european urology 55 (2009)

european urology 55 (2009) 261 270 available at www.sciencedirect.com journal homepage: www.europeanurology.com Platinum Priority Prostate Cancer Editorial by George N. Thalmann on pp. 271 272 of this issue Two Positive Nodes Represent a Significant Cut-off Value for Cancer Specific Survival in Patients with Node Positive Prostate Cancer. A New Proposal Based on a Two-Institution Experience on 703 Consecutive N+ Patients Treated with Radical Prostatectomy, Extended Pelvic Lymph Node Dissection and Adjuvant Therapy Alberto Briganti a, *, Jeffrey R. Karnes c, Luigi Filippo Da Pozzo a, Cesare Cozzarini d, Andrea Gallina a, Nazareno Suardi a, Marco Bianchi a, Massimo Freschi b, Claudio Doglioni b, Ferruccio Fazio d, Patrizio Rigatti a, Francesco Montorsi a, Michael L. Blute c a Department of Urology, Vita-Salute University, Milan, Italy b Department of Pathology, Vita-Salute University, Milan, Italy c Department of Urology, Mayo Medical School and Mayo Clinic, Rochester, Minnesota, USA d Department of Radiotherapy, Vita-Salute University, Milan, Italy Article info Article history: Accepted September 18, 2008 Published online ahead of print on October 1, 2008 Keywords: Prostate cancer Radical prostatectomy Lymph node metastasis Number of positive nodes Please visit www.eu-acme.org/ europeanurology to read and answer questions on-line. The EU-ACME credits will then be attributed automatically. Abstract Background: Currently, the 2002 American Joint Committee on Cancer (AJCC) staging system of prostate cancer does not include any stratification of patients according to the number of positive nodes. However, node positive (N+) patients share heterogeneous outcomes according to the extent of lymph node invasion (LNI). Objective: To test whether the accuracy of cancer specific survival (CSS) predictions may be improved if node positive patients are stratified according to the number of positive nodes. Design, Setting, and Participants: The study cohort included 703 N+ M0 patients treated with radical prostatectomy (RP) and extended pelvic lymph node dissection (eplnd) between September 1988 and January 2003 at two large Academic Institutions. Number of positive nodes was dichotomized according to the most informative cut-off predicting CSS. Kaplan-Meier curves assessed cancer specific survival rates. Predictive accuracy of the current N stage and of the new N classification in predicting CSS was quantified with Harrell s concordance index after adjusting for pathological (T) stage and internally validated with 200 boostraps resamples. Differences in predictive accuracy were compared with the Mantel-Haentzel test. * Corresponding author. Department of Urology, Vita-Salute University, Via Olgettina 60, 20132 Milan, Italy. Tel. +39 02 26437286; Fax: +39 02 26437298. E-mail address: briganti_alberto@yahoo.it, briganti.alberto@hsr.it (A. Briganti). 0302-2838/$ see back matter # 2008 European Association of Urology. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.eururo.2008.09.043

262 european urology 55 (2009) 261 270 Results and Limitations: Mean follow-up was 113.7 months (median: 112.5, range 3.5 243). The mean number of nodes removed was 13.9 (range: 2 52). The mean number of positive nodes was 2.3 (range: 1 31). The most informative cut-off of positive nodes in predicting CSS was 2. Of all, 532 (75.7%) patients had 2 or less positive nodes, while 171 (24.3%) had more than 2 positive nodes. Patients with 2 or less positive nodes had significantly better CSS outcome at 15 year follow-up compared to patients with more than 2 positive nodes (84% vs 62%; p < 0.001). After adjusting for pathological stage, multivariable predictive accuracy of the new N staging (2 or >2 positive nodes) was 65.0% vs 60.1% when the number of positive nodes was not considered (4.9% gain; p < 0.001). Conclusions: We demonstrated that patients with up to 2 positive nodes experienced excellent CSS, which was significantly higher compared to patients with more than 2 positive nodes. Moreover, a significant improvement in CSS prediction was reached when the number of positive nodes was considered. Thus, our results reinforce the need for a stratification of node positive patients according to the number of positive nodes and may warrant consideration in the next revision of the pathologic TNM classification. # 2008 European Association of Urology. Published by Elsevier B.V. All rights reserved. 