Clinical Chemistry 47:8

13531363 (2001)

Review

Direct and Indirect Free Thyroxine Assay

Methods: Theory and Practice
John E.M. Midgley
Background: For the diagnosis of thyroid disease, measurement of free hormone is generally accepted as an
appropriate measure. However, valid assays measuring
the free fraction of thyroxine (FT4) ideally must perform
without bias, despite large variations in the concentrations and affinities of serum T4-binding proteins in the
population. Several approaches have been taken to
overcome such bias, and these have created considerable
controversy in the field over the past decade.
Approach:This review, from both a historical and an
analytical standpoint, charts the progress made over
more than 30 years in improvements to the performance
of assays in common use for the measurement of FT4 in
serum or plasma. It reexamines the theory behind early
approaches to such assays [for example, the free thyroxine index (FTI) method], that preceded more accurate,
two-step immunoassays or one-step analog techniques.
It evaluates the continuous refinements to the latter
assays that by now have largely supplanted the FTI
approach and where the deficiencies that so exercised
clinical chemists in the past have been virtually eliminated in the leading assays.
Content:The basic Mass Action theory underpinning all
such methods is discussed by assessing how far each
particular approach obeys the criteria the theory imposes. In this, it is not the intention of the review to
dissect individual commercial or academic assays, but
rather to give guidance where appropriate as to how any
assay said to measure FT4 can be conveniently evaluated
by those intending to use it. Examples are given where
inappropriate tests may wrongly imply assay invalidity
by misinterpreting how FT4 assays work.
Summary: Detailed knowledge of the underlying theory is essential when devising tests for direct FT4 assays,
to ensure that such tests do not overstep the practical
limits of assay validity.

The Theory Underlying Free Thyroxine Assays

general background
For the diagnosis of thyroid disease, the free hormone
concept (1 ) is generally accepted [with exceptions (2 4 )]
as an appropriate measure. This implies that the supply of
thyroxine (T4)1 or triiodothyronine (T3) into cells is governed by their unbound (free) concentrations in serum
rather than the protein-bound fractions. Valid assays
measuring the free fraction of T4 (FT4) ideally must
perform without bias, despite large variations (both absolute and relative) in the concentrations (and affinities) of
serum T4-binding proteins found in the population.
No assays measure the actual unbound molecules of T4
in serum. Regardless of the probe used, be it a dialysis
membrane, a nonspecific binder, or an antibody, unbound
T4 is abstracted or dissociated in excess of the original
unbound hormone in the sample. This does not matter as
long as the amount of hormone sampled is so small a
fraction of the total (bound and unbound) that the equilibrium between bound and free forms in serum is minimally disturbed.
For a euthyroid subject with serum binding protein
concentrations within the appropriate reference intervals,
0.01 0.02% of the total serum T4 is present as FT4. The
Mass Action relationship at equilibrium between the two
fractions of bound and free hormone for a single protein
is:
K equil

Concentration of T4 on occupied sites of binding protein

[FT4] Concentration of unoccupied sites

where Kequil is the equilibrium binding constant of the

protein for T4.
This equation transforms to:
[FT4]

Concentration of T4 on occupied sites

K equil Concentration of unoccupied sites

2001 American Association for Clinical Chemistry

North Lakes Clinical, 6 High Wheatley, Ilkley, West Yorkshire LS29 8RX,
United Kingdom. E-mail je.midgley@virgin.net.co.uk.
Received November 8, 2000; accepted May 15, 2001.

1
Nonstandard abbreviations: T4, thyroxine; T3, triiodothyronine; FT4, free
T4; TBG, thyroxine-binding globulin; FTI, free thyroxine index; NEFA, nonesterified fatty acid; NTI, nonthyroidal illness; BP, binding potential; and TSH,
thyrotropin.

1353

1354

Midgley: Analysis of Free Thyroid Hormone Assays

Consequently, the measured FT4 concentration will remain essentially unaltered, provided that the ratio of
occupied/unoccupied sites holds virtually constant.
Hence, measurements of FT4 are valid only with minimum dissociation of bound T4 from the serum proteins.
From the known equilibrium constants for the three
serum binding proteins, thyroxine-binding globulin
(TBG), transthyretin, and albumin (5, 6 ), a maximum of
5% of the protein-bound T4 in serum can be dissociated
into the free form, but preferably much less than this (7 ).
Any valid assay must encompass the whole physiologic range of serum binding protein concentrations. This
can be defined as the ideal window of validity for any
FT4 assay. Within this window, no serum presented to the
assay is affected (with regard to disturbance of the boundfree equilibrium for T4) so as to be significantly influenced
by either the T4 sampling process or by procedures that
might otherwise affect the equilibrium. Minimization of
disturbance is especially difficult for sera with very low
binding capacities for T4 (e.g., with low concentrations of
TBG and/or albumin). When operating outside their
windows, assays no longer measure FT4. Various methods adhere to the above requirement to different extents,
and even within any given common technology, individual assays may differ detectably in their compliance.
Hence, uniquely in immunoassay technology, the mode of
operation and window of validity of any given FT4 assay
(or its equivalent) are individually defined, so that subtle
differences arise in the assays performance when compared with any other. Several factors cause this. These do
not merely concern arbitrary judgments by commercial
manufacturers, or perhaps lack of awareness, as to how
extensively the wide variations in serum in the human
population should be examined during development of
their assays. They also include differences in concentration, specificity, and affinity for T4 of any antibodies used;
the extent of serum dilution by assay ingredients; the
choice of assay reagents such as buffers to mimic physiologic conditions; the choice and concentration of analog
or competitive binding reagent; and the time and temperature of the assay incubation process (714 ).
Given complexity of this magnitude and the failure of
many producers to fully grasp its implications, it is
unsurprising that, even after 20 years of development,
individual FT4 methods still vary considerably in performance (i.e., their windows of validity fall short of the
ideal). However, regardless of these differences, routine
laboratories can perform useful investigations to test the
validity of each assay method. This review, in choosing
the most commonly used methods for the measurement
of FT4 in serum, aims to reevaluate their validity and to
suggest ways in which the limits of this validity can be
simply tested. Individual named assays will not be examined in detail. There are too many to be evaluated in a
limited space, and the continual appearance and replacement of commercial assays would render any review
rapidly obsolete.

The following analysis of methods for measuring FT4

includes an examination of the T3-uptake test [and thus
the indirect free thyroxine index (FTI) approach], and
explains why shortcomings in performance can occur
(9, 1518 ). Direct one- and two-step immunoassays are
then considered, comparing these in concept and performance with assays ranging from the FTI to the so-called
gold-standard methods of equilibrium dialysis and
ultrafiltration.
The direct one-step assays have recently been criticized
as both failing to correctly measure FT4 concentrations in
protein-free and other artificial solutions and to obey
classic tests for recovery of analyte (19 24 ). Many of these
experiments have, however, been conducted outside the
window of validity of the assays. For example, if the
antibody in valid FT4 assays can be allowed to bind
0.011% of the total hormone in a serum sample, recovery
estimations, in the classic sense, have no meaning. It is not
surprising, therefore, that studies of artificial solutions
containing serum T4-binding proteins and T4 showed
apparent correlations of FT4 with binding protein concentrations (19 24 ). These tests were also performed under
conditions in which the equilibrium between bound and
free hormone was sufficiently perturbed to disobey the
essential criterion for a valid assay. Such experiments, in
which (unlike the case with equilibrium dialysis or ultrafiltration) the serum binding proteins, their T4 load, and
the antibody probe interact in solution, teach the need for
caution in devising suitable tests for assay validity.

