Key words: prostatic neoplasms, risk groups, biochemical recurrence, clinical progression, prostate cancer specific mortality

JJCO Japanese Journal of Clinical Oncology Japanese Journal of Clinical Oncology, 2016, 46(8) 762 767 doi: 10.1093/jjco/hyw061 Advance Access Publication Date: 20 May 2016 Original Article Original Article Subclassification of high-risk clinically organ-confined prostate cancer for early cancer-specific mortality after radical prostatectomy Takashi Kobayashi 1, Takahiro Kimura 2, Chunwoo Lee 3, Takahiro Inoue 1, Naoki Terada 1, Yuka Kono 1, Tomomi Kamba 1, Choung-Soo Kim 3, Shin Egawa 2, and Osamu Ogawa 1, * 1 Department of Urology, Kyoto University Graduate School of Medicine, Kyoto, Japan, 2 Department of Urology, Jikei University School of Medicine, Tokyo, Japan, and 3 Department of Urology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea *For reprints and all correspondence: Osamu Ogawa, Department of Urology, Kyoto University Graduate School of Medicine, 54 Shogoinkawahara-cho, Sakyo-ku, Kyoto, 606-8507, Japan. E-mail: ogawao@kuhp.kyoto-u.ac.jp Received 10 February 2016; Accepted 14 April 2016 Abstract Objective: High-risk clinically localized prostate cancer is seen in a highly heterogeneous population with a wide variation of clinical aggressiveness and a novel subclassification for the better prediction of clinical outcomes is needed. The aim of this study is to validate a modified D Amico risk criteria for substratification of high-risk prostate cancer with regard to the prediction of biochemical recurrence, clinical progression-free survival or prostate cancer specific mortality after radical prostatectomy. Methods: We conducted a retrospective multicenter cohort study including 461 clinically organconfined (ct1-2), D Amico high-risk prostate cancer patients who underwent radical prostatectomy with pelvic lymph node dissection. The modified criteria subclassified D Amico high-risk patients into high-risk (n = 189, single high-risk parameter and two low-risk parameters) and very high risk (n = 272, at least one more intermediate or high-risk parameter in addition to the qualifying highrisk parameter) groups. Biochemical recurrence-free survival, clinical progression-free survival, prostate cancer specific mortality and overall survival were analyzed. Results: The very high risk group, compared with high-risk group, had significantly poorer biochemical recurrence (5- and 10-year biochemical recurrence-free rates: 52.8 vs 73.9% and 42.1 vs 61.7%, respectively, P < 0.0001), clinical progression-free survival (5- and 10-year survivals: 91.8 vs 98.2% and 80.5 vs 98.2%, respectively, P = 0.0013) and prostate cancer specific mortality (5- and 10-year mortalities: 2.5 vs 0.0% and 6.7 vs 0.0%, respectively, P = 0.0124). Conclusion: D Amico high-risk patients can achieve very favorable outcomes unless they are classified as very high risk. Our novel subclassification method is very simple and useful for better patient counseling and decision-making in the pretreatment setting. Key words: prostatic neoplasms, risk groups, biochemical recurrence, clinical progression, prostate cancer specific mortality The Author 2016. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com 762

Jpn J Clin Oncol, 2016, Vol. 46, No. 8 763 Introduction Radical prostatectomy (RP) is a standard treatment for patients with clinically localized prostate cancer (PCa), and a number of risk classifications or nomogram models have been proposed to predict oncological outcomes after RP (1). Among them, D Amico risk stratification (2) is one of the pioneer and most widely accepted models. It has been externally validated and shown to be useful also for the prediction of post-rp survival (3), although it was initially designed for the prediction of biochemical recurrence (BCR) after definitive treatment including RP. While excellent outcomes were achieved in low or intermediate risk cases, relatively poorer clinical outcomes were reported in high-risk patients (3). Detailed analyses of post-rp treatment outcome have raised another