1. Introduction Radical prostatectomy (RP) is a commonly performed and effective treatment for patients with organ confined prostate cancer [1]. Large clinical series have demonstrated that RP represents also a valid treatment option for patients with locally advanced prostate cancer [2 8]. In this context, great interest has been focused on patients with prostate cancer and lymph node invasion (LNI). Although the diagnosis of prostate cancer has shifted to early clinical stages in the PSA era, nodal metastases are indeed still diagnosed in a wide range of patients [9 13]. Historically, patients with nodal metastases were not considered surgical candidates and were usually treated with hormonal therapy (HT) and/or radiotherapy (RT). However, recent studies reported excellent cancer specific outcomes of patients with histologically proven nodal metastases submitted to RP, with or without adjuvant HT [14 18]. Moreover, patients with LNI are not all at the same risk of cancer recurrence and death. Indeed, patients with a low volume of nodal disease have significantly higher survival rates compared to patients with higher volume of LNI, regardless of adjuvant treatment administration [14 18]. Despite these evidences, the 2002 American Joint Committee on Cancer (AJCC) staging system of prostate cancer does not include any stratification of patients according to the number of nodes involved by cancer [19]. This reflects the primary idea that node positive prostate cancer is always a systemic disease associated with overall long-term poor prognosis. In contrast, a recent study has shown that patients with a single nodal metastasis showed long-term outcome as favourable as those without nodal involvement [14]. Based on this data, we evaluated the long term outcome in the largest node positive (N+) prostate cancer series that has been published to date. We stratified node positive patients undergoing RP, extended pelvic lymph node dissection (eplnd) and adjuvant treatments according to the extent of LNI. We hypothesized that the accuracy of cancer specific survival (CSS) predictions may be improved if node positive patients are stratified according to the number of positive nodes. 2. Materials and methods Between September 1988 and January 2003, 703 consecutive N+ M0 patients were treated with RP, eplnd and adjuvant treatments at two large Academic Institutions. Pre-operative staging included negative pelvic/abdominal computerized tomography or ultrasound, bone scan and chest x-ray. All patients underwent extended pelvic lymph node dissection (including removal of obturator, external iliac and hypogastric nodes) at the time of RP. All patients included in this study had complete clinical and pathological data including age at surgery, pre-operative PSA, pathological stage defined according to the 2002 AJCC staging system [19], pathological Gleason score, surgical margin status, number of nodes removed as well as number of positive nodes. Moreover, all patients received adjuvant treatment, including hormonal blockade alone or in combination with RT administered within 90 days after RP. The decision to administer one or both adjuvant treatments followed the surgeon and patient discussion about possible treatment options. Medical hormone deprivation

european urology 55 (2009) 261 270 263 therapy was generally intended to be lifelong. However, given the retrospective nature of this study, it is uncertain whether patients discontinued treatment after a period of androgen deprivation therapy. Kaplan-Meier analysis was used to estimate BCR free survival, overall survival (OS) and CSS rates at 5, 8, 10 and 15 years after surgery. Biochemical failure was defined as 2 consecutive PSA values >0.4 ng/ml. Cause of death was identified from death certificates or physician correspondence. Number of positive nodes was dichotomized according to the most informative cut-off predicting CSS. This was obtained applying ANOVA test for every possible cut-off value and choosing the lowest p-value. The log rank test was used to compare CSS rates of patients sub-divided according to the dichotomized number of positive nodes. Moreover, the association between the number of positive nodes as well as lymph node density (number of positive nodes over number of total nodes removed [LND]) and CSS was tested in univariable and multivariable Cox regression models after accounting for pre-operative PSA, pathological stage, pathological Gleason score, surgical margin status, year of surgery (namely, 1993 1997 vs 1988 1992 and 1998 2003 vs 1988 1992) and adjuvant RT. Furthermore, the accuracy of each mentioned variable for CSS prediction was assessed with the Harrell s concordance index. Two hundred bootstrap resamples were used to reduce overfit bias. Finally, the accuracy of pathological T stage alone as well as of multivariable models predicting CSS not accounting for the extent of LNI (based on pre-operative PSA, pathological T stage, pathological Gleason score, surgical margin status and adjuvant RT) was compared to the predictive accuracy (PA) of multivariable models including either the new N staging based on the number of positive nodes or LND. The statistical significance in PA gain induced by the inclusion of the number of positive nodes or LND was assessed with the Mantel Haenszel test. All statistical tests were performed with the use of S-PLUS Professional, version 1 (MathSoft Inc, Seattle, WA, USA). All tests were twosided with a significance level at 0.05. 