| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Professor of Clinical Biochemistry, Emeritus, University of Surrey, Guildford GU2 7XH, United Kingdom.
Address for correspondence: Oriel House, Derby Road, Haslemere GU27 1BP, United Kingdom. Fax 44-1428-652-893; e-mail vincentmarks{at}bigfoot.com.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Method: Blood was collected from 10 donors with illnesses known to be associated with rheumatoid factor. Immunoassays for 74 analytes were performed in 66 clinical laboratories. Each sample was measured in duplicate, and again in duplicate after the addition of a proprietary heterophil blocking reagent, with the laboratory’s routinely used reagents and equipment. Reagents were typically supplied by the manufacturers of the closed analytical systems. Both competitive and sandwich immunoassays were investigated.
Results: Overall 
8.7% of the 3445 results were considered to be "false positive", many of them seriously so. Twenty-one percent of the erroneously high results (1.8% of all results) were potentially misleading and were corrected by blocking reagent, although 49% of such seriously high results (4.2% of all results) were not. A further 39% of the false-positive results (2.6% of all results) would not necessarily have appeared likely to produce adverse clinical consequences but were substantially lowered by the addition of the blocking reagent.
Conclusions: Random errors can occur with all types of immunoassays tested and can be difficult to identify even when repeated in another laboratory. Clinicians need to be aware of these limitations.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The presence in a patient’s serum of anti-animal antibodies, more particularly those directed against mouse (monoclonal) immunoglobulins, is especially important because, for various reasons, monoclonal antibodies have largely replaced (polyclonal) antisera. Other well-recognized causes of erroneous immunoassay results are the presence in a patient’s serum of auto- and heterophilic antibodies (6)(7)(8), especially so-called rheumatoid factor (Rf).1 Rf, so named because of its early association with rheumatoid arthritis, has a prevalence in the adult population of up to 10% and is not necessarily associated with overt disease.
Most immunoassay manufacturers allude in their package inserts to difficulties with immunoassays arising from the presence of auto- and heterophilic antibodies, and most include one or more interference blocking agents in their reagent formulations in an attempt to overcome these problems. In the US, to meet Food and Drug Administration requirements manufacturers must draw specific attention to problems arising from the presence of human anti-mouse antibodies (HAMAs) and the like in a patient’s sample. They also invariably issue warnings against undue clinical reliance on the unsupported results of immunoassay procedures or on the validity of decision points (reference values). In practice these warnings may go unheeded, possibly because of the perceived rarity of the interferences or because the warnings do not reach the physician caring for a patient.
The present study was not designed to identify individual laboratories or manufacturers whose products were especially prone to error, but only to determine the frequency and magnitude of the problem as exemplified by results produced in everyday clinical laboratory practice by several European and US laboratories on samples obtained from 10 selected patients.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
All 66 laboratories participating in the study agreed to run the 10 selected patient samples blind on their routine clinical analyzers, and each was responsible for performing HBT pretreatment to generate their "with-treatment" immunoassay results.
Forty-eight laboratories were located in the mainland US, 13 laboratories were located in France, and 1 each was located in Italy, Germany, Austria, The Netherlands, and Costa Rica. Thirty-five of the US laboratories were in hospitals or medical centers, 8 were university clinical laboratories, and 5 were commercial clinical laboratories. Four of the participants were clinical laboratory reagent manufacturers. All but four of the non-US laboratories were hospital laboratories.
samples used for the study
The 10 donors used in the study were recruited by advertisements in local (Californian) newspapers for persons diagnosed as suffering from several illnesses known to be associated with the presence of Rf in their serum but who were otherwise in good health. Each donor signed an informed consent agreement and received appropriate payment for participation in the study. Multiple sessions (n = 1–7; median, 2) of plasmapheresis on different dates were performed on each of the donors to generate adequate volumes (500–700 mL) of plasma for the study.
