- Email: [email protected]
Identifying the Best Candidate for Radical Prostatectomy Among Patients with High-Risk Prostate Cancer
EUROPEAN UROLOGY 61 (2012) 584–592
available at www.sciencedirect.com journal homepage: www.europeanurology.com
Prostate Cancer
Identifying the Best Candidate for Radical Prostatectomy Among Patients with High-Risk Prostate Cancer Alberto Briganti a,*,1, Steven Joniau b,1, Paolo Gontero c, Firas Abdollah a, Niccolo` M. Passoni a, Bertrand Tombal d, Giansilvio Marchioro e, Burkhard Kneitz f, Jochen Walz g, Detlef Frohneberg h, Chris H. Bangma i, Markus Graefen j, Alessandro Tizzani c, Bruno Frea k, R. Jeffrey Karnes l, Francesco Montorsi a, Hein Van Poppel b, Martin Spahn f a
Department of Urology, Vita-Salute University San Raffaele, Milan, Italy; b University Hospitals Leuven, Department of Urology, Leuven, Belgium; c University of Turin,
Department of Urology, Torino, Italy; d Universite´ Catholique De Louvain Department of Urology, Brussels, Belgium; e University of Piemonte Orientale, Department of Urology, Novara, Italy; f University Hospital Wu¨rzburg, Department of Urology and Pediatric Urology, Wu¨rzburg, Germany; g Department of Urology, Institut PaoliCalmettes Cancer Centre, Marseille, France; h Community Hospital Karlsruhe, Department of Urology, Karlsruhe, Germany; i Department of Urology, Erasmus MC, University Medical Centre, Rotterdam, the Netherlands; j Martini-Clinic, Prostate Cancer Centre, University Medical Centre Hamburg-Eppendorf, Hamburg, Germany; k
Department of Urology, University of Udine, Udine, Italy; l Department of Urology, Mayo Medical School and Mayo Clinic, Rochester, Minnesota, USA
Article info
Abstract
Article history: Accepted November 22, 2011 Published online ahead of print on December 1, 2011
Background: The current role of radical prostatectomy (RP) in patients with high-risk disease remains controversial. Objective: To identify which high-risk prostate cancer (PCa) patients might have favorable pathologic outcomes when surgically treated. Design, setting, and participants: We evaluated 1366 patients with high-risk PCa (ie, at least one of the following risk factors: prostate-specific antigen [PSA] >20 ng/ml, cT3, biopsy Gleason 8–10) treated with RP and pelvic lymph node dissection (PLND) at eight European centers between 1987 and 2009. A favorable pathologic outcome was defined as specimen-confined (SC) disease—namely, pT2–pT3a, node negative PCa with negative surgical margins. Intervention: All patients underwent radical retropubic prostatectomy and PLND. Measurements: Univariable and multivariable logistic regression models tested the association between predictors and SC disease. A logistic regression coefficient-based nomogram was developed and internally validated using 200 bootstrap resamples. The Kaplan-Meier method was used to depict biochemical recurrence (BCR) and cancerspecific survival (CSS) rates. Results and limitations: Overall, 505 of 1366 patients (37%) had SC disease at RP. All preoperative variables (ie, age and PSA at surgery, clinical stage, and biopsy Gleason sum) were independent predictors of SC PCa at RP (all p 0.04). Patients with SC disease had significantly higher 10-yr BCR-free survival and CSS rates than patients without SC disease at RP (66% vs 47% and 98 vs 88%, respectively; all p < 0.001). A nomogram including PSA, age, clinical stage, and biopsy Gleason sum demonstrated 72% accuracy in predicting SC PCa. This study is limited by its retrospective design and by the lack of an external validation of the nomogram. Conclusions: Roughly 40% of patients with high-risk PCa have SC disease at final pathology. These patients showed excellent long-term outcomes when surgically treated, thus representing the ideal candidates for RP as the primary treatment for PCa. Prediction of such patients is possible using a nomogram based on routinely available clinical parameters. # 2011 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Keywords: Prostate cancer Radical prostatectomy High risk Specimen confined disease Nomogram
1
Both authors contributed equally to the manuscript. * Corresponding author. Department of Urology, Vita Salute University – San Raffaele Hospital, Via Olgettina 60 20132 Milan, Italy. Tel. +39 02 26437286; Fax: +39 02 26437298. E-mail addresses: [email protected], [email protected] (A. Briganti).
0302-2838/$ – see back matter # 2011 European Association of Urology. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.eururo.2011.11.043
EUROPEAN UROLOGY 61 (2012) 584–592
1.
Introduction
585
dedicated genitourinary pathologists. All patients were preoperatively staged for metastases with a contrast-enhanced abdominopelvic
Radical prostatectomy (RP) is considered the first-line treatment of patients with localized prostate cancer (PCa) and a life expectancy of at least 10 yr [1]. However, the current role of RP in patients with high-risk disease remains controversial [2–6]. Historically, RP has been discouraged in this patient category because of the assumption that surgery alone might not be curative in these individuals. Therefore, patients with high-risk disease were selected for other treatment modalities, such as hormone therapy (HT) alone or in combination with radiation therapy (RT) [7]. However, several recent reports have shown excellent cancer-specific outcomes for high-risk patients when treated with RP either alone or in combination with adjuvant treatments [2,3,6,8–13], primarily because 22–63% of these patients harbor favorable pathologic characteristics at RP according to the high-risk definition used [8,11–13]. On the basis of these data, even in the absence of prospective randomized trials, the currently available guidelines consider RP optional for selected patients with high-risk disease [1,14,15]. Despite this indication, it may still be argued that not all patients with high-risk PCa would benefit from RP. Intuitively, surgery might indeed be more beneficial when a complete removal of the disease is possible, that is, in patients with tumors confined to the RP specimen. This hypothesis is supported by previous data reporting more favorable survival rates in high-risk PCa patients harboring organ-confined disease at RP [11–13]. However, there is no preoperative tool aimed at predicting the presence of specimen-confined (SC) disease at RP in patients with preoperative high-risk cancer features. Such a tool might be useful for selecting those patients suitable for surgery, thus decreasing the rate of patients potentially exposed to unnecessary treatment-related complications. To address this void, we developed and internally validated a novel tool for predicting favorable pathologic outcome in high-risk patients treated with surgery. 2.
Materials and methods
We evaluated data on 1748 patients with nonmetastatic, clinical highrisk PCa treated with RP and bilateral pelvic lymph node dissection (PLND) between 1987 and 2009 at eight European tertiary care centers. Of these, 382 were excluded because of neoadjuvant therapy (n = 198), prostate-specific antigen (PSA) >100 ng/ml (n = 69), clinical stage T4 (n = 14), as well as lack of biopsy Gleason data (n = 2), preoperative PSA (n = 2), pathologic T stage (n = 2), neoadjuvant therapy information (n = 80), surgical margin (n = 4), and lymph node status (n = 11). This modification resulted in 1366 evaluable patients with at least one of the following risk factors according to the National Comprehensive Cancer Network (NCCN) and the European Association of Urology (EAU) guidelines: preoperative PSA value >20 ng/ml, cT3 clinical stage, and biopsy Gleason score 8 [1,14]. At all centers, clinical stage was assigned according to the 2002 TNM staging system, prostate biopsy cores were obtained under transrectal ultrasound guidance, and pretreatment PSA was measured before digital rectal examination or prostate ultrasound. Biopsy and pathologic grading were assessed according to the Gleason grading system by
computed tomography (CT) and bone scan. SC disease was defined as pT2–pT3a PCa with negative surgical margins (NSM; R0) and no lymph node invasion (LNI; pN0). Biochemical recurrence (BCR) was defined as a PSA value 0.2 ng/ml after RP. Prostate cancer–specific mortality (CSM) was defined as the time from RP to death attributed to PCa. The study was approved by institutional board review at each contributing institution, where required.
2.1.
Statistical analyses
The statistical methods consisted of two steps: (1) development and internal validation of a novel nomogram predicting SC disease at RP and (2) survival analyses in the overall population as well as according to the SC status of the disease. First, univariable and multivariable logistic regression analyses were used to predict the presence of SC disease at RP. Age and PSA at surgery, clinical stage, and biopsy Gleason sum with or without the variable coding for year of surgery were used as predictors. Regression-based coefficients were then used to develop a novel nomogram for predicting SC PCa. The discrimination accuracy of the nomogram was quantified using the receiver operating characteristics– derived area under curve (AUC), where 100% indicates a perfect prediction and 50% is considered equivalent to the toss of a coin. Two hundred bootstrap resamples were used for internal validation and to reduce overfit bias. Bootstrapping is a general approach to statistical inference based on building a sampling distribution for a statistic by resampling from the data at hand. The extent of over- or underestimation of pathologically confirmed SC disease rates versus the nomogram-predicted probabilities were explored graphically using calibration plots. As a second step, KaplanMeier plots were used to depict BCR-free survival and CSM-free survival rates in the entire cohort as well as according to the pathologic status of the disease (ie, SC vs non-SC). Kaplan-Meier plots were also used to graphically explore BCR-free survival rates in patients with SC disease according to the number of preoperative risk factors (ie, 1 vs 2 vs 3). Differences in patient survival were tested using the log-rank test. All statistical analyses were performed using R statistical package system (R Foundation for Statistical Computing, Vienna, Austria), with a two-sided significance level set at p < 005.
3.