3. Results Clinical and pathological characteristics of patients included in the study are shown in Table 1. Of the Table 1 Clinical characteristic of 703 patients with lymph node invasion (LNI) submitted to radical prostatectomy, extended pelvic lymph node dissection and adjuvant therapy Variables All patients (n = 703) No (%) Patients treated at Vita-Salute University, Milan, Italy (n = 250) No (%) Patients treated at Mayo Clinic, Rochester, Minnesota, US (n = 453) No (%) p value Age 0.004 Mean (median) 65 (66) 65.9 (66) 64.4 (65) Range 47 80 47 80 47 79 Pre-operative PSA 0.046 Mean (median) 28.9 (17.2) 24.7 (15.6) 31.3 (17.9) Range 0.9 616 2.8 148 0.9 616 Pathological stage (TNM 2002) <0.001 pt2a/b/c 82 (11.7) 21 (8.4) 61 (13.5) pt3a 118 (16.8) 39 (15.6) 79 (17.4) pt3b 452 (64.3) 151 (60.4) 301 (66.4) pt4 51 (7.3) 39 (15.6) 12 (2.6) Pathological Gleason score 0.24 6 or less 148 (21.1) 53 (21.2) 95 (21) 7 347 (49.4) 114 (45.6) 233 (51.4) 8 10 208 (29.5) 83 (33.2) 125 (27.6) Surgical margin status 0.56 Positive 444 (63.2) 154 (61.6) 290 (64) Negative 259 (36.8) 96 (38.4) 163 (36) Number of lymph nodes removed and examined Mean (median) 13.9 (13) 16.6 (15.5) 12.4 (12) Range 2 52 3 52 2 37 <0.001 Number of positive lymph nodes Mean (median) 2.3 (1) 2.5 (1) 2.1 (1) 0.07 Range 1 31 1 31 1 19 2 532 (75.7) 183 (73.2) 349 (77) >2 171 (24.3) 67 (26.8) 104 (23) 0.27 Data are tabulated according to the treatment Institution. PSA: Prostate Specific Antigen.

264 european urology 55 (2009) 261 270 Table 2 Clinical characteristic of 703 patients with lymph node invasion (LNI) submitted to radical prostatectomy, extended pelvic lymph node dissection and adjuvant therapy Variables Adjuvant RT + HT (n = 171) No (%) Adjuvant HT (n = 532) No (%) p value Age 0.06 Mean (median) 64.1 (64.5) 65.3 (66) Range 47 80 47 80 Pre-operative PSA 0.55 Mean (median) 30.6 (19) 28.4 (16.2) Range 0.9 321 1.6 616 Pathological stage (TNM 2002) <0.001 pt2a/b/c 6 (3.5) 76 (14.3) pt3a 20 (11.7) 98 (18.4) pt3b 115 (67.3) 337 (63.3) pt4 30 (17.5) 21 (3.9) Pathological Gleason score 6 or less 29 (17.0) 119 (22.4) 0.02 7 77 (45.0) 270 (50.8) 8 10 65 (38.0) 143 (26.8) Surgical margin status <0.001 Positive 131 (76.6) 313 (58.8) Negative 40 (23.4) 219 (41.2) Number of lymph nodes removed and examined <0.001 Mean (median) 16.4 (15) 13.1 (12) Range 2 52 2 44 Number of positive lymph nodes 0.36 Mean (median) 2.4 (2) 2.2 (1) Range 1 16 1 31 2 121 (70.8) 411 (77.3) >2 50 (29.2) 121 (22.7) 0.10 Data are tabulated according to the type of adjuvant treatment administered. RT: radiotherapy; HT: hormone therapy; PSA: Prostate Specific Antigen. 703 patients included, 453 (64.4%) were treated at Mayo Clinic, Rochester, Minnesota, U.S., while the remaining 250 (35.6%) were treated at Vita-Salute University, Milan, Italy (Table 1). Significant differences in terms of mean age, pre-operative PSA, pathological stage as well as number of lymph nodes removed and examined were found between the two treatment Institutions (all p 0.04; Table 1). All patients received adjuvant treatments: 532 out of 703 (75.7%) received adjuvant HT alone, while the remaining 171 (24.3%) received a combination of adjuvant RT and HT. As shown in Table 2, patients treated with adjuvant RT + HT had significantly higher rate of locally advanced prostate cancer, higher pathological Gleason score, higher rate of positive surgical margins as well as higher mean number of lymph nodes removed (all p 0.02). Conversely no difference has been found between the two groups in terms of mean age at surgery, preoperative PSA and number of positive nodes (all p 0.06). The most-informative positive lymph-node cutoff in predicting prostate CSS was 2 ( p < 0.001). Thus, patients were divided into two groups according to this cut-off: 2 positive nodes (n = 532; 75.7%) and more than 2 positive nodes (n = 171; 24.3%). Interestingly, significant differences in terms of mean PSA, pathological stage and Gleason score as well as surgical margin status were found between the two groups, in favour of patients with lower volume of nodal disease (all p 0.04; Table 3). Moreover patients with more than 2 positive nodes were significantly younger and received a more extensive PLND at the time of RP (all p < 0.001; Table 3). Figs. 1 3 show Kaplan Meier curves assessing BCR free survival, OS and CSS rates after surgery, respectively. Mean follow-up was 113.7 months (median: 112.5, range 3.5 243). Overall, 109 out of 703 (15.5%) patients died of prostate cancer, while 140 (19.9%) died of other causes. Biochemical recurrence free-survival rate at 5, 8,10 and 15 years was 71, 62, 58 and 49%, respectively.

