When Should a Positive Surgical Margin Ring a Bell? An Analysis of a Multi-Institutional Robot-Assisted Laparoscopic Radical Prostatectomy Database

Abstract

Objective:

The impact of positive surgical margin (SM) on cancer control outcomes in prostate cancer patients is a subject of continuous debate. We test the hypothesis that the impact of SM on clinical recurrence (CR) rate may vary based on the other clinical/pathologic characteristics of the tumor.

Materials and Methods:

We focused on 5290 patients treated with robot-assisted radical prostatectomy and pelvic node dissection, between 2002 and 2013, at three tertiary care centers. Regression tree analysis stratified patients into risk groups based on their tumor characteristics and the corresponding CR rate. Kaplan–Meier log-rank and multivariable Cox regression models tested the relationship between SM status and CR rate in each tree-generated risk group.

Results:

Mean (median) follow-up time was 47.7 (39.0) months. Regression tree analysis that considered all available covariates, except SM status, divided patients based on their CR risk into the following risk groups: (1) high risk (any pT3b/pT4 disease); (2) intermediate risk (≤pT3a disease and pGS 8–10); (3) low risk (≤pT3a, pGS ≤7, and prostate-specific antigen [PSA] >9 ng/mL); and (4) very low risk (≤pT3a, pGS ≤7, and PSA ≤9 ng/mL). Positive SM had a significant detrimental impact on CR risk only in two groups: intermediate risk (p < 0.001) and high risk (p = 0.01). These observations were confirmed by multivariable analyses.

Conclusions:

Our findings show that positive SM had a detrimental impact on CR only in a minority of patients (15%), specifically in those with advanced pathologic stage and/or pathologically poorly differentiated tumor. For all the remaining patients (85%), positive SM by itself did not increase the risk of CR.

Introduction

The presence of positive surgical margin (SM) on radical prostatectomy (RP) specimen is considered as an adverse pathologic event.^1,2 Despite stage migration in prostate cancer (PCa), between 11% and 48% of contemporary patients still harbor positive SM at surgery.^3

–6 The impact of the latter on cancer control outcomes is highly controversial.^7
–9

A recent multi-institutional study showed that positive SM has no detrimental impact on cancer-specific mortality (CSM) rate.¹⁰ However, many previous reports have shown the contrary.^7,11 Several reasons could contribute to this discrepancy. For example, some authors suggested that the lack of a relationship in the multivariate model between SM and survival is probably due to the strong relationship between positive SM and other adverse pathologic features such as advanced pathologic stage,^1,2,7 large tumor volume,² and advanced pathologic Gleason score.^1,2,7 These variables are stronger predictors of adverse survival outcomes than positive SM and might mask the impact of positive SM in multivariate models.² Likewise, there is a profound heterogeneity of positive SM population in terms of location,³ number,^3,9 and pathologic Gleason score,^12,13 suggesting that substratifications are required in this population.

With this in mind, it might be argued that the potential detrimental impact of positive SM on cancer control outcomes is not necessarily an all or none phenomenon. Indeed, it is possible that this relationship (if any) is largely influenced by other tumor characteristics. To test this hypothesis, we set to evaluate the impact of positive SM on the clinical recurrence (CR) rate, after stratifying patients into risk groups based on their survival probability and tumor characteristics. Moreover, most previous reports based their findings mainly on patients treated with open surgery, which has been largely replaced with robot-assisted laparoscopic radical prostatectomy (RALP) in contemporary patients. Our study circumvents this limitation by focusing exclusively on RALP patients originating from a contemporary multi-institutional cohort.

Materials and Methods

A total of 5290 PCa patients treated with RALP and pelvic lymph node dissection (PLND), between 2002 and 2013, at three tertiary care centers (Vattikuti Urology Institute, San Raffaele Hospital, and Martini-Clinic) were included in the study. Patients with missing clinical and/or pathologic data were excluded. It is noteworthy that the first structured program of urologic robotic surgery was started in 2000 by Menon et al.¹⁴ in Detroit at the Vattikuti Urology Institute. San-Raffaele Hospital program was started in 2006, and Martini-Clinic performed its first RALP in 2005. When necessary, patients were staged preoperatively with pelvic/abdominal CT or abdominal ultrasound, bone scan, and chest X-ray. Several surgeons performed RALP by using a standard technique with PLND.^15,16,14 Additional therapy was defined as a receipt of either adjuvant radiation or hormonal treatment (within 6 months after surgery and in the absence of biochemical recurrence) or either salvage radiation or hormonal treatment (>6 months following surgery and/or in the presence of biochemical recurrence) and was decided by the treatment physician based on patient preferences and tumor characteristics. Biochemical recurrence (BCR) was defined as two consecutive prostate-specific antigen (PSA) values ≥0.2 ng/mL after RP. Follow-up consisted of quarterly examinations with a PSA for the first 2 years and then declining in frequency to semiannually and eventually annually in the absence of recurrence. Imaging was clinically driven based on symptoms and suspicion for recurrent disease. The institutional review board of each center approved the study. Given the initial agreement between the participating institutions, center identifiers were removed from the database.