important factors affecting ft4 measurements

As emphasized earlier, the primary rule for valid measurement of FT4 in any assay is that, by the assay process,
serum protein-bound T4 is minimally displaced into the
free phase, with insignificant disturbance of the endogenous boundfree equilibrium for T4 in the sample (7, 9 ).
The permissible amount of hormone so displaced can be
many times that of the concentration of endogenous FT4
(7, 9 ). Disturbances of the equilibrium can take several
forms. T4 can be sequestrated, through the free fraction,
onto the binding sites of a probing antibody. Simultaneously, the bound hormone is further partially dissociated simply by dilution of the serum when assay ingredients for FT4 measurement are added. In gold-standard
methods, such as equilibrium dialysis, dilution is the key
function to control, because the serum is first dialyzed
before the separated dialysate is assayed for T4 by a
sensitive total-hormone assay (2527 ). Conversely, in direct one- or two-step immunoassays for FT4, both dilution
of serum and simultaneous sequestration of T4 by the
antibody probe are equally relevant. These influences are,
in the latter case, approximately additive. If the antibody
sequestrates 1% of the total T4 from a sample containing
0.02% free in the original serum, this is equivalent to a
dilution of 50-fold. If the addition of assay ingredients to
the sample dilutes it by 10-fold, then the overall effect is
approximately equivalent to a 60-fold dilution. If, how-

Clinical Chemistry 47, No. 8, 2001

ever, the original serum sample was first diluted 2-fold

(e.g., to conduct a dilution test), then the fixed amount of
antibody will remove 2% of the total hormone from the
serum proteins and this, combined with the 10-fold dilution of the assay process itself, will approximate to a
110-fold dilution. Equivalent effects occur in a serum
naturally containing one-half the normal binding capacity
for T4 (i.e., sera with lower endogenous binding protein
concentrations should act in a given assay as if a serum
with concentrations within the reference intervals had
been diluted appropriately).
In some FT4 immunoassays, one of the assay ingredients may be albumin, which was initially added to
modulate the residual binding of analog to the serum
T4-binding proteins (7 ) but was subsequently used to
mitigate the effects of substances such as nonesterified
fatty acids (NEFAs) on T4 binding to the serum binding
proteins. However, exogenous albumin may also distort
(lower) true serum FT4 values by its own sequestration of
T4 from the serum equilibrium system. (A detailed discussion will be presented later in the text.)
For the earlier T3-uptake test used in determining FTI,
two simultaneous influences are at work. One is the
potential of the competing binder (often nonspecific),
used to bind the excess T3 not bound to the serum binding
proteins, to also bind T4 (18 ). The other is the potential for
the high concentration of the T3 label, necessary to saturate the unoccupied binding sites, to dissociate some of
the bound T4 from its bound sites on the serum proteins
by direct competition. From Mass Action equations, relatively little of the bound T4 need be dissociated before
there is unacceptable distortion of the FTI result away
from a valid FT4 measurement.
Finally, factors either endogenous in the sample in vivo
or developed in vitro during assay incubation may interfere with T4 binding in serum, again displacing enough
bound T4 to seriously affect (increase) the FT4 or FTI
measurement. This is especially noticeable when certain
drugs or other substances are present that compete with
T4 for some of its serum protein binding sites, or where
similarly competing NEFAs are produced in vitro in
heparinized patients (28 33 ).
All of these factors must be evaluated and their effects
recognized in the development of valid assays for FT4.
Failure to do so, especially with regard to antibody
sequestration of T4 and the extent of intrinsic serum
dilution by an assay, can compromise the validity of the
measurements, so that FT4 values may correlate with one
or more of the serum binding protein concentrations. The
same phenomenon will occur with the T3-uptake test if
the amount of label and the effect of the competing binder
unduly disturb the natural boundfree equilibrium of the
serum proteins.

Early Approaches to FT4 Measurement

the fti
The FTI approach (34, 35 ) provided the first convenient
estimation of FT4 in serum in the routine laboratory. This

1355

method combines two measurements: total T4 by immunoassay and T3 uptake. From the basic Mass Action
equation described earlier, total T4 closely approximates
the serum protein-bound T4 (occupied sites), and T3
uptake represents the concentration of unoccupied sites.
FTI, as an index of FT4, is therefore expressed as the
product of the two tests/100. The values obtained do not,
of course, represent actual FT4 concentrations (34 ).
The total T4 measurement of the FTI is perfectly valid.
However, the commonly held interpretation of the way in
which the T3-uptake test works (34 ) is probably incorrect.
This holds that the T3 added merely occupies the vacant
binding sites on the serum T4-binding proteins, the excess
spilling over for collection by a competing binding agent.
Scrutiny of the Law of Mass Action casts doubt on this
simple interpretation.
Serum contains two classes of T4-binding proteins:
TBG, as a low-concentration, high-affinity protein; and
transthyretin and albumin as high-concentration, lowaffinity binders. For a normal euthyroid serum, the TBG
concentration is 2000-fold lower than the combined
concentrations of transthyretin and albumin, but TBG
binds 60% of the total T4 present (5, 6 ). Whereas most of
the TBG binding sites are occupied, 1% of the transthyretin and albumin sites bind T4. When excess T3 is added
in the T3-uptake test, enough must be added to both
occupy high- and low-affinity binding proteins and provide sufficient excess for sequestration by the added
binding probe. This creates a dilemma. If enough T3 must
be added to fill the large excess of high-concentration,
low-affinity binding sites (virtually all of which are unoccupied), then this concentration will be high enough to
disturb T4 binding to TBG. The affinity of TBG for T3 is
only 10% that for T4 (36 38 ), but the quantity of T3
necessary for the assay will outweigh this smaller affinity,
allowing T3 to significantly displace T4 from the TBG
binding sites. The effect of the fixed amount of added T3
on the T4 displacement phenomenon will be proportionately smaller for high-TBG sera and proportionately
greater for low-TBG sera. The T3-uptake test (through the
FTI) thus inevitably displays a TBG dependency (14
16, 18 ). Thus, because the added labeled T3 is preferentially bound to the large excess of unoccupied transthyretin and albumin sites, the T3-uptake test should also
produce falsely low FTI results in nonthyroidal illness
(NTI) where the concentrations of such proteins are often
reduced (39 42 ).
The nonspecific binders used in many T3-uptake assays
(18 ) can also create additional problems by directly sequestrating some T4 as well as T3. With such a variety of
effects (14, 16 18 ), it is not surprising that, for example,
the imperfect normalization of, e.g., high-TBG late-pregnancy sera into the mid- or high-normal range for FTI
should have been accepted historically as the correct
placement, when more accurate measurements lacking
TBG distortions place such sera lower in the reference
interval (43 45 ).

