Redesigned Proficiency Testing Materials Improve Survey Outcomes for Prostate-Specific Antigen: A College of American Pathologists Ligand Assay Survey Study

Large disparities in prostate-specific antigen (PSA) results from different assays have been observed in the College of American Pathologists (CAP) Ligand Assay Survey. Interassay results have been shown to vary severalfold.1 The PSA survey specimens have been prepared by adding semen, primarily composed of free (uncomplexed) PSA, to a processed human plasma base. The PSA from semen was used in proficiency and quality control materials at that time, because it was a convenient source of concentrated PSA. In addition, the different circulating molecular forms of PSA were not well understood.

Using gel-filtration chromatography to fractionate survey specimens, Garg et al2 verified that most PSA in CAP survey specimens is in fact free PSA. It is now well recognized that assays that do not equally recognize the free and α₁-antichymotrypsin (ACT)–bound forms of PSA can have a skewed response compared with equimolar assays.3 Several studies have shown that reactivities of different PSA assays to free PSA and PSA-ACT are a major contributing factor to disparate proficiency survey results.2,4,5 

Factors other than equimolarity can also contribute to differences among assays. These factors include the following: assay design and kinetics, antibody specificity, and calibrator composition and assignment.3,6–8 Recognition of PSA assay differences has prompted a number of efforts to standardize PSA assays. One such effort was the development of PSA reference material by CAP following a CAP and American Cancer Society conference on PSA in 1992.2,9 This material, prepared by pooling patient serum samples, was intended to reflect the fact that, in contrast to the distribution of the molecular forms of PSA present in semen, PSA in blood is primarily bound to ACT (∼80%), whereas only a small proportion is unbound.10,11 Analysis of the reference material revealed a percentage of free PSA of approximately 10% and close agreement among methods when these specimens were included in the 1994 CAP ligand survey.2 

The purpose of the CAP reference material was to provide an assayed material representative of clinical specimens for laboratories and manufacturers to evaluate the performance of PSA assays and to provide materials to aid in the standardization process.9 Following development of this material, the CAP Ligand and Therapeutic Drug Monitoring Resource Committee endeavored to introduce proficiency testing specimens for the ligand survey that would be representative of clinical specimens. Thus, based on recommendations from this committee, a pilot material was developed for CAP in which pooled serum samples were spiked with purified PSA and PSA-ACT at targeted concentrations. The purpose of these studies was to characterize the pilot material and the CAP survey and reference materials and to compare the performance of these materials in a number of total and free PSA assays.

Specimens analyzed in this study consisted of (1) CAP 1997 Ligand Assay Survey specimens K-01, K-02, and K-03, (2) CAP PSA reference material levels I, II, and III, and (3) a pilot material produced for CAP by Chiron Diagnostics (Emeryville, Calif). Preparation of the CAP survey and reference materials has been described in detail previously.2 Briefly, the Ligand Assay Survey material was prepared by supplementing a processed human plasma base with semen. In contrast, the PSA reference material consisted of human serum pools that contained targeted concentrations of endogenous PSA. The PSA pilot material was prepared by supplementing a human serum base with purified seminal fluid PSA and PSA-ACT conjugate.

Specimens were analyzed with 10 total PSA assays and 7 free PSA assays. These assays included both Food and Drug Administration–approved assays and research use only assays. Total PSA was analyzed with the following assays: Abbott ARCHITECT (Abbott Laboratories, Abbott Park, Ill) and AxSYM PSA, both polyclonal/monoclonal and monoclonal/monoclonal versions; Bayer Technicon Immuno 1 PSA (Bayer Corporation, Tarrytown, NY); Chiron (currently Bayer) ACS:180 PSA2; Diagnostic Products Corporation (DPC) Immulite 2000 PSA (DPC, Los Angeles, Calif); Hybritech Tandem-R and Tandem-MP PSA (Beckman Coulter, Brea, Calif); Roche/Boehringer Mannheim Elecsys 2010 PSA (Roche Diagnostics, Indianapolis, Ind); and Tosoh 1200 DX AIA-PACK PA (Tosoh Medics, South San Francisco, Calif). Free PSA was analyzed with the following assays: Abbott ARCHITECT and AxSYM, Bayer Technicon Immuno 1, DPC Immulite, Hybritech Tandem-R and Tandem-MP, and Roche/Boehringer Mannheim Elecsys 2010.

The lyophilized specimens were reconstituted according to the manufacturers' instructions and analyzed using all methods, with the exception of the Abbott ARCHITECT, on the same day. Analyses on the ARCHITECT were performed at a later time point using additional vials of material that were freshly reconstituted. Specimens were analyzed in duplicate, and results presented are the mean of the 2 values. The CVs for the CAP 1997 Ligand Assay Survey specimens and PSA reference material specimens were calculated from analyses performed for this study and do not represent all-method CVs observed in previous surveys.

