Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

Contents lists available at ScienceDirect

Best Practice & Research Clinical Endocrinology & Metabolism

journal homepage: www.elsevier.com/locate/beem

Medications that distort in vitro tests of thyroid function, with particular reference to estimates of serum free thyroxine
Jim R. Stockigt, MD, FRACP, FRCPA, Professor of Medicine, Consultant Endocrinologist a, b, *, Chen-Fee Lim, MAACB, MAgrSci, PhD, Head of Immunoassay Laboratory c
a

Monash University, Melbourne, Australia Epworth and Alfred Hospitals, Melbourne, Australia c Dorevitch Pathology, Melbourne, Australia
b

Keywords: free T4 interference sample dilution aspirin furosemide carbamazepine non-steroidal anti-inammatory drugs non-esteried fatty acids heparin serum total T4

The combination of serum thyroid-stimulating hormone (TSH) with measurement of circulating thyroid hormones greatly improves sensitivity and specicity of thyroid diagnosis, but these assays are not impeccable. Estimation of serum free T4 conveniently accommodates variations in the concentration of thyroxine-binding globulin (TBG), but no current technique reliably reects the in vivo free T4 concentration in numerous other situations. The effect of circulating competitors that increase T4 and T3 in vivo, in particular, many medications, is under-estimated by current free hormone estimates that involve sample dilution. Non-esteried fatty acids generated during sample storage and incubation can spuriously increase the measured free T4 estimate, especially after in vivo treatment with heparin. These artefacts are unlikely to be overcome by current assay strategies. Total serum T4, corrected for alterations in TBG concentration, gives a more robust estimate of thyroxine concentration than current methods of free hormone estimation and should now be reintroduced as the gold standard. 2009 Elsevier Ltd. All rights reserved.

* Corresponding author. Monash University, Melbourne, Australia. E-mail address: jrs@netspace.net.au (Stockigt). 1521-690X/$ see front matter 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.beem.2009.06.004

754

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

Introduction The aim of in vitro testing of thyroid function is to accurately reect the in vivo activity of key analytes such as thyroid-stimulating hormone (TSH), thyroxine and triiodothyronine. For these key laboratory parameters, it is highly desirable for results to be interpretable in relation to consensus reference intervals that do not vary depending on the choice of method. In this article, we consider whether current assays satisfy these objectives and outline situations in which the results may be potentially misleading, with particular attention to the in vitro effects of medications. Drugs that alter the secretion of TSH or thyroid hormones in vivo are excluded, as are the syndromes of thyroid hormone resistance1 and situations in which immunoreactivity of TSH may not accurately reect the biologic activity of this glycoprotein.2,3 The major part of this review will consider how free T4 and T3 estimates are inuenced by serum constituents that cause discrepancies between the in vivo hormone level and the apparent concentration in the assay tube. These effects reect the difculty of estimating free hormone concentrations in the presence of proteins that bind numerous other ligands in addition to T4 and T3. Free hormone estimates may be spuriously high if components that displace T4 and T3 from protein binding, such as non-esteried fatty acids (NEFA), are generated during sample storage or incubation, a particular problem in heparin-treated patients. Conversely, the measured free hormone concentration may be falsely low in diluted samples that contain medications that compete for protein binding of T4 or T3. Assays for TSH can be subject to interference by heterophilic antibodies, which may also distort the measurement of other analytes. Assays for T4 and T3, either total or free, are vulnerable to interference from circulating auto-antibodies, which produce divergent artefacts in the assay, depending on the technique that is used to separate bound from free hormone (see below). It is important to re-emphasise that all current assays for thyroid hormones and TSH involve the comparison of known and unknown samples, based on the assumption that the assay signal is solely a function of the concentration of analyte in the sample at the moment of collection. If any of the following three conditions is breached, the assay is likely to give a spurious result. a. Competition for antibody binding between the assay tracer and free hormone is identical in samples and standards. b. Any constituent, generated during sample storage and incubation, that inuences the concentration of analyte or the assay signal, is similar in samples and standards. c. Dilution-dependent dissociation of bound hormone is similar in samples and standards. The impact of misleading individual assay results can be minimised by considering serum T4 and TSH together, taking account of their characteristic feedback relationship and the assumptions and limitations that underpin this diagnostic relationship.4 While there is no single artefact that concurrently distorts the results for serum free T4 and TSH to simulate a false diagnosis, the combination of a low free T4 estimate with suppressed serum TSH is a challenging and ambiguous nding that may be attributable to severe illness, or to the in vivo and in vitro effects of medications. While true central hypothyroidism can occur as a neuroendocrine effect of critical illness5, a similar combination of results can be induced by medications, for example, when high dose furosemide is given together with dopamine or glucocorticoids. The latter two each suppress serum TSH4, while the former displaces T4 from its binding proteins6 and also leads to spuriously low estimates of free T4 due to a dilutiondependent in vitro artefact (see below). Whenever the cause of an apparent thyroid abnormality is unclear, it is crucial to take account of medications.6

Effects of sample storage It is widely accepted that thyroid hormones and TSH are stable when serum is stored at 4  C for up to 1 week.7 Stability on T4 and TSH of lter paper spots used for neonatal screening when stored at 4  C has been conrmed for at least 1 year.8

