PERCENT FREE PROSTATE SPECIFIC ANTIGEN IS NOT AN INDEPENDENT PREDICTOR OF ORGAN CONFINEMENT OR PROSTATE SPECIFIC ANTIGEN RECURRENCE IN UNSCREENED PATIENTS WITH LOCALIZED PROSTATE CANCER TREATED WITH RADICAL PROSTATECTOMY

0022-5347/02/1673-1306/0 THE JOURNAL OF UROLOGY® Copyright © 2002 by AMERICAN UROLOGICAL ASSOCIATION, INC.®

Vol. 167, 1306 –1309, March 2002 Printed in U.S.A.

PERCENT FREE PROSTATE SPECIFIC ANTIGEN IS NOT AN INDEPENDENT PREDICTOR OF ORGAN CONFINEMENT OR PROSTATE SPECIFIC ANTIGEN RECURRENCE IN UNSCREENED PATIENTS WITH LOCALIZED PROSTATE CANCER TREATED WITH RADICAL PROSTATECTOMY MARKUS GRAEFEN, PIERRE I. KARAKIEWICZ, ILIAS CAGIANNOS, PETER G. HAMMERER, ¨ RI PALISAAR, EDITH HULAND, PETER T. SCARDINO, ALEXANDER HAESE, JU MICHAEL W. KATTAN AND HARTWIG HULAND From the Departments of Urology, University Hospital Hamburg-Eppendorf, Germany, and Memorial Sloan-Kettering Cancer Center, New York, New York

ABSTRACT

Purpose: We studied the controversial relationship of percent free prostate specific antigen (PSA) with organ confined prostate cancer and PSA failure after radical prostatectomy. Materials and Methods: We tested the characteristics of the percent free PSA monoclonal Immulite DPC Immunoassay (Diagnostic Products Corp., Los Angeles, California) for predicting organ confinement in 698 consecutive unscreened men treated only with radical prostatectomy between 1995 and 2000. In addition, we assessed the ability of percent free PSA to predict post-radical prostatectomy PSA failure, defined as PSA 0.1 ng./ml. or greater, in a subset of 581 men in whom followup was available. All statistical analyses were repeated for stage T1c cancer with PSA between 2 and 10 ng./ml. Results: On univariate analyses percent free PSA achieved significance for predicting organ confined disease at all clinical stages (p ⬍0.001) and for stage T1c cancer with PSA between 2 and 10 ng./ml. (p ⫽ 0.012). However, significance dissipated after controlling for total PSA, biopsy Gleason sum and clinical stage (p ⫽ 0.135 and 0.851, respectively). In univariate Cox models percent free PSA failed to predict PSA failure across stages (p ⫽ 0.341), as well as in stage T1c cancer (p ⫽ 0.93). In multivariate analyses controlling for traditional PSA biopsy grade and clinical stage (p ⬍0.001), percent free PSA failed to contribute to predicting PSA failure (p ⫽ 0.342). In the stage T1c subset biopsy Gleason sum (p ⬍0.001) and PSA (p ⫽ 0.018) remained significant, in contrast to percent free PSA (p ⫽ 0.237). Conclusions: Percent free PSA has no independent, statistically significant association with organ confined status or posttreatment PSA outcome in unscreened patients who undergo radical prostatectomy for localized prostate cancer. KEY WORDS: prostate, prostatic neoplasms, prostate-specific antigen, prostatectomy

Percent free prostate specific antigen (PSA) is becoming an established clinical tool to aid in the detection of prostate cancer. Percent free PSA enhances the specificity of PSA and is clinically valuable for distinguishing cancer from benign prostate hyperplasia (BPH).1, 2 However, controversy exists regarding its usefulness for predicting pathological stage or prognosis after treatment for clinically localized prostate cancer. While some studies have shown an association of lower percent free PSA with higher stage and grade, others have failed to indicate such an association.3–12 These series focused on pathological findings after radical prostatectomy. To our knowledge no study to date has examined the relationship of percent free PSA with outcome. We investigated the association of preoperative percent free PSA with organ confined disease and tumor progression after radical prostatectomy. MATERIALS AND METHODS

