Received: March 6, 2013 Accepted after revision: July 9, 2013 Published online: January 25, 2014
Urol Int 2014;92:300–305 DOI: 10.1159/000354323
Validation of Nomograms Predicting Lymph Node Involvement in Patients with Prostate Cancer Undergoing Extended Pelvic Lymph Node Dissection Alexander I. Hinev a Deyan Anakievski a Nikolay H. Kolev b Vesselin I. Hadjiev c a
Clinic of Urology, ‘St. Marina’ University Hospital, Varna, b Clinic of Urology, ‘G. Stranski’ University Hospital, Pleven, and c Department of Statistics, University of Economics, Varna, Bulgaria
Key Words Prostate cancer · Extended pelvic lymph node dissection · Lymph node involvement · Nomograms · Validation
Abstract Our aim was to validate Briganti’s nomograms predicting the probability of lymph node involvement (LNI) in prostate cancer (PCa). Clinicopathological data of 256 PCa patients who underwent extended pelvic lymph node dissection (ePLND) and radical prostatectomy (RP) were obtained from two Bulgarian institutions. Predicted probabilities of LNI were assessed using Briganti’s nomograms based on ePLND. In addition to the established basic LNI predictors, Briganti’s nomograms included the number of lymph nodes removed (version 2006) and the number and percentage of positive biopsy cores (versions 2007 and 2012). The accuracy of these nomograms was compared with the updated Memorial Sloan-Kettering Cancer Center (MSKCC) nomogram (version 2011). Receiver-operating characteristics analysis was done to assess the discriminative ability of each of the nomograms applied. All of Briganti’s nomograms showed a higher predictive accuracy as compared with the updated MSKCC nomogram. The respective AUC values were calculated as 0.847, 0.837, 0.858 and 0.875 for the four Briganti nomograms, and 0.770 for the updated MSKCC nomogram, respectively. De-
© 2014 S. Karger AG, Basel 0042–1138/14/0923–0300$39.50/0 E-Mail [email protected]
spite the potential for heterogeneity in patient selection and management, all predictions demonstrated high concordance with actual observations. Compared with other similar prognostic tools the updated Briganti nomogram (version 2012) showed the highest predictive accuracy and should therefore be preferred. © 2014 S. Karger AG, Basel
Accurate assessment of lymph node involvement (LNI) is of paramount importance for all newly diagnosed prostate cancer (PCa) patients. It is well known that radical prostatectomy (RP) is most effective when the disease is confined to the prostate, while the presence of LNI is related to disease aggressiveness and poor oncological outcome [1–3]. On the other side, patients who have a low probability of having LNI might be spared from pelvic lymph node dissection (PLND) and from the additional morbidity related to it .
This study was presented at the 5th Uro-Oncology Winter Congress, Skopje, Macedonia, January 30–February 2, 2013, and published as an abstract in Yeni Uroloji Dergisi (The New Journal of Urology), Special Edition, 2013, pp 147–148.
Alexander I. Hinev Clinic of Urology ‘St. Marina’ University Hospital BG–9010 Varna (Bulgaria) E-Mail ahinev @ yahoo.com
The aim of the present bi-institutional study was to validate all of Briganti’s nomograms, predicting the probability of LNI in a mixed cohort of Bulgarian PCa patients.