1356

Midgley: Analysis of Free Thyroid Hormone Assays

The FTI approach is therefore one in which the rules

for valid FT4 assays are not met. It produces a hybrid
result, with a very narrow window of validity, between a
total T4 measurement and a true FT4 measurement. The
positioning of a given FTI in the scale between these two
extremes depends on the conditions defining the T3uptake test (45 ). These include the concentration of T3
used relative to the volume of serum tested, the degree of
dilution of serum in the assay conditions, and the amount
and specificity of the competing binder used to sequestrate the excess T3. Assays using more specific secondary
binders that do not incidentally sequestrate T4 from
serum should be superior to those that use nonspecific
binders.
The routine laboratory can probe the relationship of a
given FTI with TBG concentrations in serum (which
should be carried out only with a panel of euthyroid sera
having a wide range of TBG concentrations from otherwise healthy subjects, and excluding either pregnant or
nonthyroidally ill subjects). It can also place the FTI
measurements along the arbitrary scale from total T4 to
FT4 (46 ). One need only measure total T4, FTI, and FT4 by
selected methods on a panel of 150 200 routine (not
severely ill or pregnant) subjects (9 ), including some
hypo- and hyperthyroid patients, to encompass virtually
the whole range of expected values for each marker. It is
assumed that each assay method has approximately the
same imprecision. Any pair of the three markers is now
correlated. If the FTI is acting as a hybrid assay, imperfectly correcting for the natural variation in serum binding protein concentrations in the panel, then it will
correlate more closely with total T4 than will the FT4 assay
(9 ). The extent to which the correlations differ is a rough
measure of how effectively the FTI approach corrects for
binding-protein differences. If changing from FTI to an
FT4 assay is contemplated, then several FT4 assays could
be used as comparators. In general, given equivalent
imprecision, the smaller the correlation coefficient with
total T4, the more likely is it that the assay in question is
correcting completely for serum binding protein variations.
In addition, a valid FT4 assay should show better
discrimination than FTI at the euthyroid-hyperthyroid
borderline of the reference interval (9, 46 ). Performance at
the hypothyroid borderline may not be so clearly altered
(9, 46 ).

thyroxine (t4) uptake (t-uptake) assays

In this method, labeled T4 (or a suitable analog) is added
in large excess to serum (47 ). The excess hormone (or
analog) fills the hitherto unoccupied protein T4-binding
sites and spills over into a free fraction, which can be
measured. Just as for T3-uptake methods, T4-uptake techniques (47 ) may suffer from similar distortions arising
from the very different concentrations and binding affinities of TBG on the one hand vs transthyretin and albumin
on the other (see above). Before addition of additional

hormone, TBG sites are already occupied by endogenous

T4 to a large extent, whereas transthyretin and albumin
sites are not. Much of the added hormone is thus bound to
transthyretin and albumin. This means that the distribution of added hormone is more sensitive to differences,
both relative and absolute, in the concentrations of these
binding proteins. A decrease in such concentrations relative to TBG, especially in NTI, may strongly affect the
T4-uptake value, giving, in turn, low index values (48 ).
These problems, as with T3 uptake, again produce a
hybrid result, with a truncated window of validity, in
which serum T4-binding protein effects distort the values.
The only advantage over the T3-uptake test is that when
labeled T4 is used, the same hormone is being used to
assess unoccupied sites, with the same affinities as the
hormone already occupying the serum binding protein
sites. However, if a labeled T4 analog is used instead
(T-uptake), its affinity for TBG (relative to T4) may be
attenuated more than for transthyretin and albumin. This
will exaggerate the influence of transthyretin and albumin
even further and may lead to expected falsely low results
in NTI, where concentrations of these proteins usually are
low.

Direct FT4 Immunoassays

two-step immunoassays
In the period 1978 1980, challenges arose to the hegemony of FTI, which fell increasingly under suspicion of
inadequate performance. These included a variety of
so-called kinetic FT4 assays (49, 50 ), followed by a twostep approach (51 ). The former assays had only a relatively brief existence and are now obsolete. However, the
two-step approach to FT4 was basically valid (51 ) and was
a major advance over what was then available commercially. In this test, serum is first incubated with a small
quantity of immobilized T4-specific antibody, which takes
up a very small quantity of T4 from the sample, minimally
disturbing the original serum boundfree equilibrium.
After equilibrium is reached, the serum and antibody are
separated (51 ). The immobilized antibody containing
bound T4 is then washed repeatedly to remove traces of
serum. It is then incubated with a labeled probe (e.g., a
suitable T4 derivative or, initially, 125I-labeled T4), and the
occupancy of antibody binding sites is measured as for a
typical immunoassay.
The two-step assay no doubt is independent, by definition, of the influence of the serum proteins and their
bound T4. There are, however, potential problems that
must be addressed. The first is the need to remove
rigorously all traces of serum proteins before the second
incubation is begun. Even minor amounts of binding
proteins could sequestrate the labeled probe because of
their high affinity. Modern two-step assays avoid this
problem by use of T4 conjugated with macromolecules
(e.g., enzymes). Such conjugates usually have negligible
affinity for serum protein binding sites. A second problem
is the possibility of back displacement of T4 bound in

Clinical Chemistry 47, No. 8, 2001

the first incubation by competition with labeled probe in

the second, and by the resulting nonequilibrium conditions for T4 binding (i.e., the phase of serum and its
remaining T4 has been washed away). Again, macromolecular T4 conjugates can minimize the first effect because
usually the conjugate will bind much less avidly to the
antibody than does T4. The use of T3 conjugates may
reduce competition even further. The potential for reequilibration of antibody-bound T4 remains however, and
suitably avid antibodies should be used to minimize this
effect. Short (nonequilibrium) second incubations can also
help.

one-step analog immunoassays

To avoid the inconvenience of using two separate processes, with their potential for greater overall imprecision,
Dr. T.A. Wilkins and I invented and developed the
one-step analog technique for the measurement of FT4
(8, 9, 52 ) and FT3 (53 ). In these assays, the labeled T4 (T3)
then used as a competing species for antibody binding
was replaced by a chemically modified T4 (T3) analog.
This had the property of binding avidly to a T4 (T3)specific antibody, but much less avidly than the native
hormone to the serum binding proteins (79, 54 ). The aim
of the system was to improve the convenience of two-step
approaches while maintaining validity.
Accordingly, one-step approaches made redundant the
initial separation of serum binding proteins required by
the two-step approach (51 ) and superficially resembled
the corresponding total-hormone assays. Early analog
methods, although they performed adequately in correcting for the physiologically encountered differences in
TBG and transthyretin concentrations in serum, were
correlated with serum albumin concentrations (7, 5558 ).
Although the assays were at the forefront compared with
anything else then commercially available, they were
criticized for imperfect performance (55 60 ). Vehement
controversy between us and others (2 4, 7, 59 63 ) as to
the exact working and validity of such assays regrettably
generated far more heat than light and led to hesitation in
some countries (especially the US) in adopting the methodology. The critics, however, paid little attention to the
fact that the alternative FTI methods, the most frequently
used assays at the time, had greater shortcomings in the
same and other areas. As an example of this critical
misinterpretation, the correlation of late-pregnancy sera
that produced low-normal analog FT4 values with serum
albumin concentrations was alleged to demonstrate the
invalidity of these assays (55 ). It is now known that this
correlation is merely a physiologic one and not an assay
shortcoming (45 ). Other studies alleged, wrongly, that the
assays were merely measuring the albumin-bound fraction of T4, and not FT4 (57 ). Practical experience in the
development of the modern one-step assays lacking interferences also showed that the stringent criteria laid down
by theorists (2 4, 59, 60 ) for the permitted amounts and
affinities of antibodies in valid FT4 assays could be