Our study included 698 consecutive patients with clinically localized prostate cancer who were treated with radical pros-

tatectomy only at the University Hospital HamburgEppendorf. Before prostatic manipulation serum was obtained for PSA testing. Total and free PSA was determined by the monoclonal Immulite DPC immunoassay. The ratio of free-to-total PSA was calculated using the formula, (free PSA/total PSA) ⫻ 100. Free and total PSA were measured simultaneously on the day the blood samples were obtained. Samples were immediately stored at 4C after drawing for a maximum of 4 hours. In these men complete information on preoperative free and total PSA, clinical stage and biopsy Gleason grade were available for analysis. Patients who received hormonal therapy before radical retropubic prostatectomy were excluded from the study. In 581 cases followup information was available. Cases of positive lymph nodes and consecutively abandoned radical prostatectomy were considered immediate failures. In patients who underwent radical prostatectomy a postoperative PSA level of 0.1 ng./ml. or greater was considered as biochemical recurrence. All radical retropubic prostatectomy specimens were prospectively processed according to the Stanford protocol.13 After surface marking and fixation the prostate was sectioned at 3 mm. and paraffin embedded. Slides (5 ␮m.) were cut and stained with hematoxylin and eosin. For histological

Accepted for publication October 5, 2001. Supported by Deutsche Forschungsgemeinschaft GR 1866/1–1, Deutsche Krebshilfe, American Foundation for Urologic Diseases, National Cancer Institute of Canada and Medical Research Council of Canada. 1306

1307

PERCENT FREE PROSTATE SPECIFIC ANTIGEN AND PROSTATE CANCER

grading we used the Gleason system14 and for pathological staging we used the TNM classification.15 The association of percent free PSA with organ confined disease was assessed by univariate analyses using logistic regression models. Analyses were performed across all clinical stages, and separately in 301 men with clinical stage T1c disease and a total PSA 2 to 10 ng./ml. In addition, a multivariate model was constructed for all clinical stages and for stage T1c cancer with total PSA between 2 and 10 ng./ml. This clinical stage was analyzed separately because there is evidence in the literature that percent free PSA adds independent predictive information in clinical subgroups.3 Furthermore, free PSA is used to improve diagnostic ability for detecting prostate cancer, especially in this subgroup of patients. Information on outcome after radical prostatectomy was available on 581 men and the association of percent free PSA with outcome, defined as biochemical recurrence, was determined in this group. Separate analysis was performed again in men with clinical stage T1c cancer and total PSA between 2 and 10 ng./ml. For univariate comparison of preoperative percent free PSA and outcome a univariate Cox model was constructed. All analyses were performed in a multivariate fashion controlling for total PSA, biopsy Gleason grade and clinical stage. To assess accuracy of a model including traditional total PSA, clinical stage and biopsy Gleason grade with and without percent free PSA we performed receiver operating characteristics (ROC) curve analyses. Similar analyses were performed using PSA recurrence as an end point. For censored data, the ROC calculation is modified from its conventional form. Instead Harrell’s transformation was calculated.16 Its interpretation is similar to conventional ROC analysis. For statistical analysis we pooled patients across pathological stages into organ confined (pT2a and pT2b), and nonorgan confined disease (pT3a to pT4 or positive lymph nodes) to avoid forming categories with small numbers. Similarly, biopsy Gleason sum in the predictive models was classified as 6 or less and 7 or greater. RESULTS

In 687 patients intraoperative frozen section revealed negative lymph nodes and radical prostatectomy was performed. Organ confined disease was detected in 418 men (60.8%). Followup information was available on 581 men. Followup was 1 to 46 months (mean 131 months). In 11 men radical prostatectomy was abandoned due to positive frozen section. These cases were considered immediate failures in the Cox models addressing outcome. Of the cases 85 (14.6%) failed after radical prostatectomy, 74 (87.1%) were PSA failures and 11 (12.9%) were considered failures due to lymph node metastasis. Table 1 lists the clinical and pathological characteristics of the patients across all clinical stages and separately for clinical stage T1c cancer with a total PSA between 2 and 10 ng./ml. Percent free PSA achieved significance for predicting organ confined disease across all stages (odds ratio 0.957, p ⫽ 0.006) and for clinical stage T1c cancer with total PSA between 2 and 10 ng./ml. (odds ratio 0.957, p ⫽ 0.013) on univariate logistic regression analysis. Significance dissipated after controlling for traditional PSA, Gleason sum and clinical stage (table 2). In univariate Cox models percent free PSA failed to predict PSA failure across stages (odds ratio 0.988, p ⫽ 0.342), as well as for clinical stage T1c cancer with total PSA between 2 and 10 ng./ml. (odds ratio 0.972, p ⫽ 0.93). On multivariate analysis when controlling for traditional PSA (p ⬍ 0.004), biopsy grade (p ⬍ 0.001) and clinical stage (p ⬍0.001), percent free PSA failed to contribute to predicting PSA failure (table 3). In the stage T1c subset biopsy Gleason sum (p ⬍0.001) and PSA (p ⫽ 0.018) remained significant, in contrast to percent free PSA (p ⫽ 0.237). Tables 2 and 3 show odds ratios for the logistic regression models predicting organ confined disease and Cox