Numerous pre- and postoperative nomograms published in the last two decades [1–10] were based on multivariate logistic regression analyses of large series of patients with known follow-up. Multiple clinical variables were used for the purpose, e.g. serum prostate-specific antigen (PSA), Gleason score, clinical stage, cancer volume, PSA density, etc. Some of those most commonly used are the preoperative nomograms of the Memorial Sloan-Kettering Cancer Center (MSKCC) in New York [5, 9]. These nomograms can be found on the respective MSKCC website (http://mskcc.org/applications/nomograms/ prostate/ index.aspx). Nomograms predicting the probability of LNI have recently been developed in Europe as well. The most popular of them are those created by Briganti and his team [4, 6, 7]. Unlike most of the previous predictive tools, they had been based on the results of extended PLND (ePLND), supported by the current EAU guideline committee as the preferable template of lymph node (LN) dissection in PCa . In addition to the established basic LNI predictors (PSA, clinical stage and biopsy Gleason score), Briganti’s nomograms included the total number of LNs removed  and the number and percentage of positive biopsy cores [4, 7]. The newest version of Briganti’s nomograms (version 2012)  was included in the most recent EAU guidelines, version 2013 . A calculated probability of LNI 100.0 Clinical stage (1992), n (%) T1 /a,b,c/ T2 /a,b,c/ T3 /a,b,c/ Biopsy Gleason score, n (%) 5–6 7 (3+4) 7 (4+3) 8–10 Number of biopsy cores, mean, median (range) Number of positive cores, mean, median (range) Percentage of positive cores, mean, median (range) Number of LNs harvested, mean, median (range) Gleason score at surgery, n (%) 5–6 7 (3+4) 7 (4+3) 8–10 RP pathological details, n (%) OCD EPE SVI LNI PSM
180 64.9, 66 (46–78)
256 64.8, 66 (46–78)
– 15 (19.7) 39 (51.3) 20 (26.3) 2 (2.6)
13 (5.1) 67 (26.2) 92 (35.9) 74 (28.9) 10 (3.9)
22 (12.2) 119 (66.1) 39 (21.7)
6 (7.9) 52 (68.4) 18 (23.7)
28 (10.9) 171 (66.8) 57 (22.3)
85 (47.2) 29 (16.1) 18 (10.0) 48 (26.7) 10.5, 10 (6–14) 4.9, 4 (1–14) 44.5, 36.4 (8.3–100.0)
12 (15.8) 25 (32.9) 25 (32.9) 14 (18.4) 6.2, 8 (6–12) 4.2, 3 (1–10) 38.2, 25.0 (8.3–100.0)
97 (37.9) 54 (21.1) 43 (16.8) 62 (24.2) 9.2, 9 (6–14) 4.7, 4 (1–14) 42.4, 33.3 (8.3–100.0)
13.5, 13 (6–38)
11.2, 11 (5–28)
12.8, 12 (5–38)
75 (41.7) 34 (18.9) 21 (11.7) 50 (27.8)
9 (11.8) 20 (26.3) 11 (14.5) 36 (47.4)
84 (32.8) 54 (21.1) 32 (12.5) 86 (33.6)
92 (51.1) 88 (48.9) 63 (35.0) 48 (26.7) 37 (20.6)
36 (47.4) 32 (42.1) 23 (30.3) 16 (21.0) 19 (25.0)
128 (50.0) 120 (46.9) 86 (33.6) 64 (25.0) 56 (21.9)
Patient Characteristics Patient age varied from 46 to 78 years (mean 64.8). Most patients and tumor parameters showed values indicating an aggressive tumor potential, placing those patients at high risk for having LNI. Less than one third of our patients, 80 (31.2%), had a preoperative serum PSA value 7). The average percentage of positive cores at biopsy in our series was 42.4% – a fact that also demonstrates the aggressive potential of our tumors. There was a slight upgrading after prostatectomy in our series – the low-grade tumors (Gleason score 5–6) had decreased slightly in 84 cases (32.8%) while the high-grade tumors (Gleason score >7) increased in 86 cases (33.6%). The pathological review of the prostatectomy specimens showed that only half of the patients in our series had organ-confined disease. The percentage of EPE, SVI and LNI was 46.9, 33.6 and 25.0%, respectively. Positive surgical margins were detected in 56 patients (21.9%) of the entire cohort. This percentage varied between the two Hinev/Anakievski/Kolev/Hadjiev
ROC Curve 1.0
Nomogram Evaluation All of Briganti’s nomograms showed higher predictive accuracy as compared with the updated MSKCC nomogram. The respective AUC values for LNI were calculated as 0.847, 0.837, 0.858 and 0.875 for the four Briganti nomograms (versions 2006, 2007 and 2012), and 0.770 for the updated (version 2011) MSKCC nomogram, respectively (fig. 1).
