Immunoassay Standardization: Is It Possible, Who Is Responsible, Who Is Capable?

Standardization in clinical chemistry aims at improving trueness, i.e., lack of bias, but it does not directly affect assay imprecision. Standardization is important because it facilitates clinical interpretation and comparison of results from various studies (1)(2). This is especially important with analytes used for screening, and this was one reason that cholesterol assays have been standardized in the US (3). Standardization of assays for prostate-specific antigen (PSA) is important for the same reason, and also because these assays are important for monitoring of disease progression in prostate cancer. Large differences in assay calibration lead to misinterpretations of the clinical course when different assays are used (4). Variation in assays for chorionic gonadotropin (CG) (5) cause similar problems when women with pregnancy-related disorders are monitored by determination of CG in serum (6). It is a testimony of the clinical importance of standardization that the preparation of standards for PSA was initiated and carried out by a urologist, Thomas Stamey from Stanford (7). Standardization is a fairly new concept in clinical chemistry. The notion of “comprehensive measurement systems” and various levels of reference methods evolved in the 1970s (1)(2) and was formulated by Tietz in 1979 (8). Pure standards, reference methods, and standard reference materials (SRMs) are now available for many of the most important analytes determined by conventional chemical methods, but immunoassay standardization is less well developed. Immunoassay standardization was extensively discussed at two meetings organized in 1990 and 1992 by the IFCC. At these, two standardization projects were initiated: the one for cortisol was intended to serve as a model for hapten immunoassays, the other project, for standardization of assays for CG, was intended to serve as a pilot study for protein immunoassays. The problems associated with standardization of protein and hapten assays are different. Pure standards and reference methods based on mass spectrometry (MS) are available for many steroid hormone assays (9). For peptide and protein immunoassay standardization there are no reference methods, value assignment of the standard is problematic, and the analyte is mostly heterogeneous and differs from the standard. Therefore, progress in standardization of protein immunoassays has been slow. I have served as the chairman of the Working Group for Standardization of CG (5) and been involved in other projects aimed at immunoassay standardization (10)(11). This experience has affected my thinking, but I am solely responsible for the views expressed here. Problems associated with standardization of hapten assays have been dealt with extensively in the recent literature (9)(12) and will only briefly be touched on here. The aim of this opinion is to stimulate discussion rather than to present a comprehensive and balanced review. Roger Ekins, one of the inventors of immunoassays, claims that standardization of immunoassays for heterogeneous antigens is impossible (13). It is therefore necessary to examine the basic problems before discussing possible solutions. The prerequisites for standardization are availability of standards and reference methods (8)(14). The primary standard should be homogeneous, pure, and identical to the corresponding substance in the sample that is measured. It is obvious that these requirements can be met only for some hapten immunoassays. Reference methods often are classified as primary and secondary (8). The primary method should preferably be a “definitive method”, e.g., isotope dilution MS. Such methods are available for steroids (9) and thyroid hormones. Recent developments in MS have made it possible to analyze peptides and proteins, and isotope dilution MS has been used to develop reference methods for two proteins (15)(16) and thyroid hormones (17), but not for the free hormones, which have largely replaced assays for total thyroxine and triiodothyronine. However, MS is not useful for proteins and peptides occurring at low concentrations. In the absence of definitive methods, the best method available may be defined as the primary reference method, but this approach has not been applied in practice. The problems hampering immunoassay standardization are different for different analytes. With steroid hormones, cross-reactions between closely related forms and complex formation with binding proteins in the sample are major problems. Previously used extraction methods are not compatible with automation, and the use of direct methods with short incubation times on automatic analyzers has actually aggravated the problem (5). Although reference methods are available and standardization possible (12), the accuracy and sensitivity of many automatic steroid assays are disappointing (18). Immunoassay of urinary free cortisol is especially problematic because of the large excess of cross-reacting substances in urine. The use of immunoassay for this purpose is dubious, but urinary free cortisol can be accurately determined by HPLC (19). However, this method is used by only 3 of 118 laboratories participating in the British quality assessment scheme (UKNEQAS) (20). The clinicians in our hospital accept the higher costs of the HPLC assay because falsely increased immunoassay results often lead to much higher additional costs in the form of imaging examinations (21). I believe that we should not give in to pressures to reduce laboratory costs when this leads to unacceptable compromises. Peptide and protein antigens will always represent a problem for immunoassay standardization because of the