TABLE 1. Clinical characteristics of the 698 patients Characteristics

No. Pts. (%)

Clinical stage: T1c T2a T2b T2c T3a PSA (ng./ml.): 0–4.0 4.1–10 10.1–20 Greater than 20 % Free PSA (ng./ml.): 1–10 10.1–15 15.1–20 Greater than 20 Biopsy Gleason score: 5–6 7 8–10 Prostate specimen Gleason score: 2–4 5–6 7 8–10 Extracapsular extension: No Yes Seminal vesicle invasion: No Yes Surgical margin: Neg. Pos. Lymph node status: Neg. Pos.

416 (59.6) 105 (15) 115 (16.5) 28 (4) 28 (4) 64 (9.2) 375 (53.7) 167 (23.9) 92 (13.2) 325 (46.4) 190 (27.2) 89 (12.8) 94 (13.5) 447 (65.1) 215 (31.3) 25 (3.6) 1 (0.1) 299 (43.5) 371 (54) 16 (2.3) 418 (60.8) 269 (39.2) 569 (82.8) 118 (17.2) 551 (80.3) 135 (19.7) 655 (93.8) 43 (6.2)

models predicting outcome on multivariate analysis. In none of the scenarios did percent free PSA attain independent significant predictive ability. Furthermore, additional ROC curve analysis showed that the predictive accuracy of a model using traditional predictors of organ confined disease and outcome would not be enhanced by adding percent free PSA as a predictor variable. DISCUSSION

Immunoassays have been developed that provide sensitive and specific measurement of total and free PSA. In patients with prostate cancer a significantly higher proportion of serum PSA is complexed to ␣1-antichymotrypsin than in those with BPH. Several studies have shown that percent free PSA provides additional benefit over total PSA alone for differentiating prostate cancer from BPH.1 Catalona et al reported that a percent free PSA cutoff point of 25% detected 95% of cancers, while avoiding 20% of unnecessary biopsies.2 Others have had similar results using slightly different cutoff values and there is now general agreement that percent free PSA is useful for the early detection and diagnosis of prostate cancer.1 However, controversy exists on the usefulness of percent free PSA as a staging or prognostic tool for clinically localized prostate cancer. In our study we examined the relationship of percent free PSA with prostate pathology as well as with the prognosis after radical prostatectomy. We performed univariate and multivariate analyses in patients across all clinical stages as well as separate analyses in those with clinical stage T1c cancer and total PSA between 2 and 10 ng./ml. Univariate analyses revealed a significant relationship of percent free PSA with organ confinement regardless of whether all patients were considered or only those with clinical stage T1c disease. However, in all cases this relationship dissipated when percent free PSA was considered with biopsy Gleason sum, pretreatment PSA and clinical stage. Percent free PSA

1308

PERCENT FREE PROSTATE SPECIFIC ANTIGEN AND PROSTATE CANCER TABLE 2. Multivariate analysis of the association of pretreatment percent free PSA with organ confined prostate cancer p Value

