| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
Divisions of Endocrinology and
2
Informatics, National Institute for Biological Standards and Control, Blanche Lane, South Mimms, Potters Bar, Herts EN6 3QG, UK.
a Author for correspondence. Fax 44-1707-646730; e-mail brafferty{at}nibsc.ac.uk
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Coded samples were provided to the 33 laboratories in the study, and participants were asked to perform TSH assays currently in use in their laboratories. Twenty-eight laboratories contributed 93 immunoassays in 41 different method-laboratory combinations, and an additional 5 laboratories contributed bioassay data. All data were analyzed centrally at the National Institute for Biological Standards and Control.
Results: The results obtained in different laboratories and with different assay systems revealed significant variability between estimates of rTSH relative to the IRP. These ranged from 5.51 mIU (95% limits, 3.95–7.67 mIU) per ampoule by RIA to 7.15 mIU (95% limits, 6.7–7.63 mIU) per ampoule by immunofluorometric assay. However, the results showed that the assignment of a value of 6.70 mIU per ampoule of 94/674 would give reasonable continuity with the IRP in many assay systems.
Conclusions: The preparation was established as the First WHO Reference Reagent for TSH, human, recombinant, to provide a means of validating assay performance and to maintain continuity with the IRP without compromising clinical data.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To address this point, a candidate preparation of rhTSH was obtained and a collaborative study organized to establish whether it could be calibrated adequately by immunoassay in terms of the existing Second IRP of pituitary TSH (8). The aims of the study were: (a) to compare rhTSH and pituitary TSH in a variety of immunoassay systems; (b) to assess the suitability of the candidate preparation of rhTSH to serve as a standard for the calibration of diagnostic immunoassays; (c) to assess the stability of the candidate preparation of rhTSH after accelerated thermal degradation; and (d) to confirm the bioactivity of rhTSH after ampouling and lyophilization procedures.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
. In addition, participants were asked to include their in-house
standards in the assays. For immunoassay, each participant received a
basic set of ampoules containing the Second IRP, rhTSH (94/674), and
pituitary TSH preparation 81/502, and in addition they received
accelerated degradation samples selected on the basis of assay
capacity, sample availability, and a design giving information across
as wide a range of assay systems as feasible. For bioassay, the Second
IRP was replaced with the pituitary-derived bioassay standard, 84/703
[National Institute for Biological Standards and Control (NIBSC) Std
B]. The Second IRP and pituitary TSH preparations 81/502 and 84/703
have been described (1). The rhTSH 94/674 was ampouled at
NIBSC, following procedures described by WHO (9). Highly
purified (>95% by sodium dodecyl sulfate-polyacrylamide gel
electrophoresis) rhTSH, expressed in CHO cells (2), was
kindly donated to WHO by the Genzyme Corporation, Framingham, MA. The
preparation was received as a lyophilized white powder (nominal 2 mg)
that, after dilution in 2 L of a solution containing 1 g/L human serum
albumin and 5 g/L lactose, was distributed into ampoules as 1.0-mL
aliquots. Each ampoule (coded 94/674) has a nominal content of 1 µg
of TSH, 1 mg of human plasma albumin, and 5 mg of lactose.
|
participants in the study
Thirty-three laboratories in 11 countries took part in the study
and are listed alphabetically, by country, below. Throughout this
report, each participating laboratory is referred to by a code number,
which was assigned randomly and which does not reflect the order of
listing:
design of the study and assay methods contributed
Participants were asked to perform the TSH immunoassay(s)
currently in use in their laboratory and to carry out at least two
independent assays (i.e., using freshly reconstituted ampoules) of each
assay type, each assay to include a set of ampouled preparations
selected with regard to assay system capacity and to provide maximum
information for the study. Each set of preparations included a
coded duplicate preparation to provide an independent assessment of
assay accuracy and precision. All preparations and any local standards
were to be included at several doses to provide information on
linearity and parallelism of the dose–response relationship. Selected
participants were invited to provide bioassay data, following as far as
practical, the broad design suggested for immunoassays. All raw data
were reported to NIBSC for analysis.
statistical analysis
As far as possible for this study, an "assay" was defined as
an independent test for each preparation beginning with freshly
reconstituted ampoules.