Before acceptance, each donor was tested for the ability of his or her plasma to produce a falsely increased immunoassay result in at least one immunoassay method. A result was deemed a false positive if it was in the "pathologic range" for the method used and was restored to a value within the reference interval by pretreatment with heterophil-blocking reagent (HBR). The HBR pretreatment used in this study was in the form of a tube (HBT) containing a fixed amount (4 mg) of HBR-3. A set amount of sample was added to each tube and mixed for 60 min before subjecting it to immunoassay.
This study analyzes the results obtained from the plasma samples of the 10 selected donors, brief details of whom are given in Table 1
. All donors were considered to be clinically well apart from the illnesses specified in Table 1
. Routine clinical chemistry analyses, including kidney and liver function tests, were performed on all donors at the time of sample collection and confirmed as being within the appropriate reference intervals. No donor was known to be suffering from neoplastic or endocrine disease or to have suffered a recent myocardial infarction. Each plasma collection was assigned a unique identification number. All samples were analyzed by the organizers for Rf and HAMA content as well as being subjected to "routine screening" for potentially infectious viruses. All donors were negative for HIV antibody, hepatitis C antibody, hepatitis B surface antigen, and a serologic test for syphilis.
|
All participating laboratories received samples from 9 of the donors, and a few laboratories received samples from a 10th donor (labeled as donor 7, but whose details have been mislaid). Forty-two percent of the laboratories assayed all of the samples they received. With few exceptions, participating laboratories used their solitary routine "closed" analytical systems (i.e., a single instrument using reagents supplied exclusively by its manufacturer). A minority of laboratories used two or more analyzers or carried out manual assays with commercially available reagents or both.
All laboratories were asked to perform, as part of a routine analytical run, a quantitative immunoassay on each sample with and without pretreatment with HBT. Each laboratory recorded the unique sample number of the sample sent to them by the organizers as well as the lot numbers of the various reagents with which it was measured. The results obtained for each assay were returned to the organizers for data analysis. None of the laboratories was aware of the results obtained by the other participating laboratories.
Some, although not all manufacturers were asked to provide "reference ranges" for selected analytes using their reagents. These were used to complement reference intervals obtained from the literature and, more particularly, from two reference sources (9)(10).
interpretation of the data
Interpretation of laboratory assay results usually depends on comparison of the numerical result with an appropriate ethnic, sex, and age reference interval and, most importantly, compatibility with the patient’s clinical history, physical examination, and other test results.
In the absence of comparison methods such as mass spectrometry with the use of a stable-isotope internal standard (11), reference values for most polypeptides and proteins are themselves "method dependent". I have therefore, where possible, compared the numerical results obtained by individual laboratories with the suggested reference interval appropriate for the method used, as provided by the manufacturer, or, where these were not available, values obtained from standard reference tables (9)(10). For analytes not usually found in more than infinitesimally low concentrations except in disease, especially those used predominantly or exclusively as cancer markers or markers of myocardial infarction, an upper limit of normality or decision point is often used. Values exceeding the decision point are described as "positive".
For some analytes, such as therapeutic drugs in persons not receiving them and for some hormones, as well as antibodies to infectious agents, no reference value is realistic without additional relevant information on the donor. Results obtained for such analytes are, however, amenable to comparison with those obtained for the same sample in different laboratories using ostensibly the same or similar analytical methods.
This study was designed to determine the frequency at which clinically erroneous false-positive immunoassay results were obtained that in addition were inconsistent with the donor’s physical condition or were substantially altered by HBT treatment. Consequently, although this report is concerned mainly with analyses in which immunoassay results obtained by various laboratories using diverse analytical systems were abnormally high and restored to within the appropriate reference interval by pretreatment with HBT, other issues are also addressed. However, because the study was not designed to detect the relative occurrence of false results with individual assay systems, within a population of healthy or disease-specific individuals, anonymity of both the systems used and laboratories performing the assays has been preserved.
