical features suggest thyroid dysfunction, (3) evaluation of the response to treatment, and (4) assessment when associated illness, drug therapy, or pregnancy are likely to complicate clinical and laboratory assessment.

In the absence of associated disease, where there are no clinical features to suggest thyroid dysfunction, a normal serum TSH concentration has over 99% negative predictive value in ruling out primary hypothyroidism or thyrotoxicosis (3). Assessment of untreated subjects who have no features of thyroid dysfunction now commonly begins with measurement of TSH alone, with T₄ and/or T₃ assays added only if TSH is abnormal, or if an abnormality of TSH secretion is suspected (Fig. 3.1.4.2). According to this algorithm, free T₄ is measured to distinguish between overt and subclinical hypothyroidism when serum TSH is elevated, while a suppressed or subnormal TSH level should be followed by assay of both free T₄ and free T₃ to distinguish subclinical from overt thyrotoxicosis and to identify T₃ toxicosis.

The use of serum TSH as the sole initial test of thyroid function may lead to incorrect or incomplete assessment of thyroid status in a number of situations, as summarized in Table 3.1.4.1. Initial measurement of both T₄ and TSH is appropriate whenever thyroid dysfunction is clinically suspected, because thyroid dysfunction due to pituitary disease, either hypopituitarism, or the less common situation of TSH-dependent hyperthyroidism, may be missed if TSH alone is used for initial assessment (9). The far-reaching consequences of missing these disorders are not reflected by a small percentage deficit in diagnostic sensitivity!

In patients with newly treated thyrotoxicosis, TSH may remain suppressed for months after normalization of serum T₄ and T₃; serious overtreatment may result if TSH alone is used for adjustment of antithyroid drug dosage. Further, during drug treatment, thyrotoxicosis may persist due solely to T₃ excess. A reassessment of serum free T₄ and free T₃ levels is recommended after about 3 weeks drug treatment of thyrotoxicosis to allow appropriate dose adjustment. During long-term drug treatment of thyrotoxicosis, serum TSH may give a reliable guide to optimal drug dosage.

Serum TSH is the best single index of appropriate replacement, or suppressive therapy, during long-term treatment with thyroxine, but during the early phase of treatment of hypothyroidism, free T₄ should also be measured, because TSH may remain inappropriately elevated for several months after normalization of T₄ (Fig. 3.1.4.3). In elderly patients, especially those with cardiac ischaemia, dose adjustment is a clinical decision that need not be determined by serum TSH. During long-term replacement therapy, the best indicator of optimal dosage is a low-normal value for serum TSH, often associated with a slightly increased level of serum free T₄ that may vary depending on the time interval between dose and sampling. During suppressive therapy with T₄, periodic assessment of free T₄ and free T₃, in addition to TSH, is appropriate to identify and avoid thyroid hormone excess that may have adverse effects on the cardiovascular system or bone density.

In the treatment of hypothyroidism due to pituitary or hypothalamic disease, serum TSH is of no value in assessing T₄ dosage, which should be judged on the basis of serum free T₄ and clinical response.

Interpretation of thyroid function tests may be compromised by intercurrent illness and medications. There is a high prevalence of abnormal serum free T₄ or TSH values in patients with acute medical illness (10) and in some studies of acute psychiatric illness (11). However, when TSH and free T₄ are considered together, as in Fig. 3.1.4.1, most of these abnormalities do not indicate true thyroid dysfunction. Because clinical assessment of thyroid status is difficult in the face of associated disease, some have advocated widespread testing (10), but the consensus has moved away from routine testing during critical illness without some clinical indication (12). If not due to medications (see below), the combination of low serum free T₄ and low TSH indicates a poor prognosis in critically ill patients (13), although there is no evidence that these findings can usefully influence management decisions for individuals.

During any severe illness, one or more of the assumptions outlined in Box 3.1.4.2 may not be valid, e.g. when there are wide deviations from the steady state due to acute inhibition of TSH secretion or abnormally rapid T₄ clearance. Serum TSH values are frequently subnormal in the absence of thyrotoxicosis, although highly sensitive TSH assays show higher levels than are typical of thyrotoxicosis (see below). Serum free T₄ estimates during critical illness are prone to multiple method-dependent interfering influences, e.g. due to heparin and other medications (see below). Serum total T₄ measurements are less prone to such artefacts (8).