1357

considerably relaxed and still give acceptable dose

response values (64, 65 ).
At first (8, 9 ), the enforced use of radioactive labeling
probes limited analog immunoassay design. Assay development was constrained by the maximum specific activity
of the label compatible with the analogs chemical stability. Detection of adequate amounts of signal for convenient analysis required the antibody to sequestrate
slightly 1% of the total hormone in euthyroid sera with
reference binding protein concentrations (79 ), still well
within the window of validity for the assay (79 ). Albumin was also added to modulate the biasing effects of
variable serum albumin concentrations (7 ).
Since then, improvements in the performance of analog
FT4 and FT3 assays (64, 65 ) have eliminated interactions of
the labeled probe with any of the common serum binding
proteins. Newer nonradioactive labels, with higher detection sensitivity, use smaller amounts of antibody, consequently extracting a lower percentage of hormone
(64, 65 ). This gives three advantages. The first advantage
is that when measuring sera with abnormal T4-binding
capacities, intrinsic bias is minimized. The second advantage (65 ) is that, when they lack added albumin, assays
are more robust to progressive serum dilution (a useful
technique, although sometimes overinterpreted as a test
for assay validity and absence of bias). The third advantage is that FT4 results no longer significantly correlate
with serum binding protein concentrations, even albumin
(64 66 ).
Comparison of the performance of the newer and
original assays confirmed earlier forecasts (7 ) that the
need for improvement lay in the abolition of biasing
effects attributable to albumin. This comprises the
hypoalbuminemia of severe NTI (still a controversial area
in regard to the value of thyroid function tests) and the
rare conditions of analbuminemia, familial dysalbuminemic hyperthyroxinemia, and the presence in serum of
strongly binding T4 or T3 autoantibodies. The first three
have been addressed with good analog one-step assays
(67, 68 ), and all, including the vast majority of sera
containing autoantibodies, have been addressed successfully in the newer immunometric assays, where the antibody rather than the analog is labeled (64, 65 ).
With modern one-step assays, therefore, comparing
new methods with earlier ones is more difficult than
comparing an FTI with any FT4 method. This is because,
as stated above, the possible improvements in diagnostic
performance become increasingly marginal as the learning curve in development progresses. For the routine
laboratory, the most relevant test of the validity of any FT4
assay is to correlate, in euthyroid sera with wide variations in binding protein concentrations, the FT4 value
with each of the proteins TBG, transthyretin, and albumin. Once again, subjects receiving drugs such as dilantin
or aspirin, and nonthyroidally ill and pregnant subjects
should not be included in the correlations because these
may have altered physiology dependent on their condi-

1358

Midgley: Analysis of Free Thyroid Hormone Assays

tion. Specificity and sensitivity measurements are also

recommended for discrimination of hypo- and hyperthyroidism from the reference interval.
Dilution tests on sera are valuable, but require restraint. The useful limits of dilution experiments can be
calculated from the affinity constants and concentrations
of each of the serum T4-binding proteins, for which the
binding potential (BP) is given as the product of the two
parameters (K1,2,3P1,2,3). For a normal euthyroid serum,
BP is 6000 (4000 from TBG, 1300 from transthyretin, and
700 from albumin). Sera such as those in NTI could have
a BP value 1520% of this (i.e., 900 1200). Therefore,
dilution tests on a normal serum need only be pursued
out to a dilution (over and above the dilution of serum
intrinsic in the assay itself) of no more than five- to
sixfold. If the assay in question is sufficiently (not necessarily perfectly) robust out to this dilution factor, then it is
likely that sera with BP values of 900 1200 are being
measured without undue bias. Sera with BP values lower
than this are relatively rare. Dilution of sera with low BP
values is unnecessary except as an academic exercise to
probe the limits of validity of the assay. The methodology
is then being stretched into nonphysiologic regions that
may exceed its window of validity, and false indications
of assay failure may ensue. It may well be that an assay is
designed to encompass, without undue bias, the BP
values for the vast majority of sera encountered, but can
rapidly fail outside this range.
Comparison of results with those from a gold-standard
method such as equilibrium dialysis or ultrafiltration is
also valuable, if such facilities are available. However,
careful choices of sera should be made, and nonthyroidally ill subjects should not generally be included, nor
should serum from patients known to have taken drugs
that can affect the binding of T4 to the serum binding
proteins (see below).

immunometric one-step analog methods

More recently, further advances have been made in onestep FT4 analog technology, with two key innovations
(64, 65 ). The first innovation involved transferring the
detection label from the analog itself to a specific T4binding monoclonal antibody and immobilizing the analog as a complex insoluble matrix. In the second innovation, a cross-reacting hormone (T3) was used rather than
T4 itself for binding to antibody as the basis of insoluble
matrix formation (insoluble T3-matrix) (64, 65 ). This invention confronted the outstanding problem of how to
measure FT4 in sera containing avid autoantibodies
against T4 or T3, as well as finally preventing the interference of any other serum binding protein in the assay
(64 66 ). In this assay, the affinity of the insoluble T3matrix for the antibody is manyfold less than for FT4.
Thus, so that the relative rates of binding of the insoluble
T3-matrix and FT4 to the labeled antibody are roughly
equivalent, the concentration of this matrix must be
correspondingly manyfold higher than FT4 (64, 65 ).

Hence, occupation of the immobilized T3-matrix by a

small quantity of the antibody probe is negligible. This
allows the gross excess of unoccupied insoluble T3-matrix
binding sites to bind large amounts of extraneous autoantibodies without significantly affecting the binding ability of the labeled antibody probe (64, 65, 69, 70 ). The
method has been shown to accommodate the vast majority of such sera within the correct range for their underlying functional state (69, 70 ), unlike labeled-analog
methods, which are still often susceptible to interference
(71 ).
In addition, it was realized that the basic mode of
action of direct FT4 immunoassays, using avid antibodies,
could be expressed as a simple competition for binding of
FT4 and analog (or insoluble analog-matrix) until an
apparent equilibrium is reached (64, 65 ). The small, extremely slow dissociation of antibody-bound analyte and
analog thus greatly simplified the original complex arguments (7 ) that sought to explain the assay system (64, 65 ),
permitting the amounts and affinities of antibody and
analog to be more variable than the original, more rigid
theory allowed (2 4, 7, 59, 60 ).

albumin in the assay ingredients in one-step

assays
Initially, extra albumin was provided in the assay ingredients of the first one-step assays to ameliorate the biasing
effects of residual binding of the analog tracer to endogenous albumin in the serum sample (7 ). This inevitably
inserted an additional bias. The smaller the binding
protein concentrations in any given serum, the greater
was the influence of the added albumin on the assay
through its own T4 sequestration. However, direct experiments, using added albumin, showed that (a) the binding
properties of added albumin were variable and (b) the
bias was important only when serum albumin concentrations (but not those of TBG or transthyretin) differed
significantly from the norm (7, 58 ). Accordingly, in these
assays, FT4 measurements correlated only with serum
albumin concentrations (7, 66 ). Even then, the effects
were not marked until albumin concentrations had fallen
well below the reference interval for that protein [e.g., in
conditions ranging from severe NTI to analbuminemia
(7, 58 )]. Observed bias within the reference interval for
albumin concentrations did not have significant consequences for diagnostic discrimination (66 ).
Over the next 20 years, continual improvements in the
properties of the analogs eventually led to the abolition of
their binding to serum proteins (64, 65 ). Accordingly,
there was no need to add albumin expressly to counteract
bias from residual binding effects. However, another
problem manifested itself, which extra albumin in the
assay ingredients helped to ameliorate. In patients receiving heparin injections, blood lipoprotein lipase is activated (72, 73 ). When blood is withdrawn from such
patients, NEFAs are produced in vitro, often to very high
concentrations (74 78 ). The NEFAs in turn are capable of