Odds Ratio

p Value

Odds Ratio

All Ca stages: PSA ⬍0.0001 1.061 ⬍0.0001 1.057 Clinical stage ⬍0.0001 2.596 ⬍0.0001 2.642 Gleason sum ⬍0.0001 4.698 ⬍0.0001 4.588 % Free PSA – 0.135 0.98 Stage T1c Ca, PSA 2—10 ng./ml.: PSA 0.001 1.055 0.001 1.054 Gleason sum ⬍0.0001 5.366 ⬍0.0001 5.281 % Free PSA – 0.851 0.996 Univariate models tested the contribution of percent free PSA alone in predicting pathologically organ confined disease (pT2 versus pT3 or higher), whereas the multivariate model addressing the entire cohort tested the independent contribution of percent free PSA after controlling for total PSA, biopsy Gleason sum and clinical stage. The multivariate model addressing T1c cancers with PSA levels between 2 and 10 ng./ml. controlled for total PSA and biospy Gleason sum. The nomogram predictions were derived from logistic regression based coefficients. The nomogram addressing the entire cohort was based on the aggregate contribution of total PSA, clinical stage and biopsy Gleason sum. Its predictive accuracy, expressed under the area under the ROC curve, was compared to a nomogram that added percent free PSA to the other 3 predictors. The nomogram addressing T1c cancers with PSA levels between 2 and 10 ng./ml. was based on total PSA and biopsy Gleason sum. Its predictive accuracy was tested against nomogram adding percent free PSA to the other 2 predictors. * ROC 0.789, area under curve .79. † ROC 0.717, area under curve 0.716.

TABLE 3. Multivariate analysis of the association of pretreatment percent free PSA with biochemical recurrence after radical prostatectomy p Value

Odds Ratio

p Value

Odds Ratio

All Ca stages: PSA ⬍0.0001 1.032 ⬍0.0001 1.022 Clinical stage ⬍0.0001 2.12* ⬍0.0001 2.186 Gleason sum ⬍0.0001 5.296 ⬍0.0001 5.336 % Free PSA 0.342 0.982 Stage T1c Ca, PSA 2–10 ng./ml.: PSA 0.036 1.031 0.018 1.034 Gleason sum ⬍0.0001 8.78† ⬍0.0001 9.788 % Free PSA 0.237 1.045 Univariate models tested the contribution of percent free PSA alone in predicting PSA nonprogression rates, whereas the multivariate model addressing entire cohort tested the independent contribution of percent free PSA after controlling for total PSA, biopsy Gleason sum and clinical stage. Multivariate models addressing T1c cancers with PSA levels between 2 and 10 ng./ml. controlled for total PSA and biopsy Gleason sum. The nomogram addressing the entire cohort was based on the aggregate contribution of total PSA, clinical stage and biopsy Gleason sum. Its predictive accuracy, expressed under the concordance index, was compared to a nomogram that added percent free PSA to other 3 predictors. Nomogram addressing T1c cancers with PSA levels 2 to 10 ng./ml. was based on total PSA and biopsy Gleason sum. Its predictive accuracy was tested against nomogram adding percent free PSA to other 2 predictors. * ROC 0.858, concordance index 0.851. † ROC 0.795, concordance index 0.763.

was not an independent predictor of outcome after radical prostatectomy on all univariate and multivariate analyses. In addition to standard Cox multivariate analysis, we also determined the overall accuracy of 2 multivariate models for predicting organ confinement and disease outcome. Model 1 incorporated clinical stage, pretreatment PSA and biopsy Gleason sum, while model 2 contained the same 3 variables as well as percent free PSA. The areas under the curve in the 2 models were approximately equal for each end point regardless of whether all patients or only those with clinical stage T1c cancer and total PSA between 2 and 10 ng./ml. were considered. This observation indicates that no additional pathological or prognostic information was gained by adding percent free PSA over that of standard preoperative parameters. When analysing treatment outcome, the area under the curve of the predictive model decreased after adding percent free PSA. This finding indicates that the addition of percent free PSA can confound the prediction obtained from the combination of 3 accepted parameters. Therefore, our results indicate that percent free PSA does not independently predict organ confined cancer or postoperative freedom from PSA recurrence. To our knowledge, our study is the first to correlate percent free PSA directly with prognosis after treatment for localized prostate cancer. Previously others have examined the association with the final pathological condition as a surrogate of outcome. These studies have conflicting results and, therefore, there is no consensus on the usefulness of percent free PSA as a staging or prognostic marker. Southwick et al reported a multicenter study of screened patients, in which higher percent free PSA was associated with more favorable pathological findings in prostatectomy specimens.3 Specifically 75% versus 34% of patients with free PSA greater versus less than 15% had organ confined cancer, Gleason

