ML: VALUE OF PERCENT FREE PROSTATE SPECIFIC ANTIGEN ON TUMOR DETECTION AND TUMOR AGGRESSIVENESS

0022-5347/04/1716-2245/0 THE JOURNAL OF UROLOGY® Copyright © 2004 by AMERICAN UROLOGICAL ASSOCIATION

Vol. 171, 2245–2249, June 2004 Printed in U.S.A.

DOI: 10.1097/01.ju.0000127731.56103.50

PROSTATE CANCER DETECTION IN THE PROSTATE SPECIFIC ANTIGEN RANGE OF 2.0 TO 3.9 NG/ML: VALUE OF PERCENT FREE PROSTATE SPECIFIC ANTIGEN ON TUMOR DETECTION AND TUMOR AGGRESSIVENESS ´ RAAIJMAKERS,* BERT G. BLIJENBERG, JUDITH A. FINLAY,† HARRY G. RITTENHOUSE,† RENE ¨ DER MARK F. WILDHAGEN, MONIQUE J. ROOBOL AND FRITZ H. SCHRO From the Departments of Urology (RR, MFW, MJR, FHS) and Clinical Chemistry (BGB), Erasmus MC, Rotterdam, The Netherlands, and Hybritech, Inc., a subsidiary of Beckman Coulter, Inc. (JAF, HGR), San Diego, California

ABSTRACT

Purpose: We evaluated the positive predictive value and cancer detection rate in the prostate specific antigen (PSA) range of 2.0 to 3.9 ng/ml and assessed the value of percent free (F) PSA (FPSA) on tumor detection and tumor aggressiveness in this low PSA range. Materials and Methods: Of 3,623 men who were attending the second round of screening within the European Randomized Study of Screening for Prostate Cancer, section Rotterdam 883 had PSA values of 2.0 to 3.9 ng/ml. These men were offered laterally directed sextant biopsy. FPSA was prospectively determined from pretreatment serum. Cancers were classified as prognostically favorable and unfavorable using biopsy results and other pretreatment diagnostic features. Results: Using the PSA range of 2.0 to 3.9 ng/ml as a biopsy indication 126 cancers were detected, resulting in a positive predictive value of 17.1% and a cancer detection rate of 14.3%. By using percent FPSA and setting relative sensitivity at 95% 9% of biopsies could have been avoided. Unfavorable tumor characteristics were found in 46.9% of the men with T1C tumors. Mean percent FPSA was significantly lower in such men compared to men with favorable tumor characteristics. Of the men with percent FPSA lower than 10% 90% had unfavorable tumor characteristics. Conclusions: The PSA range 2.0 to 3.9 ng/ml is accessible for prostate cancer screening. Percent FPSA is of moderate value in avoiding unnecessary biopsies in the PSA range of 2.0 to 3.9 ng/ml. However, when assessing tumor aggressiveness in biopsy results, percent FPSA is predictive and can be used to select treatment options, such as watchful waiting. KEY WORDS: prostate, prostatic neoplasms, prostate-specific antigen, mass screening, biopsy

Two important goals of the European Randomized Study of Screening for Prostate Cancer (ERSPC) are to establish or disprove the effect of screening on prostate cancer (PCa) mortality and evaluate screening tests. To decrease cancer specific mortality it is important to detect cancers that are curable and pose a potential threat to the life of the participant. By lowering the prostate specific antigen (PSA) threshold more curable cancers are detected and the lead time is assumed to be longer. On the other hand, the risk of diagnosing and treating cancers that may not pose a threat to patient life will be higher. Based on this background an evaluation and optimization of screening tests is desirable. Earlier studies showed a low predictive value of digital rectal examination (DRE) and/or transrectal ultrasonography (TRUS) for detecting prostate cancer in men with a PSA of less than 4.0 ng/ml.1 Also, logistic regression analysis revealed that DRE and/or TRUS misses 63.5% of prostate can-

cers in this PSA range. Of the cancers found in this PSA range on abnormal DRE and/or TRUS alone 84% were histologically organ confined and approximately half of the tumors with a PSA lower than 4.0 ng/ml had aggressive characteristics on radical prostatectomy specimens. We report the results of a side study in the second round of the Rotterdam section of the ERSPC in which each consecutive participant with a PSA of 2.0 ng/ml or greater had a biopsy indication. We report the cancer detection rate (CDR) and positive predictive value (PPV) in this low PSA range. The value of percent free PSA (FPSA) for tumor detection and tumor aggressiveness were tested within the PSA range of 2.0 to 3.9 ng/ml. To our knowledge this is the first study to evaluate prospectively the detection rates and value of percent FPSA in these low PSA ranges in a large number of consecutive participants in the setting of a randomized, controlled trail.