In late pregnancy, there are clearly unresolved methodological problems in estimating serum free T₄, with strong negative bias in some methods (14, 15). A recent study has questioned the wisdom of continuing to rely on free thyroxine estimates during pregnancy (15). In contrast to various free T₄ methods (14, 15) total serum T₄ and its derivative, the free thyroxine index, showed a more robust inverse relationship with serum TSH, with consistent results in numerous reports (15). Thus, total T₄ measurement may be superior to free T₄ estimates as a guide to therapy during pregnancy, provided that the reference values are modified to take account of the normal oestrogen-induced increase in TBG. If free T₄ estimates continue to be used in pregnancy, clinicians should interpret results in relation to reference intervals that are both trimester specific and method specific. It remains to be established whether problems inherent in free thyroxine measurement during pregnancy can be resolved by using isotope dilution liquid chromatography tandem mass spectrometry after ultrafiltration (16).

The secretion of TSH, a 24–30 kDa glycoprotein composed of two subunits, from the thyrotropic cells of the anterior pituitary is regulated by negative feedback from the serum free T₄ and free T₃ concentrations. In normal subjects, the serum TSH concentration shows both pulsatile and diurnal variation, with mean maximum concentrations of approximately 3 mU/l at about 02.00 with nadir values of about 1 mU/l at about 16.00; there is no significant sex difference in reference values (17). Because serum TSH fluctuates with an amplitude of 20–50% around the mean (17), it can be difficult to establish whether serial changes are relevant in follow-up studies of patients with subclinical hypothyroidism, because a change of up to 40% could reflect pulsatile secretion rather than progression of disease (18).

Between 08.00 and 21.00, reference values for serum immunoreactive TSH are generally in the range 0.3–4 mU/l (Table 3.1.4.2), with higher values in the immediate postnatal period when there is a surge of TSH secretion. The reference range should be calculated after logarithmic transformation of control values to achieve a valid estimate of the lower normal limit. Although not perfect, logarithmic transformation brings TSH reference values closer to a normal distribution that can be statistically assessed. Median values are generally at about 1 mU/l with a long tail to the right, so that the upper limit of the reference range is contentious (see below).

The introduction of immunometric assays that use two antibodies against different epitopes on the α- and β-subunits of TSH has greatly improved assay sensitivity (19). With the best current techniques, serum TSH can be precisely measured at least to 0.03 mU/l, so that the lowest concentrations in normal subjects are clearly distinguishable from those found in thyrotoxicosis. Factors that become important when clinical decisions are based on values close to the limit of detection include between-assay reproducibility or precision profile, composition of the assay matrix, possible appearance of nonspecific interference during sample storage, as well as possible carryover from one sample to the next during automated sampling (20). Analytical sensitivity can be defined from the dose response characteristics of a single assay by expressing sensitivity as 2 or 3 SD above the zero point, but this estimate is often too optimistic (19). A definition of functional sensitivity as the 20% between-assay coefficient of variation has become accepted (19). Manufacturers’ estimates of functional sensitivity are often not confirmed on clinical testing, and assay performance may vary between laboratories despite apparently identical technique. Laboratories should establish their own detection limit from the between-assay precision profile in the subnormal range.

While immunometric TSH assays offer enhanced sensitivity, there can be important problems with nonspecific interference, e.g. in methods that use mouse monoclonal antibodies. An antimouse immunoglobulin in the test serum allows the formation of a false bridge between the solid phase and the signal antibody, thus generating a spuriously high assay value (21). Inclusion of nonspecific mouse immunoglobulin in the assay usually blocks this effect, although persistent false-positive detectable serum TSH values are still found in some samples (22).

Widespread application of thyroid function testing has identified large numbers of asymptomatic subjects with abnormal TSH, with normal serum T₄, who may merit the designation ‘subclinical thyroid dysfunction’ (23). A sustained abnormality should be demonstrated before definite categorization (24). The merits and limitations of initiating therapy for these individuals are discussed in Chapters 3.3.4 and 3.4.4.

