OBJECTIVE. significantly different classification rates of patients. These differences were even larger when using fixed %fpsa thresholds.

Transcription

1 Original Article CLINICAL IMPACT OF DIFFERENT PSA ASSAYS STEPHAN et al. Different prostate-specific antigen assays give different results on the same blood sample: an obstacle to recommending uniform limits for prostate biopsies Carsten Stephan, Jürgen Kramer, Hellmuth-Alexander Meyer, Glen Kristiansen*, Sabine Ziemer, Serdar Deger, Michael Lein, Stefan A. Loening and Klaus Jung Departments of Urology, *Pathology, and Clinical Chemistry and Laboratory Medicine, Charité Universitätsmedizin Berlin, CCM, Germany Accepted for publication 21 December 2006 CS and JK contributed equally to this work. All authors are members of the Biomarkers of Prostate Cancer Group, CCM OBJECTIVE To show the effect of different results for total prostate specific antigen (tpsa) and percentage free/total PSA (%fpsa) obtained with different assays for differentiating between benign and malignant prostate diseases. PATIENTS AND METHODS Data were used for tpsa and fpsa levels from 596 patients with prostate cancer (314) or no evidence of cancer (282) within the PSA range ng/ml, analysed with assays from Abbott (AxSYM), Beckman Coulter (Access), DPC (Immulite 2000), and Roche (Elecsys 2010), and with tpsa and complexed PSA (cpsa) assays from Bayer (ADVIA Centaur), as already reported. Receiver operating characteristics (ROC), specificities at assaydependent and fixed thresholds, and the percentages of correct classification rates of patients were calculated. RESULTS Whereas the areas under the ROC curves were no different among all tpsa assays, the assayspecific thresholds at 90% sensitivity were ng/ml. When using fixed 2.5 or 4 ng/ ml tpsa thresholds there was a wide sensitivity range, with significant differences among almost all assays, resulting in significantly different classification rates of patients. These differences were even larger when using fixed %fpsa thresholds. CONCLUSIONS The current situation of differences among PSA values measured with different assays do not allow the recommendation of uniform PSA limits as biopsy criteria. For that purpose, better harmonization of PSA values between the different PSA test systems must be realized. KEYWORDS PSA, prostate cancer, assays, variation INTRODUCTION PSA is accepted as adding significant information for the early detection of prostate cancer, but the low positive predictive value and the low specificity are major drawbacks with this marker [1]. The use of molecular forms of PSA such as free PSA (fpsa) and its ratio to total PSA (%fpsa) or complexed PSA (cpsa) has been proposed to enhance the diagnostic specificity for detecting prostate cancer [2,3]. However, tpsa and especially fpsa concentrations obtained with different assays are very discordant [4]. These assay-dependent variations could result in misinterpretations of individual tpsa and %fpsa values, with possibly erroneous clinical decisions for detecting prostate cancer. Non-uniform calibration of the assays and nonequimolar detection of tpsa were suggested to be the essential reasons for this discordance [5,6]. Consequently, PSA reference materials were developed [7]. Promising efforts have been made in adjusting PSA calibrations using the WHO PSA reference materials and with the introduction of equimolar-response assays [8,9]. The USA Food and Drug Administration approval of %fpsa was initially limited to the Access assay (Hybritech, Beckman Coulter, Brea, CA, USA) with extension to the Abbott (Abbott Park, IL, USA) and Roche (Roche Diagnostics Corp., Indianapolis, IN, USA) assays only recently. Although the insufficient comparability between various PSA test systems was already discussed previously [4,9,10], there are only a few studies, with few assays and few samples, that do not fully describe the current situation, especially the clinical