european urology 55 (2009) 261 270 265 Table 3 Clinical characteristic of 703 patients with lymph node invasion (LNI) submitted to radical prostatectomy, extended pelvic lymph node dissection and adjuvant therapy Variables Patients with 2 positive nodes (n = 532) Patients with >2 positive nodes (n = 171) p value Age <0.001 Mean (median) 65.5 (66) 63.4 (65) Range 48 80 47 79.3 Pre-operative PSA 0.04 Mean (median) 27.1 (15.1) 34.6 (22.4) Range 0.9 616 2.9 248 Pathological stage (2002 AJCC) 0.004 pt2a/b/c 73 (13.7) 9 (5.3) pt3a 95 (17.9) 23 (13.5) pt3b 330 (62.0) 122 (71.3) pt4 34 (6.4) 17 (9.9) Pathological Gleason score 0.001 6 or less 126 (23.7) 22 (12.9) 7 264 (49.6) 83 (48.5) 8 10 142 (26.7) 66 (38.6) Surgical margin status 0.001 Positive 318 (59.8) 126 (73.7) Negative 214 (40.2) 45 (26.3) Number of lymph nodes removed and examined Mean (median) 13.1 (12) 16.3 (16) Range 2 52 4 43 <0.001 Number of positive lymph nodes Mean (median) 1.3 (1) 5.3 (4) <0.001 Range 1 2 3 31 Data are tabulated according to the number of positive nodes (2 vs 2 positive nodes). PSA: Prostate Specific Antigen. Overall survival rate at 5, 8,10 and 15 years was 84, 74, 67 and 49%, respectively. Cancer specific survival rate at 5, 8,10 and 15 years was 90, 85, 82 and 78%, respectively. Patients with 2 or less positive nodes had significantly better CSS outcome at 15 year follow-up compared to patients with more than 2 positive nodes (84% vs 62; p < 0.001; Fig. 4). Table 4 shows the univariable and multivariable Cox regression models predicting CSS. At univariable analysis, all variables, except for year at surgery ( p = 0.88), adjuvant RT ( p = 0.19) and pre-operative PSA ( p = 0.18), were significantly associated with CSS (all p values0.001). The number of positive nodes dichotomized according to the most-infor- Fig. 1 Kaplan-Meier estimates of biochemical recurrence (BCR) free survival after surgery in the entire population (n = 703; dotted lines represent 95% confidence intervals). Fig. 2 Kaplan-Meier estimates of overall survival (OS) after surgery in the entire population (n = 703; dotted lines represent 95% confidence).

266 european urology 55 (2009) 261 270 Fig. 3 Kaplan-Meier estimates of cancer specific survival (CSS) after surgery in the entire population (n = 703; dotted lines represent 95% confidence). mative cut-off was highly significantly associated with CSS at univariable analysis ( p < 0.001). Interestingly, the number of positive nodes (2 vs >2 positive nodes) maintained a highly significant association with CSS even at multivariable analyses ( p = 0.002), as well as pathological stage ( p = 0.03), pathological Gleason score ( p < 0.001) and surgical margin status ( p = 0.008). Patients with more than 2 positive nodes had a 1.9-fold higher risk of dying for prostate cancer compared to patients with 2 or less positive nodes after accounting for all the other predictors. Of note, adjuvant RT reached a highly independent predictor status for CSS prediction ( p = 0.002) at multivariable analysis. Similarly, LND was highly significantly associated with CSS at univariable and multivariable analyses (all p < 0.001). Finally, in the univariable PA analyses, the number of positive nodes was the third most informative predictor of CSS (PA: 60.3%), after pathological Gleason score (PA:63.8%) and LND (61.8%; Table 3). Moreover, when the variable accounting for the number of positive nodes (2 vs >2 positive nodes) was added to pathological stage, the multivariable PA reached up to 65.0% and significantly exceeded that of pathological T stage alone (65.0% vs 60.1%; 4.9% gain; p < 0.001; Table 5). The inclusion of the new N staging also significantly increased the PA of multivariable models based on PSA, pathological stage, Gleason score, surgical margin status and adjuvant RT (72.1% vs 69.9%, 2.2% gain, p < 0.001). Similarly, when LND was included in multivariable models predicting CSS significant increases in PA were reached (all p < 0.001). However, the effect of LND on CSS predictions was slightly inferior compared to the new N staging based on the number of positive nodes (Table 5). 4. Discussion The incidence of nodal metastases in patients affected by prostate cancer and submitted to RP has dramatically decreased in the PSA era [1]. Nevertheless, even in the most selected clinical series, nodal metastases are still diagnosed in up to 40% of patients submitted to eplnd [11]. However, although the presence of nodal metastases represents an adverse pathological characteristic, not all patients with prostate cancer and LNI are at the same risk of BCR ad cancer specific death [14 18]. Several trials have indeed shown that patients with Fig. 4 Kaplan-Meier estimates of cancer specific survival (CSS) after surgery according to the number of positive nodes (1 vs 2 positive nodes; log rank test: p = 0.29; =2 vs more than 2 positive nodes; log rank test: p < 0.001).