Variables and endpoint definitions

Our variables consisted of age at surgery, years of surgery, body mass index (kg/m²), preoperative PSA value (ng/mL), Gleason score on biopsy (2–6 vs 7 vs 8–10), clinical tumor stage (cT1 vs cT2 vs ≥cT3), pathologic Gleason score (2–6 vs 7 vs 8–10), pathologic tumor stage (pT2 vs pT3a vs ≥T3b), SM status (negative vs positive), lymph node invasion (LNI) status (pN0 vs pN1), number of positive nodes, number of nodes removed (RLNs), adjuvant radiotherapy (aRT) status, adjuvant hormonal therapy (aHT) status, and salvage therapy administration. The primary endpoint was to analyze the impact of SM status on CR rate after RP. CR was defined as radiographic evidence for PCa with or without the presence of symptoms. Radiographic studies included bone scintigraphy, PET scans, CT scans, and MRIs. Biopsy confirmation of PCa recurrence was only pursued in the presence of significant diagnostic uncertainty. We opt to address CR, rather than BCR, as an endpoint, because BCR does not necessarily represent a good proxy for CSM rate.¹⁷

Statistical analyses

Descriptive statistics of categorical variables focused on frequencies and proportions. Means, medians, and interquartile ranges were reported for continuously coded variables. Chi-square and Mann–Whitney tests were used to compare the statistical significance of differences in proportions and medians, respectively.

Our statistical analyses consisted of several steps. First, we used a regression tree approach for censored data to predict the CR rate. The latter was defined as either local recurrence or distant metastasis captured by histology and/or imaging. Regression tree approach uses a standard and a recursive algorithm to sequentially divide a group of patients into two subgroups, in which the separation between the two class-specific survival curves in a pair is maximized. The algorithm selects the optimal sequence of classifications, as defined by a hierarchy of prognostic factors and associated cut points. All available predictors, except for positive SM, were included in the regression tree. The results of these analyses were graphically represented. Second, we used Kaplan–Meier curves to estimate CR-free rates in the entire cohort and in each regression tree-generated risk group, after stratifying patients according to SM status. Likewise, Cox regression analyses tested the relationship between SM status and CR rate in the entire cohort and in each regression tree-generated risk group. Multivariable regression models were adjusted for all available confounders. Finally, a multivariable model was built considering BCR as the outcome.

Statistical analyses were performed using the R Statistical Package (R Foundation for Statistical Computing, Vienna, Austria) and SPSS v. 20.0 (IBM Corp., Armonk, NY), considering a statistical significance at p < 0.05.

Results

Baseline characteristics

Clinical and pathologic characteristics of our cohort, stratified by margin status, are reported in Table 1. Overall, 1327 (25.1%) patients harbored positive SM at the final pathology report. Patients with positive SM were older (median: 62 vs 61, p < 0.001), had a higher preoperative PSA value (median: 9.0 vs 6.4, p < 0.001), higher pathologic stage (≥pT3b: 18.6% vs 5.1%, p < 0.001), higher pathologic grade (Gleason sum 8–10: 21.3% vs 7.7%, p < 0.001), higher LNI rate (pN1: 8.5% vs 2.8%, p < 0.001), and higher rates of adjuvant (aRT: 6.4% vs 1.0% and aHT: 1.3% vs 0.3%; p < 0.001) and salvage therapies (any salvage therapy: 24.0% vs 6.7%; p < 0.001) compared to those without positive SM.

Table 1.

Descriptive Statistics of 5290 Patients with Prostate Cancer Treated with Robot-Assisted Laparoscopic Radical Prostatectomy and Pelvic Lymph Node Dissection, Between 2002 and 2013, at Three Tertiary Care Centers