Clinical Chemistry 47, No. 8, 2001

displacing substantial quantities of bound T4 from the

serum protein T4-binding sites and can increase the measured FT4 considerably, often grossly out of the reference
interval (30, 31, 74 78 ). Albumin in the assay ingredients
opposed this artificial increase in FT4 by mopping up
the NEFAs produced, thereby restoring most values to
normality (30, 31 ). Thus, assays containing albumin can
be used with heparinized subjects (with caution).
In severe NTI, there are additional problems. Initially,
in the days of FTI supremacy, it was believed that without
overt primary thyroid dysfunction, such patients were
essentially euthyroid, although there were considerable
decreases in the concentration of total T4 into the hypothyroid range, not wholly corrected by FTI (40 42, 79, 80 ).
Gold-standard methods for FT4 measurement variously gave results below, within, and above the euthyroid
reference interval (27, 41, 42, 81 85 ). Simultaneously, thyrotropin (TSH) concentrations could also be low, normal,
or occasionally above normal (86 89 ), although not necessarily in harmony with the FT4 values. It was postulated
that in NTI, the serum contained nondialyzable substances that could bind competitively to the T4-binding
proteins, displacing bound T4 into the free phase (90 92 ).
This explanation rationalized how the gold-standard FT4
measurements could be so out of line with total T4 values,
even when the serum binding protein concentrations
(especially transthyretin and albumin) were also severely
decreased. Nevertheless, FT4 estimates straddling the
reference interval also create problems in discriminating
true primary hypo- and hyperthyroidism in subjects with
NTI from subjects with no apparent thyroid dysfunction.
It was, no doubt, hoped that when new, more convenient assays were available, these anomalies would disappear. However, over the past 20 years, the exact status
of the NTI subject with regard to FT4 in serum has
remained as controversial as ever and certainly cannot be
conceived as steady-state with regard to thyroid function indicators (93 ). In NTI, overlaying any general level
of thyroid function, there seem to be continuous (independent) ebbs and surges in the production and retention
of substances either directly or indirectly related to FT4 as
the disease progresses (8192 ). These can include serum
binding proteins (94 ); interfering substances, including
administered drugs (32, 33 ); TSH (90 92 ); and probably
production of T4 and its conversion to T3 (95, 96 ). The true
thyroidal state of the subject can thus be obscured by
relatively short-lived changes in measured indicators,
which may, at random, show up in single samples taken
from a panel of NTI patients. This highlights the danger of
diagnosing NTI patients by single samples, making it
difficult to compare assay performance if the panel of sera
inevitably contains examples of these random events.
Indeed, because of the tendency of FT4 and TSH concentrations independently to enter and leave the euthyroid
reference interval, it is recommended that both markers
be measured in NTI and that careful monitoring is done to
rule out effects that might alter the concentration of either

1359

marker, e.g., administration of albumin-binding drugs

(66, 97 ). Only in this way can we assess the underlying
integrated level of thyroid function and discount transient
anomalies. The difficulties in diagnosing hypothyroidism
accompanying NTI have been vividly described (98 ).
Most studies comparing FT4 assay performance in NTI
have not taken this complexity fully into account and are
generally poorly designed. Assays with biases attributable to added albumin have often been criticized for
giving below-normal values when TSH is presumably
within the reference interval (31, 32, 96 ). However, because gold-standard assays lacking added albumin can
equally give diagnostically meaningless above-normal
values, misplaced confidence in their validity has often
led to support for assays that show the same effect (94 ).
Here there is a logical inconsistency. If albumin-containing FT4 assays can read too low and TSH is, in a random
NTI sample, transiently high, a misdiagnosis of hypothyroidism could occur. Conversely, if albumin-free assays
give above-normal values when TSH is transiently low, a
misdiagnosis of hyperthyroidism could equally be made.
Neither methodology thus can be completely trusted in
this situation. Additionally, albumin-containing assays
may mitigate the transient effects on FT4 of albuminbinding drugs in serum and depict the underlying thyroidal function more exactly.
Finally, the potential effects of assay incubation on the
concentration of interfering substances in serum should
not be underestimated. In an albumin-free assay using
a serum whose T4-binding sites are already saturated with
other molecules, the FT4 measurement is exquisitely sensitive to any further production of interference in vitro.
Published studies show that FT4 values are increased
when NEFA concentrations are 2.5 mmol/L (78, 99
101 ). The rate of further increase of FT4 on increases in
NEFAs is 710 pmol/L per mmol/L additional NEFAs
(7, 30, 78 ). Therefore, an increase in NEFAs (or the combination of some other interferent NEFAs) of 0.2
mmol/L will cause an (in vitro-dependent) increase in
measured FT4 of 1.52 pmol/L. Albumin-containing
assays can at least partially mitigate this effect (30 ). The
amount of exogenous albumin in the assay buffers that
can be used to minimize interference in FT4 assays depends very much on its properties of T4 binding as well as
its concentration. It is well known that commercial
sources of albumin can vary markedly in their T4-binding
characteristics (58 ). Such sources are properly termed
preparations rather than well-defined purified compounds because denaturation of albumin can occur to
various degrees during production. However, a calculation can be made of how large a BP for added albumin
will distort the FT4 values of most sera to an acceptably
minimal extent relative to the assay calibrators. This
would approximate, at most, to 10 25% of that found
endogenously in a serum with binding protein concentrations within the reference intervals. Thus, in an assay

1360

Midgley: Analysis of Free Thyroid Hormone Assays

using a 10-L sample and 100 L of reagent, the bovine

serum albumin content of the reagent should be 1.2 g/L.
Hence, FT4 values are not grossly affected until endogenous albumin concentrations are less than half-normal
and TBG concentrations are also substantially diminished.
Albumin preparations with weaker T4-binding affinities
can be used in greater quantities because the smaller
affinity offsets the effects of higher concentration in producing a suitable BP factor. Supplemental data on this
topic, in the form of an Appendix, are available in the
on-line version of this journal (http://www.clinchem.
org/content/vol47/issue8/).
Selecting assays that either do or do not contain
albumin in their buffers appears to be a trade-off of
intrinsic bias in FT4 values downward in NTI (and in all
other situations where the binding capacity is reduced),
caused by the added albumin, against increased sensitivity of an albumin-free assay to extraneous in vivo and in
vitro effects. Depending on their experience with discrimination of thyroidal disease in NTI subjects, assay users
must decide for themselves which to use.
A carefully controlled study on large numbers of
subjects, with a defined and graded severity of systemic
illness, is urgently needed, with full knowledge of medication taken and with multiple sampling, in which outcomes are known and where substantial numbers of NTI
patients with primary hypo- and hyperthyroidism are
also included. Such a study, using an albumin-containing
and an albumin-free assay for comparison, together with
TSH measurements, would more usefully evaluate the
sensitivity and specificity of each assay in diagnosing
thyroid disease in these conditions. In this way, one could
offset the misdiagnosis frequency from the inevitable
biases in the albumin-containing assay against the misdiagnosis frequency in albumin-free assays.
It is unfortunate that albumin is the only useful (and
inexpensive) protein able to mitigate the effects of interfering substances on FT4 values in heparinized, drugaffected, or nonthyroidally ill patients. However, chemical modification of albumin may greatly attenuate or
destroy its binding of T4 without destroying its ability to
bind other substances, such as NEFAs (102 ). The binding
sites for NEFAs and other substances are manyfold more
than for T4. Judicious protein modification might produce
a substance still able to quench interference without
causing significant bias through extra T4 binding. This
might yield an assay giving superior discrimination in
NTI through nullification of the influence of interfering
substances while not otherwise suffering from bias from
albumin addition.