sum less than 7 and tumors involving less than 10% of the prostate. Although percent free PSA was an independent predictor of pathological outcome, by limiting their evaluation only to patients with clinical stage T1c disease and pretreatment PSA 4 to 10 ng./ml. some of the effect of these variables may have been removed. To make the populations more comparable we also assessed clinical stage T1c cancer with total PSA between 2 and 10 ng./ml. This subset of our patients had pathological features similar to those of the patients of Southwick et al (extracapsular extension 70.3% versus 71.6%). Despite these similarities, in our study no independent prediction was obtained using percent free PSA. In another study, Arcangeli et al showed that on multivariate analysis percent free PSA predicted capsular penetration and cancer volume.4 The adjusted odds ratio was 1.6 for each 5% decrease in percent free PSA for each of these pathological features. However, they reported that their population did not represent a complete series of consecutive cases nor were they randomly selected. Therefore, a bias may have been created that confounded the outcome. Grossklaus et al from Vanderbilt University also identified a significant inverse relationship of the free-to-total PSA ratio with cancer volume when total PSA was elevated at greater than 4 ng./ml.5 Furthermore, there was a significant inverse relationship among the free-to-total PSA ratio, extracapsular disease and Gleason score. However, these results were evident only on univariate analysis. Multivariate analysis was not performed. These results contrast with those in other studies. In an unscreened population, Lerner et al did not note a relationship of the free-to-total PSA ratio with pathological stage or grade.8 They detected significant overlap of free PSA among all stages, grades and ploidy values that precluded the usefulness of percent free PSA for staging purposes. Noldus et al

PERCENT FREE PROSTATE SPECIFIC ANTIGEN AND PROSTATE CANCER

observed no significant correlation of percent free PSA with total and high grade cancer volume in 170 men.9 In addition, no significant difference was noted for percent free PSA in patients with stages pT2 and pT3 cancer. In a series of 130 men, Henricks et al identified considerable overlap in percent free PSA in patients with organ confined disease versus extraprostatic spread.10 Bangma et al observed no additional value of percent free PSA for clinical staging of prostate cancer compared to total PSA in 123 men.11 Morote et al, also identified no additional value in clinical prostate cancer staging using percent free PSA over that obtained using PSA.12 Therefore, our results support the conclusion of these latter studies, indicating no independent relationship of percent free PSA with pathological findings. Furthermore, we directly examined the outcome after radical prostatectomy and again detected no independent prediction using percent free PSA. Although our followup was short, disease outcome represents a more important end point than the final pathological condition. Hull et al reported that patients with unfavorable pathological features do not necessarily have disease progression.17 There are several possibilities why some previous groups have observed a correlation of percent free PSA with pathological outcome, while we and others have not. Positive studies often enrolled patients primarily through prostate cancer screening programs.3, 4 Clearly this patient population differs from an unscreened patient sample. The median tumor volume in the study of Arcangeli et al was 1.1 cc and 30.6% of tumors were insignificant at less than 0.5 cc.4 In the study of Grossklaus et al mean tumor volume was 2.26 cc.5 This value compares to the median tumor volume of 4.98 cc in our cases with only 1.8% of tumors less than 0.5 cc. This observation implies that any benefit of percent free PSA may be evident only for the smallest tumors. It should be considered that radical prostatectomy series showing the usefulness of percent free PSA often included a significant number of patients with cancer not requiring aggressive treatment. An additional confounder may be the type of PSA assay used. Semjonow et al noted a significant and clinically relevant discordance of reported total PSA for identical samples depending on the assay used and on the concentration of free PSA in the sample.18 In addition to the potential impact of patient selection, interpretating percent free PSA for staging purposes must consider the assay. Also, we did not restrict our analysis to any given subset of patients but analyzed all patients with clinically localized prostate cancer treated with radical prostatectomy. By restricting analysis to include only a subset of patients the variability and, therefore, the influence of other factors are limited. Notably it was not our purpose to assess the improvement of percent free PSA for diagnosing prostate cancer. Percent free PSA should not be considered as an indication for prostate biopsy in men with PSA less than 2 ng./ml. (as a low percent free PSA is assay noise) or in men with (PSA greater than 10 ng./ml. (as a biopsy is indicated regardless of the percent free PSA level). In our series we assessed the question of whether percent free PSA can improve the prediction of the prognosis in men with proved prostate cancer. There are some limitations of our study. Followup was short at a mean of 13 months. However, 581 patients had followup information available and 85 events (14.6%) occurred in that period using the strict definition of biochemical disease recurrence of greater than 0.1 ng./ml. Since percent free PSA is a recent test that was not widely until the mid 1990s, our results must be reevaluated as followup matures. However, the lack of a correlation of percent free PSA with pathological findings on multivariate analysis would indicate that our results are unlikely to change with longer followup. Although more patients may have treatment failure, the lack of a pathological correlation would indicate that percent free PSA would not aid in the prediction of those who would eventually experience failure.