Results (Table 2
) are described as heterophilic false positive (HFP) when the immunoassay result was above the reference interval (or decision point) and inconsistent with the patient’s clinical condition and was restored to within the reference interval by pretreatment with HBT. Abnormally increased results obtained on samples that were inconsistent with the donor’s clinical condition and at variance with results produced by other assay systems but were not themselves restored to within the reference interval by HBT are referred to as false-positive results of uncertain etiology. Immunoassay results that were misleading, although not diagnostic of disease (i.e., they were within the reference interval for the analyte or below the decision point) but nevertheless were reduced by >30% by pretreatment with HBT are described as falsely increased. In a very small number of samples (0.3%), pretreatment with HBT turned a normal into an abnormal result (see Fig. 7
for example). These results are not considered in detail in this communication and may be attributed to interaction between HBT and other components present in the formulation of the immunoassay reagents, e.g., anti-mouse antibodies on the solid phase of the assay.
|
|
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Results were reported on 74 different analytes varying in frequency from 281 analyses for human chorionic gonadotropin (hCG and ß-hCG) measured in 40 laboratories with 15 systems to only 2 results per analyte from a single laboratory (Table 2
). Twenty-one analytes were arbitrarily classified as hormones, 18 as tumor markers, 8 as therapeutic drugs, 5 as cardiac markers, 4 as proteins, 2 as vitamin, and 16 as miscellaneous, mostly antiviral or autoantibodies.
The complete list of analytes tested, together with the number of analytical systems yielding one or more false analytical results, is shown in Table 2
. In total, 3445 analyses were performed: in <5%, and only when the initial result was within the reference interval, were repeat tests with HBT pretreatment not reported. Sixty-five results (1.8%) involving 13 analytes were deemed to be HFP results, and 146 results (4.2%) involving 17 analytes were classified as false positives of uncertain etiology. A further 89 results (2.6%) involving 24 analytes were falsely increased but not above the reference intervals or decision points. In summary, for these selected samples, 6% of analyses gave "false positive" results with potential for incorrect clinical diagnosis.
One or more numerically erroneous results were seen for many analytes, often with more than one system and in more than one laboratory. None of the assays used for the measurement of therapeutic drugs or vitamins, however, yielded numerically erroneous results, although the small number of results for these analytes may have contributed to this.
The distribution of erroneous results was not confined to any individual donor, immunoassay reagent manufacturer, or analytical system. Some donors, especially donor 1, and to a lesser extent donor 2, produced disproportionately large numbers of false-positive results relative to the other eight. In general, however, false-positive results were distributed randomly among laboratories, analytical systems, and donors. In instances where erroneous test results were repeated within the same laboratory on another occasion they were remarkably reproducible. An example is shown in Fig. 1
.
|
The findings overall in this selected group of donors revealed a disconcertingly high number of analytically and clinically erroneous false-positive results, only half of which were corrected by HBT.
examples of erroneous results
Specific examples have been chosen to illustrate the variety of problems encountered.
hCG was the analyte most commonly measured in this group of clinical laboratories. Presumably its use as a diagnostic test for pregnancy in women undergoing radiologic examinations rather than as a tumor marker for choriocarcinoma accounts for its high use, especially in the US.
Shown in Fig. 2
are the results of hCG measurements made with and without HBT pretreatment in 38 laboratories on samples of plasma collected within a 3-month period from donor 1. Fifteen laboratories used the same analytical system supplied by a single manufacturer. Despite large interlaboratory differences in their results for donor 1, all were falsely positive and none was corrected by HBT. Three laboratories repeated their analyses with acceptable, but still clinically erroneous, replication. Laboratory 1, using reagents supplied by the same manufacturer but different instrumentation, showed virtually complete abolition of falsely high hCG results with HBT pretreatment. Fig. 3
shows the effect of HBT on the hCG results produced by a manual RIA in plasmas from nine donors, two of which had elicited a clinically false-positive result and six others that appeared to give incorrect results before HBT pretreatment.