Summary
The performance of convenient assays for FT4 has improved greatly since the first attempts (FTI) in 1965. Index
methods were accepted until it became clear that they
were still strongly influenced by TBG. In the meantime,
certain truths had come to be accepted by many work-

ers, not the least that in late pregnancy, FT4 results should
lie in the middle of or in the upper regions of the
euthyroid reference intervals. Better assays, correcting
more completely for TBG concentrations, forced a sometimes painful period of readjustment in this and other
areas, which has taken 20 years or more to reconcile in
some regions of the world. By now, the better assays for
FT4 have eliminated the biasing effects that arose in the
early assays from residual binding of labels to serum
albumin. The most recent improvements have also largely
addressed the remaining problem of T4 autoantibodies in
serum. The main problem outstanding in using FT4 assays
is that of diagnosis in NTI. This is a problem that lies more
in the vagaries of patient physiology than in any hope of
a final, simple, and definitive answer in an ideal assay.
The scope for further improvement in FT4 assay methodology is very limited. Most of the residual problems
(albumin dependency, dilution performance) have been
addressed, and a final decision can be taken by diagnosticians, not merely whether to terminate their use of FTI,
with all its shortcomings, but which direct FT4 assay to
use, based on simple comparisons of each assays diagnostic properties with those of other assays.

References
1. Robbins J, Rall JE. The iodine-containing hormones in blood. In:
Gray CH, James VHT, eds. Hormones in blood. London: Academic
Press, 1979:576 688.
2. Ekins R. Measurement of free hormones in blood [Review].
Endocr Rev 1990;11:5 46.
3. Ekins R. The free hormone hypothesis and measurement of free
hormones [Editorial]. Clin Chem 1992;38:1289 93.
4. Ekins R. The free hormone hypothesis and measurement of free
hormones [Letter]. Clin Chem 1993;39:1343 4.
5. Tabachnick M. Thyroxine-protein interactions. IV. Thermodynamic
values of thyroxine with human serum albumin. J Biol Chem
1967;242:1646 50.
6. Prince H, Ramsden DB. A new theoretical description of the
binding of thyroid hormones by serum proteins. Clin Endocrinol
1977;7:30724.
7. Wilkins TA, Midgley JEM, Barron N. Comprehensive study of a
thyroxin-analog-based assay for free thyroxin (Amerlex FT4).
Clin Chem 1985;31:1644 53.
8. Midgley JEM, Wilkins TA, Inventors. A method for determining the
free portions of substances in biological fluids. European Patent
No. 0026103, 1981.
9. Midgley JEM, Wilkins TA. The direct estimation of free hormones
by a simple equilibrium radioimmunoassay. Amersham, UK:
Amersham International Ltd, 1982.
10. Nelson JC, Weiss RM. The effect of serum dilution on free
thyroxine (FT4) concentrations in the low-T4 syndrome of nonthyroidal illness. J Clin Endocrinol Metab 1985;61:239 46.
11. Midgley JEM, Wilkins TA, Moon CR. Validity of analog free
thyroxin immunoassays. Part II [Opinion]. Clin Chem 1987;33:
2145 8.
12. Ross HA, Benraad TJ. Is free thyroxine accurately measured at
room temperature? Clin Chem 1992;38:880 7.
13. van der Sluijs Veer G, Vermes I, Bonte HA, Hoorn RKJ. Temperature effects of free thyroxine measurements: analytical and
clinical consequences. Clin Chem 1992;38:132731.
14. Docter R, van Toor H, Krenning EP, de Jong M, Hennemann G.

Clinical Chemistry 47, No. 8, 2001

15.

16.

17.

18.

19.

20.

21.

22.

23.
24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

Free thyroxine assessed with three assays in sera of patients

with nonthyroidal illness and of subjects with abnormal concentrations of thyroxine-binding proteins. Clin Chem 1993;39:
1668 74.
Konno N. Serum thyrotropin response to thyrotropin-releasing
hormone and free thyroid hormone indices in patients with
familial thyroxine-binding globulin deficiency. Endocrinol Jpn
1976;23:3137.
Roosdorp N, Joustra N. A numerical comparison of the use of
T3-uptake values and of TBG levels for the estimation of free
thyroxine in serum. Clin Chim Acta 1979;98:2733.
Wilke TJ. A challenge of several concepts of free thyroxin index
for assessing thyroid status in patients with altered thyroidbinding protein capacity. Clin Chem 1983;29:56 9.
Nelson JC, Tomei RT. Dependence of the thyroxin/thyroxinbinding globulin (TBG) ratio and the free thyroxin index on TBG
concentrations. Clin Chem 1989;35:541 4.
Nelson JC, Tomei RT. Direct determination of free thyroxine in
undiluted serum by equilibrium dialysis/radioimmunoassay. Clin
Chem 1988;34:1737 44.
Nelson JC, Wilcox RB. Protein bound T4 dependence: the uncontrolled variable in free T4 assays. Exp Clin Endocrinol 1994;102:
1029.
Nelson JC, Nayak SS, Wilcox RB. Variable underestimates by
serum free thyroxine (T4) immunoassays of free T4 concentrations in simple solutions. J Clin Endocrinol Metab 1994;79:
13735.
Nelson JC, Weiss RM, Wilcox RB. Underestimates of serum free
thyroxine (FT4) concentrations by free T4 immunoassay. J Clin
Endocrinol Metab 1994;79:76 9.
Nelson JC, Wilcox RB. Analytical performance of free and total
thyroxine assays. Clin Chem 1996;42:146 54.
Wang R, Nelson JC, Weiss RM, Wilcox RB. Accuracy of free
thyroxine measurements across natural ranges of thyroxine
binding to serum proteins. Thyroid 2000;10:319.
Ellis M, Ekins RP. The radioimmunoassay of serum free triiodothyronine and thyroxine. In: Pasternak CA, ed. Radioimmunoassay in clinical chemistry. London: Heyden, 1975:299 306.
Giles AF. An improved method for the measurement of free
thyroxine in serum dialysates. Clin Endocrinol (Oxf) 1982;16:
1015.
Helenius T, Liewendahl K. Improved dialysis method for free
thyroxin in serum compared with five commercial radioimmunoassays in nonthyroidal illness and subjects with abnormal concentrations of thyroxin-binding globulin. Clin Chem 1983;29:
816 22.
Larsen PR. Salicylate-induced increases in free triiodothyronine
in human serum. Evidence of inhibition of triiodothyronine binding to thyroxine-binding globulin and thyroxine-binding prealbumin. J Clin Invest 1972;51:112534.
Tabachnick M, Hao Y-L, Korcek L. Effect of oleate, diphenylhydantoin and heparin on the binding of 125I-thyroxine to purified
thyroxine-binding globulin. J Clin Endocrinol Metab 1972;36:
392 4.
Midgley JEM, Wilkins TA, Giles AF. Treatment with heparin and
results for free thyroxin: an in vivo or an in vitro effect? Clin Chem
1982;28:24413.
Bayer MF. Effects of heparin on serum free thyroxine linked to
post-heparin lipolytic activity. Clin Endocrinol (Oxf) 1983;19:
591 6.
Stockigt JR, Lim CF, Barlow JW, Stevens V, Topliss DJ, Wynne
KN. High concentrations of furoseamide inhibit serum binding of
thyroxine. J Clin Endocrinol Metab 1984;59:62 6.
Lim CF, Bai Y, Topliss DJ, Barlow JW, Stockigt JR. Drug and fatty

34.