Variables	Overall cohort, n = 5290 (100%)	Patients with PSM, n = 1327 (25.1%)	Patients without PSM, n = 3963 (74.9%)	p-Value
Age at surgery, years
Mean	61	61	60	<0.001
Median (IQR)	61 (56–66)	62 (56–67)	61 (55–66)
Years of surgery
2001–2005	1269 (23.9)	350 (26.4)	915 (23.1)
2006–2009	2292 (43.3)	572 (43.1)	1720 (43.4)	0.03
2010–2013	1733 (32.8)	405 (30.5)	1328 (33.5)
BMI, ng/mL
Mean	27.6	28.0	27.4	<0.001
Median (IQR)	27 (25–30)	27 (25–30)	27 (25–29)
Preoperative PSA, ng/mL
Mean	7.1	9.0	6.4	<0.001
Median (IQR)	5.4 (4.2–7.7)	6.2 (4.7–9.6)	5.2 (4.1–7.2)
Clinical tumor stage
T1	3773 (71.3)	894 (67.4)	2879 (72.6)
T2	1402 (26.5)	384 (28.9)	1018 (25.7)	<0.001
≥T3	115 (2.2)	49 (3.7)	66 (1.7)
Gleason score on biopsy
2–6	2162 (40.9)	436 (32.9)	1726 (43.6)
7	2523 (47.7)	672 (50.6)	1851 (46.7)	<0.001
8–10	605 (11.4)	219 (16.5)	386 (9.7)
Pathologic tumor stage
pT2	3256 (61.6)	373 (28.1)	2883 (72.7)	<0.001
pT3a	1585 (30.0)	707 (53.3)	878 (22.2)
≥pT3b	449 (8.5)	247 (18.6)	202 (5.1)
Pathologic Gleason sum
2–6	1305 (24.7)	138 (10.4)	1167 (29.4)	<0.001
7	3397 (64.2)	906 (68.3)	2491 (62.9)
8–10	588 (11.1)	283 (21.3)	305 (7.7)
Lymph node status
pN0	5068 (95.8)	1214 (91.5)	3854 (97.2)	<0.001
pN1	222 (4.2)	113 (8.5)	109 (2.8)
Nodes removed, number
Mean	6.8	7.0	6.7	0.1
Median (IQR)	5 (3–9)	5 (3–9)	5 (3–9)
Positive nodes, number
Mean	0.1	0.2	0.04	<0.001
Median (IQR)	0 (0–0)	0 (0–0)	0 (0–0)
Adjuvant radiotherapy
No	5165 (97.6)	1242 (93.6)	3963 (99.0)	<0.001
Yes	125 (2.4)	85 (6.4)	40 (1.0)
Adjuvant hormonal therapy
No	5262 (99.5)	1310 (98.7)	3952 (99.7)	<0.001
Yes	28 (0.5)	17 (1.3)	11 (0.3)
Any salvage therapy
No	4707 (89.0)	1009 (76.0)	3698 (93.3)	<0.001
Yes	583 (11.0)	318 (24.0)	265 (6.7)

IQR = interquartile range; BMI = body mass index; PSA = prostate-specific antigen; PSM = positive surgical margin.

Regression tree

All available potential predictors of CR, except for SM status, were included in the regression tree analyses. The regression tree analysis identified three variables to stratify patients according to their CR risk and indicated the cutoffs that maximized the separation in class-specific survival. These variables consisted of preoperative PSA, pathologic Gleason score, and pathologic tumor stage. On the basis of these variables, the cohort was stratified into four risk groups: very low risk (≤pT3a disease, pathologic Gleason score ≤7, and PSA ≤9), low risk (≤pT3a disease, pathologic Gleason score ≤7, and PSA >9), intermediate risk (≤pT3a disease and pathologic Gleason score = 8–10), and high risk (pT3b–pT4 disease). The characteristics of each risk group, as well as the corresponding 5-year CR-free rates, are summarized in Figure 1.

FIG. 1.

A novel clinical recurrence risk-stratification tree based on the data of 5290 patients with prostate cancer treated with robot-assisted laparoscopic radical prostatectomy and pelvic lymph node dissection, between 2002 and 2013, at three tertiary care centers. CR = clinical recurrence; 95% CI = 95% confidence interval.

Survival estimates and Cox regression analyses

During a mean (median) follow-up time of 47.7 (39.0) months, 797 (15.1%), and 91 (1.7%) patients incur BCR and CR, respectively. We evaluated factors associated with BCR or CR after RP using multivariable Cox regression models, depicted in Table 2. Positive SM was a significant predictor of BCR (hazard ratio [HR]: 1.74) and CR (HR: 2.44), all p < 0.001. Other predictors were represented by PSA, number of positive nodes, pathologic stage, and pathologic Gleason score (all p < 0.05).

Table 2.

