Extended pelvic lymph node dissection in patients with prostate cancer previously treated with surgery for lower urinary tract symptoms Nicola Fossati*†, Daniel D. Sjoberg†, Umberto Capitanio*, Giorgio Gandaglia*, Alessandro Larcher*, Alessandro Nini*, Vincenzo Mirone‡, Andrew J Vickers†, Francesco Montorsi* and Alberto Briganti* *Division of Oncology/Unit of Urology, URI, IRCCS Ospedale San Raffaele, Milan, ‡Unit of Urology, Federico II University, Naples, Italy, and †Department of Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, NY, USA
Objectives To evaluate the effect of previous prostate surgery performed for lower urinary tract symptoms (LUTS) on the ability to predict lymph node invasion (LNI) in patients subsequently diagnosed with prostate cancer, testing two widely used LNI predictive models.
Patients and Methods From 1990 to 2012, we collected data on 4734 patients with prostate cancer treated with radical prostatectomy and extended pelvic LN dissection (ePLND). Of these, 4453 (94%) had no prior prostate surgery (‘naïve patients’), while 286 (6%) had previously undergone surgery for LUTS. Two LNI prediction models based on patients treated with ePLND were evaluated using the receiver operating characteristic-derived area under the curve (AUC), the calibration plot method, and decision-curve analyses.
Results The rate of LNI was 12%, while the median number of LNs removed was 15 in both groups (P = 0.9). The two tested
Introduction Lymph node invasion (LNI) represents a poor prognostic factor in patients with prostate cancer, affecting their cancer control outcomes [1,2]. To date, pelvic LN dissection (PLND) represents the most accurate staging procedure to assess the presence of LNI in patients with prostate cancer treated with radical prostatectomy (RP) [3,4] and may also be associated with a more favourable cancer control outcome [5]. Moreover, a general agreement has been reached on the statement that whenever a PLND is indicated, it should be anatomically extended (ePLND) [3,4,6]. On the other hand, ePLND is associated with some disadvantages, e.g. morbidity, longer
BJU Int 2015; 116: 366–372 wileyonlinelibrary.com
nomograms provided more accurate prediction in naïve patients than for those previously treated with prostate surgery for LUTS (AUC: 82% and 81% vs 68% and 71%, P = 0.01 and P = 0.04, respectively). In naïve patients the surgeon would have missed one LNI for every 53 and 34 avoided ePLND using the Briganti and Godoy nomograms, respectively; in patients previously treated with surgery for LUTS, a LNI would have been missed in 13 and 21 patients not undergoing ePLND.
Conclusion The accuracy and the clinical net-benefit of LNI prediction tools decrease significantly in patients with prior prostate surgery for LUTS. These models should be avoided in such patients, who should undergo routine ePLND.
Keywords prostate cancer, extended pelvic lymph node dissection, surgery for LUTS, lymph node involvement, predictive models
hospital stay and higher hospital costs [7,8]. For these reasons, several predictive models assessing the preoperative risk of LNI have been developed, although only three have been based on ePLND series [9–11]. The currently available European Association of Urology (EAU) Prostate Cancer Guidelines [4] assert that ePLND should be performed if the estimated risk for LNI, determined by the Briganti nomogram [11] exceeds 5%. In all other cases ePLND can be omitted, which means accepting a low risk of missing positive LNs. Similarly, the National Comprehensive Cancer Network (NCCN) Prostate Cancer Guidelines [12] state that PLND can be excluded in patients with 2 cm) were sampled in multiple blocks. If no LNs were macroscopically detected, all fat tissue was processed. Clearing solution was generally not used for PLNs. All blocks were embedded in paraffin, cut at 3 mm, and stained with haematoxylin and eosin. Our Institutional Review Board approved this study, and all patients signed an informed consent in order to be included in our prospectively collected database. Variables Definitions We included age at surgery, preoperative PSA level, clinical stage, biopsy Gleason score as clinical data and pathological stage, pathological Gleason score, LNI, and number of LNs removed as pathological data. The preoperative PSA level (AxSYM PSA assay; Abbott Laboratories, Abbott Park, IL, USA) was measured before DRE and TRUS. Clinical stage was assigned by the attending urologist according to the 2009 American Joint Committee on Cancer staging system. LNI was considered if one or several PLNs were involved with cancer. Outcome Definition The outcome of this study was to evaluate the effect of previous prostate surgery for LUTS on the ability to predict LNI, testing two widely used LNI predictive models [9,10]. We tested the original Briganti nomogram [9] and not the updated version [11], due to lack of percentage of positive biopsy cores data in a substantial proportion of the patient population (2923 patients, 61%). Statistics Our statistical analyses consisted of three steps. First, we validated the two predictive models [9,10] in our patient population, using the predetermined regression coefficients. As the Briganti nomogram was originally developed using patients from our cohort treated before 2006, we decided to validate that predictive model on the contemporary portion of our population (year of surgery ≥2006). The discrimination accuracy of the models to predict LNI was quantified using the receiver operating characteristic-derived area under the curve (AUC). Differences among AUCs were compared using the DeLong method [23]. The extent of overestimation or underestimation of the histologically confirmed vs the nomogram predicted LNI rates was graphically explored using a calibration plot. © 2014 The Authors BJU International © 2014 BJU International
367
Fossati et al.