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

755

Whether binding proteins remain stable over long periods is less certain, although there is no evidence to the contrary. The storage artefact that is most likely to inuence free hormone estimates is the generation of NEFA from serum triglyceride, particularly in heparin-treated patients (see below). Sample carry-over When serum is sampled on automated assay platforms, sample carry-over has the potential to distort results. This issue is particularly relevant for an analyte such as TSH, where the concentration may vary at least 1000-fold between severe primary hypothyroidism and hyperthyroidism. The limited information that is available is reassuring. For example, it was estimated that carry-over in TSH assays was <1:10 000 on two automated platforms.9 Nevertheless, results that are clinically inconsistent should always be conrmed with a further sample (rather than a repeat assay of the same sample), before committing a person to long-term treatment. Heterophilic antibody interference Interference by heterophilic antibodies, produced either as a result of close contact with animals (e.g., mice, rabbits or goats) or after injection of mouse monoclonal antibodies for diagnostic or therapeutic purposes10, can lead to spurious results in two-site immunometric assays for TSH. Interference is caused by generation of a false assay signal by the formation of a bridge between the capture and signal antibodies in the absence of the analyte, which usually gives falsely elevated results. It is also possible that antibody binds only to the capture antibody, thus sterically blocking the binding of the analyte to the capture antibody, causing artefactually low results. Immunometric assays are usually more susceptible to heterophilic antibodies interference than competitive assays, probably due to the use of lower afnity antibodies in immunometric assays.11 Numerous other immunoassays, including cardiac and tumour markers, and human chorionic gonadotrophin can also be affected by heterophilic antibodies.11 While heterophilic antibodies are frequently detectable in routine clinical specimens, assay interference is rare. Ward et al.12 reported a prevalence of 30%, but found discordant results attributable to heterophile antibodies in only seven of 21 000 patients (0.033%). The standard method to minimise heterophilic antibody effects is to add puried non-immune globulin or normal serum from the species used to produce the antibodies.13 Specic blocking agents for heterophilic antibodies have been developed, such as a commercially available heterophile blocker (Scantibodies), which contain freezedried immunoglobulins that bind and neutralise the heterophile antibodies.14 While this is a useful approach, it is not awless.15 Ross and Menheere16 demonstrated that sample dilution and use of Scantibodies blocking tubes did not reliably identify some heterophilic TSH interference that was removed by protein G afnity chromatography. The nding of clinically discordant results, in particular, an anomalous relationship between the free T4 estimate and serum TSH that is out of line with the clinical context, is usually the initial clue to such interference. Non-linearity with sample dilution is the most common laboratory procedure in screening for potential heterophilic antibody interference, but this procedure lacks sensitivity.15,16 If interference is suspected, re-testing of the sample with antibodies sourced from another species on an alternative assay platform may be denitive.16 Heterophilic antibody interference with assays for total and free T4 has also occasionally been reported.17 Endogenous T4- and T3-binding antibodies Circulating T3- or T4-binding auto-antibodies may result in artefactual total and free T4 and free T3 measurements18 and can occasionally affect the carriage of circulating T4 and T3 in vivo.18 Tracer T4 or T3 bound to the endogenous antibody will be falsely classied as bound in adsorption separation methods, or free in double antibody methods, leading to falsely low or falsely high T4 or T3 values, respectively. Notably, analogue tracers may also bind to these antibodies, leading to spuriously high serum free T4 estimates.19 Where the total concentration of hormone is in doubt because of potential

756

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

antibody interference, assay of total hormone in suspect and control samples after ethanol extraction is denitive.18 Anomalous protein binding of assay tracer Labelled analogues of T4, designed to minimise their protein binding in serum, have been used as tracers in free T4 assays for over 25 years. That objective is readily achieved for thyroxine-binding globulin (TBG), but not for albumin.20 When the labelled analogue is protein-bound to a greater extent in the unknown than in the standard serum, less tracer is available to compete for the assay antibody, giving a falsely high free T4 estimate, as in familial dysalbuminemic hyperthyroxinemia20 and with iodothyronine-binding immunoglobulins.19 Conversely, if serum albumin is subnormal, or its binding sites are occupied by other ligands21, free T4 estimates by analogue tracer methods tend to be low, because more tracer is available in the sample than in the standard. The early analogue free T4 methods were so highly dependent on albumin that the free T4 estimate was virtually zero in euthyroid subjects with hereditary analbuminaemia.20 In renal failure, analogue assays give low serum free T4 values by up to 40% in pre-dialysis samples21 (see below). The term T4 or T3 analogue is used in two different senses. The term analogue tracer assay is best conned to methods where a labelled, chemically modied T4 molecule and antibody are incubated with the serum sample; serum free T4 and the labelled analogue compete for a limited number of antibody sites. In a different sense, labelled analogues may be used to quantify unoccupied solid-phase binding sites after the serum component has been removed. Two-step assays for free T4, in which a tracer is added after the serum component has been removed, are generally free of this artefact. Competitor inuences on measured serum free T4 and T3 concentrations In contrast to the binding proteins that carry steroid hormones, which show little interaction with other ligands, binding sites for circulating thyroid hormones show broad interaction with NEFAs and numerous common medications22,23 (Table 1). Displacement of T4 and T3 from binding proteins can distort diagnostic tests and may inuence hormone delivery and clearance. Despite the fact that important drug competitors for T4 binding to TBG have afnities between three orders of magnitude (furosemide) and almost seven orders of magnitude (aspirin) less than that of T4 itself22,23, they may cause signicant displacement of T4. Afnity for the specic binding site is only one of three major

Table 1 T4 displacement in vitro by drug competitors at relevant therapeutic concentrations; undiluted serum, 37 C. Medication Fenclofenac Aspirin Meclofenamic acid Diunisal Mefenamic acid Phenylbutazone Flufenamic acid Diclofenac Carbamazepine Phenytoin Furosemidee Iopanoic acidf Heparin
a b c d e f a Afnity for TBG (relative to T4 103) b

Free Fraction percent

Therapeutic Concentration umol/l

T4% displacement (undiluted serum)

1.5 0.03 8 0.07 2.7 0.04 0.6 3.9 0.3 11 0.1 <0.01

0.5 11 0.2 0.7 0.6 0.9 0.2 0.5 5.8 1.8 1.2

270 1800 60 320 80 320 70 10 40 80 330 200 2

90 62 39 37 31 31 10 7 44d 1345d 530e 8 <1

Afnity for TBG in isolation relative to thyroxine at 4C (103) (data from ref 24). Undiluted serum, spectrophotometric method, data from ref 34. Data from refs 22, 27, 31. Data from ref 27. Effect at high dosage >500 mg/day; effect at lower doses if renal function impaired. Concentrations during oral cholecystography.