0.4 BR_1_2006 BR_2_2007 BR_3_2007 BR_4_2012 MSKCC_2011 Reference line
Nomograms are graphical statistical tools that use different variables to calculate the potential risk. They provide estimates based on the specific characteristics of the individual patient. The median values of these estimates, however, were derived from hundreds or even thousands of patients with the same disease. The ideal nomogram should be accurate, not influenced by cultural and socio-economic differences, not be dependent on the pathologist who made the pathomorphological assessment, or by the surgeon who had carried out the operation, and should generally be applicable to all patients, and not only to those operated in elite institutions using exclusively strict selection criteria. The first prognostic tool in PCa appeared in 1993 when Partin et al.  published predictive tables for assessing the pathological stage, using a simple combination of clinical parameters – the preoperative serum PSA, the biopsy Gleason score, and the clinical stage. The Partin tables found wide clinical application and were not only validated in the USA [2, 8, 10] but in other European and Asian countries as well [13, 18]. Our team was the first in Bulgaria to successfully validate and assess the clinical significance of Partin tables . Despite the popularity gained throughout the years, the Partin tables, and the other predictive tools which soon followed them, are not devoid of limitations. Most of these predictive tools are based on a historical dataset of patients treated in elite institutions under very strict inclusion criteria. Therefore, the majority of these patients fell into the low- and intermediate-risk groups while only a small number of them, whose tumors showed aggressive biological potential, had been primarily treated by RP. On the contrary, the majority of our patients belongs to the high-risk group and significantly differs from average American patients who formed the database for the develPCa Patients Undergoing Extended Pelvic Lymph Node Dissection
Color version available online
institutions – it was lower in Varna (20.6%) and higher in Pleven (25.0%). Patient characteristics and tumor parameters are presented in table 1.
0.4 0.6 Specificity
Fig. 1. ROC curve analysis comparing predicted probabilities of
LNI by Briganti’s nomogram, version 2006 (basic parameters (BP) + total LNs, BR_1_2006) [AUCBR1 = 0.847 (95% CI 0.792– 0.901), p < 0.001]; Briganti’s nomogram, version 2007 (BP + number of positive cores, BR_2_2007) [AUCBR2 = 0.837 (95% CI 0.781–0.892), p < 0.001]; Briganti’s nomogram, version 2007 (BP + percentage of positive cores, BR_3_2007) [AUCBR3 = 0.858 (95% CI 0.807–0.909), p < 0.001]; updated Briganti’s nomogram, version 2012 (BP + percentage of positive cores, BR_4_2012) [AUCBR4 = 0.875 (95% CI 0.829–0.921), p < 0.001], and the updated MSKCC nomogram, version 2011 (MSKCC_2011) [AUCMSKCC = 0.770 (95% CI 0.709–0.831), p < 0.001].
opment of the American predictive tools and nomograms. A similar risk distribution might also be seen among many contemporary series of patients, both in USA and in Europe, due to the increasing popularity of the current tendency of treating low-risk patients by active surveillance and reserving radical surgery for the intermediate- and high-risk patient groups. Besides, most of the previous nomograms had been based on the results of limited PLND, which did not include the nodes around the internal iliac artery, recently discriminated as the first echelon of nodes staying on the primary route of lymphatic drainage from the prostate. Using a radioisotope agent and a gamma probe, we recently proved that 41.2% of all metastatic LNs in clinically localized PCa might be found in this region – therefore, they should not be missed during a routine PLND . The EAU guidelines currently recommend that when PLND is indicated, an extended dissection template Urol Int 2014;92:300–305 DOI: 10.1159/000354323