significance [11,12]. Two recent studies were likewise solely focused on the analytical aspects without considering the diagnostic consequence [13,14]. Recently, we compared five different tpsa and fpsa assays (one with cpsa) frequently used in clinical practice in 596 sera from patients with no evidence of prostate cancer and patients with cancer, from an analytical perspective for the interchangeability of PSA values among the assays and their equimolar characteristics [15]. There was inadequate comparability of tpsa, fpsa and %fpsa values measured by the different assays [15]. In the present study the main focus was directed on the clinical impact of these assay-dependent differences in tpsa and especially %fpsa values. We evaluated the assay-related absolute thresholds at 90% and 95% sensitivity, and the impact of fixed thresholds for tpsa (2.5 and 4 ng/ml) and %fpsa (15%, 20%, 25%) on the correct classification rate when using the five different assays BJU INTERNATIONAL 99, doi: /j x x 1427

2 STEPHAN ET AL. TABLE 1 The median (95% CI) tpsa and %fpsa values, AUC and specificities at fixed tpsa and %fpsa 90% and 95% sensitivity thresholds, with the corresponding tpsa thresholds for the five tpsa and %fpsa assays, calculated from with 314 patients with prostate cancer and 282 without Assay Median, ng/ml or % AUC 90% sensitivity 95% sensitivity Threshold, ng/ml or % Specificity, % Threshold, ng/ml or % Specificity % tpsa* AxSYM 4.6 ( ) 0.72 ( ) ( ) ( ) Access 5.14 ( ) 0.70 ( ) ( ) ( ) Centaur 4.61 ( ) 0.71 ( ) ( ) ( ) Immulite 5.70 ( ) 0.71 ( ) ( ) ( ) Elecsys 5.48 ( ) 0.70 ( ) ( ) ( ) %fpsa* AxSYM 17.9 ( ) 0.80 ( ) ( ) d ( ) e Access 14.5 ( ) 0.81 ( ) c,e ( ) a,c,d ( ) e Centaur 17.7 ( ) 0.77 ( ) a,b,d ( ) b ( ) Immulite 11.4 ( ) 0.81 ( ) c ( ) b ( ) Elecsys 13.8 ( ) 0.79 ( ) b ( ) ( ) a,b *The median values of all five assays were significantly different from each other (Wilcoxon test of paired samples; P < 0.001) except for the tpsa test between AxSYM and Centaur (P = 0.212). The %fpsa for the cpsa assay was calculated as (1 cpsa/tpsa) 100. All significances for %fpsa (P < ) were Bonferroni corrected. Significantly different from: a AxSYM; b Access; c Centaur; d Immulite; e Elecsys. PATIENTS AND METHODS The study group comprised 596 untreated White men with a tpsa concentration of ng/ml, as determined with the Access PSA system (Hybritech, Beckman Coulter) for this investigation. All men were referred to our clinic, thus resulting in an over-representation of prostate cancer compared with a screening population. The patients were classified into two groups: 314 men had a histologically confirmed diagnosis of prostate cancer (median age 66 years, range 38 85) and 282 men did not (median age 63 years, range 43 79) on prostate biopsy (8 12 cores). Blood samples were taken before any diagnostic or therapeutic procedures involving the prostate, and 4 weeks after a DRE, prostatic biopsy or TRUS. The samples were collected between 2001 and 2004, and sera were stored at 80 C and analysed retrospectively. The measurements were made according to the manufacturer s instructions on the analysers AxSYM (Abbott: tpsa, cat. no. 3C19-20; fpsa, cat. no. 3C 20 20), Elecsys 2010 (Roche: tpsa, cat. no ; fpsa, cat. no ), and ADVIA Centaur (Bayer Diagnostics, Leverkusen, Germany: tpsa, cat. no ; cpsa, cat. no ), Access (Beckman Coulter: Hybritech PSA, cat. no ; Hybritech fpsa, cat. no ), and Immulite 2000 systems (Diagnostic Product Corp., Los Angeles, CA, USA; tpsa, cat. no. L2KPS6; fpsa, cat. no. L2KPF2) as already described [15]. The cpsa values were transformed into fpsa concentrations as tpsa cpsa and the %fpsa values calculated as percentage ratios of fpsa to tpsa. The between-run imprecision profiles of the measurements were estimated using control materials supplied by the manufacturers, commercial control materials, and in-house serum pools; all interassay coefficients of variation were <8% (over