Mean total PSA concentrations for the 1997 CAP Ligand Assay Survey specimens, the CAP reference material specimens, and the CAP pilot material analyzed by the 10 total PSA methods are shown in Table 1. Variability among methods was greatest for the 1997 ligand survey material (CV range, 56%–65%) followed by the pilot material (CV range, 10%–29%) and the reference material (CV range, 6%–13%). Table 2 illustrates variability patterns for the 10 methods comparing specimens representing the 3 types of materials with similar PSA concentrations. Results observed for individual PSA methods from ligand survey specimen K-03 are illustrative of patterns in the 3 1997 ligand survey specimens. Compared with the mean of all methods, PSA values for the Chiron ACS:180 PSA2 and the Abbott AxSYM polyclonal/monoclonal assays were 119% and 72% higher than the mean, respectively, whereas values for the Bayer Immuno 1 and DPC Immulite 2000 assays were 59% and 58% lower, respectively. The overall CVs for the reference (III) and pilot (3) materials were considerably lower than the CV for the survey material, 6% and 10% versus 57%, although within the reference material specimen the Chiron ACS:180 PSA2 and the Abbott AxSYM polyclonal/monoclonal methods again had the greatest increases from the mean.

Mean free PSA concentrations in the same specimens analyzed using 7 free PSA methods are shown in Table 3. In contrast to total PSA, variability among the free methods was similar among the 3 proficiency testing materials overall and when specimens with similar concentrations were compared (Table 4). Increased variability in reference material I can be attributed to concentrations close to the sensitivities of the assays.

The composition of the analyzed specimens in terms of the percentage of free PSA is presented in Table 5. Percentage of free PSA was calculated by pairing total and free methods from each manufacturer on the same platform (percentage of free PSA = [free PSA (Table 3) / total PSA (Table 1)] × 100). Using the Hybritech Tandem-R results, the ligand survey specimens were essentially composed of all free PSA, whereas the reference and pilot materials were composed of approximately 10% and 35% free PSA, respectively. Large differences among methods in the composition of survey material specimens reflect variability in both total and free PSA concentrations (Tables 2 and 4) related to measurement inaccuracies. The relationship between among-methods CVs and specimen composition is illustrated in Figure 1.

The 1998 Ligand Assay Survey Set A consisted of 3 challenges, 1 challenge similar to past surveys (K-03) and 2 challenges (K-16 and K-17) produced in a manner similar to the pilot material. Mean PSA concentrations for specimens K-03 and K-17 for the survey participants are shown in Figure 2.12 Variability among methods for specimen K-03 was 59%, with differences from the mean (7.28 ng/mL) for specific methods of −57% to 116%. In contrast, interassay variability for specimen K-17 was 13%, with differences from the mean (8.84 ng/mL) ranging from −12% to 26%. Based on participant results for free and total PSA analyzed by the Hybritech Tandem-R method, the percentages of free PSA in specimens K-03 and K-17 were 89% and 24%, respectively.

Differences in values among PSA assays have been observed, commencing with the introduction of total PSA assays in the late 1980s.13,14 It is now known that PSA in serum is composed of a number of different forms of the molecule, both free and complexed to protease inhibitors.15,16 It is, therefore, not unexpected that variability among PSA assays will exist as a result of assay design, including choice of antibodies, both polyclonal and monoclonal, and antibody specificities to the different forms of PSA and cross-reactivity to other molecules present in serum such as hK2,17 another member of the kallikrein gene family that shares 80% homology with PSA (hK3).

Participant results from the CAP ligand survey, as well as results from these and other studies2,4 using CAP survey material, demonstrate a large degree of variability among total PSA assays, with variability in some specimens showing severalfold differences. Similar large discrepancies have been observed in specimens analyzed in the New York State Proficiency Program.5 Specimens from both programs were prepared in a similar fashion with semen or seminal fluid PSA added to a plasma or serum base. Unlike serum, in which PSA is present predominately in complex with protease inhibitors such as ACT, α₂-macroglobulin, and α₁-protease inhibitor, seminal fluid PSA is composed of free, uncomplexed PSA.15 Therefore, as expected, when CAP survey specimens were analyzed to determine the free component in a previous study using gel filtration,2 and in this study using immunoassays specific for free PSA, they were found to be primarily composed of free PSA. Although the survey specimens were atypical in composition compared with human specimens, they have proven to provide scientifically valuable information by magnifying the performance of the different PSA assays.