clinically important question; high-risk cases represent a heterogeneous population that includes patients who present early relapse and those who achieve long-term disease-free status (4,5). Indeed, a contemporary review of the literature concluded that no uniform management can be recommended for high-risk PCa (6). Recent effort to stratify high-risk patients according to prognostic risks includes subclassifications using pretreatment parameters (7 12), nomograms (13 15), scoring models using postsurgical pathological information (16), genomic classifier based on gene expression signature of primary tumor (17) and their combinations (18). For pretreatment patient counseling and decision-making, simple prognostic tools based on presurgical information are feasible. Additionally, there have been very few reports on risk stratifications for Asian populations. Previously, our group proposed modified D Amico criteria that subclassifies high-risk category into two subclasses with distinct post-rp BCR rate (19). The modified substratification is characterized by a very simple and user-friendly subgrouping method that requires only the three existing preoperative parameters used in D Amico criteria. A D Amico high-risk patient is subclassified as high risk if he has only a single high-risk parameter and two lowrisk parameters, while a patient is subclassified as very high risk if he has at least one or more intermediate or high-risk parameters in addition to the qualifying high-risk parameter. Here, we have investigated whether the modified criteria is useful for the prediction of clinically more relevant outcomes namely clinical progression-free survival (cpfs), early PCa-specific mortality (PCSM) and overall survival (OS) using a data set derived from the international multiinstitutional collaboration in Asia. Patients and methods Model and study design According to our modified D Amico criteria, high-risk patients were substratified into two subgroups namely high-risk cases, who have a single high-risk (prostate-specific antigen (PSA) > 20 ng/ml, biopsy Gleason score (bgs) 8orcT 2c) and two low-risk features (PSA < 10 ng/ml, bgs < 7 and ct < 2b), and very high risk cases, who have at least one extra intermediate or high-risk feature (PSA 10 ng/ml, bgs 7orcT 2b) in addition to their qualifying high-risk parameter (19). We conducted multi-institutional international collaborative study to establish a database of PCa patients based on the approval by the institutional review board at each institute. We analyzed pooled data on patients who underwent RP for clinically organconfined ( ct2) PCa between 1997 and 2014 at Jikei Medical University Hospital (Tokyo, Japan), Asan Medical Center (Seoul, Korea) and Kyoto University Hospital (Kyoto, Japan). Patients who received neoadjuvant or adjuvant androgen deprivation or radiotherapy, who did not have pelvic lymph node dissection at the time of RP or who had any other missing data were excluded from this study. Consequently, 461 D Amico high-risk patients were studied (Table 1). Using pretreatment ct stage (based on TNM 7th edition, 2010), PSA levels and bgs, the patients were stratified into either of high-risk (n = 189) or very high risk (n = 272) group. As for institutional variation, there were no significant differences in the proportion of very high risk patients (62.3% in Jikei, 57.3% in Asan and 50.9% in Kyoto, P = 0.26, chi-square test) and patient age (65.6 ± 6.5 years for Jikei, 65.1 ± 7.5 for Asan and 67.4 ± 5.6 for Kyoto, P = 0.14, Kruskal Wallis test). However, there were significant institutional variations in the proportions of ct1c tumor (54.3% in Jikei, 29.2% in Asan and 60.4% in Kyoto, P < 0.0001, chi-square test), bgs 8 10 disease (76.2% in Jikei, 64.9% in Asan and 58.5% in Kyoto, P = 0.004, chi-square test) and preoperative PSA (14.9 ± 10.6 ng/ml for Jikei, 14.1 ± 12.5 for Asan and 20.5 ± 19.1 for Kyoto, P = 0.003, Kruskal Wallis test). Preoperative information Preoperative clinical staging was mainly based on digital rectal examination, computed tomography and bone scan, while