1309

CONCLUSIONS

Our study showed no association of percent free PSA with pathological or disease outcome of unscreened prostate cancer treated with radical prostatectomy on multivariate analysis. Furthermore, the predictive accuracy of statistical models using clinical stage, pretreatment PSA and biopsy Gleason sum were not improved by adding percent free PSA. This observation questions the usefulness of percent free PSA beyond its role in the detection of prostate cancer. REFERENCES

1. Lilja, H., Piironen, T. P., Rittenhouse, H. G. et al: Value of molecular forms of prostate specific antigen and related kallikrein, hK2, in diagnosis and staging of prostate cancer. In: Comprehensive Textbook of Genitourinary Oncology, 2nd ed. Baltimore: Lippincott Williams & Wilkins, p. 638, 2000 2. Catalona, W. J., Partin, A. W., Slawin, K. M. 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, 279: 1542, 1998 3. Southwick, P. C., Catalona, W. J., Partin, A. W. et al: Prediction of post-radical prostatectomy pathological outcome for stage T1c prostate cancer with percent free prostate specific antigen: a prospective multicenter clinical trial. J Urol, 162: 1346, 1999 4. Arcangeli, C. G., Humphrey, P. A., Smith, D. S. et al: Percentage of free serum prostate-specific antigen as a predictor of pathological features of prostate cancer in a screening population. Urology, 51: 558, 1998 5. Grossklaus, D. J., Shappell, S. B., Gautam, S. et al: Ratio of free-total prostate specific antigen correlates with tumor volume in patients with increased prostate specific antigen. J Urol, 165: 455, 2001 6. Pannek, J., Rittenhouse, H. G., Chan, D. W. et al: The use of percent free prostate specific antigen for staging clinically localized prostate cancer. J Urol, 159: 1238 1998 7. Epstein, J. I., Chan, D. W., Sokoll, L. J. et al: Nonpalpable stage T1c cancer: prediction of insignificant disease using free/total prostate specific antigen levels and biopsy findings. J Urol, 160: 2407, 1998 8. Lerner, S. E., Jacobsen, S. J., Lilja, H. et al: Free, complexed, and total serum prostate-specific antigen concentrations and their proportions in predicting stage, grade, and deoxyribonucleic acid ploidy in patients with adenocarcinom of the prostate. Urology, 48: 240, 1996 9. Noldus, J., Graefen, M., Huland, E. et al: The value of the ratio of free-to-total PSA for staging purposes in previously untreated prostate cancer J Urol, 159: 2004, 1998 10. Henricks, W. H., England, B. G., Giacherio, D. A. et al: Serum precent free PSA does not predict extraprostatic spread of prostate cancer. Am J Clin Pathol, 109: 533, 1998 11. Bangma, C. H., Kranse, R., Blijenberg, B. G. et al: The free-tototal serum prostate specific antigen ratio for staging prostate carcinoma. J Urol, 157: 544, 1997 12. Morote, J., Raventos C. X., Lorente, J. A. et al: Measurement of free PSA in the diagnosis and staging of prostate cancer. Int J Cancer, 71: 758, 1997 13. McNeal, J. E., Villers, A. A., Redwine, E. A. et al: Histologic differentiation, cancer volume, and pelvic lymph node metastasis in adenocarcinoma of the prostate. Cancer, 66: 1225, 1990 14. Gleason, D. F.: Histologic grading and clinical staging of prostatic carcinoma. In: Urologic Pathology: The Prostate. Edited by M. Tannenbaum. Philadelphia: Lea & Ferbiger, chapt. 9, pp. 171– 197, 1977 15. Schro¨ der F. H., Hermanek P. Denis L. et al. The TNM classification of prostate cancer. Prostate, suppl., 4: 129, 1992 16. Harrell, F. E., Jr. and Califf, R. M.: Evaluating the yield of medical tests. JAMA, 247: 2543, 1982 17. Hull, G. W., Rabbani, F., Abbas, F. A. et al: Cancer control with radical prostatectomy alone in 1000 consecutive patients. J Urol, 167: 000, 2002 18. Semjonow, A., Oberpenning, F., Brandt, B. et al: Impact of free prostate-specific antigen on discordant measurement results of assays for total prostate-specific antigen. Urology, suppl., 48: 10, 1996