Results
Of 1366 patients included in the study, 505 (37%) had SC disease at RP (Table 1). The number of risk factors of highrisk PCa was 1, 2, and 3 in 67%, 27%, and 6% of cases, respectively. The rate of SC disease significantly decreased with the increasing number of risk factors. Patients with 1, 2, and 3 risk factors for high-risk PCa harbored SC disease in 45.2%, 22.6%, and 9% of cases, respectively ( p < 0001; Table 1). Except for the extent of PLND ( p = 0.06), significant differences were found regarding all clinical and pathologic characteristics between patients with SC and those without SC disease at RP (all p 004; Table 1). Similarly, significant clinical and pathologic differences were found according to treating institution (all p 0.001; Table 2). Table 3 shows the univariable and multivariable logistic regression models testing the association between predictors and SC disease at RP. At multivariable analyses, all variables tested were significantly associated with the presence of SC disease (all p 004). Figure 1 shows the regression coefficient–based nomogram derived from
586
EUROPEAN UROLOGY 61 (2012) 584–592
Table 1 – Descriptive statistics of 1366 patients with clinical high-risk prostate cancer treated with radical prostatectomy and pelvic lymph node dissection between 1987 and 2009 at eight European institutions; data were stratified according to tumor extent (ie, non–specimenconfined versus specimen-confined disease) Overall (n = 1366 [100%]), no. (%) Age, yr: Mean Median Range PSA, ng/ml: Mean Median Range PSA, ng/ml: 10 10–20 20–50 50–100 Clinical tumor stage: cT1 cT2 cT3 Biopsy Gleason score: 6 7 8 NCCN/EAU risk factors: One Two Three Pathologic tumor stage: pT2 pT3a pT3b pT4 PSM: Overall pT2 disease Pathologic Gleason score: 6 7 8 Downstaging (cT3 to pT2) Upstaging (cT1/cT2 to pT3) LNI No. of lymph nodes removed: Mean Median Range Overall adjuvant therapies: Missing Adjuvant RT alone Adjuvant HT alone Adjuvant RT plus HT Year of surgery: 1987–1994 1995–1999 2000–2004 2005–2009
Patients with non–specimen-confined disease (n = 861 [63.0%]), no. (%)
Patients with specimen-confined disease (n = 505 [37.0%]), no. (%)
p value
65.4 66.0 41–88.9
65.7 66.4 41.0–88.9
64.9 66 46–84.2
0.04
24.9 21.3 0.5–100.0
28.1 23.6 0.5–100
19.4 15.1 1.0–100
<0.001
168 167 402 124
(19.5) (19.4) (46.7) (14.4)
184 (36.4) 102 (20.2) 196 (38.8) 23 (4.6)
<0.001
163 (11.9) 424 (31.0) 779 (57.0)
90 (10.5) 280 (32.5) 491 (57.0)
73 (14.5) 144 (28.5) 288 (57.0)
0.05
576 (42.2) 443 (32.4) 347 (25.4)
302 (35.1) 298 (34.6) 261 (30.3)
274 (54.3) 145 (28.7) 86 (17.0)
<0.001
916 (67.1) 372 (27.2) 78 (5.7)
502 (58.3) 288 (33.4) 71 (8.2)
414 (82.0) 84 (16.6) 7 (1.4)
<0.001
341 478 448 99
(25.0) (34.8) (32.9) (7.2)
69 (8.0) 245 (28.5) 448 (52) 99 (11.5)
272 (53.9) 233 (46.1) – –
N/A
613 (44.9) 60 (17.6)
613 (71.2) N/A
– –
N/A N/A
445 524 397 151 397 313
217 330 314 N/A N/A 313
228 (45.1) 194 (38.4) 83 (16.4) N/A N/A –
<0.001
352 269 598 147
(25.8) (19.7) (43.8) (10.8)
(32.6) (38.4) (29.1) (11.1) (29.1) (22.9)
(25.2) (38.3) (36.5)
(36.4)
11.5 10.0 2.0–49.0 657 (48.1) 121 (8.9) 112 (8.2) 406 (29.7) 139 (10.2)
11.8 10.0 2.0–49.0 575 (66.4) 89 (10.3) 98 (11.4) 345 (40.1) 129 (15)
11.0 9.0 2.0–37.0 85 (16.8) 32 (6.3) 14 (2.8) 61 (12.1) 10 (2)
226 489 471 180
164 288 295 114
62 (12.3) 201 (39.8) 176 (34.9) 66 (13.1)
(16.5) (35.8) (34.5) (13.2)
(19) (33.4) (34.3) (13.2)
N/A N/A N/A 0.06
<0.001 <0.001 <0.001 <0.001 0.006
PSA = prostate-specific antigen; NCCN = National Comprehensive Cancer Network; EAU = European Association of Urology; N/A = not applicable; PSM = positive surgical margin; LNI = lymph node invasion; RT = radiation therapy; HT = hormone therapy.
the multivariable model. Analysis of nomogram axes revealed that all variables were inversely related to risk points. After 200 bootstrap resamples, the nomogram demonstrated 72% accuracy. The calibration plot showed perfect concordance between observed and predicted probabilities of SC disease (Fig. 2). Figures 3–6 show
Kaplan-Meier curves depicting BCR- and CSM-free survival rates in the entire cohort (Fig. 3 and 4, respectively) and according to SC disease at RP (Fig. 5 and 6, respectively). Mean follow-up time for BCR and cancer-specific survival (CSS) analyses was 126.9 and 186.6 mo, respectively (median: not achieved). Patients with SC disease had
587
EUROPEAN UROLOGY 61 (2012) 584–592
Table 2 – Descriptive statistics of 1366 patients with clinical high-risk prostate cancer treated with radical prostatectomy and pelvic lymph node dissection between 1987 and 2009 at eight European institutions; data were stratified according to treating institution Wurzburg (n = 488 [35.7%]), no. (%) Age, yr: Mean 66.1 Median 67.0 Range 43.0–80.0 PSA, ng/ml: Mean 29.4 Median 24.0 Range 1.4–100.0 PSA, ng/ml: 10 95 (19.5) 10–20 94 (19.3) 20–50 213 (43.6) 50–100 86 (17.6) Clinical tumor stage: cT1c 22 (4.5) cT2 149 (30.5) cT3 317 (65.0) Biopsy Gleason score: 6 230 (47.1) 7 143 (29.3) 8 115 (23.6) NCCN/EAU risk factors: One 281 (57.6) Two 162 (33.2) Three 45 (9.2) Pathologic tumor stage: pT2 72 (14.8) pT3a 181 (37.1) pT3b 176 (36.1) pT4 59 (12.1) PSM: Overall 250 (51.2) pT2 disease 9 (12.5) Pathologic Gleason score 6 222 (45.5) 7 145 (29.7) 8 121 (24.8) Downstaging 36 (7.4) (cT3 to pT2) Upstaging 135 (27.7) (cT1/cT2 to pT3) LNI 141 (28.9) No. of lymph nodes removed: Mean 9.0 Median 8.0 Range 2.0–30.0 Year of surgery: 1987–1994 99 (20.3) 1995–1999 230 (47.1) 2000–2004 157 (32.2) 2005–2009 2 (0.4)
Milan (n = 287 [21%]), no. (%)
Leuven (n = 282 [20.6%]), no. (%)
Brussels (n = 135 [10.0%]), no. (%)
Novara (n = 57 [4.2%]), no. (%)
Rotterdam (n = 48 [3.5%]), no. (%)
Hamburg (n = 42 [3%]), no. (%)
Turin (n = 27 [2.0%]), no. (%)
p value
66.1 67.5 44.4–88.9
64.3 65.0 41–79
65.2 67.2 46.3–75.1
65.7 66.0 48.0–75.0
64.4 64.0 53.0–73.0
62.6 62.0 51.0–73.0
63.8 65.0 51.0–71.0
<0.001
23.6 19.6 0.5–100.0
18.7 13.0 1.0–100.0
22.5 21.4 2.1–74.8
31.1 27.6 20.0–78.2
7.5 5.8 2.1–25.1
34.6 26.0 20.6–99.2
36.3 29.8 18.5–98.6
<0.001
(38.7) (24.1) (32.3) (5.0)
23 (17) 33 (24.4) 75 (55.6) 4 (3.0)
0 6 47 4
36 (75.0) 10 (20.8) 2 (4.2) 0 (0)
0 (0) 0 (0) 35 (83.3) 7 (16.7)
0 (0) 3 (11.1) 20 (74.1) 4 (14.8)
<0.001
83 (28.9) 101 (35.2) 103 (35.9)
22 (7.8) 35 (12.4) 225 (79.8)
19 (14.1) 74 (54.8) 42 (31.1)
3 (5.3) 18 (31.6) 36 (63.2)
0 (0) 0 (0) 48 (100.0)
7 (16.7) 30 (71.4) 5 (11.9)
7 (25.9) 17 (63.0) 3 (11.1)
<0.001
73 (25.4) 85 (29.6) 129 (44.9)
131 (46.5) 103 (36.5) 48 (17.0)
69 (51.1) 38 (28.1) 28 (20.7)
21 (36.8) 24 (42.1) 12 (21.1)
23 (47.9) 18 (37.5) 7 (14.6)
16 (38.1) 23 (54.8) 3 (7.1)
13 (48.1) 9 (33.3) 5 (18.5)
<0.001
209 (72.8) 68 (23.7) 10 (3.5)
191 (67.7) 83 (29.4) 8 (2.8)
119 (88.1) 14 (10.4) 2 (1.5)
20 (35.1) 26 (45.6) 11 (19.3)
40 (83.3) 7 (14.6) 1 (2.1)
34 (81.0) 8 (19.0) 0 (0)
22 (81.5) 4 (14.8) 1 (3.7)
<0.001
101 59 115 12
70 135 65 12
(24.8) (47.9) (23.0) (4.3)
49 (36.3) 39 (28.9) 47 (34.8) 0 (0)
14 19 19 5
(24.6) (33.3) (33.3) (8.8)
22 (45.8) 18 (37.5) 4 (8.3) 4 (8.3)
6 (14.3) 15 (35.7) 17 (40.5) 4 (9.5)
7 (25.9) 10 (37.0) 7 (25.9) 3 (11.1)
<0.001
115 (40.1) 20 (19.8)
108 (38.3) 10 (14.3)
58 (43.0) 11 (22.4)
40 (70.2) 4 (28.6)
17 (35.4) 3 (13.6)
12 (28.6) 0 (0)
13 (48.1) 3 (42.9)
<0.001 0.27
38 124 125 28
67 129 86 52
(23.8) (45.7) (30.5) (18.4)
70 (51.9) 38 (28.1) 27 (20.0) 9 (6.7)
15 20 22 3
22 (45.8) 23 (47.9) 3 (6.2) 22 (45.8)
6 (14.3) 35 (83.3) 1 (2.4) 0 (0)
5 (18.5) 10 (37.0) 12 (44.4) 1 (3.7)
<0.001
111 (38.7)
39 (13.8)
53 (39.3)
10 (17.5)
0 (0)
31 (73.8)
18 (66.7)
<0.001
98 (34.1)
29 (10.3)
17 (12.6)
15 (26.3)
2 (4.2)
4 (9.5)
7 (25.9)
<0.001
N/A
13.0 12.0 4.0–24.0
N/A
N/A
N/A
35 41 46 13
0 6 38 13
9 (18.8) 31 (64.6) 8 (16.7) 0 (0)
14 (33.3) 28 (66.7) 0 (0) 0 (0)
0 (0) 0 (0) 22 (81.5) 5 (18.5)
89 55 115 28
(31) (19.2) (40.1) (9.8)
(35.2) (20.6) (40.1) (4.2)
(13.2) (43.2) (43.6) (9.8)
17.0 16.0 2.0–49.0 11 25 109 142
(3.8) (8.7) (38) (49.5)
109 68 91 14
9.0 8.0 2.0–36.0 58 128 91 5
(20.6) (45.4) (32.3) (1.8)
(25.9) (30.4) (34.1) (9.6)
(0) (10.5) (82.5) (7.0)
(26.3) (35.1) (38.6) (5.3)
(0) (10.5) (66.7) (22.8)
<0.001
0.001
<0.001
PSA = prostate-specific antigen; NCCN = National Comprehensive Cancer Network; EAU = European Association of Urology; PSM = positive surgical margin; LNI = lymph node invasion; N/A = not available.
significantly higher 10-yr BCR- and CSM-free survival rates than patients without SC PCa at RP (65.6% vs 47.4% and 98.2% vs 87.6%, respectively; Fig. 5 and 6, respectively; all p < 0001). Interestingly, even higher 10-yr BCR-free survival rates were noted when patients with SC disease were stratified according to the number of preoperative risk factors (ie, 1 vs >1; 68.2% vs 56.8, respectively; p = 003; Fig. 7).