days). The study was carried out in accordance with ethical standards of the local ethics board and the Helsinki Declaration of 1996, including informed consent of all participants. Statistical calculations were performed with SPSS 13.0 for Windows (SPSS, Chicago, USA) and GraphPad Prism 4.03 (GraphPad Software, San Diego, USA). The diagnostic accuracy was evaluated using receiver operating characteristic (ROC) curve analysis with calculations of the area under the curve (AUC) using MedCalc (MedCalc Software, Mariakerke, Belgium). Significance levels at 95% and 90% sensitivity were calculated by using GraphROC 2.1 for Windows [16]. P < 0.05 was considered to indicate statistical significance, using the Bonferroni correction if there were multiple comparisons (10 comparisons). RESULTS The analytical data of the comparisons among the various assays using the Access Hybritech assay as the comparison method, were reported previously [15]. The main differences are shown in Table 1; the median tpsa concentrations for the five assays was ng/ml, with significant differences among the assays. These differences are also reflected by the different values at ng/ ml tpsa at a threshold of 90% sensitivity, and at ng/ml tpsa at the corresponding level of 95% sensitivity, respectively (Table 1). However, when comparing the AUC for tpsa, there was no difference among the assays (P = ). The specificities at 90% and 95% sensitivity also did not differ (P = ) among the assays (Table 1). These equivalent data for the diagnostic validity among the assays confirm that all tpsa assays are comparable in their diagnostic validity if the assay-dependent thresholds are considered. Consequently and when using fixed tpsa thresholds of 2.5 or 4 ng/ml, the number of patients with or without prostate cancer who were classified as true-positive (cancer) or truenegative (no cancer) significantly differed among the five assays (Table 2). For example, when using 4 ng/ml as the threshold, the true-positive results were different in eight of the 10 comparisons, corresponding to BJU INTERNATIONAL

3 CLINICAL IMPACT OF DIFFERENT PSA ASSAYS TABLE 2 Classification of 314 patients with and 282 without malignancy using fixed tpsa thresholds of 2.5 and 4.0 ng/ml for the five tpsa assays, and fixed %fpsa thrsholds of 15%, 20%, and 25% for the five %fpsa assay combinations Selected thresholds sensitivities of 70% (Centaur) to 85% (Immulite). N (%) of classified patients AxSYM Access Centaur Immulite Elecsys tpsa 2.5 ng/ml True positive 288 (91.7) d 288 (91.7) d 283 (90.1) d,e 297 (94.6) a,b,c 292 (93.0) c True negative 116 (41.1) b,d,e 95 (33.7) a,c 113 (40.1) b,d,e 89 (31.6) a,c 96 (34.0) a,c 4.0 ng/ml True positive 228 (72.6) b,d,e 246 (78.3) a,c,d 221 (70.4) b,d,e 268 (85.4) a,b,c,e 255 (81.2) a,c,d True negative 165 (23.0) b,d,e 143 (50.7) a,c 162 (57.4) b,d,e 138 (48.9) a,c 143 (50.7) a,c %fpsa <15% True positive 183 (58.3) b,d,e 239 (76.1) a,c,d 182 (58.0) b,d,e 279 (88.9) a,b,c,e 240 (76.4) a,c,d True negative 238 (84.4) b,d,e 203 (72.0) a,c,d,e 228 (80.9) b,d,e 141 (50.0) a,b,c,e 181 (64.2) a,b,c,d <20% True positive 256 (81.5) b,d,e 293 (93.3) a,c,d 249 (79.3) b,d,e 307 (97.8) a,b,c,e 297 (94.6) a,c,d True negative 180 (63.8) b,d,e 127 (45.0) a,c,d,e 177 (62.8) b,d,e 79 (28.0) a,b,c,e 103 (36.5) a,b,c,d <25% True positive 285 (90.8) b,d,e 305 (97.1) a,c 290 (92.4) b,d,e 310 (98.7) a,c 307 (97.8) a,c True negative 121 (42.9) b,d,e 65 (23.0) a,c,d,e 126 (44.7) b,d,e 39 (13.8) a,b,c 49 (17.4) a,b,c *Significant differences of the classifications (McNemar test of paired samples; Bonferroni corrected; P < 0.05 <0.001). Significantly different from the: a AxSYM; b Access; c Centaur; d Immulite; e Elecsys. Corresponding calculations for %fpsa obtained with the five assays showed that the median values for all assays were also significantly different from each other (Table 1). Again, these differences among the assays were also reflected