In all immunoassays, the dose–response relationships could be reasonably described using a four-parameter logistic function. The fitted asymptotic values of the logistic function were used to transform the reported responses to logits. Logit responses were analyzed using weighted linear regression and an in-house program (10), and the linearity and parallelism of the log dose-logit response lines were assessed. In assays showing apparently significant (P <0.05) deviations from parallelism, the differences in slopes between different preparations were compared to the differences in slopes between the coded duplicate preparations; differences between different preparations that were less than the observed differences between the coded duplicate preparations were not considered to reflect consistent nonparallelism. Relative potencies were determined as the displacement of fitted parallel log dose-logit response lines. For some bioassays, there were insufficient data for satisfactory fitting of a four-parameter logistic relationship; for these assays, log dose-response data were analyzed using the methods for parallel line assays. Except as otherwise noted, there was no exclusion of data.
Groups of potency estimates were compared using unweighted analysis of variance of log potency estimates. Potency estimates were combined as unweighted geometric means, and 95% fiducial limits for the combined estimate were calculated using the variance of the logs of the estimates combined. In some cases, the variation between estimates was expressed as the geometric CV {gCV = [exp (standard deviation of log potency estimates) - 1] multiplied by 100}.
Estimates of the relative activity (to ampoules of the same preparation stored continuously at -20 °C) remaining in the ampoules of the TSH preparations after storage at increased temperatures were used to fit an Arrhenius equation relating degradation rate to absolute temperature assuming first-order decay (11) and, hence were used to predict the degradation rate of the preparations when stored at -20 °C.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
, 2
, and 3
) may be obtained from the authors upon request.
|
|
|
variability of immunoassay estimates
Estimates of the activity of the coded duplicates of 94/674
relative to one another provided an independent measure of assay
variability. Individual estimates were between 0.64 and 1.34, and the
geometric mean (95% limits) of all individual estimates was 0.98
(0.96–1.01), in good agreement with the value of 1 expected for
identical preparations.
Deviations of the observed individual values from the expected relative activity of 1.0 gave a pooled (over all immunoassay methods and laboratories) within-assay mean square for log potency estimates of ~0.01, corresponding to an ~95% interval of 82–122% for comparison of two identical materials within an assay in these systems. The between-assay mean square (also pooled over all methods and laboratories) was smaller than this, suggesting that there may be a tendency for greater similarity of estimates for the same preparations between assays than for independent tests of the same preparations within the same assay. It was noted that the within-and between-assay variability, pooled over the broad groupings of methods, was 2- to 10-fold larger for ELISAs than for the other methods.
comparison of rTSH 94/674 AND THE PITUITARYhTSH IRP
The laboratory geometric mean estimates for each of the duplicate
ampoules of 94/674, expressed as mIU of the IRP per ampoule, are shown
in Fig. 1
. The between-laboratory mean square for these estimates was
approximately three- to fourfold larger than that obtained for the
direct comparison of the identical preparations, reflecting the
markedly different estimates obtained in the different laboratories.