These considerations have become complicated because of lack of consensus on the limits of the TSH range (25, 26, 27). There is ongoing debate (26, 27) on whether the upper limit of the TSH reference range should be reduced from about 4 mU/l to 3 mU/l or even lower, based on exclusion criteria for the reference population, statistical treatment of data, inference of adverse outcome, or prospect of benefit from intervention. Similarly, for subclinical hyperthyroidism there is lack of consensus as to how subnormal TSH values should be classified. The NHANES III study (28) reserves the designation ‘subclinical hyperthyroidism’ for serum TSH values below 0.1 mU/l. By contrast, other guidelines for the diagnosis and management of subclinical thyroid disease classify values below the lower normal limit of 0.45 mU/l as indicating subclinical hyperthyroidism (29). Such a difference in classification may affect the health classification of up to 1% of any population. Since the gradation from normality to severe thyroid dysfunction is a continuum, studies of adverse outcomes or benefits from intervention will be critically dependent on uniform cut-off points and terminology.

Until these uncertainties are resolved, it is likely that most clinicians will recommend a period of observation rather than immediate intervention. If a trend towards overt disease is to be the cue to intervention, it is critical to establish what constitutes a significant change in serum TSH value, a hormone that is pulse-secreted and shows diurnal variation. From an analysis of serial individual variation over 1 year, the difference required for two test results to be convincingly different was 40% for TSH and 15% for free T₄ and free T₃ (18).

During standard T₄ replacement therapy a TSH value in the lower normal range usually coincides with an optimal symptomatic response. When the aim of T₄ suppressive therapy is regression of benign thyroid tissue, it may be appropriate to give sufficient T₄ to reduce serum TSH to 0.1–0.3 mU/l. In the follow-up treatment of high-risk patients with thyroid cancer, further TSH suppression is generally advocated, although the benefit of sufficient T₄ to suppress TSH to less than 0.1 mU/l remains unproven.

Critically ill patients frequently have subnormal levels of serum TSH, but with a sufficiently sensitive assay these values can be distinguished from the typical values found in thyrotoxicosis. The large majority of thyrotoxic subjects have values below 0.01 mU/l, whereas hospitalized patients with nonthyroidal illness do not show this degree of TSH suppression (30).

The need for TRH testing in clinical practice has almost been eliminated by the development of highly sensitive TSH assays. However, measurement of serum TSH 20–30 min after intravenous injection of 200–500 μg TRH is still useful for some purposes: (1) to assess patients whose basal serum TSH values are out of context (TSH assay artefacts, e.g. those due to heterophilic antibodies, generally fail to show a physiological response), (2) to investigate apparent thyroid hormone resistance or pituitary-dependent thyrotoxicosis (most patients with thyrotoxicosis due to TSH-secreting pituitary tumours show no increase in serum TSH after TRH (31), while those with thyroid hormone resistance usually show an increase), and (3) to identify central hypothyroidism in which a low serum free T₄ value may be associated with a normal amount of serum immunoreactive TSH that has impaired biological activity (32).

Most patients with TSH-secreting pituitary tumours have increased serum α-subunit concentrations (31), but values can also be elevated in postmenopausal women and in hypogonadal men.

Concentrations of total serum T₄ and T₃ reflect not only hormone production, but also the number and affinity of plasma protein binding sites. Total concentrations vary in direct relationship to protein binding, while serum free T₄ and free T₃ concentrations should not, if measured by valid methods. Serum total and free T₃ concentrations are somewhat higher in children (33). Typical reference ranges for serum total and free T₄ and T₃ are shown in Table 3.1.4.2. In late pregnancy, reference ranges for free T₄ show marked method-dependent variation; quoted ranges should be both trimester specific and method specific.

There have been many approaches to the assay of serum free T₄ and free T₃ concentrations, with detailed analysis of the validity of various methods (34). Two-step methods that separate a fraction of the free T₄ pool from the binding proteins as a preliminary before assay are generally least prone to analytical artefacts. Figure 3.1.4.4 outlines a two-step free T₄ method based on incubation of serum with a solid-phase T₄ antibody, followed by back titration of unoccupied antibody with labelled T₄.