4.
Discussion
RP as a treatment option for patients with high-risk disease has witnessed an increasing interest over the past decade. Such trend stems from the growing evidence that surgery either alone or in combination with adjuvant treatments is associated with favorable cancer control outcomes [2,3, 6,9–13]. However, despite this evidence, the optimal
588
EUROPEAN UROLOGY 61 (2012) 584–592
Table 3 – Univariable and multivariable analyses predicting specimen-confined disease (pT2–3a R0 pN0) in 1366 patients with clinical highrisk prostate cancer treated with radical prostatectomy and pelvic lymph node dissection between 1987 and 2009 at eight European institutions* Univariable logistic regression analyses Variables Age at surgery, yr PSA, ng/ml Clinical tumor stage: cT1c cT2 cT3 Biopsy Gleason score: 6 7 8 Yr of surgery: 1987–1994 1995–1999 2000–2004 2005–2009 Multivariable accuracy
Multivariable logistic regression analyses (Model 1)
Multivariable logistic regression analyses (Model 2)
OR (95% CI)
p value
OR (95% CI)
p value
0.98 (0.97–1.00) 0.97 (0.97–0.98)
0.04 <0.001
0.98 (0.97–1.00) 0.97 (0.96–0.98)
0.05 <0.001
0.98 (0.97–1.00) 0.97 (0.96–0.97)
0.04 <0.001
1.00 (Ref) 0.63 (0.44–0.92) 0.72 (0.51–1.02)
0.04 0.05 0.01
1.00 (Ref) 0.7 (0.47–1.04) 0.52 (0.35–0.76)
0.002 0.07 0.001
1.00 (Ref) 0.65 (0.44–0.95) 0.51 (0.35–0.73)
0.001 0.02 <0.001
1.00 (Ref) 0.54 (0.41–0.69) 0.36 (0.27–0.49)
<0.001 <0.001 <0.001
1.00 (Ref) 0.54 (0.41–0.71) 0.29 (0.21–0.41)
<0.001 <0.001 <0.001
1.00 (Ref) 0.55 (0.42–0.73) 0.31 (0.22–0.42)
<0.001 <0.001 <0.001
1.00 1.85 1.58 1.53 –
0.03 0.006 <0.001 0.01 –
1.00 (Ref) 1.71 (1.18–2.48) 1.47 (1.01–2.15) 1.72 (1.08–2.73) 71.0%
0.03 0.004 0.04 0.02 –
(Ref) (1.31–2.6) (1.12–2.23) (1.01–2.33)
OR (95% CI)
–
72.0%
p value
–
–
OR = odds ratio; CI = confidence interval; PSA = prostate-specific antigen. Multivariable analyses included age, PSA, clinical stage, biopsy Gleason sum with (Model 1) or without (Model 2) year of surgery as covariate.
*
[(Fig._1)TD$IG]
Fig. 1 – Nomogram predicting the probability of specimen-confined (SC) disease (pT2–3a,R0,N0) in patients undergoing radical prostatectomy and pelvic lymph node dissection based on age and prostate-specific antigen level at surgery, clinical stage, and biopsy Gleason score. Instructions: Locate the patient’s age at surgery on the age-at-surgery axis. Draw a line straight upward to the point axis to determine how many points toward the probability of SC disease the patient receives for his age. Repeat the process for each additional variable. Sum the points for each predictor. Locate the final sum on the total-point axis. Draw a line straight down to find the patient’s probability of having SC disease.
EUROPEAN UROLOGY 61 (2012) 584–592
[(Fig._2)TD$IG]
[(Fig._4)TD$IG]
589
Fig. 4 – Kaplan-Meier estimates of overall cancer-specific mortality–free survival. Fig. 2 – Nomogram calibration plot. The dotted line indicates the location of the ideal nomogram, in which predicted and actual probabilities are identical. The broken line indicates actual nomogram performance. Expected performance on future data is represented through the solid line.
[(Fig._5)TD$IG]
[(Fig._3)TD$IG]
Fig. 3 – Kaplan-Meier estimates of overall biochemical recurrence–free survival. CE = cumulative number of events; NR = number of patients at risk.
management of patients with high-risk PCa is still controversial. Indeed, no prospective evidence exists supporting the higher effectiveness of RP compared to other treatment modalities. Moreover, several studies have shown that men with high-risk PCa represent a highly heterogeneous patient category with no uniform prognosis [9–13]. Therefore, it might be postulated that even in the high-risk setting, the optimal treatment modality should be tailored to individual patient characteristics. For instance, it is likely that not all high-risk PCa patients will benefit from surgery because of the inability to
Fig. 5 – Kaplan-Meier estimates of biochemical recurrence–free survival according to pathologic status (ie, specimen confined vs non–specimen confined; log-rank test: p < 0.001). SCD = specimen-confined disease; NSCD = non–specimen-confined disease.
completely eradicate the primary tumor. Conversely, those patients with favorable pathologic characteristics at RP will obtain the greatest benefit from surgery alone [11]. However, which high-risk patients will harbor favorable PCa at pathologic evaluation is currently unknown. This is key, because patients in whom total excision of malignant cells is potentially achievable would be ideally considered as the best candidates for surgery. To address this limitation, we set out to develop a novel tool for predicting the presence of SC disease (pT2–pT3a, pN0 with NSM) at RP. Such an end point was used on the basis of previous
590
[(Fig._6)TD$IG]
EUROPEAN UROLOGY 61 (2012) 584–592
Fig. 6 – Kaplan-Meier estimates of cancer-specific mortality–free survival according to pathologic status (ie, specimen confined vs non–specimen confined; log-rank test: p < 0.001). SCD = specimen-confined disease; NSCD = non–specimen-confined disease; CE = cumulative number of events; NR = number of patients at risk.
[(Fig._7)TD$IG]
Fig. 7 – Kaplan-Meier estimates of biochemical recurrence–free survival in patients with specimen-confined disease (n = 505) according to the number of preoperative risk factors (ie, 1 vs >1; log-rank test: p < 0.001). RF = risk factor; CE = cumulative number of events; NR = number of patients at risk.
evidence reporting no differences in cancer progression between pT2 and pT3a disease [16]. Our cohort consisted of 1366 patients with high-risk PCa, treated with RP and PLND dissection at eight European tertiary care centers. Of these, 37% had SC disease at RP specimen. This rate increased to 45% in patients with only one preoperative high-risk feature (either PSA >20 ng/ml, biopsy Gleason score 8–10, or
clinical stage T3 or higher disease), while the rate decreased to only 9% in patients sharing all three preoperative highrisk features ( p < 0.001). To facilitate the task of selecting high-risk PCa patients who may represent the best candidates for surgery, we developed and internally validated a nomogram for predicting the presence of SC disease at RP. Our novel nomogram was based on routinely available clinical variables, such as age and PSA at surgery, clinical tumor stage, and biopsy Gleason score (Fig. 1). The nomogram showed favorable discrimination accuracy (72%) and virtually perfect calibration characteristics (Fig. 2). Therefore, it might be useful in the context of selecting those high-risk patients suitable for an extensive surgical approach. Moreover, when long-term oncologic outcomes were evaluated, significantly higher BCR-free survival rates in men with SC versus non-SC disease were found (10-yr BCR: 65.6% vs 47.4% for non-SC patients; p < 0001). Similar trends were observed when CSM-free survival rates were analyzed. Specifically, the 10-yr CSM-free survival rate was 98% in patients with SC disease versus 88% in patients with non-SC disease ( p < 0.001). Such differences in patient outcome between the two groups are even more significant if we consider that only a minority (16.8%) of men with SC disease was treated with adjuvant therapy after RP (Table 1). Conversely, two-thirds of men with non-SC disease received adjuvant HT and/or RT (Table 1). Therefore, even in the absence of a control group of patients untreated for PCa, these results strongly support the impact of the curative role of surgery alone in the vast majority of men with high-risk PCa when complete tumor excision can be achieved during RP. As a consequence, these individuals represent the optimal candidates for RP. Several results of our study are noteworthy. First, we corroborated previous findings demonstrating that patients with high-risk PCa have excellent favorable cancer control outcomes after surgery with or without adjuvant therapies. The overall CSS was as high as 91% at 10-yr follow-up. Several previous studies reported similar findings [9–13]. However, many of these reports were limited by their relatively small sample size as well as by limited follow-up [10,12,13]. We circumvented these limitations, evaluating the largest series available of clinically high-risk PCa patients assessed at a mean follow-up of roughly 15 yr after surgery. Second, our results reiterate that patients classified as high risk by the NCCN and EAU guidelines represent a heterogeneous group with different cancer control outcomes. In consequence, including all these patients in one category may decrease our ability to provide the most suitable treatment option to these individuals. Among patients with high-risk disease, older men had a higher risk of being diagnosed with non-SC disease even after accounting for all the other predictors (Table 3). These data confirm previous findings reporting age as an independent predictor of adverse pathologic outcomes after RP [17]. Third, we were able to identify a subgroup of patients with high-risk PCa (ie, patients with SC disease) that benefit the most from surgery. Thus, providing RP to these
EUROPEAN UROLOGY 61 (2012) 584–592
individuals may maximize their chances of cure. Fourth, we developed and internally validated the first tool aimed at accurately predicting the presence of SC disease in patients with preoperative high-risk features. The use of this novel tool might help physicians in the clinical decision-making process. Identification of high-risk patients more likely to be cured by surgery can also be implemented by different imaging modalities, such as 11C-choline positron emission tomography/CT and magnetic resonance imaging (MRI). Although these sophisticated approaches have failed to reach high sensitivity in the assessment of nodal status prior to surgery [18], increasing evidence supports the role of MRI using functional imaging techniques in the local staging of PCa [19]. Therefore, the use of such an approach might increase the accuracy of pathologic predictions, thus optimizing the selection of patients in whom primary cancer control is possible. However, even when optimal predictions are made, those high-risk patients showing adverse pathologic outcomes should still be considered as candidates for a multimodal, combined approach. In these cases, cancer control after surgery should be optimized by either adjuvant RT, HT, or a combination of both treatments [20–24]. In our study, roughly two-thirds of patients with non–specimen-confined disease at RP received combined treatment modalities. Although a proportion of these patients might have been potentially overtreated, multimodality still represents a key factor in the management of high-risk patients treated with curative intent, as shown by previous prospective, randomized trials [20–23,25]. Despite these strengths, our study is not devoid of limitations. First, it might be argued that the discrimination accuracy of our model (AUC: 72%) is far from being perfect, despite excellent performance characteristics of the model across all risk ranges (Fig. 2). However, men with high-risk PCa represent a wide continuum of patients with heterogeneous clinical characteristics. In consequence, it may be more challenging to accurately predict patient outcome in these patients compared to more homogeneous patient cohorts. Moreover, the accuracy of this tool is in line with that of previous models addressing pathologic stage at RP in high-risk disease [26]. Moreover, it is likely that inclusion of further clinical data not systematically available in our database, such as clinical T3 substages, prostate volume, and number and percentage of positive cores, may increase the accuracy of our model [27,28]. Furthermore, lack of data on tumor volume in the RP specimen represents another limitation of our study. Second, the population included was treated across a wide time period, ranging between 1987 and 2009. This is key because patient characteristics as well methods for cancer grading and assessment might have significantly changed over such a period of time. However, we accounted for this potential confounding factor by including the variable coding for year of surgery as co-covariate in multivariable models. Interestingly, it demonstrated a significant association with the presence of SC disease at RP. Despite this association, its inclusion failed to increase the predictive accuracy of multivariable models (72% vs 71%, respectively; Table 3). Therefore, it did not represent an
591
informative variable for the aim of our study [29]. Moreover, patients were treated at eight different institutions where differences in tumor assessment and grading might have biased the results of our analyses (Table 2). Lack of central pathologic assessments indeed represents a strong limitation of our retrospective series. However, all recruiting institutions represented tertiary referral centers, and pathologic assessment was performed by dedicated uropathologists. Moreover, when the nomogram was applied separately to each high-volume contributing institution (>100 patients treated), it maintained comparable accuracy (AUC range: 71–74%). Finally, it was not possible to account for important surgeon-related factors such as surgical skill and expertise, which represent important determinants of cancer control in PCa [30]. Despite these limitations, our nomogram is the first tool addressing the probability of SC disease in a large, surgically treated, high-risk population. This nomogram might be useful in the clinical decision-making process of patients diagnosed with high-risk PCa. However, a formal external validation is needed prior to suggesting the use of our tool in the everyday clinical practice. 5.