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

757

factors that inuence competitor potency. Total therapeutic concentrations of important drug competitors range from almost 2000 mM for aspirin and 330 mM for furosemide.22,23 In addition to afnity and total concentration, the potency of a competitor will be inuenced by its own free fraction, a variable that is highly dependent on assay dilution (see below). Relative displacement of T4 and T3 is generally equivalent23, as predicted for two ligands that share the same specic binding sites. Potentially important interactions of drug competitors with the T4 binding sites on transthyretin can inuence free T4 values when TBG values are very low, or serum T4 is extremely high.24 Principles of competition When there is competition between several ligands for a shared specic binding site, competition is a function of their relative free concentrations in that milieu. Assessment of the potency of substances that compete for thyroid hormone binding in serum is complicated by the fact that these competitors are themselves 9099% bound to albumin in undiluted serum. These albumin sites are shared with other ligands; occupancy of those sites becomes an additional key factor in determining competition. Dilution effects For highly bound ligands such as T4, it is technically easier to perform free hormone estimates with diluted serum. In the absence of competitors, such measurements give useful comparisons between samples with high, normal and low free hormone concentrations. However, it is a problem inherent in all assays for free T4 and T3, that the free concentrations of hormone and binding competitors do not change in parallel with progressive sample dilution.23 In reality, it is extremely difcult to establish systems in which the relative concentrations of free hormone, competitor(s) and unoccupied binding sites remain in the relationship that applies in vivo. Mass action dictates that there is dissociation of a bound ligand with progressive sample dilution, so that free concentration is at rst well maintained, but then decreases signicantly as the reservoir of bound ligand becomes depleted.23,25 Because important drug competitors have a much smaller reservoir of bound ligand, their free concentration declines with progressive dilution before the free T4 concentration alters.23,25 For a hormone such as T4, with a free fraction of about 1:4000, progressive dissociation will sustain the free T4 concentration at 1:100 dilution, while a drug that has a free fraction of 1:50 will show a marked decrease in free concentration after only 1:10 dilution. Thus, the effect of a competitor to increase free T4 can be under-estimated, the error being greatest in assays with the highest sample dilution. Use of undiluted or minimally diluted serum in ultraltration or equilibrium dialysis systems minimises this artefact.26,27 It should be noted that when undiluted serum is dialysed or ltered against a volume of buffer, that volume should be included in the dilution factor. As emphasised by Ekins28, the position of the dialysis membrane, within a constant total volume, is irrelevant in determining either the nal free ligand concentration or the effects of dialysable competitors. Co-dilution and pre-dilution Data on competitor effects are confusing unless details such as dilution and albumin concentration are clearly dened. The terms pre-dilution and co-dilution are useful in considering potential artefacts.29 Pre-dilution occurs when the concentration of binding proteins is progressively decreased before test concentrations of competitor are added (Fig. 1). The lower the concentration of albumin, the higher the occupancy of available binding sites, thus leading to a disproportionate increase in the free competitor concentration; this effect magnies competitor potency.30 Thus, when potential competitors are added to pre-diluted serum, their effect is prone to over-estimation. For example, in the case of a highly albumin-bound NEFA, the T4-displacing effect of 5 mM oleic acid in undiluted serum could be matched almost exactly by 0.5 mM oleic acid in serum diluted 1:1031 (Fig. 2). The reverse artefact is seen with co-dilution, which occurs when whole serum, containing a competitor or test additive, is serially diluted so that total concentrations of binding proteins,

758

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

Fig. 1. When serum is pre-diluted, followed by addition of a particular concentration of competitor, the effect of the competitor will be over-estimated as albumin occupancy increases. If a competitor is added to serum, followed by simultaneous co-dilution of binding proteins, hormone and competitors, its effect on T4 binding will be progressively be under-estimated with dilution.

hormone and competitors diminish in parallel. If a competitor is less highly protein bound than the hormone itself, its potency will be progressively under-estimated with sample dilution. A co-dilution effect was the clue that led to the recognition of furosemide as an important inhibitor of T4 binding in serum.32 When T4 binding was studied in serial dilution in critically ill

Fig. 2. Effect of oleic acid increments on T4 free fraction with progressive serum dilution (equilibrium dialysis, 37 C, Tris chloride buffer, pH 7.4). The effect of 5 mmol/l oleic acid in undiluted serum is reproduced by 0.5 mmol/l oleic acid in serum diluted 1:10. (Reference 31).

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

759

Fig. 3. The furosemide-induced increase in free T4 fraction is progressively lost with sample dilution. (Equilibrium dialysis, 37 C, 1 ml serum compartment, 20 ml total volume). Serum dilution of 1:50 is equivalent to 1:1000 dilution of free ligand. (Reference 32).

hypothyroxinaemic patients, those who had received high-dose furosemide showed a marked increase in free T4 fraction that was obscured with sample dilution (Fig. 3). The importance of a dilution artefact was demonstrated in comparing the T4-displacing effect of furosemide in three commercial free T4 assays; the effect was most obvious in the method with least sample dilution33 (Fig. 4). Similarly, therapeutic concentrations of phenytoin and carbamazepine increased the free concentration of T4 by 4050% using ultraltration of serum that had not been diluted, while no increase in T4 free fraction was seen (i.e., the free hormone estimate was spuriously low) using a commercial single-step free T4 assay after 1:5 serum dilution.27 During continuing drug therapy, total T4 was lowered by 2550%; measured free T4 concentrations were normal in the ultraltration assay and spuriously low in the assay that used diluted serum.27

Fig. 4. Effect of increasing concentrations of serum furosemide on apparent free T4 concentration, by three commercial methods that vary in sample dilution. The T4-displacing effect of the competitor is progressively obscured with increasing sample dilution. (Redrawn from ref 33, Hawkins RC. Clin Chem 1998; 44:2550).