european urology 55 (2009) 261 270 267 Table 4 Univariable and Multivariable Cox regression models predicting prostate cancer specific survival and corresponding predictive accuracy. Univariable Analysis HR; p value Multivariable Analysis HR; p value Multivariable Analysis HR; p value Univariable Predictive Accuracy Pathological stage (2002 AJCC) ; 0.001 ; 0.03 ; 0.06 60.1% T3a vs T2a/b/c 2.0; 0.2 1.3; 0.6 1.4; 0.6 T3b vs T2a/b/c 3.8; 0.01 2.1; 0.1 2.2; 0.14 T4 vs T2a/b/c 7.1; 0.001 4.0; 0.02 3.7; 0.03 New N staging 60.3% >2 vs 2 positive nodes 2.4; <0.001 1.9; 0.002 ; Pre-operative PSA 1.002; 0.18 1.0; 0.7 1.0; 0.6 58.5% Pathological Gleason score ; <0.001 ; <0.001 ; <0.001 63.8% 7 vs 2 6 2.4; 0.01 1.8; 0.09 1.8; 0.10 8 10 vs 2 6 5.4; <0.001 3.9; <0.001 4.0; <0.001 Surgical margin status (positive vs negative) 2.4; <0.001 1.9; 0.008 1.7; 0.02 59.7% Adjuvant RT 0.7; 0.19 0.42; 0.002 0.49; 0.01 54.8% Year of surgery ; 0.88 ; 0.7 ; 0.48 58.4% 1993 1997 vs 1988 1992 0.91; 0.66 0.8; 0.4 0.8; 0.23 1998 2003 vs 1988 1992 0.89; 0.68 0.9; 0.87 0.8; 0.48 Lymph node density 13.7; <0.001 ; 6.3; <0.001 61.8% PSA: Prostate Specific Antigen; RT: radiotherapy; HR: hazard ratio. Lymph node density: number of positive nodes over number of nodes removed. Table 5 Accuracy of models predicting prostate cancer specific survival Baseline Predictive Accuracy Predictive accuracy after inclusion of new N staging (>2 vs 2 positive nodes) Predictive accuracy after the inclusion of lymph node density Pathological stage alone (2002 AJCC) 60.1% 65.0% (+4.9%^) 64.9% (+4.8%^) Pre-operative PSA + Pathological stage + Pathological 69.9% 72.1% (+2.2%^) 71.9% (+2.0%^) Gleason score + Surgical margin status + Adjuvant RT ^Mantel Haenszel test after bootstraping: p < 0.001. Lymph node density: number of positive nodes over number of nodes removed. PSA: Prostate Specific Antigen; RT: radiotherapy. low volume of lymph node metastases have significantly higher CSS rates compared to patients with more extensive LNI [14 18]. Thus, the number of positive nodes is key for determining patient prognosis. Despite these evidences, the revised 2002 AJCC staging system for prostate cancer does not provide any stratification of patients with nodal metastases according to the number of positive nodes. Patients are indeed classified as having (N1) or not (N0) nodal metastases [19]. Conversely, the N classification of the 1992 prostate cancer AJCC staging system included a stratification of node positive prostate cancer patients into two different categories (N1 and N2) according to the number and size of nodes involved by metastatic spread [20]. Therefore, the AJCC lymph node staging has been simplified in 2002. However, these changes might not translate into comparable CSS predictive accuracies due to the different outcomes of node positive patients according to the extent of LNI [14 18]. Based on these controversies, we aimed at quantifying the accuracy of the revised 2002 AJCC staging system in predicting CSS of node positive patients and to compare it with the accuracy of multivariable models including the number of positive nodes as a covariate. We performed our analysis in the largest series of node positive prostate cancer available (n = 703) treated with RP, eplnd and adjuvant treatments. Patients were sub-divided into two groups according to the most-informative cut-off of positive nodes for CSS prediction (namely, 2 vs >2 positive nodes). Interestingly, we confirmed the overall excellent

268 european urology 55 (2009) 261 270 cancer specific outcome of LNI patients. Prostate CSS rate at 15 years was as high as 79%. Similarly, about 50% of patients treated with a multimodal approach were free of BCR at 15-year follow-up. Moreover, we confirmed the key role represented by the number of positive nodes in predicting CSS of patients with LNI. Men with more than 2 positive nodes had significantly less favourable cancer characteristics at diagnosis (all p 0.04; Table 3) and lower CSS rates at 15 year follow-up compared to patients with 2 positive nodes (62% vs 84%, respectively; p < 0.001; Fig. 4). This was confirmed by univariable and multivariable Cox regression models, where patients with more than 2 positive nodes had significantly lower CSS rates, even after accounting for all other key CSS predictors (such as pre-operative PSA, pathological stage and Gleason score, year of surgery, surgical margin status and adjuvant RT). Patients