Conclusions
Based on our results, we can conclude that roughly 40% of patients with high-risk PCa harbor SC disease at RP. We developed and internally validated the first nomogram for predicting SC disease in clinical high-risk PCa. This nomogram might significantly contribute to the selection of patient candidates to RP as a primary treatment for PCa. Patients with SC disease indeed showed excellent cancer control rates at long-term follow-up achieved by RP alone in the vast majority of the cases. Despite excellent performance characteristics when internally validated, the nomogram needs external validation to test its applicability in other clinical settings. 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, Joniau, Spahn, Gontero. Acquisition of data: Passoni, Abdollah, Tombal, Marchioro, Kneitz, Walz, Frohneberg, Bangma, Graefen, Briganti, Joniau, Spahn, Gontero. Analysis and interpretation of data: Briganti, Joniau, Spahn, Gontero, Karnes. Drafting of the manuscript: Briganti, Joniau, Spahn, Gontero, Abdollah, Passoni. Critical revision of the manuscript for important intellectual content: Briganti, Joniau, Gontero, Abdollah, Passoni, Tombal, Marchioro, Kneitz, Walz, Frohneberg, Bangma, Graefen, Tizzani, Frea, Karnes, Montorsi, Van Poppel, Spahn. Statistical analysis: Briganti, Joniau, Spahn, Gontero, Abdollah, Passoni. Obtaining funding: None. Administrative, technical, or material support: None. Supervision: Montorsi, Van Poppel, Spahn, Frea, Tizzani. 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 in the manuscript (eg, employment/ affiliation, grants or funding, consultancies, honoraria, stock ownership or
592
EUROPEAN UROLOGY 61 (2012) 584–592
options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
[16] Hsu C-Y, Joniau S, Oyen R, et al. Outcome of surgery for clinical unilateral T3a prostate cancer: a single-institution experience. Eur Urol 2007;51:121–9.
Funding/Support and role of the sponsor: None.
[17] Richstone L, Bianco FJ, Shah HH, et al. Radical prostatectomy in men aged >or = 70 years: effect of age on upgrading, upstaging, and the accuracy of a preoperative nomogram. BJU Int 2008;101:
References
541–6.
[1] Heidenreich A, Bellmunt J, Bolla M, et al. EAU guidelines on prostate
[18] Budiharto T, Joniau S, Lerut E, et al. Prospective evaluation of
cancer. Part 1: screening, diagnosis, and treatment of clinically
11
localised disease. Eur Urol 2011;59:61–71.
and diffusion-weighted magnetic resonance imaging for the
[2] Boorjian SA, Karnes RJ, Viterbo R, et al. Long-term survival after radical prostatectomy versus external-beam radiotherapy for
C-choline positron emission tomography/computed tomography
nodal staging of prostate cancer with a high risk of lymph node metastases. Eur Urol 2011;60:125–30.
patients with high-risk prostate cancer. Cancer 2011;117:2883–91.
[19] Hoeks CM, Barentsz JO, Hambrock T, et al. Prostate cancer: multi-
[3] Abdollah F, Sun M, Thuret R, et al. A competing-risks analysis of
parametric MR imaging for detection, localization, and staging.
survival after alternative treatment modalities for prostate cancer patients: 1988–2006. Eur Urol 2011;59:88–95. [4] Eastham JA, Evans CP, Zietman A. What is the optimal management of high risk, clinically localized prostate cancer? Urol Oncol 2010;
Radiology 2011;261:46–66. [20] Bolla M, van Poppel H, Collette L, et al. Postoperative radiotherapy after radical prostatectomy: a randomised controlled trial (EORTC trial 22911). Lancet 2005;366:572–8. [21] Thompson IM, Tangen CM, Paradelo J, et al. Adjuvant radiotherapy
28:557–67. [5] Yossepowitch O, Eastham JA. Role of radical prostatectomy in the
for pathological T3N0M0 prostate cancer significantly reduces risk
treatment of high-risk prostate cancer. Curr Urol Rep 2008;9:203–10.
of metastases and improves survival: long-term followup of a
[6] Cooperberg MR, Vickers AJ, Broering JM, et al. Comparative risk-
randomized clinical trial. J Urol 2009;181:956–62.
adjusted mortality outcomes after primary surgery, radiotherapy,
[22] Wiegel T, Bottke D, Steiner U, et al. Phase III postoperative adjuvant
or androgen-deprivation therapy for localized prostate cancer.
radiotherapy after radical prostatectomy compared with radical
Cancer 2010;116:5226–34.
prostatectomy alone in pT3 prostate cancer with postoperative
[7] Hamilton AS, Albertsen PC, Johnson TK, et al. Trends in the treatment of localized prostate cancer using supplemented cancer registry data. BJU Int 2011;107:576–84.
undetectable prostate-specific antigen: ARO 96-02/AUO AP 09/95. J Clin Oncol 2009;27:2924–30. [23] Messing EM, Manola J, Yao J, et al., Eastern Cooperative Oncology
[8] Yossepowitch O, Eggener SE, Bianco Jr FJ, et al. Radical prostatec-
Group Study EST 3886. Immediate versus deferred androgen dep-
tomy for clinically localized, high risk prostate cancer: critical
rivation treatment in patients with node-positive prostate cancer
analysis of risk assessment methods. J Urol 2007;178:493–9.
after radical prostatectomy and pelvic lymphadenectomy. Lancet
[9] Spahn M, Joniau S, Gontero P, et al. Outcome predictors of radical
Oncol 2006;7:472–9.
prostatectomy in patients with prostate-specific antigen greater
[24] Briganti A, Karnes RJ, Da Pozzo LF, et al. Combination of adjuvant
than 20 ng/ml: a European multi-institutional study of 712
hormonal and radiation therapy significantly prolongs survival of
patients. Eur Urol 2010;58:1–7.
patients with pT2-4 pN+ prostate cancer: results of a matched
[10] Miocinovic R, Berglund RK, Stephenson AJ, et al. Avoiding androgen
analysis. Eur Urol 2011;59:832–40.
deprivation therapy in men with high-risk prostate cancer: the role
[25] Bolla M, de Reijke TM, Van Tienhoven G, et al., EORTC Radiation
of radical prostatectomy as initial treatment. Urology 2011;77:
Oncology Group and Genito-Urinary Tract Cancer Group. Duration
946–50.
of androgen suppression in the treatment of prostate cancer. N Engl
[11] Walz J, Joniau S, Chun FK, et al. Pathological results and rates of treatment failure in high-risk prostate cancer patients after radical prostatectomy. BJU Int 2011;107:765–70. [12] Loeb S, Schaeffer EM, Trock BJ, et al. What are the outcomes of radical prostatectomy for high-risk prostate cancer? Urology 2010;
J Med 2009;360:2516–27. [26] Gallina A, Jeldres C, Chun FK-H, et al. Prediction of pathological stage is inaccurate in men with PSA values above 20 ng/mL. Eur Urol 2007;52:1374–80. [27] Briganti A, Chun FK, Suardi N, et al. Prostate volume and adverse prostate cancer features: fact not artifact. Eur J Cancer 2007;43:
76:710–4. [13] Donohue JF, Bianco Jr FJ, Kuroiwa K, et al. Poorly differentiated
2669–77.
prostate cancer treated with radical prostatectomy: long-term out-
[28] Briganti A, Chun FK-H, Hutterer GC, et al. Systematic assessment of
come and incidence of pathological downgrading. J Urol 2006;176:
the ability of the number and percentage of positive biopsy cores to
991–5.
predict pathologic stage and biochemical recurrence after radical
[14] NCCN guidelines: prostate cancer 2011. National Comprehensive
prostatectomy. Eur Urol 2007;52:733–45.
http://www.nccn.org/clinical.asp.
[29] Kattan MW. Judging new markers by their ability to improve
[15] Thompson I, Thrasher JB, Aus G, et al., AUA Prostate Cancer Clinical
[30] Bianco Jr FJ, Vickers AJ, Cronin AM, et al. Variations among experi-
Guideline Update Panel. Guideline for the management of clinically
enced surgeons in cancer control after open radical prostatectomy.
localized prostate cancer: 2007 update. J Urol 2007;177:2106–31.
J Urol 2010;183:977–82.
Cancer
Network
Web
site.
