Thyroid Function Tests

Over the past forty years, improvements in the sensitivity and specificity of thyroid test methodologies have dramatically impacted the clinical strategies for detecting and treating thyroid disorders. In the 1950s, only one thyroid test was available - an indirect estimate of the serum total (free + protein-bound) thyroxine (TT4) concentration, using the protein bound iodine (PBI) technique ^1,2. Since 1970, technological advances in radioimmunoassay (RIA) and immunometric assay (IMA) methodologies have progressively improved the specificity and sensitivity of the methods ^3,4. Currently, serum-based methods are available for measuring both total (TT4 and TT3) and free (FT4 and FT3) thyroid hormone concentrations ⁵. In addition, measurements can be made of the thyroid hormone binding proteins, Thyroxine Binding globulin (TBG), Transthyretin (TTR)/Prealbumin (TBPA) and Albumin, as well as for the pituitary thyroid stimulator, thyrotropin (thyroid stimulating hormone, TSH) and the thyroid hormone precursor protein, Thyroglobulin (Tg) ^4,6,7. The recognition that autoimmunity is a major cause of thyroid dysfunction has led to the development of tests for thyroid autoautoantibodies - thyroid peroxidase antibodies (TPOAb), thyroglobulin antibodies (TgAb) and TSH receptor antibodies (TRAb) ⁸. Currently, thyroid testing is performed on serum specimens using either manual or automated methods employing specific antibodies ^3,4. Methodology is still evolving as performance standards are established by the professional organizations and new technology and instruments are developed by manufacturers. This chapter is designed to give an overview of the current status and limitations of the thyroid testing methods most commonly used in clinical practice and as recommended by the new NACB guidelines ³⁵³.

2. Total Hormone Measurements (TT4 and TT3)

Thyroxine (T4) circulates ~ 99.97% bound to the plasma proteins: TBG (60-75%); TTR/TBPA (15 -30%) and Albumin (10%); whereas triiodothyronine (T3) is ~ 99.7% bound, primarily to TBG ⁶. Technically, it has been easier to develop methods to measure the concentrations of total (free + protein-bound) thyroid hormones (TT4 and TT3) that circulate at nanomolar concentrations, in contrast to the free hormones (FT4 and FT3) that are measured in the picomolar range. Serum TT4 measurement has evolved through a variety of technologies over four decades. The PBI tests of the 1950s were replaced first by competitive protein binding methods in the 1960s, and later superseded by RIA methods in the 1970s ^9-12. Currently TT4 and TT3 concentrations are primarily measured by competitive immunoassay techniques that use radioactivity, enzymes, fluorescence or chemiluminescent molecules as signals ^11,13-16. Total hormone methods require the inclusion of inhibitors, such as 8-anilino-1-napthalene-sulphonic acid, to block hormone binding to serum proteins in order to facilitate hormone binding of to the antibody reagent ^17-19. Serum TT3 method development has paralleled that of TT4, the ten-fold lower TT3 concentration presents a sensitivity and precision challenge despite the use of higher specimen volumes ^20-27. The inter-method variability for total hormone measurements (TT4 and TT3) is surprisingly large (10-17% and 20-30%, TT4 and TT3, respectively) considering that highly purified preparations of crystalline L-thyroxine and L-triiodothyronine are readily available and a reference technique has been established ²⁸. This variability most likely relates to matrix differences between calibrators and the efficiency of the blocking agent employed by different manufacturers ^25,29.

The diagnostic accuracy of total hormone measurements would equal that of free hormone if all patients had similar binding protein concentrations ⁵. Unfortunately, serum TBG abnormalities that distort the total:free hormone relationship, are commonly encountered in clinical practice (Table 2). Additionally, some patients have abnormal thyroid hormone binding proteins or autoantibodies that render total hormone measurements diagnostically unreliable [Table 1 & section 5(b)] ^30,31. Consequently, TT4 and TT3 measurements are rarely used as stand-alone tests, but are employed in conjunction with a binding protein estimate test [i.e. thyroid hormone binding ratio, THBR, see section 3A(b)] to form a free hormone index (FT4I or FT3I). The index approach corrects for common abnormalities in the proteins that bind thyroid hormones in the circulation (see section 3A) ^32,33. Total T4 reference ranges vary to some extent between methods but in general have approximated 58 to 160 nmol/L (4.5-12.6 µg/dL) for more than two decades. There is a new trend to use TT4 measurements in preference to FT4 estimate tests to monitor pregnancy, provided that the TT4 reference range is adjusted by a factor of 1.5 to accommodate the effects of the predictable TBG elevation ^261,354,³⁵⁵. Likewise, serum TT3 reference values are also method dependent, with ranges approximating to 1.2 - 2.7 nmol/L (80 –180 ng/dL) ²⁶.

Table 2.
Method Standardization	Methodologic Principle	Nomenclature	Limitations	References
A. Index Methods • Empiric standardization	Mathematical "correction of total hormone concentration for TBG by: (a) TBG measurement (b) Thyroid Hormone Binding Ratio (THBR) or "Uptake" test (c) Isotopic free hormone fraction	• Free T4 Index (FT4I) • Free T3 Index (FT3I) • T7 • Effective Thyroxine Ratio (ETR) • Tracer ultrafiltration • Tracer equilibrium dialysis • Sephadex column	•Two tests are not cost effective •THBR/Uptake tests do not fully correct for congenital abnormalities in TBG •High values in FDH •TBG Dependent •Tracer impurities	53 60,61,62,63 51,63,68 61,70 66,67
B. Absolute Methods • use gravimetric standards	Physical Separation of free from bound hormone with subsequent RIA of FT4 in the isolated free fraction.	(a) Equilibrium Dialysis (b) Ultrafiltration (c) Gel Filtration/Adsorption	• Limited availability • Technically demanding • Expensive • Limited availability • Technically demanding • Expensive • Limited availability • Expensive	73-75 37,38,77 76
C. Comparative Immunoassays • Use calibrants with values assigned by a direct absolute method (a)	(a) Two-Step labeled Hormone Serum is incubated with solid-phase anti-hormone antibody. After washing, unoccupied antibody binding sites are quantified with labeled hormone. (b) One-step Labeled Hormone Analog A labeled hormone analog, designed to to be nonreactive with plasma proteins, competes with free hormone in serum for a limited amount of anti-hormone antibody (c) Labeled Antibody	• "two step" • "back titration" • "sequential" • Hormone Analog • One Step • Analog • Labelled Antibody • Antibody Analog • Analog	• Manual tests • Two wash steps • Poor precision (some methods) • High values in FDH • Prone to interference with albumin and NEFA • Low values with NTI • High values with FDH • high values with T4 and T3 autoantibodies • Limited testing with NTI and FDH	84-87 299 299 92,94,100-104 196, 34,104

3. Free Hormone Tests (FT4 and FT3)

In accord with the free hormone hypothesis, it is believed that the minute free fraction of hormone (0.02% versus 0.2%, FT4 versus FT3, respectively) is responsible for the biologic activity of thyroid hormones at the cellular level ⁶. It follows that free hormone measurement will better reflect the physiological effects of the hormone than total hormone measurements, especially when binding proteins are abnormal ^6,34-36. Unfortunately, techniques for physically separating the exceeding small amount of free hormone from the dominant protein-bound moiety are too technically demanding, inconvenient and expensive for routine clinical laboratory use. Such physical separation methods, which employ equilibrium dialysis, ultrafiltration and gel filtration, are typically only available in reference laboratories. Clinical laboratories prefer to use one of a number of single-test immunosequestration techniques that are designed to estimate free hormone concentration in the presence of protein-bound hormone ^3,37. In reality, despite manufacturers claims, all of the FT4 and FT3 estimate tests are binding-protein dependent to some extent ³⁸. The current trend is to use one of the newer single-test free hormone immunoassay methods [Section 3C] in preference to the older two-test (total hormone + THBR) index approach [Section 3A].