760

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

Estimation of competitor potency in serum under simulated in vivo conditions The potency of competitors for T4 binding under in vivo conditions can be assessed by equilibrium dialysis of undiluted serum, adding the test substances in amounts that achieve appropriate nal concentrations of competitor in the serum compartment at equilibrium.31 Addition of each test substance was based on its serum binding, measured by a non-isotopic spectrophotometric method in which transit of drug from buffer to serum compartment is proportional to serum binding.34 In undiluted serum at 37  C, in vitro T4 displacement >20% occurs with numerous medications at relevant concentrations in the serum compartment at equilibrium (Table 1). Notably, the hierarchy of potency of various competitors at their relevant therapeutic concentrations in whole serum is strikingly different from their afnity for TBG in isolation.29,31 The effect of continuing therapy in vivo will be inuenced by the kinetics of the competitor (see below). Measurement of the T4 free fraction in undiluted serum is laborious, because even minimal free iodine contamination of the iodinated T4 tracer leads to major over-estimates of the free thyroxine fraction. This artefact can be avoided by magnesium chloride precipitation of authentic tracer in each dialysate in the presence of excess unlabelled hormone as carrier.35 Competition by non-esteried fatty acids (NEFA) A diversity of NEFA, each intensely albumin-bound at multiple sites, normally circulates at concentrations up to 0.5 mmol/l, comparable to the normal serum albumin concentration of about 0.6 mmol/l. It is well known that NEFA can inhibit serum binding of T4, but the relevance of this phenomenon is complex, for three reasons. First, the competitor potency of NEFA is markedly overestimated if test concentrations are added to pre-diluted serum31 (see above, Fig. 2). Second, there is the possibility that the concentration of NEFA present in the assay tube may be greater than the in vivo concentration, due to generation of NEFA from triglyceride during sample storage and incubation, an artefact is especially important in samples taken during heparin treatment (see below). Third, the T4-displacing effect of NEFA in vivo may not occur from direct interaction with TBG, but by cascade effects as a result of the displacement of other ligands from the albumin sites that normally limit their free concentration (see below). The heparin artefact Heparin treatment in vivo can lead to an unusual in vitro artefact that gives spuriously high estimates of circulating free T4. In the presence of a normal serum albumin concentration, NEFA concentrations >3 mmol/l will increase free T4 by displacement from TBG30,36, but these concentrations are uncommon in vivo. However, in heparin-treated patients, serum NEFA may increase to these levels as a result of heparin-induced mobilisation of lipoprotein lipase from endothelium in vivo that leads to increased NEFA in vitro during sample storage or incubation36 (Fig. 5). (Samples collected with heparin are unaffected, as heparin itself has negligible direct effect on T4 binding.) As a result of this artefact, serum NEFA concentrations at the time of assay can be much higher than they were in vivo.37 As shown in Table 2, both incubation for 22 h at 37  C and 7 days storage of serum at 4  C increases serum concentrations of NEFA to levels that will inhibit T4 binding in serum taken after full anticoagulant doses of heparin; a similar effect occurs without heparin if serum triglyceride is markedly increased. This phenomenon could account for some reports of apparent increases in free T4 and free T3 fraction in critically ill subjects. Because of the incubation at 37  C that is involved in equilibrium dialysis and ultraltration methods, it is an exaggeration to regard these techniques as impeccable gold standard techniques of free T4 estimation, particularly in subjects who have received heparin. Where heparin has been given, assays of total T4 and T3 are likely to be more informative than free T4 and T3 estimates, unless the serum is treated with additives to avoid in vitro generation of NEFA.37 The heparin-induced artefact is accentuated if serum triglyceride concentrations are increased, if serum albumin concentration is low, and particularly by prolonged incubation at 37  C (Table 2). Under

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

761

Fig. 5. Effect of heparin in vitro to increase the apparent concentration of serum free T4. Heparin acts in vivo (left) to liberate lipoprotein lipase from vascular endothelium. Lipase acts in vitro (right) to increase the concentration of non-esteried fatty acids (NEFA).

these conditions, doses of heparin as low as 10 units may be responsible for in vitro increases in serumfree T4.38 Low-molecular-weight heparin preparations appear to have a similar effect.39 Interaction between competitors Increasing concentrations of any substance that shares albumin-binding sites with a competitor could increase the free concentration of that competitor. Both oleic acid40 and 3-carboxy-4-methyl-5propyl-2-furanpropanoic acid (CMPF)41, a naturally occurring furanoid acid that accumulates in renal failure, have the potential to exert indirect or cascade effects on T4 binding in serum. At concentrations that caused only minimal direct inhibition of T4 binding in undiluted normal serum, oleic acid and CMPF each augmented the T4-displacing effect of several drug competitors for T4 binding in undiluted serum40,41 (Fig. 6). By such a mechanism, free hormone concentrations can be inuenced by substances that have little direct interaction with hormone-binding sites. Kinetics of the competitor The kinetics of the competitor itself will determine how it inuences hormone binding in vivo. Competitors of short half-life such as furosemide or salicylate will show uctuating effects on hormone binding so that free hormone estimates may vary depending on the time between dosage and sampling.42,43 By contrast, competitors of long half-life, such as phenytoin or carbamazepine27, will
Table 2 Effect of heparin, time and temperature on in vitro generation of non-esteried fatty acids (NEFA). Sample storage Normal subjects Serum NEFA mmol/l Heparin 10 IU Thawed from 80 C 4 C, 7 days 22 C, 4 h 22 C, 20 h 37 C, 4 h 37 C, 22 h 0.24 0.03 0.41 0.04* 0.17 0.04 0.23 0.05 0.19 0.05 0.35 0.05* 0.30 0.14 0.36 0.15 Heparin 25,000 U/day 0.77 0.11 2.90 1.10* 1.30 0.67* 3.29 1.33* 2.88 1.17* 3.49 1.18* 0.53 0.23 1.10 0.33* 1.31 0.48* 1.03 0.16* 2.23 1.10* Diabetic (Serum Tg >5 mmol/l)

0.48 0.21*

Serum was frozen at 80 C after sampling; mean SD; n > 4; *p < 0.05.