35.

36.

37.

38.

39.
40.
41.

42.

43.
44.

45.

46.
47.

48.

49.

50.

51.

52.
53.

1361

acid effects on serum thyroid hormone binding. J Clin Endocrinol

Metab 1988;67:682 8.
Clark F, Horn DB. Assessment of thyroid function by the combined use of the serum protein bound iodine and resin uptake of
131I-triiodothyronine. J Clin Endocrinol Metab 1965;25:39 45.
Sutherland RL, Simpson-Morgan MW. The thyroxine-binding properties of serum proteins. A competitive binding technique employing Sephadex G-25. J Endocrinol 1975;65:319 22.
Nilsson SF, Peterson PA. Studies on thyroid-hormone-binding
protein. I. The subunit structure of human thyroxine-binding
globulin and its interaction with ligands. J Biol Chem 1975;250:
854353.
Korcek L, Tabachnick M. Thyroxine-protein interactions. Interaction of thyroxine and triiodothyronine with human TBG. J Biol
Chem 1976;251:3558 62.
Snyder SM, Cavalieri RR, Goldfine D, Ingbar SH, Jorgensen EC.
Binding of thyroid hormones and their analogues to thyroxinebinding globulin. J Biol Chem 1976;251:6489 94.
Carter JN, Eastman CJ, Corcoran JM, Lazarus L. Effect of severe,
chronic illness on thyroid function. Lancet 1974;2:971 4.
Wood DG, Cyrus J, Samols E. Low T4 and low FT4I in seriously ill
patients [Concise Communication]. J Nucl Med 1980;21:4325.
Chopra IJ, Solomon DH, Gershon W, Hepner G, Morgenstein AA.
Misleadingly low free thyroxine index and usefulness of reverse
triiodothyronine measurement in non-thyroidal illnesses. Ann
Intern Med 1979;90:90512.
Chopra IJ, Van Herle AJ, Chua Teco GN, Nguyen AH. Serum free
thyroxine in thyroidal and nonthyroidal illnesses: a comparison of
measurements by radioimmunoassay, equilibrium dialysis and
free thyroxine index. J Endocrinol Metab 1980;51:135 43.
Kurtz A, Dwyer K, Ekins RP. Serum free thyroxine in pregnancy
[Letter]. Br Med J 1979;ii:550 1.
Whitworth AS, Midgley JEM, Wilkins TA. A comparison of free T4
and the ratio of total T4 to T4-binding globulin in serum through
pregnancy. Clin Endocrinol (Oxf) 1982;17:30713.
Ball R, Freedman DB, Holmes JC, Midgley JEM, Sheehan CP.
Low-normal concentrations of free thyroxin in serum in late
pregnancy: physiological fact not technical artifact. Clin Chem
1989;35:1891 6.
Midgley JEM. Methodological background of the Amersham
Amerlex free thyroxine RIA. Nuklearmedizin 1982;21:174 83.
Symons RG, Vining RF. An evaluation of a fluorescence polarization immunoassay of thyroxine and thyroxine-uptake. Clin Chem
1985;31:1342 8.
Mendel CM. Preliminary evaluation of a fluorescence polarization
immunoassay (Abbott TDx) for estimating serum free thyroxine
concentrations in patients with critical nonthyroidal illness and
low total thyroxine concentrations in serum [Technical Brief]. Clin
Chem 1992;38:1916 7.
Hertl W, Odstrchel G. Kinetic and thermodynamic studies of
antigen-antibody interactions in heterogeneous reaction phases.
I. L-Thyroxine (T4) with specific antibody immobilised on controlled pore glass. Mol Immunol 1979;16:173 8.
Ashkar FS, Buehler RJ, Chan T, Howrani M. Radioimmunoassay
of free thyroxine with pre-bound anti-T4 microcapsules. J Nucl
Med 1979;20:956 60.
Ekins RP, Filetti S, Kurtz AB, Dwyer K. A simple general method
for the assay of free hormones (and drugs); its application to the
measurement of serum free thyroxine levels and the bearing of
the results on the free thyroxine concept. J Endocrinol 1980;
85:29P30P.
Midgley JEM, Wilkins TA. Hypothyroidism in patients on fenclofenac. Lancet 1980;ii:704 5.
Wilkins TA, Midgley JEM, Stevens RAJ, Caughey I, Barron N.

1362

54.

55.

56.

57.

58.

59.
60.
61.

62.
63.

64.

65.

66.

67.

68.
69.

70.

71.

72.

73.
74.

Midgley: Analysis of Free Thyroid Hormone Assays

Assay performance and tracer properties for two analog-based

assays of free triiodothyronine. Clin Chem 1986;32:4659.
Midgley JEM, Wilkins TA. An improved method for the estimation
of the relative binding constants of T4 and its analogues with
serum proteins. Clin Endocrinol (Oxf) 1982;17:523 8.
Amino N, Nishi K, Nakatani K, Mizuta H, Ichihara K, Tanizawa O,
Miyai K. Effect of albumin concentration of the assay of serum
free thyroxine by equilibrium radioimmunoassay with labeled
thyroxin analog (Amerlex free T4). Clin Chem 1983;29:3215.
Bayer MF. Free thyroxine results are affected by albumin concentrations and non-thyroidal illness. Clin Chim Acta 1983;130:
391 6.
Stockigt JR, Stevens V, White EL, Barlow JW. Unbound analog
radioimmunoassays for free thyroxine measure the albuminbound hormone fraction. Clin Chem 1983;29:1408 10.
Midgley JEM, Winton MRJ, Wilkins TA. Relationship between
effects of added albumin, initial free thyroxine value and endogenous serum-binding protein concentrations on Amerlex free
thyroxine estimations. Clin Chim Acta 1987;167:6779.
Ekins RP. Free hormones in blood: the concept and the measurement. J Clin Immunoassay 1984;7:163 80.
Ekins R. Validity of analog free thyroxin immunoassays [Opinions
and Responses]. Clin Chem 1987;33:213752.
Midgley JEM, Wilkins TA. Effect of albumin concentration on
equilibrium radioimmunoassay of serum free thyroxin with labeled thyroxin analog (Amerlex Free T4) [Letter]. Clin Chem
1983;29:18613.
Wilkins TA, Midgley JEM. Analysis of Mass Action model for free
thyroxin assays. Clin Chem 1986;32:566 8.
Midgley JEM. The free thyroid hormone hypothesis and measurement of free hormones [Letter and Response]. Clin Chem
1993;39:1342 4.
Christofides ND, Sheehan CP, Midgley JEM. One-step, labeledantibody assay for measuring free thyroxin. I. Assay development
and validation. Clin Chem 1992;38:11 8.
Christofides ND, Sheehan CP. Enhanced chemiluminescence
labeled-antibody immunoassay (Amerlite-MAB) for free thyroxine:
development and technical validation. Clin Chem 1995;41:17
23.
Midgley JEM, Sheehan CP, Christofides ND, Fry JE, Browning D,
Mardell R. Concentrations of free thyroxin and albumin in serum
in severe nonthyroidal illness: assay artifacts and physiological
influences. Clin Chem 1990;36:76571.
Seghers J, De Nayer P, Beckers C. Clinical evaluation of Amerlites one-step immunoassay for determining free thyroxin [Technical Brief]. Clin Chem 1988;34:2160.
Reilly CP, Adamek KJ, Wellby ML. Critical assessment of the
Amerlite free T4 assay. Ann Clin Biochem 1989;26:51721.
John R, Henley R, Shankland D. Concentrations of free thyroxine
and free triiodothyronine in serum of patients with thyroxin- and
triiodothyronine-binding autoantibodies. Clin Chem 1990;36:
470 3.
Sapin R, Gasser F, Schlienger JL, Chambron J. Analytical and
clinical evaluation of a new one-step non-analogue radioimmunoassay for serum free thyroxine. Eur J Nucl Med 1989;17:1115.
Sapin R, Gasser F, Schlienger JL. Familial dysalbuminemic
hyperthyroxinemia and thyroid hormone autoantibodies: interference in current free thyroid hormone assays. Horm Res 1996;
45:139 41.
Giacomini KM, Swezey SE, Giacomini JC, Blaschke TF. Administration of heparin causes in vitro release of nonesterified fatty
acids in human plasma. Life Sci 1980;27:771 80.
Riemersma RA, Russell DC, Oliver MF. Heparin-induced lipolysis;
an exaggerated risk. Lancet 1981;ii:471.
Mendel CM, Frost PH, Cavalieri RR. Effect of free fatty acids on