Considerable confusion still surrounds the nomenclature of the free hormone tests and controversy continues regarding the technical validity of the measurements themselves, and their clinical utility in conditions associated with binding protein abnormalities of pregnancy and non-thyroidal illness (NTI) ^37,39-46, ^356, ³⁵⁷. Methods that are sensitive to albumin concentrations, the presence of certain drugs, high free fatty acid (FFA) levels or hormone binding inhibitors are reported to give inappropriately abnormal results for such patients ^41,47-51.

As shown in Table 2, free hormone measurements (FT4 and FT3) are made either by two-test index methods or single-test methods that include reference techniques that employ physical separation of free from bound hormone, and immunoassay “sequestration” methods that are usually automated. The single-test methods are either standardized with gravimetric preparations or use calibrators with values assigned by a reference method ³⁷. Since the reference methods are manual, technically demanding techniques they are too inconvenient and expensive for routine clinical use. Clinical laboratories typically use either two-test index or single-test immunoassays ¹⁶. Unfortunately, a confusing plethora of terms have been used to distinguish the different free hormone methods, and the literature abounds with inconsistencies in nomenclature of these tests (Table 2). Currently, the methodologic distinction between terms such as "T7", "effective thyroxine ratio", "one-step", "analog", "two step", "backtitration", "sequential" and "immunoextraction" has become blurred, as manufacturers have modified the original techniques or adapted them for automation ^3,37. Following the launch of the original one-step "analog" tests in the 1970s, the term "analog" became mired in confusion ³⁷. This first generation of hormone-analog tests were shown to be binding-protein dependent and have since been replaced by a new generation of labeled-antibody "analog" tests which are more resistant to the presence of abnormal binding proteins ^37,49,52. Unfortunately, manufacturers rarely disclose all assay constituents or the number of steps involved in an automated procedure. Thus, it is not possible to use the method’s nomenclature to predict its diagnostic accuracy for assessing patients with the binding protein abnormalities shown in Table 1 ³⁵³.

Free Hormone Indexes (FT4I and FT3I) are based on simple calculations that modify the total hormone value to provide an approximation of the free hormone concentration in the presence of abnormal binding proteins. These indexes have been used to estimate free hormone concentrations for 40 years and require two separate measurements. One test is a measurement of total hormone concentration (TT4 or TT3) [section 2], the other, an assessment of binding protein concentration using either (a) a TBG immunoassay, (b) a Thyroid Hormone Binding Ratio (THBR) or "Uptake’ test, or (c) an estimate of the free hormone fraction determined by isotopic dialysis or ultrafiltration. The purity of the hormone tracer critically impacts the diagnostic accuracy of indexes calculated using isotopic THBR tests or free fraction assessments ^53-55. Current THBR tests are usually able to produce normal FT4I and FT3I values when TBG abnormalities are mild (i.e. pregnancy and estrogen therapy). However, these tests often fail to normalize FT4I and FT3I values in euthyroid patients with grossly abnormal binding proteins (congenital TBG extremes, familial dysalbuminemic hyperthyroxinemia (FDH), thyroid hormone autoantibodies and NTI) ^56-58. These tests also have impaired diagnostic accuracy in the presence of certain drugs that influence thyroid hormone protein binding ⁵⁹.

There is no improvement in diagnostic accuracy of free hormone indexes (TT4/TBG) calculated using direct TBG measurement in preference to a THBR (“uptake”) test (Table 2). Further, the TT4/TBG index approach is not independent of the TBG concentration, nor does it correct for non TBG-related binding protein abnormalities (Table 1) ^32,33,60-62. Thus, despite the theoretical advantages of using direct TBG measurements, indexes employing THBR ("uptake tests") appear to be diagnostically superior to the TT4/TBG index approach that is now rarely used ⁶³.

The first "T3 uptake" tests developed in the 1950s employed the partitioning of T3-I¹³¹ tracer between the plasma proteins in the specimen and an inert scavenger (red cell membranes, talc, charcoal, ion-exchange resin or antibody) ^64-66. The scavenger "uptake" of T3 tracer was an indirect, reciprocal estimate of the TBG concentration of the specimen. Initially, T3 uptake tests were reported as percent uptakes (free/total tracer). Sera with normal TBG concentrations typically exhibited uptakes that were 30 ± 5 percent of the T3 tracer. Subsequently, T3-I¹²⁵ replaced T3-I¹³¹ tracers and the American Thyroid Association recommended that the calculation should be based on the ratio between absorbant counts divided by the total, minus absorbant counts, rather than the ratio between absorbant counts and total counts ^54,67. Further, it was recommended that uptake methods be renamed Thyroid Hormone Binding Ratio (THBR) tests and expressed as a ratio with normal sera, having an assigned value of 1.00 ^54,67. Historically, the use of T3 tracer, as opposed to T4 tracer, was made for practical reasons. Specifically, T3 binding to TBG is ten-fold lower than T4-TBG binding, resulting in a greater percentage of unbound T3 tracer being available for scavenger pick up, and consequently required shorter gamma counting times. As non-isotopic technology became more widely used, labeled T4 became the preferred hormone for the THBR testing because a T4 “uptake” more appropriately correct for T4-binding protein effects ⁶⁸. Current THBR tests usually produce normal FT4I and FT3I values when TBG abnormalities are mild (i.e. pregnancy). However, these tests may not produce normal index values when patients have congenital TBG extremes, familial dysalbuminemic hyperthyroxinemia (FDH), thyroid hormone autoantibodies, nonthyroidal illness or medications that directly or indirectly influence thyroid hormone binding to plasma proteins ^69-72.

The first free hormone tests developed in the 1960s were indexes, calculated from the product of the free hormone fraction, measured isotopically by dialysis, and TT4 measured by PBI and later, by RIA ^73-75. The early isotopic detection systems were technically demanding and included paper chromatography, electrophoresis, magnesium chloride precipitation and column chromatography ^73,76-79. The free fraction index approach was later extended to ultrafiltration and symmetric dialysis, the latter measuring the rate of transfer of isotopically-labeled hormone across a membrane separating two chambers containing the same undiluted specimen ^43,80,81. These techniques eliminated the dilution effects that had been shown to influence tracer dialysis values ^82,83. The free hormone indexes calculated with isotopic free fractions are not completely independent of the TBG concentration and furthermore, are influenced by tracer purity and the buffer matrix employed by the method ^{50,55,76,77,84,85}.

These methods fall into two categories –methods employing physical separation of free from protein-bound hormone, and immunoassays that employ sequestration.

These methods are generally considered reference methods. A variety of techniques are employed to physically isolate free from protein-bound hormone before a sensitive immunoassay is used to measure the free hormone concentration in the free hormone fraction using gravimetrically prepared standard solutions ^85-88. The physical isolation of the free from protein-bound hormone is accomplished with either (a) dialysis through a semi-permeable membrane, (b) ultrafiltration, or (c) Sephadex LH-20 resin adsorption chromatography ^{42,43,80,85,87,89}. Although these methods are generally considered “gold standard” methods they are prone to dilutional errors that may cause underestimation of FT4 when binding inhibitors are present in the specimen. Further, such methods are too labor intensive and expensive for use in the routine clinical laboratory setting.

Equilibrium Dialysis and Ultrafiltration Methods. Ellis and Ekins first reported the direct measurement of FT4 in dialysates in 1973 ⁸⁷. Subsequently, the technique has been applied to serum ultrafiltrates ^43,88. In the last decade, such physical separation methods have been refined by the adoption of more physiologic buffer diluents and improvements to the design of dialysis cells and ultrafiltration devices ^83,85,90.

Gel Absorption Methods. These methods are based on the ability of free hormone to penetrate the internal compartments of Sephadex LH-20 columns. Protein-bound hormone is excluded and eluted before the free fraction, which is assayed for hormone by sensitive RIA. Commercial kits have been based on this methodologic approach but currently are not widely used ⁸⁹.