magnetic resonance imaging was used in a part of the patients. The majority of the patients underwent open retropubic RP, while the rest did laparoscopic RP including robotic surgery. Most cases underwent at least limited lymph node dissection concomitantly with RP. As surgical specimen was processed routinely in a similar manner in the Table 1. Clinical characteristics of study patients with high-risk prostate cancer Parameter High risk Very high risk P value Patients, n 189 272 Median age, years (IQR) 66 (60 70) 67 (62 71) 0.069* Clinical T stage, n (%) <0.0001 T1a-b 1 (0.5) 1 (0.4) T1c 111 (58.7) 94 (34.6) T2a 55 (29.1) 48 (17.6) T2b 0 (0) 74 (27.2) T2c 22 (11.6) 55 (20.2) Biopsy Gleason score, <0.0001 n (%) 6 47 (24.9) 11 (4.0) 7 0 (0) 82 (30.1) 8 10 142 (75.1) 179 (65.8) Median preoperative 7.2 (5.5 8.9) 14.9 (10.7 24.0) <0.0001* PSA, ng/ml (IQR) Pretreatment PSA range, <0.0001 ng/ml, n (%) <10 164 (86.8) 54 (19.9) 10 20 0 (0) 119 (43.8) >20 25 (13.2) 99 (36.4) Surgical procedure, n (%) 0.303 Open 140 (74.1) 210 (77.2) Laparoscopic 49 (25.9) 62 (22.8) Median postoperative follow-up (IQR), months 58 (44 91) 65 (44 99) 0.200* IQR, interquartile range; PSA, prostate-specific antigen. *Mann Whitney U-test; Chi-square test; Fisher s exact test.

764 Subclassification of high-risk prostate cancer three institutes, pathological information on biopsy and prostatectomy specimen was collected based on pathological reports from each institute. Postoperative follow-up Postoperatively, patient follow-up generally comprised physical examination and serum PSA measurement on every 3 6 months during the first 2 years and every 6 12 months thereafter. Imaging studies including CT and bone scans were obtained at the physician s discretion. BCR was defined as two consecutive measurements of PSA levels 0.2 ng/ml and the date of BCR was defined as the time of the first measurement of PSA level 0.2 ng/ml. When PSA levels after surgery, the date of BCR was defined as the time of surgery. The treatment options after BCR or clinical progression were determined according to the physician s preference. The patients who kept no biochemical evidence of disease following local salvage radiotherapy after BCR were defined as having BCR but no clinical progression. Clinical progression was defined as manifestation of any symptomatic or radiologically evident lesion of relapsed disease. When clinical progression without PSA elevation was observed (although only two patients), the date of BCR was defined as the time of clinical progression. The recurrence site of clinically progressed disease was also reported. PCa-specific death was clinically determined by the attending urologist who followed the patients until death. There had to be a documentation of uncontrollable disease progression against androgen deprivation or chemotherapy. Statistical analysis Probabilities of BCR-free survival, cpfs, disease-specific survival (DSS) and OS were estimated using the Kaplan Meier method, and statistical significance was examined using log-rank test. Concordance index was also reported with the standard error (±SE). Cox proportional hazard model was used for multivariate analyses for survival. Time to BCR, clinical progression and mortality were calculated from the date of surgery. Additionally, time from BCR to clinical progression and mortality were also calculated. All P values were two-sided, and P < 0.05 were considered statistically significant. Results Pathological outcomes Compared with high-risk patients, very high risk patients had significantly higher probability for having pt3-4 disease, lymph node involvement, higher pathological Gleason score and positive resection margin on RP (Table 2). Postoperative survivals BCR-free survival was significantly better in high-risk patients than in very high risk patients (5- and 10-year BCR-free rates: 73.9 vs 52.8% and 61.7 vs 42.1%, respectively, P < 0.0001, concordance index: 0.592 ± 0.020, Fig. 1A). Of note, statistically significant differences in BCR-free survival between the two groups were robustly observed