should be offered. Limited LND should no longer be performed because it misses at least half the nodes involved . As most of the predictive tools are based on limited PLND, firstly they probably underestimate the real percentage of the regional LNs involved, and secondly the probabilities of LNI predicted by them will not be applicable to the new candidates for RP who would be subjected to an ePLND as well. The only current nomograms based on the results of ePLND are those developed in Italy by Briganti and his team [4, 6, 7] and the updated MSKCC nomogram, recently published by Godoy et al. . It is important to note that all Bulgarian patients who were enrolled in the present study had undergone ePLND. Due to this reason, our database set was extremely suitable to serve as a model to test the predictive accuracy of the above-mentioned contemporary nomograms based on ePLND. Also, using an extended template for LN dissection, Heidenreich et al.  recently validated the nomogram developed in 2007 by Briganti et al. . These authors confirmed that the percentage of positive cores involved with PCa was the most reliable predictor of LNM, indicating the need for ePLND. The newest version of Briganti’s nomogram (version 2012)  was externally validated for the first time by Gacci et al. . The nomogram had been applied to a large multicentric population of men who underwent ePLND at the time of RP at two tertiary referral centers, both located in Italy. The updated Briganti nomogram demonstrated excellent accuracy and calibration characteristics and a general applicability for predicting the presence of LNI. In this study, the predictive accuracy of the full model reached 79% . The current study is the first validation study outside Italy of the updated version (2012) of Briganti’s nomogram  included in the EAU guidelines (version 2013). It convincingly confirmed that the five nomograms studied are able to predict LNI with good discriminative ability, as the AUC of each one of the models exceeded 0.70. Moreover, the present study proved the superiority of the newest version of Briganti’s nomograms (version 2012)  (fig. 1). The predictive accuracy of this nomogram was 87.5%, thus well exceeding the predictive accuracy of any other tool currently used to predict LNI in PCa. Our study, however, has a few limitations that might lead to biases when interpreting the results and have to be considered too. Firstly, it comprised a limited number of cases (256) with an uneven distribution among the two institutions involved (Varna 180 and Pleven 76 patients enrolled). Secondly, some apparent differences were observed among the tumor and patient characteristics be304
Urol Int 2014;92:300–305 DOI: 10.1159/000354323
tween the two centers. Besides, there were significant differences between the tumor and patient characteristics in our study and those in comparative Italian and American studies [4, 6, 7, 9]. Generally the majority of our patients might be classified as a high-risk patient subset, showing more aggressive features of their tumors – higher values of initial serum PSA, higher clinical stage, higher Gleason score, etc., as compared to the other studies, and the MSKCC patient cohort in particular . Further limitations of the study were the 17-year recruitment period with all the respective changes that followed. For example, the biopsy protocols significantly changed throughout the years: the sextant biopsy, initially accepted as a standard, is no longer considered adequate; therefore, it was recently replaced by more extended biopsy schemes, involving 8–12 cores, in line with the current guidelines . Finally, our study was also limited by the lack of central pathological review, implying that there might be considerable variations in the interpretation of the pathological results between the two centers. Nevertheless, the present validation study confirmed that all these nomograms are adequate in predicting LNI and might be used in Bulgarian PCa patients as well. The differences observed in some of the clinical and pathological parameters reflect the national differences with regard to patients’ attitudes to screening, the availability and the routine use of serum PSA testing and the current strategy of treating localized PCa in Bulgaria.
Despite the potential for heterogeneity in patient selection and management, all predictions demonstrated high concordance with actual observations. Compared with other similar prognostic tools, also based on ePLND, the updated Briganti nomogram (version 2012)  showed the highest predictive accuracy and should therefore be preferred.
1 Partin AW, Yoo J, Carter HB, et al: The use of prostate specific antigen, clinical stage and Gleason score to predict pathological stage in men with localized prostate cancer. J Urol 1993;150:110–114. 2 Eifler JB, Feng Z, Lin BM, et al: An updated prostate cancer staging nomogram (Partin tables) based on cases from 2006 to 2011. BJU Int 2013;111:22–29. 3 Katz MS, Efstathiou JA, D’Amico AV, et al: The ‘CaP calculator’: an online decision support tool for clinically localized prostate cancer. BJU Int 2010;105:1417–1422.