Accessed April 14, 2011.
predictive accuracy. J Natl Cancer Inst 2003;95:634–5.
available at www.sciencedirect.com journal homepage: www.europeanurology.com
Prostate Cancer
Identifying the Best Candidate for Radical Prostatectomy Among Patients with High-Risk Prostate Cancer Alberto Briganti a,*,1, Steven Joniau b,1, Paolo Gontero c, Firas Abdollah a, Niccolo` M. Passoni a, Bertrand Tombal d, Giansilvio Marchioro e, Burkhard Kneitz f, Jochen Walz g, Detlef Frohneberg h, Chris H. Bangma i, Markus Graefen j, Alessandro Tizzani c, Bruno Frea k, R. Jeffrey Karnes l, Francesco Montorsi a, Hein Van Poppel b, Martin Spahn f a
Department of Urology, Vita-Salute University San Raffaele, Milan, Italy; b University Hospitals Leuven, Department of Urology, Leuven, Belgium; c University of Turin,
Department of Urology, Torino, Italy; d Universite´ Catholique De Louvain Department of Urology, Brussels, Belgium; e University of Piemonte Orientale, Department of Urology, Novara, Italy; f University Hospital Wu¨rzburg, Department of Urology and Pediatric Urology, Wu¨rzburg, Germany; g Department of Urology, Institut PaoliCalmettes Cancer Centre, Marseille, France; h Community Hospital Karlsruhe, Department of Urology, Karlsruhe, Germany; i Department of Urology, Erasmus MC, University Medical Centre, Rotterdam, the Netherlands; j Martini-Clinic, Prostate Cancer Centre, University Medical Centre Hamburg-Eppendorf, Hamburg, Germany; k
Department of Urology, University of Udine, Udine, Italy; l Department of Urology, Mayo Medical School and Mayo Clinic, Rochester, Minnesota, USA
Article info
Abstract
Article history: Accepted November 22, 2011 Published online ahead of print on December 1, 2011
Background: The current role of radical prostatectomy (RP) in patients with high-risk disease remains controversial. Objective: To identify which high-risk prostate cancer (PCa) patients might have favorable pathologic outcomes when surgically treated. Design, setting, and participants: We evaluated 1366 patients with high-risk PCa (ie, at least one of the following risk factors: prostate-specific antigen [PSA] >20 ng/ml, cT3, biopsy Gleason 8–10) treated with RP and pelvic lymph node dissection (PLND) at eight European centers between 1987 and 2009. A favorable pathologic outcome was defined as specimen-confined (SC) disease—namely, pT2–pT3a, node negative PCa with negative surgical margins. Intervention: All patients underwent radical retropubic prostatectomy and PLND. Measurements: Univariable and multivariable logistic regression models tested the association between predictors and SC disease. A logistic regression coefficient-based nomogram was developed and internally validated using 200 bootstrap resamples. The Kaplan-Meier method was used to depict biochemical recurrence (BCR) and cancerspecific survival (CSS) rates. Results and limitations: Overall, 505 of 1366 patients (37%) had SC disease at RP. All preoperative variables (ie, age and PSA at surgery, clinical stage, and biopsy Gleason sum) were independent predictors of SC PCa at RP (all p 0.04). Patients with SC disease had significantly higher 10-yr BCR-free survival and CSS rates than patients without SC disease at RP (66% vs 47% and 98 vs 88%, respectively; all p < 0.001). A nomogram including PSA, age, clinical stage, and biopsy Gleason sum demonstrated 72% accuracy in predicting SC PCa. This study is limited by its retrospective design and by the lack of an external validation of the nomogram. Conclusions: Roughly 40% of patients with high-risk PCa have SC disease at final pathology. These patients showed excellent long-term outcomes when surgically treated, thus representing the ideal candidates for RP as the primary treatment for PCa. Prediction of such patients is possible using a nomogram based on routinely available clinical parameters. # 2011 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Keywords: Prostate cancer Radical prostatectomy High risk Specimen confined disease Nomogram
1
Both authors contributed equally to the manuscript. * Corresponding author. Department of Urology, Vita Salute University – San Raffaele Hospital, Via Olgettina 60 20132 Milan, Italy. Tel. +39 02 26437286; Fax: +39 02 26437298. E-mail addresses: [email protected], [email protected] (A. Briganti).
0302-2838/$ – see back matter # 2011 European Association of Urology. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.eururo.2011.11.043
EUROPEAN UROLOGY 61 (2012) 584–592
1.
Introduction
585
dedicated genitourinary pathologists. All patients were preoperatively staged for metastases with a contrast-enhanced abdominopelvic
Radical prostatectomy (RP) is considered the first-line treatment of patients with localized prostate cancer (PCa) and a life expectancy of at least 10 yr [1]. However, the current role of RP in patients with high-risk disease remains controversial [2–6]. Historically, RP has been discouraged in this patient category because of the assumption that surgery alone might not be curative in these individuals. Therefore, patients with high-risk disease were selected for other treatment modalities, such as hormone therapy (HT) alone or in combination with radiation therapy (RT) [7]. However, several recent reports have shown excellent cancer-specific outcomes for high-risk patients when treated with RP either alone or in combination with adjuvant treatments [2,3,6,8–13], primarily because 22–63% of these patients harbor favorable pathologic characteristics at RP according to the high-risk definition used [8,11–13]. On the basis of these data, even in the absence of prospective randomized trials, the currently available guidelines consider RP optional for selected patients with high-risk disease [1,14,15]. Despite this indication, it may still be argued that not all patients with high-risk PCa would benefit from RP. Intuitively, surgery might indeed be more beneficial when a complete removal of the disease is possible, that is, in patients with tumors confined to the RP specimen. This hypothesis is supported by previous data reporting more favorable survival rates in high-risk PCa patients harboring organ-confined disease at RP [11–13]. However, there is no preoperative tool aimed at predicting the presence of specimen-confined (SC) disease at RP in patients with preoperative high-risk cancer features. Such a tool might be useful for selecting those patients suitable for surgery, thus decreasing the rate of patients potentially exposed to unnecessary treatment-related complications. To address this void, we developed and internally validated a novel tool for predicting favorable pathologic outcome in high-risk patients treated with surgery. 2.
Materials and methods
We evaluated data on 1748 patients with nonmetastatic, clinical highrisk PCa treated with RP and bilateral pelvic lymph node dissection (PLND) between 1987 and 2009 at eight European tertiary care centers. Of these, 382 were excluded because of neoadjuvant therapy (n = 198), prostate-specific antigen (PSA) >100 ng/ml (n = 69), clinical stage T4 (n = 14), as well as lack of biopsy Gleason data (n = 2), preoperative PSA (n = 2), pathologic T stage (n = 2), neoadjuvant therapy information (n = 80), surgical margin (n = 4), and lymph node status (n = 11). This modification resulted in 1366 evaluable patients with at least one of the following risk factors according to the National Comprehensive Cancer Network (NCCN) and the European Association of Urology (EAU) guidelines: preoperative PSA value >20 ng/ml, cT3 clinical stage, and biopsy Gleason score 8 [1,14]. At all centers, clinical stage was assigned according to the 2002 TNM staging system, prostate biopsy cores were obtained under transrectal ultrasound guidance, and pretreatment PSA was measured before digital rectal examination or prostate ultrasound. Biopsy and pathologic grading were assessed according to the Gleason grading system by
computed tomography (CT) and bone scan. SC disease was defined as pT2–pT3a PCa with negative surgical margins (NSM; R0) and no lymph node invasion (LNI; pN0). Biochemical recurrence (BCR) was defined as a PSA value 0.2 ng/ml after RP. Prostate cancer–specific mortality (CSM) was defined as the time from RP to death attributed to PCa. The study was approved by institutional board review at each contributing institution, where required.
2.1.
Statistical analyses
The statistical methods consisted of two steps: (1) development and internal validation of a novel nomogram predicting SC disease at RP and (2) survival analyses in the overall population as well as according to the SC status of the disease. First, univariable and multivariable logistic regression analyses were used to predict the presence of SC disease at RP. Age and PSA at surgery, clinical stage, and biopsy Gleason sum with or without the variable coding for year of surgery were used as predictors. Regression-based coefficients were then used to develop a novel nomogram for predicting SC PCa. The discrimination accuracy of the nomogram was quantified using the receiver operating characteristics– derived area under curve (AUC), where 100% indicates a perfect prediction and 50% is considered equivalent to the toss of a coin. Two hundred bootstrap resamples were used for internal validation and to reduce overfit bias. Bootstrapping is a general approach to statistical inference based on building a sampling distribution for a statistic by resampling from the data at hand. The extent of over- or underestimation of pathologically confirmed SC disease rates versus the nomogram-predicted probabilities were explored graphically using calibration plots. As a second step, KaplanMeier plots were used to depict BCR-free survival and CSM-free survival rates in the entire cohort as well as according to the pathologic status of the disease (ie, SC vs non-SC). Kaplan-Meier plots were also used to graphically explore BCR-free survival rates in patients with SC disease according to the number of preoperative risk factors (ie, 1 vs 2 vs 3). Differences in patient survival were tested using the log-rank test. All statistical analyses were performed using R statistical package system (R Foundation for Statistical Computing, Vienna, Austria), with a two-sided significance level set at p < 005.
3.
