CLIN. CHEM. 30/9, 1558-1560(1984)
Use of a Rapid, SensitiveImmunoradiometric
Assayfor Thyrotropinto
DistinguishNormalfrom HyperthyroidSubjects
William E. Cobb, R. Preston Lamberton,’ and Ivor M. D. Jackson”2
Usinga new three-siteimmunoradiometricassayforthyrotropin (TSH), we measuredconcentrationof thishormonein the
serum of 47 patients with hyperthyroidism and 46 controls.
The mean and range of serum TSH concentration was
significantly lowerin thyrotoxicthan in controlsubjects,and it
was possible to correctly identify 96% of thyrotoxic patients
onthebasisofa serum TSH concentration
less than 0.5 milli-
mt. uniVL.
We conclude
that such measurement
is highly
sensitive for distinguishing hyperthyroid from normal subjects, and that the lower limit of normal for TSH in serum is
about 0.5 milli-int.unitlL.
Addftlonal Keyphrases: thyroid status
reference interval
In accordance with the negative-feedback
regulation
of
secretion of pituitary thyrotropin (TSH, thyroid-stimulating
hormone) by thyroid hormone, the TSH in serum is suppressed in states of increased thyroid function (1), so that a
single measurement
of the basal concentration
of TSH in
serum should be a potentially useful test for distinguishing
between normal and hyperthyroid
patients.
However,
unless the assay is modified by extraction of serum or of the
radioactive label, or by prolonging the time of incubation (24), the low end of the normal range for serum TSH concentration is below the sensitivity
of most TSH radioimniunoassays (RIA) that are currently in widespread use. Accordingly, until now it has not been practical to utilize the basal
serum TSH as a diagnostic aid in the evaluation of hyperthyroidpatients.
To improve sensitivity over that available
with the conventional TSH-RJA,
a magnetic solid-phase,
three-site
immunoradiometric
assay (mr)
was recently
developed (Serono Diagnostics).
The specific aim of this study was to determine whether
this new rapid TSH-mzviA would permit us to distinguish
thyrotoxic from normal subjects on the basis of a single
measurement
of the serum TSH concentration.
Materials and Methods
We used a TSH-nu
kit (Serono Diagnostics, Randolph,
MA) to study the concentration of TSH in the sera of 47
patientswith thyrotoxicosis
and 46 controls.
Patients. All of the thyrotoxic patients had been evaluated
in the Endocrine Clinic of the New England Medical Center
Hospital, where the diagnosis was established by the characteristic clinical
and biochemical features of this disorder.
The control population consisted of patients attending the
Endocrine Clinic or individuals undergoing a pre-employment health evaluation at the New England Medical Center
Employee Health Clinic. None of the controls had a known
Department of Medicine, Division of Endocrinology, Tufts-New
England Medical Center, Boston, MA 02111.
1 Present address: Division of Endocrinology, Brown University,
Rhode Island Hospital, 593 Eddy St., Providence, RI 02902.
2 Author to whom correspondence
should be sent.
Received April 2, 1984; accepted June 11, 1984.
1558
CLINICALCHEMISTRY,Vol. 30, No. 9, 1984
hi8tory
current
tration.
at -20
study.
of thyroid disease or goiter, or any condition or
medication known to affect the serum TSH concenSera from all patients and controls had been stored
#{176}C
for intervals of a week to 18 months before the
Laboratory
methods. The assay involves three monoclonal
antibodies.
Two of these are labeled with isa1 and reportedly
bind to different sites on TSH that each contain alpha- and
beta-subunit
determinants.
The third antibody is coupled to
fluorescein isothiocyanate
and reportedly
binds only to a
TSH beta-subunit
determinant.
Standard solutvjns consist
of 0, 0.5, 1, 5, 10, 30, and 50 milli-int. units/L caF,rators,
standardized
against
International
Reference Preparation
TSH 80/558, provided by the World Health Organization.
We incubated 100 jsL of a reagent containing the three
monoclonal
antibodies
with 200 pL of either standard
solutions, serum controls, or serum unknowns, for 2 h at
room temperature.The labeledantibody-TSH
complexes
that formed were then isolated in a magnetic
solid-phase
separation step, as follows. We added 200 uL of a reagent
containing
a sheep polyclonal antibody
to fluorescein isothiocyanate
coated onto a magnetic solid-phase, incubated
the mixture for an additional 10 mm at room temperature,
and separated the solid-phase from the supernatant
fluid,
washed the sediment once with 500 L of distilled water,
and measured its radioactivity in a gamma counter.