762

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

Fig. 6. Addition of 0.3 mmol/l CMPF, which alone produces minimal inhibition of T4 binding, augments T4 displacement by four medications that interact with T4 binding to TBG (undiluted serum, equilibrium dialysis, 37 C). Drugs were added to produce relevant therapeutic concentrations in the serum compartment at equilibrium: furosemide 0.91, fenclofenac 4.7, diunisal 20, aspirin 300 umol/l. (Reference 41).

result in a new steady state with a lowered total hormone concentration with an increased free fraction, resulting in a normal free hormone estimate.27 Potential binding competitors in critical illness It has been suggested that free T4 and T3 results could be inuenced in critical illness by circulating competitors for T4 and T3 binding.44 Concentrations of NEFA above 2 mmol/l can inhibit T4 binding, and samples taken during heparin treatment can achieve these levels during storage and incubation37 (see above). Pre-dilution effects may have inuenced the earlier studies that suggested a possible direct effect of oleic acid on T4 binding, tested by addition of oleic acid to diluted serum.45 Conversely, a serum dilution factor of 1:64 (dialysate and diaysand) could have attenuated any potential effect in a study that showed no evidence of a T4-binding inhibitor in sera from a large group of critically ill patients46, that is, there may have been a substantial co-dilution effect. So far, there is a lack of studies with undiluted sera from critically ill, heparin-free subjects whose drug therapy is fully documented. Sample storage and incubation times would need to be minimised to avoid in vitro generation of NEFA.37 A further possibility that might inuence tissue delivery of thyroid hormone is the potential effect of TBG degradation at the site of sepsis or inammation.47,48 Between-method differences It is cause for concern that estimates of free T4 show large between-method differences in three situations where thyroid function is difcult to assess on clinical grounds alone: critical illness, renal insufciency and late pregnancy. By contrast, total T4 measurements in these situations give predominantly normal results (see below). Critical illness Marked method-dependent differences in apparent free T4 concentration have been demonstrated in some studies of critical illness. Sapin et al.49 used six commercial free T4 kits to study 20 previously euthyroid subjects in a clearly dened example of critical illness, 7 days after bone marrow transplantation, during multiple drug therapy, including heparin and glucocorticoids (Fig. 7). Methods that involved sample incubation at 37  C gave high free T4 values in 2040% of their study subjects, while

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

763

Fig. 7. Free T4 estimated by six different methods in 20 previously euthyroid bone marrow recipients on the seventh day after transplantation. Free T4 estimates are increased or decreased, depending on the type of assay. Serum total T4 remained normal in 19 of the 20 subjects, while serum TSH was subnormal in 11, independent of the method used. Therapy included heparin and glucocorticoids at the time of sampling. Mid-reference ranges for each method have been normalized to 100%. (Re-drawn from ref 49, Sapin R et al, Clin Chem 2000; 46:41820).

analogue tracer methods that are inuenced by tracer binding to albumin, gave subnormal estimates of free T4 in 2030%. By contrast, total T4 was normal in 19 of these 20 presumably euthyroid subjects. Serum TSH was <0.1 mU/l in half the subjects, independent of method, a change attributable to glucocorticoid treatment. Thus, free T4 methods show an unacceptable rate of false-positive results in this situation. On the basis of artefacts in free T4 methodology in the face of subnormal serum TSH, ndings could be falsely interpreted to indicate either thyrotoxicosis or secondary hypothyroidism. Renal failure The frequency of thyroid dysfunction is increased in renal failure50, but assay artefacts can compromise laboratory assessment. Iitaka et al.21 showed marked method-dependent discrepancies in apparent free T4 concentrations in patients with renal failure (Fig. 8). With an albumin-dependent analogue-tracer method, the free T4 estimate was low, as would occur if more tracers were available in a sample than in the standards, an effect accentuated by low serum albumin or occupancy of albumin by retained metabolites. Concentrations of three organic acids, indoxyl sulphate (IS), indole acetic acid (IAA) and hippuric acid (HA) were diminished by haemodialysis, but remained >10-fold above normal. When organic acids that are retained in renal failure were added to normal serum, the apparent free T4 concentration diminished using an analogue tracer method, increased in an equilibrium dialysis method and was unaltered with a labelled antibody method.21 All three methods showed an apparent rise in free T4 after heparin treatment for haemodialysis, consistent with an effect of NEFA generated during sample incubation.30,36 The effects on apparent free T3 were similar, or more marked.21 Pregnancy The demonstration that maternal hypothyroxinaemia in the rst trimester can have an adverse effect on later psychomotor development has focussed attention on precise assessment of thyroid function in pregnancy; current uncertainty in the interpretation of free T4 estimates in pregnancy is

764

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

Fig. 8. Comparison of three types of serum free T4 estimate, pre and post maintenance haemodialysis, in 11 uremic subjects. Horizontal bars indicate the median, the boxes the range from 25 to 75%. Estimates of free T4 by an analog tracer method were lower than by equilibrium dialysis. Apparent free T4 by all three methods increased after heparinization for hemodialysis. Concentrations of three organic acids, indoxyl sulphate (IS), indole acetic acid (IAA) and hippuric acid (HA) were diminished by hemodialysis, but remained >10-fold above normal. The effect on apparent free T4 of adding these organic acids to normal serum, varied widely between methods (see text). (Re-drawn from ref 21, Iitaka et al. Clin Endocrinol 1998; 48: 73946).