Because of this between-laboratory variability and because not all
laboratories contributed the same number of assays, a single geometric
mean of all assay estimates for the two ampoules of 94/674 was obtained
for each method in each laboratory, and these laboratory and method
means were used to calculate the combined estimates. Estimates by RIA
(mean, 5.51 mIU per ampoule; 95% limits, 3.95–7.67 mIU per ampoule;
n = 3) were notably smaller than estimates by the other broadly
defined methods, and this difference would be more marked if the
estimates from laboratory 17 were excluded. Estimates by ICLMA were
also generally smaller than estimates by other methods if the ICLMA of
laboratory 14 was excluded (mean, 5.55 mIU per ampoule; 95% limits,
5.20–5.92 mIU per ampoule; n = 4). Estimates by IFMA (mean, 7.15
mIU per ampoule; 95% limits, 6.7–7.63 mIU per ampoule; n = 9)
were generally larger than estimates by the other methods. Estimates by
ELISA (mean, 6.50 mIU per ampoule; 95% limits, 5.69–7.43 mIU per
ampoule; n = 8) and IRMA (mean, 6.77 mIU per ampoule; 95% limits,
6.31–7.26 mIU per ampoule; n = 13) spanned the ranges of
estimates by the various other methods, with marked differences between
estimates from different laboratories within these methods. The mean
estimate over all methods, excluding data as described previously, was
6.59 mIU per ampoule (95% limits, 6.28–6.92 mIU per ampoule; n =
38).
A limited number of bioassays were carried out as summarized in Table 3
. Laboratory 31 also reported an estimate of 2.4 mIU for an ampoule of
94/674 stored at 45 °C for 400 days.
|
comparison of the various local standards with the irp
Geometric mean estimates of the relative activity of the various
local standards expressed as IU of the IRP equivalent in activity to a
nominal IU of the local standard are shown in Table 4
. These estimates were all ~1.0, as might be expected given
the length of time for which the IRP has been in use, although there
were significant (P <0.05) differences between laboratories
when the local standards were compared directly with the IRP.
Individual estimates ranged from 0.58 to 1.31 IU per nominal IU, with a
mean of individual estimates of 0.97 IU (95% limits, 0.89–1.05 IU)
per nominal IU.
|
comparison of pituitary hTSH 81/502 AND THE IRP
Laboratory geometric mean estimates of 81/502 expressed as mIU of
the IRP per ampoule are shown in Fig. 2
. The between-laboratory mean square, pooled over all
laboratories and methods, for these estimates was approximately twofold
greater than that for the coded duplicate preparations, suggesting that
there were significant differences between laboratories. For example,
the RIAs in use in laboratories 07 and 28 gave estimates that were
notably smaller than the estimate given by the RIA in use in laboratory
17, a relationship similar to that seen for comparison of 94/674 with
the IRP. There was no clear difference between the broad categories of
assay, although estimates by IFMA tended to be slightly larger (mean,
7.60 mIU per ampoule; 95% limits, 7.02–8.22 mIU per ampoule) than
estimates by the other methods. The geometric mean of all laboratory
and method geometric means was 7.30 mIU per ampoule (7.00–7.60 mIU per
ampoule; n = 37), omitting data as described previously.
comparison of rTSH 94/674 AND PITUITARY hTSH
81/502
Laboratory geometric mean estimates for each of the duplicate
ampoules of 94/674, expressed as milliunits of 81/502 using the value
of 7.75 mIU per ampoule previously assigned to 81/502, are shown in
Fig. 3
. The between-laboratory mean square, pooled over all
laboratories and methods, for these estimates was approximately
fourfold larger than that obtained for the comparison of the coded
duplicate preparations. However, the between-laboratory mean squares
for IFMA or RIA assays considered separately were similar to those
obtained for the coded duplicate preparations, although the geometric
mean estimates by these two methods differed significantly, being 7.29
mIU (6.93–7.67 mIU) per ampoule by IFMA or 6.38 mIU (6.17–6.60 mIU)
per ampoule by RIA. In contrast, the between-laboratory mean square for
estimates by ICLMA was approximately 10-fold larger than that of the
coded duplicate preparations, and these methods gave both the smallest
and largest estimates obtained for this comparison, from 5 to 10 mIU
per ampoule. Both for this comparison and for the comparison of 94/674
with the IRP, the ICLMA in use in laboratory 14 gave substantially
larger estimates than those given by the other ICLMAs.