An exquisitely sensitive T4 RIA method is needed to measure the picomolar concentrations of FT4 in free fraction isolates as compared with total hormone that circulates at nanomolar concentrations. The requisite sensitivity has been achieved by using higher affinity T4 antibodies (>1x10¹¹ L/mol) and higher specific activity T4-I¹²⁵ tracers ³⁷. It has been more difficult to apply this approach to FT3 measurement since T3 RIA sensitivity has been limited by the requirement for T3 antibodies with affinities >3x10¹¹ L/mol ^37,86,88. A number of direct absolute FT4 methods employing equilibrium dialysis and ultrafiltration have been developed for both research and commercial use ^78-80,85. It is generally considered that the direct absolute methods are least influenced by binding proteins and by inference, provide free hormone values that best reflect circulating free hormone concentrations ^5,37. When available, they may be useful for evaluating the thyroid status of the unusual patient in whom routine FT4 test results and serum TSH values are diagnostically discordant. These methods are also important for assigning values to the calibrators used for some of the immunoassay methods [section 3C]. There is closer agreement between the reference ranges of methods employing physical separation than between the FT4 immunoassays and tracer dialysis techniques ^37,85. The reference ranges for FT4 immunoassay methods approximate to 9-23 pmol/L (0.7 –1.8 ng/dL) ³⁵⁶. In contrast, the upper FT4 limit for methods such as equilibrium dialysis that employ physical separation extends above 30 pmol/L (2.5 ng/dL) ⁸⁵. Reference ranges for FT3 immunoassay methods approximates to 3.5-7.7 pmol/L (23 –50 ng/L) ²⁶. FT3 methods that employ physical separation are only available as research assays ⁸⁶.

Free hormone immunoassays use a specific hormone antibody to sequester a small amount of the total hormone. The antibody occupancy, which is proportional to the free hormone concentration, is quantified using gravimetric standards or calibrators with free hormone values assigned by a reference equilibrium dialysis method ³⁷. The actual proportion of total hormone sequestered (up to 5%) varies with the methodologic design, but greatly exceeds the actual free hormone concentration ³⁷. The key to the validity of these methods is the use of conditions that maintain the free to protein-bound hormone equilibrium and minimize dilution effects that weaken the influence of any endogenous inhibitors present in the specimen ^35,41,81-83. Three general approaches have been used to develop comparative FT4 and FT3 immunoassay methods: (a) Two-step Labeled-Hormone; (b) One-step Labeled-Analog; and (c) Labeled Antibody.

This approach was first developed by Ekins and colleagues in the late 1970s and subsequently adapted to form the basis of many current FT4 and FT3 methods ^91-94. Two-step methods typically employ anti-hormone antibody bound to a solid support (ultrafine-sephadex, tube or particles). The solid-phase, high affinity (>1x10¹¹ L/mol) antibody sequesters a small proportion of total hormone from a diluted serum specimen during the first incubation step ^5,37,95. Unbound serum constituents are washed away before the addition of labeled hormone that is taken up by the unoccupied antibody-binding sites during a second incubation step. After washing, the amount of labeled hormone bound to the solid-phase antibody is quantified relative gravimetric standards or calibrators that have free hormone values assigned by a method employing physical separation. Following the introduction of the less labor-intensive one-step labeled hormone-analog methods [section 3C(b)] two-step techniques lost popularity, despite comparative studies showing that they were less affected by the binding protein abnormalities that were problematic for most one-step methods ^40,70,96,97. Currently, some manufacturers still base their FT4 and FT3 methods on the two-step approach.

Classic “one-step” methods use hormone analogs having molecular structures that claim to be totally non-reactive with serum proteins, but still able react with unoccupied hormone antibody sites ⁹⁸. If these methods were physiochemically valid, the hormone-analog, coupled to a signal-generating molecule such as an isotope or enzyme, could compete with free hormone for a limited number of anti-hormone antibody binding sites in a classic competitive immunoassay format. Unfortunately, although conceptually attractive, this approach was technically difficult to achieve in practice, despite early claims of success ^99,100. The hormone-analog methods were principally engineered to give normal FT4 values in high TBG states (i.e. pregnancy). They were found to have poor diagnostic accuracy in the presence of abnormal albumin concentrations secondary to Familial Dysalbuminemic Hyperthyroxinemic (FDH), NTI, high FFA concentrations or thyroid hormone autoantibodies ^{30,39,44,47,70,96,97,101-103}. During the 1980s, attempts were made to overcome the albumin dependence of these methods by adding propriety chemicals to block analog binding to albumin or by empirically adjusting calibrator values to correct for protein-dependent biases ^{29,93,100,104}. These attempts were unsuccessful and after a decade of criticism the hormone-analog methods have been abandoned in favor of the labeled antibody techniques described below ^{52,105,106-108}.

Most clinical laboratories now use free hormone methods based on this approach. Labeled antibody methods are competitive immunoassays that measure free hormone as a function the fractional occupancy of hormone-antibody binding sites ³. The methods use specific immunoabsorbants to quantify the unoccupied antibody binding sites in the reaction mixture. The physiochemical theory of analog-based labeled antibody methods would suggest that they would be as susceptible to the same errors as the labeled hormone-analog methods described above. However, the physicochemical differences arising from the binding of analog to the solid support confer kinetic differences that result in decreased analog affinity for endogenous binding proteins and more reliable free hormone measurements ^37,109. The labeled antibody approach is becoming the favored technique for automated free hormone testing ^16, ^356, ³⁵⁷.

The impetus for free hormone test development has been the high frequency of binding-protein abnormalities encountered in clinical practice. The only reason to select a free hormone method (FT4 or FT3) in preference to total hormone is to improve the diagnostic accuracy for detecting hypo- and hyperthyroidism in patients with thyroid hormone binding abnormalities that compromise the diagnostic utility of total hormone measurements. Situations where current FT4 methods may give misleading results are listed in Table 1. Unfortunately, the diagnostic accuracy of the free hormone methods cannot be predicted either from their methodologic classification (Table 2) or an in vitro test of technical validity, such as dilution ³⁷. The indirect index tests (FT4I) and the direct comparative methods are all protein dependent to some extent, and most are prone to underestimate FT4 in patients with binding protein abnormalities ^5,55,62,110. These binding protein abnormalities cause discordances between the total and free hormone concentrations. Unfortunately, no current FT4 method is universally valid in all clinical situations. However, pre-analytical or analytical assay artifacts may arise in a number of situations associated conditions not related to abnormal TBG concentrations: (a) when the binding of the assay tracer to albumin is abnormal; (b) in the presence of medications that displace T4 from TBG and (c) during critical NTI. Assay artifacts causing paradoxically abnormal FT4 values are more common than assay problems causing inappropriate TSH as shown in Table 1. Anomalous FT4 values are suspected when there is gross FT4/TT4 discordance or a history of medications affecting FT4 methodology. In these situations it is useful to check the FT4 result with another manufacturer’s FT4 test. Gross discordance between two methods suggests that the FT4 abnormality is a technical artifact and that TSH is the more reliable test, unless the discrepancy is due to central hypothyroidism(Table 1).

Common conditions that decrease the diagnostic accuracy of current free hormone tests in ambulatory patients include; a) TBG abnormalities (TBG excess or deficiency); b) Familial Dysalbuminemic Hyperthyroxinemia, FDH; c) T4 and T3 autoantibodies and d) interfering substances such as Rheumatoid Factor and Heterophilic antibodies (HAMA) [section 5] ^{30,103,111-113}. There is growing concern that the lack of method-specific, trimester-specific reference ranges for current FT4 tests renders these tests unreliable for assessing thyroid status during pregnancy when there are progressive changes in binding proteins. TT4 measurements may be a better option if the reference limits are adjusted for increased TBG by a factor of 1.5 ³⁵⁵. Additionally, free hormone tests have not received adequate testing in hospitalized patients in whom NTI and drug therapies decrease the diagnostic accuracy of both thyroid hormone and TSH testing ^44,114, ³⁵⁸. Ideally, three classes of hospitalized patients should be tested: a) patients without thyroid dysfunction but with low or high TT4 due to NTI; b) patients with hypothyroidism associated with severe NTI and, c) patients with hyperthyroidism associated with NTI ^40,44,115. However, since it is not feasible for the manufacturers to obtain such pedigreed patient specimens, it will continue to be difficult to ensure the diagnostic reliability of a FT4 test made in a sick hospitalized patient. A combination of FT4 plus TSH is recommended for assessing the thyroid status in this setting ^116, ³⁵³. In most clinical situations involving discordant FT4 and TSH results, the TSH test is usually the most reliable, provided that the patient is not receiving medications such as glucocorticoids and dopamine that directly inhibit TSH secretion, or the condition involves pituitary failure.(Table 1). When the diagnostic accuracy of the FT4 test is questionable, the demonstration of gross discordance between the relationship between FT4 and TT4 measurements may suggest problems with the FT4 and more rapidly available than an FT4 measurement made by a reference method employing physical separation.