No current method conveniently measures the free T₄ concentration in undisturbed, undiluted serum in a way that reflects in vivo conditions. Although equilibrium dialysis is widely considered the reference method for free T₄ measurement, it is also subject to error, especially as a result of generation of fatty acids during sample storage or incubation, and the inability of diluted samples to reflect the effect of binding competitors (see below). Evaluation of novel serum free T₄ methods should include testing with various protein binding abnormalities, as well as sera that contain substances that compete for serum protein binding sites. Unexpected interference may only be noted after methods have been used for some time, as in the effect of rheumatoid factor (35), heparin (36), or drug competitors for protein binding (8).

Recent reports suggest that methods based on liquid chromatography/tandem mass spectrometry after ultrafiltration (16), or equilibrium dialysis (37) may improve the measurement of free T₄. Further evaluations, in particular details of long-term reproducibility (i.e. interassay variation) of these techniques, as well as serial dilution studies to evaluate the effect of circulating inhibitors of T₄ binding (see below) are awaited.

Assays for serum total or free T₃ have no place in the diagnosis of hypothyroidism, but should be included in the diagnostic protocol in the following situations:

The serum T₃ concentration is not useful in assessing the effectiveness of T₃ replacement. Because of its short plasma half-life, the T₃ concentration is highly dependent on the interval between dosage and sampling.

Molecular changes in TBG, transthyretin (TTR, previously known as thyroxine-binding prealbumin), or albumin may result in altered serum concentrations of these binding proteins, or may alter their binding affinity for T₄ and/or T₃ (38). The X-linked structural TBG variants, some of which show abnormal heat lability, have either normal or reduced affinity for T₄; T₃ is usually similarly affected. Fifteen of at least 24 known X-linked variants of TBG cause complete TBG deficiency, while eight variants are associated with subnormal concentrations of immunoreactive serum TBG, often with reduced affinity for T₄ (38). In the total absence of TBG, total serum T₄ is reduced to 20–40 nmol/l (normal 50–140 nmol/l), whereas in hereditary TBG excess the concentration may increase up to 250 nmol/l (38); free T₄ remains normal. In general, the various methods of estimating serum free T₄ give a valid correction for TBG abnormalities, whether hereditary or acquired.

The albumin variant responsible for familial dysalbuminaemic hyperthyroxinaemia (FDH) (38), due to an Arg-His substitution at position 218, shows a selective increase in binding affinity for T₄ resulting in total serum T₄ in the range 180–240 nmol/l. The variant protein has increased affinity for some T₄-analogue tracers, resulting in spuriously high serum free T₄ estimates (38); equilibrium dialysis and various two-step free T₄ methods and serum TSH confirm that people with the FDH variant are euthyroid. TTR variants can increase total serum T₄ into the range 150–200 nmol/l, but are not reported to cause spurious free T₄ results.

Circulating T₃- or T₄-binding autoantibodies can cause methodological artefacts in both total and free measurements of T₄ and T₃ (8). Depending on the separation method that is used, tracer bound to the endogenous human antibody will be classified as ‘bound’ in absorption methods of assay separation, but falsely classified as ‘free’ in double antibody methods, leading, respectively, to spuriously low or high serum values (8). Assay after ethanol extraction of serum establishes the true total hormone concentration.

These terms are used when the total or free T₄ concentrations are increased or decreased without evidence of thyroid dysfunction. The effects of medications and alterations in the T₄ binding proteins are the commonest causes (Box 3.1.4.3, Table 3.1.4.3). Hypothyroxinaemia is a normal response when TSH secretion is inhibited by another thyromimetic such as T₃ or triiodothyroacetic acid. During critical illness serum T₄ may be subnormal due to inhibition of TSH secretion (39), decreased production of binding proteins, or accelerated T₄ clearance. Hypothyroxinaemia without the anticipated increase in TSH also is seen in very low birthweight premature infants, in whom the lack of TSH response appears to reflect hypothalamic–pituitary immaturity (40).