when they were analyzed for each institute separately (data not shown). High-risk patients showed significantly better cpfs than very high risk patients (5- and 10-year survivals: 98.2 vs 91.8% and 98.2 vs 80.5%, respectively, P = 0.0013, concordance Table 2. Pathological outcome of high and very high-risk patients Parameter index: 0.661 ± 0.051, Fig. 1B). Moreover, high-risk patients had significantly lower PCSM than very high risk patients (5- and 10-year mortalities: 0.0 vs 2.5% and 0.0 vs 6.7%, respectively, P = 0.0124, concordance index: 0.698 ± 0.087, Fig. 1C), whereas there was no statistically significant difference in OS (5- and 10-year survivals: 98.8 vs 97.2% and 98.8 vs 89.2%, respectively, P = 0.0784, concordance index: 0.600 ± 0.070, Fig. 1D). In multivariate Cox proportional hazard analysis, ct2c had no statistically significant impact on BCR-free survival, cpfs, DSS and OS after adjusting pretreatment PSA and bgs (Table 3). PSA >20 ng/ml had a significant impact on BCR-free survival, DSS and OS, while bgs 8 had a significant impact on BCR-free survival and OS. Survival after BCR Fifty (26.5%) patients had BCR in high-risk group, whereas 127 (46.7%) had BCR in very high isk group. Among them, only 3 patients (6.0%) in high-risk group developed clinical progression of PCa, whereas 26 (14.7%) developed in very high risk group at a median follow-up of 4.9 years after BCR. In high-risk group, 67% (2 of 3) of clinical progressions were identified as locoregional recurrence (prostatic bed or pelvic lymph node involvement), whereas 65% (17 of 26) of clinical progressions were manifested as skeletal or visceral metastasis in very high risk group (P = 0.279, Fig. 2A). Time from BCR to clinical progression was significantly shorter in very high-risk group as evident in the Kaplan Meier plot after BCR (P = 0.0365, Fig. 2B). Accordingly, similar tendency was observed in DSS and OS (Figs 2C and D). There was no particular difference in pathological and oncological outcomes among the three study institutes. Discussion High risk (n = 189) Very high risk (n = 272) P value Pathological T stage, n (%) 0.0027 T2 99 (52.4) 101 (37.1) T3 81 (42.9) 145 (53.3) T4 9 (4.8) 26 (9.6) Pathological N stage, n (%) 0.0009 N0 183 (96.8) 240 (88.2) N1 6 (3.2) 32 (11.8) Prostatectomy Gleason 0.0289 score, n (%) 6 16 (8.5) 8 (2.9) 7 100 (52.9) 147 (54.0) 8 10 73 (38.6) 117 (43.0) Positive resection margins, n (%) 63 (33.3) 130 (47.8) 0.0021 Chi-square test; Fisher s exact test. We previously reported that a simple subclassification method successfully predicted post-rp BCR in a heterogeneous population of men with D Amico high-risk PCa (19). This study has validated the previously reported model with regard to BCR prediction using larger patient populations. Essentially, this study demonstrated that our model is readily verifiable using existing databases without additional data collection and widely applicable to most established databases. Of particular note, this study provides risk stratification

Jpn J Clin Oncol, 2016, Vol. 46, No. 8 765 Figure 1. Kaplan Meier curves for biochemical recurrence-free survival (A), clinical progression-free survival (B), prostate cancer specific survival (C) and overall survival (D) after radical prostatectomy. Table 3. Cox proportional hazard analysis on the three parameters for clinical outcomes Clinical outcomes and variables HR 95% CI P value BCR-free survival ct2c vs ct2b 1.669 0.994 2.802 0.053 PSA > 20 ng/ml vs PSA 20 ng/ml 2.396 1.614 3.555 <0.0001 GS 8vsGS 7 1.745 1.193 2.853 0.006 cpfs ct2c vs ct2b 1.025 0.301 3.491 0.968 PSA > 20 ng/ml vs PSA 20 ng/ml 1.798 0.719 4.494 0.210 GS 8vsGS 7 2.607 0.887 7.656 0.081 DSS ct2c vs ct2b 1.043 0.113 9.603 0.970 PSA > 20 ng/ml vs PSA 20 ng/ml 5.320 1.283 22.061 0.021 GS 8vsGS 7 3.220 0.689 15.058 0.137 OS ct2c vs ct2b 1.204 0.246 5.891 0.819 PSA > 20 ng/ml vs PSA 20 ng/ml 5.568 1.879 5.568 0.002 GS 8vsGS 7 3.846 1.147 12.897 0.029 HR, hazard ratio; CI, confidence interval; BCR, biochemical recurrence; cpfs, clinical progression-free survival; DSS, disease-specific