All determinations
were made in duplicate for each assay.
Statistical
methods.
We determined
the unknown ‘NH
concentrations
by solving for each duplicate sample a computer-generated
equationthatbestfitsthelinear regression
of radioactivity
(in counts/mm)
in the calibrator range.
Within this range, the regression
was linear (in three
separate assays, r = 1.000, 1.000, and 0.997 for concentrations of 0 to 30 milli-int. unitsfL). Because we knew values
would not exceed 30 milli-int. units/L, we eliminated the 50
milli-int. unit/L calibrator when calculating
the regression
equations.
The mean detection limit calculated from the results of
the zero standard plus 2 SD (nine assays) was 0.2milli-int.
unit/L (range 0.0 to 0.5 milli-int.
unitlL).The interassay
coefficient
of variation(CV) of four controlsrun in each
assay was 7.4, 4.5, 5.2, and 3.7% at 1.6, 4.4, 7.4, and 27.8
milli-int.
unitlLconcentrations,
respectively.
To determine
assay precision at 0.5 milli-int. unitlL, we calculated the
interassay CV of (B05-B0)/B0 to be 10.4%, where B0 represents the counts bound at zero dose and B05 the counts
bound at the 0.5milli-int.
unitiLdose.
We determined the distributions and means for both the
TSH concentration
and the log TSH concentration in serum.
We computed with TSH values <0.2 milli-int. unitfL as if
they were 0.1 milli-int. unitiL.
Normality
of the distribution
was tested by calculating
the Wilk-Shapiro
statistic, and the means of the two
populations were compared by the Wilcoxon nonpaired rank
sum test. All statistical
calculations
were performed by
computer
(We used curm’o software; Bolt, Beranek
and
Newman, Inc., Cambridge, MA).
Results
The mean TSH concentration in the serum of 47 patients
with hyperthyroidism
was significantly
lower than the
mean concentration of 46 controls (p <0.001). The results
are shown in Table 1 as the mean and standard error (SE)
both of TSH and the log TSH concentrations,
because tests
for normality showed that TSH was not distributed normally in either group and that log TSH followed a normal
distribution
in controls but not among thyrotoxic patients.
However, the distribution of thyrotoxicvalues cannot be
determined
with certainty because it is heavily skewed by
the large number of thyrotoxic
patients(51%) whose TSH
was undetectable
(i.e., <0.2 milli-int. unit/L). Inspection of
the histograms
(Figure 1) suggests that the hyperthyroid
values, like the controls, might also conform to a log
distribution
if TSH values <0.2 milli-int. unitiL could be
accurately
determined,
but for the purpose of statistical
comparison,
we treated the distributions as nonparametric
data.
The serum TSH concentration was below the lower limit
of the control range (0.5 milli-int.
unit/L) in 45 of 47 patients
with hyperthyroidism.
Of the two hyperthyroid
patients
whose serum TSH overlappedthe controlrange, one had a
serum TSH of 0.5 milli-int. unitlL, the other a TSH of 0.7
milli-int.
unitlL.The range of ± 4 SD from the log mean
TSH (1.8) in the controls is 0.5 to 6 milli-int. units/L, which
includes 100% of values in the normal subjects. The serum
TSH in 45 of 47 hyperthyroid
patients fell below this range.
Among controls, the serum TSH concentration
was equal
to or >0.5 milli-int. unitlL in all, and >1 milli-int. unit/L in
43 of 46 (93%). Of the three patients whose TSH was <1
milli-int. unitiL, the values for two (0.5 and 0.7 milli-int.
unitlL) were within the range of TSH concentrations
found
among thyrotoxic patients.
These findings
indicate
that 96% of thyrotoxic patients
can be distinguishedfrom normal on the basis of a single
measurement
of the TSH concentration.
Table 1. Mean and Range of TSH and Log TSH
Concentrations in Sera from Normal and
Hyperthyroid Persons
TSH concna
Mean (and SEM)
Normal (n = 47)
Log mean (and SE range)
Range
2.0
1.8
(1,7-1.9)
0.5-5.6
(0.2)
Hyperthyroid (n = 46)
0.2
0.2
(0.0)
milli-int. units/L.
<0.2-0.7
(<0.2-0.2)
HYPERTHYROID
NORMAL
4,
a
a0
4,
n
Discussion
This study shows that a rapid, sensitive TSH-nu.
can
distinguish between the low serum TSH found inhyperthyroid patients and normal TSH concentration
in serum. As a
test for thyrotoxicosis,
measurement
of TSH in a single
serum sample in this assay had a diagnostic
precision
of
96%, which represents the percentage of thyrotoxic patients
whose serum TSH concentration
was below the range for
the euthyroid controls. Although it has been possible to
modify conventional TSH-RJAs to permit measurement
of
TSH concentrations
that are below the normal range, the
improvements
require separation
and (or) purification steps,
and long incubations under disequilibrium
conditions (2-4)
that are not feasible for most laboratories
that perform
routine clinical assays. The TSH-mr
we used is technically
simple and can be run in a single half-day in any laboratory
equipped for basic RIA determinations.
The mean and range for TSH in our patients agree closely
with those found in other studies in which highly sensitive
RIAs or cytochemical bioassays for TSH were used (Table 2).
In two studies in which RIA was used, the lower limit of
normal TSH was 0.5 milli-int. unitfL (2,3), and in a third, 1
milli-int. unit/L (4).In our study the values for both the
lowest control (0.5 milli-int. unitfL) and 4 SD below the log
mean TSH of controls (0.5 milli-int. unitlL) suggest a lower
limit for the normal TSH concentration
of approximately
0.5
milli-int.
unit.IL.
S
z
0
.2
.4
.6
.8
0
2
3
4
5
6
TSH, milli-Int. unitslL
Fig. 1. Distributionof serum TSH concentrationsamong thyrotoxic
(left) and amongcontrolsubjects(right)
patients
We conclude that measurement
of the serum TSH concentration by this new TSH-mr& is a highly specific means of
distinguishing
normal
from thyrotoxicsubjects.
A serum
TSH <0.5 milli-int. unitlL indicates a suppressed concentration consistent with thyrotoxicosis,
whereas a TSH >0.7
milli-int. unit/L excludes this diagnosis.
Table 2. Mean (and SEM) and Range of 15K Concentration in Sera of Normal and Hyperthyroid
Persons, from Various Studies Involving Highly SensItive TSH Assays
Normal
Study
(ret.)
This report
Assay
method
Mean
(and SEM)
2
RIA
3
RIA
59
4
5
RIA
49
CBA
Mean
(and SEM)
Range
n
47
14
TSH, milli-Int. unlts/L
n
46
31
IRMA
Hyperthyroid
(0.2)
1.31 (0.15)
1.49 (0.10)
2.44 (0.18)
2.Oa
0.5-5.6
0.5-4.5
0.5-4.5
1.0-6.8
10
0.93 (0.69)
a Although our study shows that TSH follows a log normal distribution, the mean
29
±
Range
TSH, milli-Int. unlts/L
0.2
0.33
(0.0)a
62
0.33
0.34 (0.02)
9
0.28 (0.46)
<0.2-0.7
<0.33
<0.33
0.1-0.85
Sensitivity
of assay,
milli-int.
units/I..
0.2
0.33
0.33
0.005
-
SEM of the untransformed data has been used here for comparison with
results of prior studies that were not expressed as log transformed values but as the mean
CBA, competitive binding assay.
±
SEM of the untransformed TSH concentration.
CLINICALCHEMISTRY,Vol. 30, No. 9,
1984
1559
We acknowledge the expert technical assistance of Mr. Tom
Conlon, cwnro Systems Manager of the General Clinical Research
Center, Tufts-New England Medical Center, and the gift of materials and supplies, including 1’SH-uu kits, from Serono Diagnostics,
Inc. This study was supported in part by NIH Grant AM 21863.
References
1. Jackson IMD. Neuroendocrine control of pituitary TSH secretion. Thyroid Today 6, 1-7 (1983).
2. Wehinann RE, Rubenstem HA, Nisula BC. A sensitive conve-
1560 CLINICALCHEMISTRY, Vol. 30, No. 9, 1984
nient radioimmunoassay procedure which demonstrates that serum
hTSH is suppressed below the normal range in thyrotoxic patients.
Endocrinol Res Commun 6, 249-255 (1979).
3. Wehmann RE, Rubenstein HA, Pugeat MM, Nisula BC. Extended clinical study of a sensitive and reliable radioimmunoassay of
thyroid-stimulating hormone. South Med J 76, 969-976(1982).
4. Spencer CA, Nicolofi’ JT. Improved radioimmunoassay for human TSH. Clin Chim Acta 108, 415-424 (1980).
5. Peterson V, Rees Smith B, Hall R. A study of thyroid stimulating
activity in human serum with the highly sensitive
cytochemical
bioassay. J Clin Endocrinol Metab 41, 199-202 (1975).