unhelpful, especially as serum TSH shows trimester-dependent variations. While it is clear that normal pregnancy is associated with a marked increase of about 40% in total T4 concentration, there is no consensus as to whether free T4 is maintained at normal levels51, or whether the mean free hormone concentration decreases slightly, with values still predominantly within the normal range.52 Interpretation is complex because, particularly in late pregnancy, there are marked discrepancies between various methods of estimating free T4, with strong negative bias in some methods.53,54 In a comparative study of free T4 by seven commercial methods in 23 euthyroid women at term, Roti et al. found that albumin-dependent methods show marked negative bias, with up to 50% subnormal values, while other methods gave values above their non-pregnant reference interval.53 A recent study that conrmed marked negative bias in free T4 estimates during pregnancy has questioned the wisdom of continuing to rely on free thyroxine estimates during pregnancy.54 In contrast to two-kit free T4 methods, total serum T4 and its derivative, the free thyroxine index, showed the anticipated inverse relationship with serum TSH, with historically consistent results in numerous reports.54 Thus, because of consistency between methods, total T4 measurement may be superior to free T4 estimation as a guide to therapy during pregnancy, provided that the reference values take account of the normal oestrogen-induced increase in TBG. If free T4 estimates continue to be used in pregnancy, clinicians will need to interpret results in relation to reference intervals that are both trimester and method specic. The reasons for wide method-dependent discrepancies in free T4 values in pregnancy remain unclear. Albumin is lower and serum triglyceride and NEFA concentrations increase in late pregnancy55, phenomena that can inuence estimates of free T4. The possibility of a pregnancy-related circulating inhibitor of thyroid hormone binding that is undetected in diluted samples may merit further study.

Implications for free and total hormone assays It is uncertain whether the limitations of current free T4 methodology, as outlined above, are likely to be overcome by further technical innovations. Whether free T4 is estimated by two-step methods that separate a fraction of the free T4 pool from the binding proteins before the T4 assay is performed,

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

765

or by one-step methods designed to give a signal that is proportional to the free T4 concentration in the presence of binding proteins, the frequency of non-specic abnormalities has become a major disadvantage of these techniques. One-step methods generally become invalid when the sample and standard differ in their binding of assay tracer20,25, but two-step methods are generally free of this artefact. However, both types of method are vulnerable to artefacts that result from generation of NEFA during sample storage and incubation and both fail to reect the in vivo effect of binding inhibitors, which is obscured by sample dilution. Especially in samples from heparin-treated patients, generation of NEFA during sample storage and incubation can be sufcient to displace T4 directly or to augment the effect of other competitors by cascade effects. The heparinNEFA effect may remain an insuperable pre-analytical problem, unless a simple, non-toxic additive can be developed that will block heparininduced lipase from the moment of sample collection. The problems documented here have not yet been addressed in a recently developed free T4 and free T3 methods56, regarded as a major advance in the routine assessment of free thyroid hormones.57 The method uses on-line solid phase extraction liquid chromatography/tandem mass spectrometry to measure the free T4 and T3 in protein-free dialysate after 20 h dialysis of serum against an equal volume of buffer at pH 7.2 (effective dilution 1:1), using a 96-well dialysis plate. Assay standards are prepared in protein-free buffer. Detection limit, internal quality controls and assay precision methods appeared to be satisfactory, but this technique gave consistently higher values than established methods. On that basis, method-specic reference intervals would need to be developed. Assessment of potential complexities related to heparin-induced lipolysis, pregnancy or critical illness has not been reported. There has been a general trend to regard equilibrium dialysis methods, however cumbersome, as the gold standard for free T4 estimation. The artefacts summarised here suggest that this view is no longer tenable and that total T4 measurement is the more robust parameter. Notably, denitive studies of the pituitarythyroid axis in critical illness58,59 have relied on total rather than free T4 as the preferred index of thyroid hormone secretion. Whatever techniques are developed to achieve precise laboratory assessment, diagnostic accuracy cannot be assured without collaboration across the laboratoryclinical interface. Numerous diagnostic errors can occur if laboratory results are classied without regard for the clinical context.60

Clinical points  Free T4 estimates are unreliable in several common diagnostically difcult situations.  Because free T4 reference intervals vary widely between methods, interpretation requires method-specic ranges.  Total serum T4 shows much less variation between methods. When corrected for variations in TBG and interpreted in relation to TSH, total T4 is the more reliable parameter in situations of diagnostic difculty.  The medication history is crucial whenever tests of thyroid function are anomalous, or when the sick euthyroid syndrome is considered.

Research agenda  Why do many free T4 estimates show strong negative bias in late pregnancy? Is this phenomenon due to a circulating binding inhibitor that is obscured by sample dilution?  Can an effective non-toxic additive be developed that will block the in vitro action of heparininduced lipase that causes spurious NEFA-induced elevation of free T4?  Can technology for rapid ultra-ltration of undiluted serum be developed for routine use?