by the different absolute %fpsa thresholds of % at 90% sensitivity and % at the corresponding 95% sensitivity level. However, the AUC were in a narrow range of , but with four significant differences, between the Centaur assays (0.77) and all others except the Elecsys assays (0.79), and the Access and Elecsys assays. At 90% sensitivity the specificity showed a significantly better performance for the Access compared to all other except the Elecsys assay. All other comparisons at this point showed no differences. At 95% sensitivity there were also almost no differences, except a significantly better performance of the AxSYM and Access compared to the Elecsys assay. Thus, the data for the diagnostic validity among %fpsa values were almost equivalent when assaydependent thresholds at 90% or 95% sensitivity were considered. By contrast, when using fixed %fpsa thresholds of 15%, 20% or 25%, the number of true-positive or true-negative patients correctly classified significantly differed among the five assays for most comparisons (Table 2). For example, when using a 15% threshold for %fpsa, the Immulite assays detected almost 100 more patients with cancer than the AxSYM and Centaur assays, but missed almost 100 more patients with no cancer than the AxSYM assay. These results would imply that the Immulite tests had a 30% higher sensitivity, but a 31 34% lower specificity than the two other assays. These differences were also evident at the two other fixed thresholds of 20% and 25% (Table 2). The differences in sensitivity among the assays understandably decrease and the corresponding specificities increase with higher thresholds. Thus, it is clear that fixed %fpsa thresholds lead to large differences among the assays, with a misclassification of more than a third of the patients only by using another assay. DISCUSSION Since the introduction of the assays for molecular forms of PSA in the early 1990s [2,3] and their routine clinical use there have been many reports on the use of %fpsa%fpsa to reduce the number of unnecessary prostate biopsies. Only a few studies have given enough attention to possible differences and misleading interpretation when using different assays for tpsa and fpsa [11,12,15,17]. The present study shows significant differences among five frequently used commercial tpsa and fpsa assays, although the WHO standard was introduced to improve the comparability of PSA measurement using different assays [15]. However, the present results show that the situation is unsatisfactory. An optimum tpsa threshold for the first decision to guide the patient remains of utmost importance. Whereas most clinicians still prefer the 4 ng/ ml threshold to direct the patient for a biopsy or further assessments, in the European Randomized Study for Prostate Cancer study the threshold was lowered to 3 ng/ml [18]. Moreover, the tendency, especially in the USA, is to lower the threshold to 2.5 ng/ml to keep at least 90% sensitivity [19]. This could be confirmed by the present results, but only when using the Centaur or AxSYM tpsa assays. Furthermore, the cancer detection rate at lower tpsa concentrations as analysed by Thompson et al. [20] is, especially at 2 4 ng/ml tpsa, similar to that of the grey zone of 4 10 ng/ml [21], so that the 4 ng/ml threshold loses any utility. However, it is surprising that these controversial discussions do not consider sufficiently the inadequate interchangeability of PSA values obtained with different assays, as shown in the present study. The misclassifications shown in Table 2 if fixed thresholds are used, or a tpsa level of ng/ml to reach 90% sensitivity (Table 1), clearly underline this issue. One possible explanation for the large variety among the five assays might be that only three (AxSYM, Centaur, Elecsys) are calibrated against the WHO standard [22]. The tpsa assays from Beckman and DPC, not calibrated against the WHO standard, tended to give generally higher tpsa concentrations. However, despite the uniform calibration against the WHO standard, the Elecsys assay resulted in higher values than the other two WHO calibrated assays (AxSYM and Centaur). Thus, calibration might be an important, but not the exclusive factor, to improve the harmonization among the various assays. The situation seems to be similar when analysing the calculated %fpsa; here the 2007 BJU INTERNATIONAL 1429