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.
92.

93.

the concentration of free thyroxine in human serum: the role of

albumin. J Clin Endocrinol Metab 1986;63:1394 9.
Mendel CM, Frost PH, Cavalieri RR. In vitro lipolysis during
equilibrium dialysis may cause overestimation of free thyroxine
levels [Abstract]. Clin Res 1986;34:429A.
Mendel CM, Frost PH, Kunitake ST, Cavalieri RR. Mechanism of
the heparin-induced increase in the concentration of free thyroxine in plasma. J Clin Endocrinol Metab 1987;65:1259 64.
Juame JC, Mendel CM, Frost PH, Greenspan FS, Laughton CW.
Extremely low doses of heparin release lipase activity into the
plasma and can thereby cause artifactual elevation in the serum
free thyroxine concentrations as measured by equilibrium dialysis. Thyroid 1996;6:79 83.
Stevenson HP, Archbold GPR, Johnston P, Young IS, Sheridan B.
Misleading serum free thyroxine results during low molecular
weight heparin treatment. Clin Chem 1998;44:10027.
Hershman JM, Krugman LG, Kapple JD, Reed AW, Azukizawa M,
Shinaberger JH. Thyroid function in patients undergoing maintenance hemodialysis; unexplained low serum thyroxine concentration. Metabolism 1978;27:7559.
Braverman LE, Abreau CM, Brock P, Kleinmann R, Fournier L,
Odstrchel G, Shoemaker HJ. Measurement of serum free thyroxine by RIA in various clinical states. J Nucl Med 1980;21:2339.
Bayer MF, McDougall IR. Radioimmunoassay of free thyroxine in
serum: comparison with clinical findings and results of conventional thyroid function tests. Clin Chem 1980;26:1186 92.
Kaptein EM, McIntyre SS, Weiner JM, Spencer CA, Nicoloff JT.
Free thyroxine estimates in nonthyroidal illness: comparison of
eight methods. J Clin Endocrinol Metab 1981;52:10737.
Kaptein EM, Grieb DA, Spencer CA, Wheeler WS, Nicoloff JT.
Thyroxine metabolism in the low thyroxine state of critical
nonthyroidal illness. J Clin Endocrinol Metab 1981;53:764 71.
Slag MF, Morley JE, Elson MK, Labrosse KR, Crowson TW, Nuttall
FQ, Shafer RB. Free thyroxine level in critically ill patients: a
comparison of currently available assays. JAMA 1981;246:
2702 6.
Melmed S, Geola FL, Reed AW, Pekary AE, Hershman JM. A
comparison of methods for assessing thyroid function in nonthyroidal illness. J Clin Endocrinol Metab 1982;54:300 6.
Wehmann RE, Gregerman RI, Burns WH, Sarai R, Santos GW.
Suppression of thyrotropin in the low thyroxine state of severe
nonthyroidal illness. N Engl J Med 1985;312:546 52.
Brent GA, Hershman JM, Braunstein GD. Patients with severe
nonthyroidal illness and serum thyrotropin concentrations in the
hypothyroid range. Am J Med 1986;81:463 6.
Bayer MF, Macoviak JA, McDougall IR. Diagnostic performance of
sensitive measurements of serum thyrotropin during severe
nonthyroidal illness: their role in the diagnosis of hyperthyroidism. Clin Chem 1987;33:2178 84.
Eggertsen R, Petersen K, Lundberg PA, Nystrom E, Lindstedt G.
Screening for thyroid disease in a primary care unit with a thyroid
stimulating hormone assay with a low detection limit. Br Med J
1988;297:1586 92.
Chopra IJ, Chua Teco GN, Nguyen AH, Solomon DH. In search of
an inhibitor of thyroid hormone binding to serum proteins in
nonthyroidal illness. J Clin Endocrinol Metab 1979;49:64 9.
Woeber KA, Maddux BA. Thyroid hormone binding in nonthyroidal
illness. Metabolism 1981;30:412 6.
Oppenheimer JH, Schwartz HL, Mariash CN, Kaiser FE. Evidence
for a factor in the sera of patients with nonthyroidal disease
which inhibits iodothyronine binding by solid matrices, serum
proteins and rat hepatocytes. J Clin Endocrinol Metab 1982;54:
757 66.
de Groot LJ. Dangerous dogmas in medicine: the nonthyroidal

Clinical Chemistry 47, No. 8, 2001

illness syndrome [Review]. J Endocrinol Metab 1999;84:151

64.
94. Csako G, Zweig MH, Glickman J, Ruddel M, Kestner J. Direct and
indirect techniques for free thyroxin compared in patients with
nonthyroidal illness. II. Effect of albumin, prealbumin and thyroxin-binding globulin. Clin Chem 1989;35:1655 62.
95. Vagenakis AG, Portnay GI, OBrian JT, Rudolph M, Arky RA, Ingbar
SH, Braverman LE. Effect of starvation on the production and
metabolism of thyroxine and triiodothyronine in euthyroid obese
patients, J Clin Endocrinol Metab 1977;45:13059.
96. Balsam A, Ingbar SH. The influence of fasting, diabetes and
several pharmacological agents on the pathways of thyroxine
metabolism in rat liver. J Clin Invest 1978;62:41524.

1363

97. Stockigt JR. Guidelines for diagnosis and monitoring of thyroid

disease: nonthyroidal illness. Clin Chem 1996;42:188 92.
98. Lamb EJ, Martin J. Thyroid function tests: often justified in the
acutely ill. Ann Clin Biochem 2000;37:158 64.
99. Schatz DL, Sheppard RH, Steiner G, Chandarlapaty CS, de Veber
GA. Influence of heparin on serum free thyroxine. J Clin Endocrinol Metab 1969;29:101522.
100. Wang YS, Hershman JM, Smith V, Pekary AE. Effect of heparin on
free thyroxin as measured by equilibrium dialysis and ultrafiltration. Clin Chem 1986;32:700.
101. Jain R, Uy HL. Increase in serum free thyroxine levels related to
intravenous heparin treatment. Ann intern Med 1996;12:74 5.
102. Spector AA. Fatty acid binding to plasma albumin [Review]. J
Lipid Res 1975;16:16579.