|
|
Measurements of plasma myoglobin showed the highest percentage of HFP results as well as the largest proportion of systems producing them (Table 2
). The results obtained in one laboratory on samples from all nine donors are shown in Fig. 4
. HBT pretreatment substantially reduced the apparent myoglobin content in all samples tested, three of which yielded falsely positive results before treatment. Fig. 5
shows the results produced by 11 laboratories using seven different analytical systems on plasma samples from donor 2. One system consistently produced erroneously high results that were corrected by HBT pretreatment.
|
|
Insulin assays were performed in three laboratories. Two laboratories (MC and DU) used the same assay system but applied it only to samples from donors 1 and 2. The results are shown in Fig. 6
. Donor 1 showed unexpectedly high insulin concentrations in both laboratories, and these were reduced by 30% by HBT pretreatment: results for donor 2 were more appropriate insulin values, but they were, nevertheless, reduced by 50% by the addition of HBT. Which, if either, insulin concentration was correct cannot be determined from these data.
|
Plasma C-peptide was measured in only one laboratory (the third laboratory that measured insulin). The results are shown in Fig. 7
. Although it had no significant effect on plasma samples from donors 1–5, HBT had a profound effect on samples collected from donors 6, 8, 9, and 10, in which it produced an inappropriately low concentration that was inconsistent with both the donor’s clinical status and the results of the insulin assays conducted simultaneously.
The results of the follicle-stimulating hormone (FSH) assays performed on donor 6 in 12 laboratories using 11 different assay systems are shown in Fig. 8
. Three laboratories confirmed the donor’s postmenopausal status, which was consistent with simultaneously high plasma luteinizing hormone (LH; not shown): the other nine laboratories reported inappropriately low LH and FSH. HBT had no perceptible effect on any of these results, and the cause of the erroneously low results obtained is unknown.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Despite the great advantages of immunoassay for measuring proteins, peptides, and smaller molecules at infinitesimal concentrations, often without prior purification, it is neither as specific nor as accurate as was originally presumed. Attention was originally directed toward the specificity of the primary antibody for the analyte in question, and problems on this issue have been largely, although still far from completely, resolved by improvements in antibody production technologies (15)(16)(17)(18).
The role of interfering substances that are present in the patient’s sample but not in the calibrators or quality-control samples continues to haunt the clinical laboratory. In addition to the large number of nonspecific variables, such as minor changes in the ionic strength, hydrogen ion concentration (pH), or osmolality of the final reaction mixture, all of which can be easily tested for and remedied in the development stage of the test, there are other causes of interference that are idiosyncratic and depend on a unique reaction between one or more constituents in the patient’s sample and one or more of the components in the immunoassay reagents (1)(2)(3)(4)(5)(6)(7)(8)(19)(20).
Plasma, once the preferred specimen for analysis, has for convenience largely been replaced by serum. Most commercial immunoassay systems, however, permit the use of either serum or plasma, the material used, for logistic reasons, in the present study.
Possibly the most important of the idiosyncratic interfering substances found in patient samples are those that are either autoantibodies against the analyte itself (e.g., insulin autoantibodies) or heterophilic (including anti-animal) antibodies that react with one or more of the assay reagents. Both types of antibody can produce falsely high or falsely low results.
The correct result for any particular analyte cannot be ascertained in the absence of analyses conducted with a reference method (21), but an approximation to it can be inferred from the collective results of other immunoassay methods and some knowledge of the donor’s clinical status. This report is concerned predominantly with the prevalence of false and misleadingly high results obtained by clinical laboratories using their routine clinical procedures and not with the large variations in absolute concentrations observed among them.
Although differing in detail, all immunoassays require the same key reagents, namely, (a) one or more antibodies raised against one or (in the case of most sandwich assays) two epitopes believed to be specific to the analyte in question; (b) a calibrator in a fluid similar to the patient’s sample to be tested; and (c) a label capable of being measured quantitatively in a clinically relevant range by an appropriate measuring instrument.