stability of rTSH 94/674 AND PITUITARY TSH
81/502
Estimates of the activity of the sample of 94/674 that had been
stored for 215 days at 20 °C relative to the samples stored
continuously at -20 °C showed no detectable loss of activity. The
geometric mean of laboratory geometric mean estimates for the sample
stored at 20 °C was 0.99 ampoules (95% limits, 0.95–1.02 ampoules;
n = 20) of 20 °C sample having equivalent activity to one
ampoule of 94/674. Samples stored for 215 days at 37 or 45 °C showed
substantial losses of activity in nearly all assay systems. Predicted
losses of activity for samples stored continuously at -20 °C vary
depending on results used and in some cases, show poor agreement with
the theoretical model. However, the apparent immunological stability of
the samples stored at 20 °C, and the predicted stability when this
information is combined with the apparently moderate losses of activity
of samples stored at 37 or 45 °C indicate that 94/674 under the
usual conditions of storage at -20 °C is likely to be sufficiently
stable to serve as a standard.
The samples of 81/502 had been stored at relatively high temperatures for a considerable time, ~15 years. The immunologically detectable loss of activity ranged from ~15% to undetectable for the samples stored at 20 °C (mean remaining activity, 95%; 95% limits, 90–99%) and from 35% to 10% for the samples stored at 37 °C (mean remaining activity, 76%; 95% limits, 71–83%), indicating that, as for 94/674, the measured loss of activity is dependent on the assay system used for detection. (This suggests that the degradative changes induced in the molecule variously affect antibody recognition). Using the remaining activity at 20 and 37 °C of either 95% and 65%, as obtained by the system seeing the largest loss of activity, or 96% and 85%, as obtained by the system seeing the least loss of activity, gives predicted immunologically detectable losses of activity of <0.01% per year for the samples stored continuously at -20 °C.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Comparison of the ampouled preparation of rhTSH (94/674) with the IRP
(80/558) showed only limited evidence of nonparallelism in the 41
method/laboratory combinations included in the study. However, there
was substantial variability between laboratory estimates of 94/674
relative to the IRP. The extent to which this was system-based is not
clear; estimates by RIA were smaller than estimates from the other
categories of assays, whereas estimates by IFMA were generally larger.
The lower immunopotency of rhTSH by RIA compared with two-site
noncompetitive systems has been described previously (7) and
may be attributable to differential recognition by the antibody of
labeled and unlabeled ligand from different sources. Other factors,
such as surface adsorption or matrix effects, may influence assay
performance by affecting recovery of material from the ampoules.
However, most participants performed the initial dilutions in a bovine
serum albumin- or serum-based diluent, and in addition, protein
protection was offered by human serum albumin as part of the ampoule
content formulation. Examination of the data from the largest category
of assays (IRMA; n = 15) showed a group of four laboratories with
a smaller mean estimate than the remaining members. However, there was
no obvious correlation of initial dilution volume and primary and
secondary diluent composition with assay estimate. During the course of
the study, reconstitution of ampoule contents in water was shown to
cause problems in several laboratories; one laboratory reported
nonlinearity of responses (data not analyzed), and estimates from two
others were substantially smaller than those obtained in the majority
of other assay systems. The incorporation of two different preparations
of pituitary TSH (IRP and 81/502) provided evidence that at least some
assay systems may distinguish between rhTSH and pituitary TSH because
the interlaboratory variability for comparison of either pituitary
preparation and the rhTSH was twofold greater than that for comparison
of the two pituitary preparations. It may be, for example, that the
large estimates obtained by laboratory 14, which used ICLMA, or
laboratory 18, which used ELISA, for 94/674 relative to either the IRP
or 81/502 reflect a relatively greater selectivity of these systems for
the rhTSH than for the pituitary TSH, whereas the reverse is true for
many of the ICLMAs and ELISAs, which gave generally lower estimates for
these comparisons (Figs. 1
and 3
).