heterogeneity of the antigens occurring in biological fluids (13). In its strictest sense, standardization is possible only if standard and analyte are identical. This is the case with only a few small peptides, e.g., angiotensin-1 and -2. With larger peptides and proteins, the problems tend to become more complicated because biological samples often contain proforms, splice variants, fragments, and complexes of the analyte. Most hormones and tumor markers are glycoproteins. These are inherently heterogeneous, and it is virtually impossible to prepare a glycoprotein standard that is identical to the circulating form. There are no universal solutions to these problems, but if we identify them, accept some compromises, and set our ambitions at a realistic level, it is possible to improve the state of standardization of most immunoassays to a clinically adequate level. It is also necessary to identify analytes that cannot be appropriately standardized and to develop other solutions for these. Many of the typical problems of immunoassay standardization have been encountered in the CG project, and the solutions that the CG working group has arrived at may be applicable to other similar assays. Intact CG is a heterodimer, but free subunits (CGα and CGβ) and partially degraded (nicked) forms (CGn and CGβn) also occur in circulation and are measured to a variable extent by immunoassays (5). Furthermore, a core fragment of CGβ (CGβcf) occurs in urine and is a potentially useful tumor marker. International standards are available for only part of these forms, and the CG standard is contaminated by nicked forms (22). Many of the original hormone standards consisted of partially purified preparations, the potencies of which were determined by bioassay and expressed in arbitrary international units (IUs). IUs have also been assigned to some analytes without known biological activity (e.g., α-fetoprotein, carcinoembryonic antigen, CGα, and CGβ). IUs based on bioactivity are important for characterization of therapeutic hormone preparations and standardization of biological assays. However, they are not necessarily well suited for immunoassays, and when new hormone standards are introduced, the ratio between biological activity and immunoreactivity is likely to change (5). Furthermore, tracing of the values back to previously used impure standards loses its scientific basis when assay specificity is changed by the introduction of methods based on monoclonal antibodies. The IFCC recommends the use of International System of Units (SI) units, and this is possible if the standard can be isolated in pure form. The primary structures of all clinically important hormones are known, and they can either be purified from natural sources or expressed by recombinant techniques. Therefore, their molar concentrations can be determined by amino acid analysis, and this should be the primary method for value assignment of new standards. The new CG standards will be assigned values by this method and expressed in substance concentrations, i.e., mol/L (5). This is based on the following arguments: The molar concentration of a pure protein can be accurately determined by amino acid analysis. Antibodies primarily recognize peptide epitopes, which are related to the amino acid composition. Physiological variation in carbohydrate composition causes variation in bioactivity and weight of the preparation, but it rarely affects immunoreactivity. Immunoassays measure immunoreactivity rather than bioactivity. Molar concentrations are logical when comparing concentrations of related compounds with large differences in molecular weight, e.g., intact CG and subunits or free and complexed PSA. The new CG standards are currently being evaluated by the National Institute of Biological Standards and Control (NIBSC; United Kingdom) for suitability as international standards. The biological activity will be determined for intact CG, and the standards will be cross-calibrated by immunoassay against the currently used standards. Thus, either IUs or molar concentrations can be used. When new standards need to be prepared, the ratio between molar concentration and immunoreactivity can be maintained, i.e., the determination of protein content by amino acid analysis will guarantee continuity of the value assignment. The molecular weight of the various CG preparations has been determined by MS, and the mass of protein in the standards can be calculated on the basis of the average molecular weight (23). Most protein antigens are now determined by sandwich assays using two monoclonal antibodies or a combination of a monoclonal antibody and a polyclonal antiserum reacting with different epitopes. This facilitates tuning of assay specificity, e.g., CG assays can be designed to detect only heterodimeric CG, only CGβ, or both components at the same time (5). In many cases, the specificity obtained by sandwich assays improves clinical utility, e.g., determination of the proportion of free or complexed PSA provides better cancer specificity than total PSA (24), and assays specific for CGβ provide better sensitivity for nontrophoblastic cancer than assays that detect both CG and CGβ (25). Assays measuring two components together tend to underestimate either one (26), but at least for PSA it has been possible to design assays that recognize free and complexed forms equally (27). Thus, standardization may be possible even when the analyte occurs in quite different forms in serum. Partial proteolytic degradation of the peptide chain or nicking is fairly common among hormones and tumor markers, and this may cause variation in immunoreactivity. Some antibodies recognize intact and nicked forms equally, whereas others underestimate nicked