RIAs (12) based on the principles of saturation, i.e., reagent-limited, analysis (22) and that included polyclonal antisera were the first type of immunoassay used for measuring hormones and small molecules in biological samples. They have largely, although not entirely, been replaced by immunometric, i.e., reagent excess, methods (23) produced by various manufacturers (24).
The recent biomedical literature includes case reports of misdiagnosis or improper treatments resulting from analytical errors (1)(5)(25)(26)(27)(28)(29). Analytical errors arising from the presence of antibodies to mouse (monoclonal) immunoglobulins in the patient’s plasma or serum have received the most attention but are just one of the many causes of interference in immunoassays. Reproducibility within a laboratory or among laboratories using the same or different analytical systems is no guarantee of the validity or correctness of the results
No consistent pattern emerges from the donor sample, analytes assayed, or analytical system to explain the high incidence of erroneous results observed apart from the fact that the donors were selected for their high probability of having interfering substances in their sera. The donors used in this study, however, were selected from within an otherwise normal or "healthy" population. Samples from a patient population (i.e., people who are clinically sick) may have an even higher potential for immunoassay interference.
The consequences of analytical errors, especially when seemingly confirmed by results from a second, third, or even fourth laboratory using the same, as is commonly the case, or a different methodology can be disastrous for the patient. They may also be comparatively trivial as, for example, in the present series in which no less than 9 of the 11 assay systems tested gave incorrect results (i.e., clinically inconsistent) for FSH and LH in samples from donor 1.
This study confirms the limitations of immunoassays, as with all other assay methodologies, in the routine clinical setting. Although the risk of diagnostic error can be reduced, it cannot be abolished by incorporating broad-spectrum heterophilic blocking agents in the assay system. Consequently, it is only by careful attention to the patient’s clinical history and the results of dynamic function tests, closer and better informed communication between clinicians and the laboratory, and sound judgment that patient catastrophes are going to be avoided.
In the past, immunoassay was often the only analytical method with the requisite specificity and sensitivity to detect and measure many substances of clinical importance in biological fluids. This is no longer the case. Specific, accurate, and precise quantification of most analytes of clinical importance is now possible by a combination of analytical methods. Some of these, such as a combination of HPLC and immunoassay with appropriate internal standards, are available to any well-appointed laboratory. Although too technically demanding for routine use, they can be applied to samples that do not accord with the clinical picture or expectations. Methods such as liquid chromatography–mass spectrometry with or without stable-isotope dilution (11)(30) can be resorted to as comparison methods, especially in forensic and other situations in which absolute identification and accuracy are essential and where erroneous conclusions, based on incorrect data that cannot be independently validated, may have devastating consequences.