Results
Of 1366 patients included in the study, 505 (37%) had SC disease at RP (Table 1). The number of risk factors of highrisk PCa was 1, 2, and 3 in 67%, 27%, and 6% of cases, respectively. The rate of SC disease significantly decreased with the increasing number of risk factors. Patients with 1, 2, and 3 risk factors for high-risk PCa harbored SC disease in 45.2%, 22.6%, and 9% of cases, respectively ( p < 0001; Table 1). Except for the extent of PLND ( p = 0.06), significant differences were found regarding all clinical and pathologic characteristics between patients with SC and those without SC disease at RP (all p 004; Table 1). Similarly, significant clinical and pathologic differences were found according to treating institution (all p 0.001; Table 2). Table 3 shows the univariable and multivariable logistic regression models testing the association between predictors and SC disease at RP. At multivariable analyses, all variables tested were significantly associated with the presence of SC disease (all p 004). Figure 1 shows the regression coefficient–based nomogram derived from
586
EUROPEAN UROLOGY 61 (2012) 584–592
Table 1 – Descriptive statistics of 1366 patients with clinical high-risk prostate cancer treated with radical prostatectomy and pelvic lymph node dissection between 1987 and 2009 at eight European institutions; data were stratified according to tumor extent (ie, non–specimenconfined versus specimen-confined disease) Overall (n = 1366 [100%]), no. (%) Age, yr: Mean Median Range PSA, ng/ml: Mean Median Range PSA, ng/ml: 10 10–20 20–50 50–100 Clinical tumor stage: cT1 cT2 cT3 Biopsy Gleason score: 6 7 8 NCCN/EAU risk factors: One Two Three Pathologic tumor stage: pT2 pT3a pT3b pT4 PSM: Overall pT2 disease Pathologic Gleason score: 6 7 8 Downstaging (cT3 to pT2) Upstaging (cT1/cT2 to pT3) LNI No. of lymph nodes removed: Mean Median Range Overall adjuvant therapies: Missing Adjuvant RT alone Adjuvant HT alone Adjuvant RT plus HT Year of surgery: 1987–1994 1995–1999 2000–2004 2005–2009
Patients with non–specimen-confined disease (n = 861 [63.0%]), no. (%)
Patients with specimen-confined disease (n = 505 [37.0%]), no. (%)
p value
65.4 66.0 41–88.9
65.7 66.4 41.0–88.9
64.9 66 46–84.2
0.04
24.9 21.3 0.5–100.0
28.1 23.6 0.5–100
19.4 15.1 1.0–100
<0.001
168 167 402 124
(19.5) (19.4) (46.7) (14.4)
184 (36.4) 102 (20.2) 196 (38.8) 23 (4.6)
<0.001
163 (11.9) 424 (31.0) 779 (57.0)
90 (10.5) 280 (32.5) 491 (57.0)
73 (14.5) 144 (28.5) 288 (57.0)
0.05
576 (42.2) 443 (32.4) 347 (25.4)
302 (35.1) 298 (34.6) 261 (30.3)
274 (54.3) 145 (28.7) 86 (17.0)
<0.001
916 (67.1) 372 (27.2) 78 (5.7)
502 (58.3) 288 (33.4) 71 (8.2)
414 (82.0) 84 (16.6) 7 (1.4)
<0.001
341 478 448 99
(25.0) (34.8) (32.9) (7.2)
69 (8.0) 245 (28.5) 448 (52) 99 (11.5)
272 (53.9) 233 (46.1) – –
N/A
613 (44.9) 60 (17.6)
613 (71.2) N/A
– –
N/A N/A
445 524 397 151 397 313
217 330 314 N/A N/A 313
228 (45.1) 194 (38.4) 83 (16.4) N/A N/A –
<0.001
352 269 598 147
(25.8) (19.7) (43.8) (10.8)
(32.6) (38.4) (29.1) (11.1) (29.1) (22.9)
(25.2) (38.3) (36.5)
(36.4)
11.5 10.0 2.0–49.0 657 (48.1) 121 (8.9) 112 (8.2) 406 (29.7) 139 (10.2)
11.8 10.0 2.0–49.0 575 (66.4) 89 (10.3) 98 (11.4) 345 (40.1) 129 (15)
11.0 9.0 2.0–37.0 85 (16.8) 32 (6.3) 14 (2.8) 61 (12.1) 10 (2)
226 489 471 180
164 288 295 114
62 (12.3) 201 (39.8) 176 (34.9) 66 (13.1)
(16.5) (35.8) (34.5) (13.2)
(19) (33.4) (34.3) (13.2)
N/A N/A N/A 0.06
<0.001 <0.001 <0.001 <0.001 0.006
PSA = prostate-specific antigen; NCCN = National Comprehensive Cancer Network; EAU = European Association of Urology; N/A = not applicable; PSM = positive surgical margin; LNI = lymph node invasion; RT = radiation therapy; HT = hormone therapy.
the multivariable model. Analysis of nomogram axes revealed that all variables were inversely related to risk points. After 200 bootstrap resamples, the nomogram demonstrated 72% accuracy. The calibration plot showed perfect concordance between observed and predicted probabilities of SC disease (Fig. 2). Figures 3–6 show
Kaplan-Meier curves depicting BCR- and CSM-free survival rates in the entire cohort (Fig. 3 and 4, respectively) and according to SC disease at RP (Fig. 5 and 6, respectively). Mean follow-up time for BCR and cancer-specific survival (CSS) analyses was 126.9 and 186.6 mo, respectively (median: not achieved). Patients with SC disease had
587
EUROPEAN UROLOGY 61 (2012) 584–592
Table 2 – Descriptive statistics of 1366 patients with clinical high-risk prostate cancer treated with radical prostatectomy and pelvic lymph node dissection between 1987 and 2009 at eight European institutions; data were stratified according to treating institution Wurzburg (n = 488 [35.7%]), no. (%) Age, yr: Mean 66.1 Median 67.0 Range 43.0–80.0 PSA, ng/ml: Mean 29.4 Median 24.0 Range 1.4–100.0 PSA, ng/ml: 10 95 (19.5) 10–20 94 (19.3) 20–50 213 (43.6) 50–100 86 (17.6) Clinical tumor stage: cT1c 22 (4.5) cT2 149 (30.5) cT3 317 (65.0) Biopsy Gleason score: 6 230 (47.1) 7 143 (29.3) 8 115 (23.6) NCCN/EAU risk factors: One 281 (57.6) Two 162 (33.2) Three 45 (9.2) Pathologic tumor stage: pT2 72 (14.8) pT3a 181 (37.1) pT3b 176 (36.1) pT4 59 (12.1) PSM: Overall 250 (51.2) pT2 disease 9 (12.5) Pathologic Gleason score 6 222 (45.5) 7 145 (29.7) 8 121 (24.8) Downstaging 36 (7.4) (cT3 to pT2) Upstaging 135 (27.7) (cT1/cT2 to pT3) LNI 141 (28.9) No. of lymph nodes removed: Mean 9.0 Median 8.0 Range 2.0–30.0 Year of surgery: 1987–1994 99 (20.3) 1995–1999 230 (47.1) 2000–2004 157 (32.2) 2005–2009 2 (0.4)
Milan (n = 287 [21%]), no. (%)
Leuven (n = 282 [20.6%]), no. (%)
Brussels (n = 135 [10.0%]), no. (%)
Novara (n = 57 [4.2%]), no. (%)
Rotterdam (n = 48 [3.5%]), no. (%)
Hamburg (n = 42 [3%]), no. (%)
Turin (n = 27 [2.0%]), no. (%)
p value
66.1 67.5 44.4–88.9
64.3 65.0 41–79
65.2 67.2 46.3–75.1
65.7 66.0 48.0–75.0
64.4 64.0 53.0–73.0
62.6 62.0 51.0–73.0
63.8 65.0 51.0–71.0
<0.001
23.6 19.6 0.5–100.0
18.7 13.0 1.0–100.0
22.5 21.4 2.1–74.8
31.1 27.6 20.0–78.2
7.5 5.8 2.1–25.1
34.6 26.0 20.6–99.2
36.3 29.8 18.5–98.6
<0.001
(38.7) (24.1) (32.3) (5.0)
23 (17) 33 (24.4) 75 (55.6) 4 (3.0)
0 6 47 4
36 (75.0) 10 (20.8) 2 (4.2) 0 (0)
0 (0) 0 (0) 35 (83.3) 7 (16.7)
0 (0) 3 (11.1) 20 (74.1) 4 (14.8)
<0.001
83 (28.9) 101 (35.2) 103 (35.9)
22 (7.8) 35 (12.4) 225 (79.8)
19 (14.1) 74 (54.8) 42 (31.1)
3 (5.3) 18 (31.6) 36 (63.2)
0 (0) 0 (0) 48 (100.0)
7 (16.7) 30 (71.4) 5 (11.9)
7 (25.9) 17 (63.0) 3 (11.1)
<0.001
73 (25.4) 85 (29.6) 129 (44.9)
131 (46.5) 103 (36.5) 48 (17.0)
69 (51.1) 38 (28.1) 28 (20.7)
21 (36.8) 24 (42.1) 12 (21.1)
23 (47.9) 18 (37.5) 7 (14.6)
16 (38.1) 23 (54.8) 3 (7.1)
13 (48.1) 9 (33.3) 5 (18.5)
<0.001
209 (72.8) 68 (23.7) 10 (3.5)
191 (67.7) 83 (29.4) 8 (2.8)
119 (88.1) 14 (10.4) 2 (1.5)
20 (35.1) 26 (45.6) 11 (19.3)
40 (83.3) 7 (14.6) 1 (2.1)
34 (81.0) 8 (19.0) 0 (0)
22 (81.5) 4 (14.8) 1 (3.7)
<0.001
101 59 115 12
70 135 65 12
(24.8) (47.9) (23.0) (4.3)
49 (36.3) 39 (28.9) 47 (34.8) 0 (0)
14 19 19 5
(24.6) (33.3) (33.3) (8.8)
22 (45.8) 18 (37.5) 4 (8.3) 4 (8.3)
6 (14.3) 15 (35.7) 17 (40.5) 4 (9.5)
7 (25.9) 10 (37.0) 7 (25.9) 3 (11.1)
<0.001
115 (40.1) 20 (19.8)
108 (38.3) 10 (14.3)
58 (43.0) 11 (22.4)
40 (70.2) 4 (28.6)
17 (35.4) 3 (13.6)
12 (28.6) 0 (0)
13 (48.1) 3 (42.9)
<0.001 0.27
38 124 125 28
67 129 86 52
(23.8) (45.7) (30.5) (18.4)
70 (51.9) 38 (28.1) 27 (20.0) 9 (6.7)
15 20 22 3
22 (45.8) 23 (47.9) 3 (6.2) 22 (45.8)
6 (14.3) 35 (83.3) 1 (2.4) 0 (0)
5 (18.5) 10 (37.0) 12 (44.4) 1 (3.7)
<0.001
111 (38.7)
39 (13.8)
53 (39.3)
10 (17.5)
0 (0)
31 (73.8)
18 (66.7)
<0.001
98 (34.1)
29 (10.3)
17 (12.6)
15 (26.3)
2 (4.2)
4 (9.5)
7 (25.9)
<0.001
N/A
13.0 12.0 4.0–24.0
N/A
N/A
N/A
35 41 46 13
0 6 38 13
9 (18.8) 31 (64.6) 8 (16.7) 0 (0)
14 (33.3) 28 (66.7) 0 (0) 0 (0)
0 (0) 0 (0) 22 (81.5) 5 (18.5)
89 55 115 28
(31) (19.2) (40.1) (9.8)
(35.2) (20.6) (40.1) (4.2)
(13.2) (43.2) (43.6) (9.8)
17.0 16.0 2.0–49.0 11 25 109 142
(3.8) (8.7) (38) (49.5)
109 68 91 14
9.0 8.0 2.0–36.0 58 128 91 5
(20.6) (45.4) (32.3) (1.8)
(25.9) (30.4) (34.1) (9.6)
(0) (10.5) (82.5) (7.0)
(26.3) (35.1) (38.6) (5.3)
(0) (10.5) (66.7) (22.8)
<0.001
0.001
<0.001
PSA = prostate-specific antigen; NCCN = National Comprehensive Cancer Network; EAU = European Association of Urology; PSM = positive surgical margin; LNI = lymph node invasion; N/A = not available.
significantly higher 10-yr BCR- and CSM-free survival rates than patients without SC PCa at RP (65.6% vs 47.4% and 98.2% vs 87.6%, respectively; Fig. 5 and 6, respectively; all p < 0001). Interestingly, even higher 10-yr BCR-free survival rates were noted when patients with SC disease were stratified according to the number of preoperative risk factors (ie, 1 vs >1; 68.2% vs 56.8, respectively; p = 003; Fig. 7).