4 STEPHAN ET AL. assays not calibrated against WHO standards show the lowest (DPC) and third lowest values (Beckman), with a range for the median %fpsa of % and significant differences among all median values (Table 1). However, the use of fixed thresholds for %fpsa (15%, 20%, 25%) gave larger problems than for tpsa, as shown by the misclassifications in Table 2. This is important from the clinical perspective, as the fpsa assays used should be given more attention as additional tools to reduce unnecessary prostate biopsies [23]. At 95% sensitivity, the respective absolute %fpsa thresholds varied from 18% to 28%, so that only assay-specific thresholds would give comparable clinical results. However, it is also obvious that the WHO calibration, used according to the information from the manufacturers of the AxSYM and Elecsys assays, does not guarantee a solution for the poor interchangeability of %fpsa results (Table 1). Medians and thresholds at 90% and 95% sensitivity were clearly different. These thresholds can only be regarded for comparative purposes and not as practical thresholds, because the patients included in the present study were not from a screening population. In conclusion, in this study with parallel measurements of five current tpsa and fpsa assays in almost 600 patients, within the relevant tpsa range, showed highly significant differences for tpsa and especially %fpsa. These differences were also obtained with assays calibrated using the WHO standard. Thus, we think that a better harmonization of PSA results among different assays is needed as an essential precondition for success when selecting final recommendations for tpsa or %fpsa thresholds as indicators for biopsy. Further investigations on the reference material or a prospective reference method, with a final aim of traceability for all tpsa and fpsa assays used, are needed to improve this difficult situation for clinicians. Currently, the use of different tpsa and fpsa assays with fixed thresholds leads to serious problems and large differences when deciding to take a prostate biopsy or not. ACKNOWLEDGEMENTS We especially thank Janett Reiche for her valuable technical assistance. This work was supported by the Deutsche Krebshilfe (Grant ST1), ProFIT (No ) the Berliner Sparkassenstiftung Medizin, and by the SONNENFELD-Stiftung. CONFLICT OF INTEREST None declared. REFERENCES 1 Polascik TJ, Oesterling JE, Partin AW. Prostate specific antigen: a decade of discovery what we have learned and where we are going. J Urol 1999; 162: Lilja H, Christensson A, Dahlen U et al. Prostate-specific antigen in serum occurs predominantly in complex with α1- antichymotrypsin. Clin Chem 1991; 37: Stenman UH, Leinonen J, Alfthan H, Rannikko S, Tuhkanen K, Alfthan O. A complex between prostate-specific antigen and α1-antichymotrypsin is the major form of prostate-specific antigen in serum of patients with prostatic cancer: assay of the complex improves clinical sensitivity for cancer. Cancer Res 1991; 51: Semjonow A, De Angelis G, Oberpenning F, Schmid HP, Brandt B, Hertle L. The clinical impact of different assays for prostate specific antigen. BJU Int 2000; 86: Blijenberg BG, Yurdakul G, Van Zelst BD, Bangma CH, Wildhagen MF, Schroder FH. Discordant performance of assays for free and total prostate-specific antigen in relation to the early detection of prostate cancer. BJU Int 2001; 88: Semjonow A, Oberpenning F, Brandt B, Zechel C, Brandau W, Hertle L. Impact of free prostate-specific antigen on discordant measurement