with more than 2 positive nodes had roughly a 2-fold higher risk of dying for prostate cancer compared to patients with 2 positive nodes after accounting for the effect of all the other mentioned predictors (Table 4). Finally, we compared the predictive accuracy of the 2002 AJCC staging system (not accounting for the number of positive nodes) in predicting CSS with the accuracy a two-variable model including both pathological T stage and the number of positive nodes (namely, 2 vs more than 2 positive nodes). Interestingly a 4.9% gain in bootstrap-corrected PA was reached when data on the number of positive nodes was included into the model (65.0 vs 60.1%; p < 0.001, Table 5). Significant gains in PA were also reached when the number of positive nodes was added to a multivariable post-operative model ( p < 0.001, Table 5). Therefore, a significant increase in the CSS predictive ability has been reached when the number of positive nodes was considered. Interestingly, despite a highly significant association with CSS, the effect of LND on CSS predictions was slightly inferior compared to the new N staging based on the number of positive nodes (Table 5). Several aspects of our study are noteworthy. First, our results reinforce the need for stratification of node positive prostate cancer patients according to the number of positive nodes. Our results are in line with previous studies supporting extremely favourable long term outcome of patients with low volume of nodal disease compared to patients with more extensive nodal invasion [14 18]. Bader et al [16] reported a 78% cause specific survival rate in patients treated with RP and eplnd and who did not undergo any adjuvant therapy until progression. Interestingly, among those with 1 positive node, 39% remained free of clinical or biochemical progression, compared to 12% of patients with 2 or more positive nodes. Similar results have been reported by Daneshmand et al [17] who assessed the long term outcome of a large retrospective node positive series treated with eplnd. Roughly 31% of patients received adjuvant HT. The 10-year clinical recurrence-free survival rate was as high as 65%. When stratified by positive lymph node density (LND), patients with a LND of 20% or greater were at higher risk for clinical recurrence compared to those with a density of less than 20% (relative risk:2.31; p < 0.001). Cheng et al [14] reported a 79% 10-year cause specific survival in a large series of 322 patients treated by RP and prolonged adjuvant HT. Interestingly, after 10 years, CSS rate was as high as 94% in patients with a single node metastasis. This rate was not significantly different from CSS of patients without nodal involvement. However, none of the previous studies assessed the predictive ability of the number of positive nodes for CSS prediction. Boorjan et al previously found > 2 positive nodes to be an independent predictor of outcome at multivariable analyses [18]. Similarly, Schumacher et al have recently reported significantly higher 10-year CSS rates in patients with 2 positive nodes compared to patients with 3 positive nodes not receiving any adjuvant therapy (78.6 vs 33.4%, respectively) [21]. However, this cut-off was arbitrarily used in both studies and the increase in PA associated with the inclusion of the number of positive nodes was not tested. Kattan [22] indeed indicated that statistical significance is not synonymous with predictive ability. A truly informative variable should increase the combined PA of base predictors, in addition to achieving independent, multivariate predictor status. Therefore, we compared bootstrap-corrected PA of the 2002 TNM staging system with and without the number of positive nodes as a predictor. Interestingly, the number of positive nodes was able to significantly increase CSS predictions (gain:4.9%; p < 0.001) when added to the base variable represented by pathological T stage. Therefore, we consider patient classification according to number of positive nodes a key variable for CSS predictions of node positive patients. Second, our results are based on the largest series of node positive prostate cancer published to date (n = 703) assessed at a long-term follow-up. Moreover, all our patients were treated with an anatomically defined extended pelvic lymph node dissection in the PSA era, which avoids any bias related to the diagnosis of LNI as well as to the current applicability of our findings [23].