The single geometric mean of all assay estimates for the coded duplicates of 94/674 relative to the IRP, excluding estimates in terms of an in-house standard or with water as primary diluent, was 6.59 mIU (6.28–6.92 mIU; n = 38) per ampoule. The mean estimate by RIA was smaller than estimates from the other categories of assays. RIA has mostly been superseded by two-site noncompetitive assays (only 3 laboratories of 28 offered RIA) and is likely to be replaced completely in the future. Removal of the RIA data gave a geometric mean of 6.70 mIU (6.38–7.02 mIU; n = 35) per ampoule.
The single geometric mean of all estimates of 94/674 relative to the
pituitary preparation 81/502 was 6.91 mIU (6.60–7.23 mIU; n = 41)
per ampoule. The potency (7.75 mIU per ampoule) of 81/502 had been
assigned previously in the international study that established the
Second IRP 80/558 (1); therefore, inclusion of 81/502,
together with samples of it stored continuously at increased
temperatures for 15 years, also allowed the possible loss of
immunoreactivity of both preparations to be monitored over this
extended period. Although there was little evidence of lack of
stability, the mean estimate of potency relative to the IRP in the
current study was 7.30 mIU per ampoule (7.00–7.60 mIU per ampoule;
n = 37). It is possible that this apparent decline in potency is a
reflection of the optimization of assays in the intervening 15 years
toward the pituitary preparation contained in the second IRP, or
although less likely, because the potency was initially assigned in
terms of both immunoassay and bioassay. Because there are significant
(P <0.05) interlaboratory differences in estimates, the
possibility cannot be excluded that inclusion or exclusion of
particular laboratories may have affected the overall mean estimate,
both in this and the previous study. For example, IRMAs, excluding
laboratories 16, 17, 22, and 28, gave a mean of 7.76 mIU per ampoule,
and estimates from several laboratories were consistent with the value
of 7.75 mIU per ampoule (Fig. 2
).
Although it appeared that some assay systems distinguished between pituitary and recombinant TSH, the results of the study showed that assignment of a value of 0.0067 IU (6.70 mIU) per ampoule of rhTSH (94/674) would give reasonable continuity with the Second IRP for TSH in many assay systems. It appeared to be sufficiently stable to serve as a reference reagent because the predicted loss of immunoreactivity when stored continuously at -20 °C was <0.01% per year. The preparation was also shown to have appropriate biological activity. At its 47th meeting, therefore, the Expert Committee on Biological Standardization of WHO established the preparation in ampoules coded 94/674 as the First WHO Reference Reagent for TSH, human recombinant, with an assigned value of 0.0067 IU per ampoule (19).
It is not intended that the recombinant preparation should replace the current IRP because there are sufficient stocks of the IRP to last for many years. Rather, it is envisaged that, as companies gradually change over to the use of recombinant DNA-derived materials as kit calibrators, a reference reagent that has been assessed in a variety of assay systems relative to the IRP [and which may have similarities to forms of circulating TSH seen in some clinical conditions (20)] will provide a means of validating assay performance of those systems that may distinguish between the native and recombinant materials. By characterizing the system in this manner, the assigned value would maintain continuity with the IRP and should not compromise the clinical data.
|
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  K. Zophel, G. Wunderlich, and J. Kotzerke Should We Really Determine a Reference Population for the Definition of Thyroid-Stimulating Hormone Reference Interval? Clin. Chem., February 1, 2006; 52(2): 329 - 330. [Full Text] [PDF]
|
|
|
|
|
|
S. Donadio, A. Pascual, J. H.H. Thijssen, and C. Ronin Feasibility Study of New Calibrators for Thyroid-Stimulating Hormone (TSH) Immunoprocedures Based on Remodeling of Recombinant TSH to Mimic Glycoforms Circulating in Patients with Thyroid Disorders Clin. Chem., February 1, 2006; 52(2): 286 - 297. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