766

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

References
1. Refetoff S & Dimitruescu AM. Syndromes of reduced sensitivity to thyroid hormone; genetic defects in hormone receptors, cell transporters and deiodination. Best Practice & Research. Clinical Endocrinology & Metabolism 2007; 21: 277305. 2. Oliviera JHA, Persani L, Beck-Peccoz P et al. Investigating the paradox of hypothyroidism and increased TSH levels in Sheehans syndrome: characterization of TSH carbohydrate content and bioactivity. The Journal of Clinical Endocrinology and Metabolism 2001; 86: 16941699. 3. Beck-Peccoz P & Persani L. Variable biological activity of thyroid stimulating hormone. European Journal of Endocrinology 1994; 131: 331340. 4. Stockigt JR. Thyroid function tests and the effect of drugs. In Wass JAH & Shalet SM (eds.). Oxford textbook of endocrinology and diabetes. OUP, 2002, pp. 306316. 5. Mebis L, Debaveye Y, Visser TJ et al. Changes within the thyroid axis during the course of critical illness. Endocrinology and Metabolism Clinics of North America 2006; 35: 807821. 6. Surks MI & Sievert R. Drugs and thyroid function. The New England Journal of Medicine 1995; 333: 16881694. 7. National Academy of Clinical Biochemistry. Laboratory support for the diagnosis of thyroid disease, http://www.aacc.org/ members/nacb/archive/lmpg/thyroiddisease/pages/default.aspx. 8. Lando VS, Batista MC, Nakamura IT et al. Effects of long-term storage of lter paper blood samples on neonatal thyroid stimulating hormone, thyroxin and 17-alpha-hydroxyprogesterone measurements. Journal of Medical Screening 2008; 15: 109111. 9. Hebles-Duvison M, Cruz-Ruiz M, Vazquez-Rubio R et al. Evaluation of AutoDela and access automated immunoassay systems for third generation assays for thyrotropin. European Journal of Clinical Chemistry and Clinical Biochemistry 1995; 33: 881885. 10. Despres N & Grant AM. Antibody interference in thyroid assays: a potential for clinical misinformation. Clinical Chemistry 1998; 44: 440454. 11. Kricka LJ. Interferences in immunoassays still a threat. Clinical Chemistry 2000; 46: 1003710038. 12. Ward G, Mckinnon L, Badrick T et al. Heterophilic antibody remains a problem for the immunoassay laboratory. American Journal of Clinical Pathology 1997; 108: 417421. 13. Hunter WM, Bennie JG, Budd PS et al. Immunometric assays using monoclonal antibodies. In Hunter WM & Corrie JET (eds.). Immunoassays for clinical chemistry. Edinburgh: Churchill-Livingston, 1983, pp. pp531pp544. 14. Scantibodies laboratories, http://www.scantibodies.com/block.html. 15. Bjerner J, Nustad K, Norum LF et al. Immunometric assay interference: incidence and prevention. Clinical Chemistry 2002; 48: 613621. 16. Ross HA, Menheere PPCA, Thomas CMG et al. Interference from heterophilic antibodies in seven current TSH assays. Annals of Clinical Biochemistry 2008; 45: 616. 17. Flad TM, Duffy J & McKenna TJ. Multiple spuriously abnormal thyroid function indices due to heterophilic antibodies. Clinical Endocrinology 1994; 41: 391395. 18. Sakata S, Nakamura S & Miura K. Autoantibodies against thyroid hormones or iodothyronine. Annals of Internal Medicine 1985; 103: 579589. 19. Beck-Peccoz P, Romelli PB, Cattaneo MG et al. Evaluation of free T4 methods in the presence of iodothyronine autoantibodies. The Journal of Clinical Endocrinology and Metabolism 1984; 58: 736739. 20. Stockigt JR, Stevens V, White EL et al. Unbound analog radioimmunoassays for free thyroxin measure the albuminbound hormone fraction. Clinical Chemistry 1983; 29: 14081410. *21. Iitaka M, Kawasaki S, Sakurai S et al. Serum substances that interfere with thyroid hormone assays in patients with chronic renal failure. Clinical Endocrinology 1998; 48: 739746. 22. Stockigt JR, Lim C-F, Barlow JW et al. Interaction of furosemide with serum thyroxine-binding sites: in vivo and in vitro studies and comparison with other inhibitors. The Journal of Clinical Endocrinology and Metabolism 1985; 60: 10251031. 23. Stockigt JR, Lim C-F, Barlow JW et al. Thyroid hormone transport. In Weetman AP & Grossman A (eds.). Pharmacotherapeutics of the thyroid gland. Heidelberg: Springer Verlag, 1997, pp. 119150. 24. Munro SL, Lim C-F, Hall JG et al. Drug competition for thyroxine binding to transthyretin (prealbumin): comparison with effects on thyroxine-binding globulin. The Journal of Clinical Endocrinology and Metabolism 1989; 68: 141147. 25. Ekins R. The free hormone hypothesis and measurement of free hormones. Clinical Chemistry 1992; 38: 12891293. 26. Surks MI, Hupart KH, Chao P et al. Normal free thyroxine in critical nonthyroidal illnesses measured by ultraltration of undiluted serum and equilibrium dialysis. The Journal of Clinical Endocrinology and Metabolism 1988; 67: 10311039. *27. Surks MI & Defesi CR. Normal free thyroxine concentrations in patients treated with phenytoin or carbamazepine: a paradox resolved. The Journal of the American Medical Association 1996; 275: 14951498. *28. Ekins R. Measurement of free hormones in blood. Endocrine Reviews 1990; 11: 546. 29. Stockigt JR. Thyroid hormone binding and variants of transport proteins. In DeGroot LJ & Jameson L (eds.). Endocrinology. 5th edn. Philadelphia: WB Saunders, 2006, pp. 22152226. 30. Mendel CM, Frost PH & Cavalieri RR. Effect of free fatty acids on the concentration of free thyroxine in human serum: the role of albumin. The Journal of Clinical Endocrinology and Metabolism 1986; 63: 13941399. *31. Lim C-F, Bai Y, Topliss DJ et al. Drug and fatty acid effects on serum thyroid hormone binding. The Journal of Clinical Endocrinology and Metabolism 1988; 67: 682688. 32. Stockigt JR, Lim CF, Barlow JW et al. High concentrations of furosemide inhibit serum binding of thyroxine. The Journal of Clinical Endocrinology and Metabolism 1984; 59: 6266. 33. Hawkins RC. Furosemide interference in newer free thyroxine assays. Clinical Chemistry 1998; 44: 25502551. 34. Lim C-F, Wynne KN, Barned JM et al. Non-isotopic spectrophotometric determination of the unbound fraction of drugs in serum. The Journal of Pharmacy and Pharmacology 1986; 38: 795800. 35. Sterling K & Brenner MA. Free thyroxine in human serum: simplied measurement with aid of magnesium precipitation. The Journal of Clinical Investigation 1966; 45: 153163.