Table 1 The descriptive characteristics of the 4734 patients with prostate cancer, treated with RP and ePLND between 1990 and 2012, stratified into naïve patients and patients who had previously undergone prostate surgery for LUTS. Variable
Naïve patients
No. of patients (%) Median (IQR) age, years Median (IQR) preoperative PSA level, ng/mL Clinical stage, n (%) T1 T2 T3 Biopsy Gleason score, n (%) ≤6 7 ≥8 Pathological stage, n (%) pT2 pT3a pT3b-pT4 Pathological Gleason score, n (%) ≤6 7 ≥8 LNI, n (%) no yes Median (IQR) no. of LNs removed
4453 (94%) 65 (60, 70) 7.0 (5.1,10.8)
281 (6%) 70 (66, 73) 5.1 (3.3,7.8)
2404 (54) 1523 (34) 526 (12)
168 (60) 89 (31) 24 (9)
2688 (60) 1371 (31) 394 (9)
165 (59) 88 (31) 28 (10)
2997 (67) 767 (17) 689 (16)
187 (67) 45 (16) 49 (17)
1610 (36) 2242 (50) 601 (14)
95 (34) 142 (50) 44 (16)
3922 (88) 531 (12) 15 (10, 21)
248 (88) 33 (12) 15 (10, 21)
Objectives To evaluate the effect of previous prostate surgery performed for lower urinary tract symptoms (LUTS) on the ability to predict lymph node invasion (LNI) in patients subsequently diagnosed with prostate cancer, testing two widely used LNI predictive models.
Patients and Methods From 1990 to 2012, we collected data on 4734 patients with prostate cancer treated with radical prostatectomy and extended pelvic LN dissection (ePLND). Of these, 4453 (94%) had no prior prostate surgery (‘naïve patients’), while 286 (6%) had previously undergone surgery for LUTS. Two LNI prediction models based on patients treated with ePLND were evaluated using the receiver operating characteristic-derived area under the curve (AUC), the calibration plot method, and decision-curve analyses.
Results The rate of LNI was 12%, while the median number of LNs removed was 15 in both groups (P = 0.9). The two tested
Introduction Lymph node invasion (LNI) represents a poor prognostic factor in patients with prostate cancer, affecting their cancer control outcomes [1,2]. To date, pelvic LN dissection (PLND) represents the most accurate staging procedure to assess the presence of LNI in patients with prostate cancer treated with radical prostatectomy (RP) [3,4] and may also be associated with a more favourable cancer control outcome [5]. Moreover, a general agreement has been reached on the statement that whenever a PLND is indicated, it should be anatomically extended (ePLND) [3,4,6]. On the other hand, ePLND is associated with some disadvantages, e.g. morbidity, longer
BJU Int 2015; 116: 366–372 wileyonlinelibrary.com
nomograms provided more accurate prediction in naïve patients than for those previously treated with prostate surgery for LUTS (AUC: 82% and 81% vs 68% and 71%, P = 0.01 and P = 0.04, respectively). In naïve patients the surgeon would have missed one LNI for every 53 and 34 avoided ePLND using the Briganti and Godoy nomograms, respectively; in patients previously treated with surgery for LUTS, a LNI would have been missed in 13 and 21 patients not undergoing ePLND.