4 Briganti A, Larcher A, Abdollah F, et al: Updated nomogram predicting lymph node invasion in patients with prostate cancer undergoing extended pelvic lymph node dissection: the essential importance of percentage of positive cores. Eur Urol 2012;61:480–487. 5 Cagiannos I, Karakiewicz P, Eastham JA, et al: A preoperative nomogram identifying decreased risk of positive pelvic lymph nodes in patients with prostate cancer. J Urol 2003;170: 1798–1803. 6 Briganti A, Chun FK, Salonia A, et al: Validation of a nomogram predicting the probability of lymph node invasion based on the extent of pelvic lymphadenectomy in patients with clinically localized prostate cancer. BJU Int 2006;98:788–793. 7 Briganti A, Karakiewicz PI, Chun FK, et al: Percentage of positive biopsy cores can improve the ability to predict lymph node invasion in patients undergoing radical prostatectomy and extended pelvic lymph node dissection. Eur Urol 2007;51:1573–1581. 8 Makarov DV, Trock BJ, Humphreys EB, et al: Updated nomogram to predict pathologic stage of prostate cancer given prostate-specific antigen level, clinical stage, and biopsy Gleason score (Partin tables) based on cases from 2000 to 2005. Urology 2007;69:1095–1101.
PCa Patients Undergoing Extended Pelvic Lymph Node Dissection
9 Godoy G, Chong KT, Cronin A, et al: Extent of pelvic lymph node dissection and the impact of standard template dissection on nomogram prediction of lymph node involvement. Eur Urol 2011;60:195–201. 10 Huang Y, Isharwal S, Haese A, et al: Prediction of patient-specific risk and percentile cohort risk of pathological stage outcome using continuous prostate-specific antigen measurement, clinical stage and biopsy Gleason score. BJU Int 2011;107:1562–1569. 11 Heidenreich A, Bastian PJ, Bellmunt J, et al: Guidelines on Prostate Cancer. EAU 2013. Available at: http://www.uroweb.org/gls/ pdf/09_Prostate_Cancer_LR.pdf. 12 Shariat SF, Karakiewicz PI, Roehrborn CG, Kattan MW: An updated catalog of prostate cancer predictive tools. Cancer 2008; 113: 3075–3099. 13 Hinev AI, Anakievski D, Kolev N, Marianovski V, Hadjiev V: Validation of pre- and postoperative nomograms used to predict the pathological stage and prostate cancer recurrence after radical prostatectomy: a multi-institutional study. J BUON 2011;16:316–322. 14 Hinev AI, Hadjiev V, Kolev N: Validation of preoperative nomograms predicting lymph node involvement in prostate cancer: a bi-institutional study. Eur Urol 2011; 60: 1310– 1311.
Urol Int 2014;92:300–305 DOI: 10.1159/000354323
15 Hinev AI, Klissarova AD, Ghenev PI, et al: Radioisotopic detection of sentinel lymph nodes in clinically localized high-risk prostate cancer. J BUON 2009;14:661–667. 16 Gleason DF: Histologic grading and clinical stage of prostatic carcinoma; in Tannenbaum M (ed): Urologic Pathology: The Prostate. Philadelphia, Lea & Ferbirger, 1997, pp 171. 17 Schroeder FH, Hermanek P, Denis L, et al: The TNM classification of prostate cancer. Prostate Suppl 1992;4:129–138. 18 Xiao WJ, Ye DW, Yao XD, Zhang SL, Dai B: Comparison of three versions of Partin tables to predict final pathologic stage in a Chinese cohort: a decision curve analysis. Urol Int 2013;91:69–74. 19 Heidenreich A, Pfister D, Thüer D, Brehmer B: Percentage of positive biopsies predicts lymph node involvement in men with lowrisk prostate cancer undergoing radical prostatectomy and extended pelvic lymphadenectomy. BJU Int 2011;107:220–225. 20 Gacci M, Schiavina R, Lanciotti M, et al: External validation of the updated nomogram predicting lymph node invasion in patients with prostate cancer undergoing extended pelvic lymph node dissection. Urol Int 2013; 90:277–282.
Copyright: S. Karger AG, Basel 2014. Reproduced with the permission of S. Karger AG, Basel. Further reproduction or distribution (electronic or otherwise) is prohibited without permission from the copyright holder.