results of assays for total prostate-specific antigen. Urology 1996; 48 (Suppl.): Stamey TA, Chen Z, Prestigiacomo AF. Reference material for PSA: the IFCC standardization study. International Federation of Clinical Chemistry. Clin Biochem 1998; 31: Rafferty B, Rigsby P, Rose M, Stamey T, Gaines Das R. Reference reagents for prostate-specific antigen (PSA): establishment of the first international standards for free PSA and PSA (90: 10). Clin Chem 2000; 46: Semjonow A, Oberpenning F, Weining C et al. Do modifications of nonequimolar assays for total prostate-specific antigen improve detection of prostate cancer? Clin Chem 2001; 47: Semjonow A, Brandt B, Oberpenning F, Roth S, Hertle L. Discordance of assay methods creates pitfalls for the interpretation of prostate-specific antigen values. Prostate Suppl 1996; 7: Link RE, Shariat SF, Nguyen CV et al. Variation in prostate specific antigen results from 2 different assay platforms: clinical impact on 2304 patients undergoing prostate cancer screening. J Urol 2004; 171: Patel D, White PA, Milford WA. A comparison of six commercial assays for total and free prostate specific antigen (PSA): the predictive value of the ratio of free to total PSA. BJU Int 2000; 85: Kort SA, Martens F, Vanpoucke H, van Duijnhoven HL, Blankenstein MA. Comparison of 6 automated assays for total and free prostate-specific antigen with special reference to their reactivity toward the WHO 96/670 reference preparation. Clin Chem 2006; 52: Roddam AW, Rimmer J, Nickerson C, Ward AM; NHS Prostate Cancer Risk Management Programme. Prostatespecific antigen: bias and molarity of commercial assays for PSA in use in England. Ann Clin Biochem 2006; 43: Stephan C, Klaas M, Muller C, Schnorr D, Loening SA, Jung K. Interchangeability of measurements of total and free prostate-specific antigen in serum with 5 frequently used assay combinations: an update. Clin Chem 2006; 52: Kairisto V, Poola A. Software for illustrative presentation of basic clinical characteristics of laboratory tests GraphROC for Windows. Scand J Clin Lab Invest Suppl 1995; 222: Jung K, Stephan C, Lein M et al. Analytical performance and clinical validity of two free prostate-specific antigen assays compared. Clin Chem 1996; 42: Schröder FH, van der Cruijsen-Koeter I, de Koning HJ, Vis AN, Hoedemaker RF, Kranse R. Prostate cancer detection at low prostate specific antigen. J Urol 2000; 163: BJU INTERNATIONAL

5 CLINICAL IMPACT OF DIFFERENT PSA ASSAYS 19 Catalona WJ, Partin AW, Finlay JA et al. Use of percentage of free prostatespecific antigen to identify men at high risk of prostate cancer when PSA levels are ng/ml and digital rectal examination is not suspicious for prostate cancer: an alternative model. Urology 1999; 54: Thompson IM, Pauler DK, Goodman PJ et al. Prevalence of prostate cancer among men with a prostate-specific antigen level < or =4.0 ng per milliliter. N Engl J Med 2004; 350: Catalona WJ, Richie JP, Ahmann FR et al. Comparison of digital rectal examination and serum prostate specific antigen in the early detection of prostate cancer: results of a multicenter clinical trial of 6630 men. J Urol 1994; 151: Stephan C, Semjonow A. Accuracy assessment of PSA methods must be based on clinical instead of artificial samples a response to AW Roddam et al. Ann Clin Biochem 2006; 43: Ann Clin Biochem 2006; 43: Catalona WJ, Partin AW, Slawin KM et al. Use of the percentage of free prostate-specific antigen to enhance differentiation of prostate cancer from benign prostatic disease: a prospective multicenter clinical trial. JAMA 1998; 279: Correspondence: Carsten Stephan, Department of Urology, Charité Universitätsmedizin Berlin, CCM, Charitéplatz 1, D Berlin, Germany. carsten.stephan@charite.de Abbreviations: ROC, receiver operating characteristic (curve); AUC, area under ROC curve; fpsa, free PSA; tpsa, total PSA; cpsa, complexed PSA; PSA, prostate specific antigen; %fpsa, percent free PSA BJU INTERNATIONAL 1431