The multiple effects of medications on the pituitary–thyroid axis (Table 3.1.4.3) have been reviewed elsewhere (1, 38, 41) Medications that present special problems include amiodarone, heparin, lithium, phenytoin, highly active antiretroviral therapy, and drugs that displace T₄ from TBG. Oestrogen, endogenous or exogenous, is the substance that most commonly affects tests of thyroid function by increasing total T₄ due to an increase in the concentration of TBG. Free T₄ remains normal. Oestrogens, including a minor effect of selective oestrogen agonists such as raloxifene (42), act to increase the glycosylation of TBG, which slows its clearance (38). Transdermal oestrogens do not show this effect (38).

Amiodarone is the most complex and difficult of the drugs that can affect thyroid status (43). The clinical entities that may result from amiodarone therapy include two forms of thyrotoxicosis, one due to iodine excess and one attributed to thyroiditis (see Chapter 3.3.10). In iodine-replete regions the predominant amiodarone-induced thyroid abnormality is hypothyroidism, which is especially prevalent in those with associated autoimmune thyroiditis (see Chapter 3.2.6). The drug also causes benign euthyroid hyperthyroxinaemia in up to 25% of treated patients. There is often poor correlation between circulating thyroid hormone levels and the clinical manifestations of amiodarone-induced thyroid dysfunction, perhaps because of interaction of this drug or its metabolites with thyroid hormone receptors. In assessing the severity of amiodarone-induced thyrotoxicosis, the extent of measured thyroid hormone excess is less relevant than criteria such as muscle weakness and weight loss.

In serum obtained from heparin-treated patients, the measured concentration of serum free T₄ may be higher than the true in vivo concentration, due to in vitro generation of nonesterified fatty acids as a result of heparin-induced lipase activity during sample storage or incubation (36) (Fig. 3.1.4.5). High serum triglyceride concentrations and sample incubation at 37°C accentuate this artefact. Low-molecular-weight heparin preparations have a similar effect (44).

Lithium, a medication used in the management of manic-depressive illness, has multiple effects on the pituitary–thyroid axis, the most important being an effect to inhibit thyroglobulin hydrolysis and hormone release (45). It can exacerbate or may initiate autoimmune thyroid disease with development of goitre and hypothyroidism; there are also rare reports of lithium-induced thyrotoxicosis (45). Serum TSH, T₄, and T₃ assays give a true index of thyroid status during lithium treatment.

The antiepileptic phenytoin and carbamazepine both commonly result in subnormal serum total T₄, with an apparent lowering of free T₄, not accompanied by the anticipated increase in TSH (46). This discrepancy, which is not easily distinguishable from central hypothyroidism due to pituitary deficiency, is a methodological artefact related to underestimation of true free T₄ in diluted serum samples that contain inhibitors of T₄ protein binding (8, 46) (see below).

Infection with HIV may influence tests of thyroid function by various mechanisms, occasionally as result of direct infection of the thyroid gland or alteration of immunological function, but more frequently from the effect of medications that alter metabolism of thyroxine or as a nonspecific effect of debilitating illness. Some studies (47) show a higher than expected prevalence of hypothyroidism, predominantly subclinical, during treatment with highly active antiretroviral therapy (HAART). There are reports of reduced effectiveness of thyroxine replacement during treatment for HIV infection as a result of accelerated thyroxine metabolism, as with lopinavir/ritonavir (48); there is one paradoxical report of transient over-replacement during treatment with indinavir (49). There is no consensus as to whether thyroid function should be routinely monitored in HIV-infected patients, but testing will frequently be required to assess features that could be due to thyroid dysfunction. During HAART, thyroxine replacement needs to be monitored and adjusted (47, 48).