Accepted for publication January 16, 2004. Study received Dutch law on population screening approval. PARTICIPANTS AND METHODS Supported by Grants EUR-94-869 and EUR-98-1757 from the Dutch Cancer Society, Grants 002-22820 and 2000-2-1016 from The The ERSPC, section Rotterdam. Within the Dutch (RotterNetherlands Organization for Health Research and Development dam) section of the ERSPC 42,376 men 55 to 74 years old (ZONMw), 5th Framework Program Grant QLRI-2000-01741 from were randomized into a screening and a control group. After the European Union, Europe Against Cancer and a grant from Hya number of pilot studies2 the study started in June 1994. britech, Inc., a subsidiary of Beckman Coulter, Inc. * Correspondence: Department of Urology, Room Z841, Erasmus Participants in the screening group were offered PSA measMC, P. O. Box 2040, 3000 CA Rotterdam, The Netherlands (tele- urement, DRE and TRUS. Initially the biopsy indication was phone: ⫹31 6 21680899; FAX: ⫹31 10 4633968; e-mail: based on PSA 4.0 ng/ml or greater, or suspicious DRE or [email protected]). 3 † Financial interest and/or other relationship with Beckman- TRUS. Based on predictions by logistic regression, the European study group decided in February 1997 to screen using Coulter. 2245

2246

PROSTATE CANCER DETECTION AND PROSTATE SPECIFIC ANTIGEN

PSA only and to perform biopsy in all men with a PSA of 3.0 ng/ml or greater. After a screening interval of 4 years all participants in the described age group were re-invited for the second round of screening. PSA measurement was done and when PSA was 3.0 ng/ml or greater, DRE, TRUS and biopsy were performed. The PSA 2.0 to 3.9 ng/ml study. The protocol of this side study was commenced in April 2001 within the second round of screening. During the first round 92.8% of these men were screened by a biopsy indication of PSA 3.0 ng/ml or greater. The remaining 63 men (7.1%) had a biopsy indication when PSA was 4.0 ng/ml or greater, DRE or TRUS was suspicious. Prospectively all consecutive participants presenting with a PSA of 2.0 ng/ml or greater were offered DRE, TRUS and laterally directed sextant biopsy. Within the PSA range of 2.0 to 3.9 ng/ml additional FPSA values were determined. Percent FPSA was calculated by dividing FPSA by total PSA values. All participants in this side study provided an informed consent form. CDRs and PPVs were determined within this PSA range. The CDR was defined as the proportion of cancers of all those screened in the PSA range. The PPV was defined as the number of men who had a positive test and cancer divided by those with a positive test and biopsy. A positive test in this study was a PSA of 2.0 ng/ml or greater. If in a given subgroup all men underwent biopsy, the PPV and CDR were identical. After blood samples were obtained serum samples were processed and refrigerated within 3 hours. Serum samples not analyzed the same day were stored at ⫺70C until analyzed. Total and free Hybritech PSA was measured on an Access immunoanalyzer (Beckman Coulter, Inc., San Diego, California). Pathological classification. The features of aggressiveness in biopsy specimens were categorized according to Epstein4 and Carter5 et al. Unfavorable tumor characteristics were defined as PSA density (PSAD) greater than 0.1 ng/ml/gm, Gleason score 7 or greater, 3 or greater tumor positive biopsy cores and/or more than 50% tumor involvement in at least 1 core. Statistical analyses. Statistical analyses were done using commercially available software. ROC curves were constructed using Analyse-it software (Analyse-It Software, Ltd., Leeds, United Kingdom). Mean values were compared with the t test for normally distributed parameters and the Mann-Whitney U test for nonnormal parameters. The predictive accuracy of the tests was assessed by ROC analyses. Multivariate logistic regression was performed using all available relevant parameters in a backward method.