european urology 55 (2009) 261 270 269 Despite several elements of interest, our study is not devoid of limitations. First, all patients included in the analyses were submitted to adjuvant treatments after surgery. This may be criticized as a confounder of the impact of RP and PLND alone on lymph node positive prostate cancer. However, the use of adjuvant HT reflects a practice that is not uncommon among practicing urologists, based on the evidence of improved overall survival, CSS and progression-free survival of node positive prostate cancer submitted to adjuvant HT [24]. However, it has to be acknowledged that our results might not applicable to node positive patients left untreated after surgery until progression. Nevertheless, even when adjuvant HT was not administered to node positive patients, patients with low volume of nodal burden had significantly better outcome compared to patients with higher number of positive nodes removed [15,16,21]. This may suggest a potential curative role of PLND in selected category of patients. However, such hypothesis cannot be confirmed in our series since all patients received adjuvant HT after surgery. Second, the differences in the population characteristics between the two contributing Institutions might represent another limitation of our study (Table 1). However, the two cohorts did not differ in terms of the extent of LNI (i.e. number of positive nodes) which represented the key variable of our study. Moreover, the effect of all these differences was accounted for by including patient characteristics as covariates in multivariable models. Third, our analysis was in part limited by the differences in the way the lymph nodes were sent for pathological assessment between the two contributing Institutions. Indeed, at Mayo Clinic lymph nodes were submitted en bloc from each side of the pelvis. Conversely, at Vita-Salute University lymph nodes were submitted in multiple packages, which has been shown to improve the quality of pathological assessment of the number of lymph nodes that may be involved with metastatic cancer [25]. This may be responsible for the difference in the mean number of lymph nodes examined between the two Centres. In addition, this may also be responsible for the overall relatively low mean number of lymph nodes examined in this series of extensive nodal dissection (Table 1). Nevertheless, the mean number of nodes removed was 13.9 which is significantly higher compared to limited nodal dissection series, where the mean number of nodes removed was as low as 5.8 [26]. Moreover, the lack of a central pathological re-assessment may represent another important limitation. Indeed, pathological evaluation of node specimens may account for different number of nodes identified and examined. For example, at some Institutions a more rigorous search for lymph nodes within the surgical specimen may yield a higher number of nodes that are submitted for microscopic examination, despite the same number of lymph nodes removed. Indeed, pathological specimen dissection based on tactile findings may miss non-palpable lymph nodes, especially when a large amount of fatty tissue surrounds them. Thus, at pathological evaluation, an LNI diagnostic bias may be introduced if size and consistency are used as criteria for identification of suspicious nodes. 5. Conclusions Our results based on 703 node positive prostate cancer patients treated with RP, eplnd and adjuvant treatments show favourable long-term outcome of patients with LNI. We have demonstrated that the number of positive nodes represents a key variable for CSS predictions. Patients with up to 2 positive nodes experience excellent CSS rate, which was significantly higher compared to patients with more than 2 positive nodes ( p < 0.001). Moreover, a significant improvement in CSS prediction is reached when the number of positive nodes is considered ( p < 0.001). These results reinforce the need for a stratification of node positive patients according to the number of positive nodes and may warrant consideration in the next revision of the pathologic TNM classification. Author contributions: Alberto Briganti had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Briganti, Karnes, Blute, Montorsi. Acquisition of data: Karnes, Gallina, Suardi, Bianchi, Briganti. Analysis and interpretation of data: Briganti, Gallina, Suardi, Da Pozzo, Karnes. Drafting of the manuscript: Briganti, Karnes, Montorsi, Blute. Critical revision of the manuscript for important intellectual content: Rigatti, Montorsi, Cozzarini, Fazio, Briganti, Karnes, Blute, Da Pozzo. Statistical analysis: Briganti, Gallina. Obtaining funding: None. Administrative, technical, or material support: None. Supervision: None. Other (specify): None. Financial disclosures: I certify that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed

270 european urology 55 (2009) 261 270 in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None. Funding/Support and role of the sponsor: None. References [1] Heidehreich A, Aus G, Bolla M, et al. EAU guidelines on prostate cancer. Eur Urol 2008;53:68 80. [2] Freedland DJ, Partin AW, Humphreys EB, et al. Radical prostatectomy for clinical stage T3 disease. Cancer 2007; 109:1273 8. [3] Yossepowitch O, Eggener SE, Serio AM, et al. Secondary therapy, metastatic progression and cancer specific mortality in men with clinically high-risk prostate cancer treated with radical prostatectomy. Eur Urol 2008;53:950 9. [4] Gontero P, Marchioro G, Pisani R, et al. Is radical prostatectomy feasible in all cases of locally advanced non-bone metastatic prostate cancer? Results of a single-institution study. Eur Urol 2007;51:922 30. [5] Morgan WR, Bergstralh EJ, Zincke H. Long-term evaluation of radical prostatectomy as treatment for clinical stage C (T3) prostate cancer. Urology 1993;41:113 20. [6] Van den Ouden D, Davidson PJ, et al. Radical prostatectomy as a monotherapy for locally advanced (stage T3) prostate cancer. J Urol 1994;151:646 51. [7] Gerber GS, Thisted RA, Chodak GW, et al. Results of radical prostatectomy in men with locally advanced prostate cancer: multi-institutional pooled analysis. Eur Urol 1997;32:385 90. [8] Hsu C-Y, Joniau S, Oyen R, Roskams T, Van Poppel H. Outcome of surgery for clinical unilateral T3a prostate cancer: a single-institution experience. Eur Urol 2007;51: 121 9. [9] Briganti A, Chun FK-H, Salonia A, et al. Validation of a nomogram predicting the probability of lymph node invasion among patients undergoing radical prostatectomy and an extended pelvic lymphadenectomy. Eur Urol 2006;49:1019 27. [10] Bader P, Burkhard FC, Markwalder R, et al. Is a limited lymph node dissection an adequate staging procedure for prostate cancer? J Urol 2002;168:514 8. [11] Heidenreich A, Ohlmann CH, Polyakov S. Anatomical extent of pelvic lymphadenectomy in patients undergoing radical prostatectomy. Eur Urol 2007;52:29 37. [12] Heidenreich A, Varga Z, Von Knobloch R. Extended pelvic lymphadenectomy in patients undergoing radical prostatectomy: high incidence of lymph node metastasis. J Urol 2002;167:1681 6. [13] Allaf ME, Palapattu GS, Trock BJ, et al. Anatomical extent of lymph node dissection: impact on men with clinically localized prostate cancer. J Urol 2004;172:1840 4. [14] Cheng L, Zincke H, Blute ML, et al. Risk of prostate carcinoma death in patients with lymph node metastasis. Cancer 2001;91:66 73. [15] Palapattu GS, Allaf ME, Trock BJ, et al. Prostate specific antigen progression in men with lymph node metastases following radical prostatectomy: results of long-term followup. J Urol 2004;172:1860 4. [16] Bader P, Burkhard FC, Markwalder R, et al. Disease progression and survival of patients with positive lymph nodes after radical prostatectomy. Is there a chance of cure? J Urol 2003;169:849 54. [17] Daneshmand S, Quek ML, Stein JP, et al. Prognosis of patients with lymph node positive prostate cancer following radical prostatectomy: long-term results. J Urol 2004;172:2252 5. [18] Boorjian SA, Thompson RH, Siddiqui S, et al. Long-Term Outcome After Radical Prostatectomy for Patients With Lymph Node Positive Prostate Cancer in the Prostate Specific Antigen Era. J Urol 2007;178:864 71. [19] Greene FL, Page DL, Fleming ID, et al. editors. American Joint Committee on Cancer Staging Manual. 6th ed. Philadelphia: Springer; 2002. [20] Beahrs OH, Henson DE, Hutter RVP, Kennedy BJ. American Joint Committee on Cancer staging manual. 4th edition. Philadelphia: J.B. Lippincott; 1992, pp. 181 3. [21] Schumacher MC, Burkhard FC, Thalmann GN, Fleischmann A, Studer UE. Good outcome for patients with few lymph node metastases after radical retropubic prostatectomy. Eur Urol 2008;54:344 52. [22] KattanMW. Judgingnew markers bytheirabilitytoimprove predictive accuracy. J Natl Cancer Inst 2003;95:634 5. [23] Catalona WJ, Smith DS, Ratliff TL, et al. Detection of organ-confined prostate cancer is increased through prostate-specific antigen-based screening. JAMA 1993; 270:948 54. [24] Messing EM, Manola J, Yao J, et al. Immediate versus deferred androgen deprivation treatment in patients with node-positive prostate cancer after radical prostatectomy and pelvic lymphadenectomy. Lancet Oncol 2006;7:472 9. [25] Bochner BH, Herr HW, Reuter VE. Impact of separate versus en bloc pelvic lymph node dissection on the number of lymph nodes retrieved in cystectomy specimens. J Urol 2001;166:2295 6. [26] Berglund RK, Sadetsky N, DuChane J, Carroll PR, Klein EA. Limited pelvic lymph node dissection at the time of radical prostatectomy does not affect 5-year failure rates for low, intermediate and high risk prostate cancer: results from CaPSURE. J Urol 2007;177:526 9.