In contrast to the steroid and vitamin D binding plasma proteins, both TBG and TTR show extensive cross-reactions with a wide range of drugs (1, 38, 41). As reviewed elsewhere (8), the failure of current free T₄ and T₃ methodology to reliably reflect the effect of drug competitors that increase free T₄ and T₃in vivo by displacement, remains a major limitation in the general applicability of free hormone assays. These effects are poorly reflected by standard free hormone assays because samples are generally assayed after dilution, resulting in underestimation of the free hormone concentration in the presence of competitors (8, 46). When measured by ultrafiltration of undiluted serum, therapeutic concentrations of phenytoin and carbamazepine showed an increase in free T₄ fraction by 45–65%, but these effects were obscured by 1:5 assay dilution of serum (46). This discrepancy occurs because of dissociation of bound ligand with progressive sample dilution, so that the free concentration, at first well maintained, declines steeply as the ‘reservoir’ of bound ligand becomes depleted (8, 46). Important drug competitors have a much smaller proportional reservoir of bound ligand than does T₄, so that their free concentration becomes negligible with progressive dilution while the free T₄ concentration remains unaltered (8). Since competition is a function of relative free ligand concentrations, the effect of a competitor to increase free T₄ is underestimated, the error being greatest in assays with the highest sample dilution (1, 8, 46).

Drug effects on thyroid function may be especially potent when several agents are given together. For example, infusion of furosemide in high dosage lowers serum T₄, while concurrent dopamine infusion inhibits TSH secretion; together they can result in profound hypothyroxinaemia. Combinations of rifampicin or ritonavir or other medications that accelerate T₄ clearance, with glucocorticoid-induced inhibition of TSH secretion can have a similar effect.

When thyroid function is abnormal, additional diagnostic information can be gained from antibody studies, imaging techniques, and measurement of thyroglobulin. The investigation of thyroid masses per se is not considered here.

In subclinical hypothyroidism, the presence of thyroid peroxidase (TPO) antibodies indicates a four- to fivefold increase in the chance of developing overt hypothyroidism (2). The presence of this antibody also indicates an increased likelihood of postpartum thyroiditis or amiodarone-induced hypothyroidism. The finding of persistently positive thyrotropin receptor antibody (TRAb) is useful in indicating that apparent remission of Graves’ disease is unlikely to be sustained. TRAb measurement can indicate the possibility of neonatal thyrotoxicosis in the infant of a mother with autoimmune thyroid disease and may also define the aetiology of atypical eye disease.

The use of isotope imaging techniques in thyrotoxicosis due to Graves’ disease varies widely between different centres. While some now regard routine radioisotope imaging as redundant in typical Graves’ disease, negligible uptake can be a key feature in confirming thyrotoxicosis due to thyroiditis, iodine contamination, and factitious ingestion of thyroid hormone. Imaging also can confirm a ‘hot’ nodule as the predominant source of thyroid hormone excess. CT is valuable in identifying the extent of retrosternal extension, but contrast agents should be avoided. Colour flow Doppler has been reported to differentiate between type 1 and type 2 amiodarone-induced thyrotoxicosis (43) (see Chapter 3.3.10).

In the follow-up of differentiated thyroid cancer, an undetectable serum thyroglobulin concentration in the presence of high serum TSH indicates effective ablation and may justify less rigorous T₄-induced suppression of TSH. Thyroglobulin is undetectable in thyrotoxicosis factitia, and generally extremely high in subacute thyroiditis and in amiodarone-induced thyrotoxicosis of the thyroiditis type.

Assay of thyroglobulin in the needle wash from suspect neck lymph nodes appears to have a higher sensitivity and specificity than cytology in establishing whether they contain metastatic thyroid tissue (50).

While there is currently no diagnostically reliable laboratory index of peripheral thyroid hormone action, some tests (51), including sex steroid binding globulin, serum ferritin, serum angiotensin-converting enzyme, as well as measurement of oxygen consumption, systolic time interval, and ultrasonographic parameters of cardiac contractility (52), may be useful in following individual response in situations of suspected thyroid hormone resistance or during long-term suppressive therapy with T₄.

When there is discordance between laboratory results and clinical findings, a distinction needs to be made between anomalous assay results due to specific or nonspecific assay interference and those that indicate previously unsuspected or subclinical disease. Consideration of the fundamental assumptions that underlie the diagnostic use of the trophic-target hormone relationship (Box 3.1.4.2) may give a clue to the discrepancy. Anomalous or unexpected assay results can be approached in the following sequence: