25 Clinica Chimica Acta, 60 (1975) 25-32 0 Elsevier Scientific Publishing Company, Amsterdam -Printed
in The Netherlands
CCA 6905
SEMI-AUTOMATED RADIOIMMUNOASSAYS THYROXINE AND TRIIODOTHYRONINE
G.S. CHALLAND, Radioimmunoassay
W.A. RATCLIFFE Unit, Stohhill
FOR TOTAL
SERUM
and J.G. RATCLIFFE
General Hospital,
Glasgow G21 3UW (U.K.)
(Received September 21, 1974)
Summary Single stage semi-automated radioimmunoassays for total serum thyroxine (T4) and triiodothyronine (T3) are described which employ an automatic pipetting station, automatic gamma counter, and a programmable calculator with paper tape reader and printing facility. Both assays require only a small volume of unextracted serum, and are specific and sensitive. Their sample capacity, precision, speed, and cost are comparable with the measurement of serum protein-bound iodine. Both assays therefore have significant advantages over previous methods for the assessment of thyroid function in the diagnostic laboratory. A simple method of automating the calculation of results is described, which is applicable to any radioimmunoassay in which the standard curve is approximately linear on a plot of the free/bound fraction against the antigen concentration. In addition, a general method is reported which reveals the relative contributions of intrinsic, systematic, and random error to radioimmunoassay precision.
Introduction The assessment of thyroid status in vitro requires the measurement of circulating thyroid hormone levels. Competitive protein binding assays and radioimmunoassays have now been described [l---16} which have advantages of specificity and sensitivity over previous methods such as protein-bound iodine. Currently, however, such methods are labour intensive and are thus unsuited to clinical biochemistry laboratories with large workloads. We report here singlestage semi-automated radioimmunoassays for total thyroxine (T4) and triiodothyronine (T3), which match the speed, precision, cheapness, and sample capacity of the automated measurement of protein-bound iodine while retaining the advantages of specificity and sensitivity.
26
Methods
A Micromedic automated pipetting station, Model 24002, adapted for duplicate sampling of sera, was used in conjunction with a Micromedic high speed automatic pipette, Model 25006 (Chemlab Instruments Ltd, Ilford, Essex). The volumes dispensed by Micromedic pumps are readily varied, and the combination of 50 ~.tland 1 ml pumps on the Station, and a 100 ~1 pump on the Automatic pipette enabled sera to be sampled in duplicate, and all reagents to be added simultaneously to the assay tubes for both assays. Sera were sampled by the 50 1.11pump, and Reagent 1 (see Table I) was dispensed from the 1 ml pump, both pumps being adjusted to the appropriate volume. Reagent 2, containing the antiserum, was dispensed from the 100 ~1 pump on the Automatic pipette acting in conjunction with the Pipetting station. All reagents could be added to 430 assay tubes in one hour by this method. T4 radioimmunoassay Thyroxine antiserum was raised in rabbits as described by Ratcliffe and co-workers [7], and was used at a final dilution of 1 : 3000. [’ ” I] T4 of specific activity 20.--50 #&‘~g was from the Radiochemical Centre, Amersham. The sodium salt pent~ydmte of thyroxine (Sigma Chemical Co. Ltd) was used as standard. &Anilinonaphthalene-l-sulphonic acid (ANS), obtained from Sigma, was used to inhibit the binding of thyroid hormones to serum binding proteins in both T3 and T4 assays. The separating agent was polyethylene glycol 6000, (PEG), obtained from British Drug Houses and used at a final concentration of 13%. T3 rad~oimmu~oassay
Triiodothyronine antiserum was raised in rabbits as described by Ratcliffe and co-workers [16], and was used at a final dilution of 1 : 2500. ,[I *’ I] T3 (Abbott Laboratories) of specific activity 40-70 pCi/pg or specific activity of up to 600 ,uCi/iug was used. Triiodothyronine (free acid, from Sigma) was used as standard. A second antibody separation method using donkey anti-rabbit serum (RD17 from Burroughs Wellcome) at a final dilution of 1 : 96 and carrier non-immune rabbit serum at a final dilution of 1 : 2000 gave maximum precipitation of the bound fraction. Standards
and quality controls
Standards for each assay were prepared in bulk by adding weighed amounts of T4 and T3 to T4-free or T3-free serum prepared from pooled serum by charcoal treatment [ 121. Commercial quality control sera, or pooled human sera were included at the beginning of each assay. In addition a pooled serum from euthyroid patients was included at regular intervals throughout each assay to monitor assay drift. All standard sera and quality control sera were stored as aliquots at -20°C. Assay procedures
Assay constituents for both T4 and T3 radioimmunoassays Table I. Assay procedures are summarised in Table II.
are shown in
27
TABLE
I
ASSAY
CONSTITUENTS
FOR
SEMI-AUTOMATED
T4
Serum
(standard
Reagent and
1 (Tracer, separating
or sample) ANS
RADIOIMMUNOASSAYS
T3
Radioimmunoassay
50 !.I1 850
~1 of
Tracer
2 (Antiserum)
T3
20 /ll T4
~1 of
(500
pg)
Tracer
ANS
(320
pg)
ANS
PEG
(106
mg)
Donkey
in diluent
buffer
buffer
8.6
pH
bovine Reagent
AND
Radioimmunoassay
680
agent)
T4
100
(0.05
M barbitone
containing albumin)
~1 of T4
antiserum
(1
:
375)
volume
of incubate
800
buffer
buffer
8.6
pH
100
serum (0.05
containing albumin)
11 of T3
antiserum rabbit
:
(1
82)
M barbitone
serum
1.00
~1
pg)
in diluent
Non-immune
buffer
50 pg)
anti-rabbit
in diluent Total
(
(100
bovine
serum
in diluent
0.1%
T3
0.1% (1
:
serum
250)
(1
:
200)
buffer
ml
Analysis of results 1. Verification of the standard curve For both assays, the plot of the free fraction divided by the bound fraction against antigen concentration [17,18] (the F/B plot) gave standard curves which were close to linear, shown in Figs 1 and 2. Little distortion was therefore involved if each standard curve was considered as a series of straight lines connecting adjacent points, and this linear interpolation method was used to derive results automatically using a programmable calculator. However, before results were calculated, the standard curve was submitted to automated verification in an attempt to exclude any unsatisfactory standard points, that is, standard points which lay some way from their expected position. After input of the series of standard values and the counts corresponding to each, the F/B fraction was stored for each standard point. The gradient of the line through each pair of standard points was usually close to the gradient between the neighbouring pair. More than 20% difference in gradient was considered unsatisfactory. This arbitrary cut-off imposed tight limits on each stanTABLE
II
ASSAY
PROCEDURES
FOR
SEMI-AUTOMATED
T4
AND
T3
Radioimmunoassay
Incubation
overnight
Separation
Centrifugation
at 4’C
minutes
or 3 h at 37’C
at 1000
Precipitate seconds
RADIOIMMUNOASSAYS
T3
at 4’C
Supernatant Counting
T4
for
30
Radioimmunoassay
overnight
at 4’C
as for
assay
T4
X g.
aspirated counted
for
on Wallac
30-60
Autogamma
Counter
Precipitate (T3 30-60 600
counted
tracer
seconds fiCi/ug)
analysis
and
By paper
of results
Counter Model
from
Autogamma
Hewlett
Packard
10 programmable
calculator Model
tape
fed into
via Hewlett
9863A
paper
Packard tape
reader
as for
T4
(T3
assay
2-4
minutes
pCi/ug) tracer
on Wallac
Counter) Calculation
for
of 40-70
or of
Autogamma
28
dard point, usually equivalent to less than 1% difference in the bound fraction. If a bad standard point was identified, the calculator derived a new point from the equation of the regression line around that section of the standard curve. In addition, standard points were examined by calculating a weighted difference between duplicates, discussed below. If significant disagreement occurred between duplicates, that standard point was rejected and a new point was derived by linear interpolation between the adjacent standard points. 2. Calculation of results Random errors were detected from the difference between the observed counts of radioimmunoassay duplicates, weighted to allow for counting error. Weighted difference
(First count - Second count) A =J(First count + Second count)
At least 95% of all duplicates for both assays usually fell within a A value of modulus 3.5. A A value of greater than modulus 3.5 was taken to indicate random error in duplicates. If this occurred in the standard curve, the standard point was rejected; in the assay run, the sample was repeated. If the A value was within acceptable limits for a duplicate pair, the result was calculated automatically from the mean count by linear interpolation using the F/B plot of the smoothed standard curve derived by the calculator. Results were printed using the Hewlett Packard calculator’s alpha printer facility. 3. Verification of results The following methods were used to verify results. i. Calculation of the midpoint of the standard curve, that is, the hormone concentration corresponding to half the difference between antibody and diluent blank counts. ii. Examination of the results of quality control samples run early in the assay. iii. Examination of the results of drift control samples placed every fourteen samples throughout the assay. iv. Plotting a histogram of the A values obtained during the assay run. The incidence of random error was indicated by the number of outliers (A values greater than modulus 3.5). The relative level of intrinsic error was revealed by the spread of A values. The presence of systematic error in the handling of duplicate assay tubes was shown by a shift in the mean A value away from zero. Results Standard curves and comparison with manual methods Standard curves for T4 and T3 assays obtained by the semi-automated method (Figs 1 and 2) closely resembled those by manual methods [16]. T4 results, y, obtained by the semi-automated method gave good correlation (r = 0.96) with those by the manual method, x(y = 1.04 x -.- 0.01). Similarly, T3 results, y, by the semi-automated method correlated well with those obtained manually, x(r = 0.95; y = 0.85 x + 0.10). The lower absolute T3 levels
29
T4 Fig.
CONCENTRATION
1. Standard
curve
(n moler’l~tre)
for the
T4
74 CONCENTRATION
(n dcr/l,tre>
radioimmunoassay
obtained by the semi-automated method resulted from the use of serum standards. In the manual method, standards were prepared from a stock solution of T3 in diluent buffer, with T3-free serum added immediately before assay. Several batches of serum standards assayed by the manual method showed a discrepancy from their expected value corresponding to the shift in absolute T3 values observed above. Precision and quality control Fig. 3 shows a quality control chart for the T4 assay for 16 consecutive assays over an eight week period. Over this period the mean between-batch coefficient of variation for all quality control samples was approximately 8% in both T4 and T3 assays. Fig. 3 also shows the standard curve midpoint for the T4 assay. In general this remained approximately constant, but shows a change at Assay 5, probably due to the introduction of a new batch of T4-free serum. Drift occurred in both T3 and T4 assays. The mean drift in the T3 assay is shown in Fig. 4. Drift was not eliminated by keeping both reagents on ice throughout the assay, nor by keeping samples capped until immediately before sampling. When drift was less than lO%, which usually occurred when assay
00.2
1.0
13 CONCENTRATION Fig.
2. Standard
curve
39
10
30
(n moks/litre> for the
T3
radioimmunoassay.
0
5
10
13 CONCENTRATION
15
20
Cnmoledlitrs)
25
30
130
.
High pool 125. MIddIe
pool
120 x _ z 3 P
Curve
modpant
/
115. 110
0 6
105
G
100.
ii 95 4
7
10
Assay
Fig. 3. Quality
control
, 16
13
I, 0
25
Number
50 Number
75 of
100
125
150
sampler
chart for the T4 radioimmunoassay.
Fig. 4. Drift in the T3 radioimmunoassay. the mean result (? S.E.M.) for successive of the first drift control result.
Drift results were collected drift control results through
over ten successive the assay is plotted
assay runs, and as a percentage
runs involved less than a hundred samples, its effect on results was ignored. However, when drift was more than lo%, results were corrected accordingly. Fig. 5 shows histograms of A values for a semi-automated T4 assay and for a manual assay. For the semi-automated assay (Fig. 5b), the distribution of values was close to that expected from counting error alone. The few outliers indicated that random errors were uncommon and the mean value close to zero suggested that systematic error was absent. For the manual assay (Fig. 5a) there was a higher incidence of random error, the mean A was not zero indicating that systematic error was present, and the spread of A values was wider than for the semi-automated assay, suggesting that intrinsic error was larger.
a) MANUAL ‘50
b)SEMI-AUTOMATED
ASSAY
h4eana:-0~9 SD.A:2.2
)5l
h&n
Fig. 5. Histograms of the weighted semi-automated T4 assay.
d
ASW
: -0.1
difference
between
duplicates.
A. (a) for a manual &%a~,
and (b) for a
31
Discussion Radioimmunoassays for total T4 and T3 have rapidly established their place in the evaluation of thyroid function in vitro [19]. Automation of these assays is essential in dealing with large numbers of samples rapidly, precisely, and cheaply since current manual methods are demanding in analyst time and skill. Full automation of radioimmunoassays has in general proved difficult [20], and semi-automated systems have been described for few assays of clinical importance (e.g. insulin, human growth hormone, human placental lactogen) [21,22]. The choice of a suitable separating agent is a critical feature of a semiautomated assay. The separating agent, PEG, in the T4 assay has significant advantages over other methods [16] . Thus, it is convenient, gives apparently complete separation of bound and free fractions and is cheap. For the semiautomated T3 assay, the second antibody method has advantages over other separating agents such as Charcoal, Dowex resin and PEG [ 161. The incorporation of all assay constituents and separating agents in only two assay reagents allows these assays to be set up in one operation, significantly reducing analyst time. In addition to the automation of pipetting operations it is important to simplify labour intensive operations such as the calculation of results. The linear interpolation method used here is simple and has the advantage that no assumptions are made about the shape of the standard curve, thus avoiding the distortions which occur when curve-fitting procedures such as the logit-log transform [23] are applied to these assays. The method is optimised by choosing standard curves which are close to linear (the F/B plots) and by identifying outlying standard points. Random errors in standards and samples are identified by the calculation of the weighted difference, A, between the observed counts of duplicate tubes, and rejection of duplicates in which the A value exceeds certain limits (which depend on the levels of intrinsic and systematic error in the particular assay). A also allows recognition of the contributions of different types of error to overall assay precision, and enables easy identification of methods giving best assay precision. This simple technique has not previously been described, and is generally applicable. Automated calculation of results reduces clerical errors and analyst time. For the T4 radioimmunoassay an assay run of fifty samples (with a total counting time of two hours) can be completed and results made available within the working day. Alternatively, with a workload of 500 samples weekly, one junior technician for each assay can easily produce results the following day. The single stage semi-automated assays for T4 and T3 described here are as sensitive and specific as manual assays and offer the same diagnostic accuracy. Only small serum volumes are required, so that the assays are well suited to paediatric practice. The costs of the assays are comparable with the measurement of PBI. These assays therefore have significant advantages, summarised in Table III, over previous methods for the assessment of thyroid status in the diagnostic laboratory.
32
TABLE
I11
COMPARISON
OF
DIFFERENT
METHODS
FOR
MEASURING
CIRCULATING
THYROID
HOR-
MONES Arrows
are used
to indicate
relative
amounts.
PBI
T4
by
CPB
T4
by
semi-
T3
automated
RI.4
Specificity
+
+
+
Sensitivity
+
+
+
r
4
I
T
1
+
+
Sample
size
t
Sample
capacity
I
Cost
1
i
Diagnostic
semiRI.4
1
+
accuracy
by
automated
A fuller description of the programs written calculator is available on application to G.S.C.
for the Hewlett
Packard
Acknowledgements We acknowledge
the excellent
technical
assistance
of Mrs D. Cooper.
References I
B.E.P.
2
R.P.
3
T.
Murphy Ekins,
Mitsuma,
(1972) I.J.
J.
Chopra,
R.P.
J. Clin.
and
S. Ellis,
L.
Shenkman
Colucci.
J. Clin.
Endocrinol. and
J. Dockalova.
Ratcliffe,
(1974) 8
Pattee,
C. Comette
Larsen.
W.A.
C.J.
Williams
Endocrinol. Clin.
Metab.,
Biochem..
and
24 (1964)
2 (1969)
C.S.
7 87
253
Hollander.
Biochem.
Biophys.
Res.
Commun..
46
2107
C. Beckers, P.R.
and
E.S.
J.G.
Metab..
34 (1972)
M. Thalasso, D. Sipula
Ratcliffe,
J. Nut.
and
A.D.
F.M.
938 Med.,
Will,
McBride,
14 (1973)
J. Clin.
W.A.
317
Endocrinol.
Harland
and
Metab., T.W.
37 (1973)
177
Clin.
Endocrinol..
Randall,
3
481 Ekins.
B.L.
Brown,
S.
Ellis
and
W.S.
Reith.
Abstract
126.
6th
Int.
Conf.,
Vienna,
Austria,
1970,
138 9
H. Gharib,
W.E.
Mayberry
10
T. Mitsuma.
N.
11
I.J.
D.H.
Chopra,
12
P.R.
13
J. Lieblich
Larsen.
14
C.J.
Nihei,
Eastman.
J.M. W.R.
and
Invest.,
R.D.
R.J.
Ryan,
J. Clin.
Gershengorn
Solomon
J. Clin. and
and
M.C.
G.N.
and Beall,
51 (1972)
Utiger,
J. Clin.
Corcoran,
Endocrinol.
C.S.
Metab.,
Hollander.
J. Clin.
J. Clin.
Invest.,
31 (1970) Invest..
50 (1971)
709
50 (1971)
26
2033
1939
Invest.,
A. Jequier,
51 (1972) R.P.
157
Ekins
and
E.S.
Williams,
Clin.
Sci.
Mol.
Med.,
45 (1973)
251 15
C.B.
Sekadde.
16
W.A.
Ratcliffe.
17
R.P.
Ekins,
(eds).
Slaunwhite
G.S. G.B.
Newman
Radioisotopes
Technical
in
Information,
18
R.P.
Ekins,
Br.
19
C.W.
Burke
and
20
H.A.
Butler.
21 22
Knight,
assay
Methods,
A.T.
Letchworth,
Commonw., 23
Methods,
R.A.
D. Rodbard
C.
Caldwell Churchill R.J.
78 (1971) and
J.E.
In
and
D.
Vitro
BuII..
Marchant
Churchill T.A.
Boardman,
1968.
19 (1973)
CIin.
Biochem.,
R.L.
Hayes,
U.S.
1016 11 (1974) E.A.
Atomic
224
Goswitz
Energy
and
B.E.P.
Commission.
Acta
of
p. 59
30 (1974)
and
A.
Livingstone,
Audhya, C.
B&tow,
in Kirkham,
Edinburgh,
in Kirkham,
Edinburgh,
93
Verdin,
1971. J.
1971,
K.E.
and
K.E.
and
Hunter,
W.M.
Endocrinol.
(eds),
p. 635
Hunter,
W.M.
(eds),
Radioimmuno-
P. 598
Landon
and
T.
Chard,
J.
Obstet.
Gynaecol.
542 LewaId.
Murphy
Division
3
Med.
Livingstone,
in
Studies,
Tennessee,
Br.
Chem..
Ann.
O’Riordan,
30 (1974)
Eastman,
CIin.
Ratcliffe.
J.L.H.
Ridge,
Bagshawe,
Radioimmunoassay
T. Aceto,
J.G.
Medicine:
Bull.,
C.J.
and and
Oak
Med.
K.D.
and
ChaIIand
(Copenhagen),
Suppl.
No.
147.
(1970)
79
Br.