4.
Discussion
RP as a treatment option for patients with high-risk disease has witnessed an increasing interest over the past decade. Such trend stems from the growing evidence that surgery either alone or in combination with adjuvant treatments is associated with favorable cancer control outcomes [2,3, 6,9–13]. However, despite this evidence, the optimal
588
EUROPEAN UROLOGY 61 (2012) 584–592
Table 3 – Univariable and multivariable analyses predicting specimen-confined disease (pT2–3a R0 pN0) in 1366 patients with clinical highrisk prostate cancer treated with radical prostatectomy and pelvic lymph node dissection between 1987 and 2009 at eight European institutions* Univariable logistic regression analyses Variables Age at surgery, yr PSA, ng/ml Clinical tumor stage: cT1c cT2 cT3 Biopsy Gleason score: 6 7 8 Yr of surgery: 1987–1994 1995–1999 2000–2004 2005–2009 Multivariable accuracy
Multivariable logistic regression analyses (Model 1)
Multivariable logistic regression analyses (Model 2)
OR (95% CI)
p value
OR (95% CI)
p value
0.98 (0.97–1.00) 0.97 (0.97–0.98)
0.04 <0.001
0.98 (0.97–1.00) 0.97 (0.96–0.98)
0.05 <0.001
0.98 (0.97–1.00) 0.97 (0.96–0.97)
0.04 <0.001
1.00 (Ref) 0.63 (0.44–0.92) 0.72 (0.51–1.02)
0.04 0.05 0.01
1.00 (Ref) 0.7 (0.47–1.04) 0.52 (0.35–0.76)
0.002 0.07 0.001
1.00 (Ref) 0.65 (0.44–0.95) 0.51 (0.35–0.73)
0.001 0.02 <0.001
1.00 (Ref) 0.54 (0.41–0.69) 0.36 (0.27–0.49)
<0.001 <0.001 <0.001
1.00 (Ref) 0.54 (0.41–0.71) 0.29 (0.21–0.41)
<0.001 <0.001 <0.001
1.00 (Ref) 0.55 (0.42–0.73) 0.31 (0.22–0.42)
<0.001 <0.001 <0.001
1.00 1.85 1.58 1.53 –
0.03 0.006 <0.001 0.01 –
1.00 (Ref) 1.71 (1.18–2.48) 1.47 (1.01–2.15) 1.72 (1.08–2.73) 71.0%
0.03 0.004 0.04 0.02 –
(Ref) (1.31–2.6) (1.12–2.23) (1.01–2.33)
OR (95% CI)
–
72.0%
p value
–
–
OR = odds ratio; CI = confidence interval; PSA = prostate-specific antigen. Multivariable analyses included age, PSA, clinical stage, biopsy Gleason sum with (Model 1) or without (Model 2) year of surgery as covariate.
*
[(Fig._1)TD$IG]
Fig. 1 – Nomogram predicting the probability of specimen-confined (SC) disease (pT2–3a,R0,N0) in patients undergoing radical prostatectomy and pelvic lymph node dissection based on age and prostate-specific antigen level at surgery, clinical stage, and biopsy Gleason score. Instructions: Locate the patient’s age at surgery on the age-at-surgery axis. Draw a line straight upward to the point axis to determine how many points toward the probability of SC disease the patient receives for his age. Repeat the process for each additional variable. Sum the points for each predictor. Locate the final sum on the total-point axis. Draw a line straight down to find the patient’s probability of having SC disease.
EUROPEAN UROLOGY 61 (2012) 584–592
[(Fig._2)TD$IG]
[(Fig._4)TD$IG]
589
Fig. 4 – Kaplan-Meier estimates of overall cancer-specific mortality–free survival. Fig. 2 – Nomogram calibration plot. The dotted line indicates the location of the ideal nomogram, in which predicted and actual probabilities are identical. The broken line indicates actual nomogram performance. Expected performance on future data is represented through the solid line.
[(Fig._5)TD$IG]
[(Fig._3)TD$IG]
Fig. 3 – Kaplan-Meier estimates of overall biochemical recurrence–free survival. CE = cumulative number of events; NR = number of patients at risk.
management of patients with high-risk PCa is still controversial. Indeed, no prospective evidence exists supporting the higher effectiveness of RP compared to other treatment modalities. Moreover, several studies have shown that men with high-risk PCa represent a highly heterogeneous patient category with no uniform prognosis [9–13]. Therefore, it might be postulated that even in the high-risk setting, the optimal treatment modality should be tailored to individual patient characteristics. For instance, it is likely that not all high-risk PCa patients will benefit from surgery because of the inability to
Fig. 5 – Kaplan-Meier estimates of biochemical recurrence–free survival according to pathologic status (ie, specimen confined vs non–specimen confined; log-rank test: p < 0.001). SCD = specimen-confined disease; NSCD = non–specimen-confined disease.
completely eradicate the primary tumor. Conversely, those patients with favorable pathologic characteristics at RP will obtain the greatest benefit from surgery alone [11]. However, which high-risk patients will harbor favorable PCa at pathologic evaluation is currently unknown. This is key, because patients in whom total excision of malignant cells is potentially achievable would be ideally considered as the best candidates for surgery. To address this limitation, we set out to develop a novel tool for predicting the presence of SC disease (pT2–pT3a, pN0 with NSM) at RP. Such an end point was used on the basis of previous
590
[(Fig._6)TD$IG]
EUROPEAN UROLOGY 61 (2012) 584–592
Fig. 6 – Kaplan-Meier estimates of cancer-specific mortality–free survival according to pathologic status (ie, specimen confined vs non–specimen confined; log-rank test: p < 0.001). SCD = specimen-confined disease; NSCD = non–specimen-confined disease; CE = cumulative number of events; NR = number of patients at risk.
[(Fig._7)TD$IG]
Fig. 7 – Kaplan-Meier estimates of biochemical recurrence–free survival in patients with specimen-confined disease (n = 505) according to the number of preoperative risk factors (ie, 1 vs >1; log-rank test: p < 0.001). RF = risk factor; CE = cumulative number of events; NR = number of patients at risk.
evidence reporting no differences in cancer progression between pT2 and pT3a disease [16]. Our cohort consisted of 1366 patients with high-risk PCa, treated with RP and PLND dissection at eight European tertiary care centers. Of these, 37% had SC disease at RP specimen. This rate increased to 45% in patients with only one preoperative high-risk feature (either PSA >20 ng/ml, biopsy Gleason score 8–10, or
clinical stage T3 or higher disease), while the rate decreased to only 9% in patients sharing all three preoperative highrisk features ( p < 0.001). To facilitate the task of selecting high-risk PCa patients who may represent the best candidates for surgery, we developed and internally validated a nomogram for predicting the presence of SC disease at RP. Our novel nomogram was based on routinely available clinical variables, such as age and PSA at surgery, clinical tumor stage, and biopsy Gleason score (Fig. 1). The nomogram showed favorable discrimination accuracy (72%) and virtually perfect calibration characteristics (Fig. 2). Therefore, it might be useful in the context of selecting those high-risk patients suitable for an extensive surgical approach. Moreover, when long-term oncologic outcomes were evaluated, significantly higher BCR-free survival rates in men with SC versus non-SC disease were found (10-yr BCR: 65.6% vs 47.4% for non-SC patients; p < 0001). Similar trends were observed when CSM-free survival rates were analyzed. Specifically, the 10-yr CSM-free survival rate was 98% in patients with SC disease versus 88% in patients with non-SC disease ( p < 0.001). Such differences in patient outcome between the two groups are even more significant if we consider that only a minority (16.8%) of men with SC disease was treated with adjuvant therapy after RP (Table 1). Conversely, two-thirds of men with non-SC disease received adjuvant HT and/or RT (Table 1). Therefore, even in the absence of a control group of patients untreated for PCa, these results strongly support the impact of the curative role of surgery alone in the vast majority of men with high-risk PCa when complete tumor excision can be achieved during RP. As a consequence, these individuals represent the optimal candidates for RP. Several results of our study are noteworthy. First, we corroborated previous findings demonstrating that patients with high-risk PCa have excellent favorable cancer control outcomes after surgery with or without adjuvant therapies. The overall CSS was as high as 91% at 10-yr follow-up. Several previous studies reported similar findings [9–13]. However, many of these reports were limited by their relatively small sample size as well as by limited follow-up [10,12,13]. We circumvented these limitations, evaluating the largest series available of clinically high-risk PCa patients assessed at a mean follow-up of roughly 15 yr after surgery. Second, our results reiterate that patients classified as high risk by the NCCN and EAU guidelines represent a heterogeneous group with different cancer control outcomes. In consequence, including all these patients in one category may decrease our ability to provide the most suitable treatment option to these individuals. Among patients with high-risk disease, older men had a higher risk of being diagnosed with non-SC disease even after accounting for all the other predictors (Table 3). These data confirm previous findings reporting age as an independent predictor of adverse pathologic outcomes after RP [17]. Third, we were able to identify a subgroup of patients with high-risk PCa (ie, patients with SC disease) that benefit the most from surgery. Thus, providing RP to these
EUROPEAN UROLOGY 61 (2012) 584–592
individuals may maximize their chances of cure. Fourth, we developed and internally validated the first tool aimed at accurately predicting the presence of SC disease in patients with preoperative high-risk features. The use of this novel tool might help physicians in the clinical decision-making process. Identification of high-risk patients more likely to be cured by surgery can also be implemented by different imaging modalities, such as 11C-choline positron emission tomography/CT and magnetic resonance imaging (MRI). Although these sophisticated approaches have failed to reach high sensitivity in the assessment of nodal status prior to surgery [18], increasing evidence supports the role of MRI using functional imaging techniques in the local staging of PCa [19]. Therefore, the use of such an approach might increase the accuracy of pathologic predictions, thus optimizing the selection of patients in whom primary cancer control is possible. However, even when optimal predictions are made, those high-risk patients showing adverse pathologic outcomes should still be considered as candidates for a multimodal, combined approach. In these cases, cancer control after surgery should be optimized by either adjuvant RT, HT, or a combination of both treatments [20–24]. In our study, roughly two-thirds of patients with non–specimen-confined disease at RP received combined treatment modalities. Although a proportion of these patients might have been potentially overtreated, multimodality still represents a key factor in the management of high-risk patients treated with curative intent, as shown by previous prospective, randomized trials [20–23,25]. Despite these strengths, our study is not devoid of limitations. First, it might be argued that the discrimination accuracy of our model (AUC: 72%) is far from being perfect, despite excellent performance characteristics of the model across all risk ranges (Fig. 2). However, men with high-risk PCa represent a wide continuum of patients with heterogeneous clinical characteristics. In consequence, it may be more challenging to accurately predict patient outcome in these patients compared to more homogeneous patient cohorts. Moreover, the accuracy of this tool is in line with that of previous models addressing pathologic stage at RP in high-risk disease [26]. Moreover, it is likely that inclusion of further clinical data not systematically available in our database, such as clinical T3 substages, prostate volume, and number and percentage of positive cores, may increase the accuracy of our model [27,28]. Furthermore, lack of data on tumor volume in the RP specimen represents another limitation of our study. Second, the population included was treated across a wide time period, ranging between 1987 and 2009. This is key because patient characteristics as well methods for cancer grading and assessment might have significantly changed over such a period of time. However, we accounted for this potential confounding factor by including the variable coding for year of surgery as co-covariate in multivariable models. Interestingly, it demonstrated a significant association with the presence of SC disease at RP. Despite this association, its inclusion failed to increase the predictive accuracy of multivariable models (72% vs 71%, respectively; Table 3). Therefore, it did not represent an
591
informative variable for the aim of our study [29]. Moreover, patients were treated at eight different institutions where differences in tumor assessment and grading might have biased the results of our analyses (Table 2). Lack of central pathologic assessments indeed represents a strong limitation of our retrospective series. However, all recruiting institutions represented tertiary referral centers, and pathologic assessment was performed by dedicated uropathologists. Moreover, when the nomogram was applied separately to each high-volume contributing institution (>100 patients treated), it maintained comparable accuracy (AUC range: 71–74%). Finally, it was not possible to account for important surgeon-related factors such as surgical skill and expertise, which represent important determinants of cancer control in PCa [30]. Despite these limitations, our nomogram is the first tool addressing the probability of SC disease in a large, surgically treated, high-risk population. This nomogram might be useful in the clinical decision-making process of patients diagnosed with high-risk PCa. However, a formal external validation is needed prior to suggesting the use of our tool in the everyday clinical practice. 5.