survival; OS, overall survival; GS, Gleason score. model validated in Asian populations, which has been reported to have distinct clinical and biological characteristics. More importantly, this study has shown that this modified D Amico criteria successfully predicts a subset of patients who are at high risk for clinical progression and early PCSM. This model is characterized by very simple criteria that make it user-friendly. There are several nomograms that were reported to show better predictive performance but they require calculation formula or scale bars. Additionally, our model will be useful for patient counseling and clinical decision-making preoperatively. Very high risk patients are less likely to be cured by RP alone, and multimodal treatment strategy should be recommended to this group of patients (5,20 22). In contrast, high-risk (not very high) patients showed relatively better outcomes that are similar to those for intermediate risk patients. Recently, a multi-institutional study stratified patients undergoing robot-assisted laparoscopic RP into five groups with distinct risks for BCR, clinical progression and consequent use of salvage therapy (7). Their model is also characterized by very simple categorization using only pretreatment PSA and bgs but not ct stage. There have been several other reports on substratification of highrisk PCa using preoperative parameters (8 12). Some of them use PSA velocity (8,9) or number of positive score (10), which makes the external validation difficult due to lack of these data in some population. Others use simply PSA, bgs and ct stage (11,12), which allows clinicians to apply those models to their daily clinical practice easily or to make a head-to-head comparison of two distinct models. Other recent efforts for post-rp outcome prediction comprise nomograms (15), a postsurgical scoring model (16) and a genomic classifier (23). Although these models yielded better predictive abilities independently (15 17) and in combination (18), they utilize postoperative pathological information or gene expression profiles and cannot be used for preoperative patient counseling and decision-making.

766 Subclassification of high-risk prostate cancer Figure 2. Distribution of clinical relapse sites (A). Kaplan Meier curves for clinical progression-free survival (B), prostate cancer specific survival (C) and overall survival (D) after biochemical recurrence (BCR) among men with BCR. In contrast with those previous studies including ct3 cases, the subjects of this study had clinically organ-confined disease ( ct2). Indeed, very few patients with locally advanced (ct3) PCa are treated with RP alone. Since our cohort consisted of ct2 patients, multivariate analysis revealed that ct stage has little impact on clinical outcomes as reported by other group (24). Little impact of ct stage might be due to large number of ct1 patients or due to difference in staging modalities. In this regard, PSA, highly standardized testing, showed the strongest significance among incorporated parameters. Thus, prognostic models comprising ct stage should be applied to another patient population with caution. In addition to the shorter postoperative survivals, patients with very high risk PCa had significantly earlier clinical progression and PCSM from the time of BCR, which is consistent with previous reports showing a significant association between shorter time to BCR and post-bcr clinical progression (25,26) or PCSM (27 29). However, our subclassification is clearly different from previously proposed ones in that it provides predictive information preoperatively, which yields strong advantages as described above. As evident by higher rate of skeletal or visceral involvement in the clinical progression setting in our series, very high risk PCa patients are more likely to have already developed occult distant metastasis at the time of RP. Therefore, it seems reasonable to recommend adjuvant systemic therapy for very high risk patients rather than differed salvage therapy after BCR. As for disease progression