A take-home message from the present study is that laboratories, and the laboratory reagent manufacturers on whom they rely, must ensure that physicians and surgeons looking after patients are made aware of the limitations of immunoassays and other analytical results if costly tragedies such as those so graphically described by the USA hCG Reference Service (31) are to be avoided in the future.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  E. Kuhn, T. Addona, H. Keshishian, M. Burgess, D.R. Mani, R. T. Lee, M. S. Sabatine, R. E. Gerszten, and S. A. Carr Developing Multiplexed Assays for Troponin I and Interleukin-33 in Plasma by Peptide Immunoaffinity Enrichment and Targeted Mass Spectrometry Clin. Chem., June 1, 2009; 55(6): 1108 - 1117. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M J Hill, S Kumar, and P G McTernan Adipokines and the clinical laboratory: what to measure, when and how? J. Clin. Pathol., March 1, 2009; 62(3): 206 - 211. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H.-L. Wong, R. M. Pfeiffer, T. R. Fears, R. Vermeulen, S. Ji, and C. S. Rabkin Reproducibility and Correlations of Multiplex Cytokine Levels in Asymptomatic Persons Cancer Epidemiol. Biomarkers Prev., December 1, 2008; 17(12): 3450 - 3456. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Carraro and M. Plebani Errors in a Stat Laboratory: Types and Frequencies 10 Years Later Clin. Chem., July 1, 2007; 53(7): 1338 - 1342. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Marks Hypoglycaemia: insulin and conflicts with the law The British Journal of Diabetes & Vascular Disease, November 1, 2006; 6(6): 281 - 285. [Abstract] [PDF]
|
|
|
|
 |
|
 T. W. Wallace, S. M. Abdullah, M. H. Drazner, S. R. Das, A. Khera, D. K. McGuire, F. Wians, M. S. Sabatine, D. A. Morrow, and J. A. de Lemos Prevalence and Determinants of Troponin T Elevation in the General Population Circulation, April 25, 2006; 113(16): 1958 - 1965. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. D. Newman, P. B. Bergman, J. C.G. Doery, and N. D.H. Balazs Factitious increase in thyrotropin in a neonate caused by a maternally transmitted interfering substance. Clin. Chem., March 1, 2006; 52(3): 541 - 542. [Full Text] [PDF]
|
|
|
|
|
|
L. J. Kricka and J. Y. Park Additive-Aggravated Assays: An Authoritative Answer Clin. Chem., October 1, 2005; 51(10): 1767 - 1767. [Full Text] [PDF]
|
|
|
|
|
|
M. L. Cawood, H. P. Field, C. G. Ford, S. Gillingwater, A. Kicman, D. Cowan, and J. H. Barth Testosterone Measurement by Isotope-Dilution Liquid Chromatography-Tandem Mass Spectrometry: Validation of a Method for Routine Clinical Practice Clin. Chem., August 1, 2005; 51(8): 1472 - 1479. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Y. Liu, A. M. Xydakis, R. C. Hoogeveen, P. H. Jones, E. O. Smith, K. W. Nelson, and C. M. Ballantyne Multiplexed Analysis of Biomarkers Related to Obesity and the Metabolic Syndrome in Human Plasma, Using the Luminex-100 System Clin. Chem., July 1, 2005; 51(7): 1102 - 1109. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. W.L. Choi, S. Srivatsa, and J. C. Ritchie Aberrant Thyroid Testing Results in a Clinically Euthyroid Patient Who Had Received a Tumor Vaccine Clin. Chem., March 1, 2005; 51(3): 673 - 675. [Full Text] [PDF]
|
|
|
|
|
|
A. A.A. Ismail A Radical Approach Is Needed to Eliminate Interference from Endogenous Antibodies in Immunoassays Clin. Chem., January 1, 2005; 51(1): 25 - 26. [Full Text] [PDF]
|
|
|
|
|
|
R. Sapin, A. Agin, and F. Gasser Misleading High Thyrotropin Results Obtained with a Two-Site Immunometric Assay Involving a Chimeric Antibody Clin. Chem., May 1, 2004; 50(5): 946 - 948. [Full Text] [PDF]
|
|
|
|
|
|
P. A. Torjesen, L. Sandnes, and P. Boudou Serum Testosterone in Women as Measured by an Automated Immunoassay and a RIA * Dr. Boudou responds for the authors of the article cited above: Clin. Chem., March 1, 2004; 50(3): 678 - 679. [Full Text] [PDF]
|
|
|
|
|
|
J. Kristiansen The Guide to Expression of Uncertainty in Measurement Approach for Estimating Uncertainty: An Appraisal Clin. Chem., November 1, 2003; 49(11): 1822 - 1829. [Abstract] [Full Text]
|
|
|
|
|
|
L. L. Leape Striving for Perfection Clin. Chem., November 1, 2002; 48(11): 1871 - 1872. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) and without (
) use of HBT on samples from nine donors.
-fetoprotein (AFP) results obtained in 11 laboratories on samples from donor 5 without (