Table 1 lists the number of detected prostate cancers, and the PPV and CDR per PSA range. The PPV in the PSA ranges 2.0 to 2.9 and 3.0 to 3.9 ng/ml was 15.7% and 19.8%, respectively. Similar results were obtained when the 63 men with a different biopsy indication in the first round were left out of the analyses. Tumor detection. Table 2 shows the characteristics of the men with biopsy in the PSA range 2.0 to 3.9 ng/ml. Mean percent FPSA in men with prostate cancer detected was 17.0% vs 19.7% for men with a negative biopsy. This difference was significant (p ⬍0.001). Mean age and PSA did not differ significantly in this population. Prostate volume was significantly larger in men with no prostate cancer detected. Table 3 shows the PPV of tumor detection in the PSA range 2.0 to 3.9 ng/ml, subdivided for the different percent FPSA ranges. In 2 men with biopsy FPSA was not determined, leaving 734. Lower percent FPSA ranges were significantly associated with a higher PPV for tumor detection (chi-square test p ⬍0.001). ROC curves showed an area under the curve (AUC) for percent FPSA of 62% compared with 55% for total PSA (fig. 1). This indicates slightly better discrimination between men with and without prostate cancer than total PSA alone in this PSA range. Only approximately 9% of the biopsies could have been avoided, detecting 95% of biopsy detectable tumors. On multivariate logistic regression percent FPSA (p ⫽ 0.001), FPSA (p ⫽ 0.006), age (p ⫽ 0.014) and prostate volume (p ⫽ 0.020) were the most important predictors for detecting prostate cancer. Tumor aggressiveness. As described, the classification of aggressiveness in biopsy specimens according to Epstein et al4 was validated in T1C tumors. Of the 126 prostate cancers found in this PSA range 104 (82.5%) were found to involve a nonpalpable (T1C) tumor. Detailed biopsy results were available on 98 participants. A total of 52 men (53.1%) were classified as having favorable tumor characteristics and 46 (46.9%) had unfavorable tumor characteristics. Mean percent FPSA was significantly lower in men with unfavorable tumor characteristics vs men with favorable tumor characteristics (15.4% vs 19.1%, p ⫽ 0.006). Table 4 shows that 9 of the 10 men (90%) with percent FPSA below 10% indeed had unfavorable tumor characteristics and the percent with unfavorable tumor characteristics decreased at higher percent FPSA ranges (p ⫽ 0.008). The AUC of ROC curves for percent FPSA for discriminating favorable and unfavorable tumor characteristics was 66% (fig. 2). On multivariate logistic regression percent FPSA (p ⬍0.001), FPSA (p ⫽ 0.005), age (p ⫽ 0.013) and prostate volume (p ⫽ 0.020) were the most important predictors of unfavorable tumor characteristics prior to biopsy.

RESULTS

A total of 3,623 men were screened within the protocol of this side study from April 2001 to December 2002. In 64% of these men PSA was lower than 2.0 ng/ml. Of the men 883 (24.4%) had PSA within the range 2.0 to 3.9 ng/ml and 419 (11.6%) had PSA greater than 4.0 ng/ml. The incidence of men with biopsy in the PSA range 2.0 to 3.9 ng/ml was 83.4%. This did not differ significantly from the higher PSA ranges.

TABLE 2. Age, prostatic volume, total PSA, FPSA and percent FPSA in men biopsied in PSA range 2.0 to 3.9 ng/ml Parameter

Mean No PCa

Mean PCa

p Value

No. men Age Prostate vol (ml) Total PSA (ng/ml) FPSA (ng/ml) % FPSA

610 66.3 43.0 2.7 0.53 19.7

126 67.0 35.0 2.8 0.48 17.0

Not significant ⬍0.001 Not significant 0.005 ⬍0.001

TABLE 1. PPV and CDR per PSA range TABLE 3. PPV per percent FPSA range on tumor detection

PSA Range (ng/ml)

No. Men/No. Biopsied (%)