pattern, whether the recurrence is local, regional or systemic is a key distinction that impacts post-bcr management (30). The present results suggest that very high risk PCa are more likely to have been already disseminated distantly at the time of RP, which must have caused shorter time from RP to BCR and shorter time from BCR to PCSM. Patients with very high risk PCa may need more frequent biochemical and radiological studies in post-rp follow-up. Several limitations also exist including retrospective nature. Pathology data were not centralized but based on the pathology report from each institute. The data lack information of the number of patients who had magnetic resonance imaging for clinical staging. Although all patients analyzed in this study had pelvic lymph node dissection (LND), the study lacks the information on how many patients had extended LND since the extent of LND was not standardized. The fact that extended LND was not performed in all patients might have caused the relatively low lymph node positivity in our cohort compared with previous reports on high risk PCa as well as racial difference and lower ct stage not including ct3 disease in this study. We are aware that about 5-year follow-up after RP is not sufficient to examine post-rp survival due to the protracted natural history of the clinical progression of PCa. However, our results clearly indicate that patients with very high-risk PCa have significantly earlier PCSM than those with high-risk PCa. It is obvious that a larger study with longer follow-up would be warranted in the future to validate the results of this study. In this regard, our subclassification model is very simple and requires no more parameters in addition to the D Amico s three factors, which hopefully enables investigators to further validate our subclassification method without difficulty. In conclusion, we have demonstrated that our modified D Amico s subclassification method has a strong potential to substratify D Amico

Jpn J Clin Oncol, 2016, Vol. 46, No. 8 767 high-risk PCa patients into two groups at distinct risks for post-rp clinical progression and early PCSM in addition to BCR. This differentiation of D Amico high-risk PCa patients is clinically relevant and immediately applicable to the daily clinical practice as a physicianand patient-friendly model for risk stratification. Funding This study was supported in part by a grant-in-aid for scientific research from the Ministry of Education, Culture, Sports, Science and Technology, Japan (26253078) (to O.O.). Conflict of interest statement None declared. References 1. Lughezzani G, Briganti A, Karakiewicz PI, et al. Predictive and prognostic models in radical prostatectomy candidates: a critical analysis of the literature. Eur Urol 2010;58:687 700. 2. D Amico AV, Whittington R, Malkowicz SB, et al. Biochemical outcome after radical prostatectomy, external beam radiation therapy, or interstitial radiation therapy for clinically localized prostate cancer. JAMA 1998;280:969 74. 3. Boorjian SA, Karnes RJ, Rangel LJ, Bergstralh EJ, Blute ML. Mayo Clinic validation of the D amico risk group classification for predicting survival following radical prostatectomy. JUrol2008;179:1354 60; discussion 60 1. 4. Yossepowitch O, Eastham JA. Role of radical prostatectomy in the treatment of high-risk prostate cancer. Curr Urol Rep 2008;9:203 10. 5. Eastham JA, Evans CP, Zietman A. What is the optimal management of high risk, clinically localized prostate cancer? Urol Oncol 2010;28:557 67. 6. Bastian PJ, Boorjian SA, Bossi A, et al. High-risk prostate cancer: from definition to contemporary management. Eur Urol 2012;61:1096 106. 7. Abdollah F, Sood A, Sammon JD, et al. Long-term cancer control outcomes in patients with clinically high-risk prostate cancer treated with robot-assisted radical prostatectomy: results from a multi-institutional study of 1100 patients. Eur Urol 2015;68:497 505. 8. D Amico AV, Chen MH, Catalona WJ, Sun L, Roehl KA, Moul JW. Prostate cancer-specific mortality after radical prostatectomy or external beam radiation therapy in men with 1 or more high-risk factors. Cancer 2007;110:56 61. 