forms of CG (28) and PSA (29). It is therefore important to know the reactivity of the antibodies used in relation to the characteristics of analyte and standard. The specificity of an assay is dependent mainly on the epitope specificity of the antibody used. Detailed epitope maps are available for CG (30)(31), and this was an important reason to choose it as a model compound for immunoassay standardization (5). The International Society for Oncodevelopmental Biology and Medicine (ISOBM) has recently organized epitope mapping of several clinically important tumor markers in a series of workshops, in which the major immunoassay manufacturers have participated (10)(32)(33)(34). These epitope maps form the basis for selection of antibodies for routine and reference methods. Reference methods have been established for most clinically important steroid hormones (9) but only for two proteins, i.e., hemoglobin A1c (16) and apolipoprotein A-1. These methods are based on proteolytic digestion of the protein in the crude sample, which releases a peptide that is determined by MS. Because this approach is applicable only to proteins occurring at high concentrations, the working group for CG has planned to develop immunological reference methods. This is possible principally because standards (23) and antibodies with well-defined epitope specificity are available (11). However, the companies supporting the CG project have not embraced this idea, and unless accepted by the industry, reference methods are of limited value. The main argument has been that an immunological reference method may be too dependent on a certain assay technology. Thus, even with the same antibodies and standards, different results could be obtained when shifting from a manual reference method to an automated routine assay. To be acceptable, a reference method would need to be insensitive to matrix effects, applicable to some generally available assays formats, and based on international standards and readily available monoclonal antibodies with known epitope specificities. It should be possible to either establish the reference method in any laboratory or to maintain a network of reference laboratories similar to that established for cholesterol (3). However, the feasibility of immunological reference methods first needs to be demonstrated in pilot studies. Panels of sera in which the content of one or several analytes has been determined by a definitive method have been prepared for many chemical and some immunochemical assays. These secondary standards are called SRMs by the NIST and Certified Reference Materials by the Institute for Reference Materials and Measurements (IRMM) (12). Standardization of serum protein determinations has been accomplished by preparation of a serum-based standard in which values were assigned to the most commonly determined serum proteins (35). The CG working group has also considered distributing calibrated serum samples (23), but this is feasible only if reference methods are available. The use of serum-based secondary standards has the advantage that the matrix is similar to that in clinical samples. However, this does not eliminate the matrix problem, which varies from one sample to another. Furthermore, different immunoassays have different matrix problems. Although assay manufacturers are skeptical about immunological reference methods, they often use a well-respected commercial assay or that of the market leader to calibrate their assays (18). Would it be possible to utilize this principle for standardization? Could an expert group investigate available assays and declare the best one a reference method? This approach has several advantages: it is economical, it could be effectuated rapidly, and the reference method would be readily available. However, it also carries risks. The life span of the method might be limited because of changes in assay technology and company management. Furthermore, having the official reference method might be considered an unfair advantage by competitors. Because of this, it appears necessary to establish reference methods that are independent of commercial companies, but the use of commercial methods could be an interim solution in selected cases. Several organizations are involved in various aspects of immunoassay standardization, but the responsibility of each of these has not been defined. At the top of the standardization hierarchy is the International Standardization Organization (ISO), to which national and regional standardization organizations that have established standards and reference materials for substances of importance in clinical chemistry, e.g., NIST in the United States and the Community Bureau of Reference (BCR) in Europe, report. The WHO issues standards for most biologically important hormones and several other clinically important antigens determined by immunoassay. These standards are validated by the NIBSC in the United Kingdom, and if found acceptable, are issued as International Standards by the Expert Committee on Biological Standards (ECBS) and distributed by NIBSC. To date, WHO deals only with standards but not with assays and reference methods. Other important organizations include the NIH, which provides hormone preparations for research purposes, and the CDC. In addition to these official bodies, several professional organizations have projects concerning immunoassay standardization, e.g., IFCC, the International Union for Pure and Applied Chemistry (IUPAC), ISOBM, the College of American Pathologists (CAP), NCCLS, and the American and Canadian societies for clinical chemistry, AACC and CFCC. Although quite a few organizations deal with standardization, it is not clear who is responsible for what. Because