No. PCa

% PPV

% CDR

2.0–2.4 2.5–2.9 2.0–2.9 3.0–3.4 3.5–3.9 3.0–3.9

330/272 (82.4) 246/206 (83.7) 576/478 (83) 173/143 (82.7) 134/115 (85.8) 307/258 (84)

41 34 75 29 22 51

15.1 16.5 15.7 20.3 19.1 19.8

12.4 13.8 13.0 16.8 16.4 16.6

883/736 (83.4)

126

17.1

14.3

Overall

% FPSA Less than 10 10–15 15–20 20–25 25 or Greater

No. Men Biopsied (%) 38 168 227 176 125

(5.2) (22.9) (30.9) (24.0) (17.0)

Overall 734 (100) Chi-square test p ⬍0.001.

No. PCa

% PPV

13 39 39 20 15

34.2 23.2 17.2 11.4 12.0

126

17.1

PROSTATE CANCER DETECTION AND PROSTATE SPECIFIC ANTIGEN

2247

FIG. 1. ROC curves for tumor detection in 734 patients (AUC PSA) 55% and AUC percent (FPSA 62%) TABLE 4. Distribution of men with unfavorable tumor characteristics per percent FPSA range % FPSA

No. PCa

No. Unfavorable (%)

Less than 10 10–15 15–20 20–25 25 or Greater

10 28 32 14 14

9 (90) 15 (53.6) 14 (43.8) 4 (28.6) 4 (28.6)

98

46 (46.9)

Totals Chi-square test p ⫽ 0.008.

DISCUSSION

The use of a lower PSA cutoff as a biopsy indication will inevitably result in a decrease in the PPV and relative sensitivity, resulting in a larger proportion of men who undergo biopsy unnecessarily. Another concern and open question is the possible detection of more clinically insignificant tumors. This problem is addressed by classifying the detected cancers into 2 groups with a favorable and unfavorable prognosis, respectively. We also tested the value of percent FPSA for discriminating men with favorable and unfavorable tumor characteristics. PPV and CDR in the PSA range 2.0 to 3.9 ng/ml. The definitions of PPV and CDR allow the differentiation of men who had a biopsy indication and the 83.4% who in fact underwent biopsy. The PPV and CDR would have been identical if all men had undergone biopsy. An accurate determination of PPV and the relative sensitivity of a test is only possible in a cohort of individuals who have undergone appropriate testing.6, 7 However, in this prospective biopsy study the proportion of men who did not undergo testing was relatively small and the results can be considered reasonably accurate with respect to the studied PSA ranges. Our data show that the PSA range 2.0 to 3.9 ng/ml is accessible for cancer screening with a total PPV of 17.1% resulting from laterally directed sextant biopsies. This translates into 5.8 biopsies needed to detect 1 prostate cancer. Table 1 shows the PPV and CDR per PSA range in steps of 0.5 ng/ml. Based on these percents one can choose cutoff values as they may appear useful.

Earlier ERSPC related studies showed that percent FPSA can decrease the rate of unnecessary biopsies in the PSA range of 4.0 to 10.0 ng/ml.8, 9 Similar results in this PSA range were reported in other studies.10, 11 It was subject to this study whether similar improvement in relative specificity could be achieved in the lower PSA ranges. Unfortunately the improvement seen was only marginal. By setting relative sensitivity at 95% (allowing 5% of cancers to be missed) only 9% of biopsies could be saved, as shown by ROC analysis (fig. 1). Others evaluated the value of percent FPSA in the PSA range 2.6 to 4.0 ng/ml.12–15 These groups also reported a significantly different mean percent PSA value in men with and without prostate cancer. This translated into a moderate gain in omitting unnecessary biopsies in this PSA range and the AUC of the respective ROC curves was between 58.5% and 74.9%. However, in these studies not all men at risk underwent biopsy. While in our study mean percent FPSA differed significantly between cancer and noncancer cases, and although there was discrimination on multivariate analysis in relation to tumor detection, percent FPSA adds only little to total PSA in improving relative specificity (saving unnecessary biopsies). When interpreting our results, it must be considered that this study was done in a second round of screening. Test characteristics changed considerably for tests that were used during the first round.16 As mentioned, 92.9% of all men underwent biopsy during the first round if they presented with a PSA of 3.0 ng/ml or higher. It is impossible to correct retrospectively for this problem and relate our findings to a previously unscreened population. Still, the indication for biopsy in men with a PSA of 3.0 to 3.9 ng/ml in the first round probably decreased the PPV of this range and of higher PSA values in the second round. In men in this study with a PSA of 2.0 to 2.9 ng/ml no systemic biopsies were done in round 1. Hence, this PSA range was not influenced in this way and it may in fact reflect the true PPV and CDR of a previously unscreened population. Favorable and unfavorable tumor characteristics. To deter-