J.R. Stockigt, C.-F. Lim / Best Practice & Research Clinical Endocrinology & Metabolism 23 (2009) 753767

767

*36. Mendel CM, Frost PH, Kunitake ST et al. Mechanism of the heparin-induced increase in the concentration of free thyroxine in plasma. The Journal of Clinical Endocrinology and Metabolism 1987; 65: 12591264. 37. Zambon A, Hashimoto SI & Brunzell JD. Analysis of techniques to obtain plasma for measurement of levels of free fatty acids. Journal of Lipid Research 1993; 34: 10211028. 38. Jaume JC, Mendel CM, Frost PH et al. Extremely low doses of heparin release lipase activity into the plasma and can thereby cause artefactual elevations in the serum free thyroxine concentrations as measured by equilibrium dialysis. Thyroid 1996; 6: 7983. 39. Stevenson HP, Archbold GPR, Johnston P et al. Misleading serum free thyroxine results during low molecular weight heparin treatment. Clinical Chemistry 1998; 44: 10021007. *40. Lim C-F, Curtis AJ, Barlow JW et al. Interactions between oleic acid and drug competitors inuence specic binding of thyroxine in serum. The Journal of Clinical Endocrinology and Metabolism 1991; 73: 11061110. *41. Lim C-F, Stockigt JR, Curtis AJ et al. Inuence of a naturally-occurring furanoid acid on the potency of drug competitors for specic thyroxine binding in serum. Metabolism 1993; 42: 14681474. 42. Newnham HH, Hamblin PS, Long F et al. Effect of oral furosemide on diagnostic indices of thyroid function. Clinical Endocrinology 1987; 26: 423431. 43. Wang R, Nelson JC & Wilcox RB. Salsalate administration as a potential pharmacological model of the sick euthyroid syndrome. The Journal of Clinical Endocrinology and Metabolism 1998; 83: 30953099. 44. Chopra IJ, Huang TS, Beredo A et al. Serum thyroid hormone binding inhibitor in non-thyroidal illness. Metabolism 1986; 35: 152159. 45. Chopra IJ, Huang TS, Solomon DH et al. The role of thyroxine-binding serum proteins in oleic acid-induced increase in free T4 in nonthyroidal illness. The Journal of Clinical Endocrinology and Metabolism 1986; 63: 776779. 46. Mendel CM, Laughton CW, McMahon FA et al. Inability to detect an inhibitor to thyroxine-serum protein binding in sera from patients with nonthyroidal illness. Metabolism 1991; 40: 491502. 47. Jirasakuldech B, Schussler GC, Yap MG et al. A characteristic serpin cleavage product of thyroxine-binding globulin appears in sepsis sera. The Journal of Clinical Endocrinology and Metabolism 2000; 85: 39963999. 48. Janssen OE, Golcher HM, Grasberger H et al. Characterization of T4-binding globulin cleaved by human elastase. The Journal of Clinical Endocrinology and Metabolism 2002; 87: 12171222. *49. Sapin R, Schlienger J-L, Gasser F et al. Intermethod discordant free thyroxine measurements in bone marrow-transplanted patients. Clinical Chemistry 2000; 46: 418422. 50. Kaptein EM. Thyroid hormone metabolism and thyroid diseases in chronic renal failure. Endocrine Reviews 1996; 17: 45 63. 51. Glinoer D. Thyroid disease manager. Chapter 14. Thyroid regulation and dysfunction in the pregnant patient, www: thyroidmanager.org. 52. Ball R, Freeman DB, Holmes JC et al. Low-normal concentrations of free thyroxine in late pregnancy; physiological fact, not technical artefact. Clinical Chemistry 1989; 35: 18911896. *53. Roti E, Gardini E, Minelli R et al. Thyroid function evaluation by different commercially available free thyroid hormone measurement kits in term pregnant women and their newborns. Journal of Endocrinological Investigation 1991; 14: 19. *54. Lee RH, Spencer CA, Mestman JH et al. Free T4 assays are awed during pregnancy. American Journal of Obstetrics and Gynecology 2009; 260. 260e1260e6. 55. Sivan E, Homko CJ, Whittaker PG et al. Free fatty acids and insulin resistance during pregnancy. The Journal of Clinical Endocrinology and Metabolism 1998; 83: 23382342. 56. Yue B, Rockwood AL, Sandrock T et al. Free thyroid hormones in serum by direct equilibrium dialysis and on-line solidphase extraction-liquid chromatography/tandem mass spectrometry. Clinical Chemistry 2008; 54: 642651. 57. Thienpont LM. A major step forward in the routine measurement of serum free thyroid hormones. Clinical Chemistry 2008; 54: 625626. 58. Van den Berghe G, de Zegher F, Baxter RC et al. Neuroendocrinology of prolonged critical illness: effects of exogenous thyrotropin-releasing hormone and its combination with growth hormone secretagogues. The Journal of Clinical Endocrinology and Metabolism 1998; 83: 309319. 59. Van den Berghe G, Wouters P, Weekers F et al. Reactivation of pituitary hormone release and metabolic improvement by infusion of growth hormone releasing peptides and thyrotropin releasing hormone in patients with protracted critical illness. The Journal of Clinical Endocrinology and Metabolism 1999; 84: 13111323. 60. Stockigt JR. Thyroid disease manager. Chapter 6b. Clinical strategies in the testing of thyroid function. Section 6. The physicianlaboratory interface, www:thyroidmanager.org.