Conclusions
Based on our results, we can conclude that roughly 40% of patients with high-risk PCa harbor SC disease at RP. We developed and internally validated the first nomogram for predicting SC disease in clinical high-risk PCa. This nomogram might significantly contribute to the selection of patient candidates to RP as a primary treatment for PCa. Patients with SC disease indeed showed excellent cancer control rates at long-term follow-up achieved by RP alone in the vast majority of the cases. Despite excellent performance characteristics when internally validated, the nomogram needs external validation to test its applicability in other clinical settings. 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, Joniau, Spahn, Gontero. Acquisition of data: Passoni, Abdollah, Tombal, Marchioro, Kneitz, Walz, Frohneberg, Bangma, Graefen, Briganti, Joniau, Spahn, Gontero. Analysis and interpretation of data: Briganti, Joniau, Spahn, Gontero, Karnes. Drafting of the manuscript: Briganti, Joniau, Spahn, Gontero, Abdollah, Passoni. Critical revision of the manuscript for important intellectual content: Briganti, Joniau, Gontero, Abdollah, Passoni, Tombal, Marchioro, Kneitz, Walz, Frohneberg, Bangma, Graefen, Tizzani, Frea, Karnes, Montorsi, Van Poppel, Spahn. Statistical analysis: Briganti, Joniau, Spahn, Gontero, Abdollah, Passoni. Obtaining funding: None. Administrative, technical, or material support: None. Supervision: Montorsi, Van Poppel, Spahn, Frea, Tizzani. 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 in the manuscript (eg, employment/ affiliation, grants or funding, consultancies, honoraria, stock ownership or
592
EUROPEAN UROLOGY 61 (2012) 584–592
options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
[16] Hsu C-Y, Joniau S, Oyen R, et al. Outcome of surgery for clinical unilateral T3a prostate cancer: a single-institution experience. Eur Urol 2007;51:121–9.
Funding/Support and role of the sponsor: None.
[17] Richstone L, Bianco FJ, Shah HH, et al. Radical prostatectomy in men aged >or = 70 years: effect of age on upgrading, upstaging, and the accuracy of a preoperative nomogram. BJU Int 2008;101:
References
541–6.
[1] Heidenreich A, Bellmunt J, Bolla M, et al. EAU guidelines on prostate
[18] Budiharto T, Joniau S, Lerut E, et al. Prospective evaluation of
cancer. Part 1: screening, diagnosis, and treatment of clinically
11
localised disease. Eur Urol 2011;59:61–71.
and diffusion-weighted magnetic resonance imaging for the
[2] Boorjian SA, Karnes RJ, Viterbo R, et al. Long-term survival after radical prostatectomy versus external-beam radiotherapy for
C-choline positron emission tomography/computed tomography
nodal staging of prostate cancer with a high risk of lymph node metastases. Eur Urol 2011;60:125–30.
patients with high-risk prostate cancer. Cancer 2011;117:2883–91.
[19] Hoeks CM, Barentsz JO, Hambrock T, et al. Prostate cancer: multi-
[3] Abdollah F, Sun M, Thuret R, et al. A competing-risks analysis of
parametric MR imaging for detection, localization, and staging.
survival after alternative treatment modalities for prostate cancer patients: 1988–2006. Eur Urol 2011;59:88–95. [4] Eastham JA, Evans CP, Zietman A. What is the optimal management of high risk, clinically localized prostate cancer? Urol Oncol 2010;
Radiology 2011;261:46–66. [20] Bolla M, van Poppel H, Collette L, et al. Postoperative radiotherapy after radical prostatectomy: a randomised controlled trial (EORTC trial 22911). Lancet 2005;366:572–8. [21] Thompson IM, Tangen CM, Paradelo J, et al. Adjuvant radiotherapy
28:557–67. [5] Yossepowitch O, Eastham JA. Role of radical prostatectomy in the
for pathological T3N0M0 prostate cancer significantly reduces risk
treatment of high-risk prostate cancer. Curr Urol Rep 2008;9:203–10.
of metastases and improves survival: long-term followup of a
[6] Cooperberg MR, Vickers AJ, Broering JM, et al. Comparative risk-
randomized clinical trial. J Urol 2009;181:956–62.
adjusted mortality outcomes after primary surgery, radiotherapy,
[22] Wiegel T, Bottke D, Steiner U, et al. Phase III postoperative adjuvant
or androgen-deprivation therapy for localized prostate cancer.
radiotherapy after radical prostatectomy compared with radical
Cancer 2010;116:5226–34.
prostatectomy alone in pT3 prostate cancer with postoperative
[7] Hamilton AS, Albertsen PC, Johnson TK, et al. Trends in the treatment of localized prostate cancer using supplemented cancer registry data. BJU Int 2011;107:576–84.
undetectable prostate-specific antigen: ARO 96-02/AUO AP 09/95. J Clin Oncol 2009;27:2924–30. [23] Messing EM, Manola J, Yao J, et al., Eastern Cooperative Oncology
[8] Yossepowitch O, Eggener SE, Bianco Jr FJ, et al. Radical prostatec-
Group Study EST 3886. Immediate versus deferred androgen dep-
tomy for clinically localized, high risk prostate cancer: critical
rivation treatment in patients with node-positive prostate cancer
analysis of risk assessment methods. J Urol 2007;178:493–9.
after radical prostatectomy and pelvic lymphadenectomy. Lancet
[9] Spahn M, Joniau S, Gontero P, et al. Outcome predictors of radical
Oncol 2006;7:472–9.
prostatectomy in patients with prostate-specific antigen greater
[24] Briganti A, Karnes RJ, Da Pozzo LF, et al. Combination of adjuvant
than 20 ng/ml: a European multi-institutional study of 712
hormonal and radiation therapy significantly prolongs survival of
patients. Eur Urol 2010;58:1–7.
patients with pT2-4 pN+ prostate cancer: results of a matched
[10] Miocinovic R, Berglund RK, Stephenson AJ, et al. Avoiding androgen
analysis. Eur Urol 2011;59:832–40.
deprivation therapy in men with high-risk prostate cancer: the role
[25] Bolla M, de Reijke TM, Van Tienhoven G, et al., EORTC Radiation
of radical prostatectomy as initial treatment. Urology 2011;77:
Oncology Group and Genito-Urinary Tract Cancer Group. Duration
946–50.
of androgen suppression in the treatment of prostate cancer. N Engl
[11] Walz J, Joniau S, Chun FK, et al. Pathological results and rates of treatment failure in high-risk prostate cancer patients after radical prostatectomy. BJU Int 2011;107:765–70. [12] Loeb S, Schaeffer EM, Trock BJ, et al. What are the outcomes of radical prostatectomy for high-risk prostate cancer? Urology 2010;
J Med 2009;360:2516–27. [26] Gallina A, Jeldres C, Chun FK-H, et al. Prediction of pathological stage is inaccurate in men with PSA values above 20 ng/mL. Eur Urol 2007;52:1374–80. [27] Briganti A, Chun FK, Suardi N, et al. Prostate volume and adverse prostate cancer features: fact not artifact. Eur J Cancer 2007;43:
76:710–4. [13] Donohue JF, Bianco Jr FJ, Kuroiwa K, et al. Poorly differentiated
2669–77.
prostate cancer treated with radical prostatectomy: long-term out-
[28] Briganti A, Chun FK-H, Hutterer GC, et al. Systematic assessment of
come and incidence of pathological downgrading. J Urol 2006;176:
the ability of the number and percentage of positive biopsy cores to
991–5.
predict pathologic stage and biochemical recurrence after radical
[14] NCCN guidelines: prostate cancer 2011. National Comprehensive
prostatectomy. Eur Urol 2007;52:733–45.
http://www.nccn.org/clinical.asp.
[29] Kattan MW. Judging new markers by their ability to improve
[15] Thompson I, Thrasher JB, Aus G, et al., AUA Prostate Cancer Clinical
[30] Bianco Jr FJ, Vickers AJ, Cronin AM, et al. Variations among experi-
Guideline Update Panel. Guideline for the management of clinically
enced surgeons in cancer control after open radical prostatectomy.
localized prostate cancer: 2007 update. J Urol 2007;177:2106–31.
J Urol 2010;183:977–82.
Cancer
Network
Web
site.
Accessed April 14, 2011.
predictive accuracy. J Natl Cancer Inst 2003;95:634–5.