2248

PROSTATE CANCER DETECTION AND PROSTATE SPECIFIC ANTIGEN

FIG. 2. ROC curve for percent FPSA on tumor aggressiveness in 98 patients (AUC 66%)

mine tumor aggressiveness we used biopsy results and other pretreatment diagnostic findings. Although it might be more reliable to use data resulting from the pathological evaluation of radical prostatectomy specimens, the evaluation of pretreatment parameters is crucial for the development of future treatment strategies. We used the classification of biopsy results for T1C tumors according to Epstein et al,4 which was validated using radical prostatectomy specimens. To be able to differentiate favorable and unfavorable tumors we applied the results of the prospective evaluation of T1C cancers, as described by Epstein4 and Carter5 et al. We used the classification in which PSAD less than 0.1 ng/ml/gm is considered predictive of favorable tumor characteristics instead of PSAD less than 0.1 ng/ml/gm and PSAD between 0.1 and 0.15 ng/ml/gm with cancer less than 3 mm in 1 biopsy core. We find this classification easier to use and more reproducible in the clinical setting without compromising the PPV and negative predictive value too much. The PPV and negative predictive value of predicting pathological features of radical prostatectomy specimens through biopsy results were 92% and 63%, respectively. We then explored the relationship of percent FPSA to cancers classified as favorable and unfavorable. The data showed us the distribution of unfavorable features within 5 groupings of percent FPSA (table 4). It can be seen that 9 of 10 men with percent FPSA below 10% had unfavorable tumor characteristics. Their proportion decreased steadily with increasing percent FPSA. This parameter may be useful to select and exclude patients for watchful waiting and other treatment regimens. The possible role of percent FPSA for selecting treatment options should be further investigated, preferably with respect to simpler and more reliable prognostic factors that may become available in the future. Further improvement may also be possible by replacing percent FPSA by one of the different molecular forms of PSA, which are currently under study. Recent data suggest the possibility that 2 or ⫺7 pro-PSA could be an active component within FPSA.17, 18 Further studies are needed to investigate the usefulness of these molecular forms of PSA for improving the test characteristics of PSA 2.0 to 3.9 ng/ml and of tumor aggressiveness parameters.

CONCLUSIONS

The PPV and CDR in the PSA range 3.0 to 3.9 ng/ml as a biopsy indication are comparable to those of higher PSA ranges in the second round of screening. Still, 15.7% of with men biopsy in the PSA range 2.0 to 2.9 ng/ml had biopsy detectable prostate cancer. Percent FPSA is of moderate value for improving test characteristics in the PSA range 2.0 to 3.9 ng/ml. However, when assessing tumor aggressiveness in biopsy results, percent FPSA is predictive and can be used to select treatment options. REFERENCES