4. TSH (Thyroid Stimulating Hormone/Thyrotropin) Assays

Over the last thirty years, the sensitivity and specificity of TSH assays have undergone dramatic improvements. The first clinical TSH assays developed in the 1960s used RIA methodology ^117-119. These methods were too insensitive to detect TSH in all euthyroid subjects and were limited to the diagnosis of primary hypothyroidism ^120-123. In 1970, the hypothalamic tripeptide, Thyrotropin Releasing Hormone (TRH) (Thyroliberin) was synthesized, and shown to stimulate serum TSH into the measurable range in all euthyroid subjects but not patients with hyperthyroidism or hypopituitarism ¹²⁴. These observations led to the practice of measuring serum TSH 15-30 minutes following a 200-500µg IV TRH dose as a way to overcome the insensitivity of the TSH RIA methodology ^124-126. The practice of TRH testing fell into decline after the development of the inherently more sensitive TSH immunometric assay (IMA) methods (also called "sandwich" or "noncompetitive" assays) that became available in the mid-1980s. These IMA techniques were based on the excess antibody approach of Miles and Hales, originally reported in the 1960s ¹²⁷. This new approach could not be universally adopted until the 1980s when monoclonal antibody technology facilitated the large-scale production of specific antibodies ^128,129. IMA methods were quickly found to be ten-fold more sensitive than the older RIA methods and as such were able to distinguish between the basal TSH levels of hyperthyroid and euthyroid subjects without TRH testing ^130-136. By 1990, IMA methodology had replaced most RIA methods and TSH euthyroid reference ranges became narrower as a result of reduced glycoprotein hormone cross-reactivity and improved precision ^137,138. In the last decade, TSH assay sensitivity has been further enhanced by the adoption of non-isotopic signals such as chemiluminescent and fluorescent molecules that are inherently more sensitive than isotopes and, in addition, are easier to automate ^4,139,145. Most current TSH methods are automated and report sensitivities down to 0.01 mU/L ^4,146-148. With this level of sensitivity, TSH measurement can be used to diagnose both overt and mild (subclinical) degrees of thyroid hormone excess without TRH stimulation ^{4,146,149,150}.

TSH assay "quality" has historically been defined by its clinical sensitivity – the ability to discriminate between hyperthyroid and euthyroid values ^151-153. During the RIA era (1965-1985) considerable efforts were made to develop methods that could achieve this clinical discrimination ^154-158. Although some success was reported using modified RIA procedures employing a pre-treatment lectin affinity extraction or a long pre-incubation period before tracer addition, these were principally research techniques ^159,157. The TSH IMA methods launched in the mid-1980s rapidly replaced the insensitive RIA tests because they could distinguish hyperthyroid from euthyroid values after only a relatively short incubation. Mechanistically, these IMA methods employed an excess of TSH monoclonal antibody, bound to a solid support (bead, tube, magnetic microparticle or adsorption gel) to capture TSH from the serum specimen during a 20 to 120 minute incubation period ^160-162. A different poly- or monoclonal TSH antibody, targeted to a different TSH epitope(s) and labeled with a signal, was included in the reaction mixture. After removing unbound constituents by washing, the signal bound to the solid support is directly proportional to the serum TSH concentration. Later modifications to this basic concept have included the use of fragmented monoclonal antibodies to reduce interference by heterophilic antibodies [section 5(c)] and the use of Avidin-Biotin and magnetic particle separation techniques ^163-165. The first IMA methods used a radioisotopic signal (I-¹²⁵) and were designated "immunoradiometric assays", or IRMAs ¹³⁰. Current TSH IMA methods use a variety of non-isotopic signals that have given rise to a lexicon of terminology to distinguish between assays using different signals. Immunoenzymometric assays (IEMA) use enzyme signals; immunofluorometric assays (IFMA) use fluorophors as signals, immunochemiluminometric assays (ICMA) use chemiluminescent molecules as signals and immunobioluminometric assays (IBMA) use bioluminescent signal molecules ^{131,140-142,166-168}.

Initially, TSH IMA methods were designated as "sensitive", "highly sensitive", "ultrasensitive" or "supersensitive" assays, terms used to distinguish the new IMA methodology from the older insensitive RIA methods then still in use ^138,169-173. This descriptive nomenclature was confusing and led to a debate concerning the meaning of "sensitivity" ^4,174. It is now accepted that TSH between-run precision should be the determinant of TSH assay sensitivity ¹⁷⁵. A new parameter “functional sensitivity”, defined as the TSH value associated with 20 percent CV established from assays run over a 6 to 8 week period, is now generally accepted as the determinant of the lower reporting limit for TSH assays ^4,353. During the last decade, a generational nomenclature system based on functional sensitivity (FS) has been widely adopted ^4,138. This nomenclature defines each generation as having a ten-fold difference in functional sensitivity. For example, the older RIA methods are designated "first generation" (FS = 1-2 mU/L); IRMA methods "second generation" (FS = 0.1 – 0.2 mU/L) and the even more sensitive current non-isotopic methods (IFMA, ICMA and IBMA) referred to as "third generation" (FS = 0.01 –0.02 mU/L) ^4,176,177.

Serum TSH values are higher in neonates and children ^178,179. Although serum TSH exhibits a diurnal variation with a peak between midnight and 0400, diagnostic information is not significantly influenced by the time of the blood draw ¹⁸⁰. The reference range for euthyroid adults has progressively contracted over the last two decades. Improvements in the sensitivity of TSH methodology over the last three decades, from first-generation (1G) radioimmunoassays to current 3G immunometric assays (IMA), have facilitated the determination of the lower TSH reference limit (~0.3-0.4 mIU/L) and are in part responsible for a contraction of the upper limit from ~10 to ~4 mIU/L (1G to 3G, respectively). This contraction also reflects the adoption of IMA methodology employing monoclonal capture antibodies that eliminate cross-reactivity with other glycoprotein hormones (a problem with the 1G RIAs) and the use of more sensitive and specific Thyroid Peroxidase Antibody (TPOAb) immunoassays that have replaced insensitive antimicrosomal antibody (AMA) tests for excluding individuals with autoimmune thyroid conditions, the inclusion of whom skew TSH reference limits ³⁵³.

There is a high degree of confidence that the mean TSH value for truly euthyroid subjects lies in the 1.0-1.5 mIU/L range, and that the lower 2.5% confidence limit approximates to 0.4 mIU/L. This being the case, the upper limit of the Gaussian distribution would be expected to approximate 2.5 mIU/L. However, a higher upper reference limit is cited for most current TSH assays (3.5 to 4.0 mIU/L) based on population studies ³⁵⁹. Possible reasons for the persistent ‘skew’ in the TSH upper limit include:

It is difficult to reliably exclude all individuals with thyroid dysfunction when determining the TSH reference range - given the high prevalence of thyroid dysfunction (especially subclinical hypothyroidism) in the general population exaccerbated by the sensitivity and/or specificity limitations of current TPOAb tests. Although a number of possible reasons for the persistent ‘skew’ in the TSH upper limit are listed above, the dominant cause is likely to be the inclusion of TPOAb-negative individuals with occult autoimmune thyroid dysfunction that is not detected as a humoral response (i.e. TPOAb) ³⁶⁰.