9. Nguyen PL, Chen MH, Catalona WJ, Moul JW, Sun L, D Amico AV. Predicting prostate cancer mortality among men with intermediate to high-risk disease and multiple unfavorable risk factors. Int J Radiat Oncol Biol Phys 2009;73:659 64. 10. Sundi D, Wang VM, Pierorazio PM, et al. Very-high-risk localized prostate cancer: definition and outcomes. Prostate Cancer Prostatic Dis 2014;17:57 63. 11. Joniau S, Briganti A, Gontero P, et al. Stratification of high-risk prostate cancer into prognostic categories: a European multi-institutional study. Eur Urol 2015;67:157 64. 12. Grossfeld GD, Latini DM, Lubeck DP, Mehta SS, Carroll PR. Predicting recurrence after radical prostatectomy for patients with high risk prostate cancer. J Urol 2003;169:157 63. 13. Kattan MW, Eastham JA, Stapleton AM, Wheeler TM, Scardino PT. A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst 1998;90:766 71. 14. Stephenson AJ, Kattan MW, Eastham JA, et al. Prostate cancer-specific mortality after radical prostatectomy for patients treated in the prostatespecific antigen era. J Clin Oncol 2009;27:4300 5. 15. Stephenson AJ, Scardino PT, Eastham JA, et al. Postoperative nomogram predicting the 10-year probability of prostate cancer recurrence after radical prostatectomy. J Clin Oncol 2005;23:7005 12. 16. Cooperberg MR, Hilton JF, Carroll PR. The CAPRA-S score: a straightforward tool for improved prediction of outcomes after radical prostatectomy. Cancer 2011;117:5039 46. 17. Karnes RJ, Bergstralh EJ, Davicioni E, et al. Validation of a genomic classifier that predicts metastasis following radical prostatectomy in an at risk patient population. J Urol 2013;190:2047 53. 18. Cooperberg MR, Davicioni E, Crisan A, Jenkins RB, Ghadessi M, Karnes RJ. Combined value of validated clinical and genomic risk stratification tools for predicting prostate cancer mortality in a high-risk prostatectomy cohort. Eur Urol 2015;67:326 33. 19. Kita Y, Inoue T, Shimizu Y, Kamba T, Yoshimura K, Ogawa O. Subclassification of the D Amico s high risk group to predict biochemical recurrence after radical prostatectomy. Hinyokika Kiyo 2012;58:319 24. 20. Perez BA, Koontz BF. Radiotherapy before and after radical prostatectomy for high-risk and locally advanced prostate cancer. Urol Oncol 2015;33:226 34. 21. Oh WK. Neoadjuvant therapy before radical prostatectomy in high-risk localized prostate cancer: defining appropriate endpoints. Urol Oncol 2003;21:229 34. 22. Skinner EC, Glode LM. High-risk localized prostate cancer: primary surgery and adjuvant therapy. Urol Oncol 2003;21:219 27. 23. Erho N, Crisan A, Vergara IA, et al. Discovery and validation of a prostate cancer genomic classifier that predicts early metastasis following radical prostatectomy. PLoS One 2013;8:e66855. 24. Klaassen Z, Singh AA, Howard LE, et al. Is clinical stage T2c prostate cancer an intermediate- or high-risk disease? Cancer 2015;121:1414 21. 25. Antonarakis ES, Feng Z, Trock BJ, et al. The natural history of metastatic progression in men with prostate-specific antigen recurrence after radical prostatectomy: long-term follow-up. BJU Int 2012;109:32 9. 26. Ross AE, Feng FY, Ghadessi M, et al. A genomic classifier predicting metastatic disease progression in men with biochemical recurrence after prostatectomy. Prostate Cancer Prostatic Dis 2014;17:64 9. 27. Pound CR, Partin AW, Eisenberger MA, Chan DW, Pearson JD, Walsh PC. Natural history of progression after PSA elevation following radical prostatectomy. JAMA 1999;281:1591 7. 28. Freedland SJ, Humphreys EB, Mangold LA, et al. Risk of prostate cancerspecific mortality following biochemical recurrence after radical prostatectomy. JAMA 2005;294:433 9. 29. Brockman JA, Alanee S, Vickers AJ, et al. Nomogram predicting prostate cancer-specific mortality for men with biochemical recurrence after radical prostatectomy. Eur Urol 2015;67:1160 7. 30. Simmons MN, Stephenson AJ, Klein EA. Natural history of biochemical recurrence after radical prostatectomy: risk assessment for secondary therapy. Eur Urol 2007;51:1175 84.