1. Schro¨ der, F. H., van der Cruijsen-Koeter, I., de Koning, H. J., Vis, A. N., Hoedemaeker, R. F. and Kranse, R.: Prostate cancer detection at low prostate specific antigen. J Urol, 163: 806, 2000 2. Schroder, F. H., Denis, L. J., Kirkels, W., de Koning, H. J. and Standaert, B.: European randomized study of screening for prostate cancer. Progress report of Antwerp and Rotterdam pilot studies. Cancer, 76: 129, 1995 3. Kranse, R., Beemsterboer, P., Rietbergen, J., Habbema, D., Hugosson, J. and Schroder, F. H.: Predictors for biopsy outcome in the European Randomized Study of Screening for Prostate Cancer (Rotterdam region). Prostate, 39: 316, 1999 4. Epstein, J. I., Walsh, P. C., Carmichael, M. and Brendler, C. B.: Pathologic and clinical findings to predict tumor extent of nonpalpable (stage T1c) prostate cancer. JAMA, 271: 368, 1994 5. Carter, H. B., Sauvageot, J., Walsh, P. C. and Epstein, J. I.: Prospective evaluation of men with stage T1C adenocarcinoma of the prostate. J Urol, 157: 2206, 1997 6. Punglia, R. S., D’Amico, A. V., Catalona, W. J., Roehl, K. A. and Kuntz, K. M.: Effect of verification bias on screening for prostate cancer by measurement of prostate-specific antigen. N Engl J Med, 349: 335, 2003 7. Schroder, F. H. and Kranse, R.: Verification bias and the prostate-specific antigen test—is there a case for a lower threshold for biopsy? N Engl J Med, 349: 393, 2003 8. Bangma, C. H., Kranse, R., Blijenberg, B. G. and Schroder, F. H.: The value of screening tests in the detection of prostate cancer. Part I: results of a retrospective evaluation of 1726 men. Urology, 46: 773, 1995

PROSTATE CANCER DETECTION AND PROSTATE SPECIFIC ANTIGEN 9. Bangma, C. H., Rietbergen, J. B. W., Kranse, R., Blijenberg, B. G., Petterson, K. and Schro¨ der, F. H.: The free-to-total prostate specific antigen ratio improves the specificity of prostate specific antigen in screening for prostate cancer in the general population. J Urol, 157: 2191, 1997 10. Partin, A. W., Catalona, W. J., Southwick, P. C., Subong, E. N., Gasior, G. H. and Chan, D. W.: Analysis of percent free prostate-specific antigen (PSA) for prostate cancer detection: influence of total PSA, prostate volume, and age. Urology, 48: 55, 1996 11. Gann, P. H., Ma, J., Catalona, W. J. and Stampfer, M. J.: Strategies combining total and percent free prostate specific antigen for detecting prostate cancer: a prospective evaluation. J Urol, 167: 2427, 2002 12. Catalona, W. J., Smith, D. S. and Ornstein, D. K.: Prostate cancer detection in men with serum PSA concentrations of 2.6 to 4.0 ng/mL and benign prostate examination. Enhancement of specificity with free PSA measurements. JAMA, 277: 1452, 1997 13. Catalona, W. J., Partin, A. W., Finlay, J. A., Chan, D. W., Rittenhouse, H. G., Wolfert, R. L. et al: Use of percentage of free prostate-specific antigen to identify men at high risk of prostate cancer when PSA levels are 2.51 to 4 ng/mL and

14.

15.

16.

17.

18.

2249

digital rectal examination is not suspicious for prostate cancer: an alternative model. Urology, 54: 220, 1999 Roehl, K. A., Antenor, J. A. V. and Catalona, W. J.: Robustness of free prostate specific antigen measurements to reduce unnecessary biopsies in the 2.6 to 4.0 ng./ml. range. J Urol, 168: 922, 2002 Djavan, B., Zlotta, A., Kratzik, C., Remzi, M., Seitz, C., Schulman, C. C. et al: PSA, PSA density, PSA density of transition zone, free/total PSA ratio, and PSA velocity for early detection of prostate cancer in men with serum PSA 2.5 to 4.0 ng/mL. Urology, 54: 517, 1999 Raaijmakers, R., Wildhagen, M. F., Ito, K., Paez, A., de Vries, S. H., Roobol, M. J. et al: Prostate-specific antigen change in the European Randomized Study of Screening for Prostate Cancer section Rotterdam. Urology, 63: 316, 2004 Mikolajczyk, S. D., Marks, L. S., Partin, A. W. and Rittenhouse, H. G.: Free prostate-specific antigen in serum is becoming more complex. Urology, 59: 797, 2002 Sokoll, L. J., Chan, D. W., Mikolajczyk, S. D., Rittenhouse, H. G., Evans, C. L., Linton, H. J. et al: Proenzyme psa for the early detection of prostate cancer in the 2.5– 4.0 ng/ml total psa range: preliminary analysis. Urology, 61: 274, 2003