In fact, it has now been shown that the inclusion of TPOAb-positive persons in the cohort of subjects used for the TSH reference range skews the upper but not the lower limit ³⁵⁹. The need for rigorous exclusion of thyroid antibody-positive individuals is reinforced by the new NACB consensus guideline #22 that states, “TSH reference intervals should be established from the 95 % confidence limits of the log-transformed values of at least 120 rigorously screened normal euthyroid volunteers who have; (a) No detectable thyroid autoantibodies, TPOAb or TgAb (measured by sensitive immunoassay); (b) No personal or family history of thyroid dysfunction; (c) No visible or palpable goiter and, (c) Who are taking no medications except estrogen” ³⁵³. Figure 1 contrasts the new reference range for euthyroid subjects and the target range for L-T4 replacement therapy with the typical laboratory reference range (0.4-4.0 mIU/L).

The new TSH upper limit of 2.5 mIU/L is supported by the data from the Whickham follow-up study that found an increased risk of hypothyroidism in individuals with serum TSH > 2.0 mU/L, especially if TPOAb was present ¹⁸². This report suggests that when longitudinal study data are available, the TSH upper reference limit may fall further.

A number of current guidelines recommend the use of TSH as the first-line test for detecting both hypo- and hyperthyroidism in ambulatory patients with stable thyroid status and intact hypothalamic/pituitary function ^116,183. This strategy is considered more cost-effective than a panel approach (TSH + FT4 or FT4 + FT3) but necessitates the use of a TSH assay with a functional sensitivity below 0.02 mU/L (i.e. third generation) ¹⁸⁴. However, this TSH-first strategy can miss patients with central hypothyroidism or TSH secreting pituitary tumors ^185-188. As shown in Figure 1, third generation sensitivity (< 0.02 mU/L) is critical for the detecting subnormal TSH values, since less sensitive second generation assays are prone to produce falsely negative (normal range) values for sera with subnormal TSH concentrations ⁴. The reliable detection of subnormal TSH is critical for evaluating both ambulatory and hospitalized patients ¹⁸⁹. In older ambulatory patients, mild (subclinical) hyperthyroidism is a risk factor for atrial fibrillation ¹⁹⁰. In hospitalized patients, it is important to be able to distinguish the mild TSH suppression typical of euthyroid patients with NTI (0.02 –0.4 mU/L) from the profound TSH suppression typical of sick hyperthyroid patients ^44,46. Additionally, the use of a wider reference limit (0.02 –10 mU/L) is thought to improve its positive predictive value of TSH measurement for evaluating sick hospitalized patients (reflecting the author’s view that the tests in this condition are ”unreliable”- Ed) ^44,191.

Schematic contrasting the typical laboratory reference range (0.4-4.0 mIU/L with the new TSH reference ranges recommended by by the NACB guidelines ³⁵³.

5. Interferences with Thyroid Test Methodologies

It is difficult for the laboratory to proactively detect interference from a single measurement i.e. a first-line TSH test. Interference is more often suspected by the physician who observes that a reported value is inconsistent with the clinical status of the patient ¹⁹². Classic laboratory checks of analyte identity, such as dilution, may not always detect an interference problem ^193,194. The most practical way to test for an interference is to measure the specimen by a different manufacturers method and check for discordance between the test results reflecting that methods vary in their susceptibility to interfering substances ^195,353. In addition, in the case of suspected TSH interferences, the physician can perform a TRH-stimulation or thyroid hormone suppression test as a biological check ¹⁵⁰. Interferences producing a falsely elevated basal TSH values will usually be associated with a blunted (<3-fold increase) TRH response, or a less than the 90 percent suppression by 48 hours following the oral administration of 1mg of L-T4 or 200µg L-T3 ¹⁵⁰.

Interferences in competitive immunoassays and non-competitive IMAs fall into four classes: (a) crossreactivity problems, (b) endogenous analyte antibodies, (c) heterophilic antibodies and (d) drug interactions ¹⁹⁶.

The specificity of an assay depends on the ability of the antibody reagent to discriminate flawlessly between analyte and it’s structurally related target ligand ¹⁹⁷. TSH assays are more likely to be affected by cross-reactivity problems than thyroid hormone tests where chemically pure iodothyronine preparations are available for selecting antibody specificity ¹⁹⁸. The use of monoclonal antibodies for current TSH IMA methods has virtually eliminated cross-reactivity with glycoprotein hormones (LH or hCG) that plagued the early TSH RIA methods. However, even when using current assays an occasional patient will be identified with an unusual cross-reacting TSH isoform ^137,199,200.

In 1956 Robbins and colleagues first reported an unusual thyroxine binding protein globulin in serum ²⁰¹. Subsequently, both T4 and T3 as well as TSH autoantibodies have been identified in sera from patients with autoimmune thyroid and nonthyroid disorders ^202-207. Although there are a number of reports of anomalous total and free thyroid hormone, and TSH test values caused by T3, T4 or TSH autoantibodies, these autoantibodies rarely cause interference with current methods ^31,111,208. ^{31,197,198,200,} ^206,209. Endogenous antibody interferences are characterized by either falsely low or falsely high values, depending on the type and composition of the assay ^{31,205,210,211}. As discussed in section 7(e), autoantibodies against thyroglobulin (TgAb) still cause major problems with serum Tg measurement.

HAMA affect IMA methods more than competitive immunoassays by forming a bridge between the capture and signal antibodies, thereby creating a false signal, resulting in a high value artifact ^212, 361. The inappropriate value may not necessarily be abnormal but may be inappropriately normal ^213,214. Since antibodies cross the placenta, they have the potential to influence the neonatal screening result ²¹⁵. Heterophilic antibody interference falls into two classes:

These are relatively weak multispecific, polyreactive antibodies that are frequently IgM rheumatoid factor are often called human anti-mouse antibodies (HAMA) ¹¹³. HAMA has been reported to interfere with both total and free T4 and T3, as well as TSH measurements ^112,216-219. Approaches to reduce this form of interference have included the use of Fab fragments and hetero-species assay configurations ^164,220-222.

Specific human anti-animal (immunoglobulins) antibodies (HAAA) are produced against well-defined specific antigens following exposure to a therapeutic agent containing animal antigens (i.e. murine antibody) or by coincidental immunization through workplace exposure (i.e. animal handlers) ^113,223,224. Although assays for HAMA have been developed, inter-method differences are so large that these tests are currently not reliable for detecting heterophilic antibody interference ^225,226.

Drug interferences may relate to in-vitro or in-vivo effects. In-vitro affects result when the specimen contains a sufficient concentration of certain therapeutic and diagnostic agents to produce methodologic interference ^59,227. Alternatively, following heparin administration, the heparin in the specimen can cause in-vitro stimulation of lipoprotein lipase that liberates free fatty acids that inhibit T4 binding to serum proteins ^71,228. In-vivo effects on thyroid test results are caused by the therapeutic agents themselves (i.e. furosemide) competitively inhibiting thyroid hormone binding to serum proteins in the specimen, thereby causing abnormal (usually low) thyroid hormone test results ^72,229,230. In certain pathologic conditions such as uremia, abnormal serum constituents such as indole acetic acid may accumulate and interfere with thyroid hormone binding ²³¹. Thyroid test methods employing fluorescent signals may be sensitive to the presence of fluorophor-related therapeutic or diagnostic agents in the specimen ^57,232,233.

6. Thyroid Specific Autoantibodies (TPOAb, TgAb and TRAb)

Tests for antibodies against thyroid-specific antigens, anti-thyroid peroxidase (TPO), thyroglobulin (Tg) and TSH receptors are used in the diagnosis of autoimmune thyroid disorders ^8,234-236. Over the last four decades, antibody measurement techniques have evolved from semi-quantitative agglutination and complement fixation tests and whole animal bioassays to specific ligand assays using recombinant antigens and cell culture systems transfected with the human TSH receptor ^8,237-239. Unfortunately, the diagnostic and prognostic value of these thyroid autoantibody measurements is hampered by differences in the sensitivity and specificity of current methods. Although autoantibody tests have inherent clinical utility in a number of clinical situations, these tests should be selectively employed.