standardization is an international rather than a national or regional problem, it is desirable that one international organization should be responsible for the coordination of various standardization projects. The requirements imposed by various bodies controlling the use of immunoassays vary from one country to another. The need to obtain certification for new assays by the Food and Drug Administration (FDA) in the US has led to considerable restrictions in assay availability, but this does guarantee that the assays are properly The Community has recently issued the In which should be by It that of values assigned to materials be available reference measurement available reference materials of a higher The Committee for Standardization or has its to develop standards. for in are by which in with the has the responsibility for and is the of the National Reference System for the However, the or responsible for of reference has not been to the may be but this is not for of a method in These requirements are determined by national I believe that standardization should be for and the have been defined and the i.e., standards and reference methods, have been made available. However, it is obvious that reference for protein immunoassays will not be available by when the should be The on and of standards and reference methods is by in working problem is that many of the either have a in antibodies and assay or they have or are we these, the of will be of all that in the industry, appears both necessary and At least for the CG project, the working group has also most of the from assay This is not a and it may be to unless is Although many aspects of standardization pure it is not to for research with standardization projects. Therefore, these should be from other Most laboratories considerable on quality which standardization the aim of improving the quality of laboratory I that would be to for better standardization if the were available and progress could be In addition to being problematic and standardization is also and The preparation of PSA standards was first discussed at two meetings in Stanford organized by Thomas Stamey in 1992 and and the on these standards in this The CG project was initiated at the same and the standards will be available this This time is the problems of immunoassay standardization will not be this or even the even if was made available. therefore need to the following interim solutions. most of the standardization problems are of by assay and this will not change standardization is an important and much on this is found In of this the results are which can be on both the lack of reference and These problems have actually become because of the of automated analyzers with to assay Therefore, standardization may not be possible for all assays on all Although this problem is it is also a reason not to the problem to the standardization is not can be used to calibrate various assays to give the same However, assays may all be would some of the clinical problems, but from a of it is dubious, and of calibration is problematic in the absence of reference methods. Therefore, standardization should be the For assays of antigens that have not been e.g., only is possible at With the present state of standardization, reference values need to be determined for each assay. The is responsible for the reference values but I believe that assay manufacturers should be to prepare these the of the IFCC. At few companies of reference values can be quite especially for hormones, the concentrations of which are dependent on and Therefore, among and assay manufacturers is of samples from reference should be established and made available to assay This could be accomplished in and it would be feasible because of the value of reference values are necessary because of the standardization, but they not The interpretation of laboratory results can be by the results not only as concentrations but also as an of the that the a certain This approach is especially useful for interpretation of results of determinations as used in for on the basis of free and total PSA in combination with clinical has been to improve the accuracy for prostate cancer of the of disease can be by the results for total and cholesterol and protein The results are not dependent on standardization and reference but the are only for the assays used to establish unless the assays are of is because it both reference of who not develop the disease and large of with accurately established Therefore, this is more than of reference and use of will rather than the need for The results of quality assessment provide about the of various methods, but they not an assay is properly is obtained if the samples have been with pure standards or by a reference method The results for samples the state of standardization, but matrix and problems it impossible to the calibration of assays on the basis of the results of a few samples I not any claims to have adequate to the in the of this but I would to present a of and problems that need to be of Because standardization of many immunoassays compromises, we need to the quality for various assays clinical needs and and Because are limited and standardization is and we need to in which assays should be of the of immunological reference methods. we reference methods accurately that they can be used or should we establish a network of reference laboratories or prepare serum-based secondary These need to be and on common for organizations working on immunoassay standardization should be established and the of the various organizations additional is not responsible for and quality assessment of assays should have an in The problems would be if a small of the on these could be to