Conclusion The accuracy and the clinical net-benefit of LNI prediction tools decrease significantly in patients with prior prostate surgery for LUTS. These models should be avoided in such patients, who should undergo routine ePLND.
Keywords prostate cancer, extended pelvic lymph node dissection, surgery for LUTS, lymph node involvement, predictive models
hospital stay and higher hospital costs [7,8]. For these reasons, several predictive models assessing the preoperative risk of LNI have been developed, although only three have been based on ePLND series [9–11]. The currently available European Association of Urology (EAU) Prostate Cancer Guidelines [4] assert that ePLND should be performed if the estimated risk for LNI, determined by the Briganti nomogram [11] exceeds 5%. In all other cases ePLND can be omitted, which means accepting a low risk of missing positive LNs. Similarly, the National Comprehensive Cancer Network (NCCN) Prostate Cancer Guidelines [12] state that PLND can be excluded in patients with 2 cm) were sampled in multiple blocks. If no LNs were macroscopically detected, all fat tissue was processed. Clearing solution was generally not used for PLNs. All blocks were embedded in paraffin, cut at 3 mm, and stained with haematoxylin and eosin. Our Institutional Review Board approved this study, and all patients signed an informed consent in order to be included in our prospectively collected database. Variables Definitions We included age at surgery, preoperative PSA level, clinical stage, biopsy Gleason score as clinical data and pathological stage, pathological Gleason score, LNI, and number of LNs removed as pathological data. The preoperative PSA level (AxSYM PSA assay; Abbott Laboratories, Abbott Park, IL, USA) was measured before DRE and TRUS. Clinical stage was assigned by the attending urologist according to the 2009 American Joint Committee on Cancer staging system. LNI was considered if one or several PLNs were involved with cancer. Outcome Definition The outcome of this study was to evaluate the effect of previous prostate surgery for LUTS on the ability to predict LNI, testing two widely used LNI predictive models [9,10]. We tested the original Briganti nomogram [9] and not the updated version [11], due to lack of percentage of positive biopsy cores data in a substantial proportion of the patient population (2923 patients, 61%). Statistics Our statistical analyses consisted of three steps. First, we validated the two predictive models [9,10] in our patient population, using the predetermined regression coefficients. As the Briganti nomogram was originally developed using patients from our cohort treated before 2006, we decided to validate that predictive model on the contemporary portion of our population (year of surgery ≥2006). The discrimination accuracy of the models to predict LNI was quantified using the receiver operating characteristic-derived area under the curve (AUC). Differences among AUCs were compared using the DeLong method [23]. The extent of overestimation or underestimation of the histologically confirmed vs the nomogram predicted LNI rates was graphically explored using a calibration plot. © 2014 The Authors BJU International © 2014 BJU International
367
Fossati et al.
Table 1 The descriptive characteristics of the 4734 patients with prostate cancer, treated with RP and ePLND between 1990 and 2012, stratified into naïve patients and patients who had previously undergone prostate surgery for LUTS. Variable
Naïve patients
No. of patients (%) Median (IQR) age, years Median (IQR) preoperative PSA level, ng/mL Clinical stage, n (%) T1 T2 T3 Biopsy Gleason score, n (%) ≤6 7 ≥8 Pathological stage, n (%) pT2 pT3a pT3b-pT4 Pathological Gleason score, n (%) ≤6 7 ≥8 LNI, n (%) no yes Median (IQR) no. of LNs removed
4453 (94%) 65 (60, 70) 7.0 (5.1,10.8)
281 (6%) 70 (66, 73) 5.1 (3.3,7.8)
2404 (54) 1523 (34) 526 (12)
168 (60) 89 (31) 24 (9)
2688 (60) 1371 (31) 394 (9)
165 (59) 88 (31) 28 (10)
2997 (67) 767 (17) 689 (16)
187 (67) 45 (16) 49 (17)
1610 (36) 2242 (50) 601 (14)
95 (34) 142 (50) 44 (16)
3922 (88) 531 (12) 15 (10, 21)
248 (88) 33 (12) 15 (10, 21)