TPOAb antibodies were initially detected as antibodies against thyroid microsomes (antimicrosomal antibody, AMA) using semi-quantitative complement fixation and tanned erythrocyte hemaagglutination techniques ^240,241. Recent studies have identified the principal antigen in the AMA tests as the thyroid peroxidase (TPO) enzyme, a 100kD glycosylated protein present in thyroid microsomes ^242,243. Currently, manual agglutination tests are being replaced by automated, more specific TPOAb immunoassay or immunometric assay methods, based on purified or recombinant TPO ^244-251. Despite the use of the same International Reference Preparation (MRC 66/387), inter-method variability of current TPOAb assays is significant (correlation coefficients 0.65 and 0.87) ⁸. It appears that both the methodologic principles of the test, and the purity of the TPO reagent used, may influence the sensitivity, specificity and normal reference range of these methods ^8,239. The variability in the sensitivities (range <0.3 to >20 kIU/L) and reference ranges of the different methods has led to different concepts regarding the "normality" of a detectable TPOAb ⁸. Specifically, assays characterized by a low detection limit (<10 kIU/L) typically report unmeasureable TPOAb values for normal subjects, suggesting that the detection of this autoantibody is a pathologic finding (²⁵²). In contrast, assays reporting higher detection limits (>10kIU/L) typically cite a TPOAb "normal range", suggesting that low levels of this autoantibody are compatible with normal physiology ²⁵³. Whether these detectable "normal" values reflect physiology or poor assay specificity remains to be determined

High between-method variability precludes the numeric comparison of serum TPOAb values measured by different tests ^{8,237,239,247,251}. An abnormal TPOAb is detected in 12 to 14 percent of "healthy" euthyroid subjects and even higher percentages of patients with various non-thyroid autoimmune disorders ¹⁸¹. Approximately 70-80 % of patients with Graves' disease and virtually all patients with Hashimoto's, atrophic thyroiditis or post-partum thyroiditis have TPOAb detected ^{239,247,251,252,254}. In fact, TPOAb is implicated as a cytotoxic agent in the destructive thyroiditic process ^255,256. TPO antibodies are involved in the tissue destructive processes associated with the hypothyroidism observed in Hashimoto’s thyroiditis. Some studies suggest that TPOAb itself may be cytotoxic to the thyroid. Estimates of TPOAb prevalence depend on the sensitivity and specificity of the method employed ³⁶². The recent NHANES survey of ~17,000 subjects without apparent thyroid disease, reported that TPOAb was detected in 12.6 % of subjects using a competitive immunoassay method ¹⁸¹. As shown in Figure 2, the odds ratio for hypothyroidism was strongly associated with the presence of TPOAb and not TgAb ¹⁸¹.

In the future, TPOAb measurement may be used as a prognostic indicator for thyroid dysfunction. Although the appearance of TPOAb usually precedes the development of thyroid dysfunction, recent studies suggest that a hypoechoic ultrasound pattern may precede a biochemical TPOAb abnormality, as shown in Figure 3 ³⁶⁰. The paradoxical absence of TPOAb in some patients with unequivocal TSH abnormalities likely reflects the suboptimal sensitivity and/or specificity of current TPOAb tests or non-autoimmune thyroid failure (atrophic thyroiditis) ^181, ^360, ³⁶¹.

Although changes in autoantibody concentrations often reflect a change in disease activity, serial thyroid autoantibody measurements are not recommended for monitoring treatment for AITD. This is not surprising since treatment of AITD addresses the consequence (thyroid dysfunction) and not the cause (autoimmunity) of the disease. The presence of TPOAb has now been established as a risk factor for developing thyroid dysfunction in patients taking Amiodarone, Interferon-alpha, Interleukin-2 or Lithium therapies ^263-266.

Longitudinal studies suggest that even low levels of TPOAb, that could not be detected by the insensitive AMA agglutination tests used in the 1970s, are likely reflect the presence of occult thyroid dysfunction and are a risk factor for the development of clinical hypothyroidism over a timescale of years or decades ³⁶⁵. Specifically, as shown in Figures 2 and 3, a detectable TPOAb is a risk factor for hypothyroidism and typically precedes the development of an elevated TSH ^359, ^360, ³⁶⁵. In addition, the presence of TPOAb has been linked to reproductive complications such as miscarriage, infertility, IVF failure, fetal death, pre-eclampisa, pre-term delivery and post-partum thyroiditis and depression ^257-262, ^366-371. It is generally considered cost-effective to replace semi-quantitative AMA agglutination tests by TPOAb immunoassay methods, since the enhanced sensitivity and specificity of the new immunoassays obviates the need for additional TgAb measurements in the routine diagnosis of autoimmune thyroid disorders ^8,181.

Antithyroglobulin autoantibodies (TgAb) were the first thyroid antibodies to be recognized to circulate in patients with autoimmune thyroid disorders. The first TgAb methods were based on tanned red cell hemagglutination ²⁴¹. Subsequently, methodologies have evolved in parallel with TPOAb methodology from semi-quantitative techniques, to more sensitive ELISA and RIA methods and more recently chemiluminescent immunoassays ^247,252,253). Unfortunately, the inter-method variability of current TgAb assays is even greater than that of the TPOAb tests discussed above ⁸. This variability likely reflects both the purity and the epitope specificity of the Tg protein reagent, as well as the inherent heterogeneity of the antibodies present in different patient sera ²⁶⁷. As with TPOAb methods, TgAb tests report a wide range of sensitivity limits (<0.3 to >20 kIU/L) as well as reference ranges, despite the use of the same International Reference Preparation (MRC 65/93). There is some data to suggest that low levels of TgAb may be present in normal individuals ²⁵³. However, studies suggest that the TgAb epitope specificity expressed by normal individuals and patients with either differentiated thyroid carcinoma or thyroid autoimmune disorders may be qualitatively different ^253,268. These differences in test specificity can impact the suitability of a TgAb method for screening sera for prior to serum Tg measurement [section 7 (e)].

Autoantibodies against Tg are encountered in autoimmune thyroid conditions, usually in association with TPOAb ²⁶⁹. However, the recent NHANES III study found that 3 % of subjects with no risk factors for thyroid disease had detectable TgAb without TPOAb ¹⁸¹. In these subjects with only TgAb detected, no association with TSH abnormalities was found so that the clinical significance of an isolated TgAb abnormality remains to be established. This suggests that it is unnecessary to measure both TPOAb and TgAb for a routine evaluation of thyroid autoimmunity ⁸. TgAb is primarily used as an adjunctive test for serum Tg measurement.

Current guidelines recommend that all sera be prescreened for TgAb by a sensitive immunoassay method, prior to Tg testing ^7, ³⁵³. Immunoassay methods detect TgAb in approximately 20% of patients with differentiated thyroid carcinoma (DTC) compared with 10 % of normal subjects ^252,270. There appears to be no threshold TgAb concentration above normal that precludes TgAb interference ²⁵². When patients with differentiated thyroid carcinoma have TgAb detected, serial measurements of TgAb concentrations serve as an independent parameter for detecting changes in tumor mass ^52,271-273, ³⁷². For example, TgAb-positive patients who are rendered athyreotic by surgery display progressively declining TgAb concentrations during the first few post-operative years. Such patients usually become TgAb-negative after four years. In contrast, a rise or appearance of serum TgAb concentrations is often the first indication of recurrence. When using serial TgAb measurements as a surrogate marker for changes in tumor burden, between-method differences necessitate the use of the same manufacturer’s TgAb test kit ^7,252. Usually, serial serum Tg RIA and TgAb changes in such patients are concordant ^252,271. However, a disparity between these two parameters (rising TgAb/declining Tg RIA) may indicate recurrence, since as tumor mass increases and more Tg is secreted more Tg-TgAb immune complexes may form, which may be cleared more rapidly ²⁷⁴. Early data suggesting that quantitation of Tg mRNA in peripheral blood might be an additional tumor parameter for detecting the presence of thyroid tissue in TgAb-positive patients with DTC has now been discredited ^275,276. Recent studies report that false positive results may arise from assay artifacts or illegitimate transcription ^382, ³⁸³. Additionally, false negative results are seen in patients with metastatic disease ^379, ^380, ³⁸¹.