Skewed responses by nonequimolar assays is a likely explanation for much of the large variability observed in the survey specimens. Nonequimolarity can result from antibodies, primarily polyclonal, that bind to epitopes on PSA that are blocked by ACT and from assay kinetic differences in which preferential binding of lower-molecular-weight free PSA may occur in assays with short incubation times.3,6,18 It is notable that the Chiron ACS:180 assay has the shortest incubation time, 7.5 minutes, and the highest CAP survey results. The Chiron ACS:180 and Abbott AxSYM/IMx assays with polyclonal/monoclonal formats have been reported to be nonequimolar,4,5,8,18–21 whereas the dual monoclonal Hybritech Tandem and Tosoh assays and the Bayer Immuno 1 assay have been reported in general to be equimolar in design.4,5,8,20,22 As shown in this study and in others,4,5,9,23 as a result of skewed assays, there is a direct relationship between assay discrepancies and the percentage of free PSA in the specimens. The survey materials were found to consist almost entirely of free PSA as determined using immunoassays, confirming previous observations by Garg et al.2 These materials displayed the greatest variability among methods compared with the reference and pilot materials that contained much lower percentages of free PSA.

Seminal fluid PSA is primarily composed of free active PSA and also contains approximately 30% nicked or clipped PSA.7,15 Different glycosylated forms are also present. Reactivities to these different forms of free PSA may also contribute to variability among the different PSA assays. A recent study by Chu et al24 showed no differences in relative immunoreactivity among PSA isoforms using the Tosoh AIA-600 assay, although differences were observed using the DPC Immulite third-generation PSA assay. Preliminary data from our laboratory assessing reactivity of seminal fluid PSA and inactive (nicked) seminal fluid PSA in the same 10 assays from this study did not suggest assays had different reactivities toward these forms of free PSA.25 Different reactivities toward additional forms of PSA not evaluated, both free and complexed, or cross-reactivities with other molecules may still be a contributing factor to assay variability.

Differences in calibration have also been proposed to contribute to variability among assays.3,13 Differences may reflect calibrator assignment and composition of the calibrators and specific molecular form(s) used.6,7 Initial assays used free PSA from seminal fluid as a calibrator material. A number of subsequently developed assays have used materials consisting of 90% PSA-ACT and 10% free PSA to more closely represent the composition of PSA in human serum, specifically those patients with prostate cancer. Assays that use this approach include Abbott AxSYM (mono/mono assay) and ARCHITECT, Chiron ACS:180 PSA2, and Roche/BMC Elecsys. Adoption of this 90:10 material has been proposed as a method of interassay standardization.26 However, use of this material will not necessarily uniformly standardize all assays, particularly those that are not equimolar.

In contrast to total PSA, interassay variability for the 7 free PSA methods was similar for the 3 different types of preparations. Variability of 10% to 23% was similar to the results found by Fox et al5 in the New York State Proficiency Survey in specimens of purified PSA forms in a nonhuman matrix and slightly lower than Zuchelli et al27 in the International External Quality Assessment Program, who found an overall between-laboratory CV of 28% in Oncocheck specimens prepared from human serum pools. Variability of reference material specimen I reflects specimen concentrations near the sensitivities of the assays. The larger CVs noted for the total assays compared with the free assays are most likely a result of measurement issues related to free and complexed components of total PSA. Differences among free PSA assays may reflect antibody reactivities to different forms of free PSA, and increased imprecision at lower concentrations may also contribute. As we have previously shown in a study comparing combinations of a number of total and free methods,28 differences observed among free PSA methods suggest the need for free PSA standardization. Standardization is an issue for free and total PSA assays individually but also for free and total assays when used together to determine the percentage of free PSA as an approach to increase the specificity of PSA for detecting prostate cancer. It has been suggested that free and total assays should be calibrated so that specimens would have the same value in both assays and thus the 2 assays would be linked.29 We and others have shown that variability in both total and free PSA assays influences the determination of the percentage of free PSA,4,5,27,28,30,31 which may affect clinical interpretation, and thus it had been recommended28 that cutoff values for discrimination should be determined for individual free and total assay combinations both within and between manufacturers.

In summary, differences in total PSA concentrations among various manufacturers was minimized with CAP reference and pilot formulations compared with the original survey material. Differences between the survey and other 2 materials was attributed to the composition of the specimens with respect to the molecular forms of PSA as a result of the method of preparation of the specimens. As shown in this study of 10 total PSA assays, the pilot material, manufactured to simulate patients specimens that typically contain percentages of free PSA in the range of 10% to 30%, was closer to the CAP reference material, which was composed of pooled patient serum samples, when the performance of the individual assays was compared. The effectiveness of the pilot material in diminishing discrepancies among PSA assays is illustrated in Figure 2, comparing participant results from 2 1998 Ligand Assay Survey challenges.12 Variability among assays was 59% for the typical survey material, which consisted of 89% free PSA compared with 13% for a pilot-like material that consisted of 24% free PSA. Based on participant results from the 1998 surveys in which this new material was tested, this new formulation of PSA using defined concentrations of purified PSA forms replaced the original approach of semen supplementation for all Ligand Assay Survey specimen challenges starting in 1999.