TSH Receptor Antibodies (TRAb) were first recognized as a long-acting thyroid stimulator (LATS) using mouse bioassays ²³⁵. These autoantibodies are directed against epitopes on the ectodomain of the TSH receptor ²⁷⁷. Studies carried out by several laboratories suggest that TRAb epitopes, as well as TSH binding domains, are distributed, throughout the extracellular parts of the receptor. However, the complex structure of the TSH receptor makes it difficult to localize the epitopes targeted by these autoantibodies. Methods for measuring TRAb are even more varied than for TPOAb and TgAb ²⁷⁸. Two classes of TRAb can be associated with autoimmune thyroid disorders – (a) thyroid stimulating autoantibodies (TSAb) that cause Graves’ hyperthyroidism and (b) thyroid stimulation-blocking antibodies (TBAb) which block receptor binding of TSH ⁸. Attempts to distinguish stimulatory from inhibitory TRAb epitopes have not provided clear results ²⁷⁹. Each class of TRAb (TSAb and TBAb) may be detected alone or in combination in Graves’ disease and Hashimoto’s thyroiditis ²⁸⁰. The relative concentrations of the two classes of TRAb may modulate the severity of Graves’ hyperthyroidism and may change in response to therapy or pregnancy ^280-282. Two different methodologic approaches have been used to quantify TRAb (283). Bioassays have been used to measure the stimulating class of antibodies (TSAb) ^284,285. In contrast, radioreceptor assays using isolated, solubilized or more recently, recombinant human TSH receptors have been used to develop thyroid-binding inhibition immunoglobulin (TBII) assays that detect both classes (stimulating and blocking) of TRAb ²⁸⁶.

These bioassay techniques use the cAMP second messenger system as the biological endpoint of activity. Assays have been based on both homogeneous (human) and heterogeneous (animal) cells expressing TSH receptors. Early homogeneous assays used surgical human thyroid specimens, whereas heterogeneous assays used mice, guinea pigs and later rat FRTL-5 cell lines ^235,284,287. Recently, assays have been developed using cell lines transfected with the human TSH receptor ^285,288,289.

These radioreceptor assays measure the ability of the serum or an immunoglobulin preparation to inhibit the binding of 125I- labeled TSH to solubilized TSH receptors ²⁸⁶. Typically, animal tissues (guinea pig fat cells or porcine thyroid membranes) have been used for TSH receptor preparation. Recently, methods using human receptor expressed in CHO cells or recombinant receptor protein have been developed Cloning of the TSH receptor greatly benefits bioassay development, as TSH receptor transfected cell lines are more useful tools than thyroid cells ^290,291. Currently, inter-method differences are wide and the interassay precision (CVs 15.2 to 21.6%) of the different tests is very variable so that is difficult to compare values from different methods ⁸. New chimeric assays are becoming available, which may be able to target the loci of TRAb epitopes and TSH binding sites may better link assay responses to clinical outcome ^292-294. New “second generation” assays employing human recombinant TSH-Receptor, are reported to offer improved sensitivity for diagnosing Graves’ disease without a loss of specificity ²⁹⁵. However, preliminary studies of disease suggest that such methods do not appear to offer improved predictability for the response to anti-thyroid drug (ATD) treatment ²⁹⁶.

TRAb tests are used in the differential diagnosis of hyperthyroidism, the prediction of fetal and neonatal thyroid dysfunction due to transplacental passage of maternal TRAb and prediction the course of Graves’ disease treated with antithyroid drugs ^{235,297,273,298}. Most measurements are made with TBII tests because these methods are available in commercial kit form ⁸. Although TBII assays do not directly measure the stimulating antibodies, these tests have comparable diagnostic sensitivity to TSAb bioassays (70-95%) for diagnosing Graves’ hyperthyroidism or detecting a relapse or response to therapy ^{280,286,287,297,299}. The second generation assays employing human recombinant TSH receptor are now becoming available and are reported to have superior diagnostic sensitivity for Graves’ disease ³⁰⁰. Current tests are manual and expensive and vary in precision, sensitivity, specificity and reference ranges ⁸. TBII test use is geographically sensitive to differences in practice patterns and reimbursement. In the United States, where reimbursement restricts biochemical testing, TRAb tests are rarely used and thyroid hormone and TSH measurements are preferred for diagnosing or monitoring treatment for Graves’ hyperthyroidism ¹⁸³. However, the TBII tests are important for evaluating pregnant patients with a history of autoimmune thyroid disease, in whom there is a risk of transplacental passage of TRAb to the infant ³⁰¹. Both stimulating and inhibiting activities should be tested since the expression of thyroid dysfunction may be different in the mother and the infant ³⁰². The lack of specificity of the TBII methods is actually an advantage in this clinical situation, since a TBII test will detect both the stimulating and blocking classes of TRAb that can produce transient hyper- or hypothyroidism, respectively, in the fetus and newborn. TRAb plays an uncertain role in thyroid-associated opthalmopathy (TAO), which appears to be exacerrbated by radioiodine therapy ³⁰³. Since TRAb and other thyroid antibodies levels increase acutely significantly after radioiodine therapy, a TRAb measurement prior to radioiodine therapy may be useful to predict risk of TAO ^304-307. However, prospective studies are needed to establish the clinical utility of TRAb measurement in this context. Patients with very high circulating concentrations of hCG due to choriocarcinoma or hydatiform mole, as well as a small number of pregnant patients, may have misleading positive results using TSAb assays ²⁷³.

7. Thyroglobulin (Tg) Methods

Serum Tg measurement is used as a tumor marker in the management of patients with differentiated thyroid carcinomas (DTC) ^308,309. Current Tg methods are based either on IMA or RIA techniques ^7,310,311. There is a trend for non-isotopic IMA methods to replace RIA methods because IMA methods are easier to automate, have shorter turn around times, wider working ranges and use reagents with a longer shelf life ^7,312,313.

Unfortunately, serum Tg measurement is technically challenging and there are five methodologic problems that impair the clinical utility of the test: (a) between-method biases; (b) suboptimal sensitivity; (c) suboptimal interassay precision; (d) "hook" problems (IMA methods) and (e) Tg autoantibody (TgAb) interference ⁷.

Current methods should be standardized on the International Reference Preparation CRM-457. Even when methods are standardized against CRM-457 there can be a two-fold difference between serum Tg values as shown in Figure 4 ^7,314-317, ^353, ³⁷³. This magnitude of between-method bias necessitates that the same Tg method be used for the serial monitoring of patients. Laboratories considering changing to a Tg method having a bias of >15% relative to their current method should alert physicians and allow them sufficient time to re-baseline patients. The causes of Tg method variability are multifactorial and include standardization differences, matrix effects and Tg epitope recognition differences. The typical serum Tg reference range for a CRM-457 standardized method approximates 3 to 40 µg/L.

Serum Tg should be detectable in the sera of all normal euthyroid subjects. As shown in Figure 4, current Tg methods can barely discriminate between the lower limit of the euthyroid reference range and the assay detection limit ^7, ³⁷³. TSH suppression, which reduces serum Tg by ~50%, exaccerbates assay sensitivity limitations ³¹⁸. Insensitive assays are unable to detect small amounts of tumor, especially when TSH is suppressed. A recent metaanalysis reports that 20% of patients with undetectable serum Tg (<1µg/L) during thyroid hormone suppression therapy (THST) display detectable serum Tg values 72 hours after recombinant human TSH (rhTSH) administration ^319, ³⁷⁴. Recent reports also suggest that more sensitive Tg assays will have improved clinical sensitivity for detecting disease without the need for TSH stimulation ³⁷⁵. As discussed relative to TSH in Section 4(a), Tg assay functional sensitivity should be calculated from the 20% between-run coefficient of variation. Recommendations are that in the case of Tg assays the human serum pools used for precision assessment should not contain any detectable TgAb and the precision should be assessed across the typical clinical interval used for the test, which in the case of serum Tg is 6-12 months ³⁵³.

Normal euthyroid individuals with stable TSH status exhibit intra-person biologic variability of ~14 % across a 4-month period ³²⁰. When serum Tg is being monitored in a DTC patient on suppression, any serum Tg increase in excess of 15% should prompt an evaluation for recurrence. Unfortunately, Tg assay precision erodes over the long interval typically used for monitoring DTC patients (6 to 12 months). Unfortunately, the between-run precision of most Tg tests exceeds 15% across this interval when measuring Tg values below the lower reference limit. This precision problem has the potential to mask the detection of clinically important changes ^7,308. One approach to eliminate interassay imprecision is to remeasure the patient’s archived specimen from the previous evaluation alongside the current specimen in the same assay ³⁰⁸.

Tumor marker IMA methodologies are prone to "high-dose hook effects" ³²¹. This is characterized by inappropriately low values in sera with high analyte concentrations. The problem is caused by an excess of analyte in the specimen overwhelming the binding capacity of the solid-phase capture antibody and inhibiting the ability of endogenous analyte to form a bridge between the two antibodies, resulting in an inappropriately low value. Some Tg IMA methods are prone to this hook problem ⁷. Manufacturers are generally aware of this problem and attempt to overcome it by designing assays using either a two-step process, or recommend that each specimen be run at two dilutions. When an unexpectedly low serum Tg value is reported for a patient with known metastases, the physician should request the laboratory to repeat the measurement at multiple log dilutions to rule out a "hook" problem.

Thyroglobulin autoantibody (TgAb) interference with serum Tg measurements remains the most serious problem limiting the clinical value of serum Tg measurement ^252, ³⁵³. It is important that the laboratory use a sensitive TgAb immunoassay for identifying sera containing TgAb. Serial TgAb measurements can be used as an independent prognostic test for the presence of Tg-secreting thyroid tissue ^{252,272,271,372}. The interpretation of a Tg value depends on the TgAb status of the patient and the class of Tg method used (RIA or IMA). The direction and magnitude the interference relates to the method and the concentration and affinity of the TgAb in the specimen. Typically, unidirectional interference (underestimation) is typical of IMA methods, which display a disproportionately high percentage of low and undetectable serum Tg values when patients are TgAb-positive ³¹¹ (Figure 4b). Some Tg IMA methods claim to overcome TgAb interference by using monoclonal antibodies directed against epitopes not involved in thyroid autoimmunity ³¹³. Although conceptually attractive, this approach does not appear to overcome interference problems in practice, possibly because less restricted TgAb epitopes are associated with thyroid carcinoma than with the autoimmune thyroid conditions ^322,323. In contrast, TgAb can cause bi-directional interference (either over- or underestimation) with serum Tg RIA measurements. The magnitude and direction of the interference is determined by the factors that affect the partitioning of the Tg tracer between Tg antibody reagent and the endogenous TgAb ³²⁴. In general, fewer low and undetectable values are seen for TgAb-positive sera measured by RIA techniques that appear to represent the measurement of both free Tg and Tg complexed with TgAb. ³¹¹.

Figure 4 shows serum Tg measurements made in TgAb-negative and TgAb-positive sera, measured by both classes of method (IMA and RIA) ³⁷². The between–method differences evident in the absence of TgAb (~30% CV) (Figure 4a) reflect the specificity differences between methods. When TgAb is present, sera typically have detectable Tg RIA values and lower or undetectable IMA values. This RIA/IMA discordance appears to be a better indicator of TgAb interference than exogenous recovery tests ³¹¹. In fact the new NACB consensus guidelines recommend abandoning recoveries and recommend the measurement of TgAb concentrations directly ³⁵³. The mechanism responsible for the low IMA values appears to be an inability of Tg complexed with TgAb from participating in the two-site reaction ^324,325. When a patient has detectable TgAb, clinical judgment should not be based on an undetectable serum Tg IMA value. It is difficult to prove which class of method (RIA or IMA) most accurately reflects the Tg status of a TgAb-positive patient. It is generally considered that serum Tg underestimation is more problematic than overestimation, since inappropriately low values may mislead the physician into believing that the patient is free of disease, even when metastatic disease is present ²⁵². In contrast, inappropriately high values that are typical of some RIA methods may prompt unnecessary imaging studies but will have the advantage of promoting a higher level of patient surveillance. A number of laboratories now believe that TgAb-positive patients are best monitored by serum Tg RIA measurements that appear to correlate better with clinical progression or regression of disease ^{252,308,323,326,327}. Many laboratories using IMA methods are currently still reporting falsely undetectable serum Tg values for TgAb-positive patients, albeit with a cautionary comment ⁷. Other laboratories are adopting a dual strategy for serum Tg measurement – i.e. restricting the use of the IMA methods to TgAb-negative sera and retaining RIA methodology for TgAb-positive sera.

Serum Tg concentrations measured in 88 TgAb-negative (panel a) and 36 TgAb-positive (panel b) euthyroid volunteers (TSH 0.5-2.5 mIU/L) using different methods. Method #1=University of Southern California RIA, Los Angeles, CA, USA; Method #2=Nichols Bead ICMA, Nichols Institute Diagnostics, San Juan Capistrano, CA, USA Method #3= Nichols Advantage ICMA, Nichols Institute Diagnostics ICMA, San Juan Capistrano, CA, USA; Method #4= Access ICMA, Beckman-Coulter, Fullerton, CA USA. Method #5= Immulite ICMA, Diagnostic Products Corporation, Los Angeles, CA, USA. The shaded area indicates the manufacturers recommended sensitivity (not functional sensitivity). Lines indicate median values. The symbols in panel b indicate the values for an individual patient measured across the methods.

The technical problems with serum Tg measurement discussed above require a strong physician-laboratory interface for the optimal management of DTC patients. Physicians need to understand the basic technical issues concerning serum Tg measurement, and develop laboratory partnerships that provide an appropriate level of support. It is important to recognize that the serum Tg concentration integrates three principal factors: (1) the mass of differentiated thyroid tissue present (normal tissue + tumor); (2) any inflammation of, or injury to thyroid tissue, such as follows fine needle aspiration biopsy, surgery, radioiodine therapy or thyroiditis; and (3) the degree of stimulation of TSH receptors (by TSH, hCG or TSAb). Serum Tg measurement is useful in four phases of management of DTC patients:

Radioimmunoassay methods are difficult to automate since they require a physical separation of antibody-bound from free tracer. Once homogeneous methods based on monoclonal antibodies were developed, significant progress was made in automating thyroid test immunoassays. The current trends in automation are geared towards high-throughput, modular, robotic systems that incorporate both immunoassay and clinical chemistry analyzers into one instrument ^329,330. Tests for TT4, TT3, THBR, TSH, Tg, TPOAb and TgAb using non-isotopic (primarily chemiluminescent) signals, are becoming available on a variety of immunoassay analyzer platforms which employ bar-coding, multiple-analyte random-access, primary tube sampling, autodilution, STAT testing and computerized data output ^16,331-333. Laboratories primarily select an analyzer to perform thyroid testing on the basis of instrument menu and operating costs, and only secondarily on considerations of functional performance. Although the move to automation is seen as cost-effective, the consolidation of a diversity of immunoassay tests onto one platform has led to a transfer of thyroid testing from small specialized laboratories to the general chemistry laboratory setting. This centralization has resulted in a loss of laboratory expertise for the clinically interpretation of thyroid tests. This has negatively impacted the ability of laboratory staff to discuss reasons for discordant test results with physicians. Current trends towards point-of-care testing using miniaturized and biosensor technology would appear to bring laboratory personnel closer to the patient ^334,335. However, near-patient testing is only more cost effective when immediate diagnosis and therapy reduces hospitalization costs ³³⁴. Since most thyroid disease is treated on an outpatient basis, point-of-care thyroid tests are unlikely to replace centralized automated thyroid testing.

FOR A DOWNLOAD OF THIS CHAPTER IN PDF OR WORD FORMAT, CLICK HERE

1. Introduction