Multivariate Cox Regression Analyses Predicting Biochemical Recurrence and Clinical Recurrence in 5290 Patients with Prostate Cancer Treated with Robot-Assisted Laparoscopic Radical Prostatectomy

		Multivariate analysis
	Biochemical recurrence		Clinical recurrence
Variables	HR (95% CI)	p	HR (95% CI)	p
Age, years	1.00 (0.99–1.01)	0.5	1.00 (0.97–1.03)	0.9
PSA, ng/mL	1.01 (1.01–1.02)	<0.001	1.02 (1.00–1.04)	0.05
Positive SM	1.74 (1.47–2.05)	<0.001	2.44 (1.38–4.29)	0.001
Number of positive nodes	1.16 (1.06–1.28)	0.001	1.19 (1.02–1.40)	0.03
Pathologic stage
pT2	Ref	Ref	Ref	Ref
pT3a	1.88 (1.54–2.31)	<0.001	1.74 (0.74–4.09)	0.2
pT3b–T4	4.12 (3.23–5.25)	<0.001	5.24 (2.17–12.61)	<0.001
GS pathology
2–6	Ref	Ref	Ref	Ref
7	2.62 (1.87–3.68)	<0.001	1.94 (0.43–8.82)	0.4
8–10	7.35 (5.10–10.60)	<0.001	11.45 (2.46–53.35)	0.001
Adjuvant or salvage therapies	0.91 (0.61–1.34)	0.6	1.43 (0.59–3.48)	0.4

HR = hazard ratio; CI = confidence interval; GS = Gleason score.

In multivariable analyses predicting CR risk, patients with positive SM had a 2.3-fold higher CR risk compared to those with negative SM (p < 0.001). When patients were stratified according to our novel risk groups, patients with positive SM had less favorable CR rates in the intermediate-risk (HR: 5.84; p = 0.003) and high-risk (HR: 1.80; p = 0.04) groups only. In all other risk groups, SM status was not an independent predictor of CR rate (all p > 0.5) (Table 3).

Table 3.

Univariable and Multivariable Cox Regression Analyses Testing the Relationship Between Positive Surgical Margin and Clinical Recurrence-Free Survival Rate in 5290 Patients with Prostate Cancer Treated with Robot-Assisted Laparoscopic Radical Prostatectomy

	Univariate analyses		Multivariate analyses
	HR (CI 95%)	p-Value	HR (CI 95%)	p-Value
Overall population (n = 5290, 100%)
Positive SMs
No	Ref.	—	Ref.	—
Yes	6.01 (3.85–9.39)	<0.001	2.32 (1.43–3.76)	0.001
Very low-risk population (n = 3831, 72.4%)
Positive SMs
No	Ref.	—	Ref.	—
Yes	3.38 (0.98–11.71)	0.05	1.57 (0.39–6.37)	0.5
Low-risk population (n = 637, 12.0%)
Positive SMs
No	Ref.	—	Ref.	—
Yes	1.27 (0.36–4.53)	0.7	1.45 (0.34–6.20)	0.6
Intermediate-risk population (n = 373, 7.1%)
Positive SMs
No	Ref.	—	Ref.	—
Yes	7.51 (2.52–22.32)	<0.001	5.84 (1.83–18.63)	0.003
High-risk population (n = 449, 8.5%)
Positive SMs
No	Ref.	—	Ref.	—
Yes	2.25 (1.19–3.26)	0.01	1.80 (0.95–3.41)	0.04

Patients were stratified based on regression tree analysis. Multivariate models adjusted for age, PSA value, pathologic stage, pathologic Gleason, number of positive nodes, adjuvant therapies, and salvage therapies.

Discussion

Although previous retrospective studies report an increased risk of BCR^3,4,6,5,18 and CSM rate⁷ after RP for patients who harbor a positive SM, a recent multi-institutional series failed to find any survival differences in patients with positive SM compared to those with negative SM.¹⁰ This controversy might be attributed, at least partially, to the heterogeneity of patients with positive SM status. Specifically, other tumor characteristics might have an important impact on the relationship between SM status and cancer control outcomes. To test this hypothesis, we set to evaluate the impact of positive SM on CR rate, after stratifying patients into risk groups based on their survival probability and tumor characteristics. It is noteworthy that most previous reports addressing RALP cancer control outcomes focused exclusively on SM status and/or BCR rate.^19

–24 However, patients with BCR are a heterogeneous group, and the natural history of BCR is variable. Indeed, BCR does not always translate CR or CSM rate.¹⁷ To circumvent this limitation, we focused on the CR rate, which represents a more significant endpoint.

In our cohort of 5290 patients, all men were treated with RP and PLND. Moreover, only 2.4% and 0.5% received aRT and aHT, respectively. Overall, 25.1% patients harbored a positive SM. At 5 years, the CR-free rate was 98.0%. When patients were stratified according to SM status, those with positive SM had less favorable CR-free survival rates (99.2% vs 94.6%, p < 0.001). This holds true, even after adjusting to tumor characteristics at multivariable analyses (HR: 2.32; p = 0.001). Thus, unlike previous reports, our study shows a detrimental impact of positive SM on CR rate and also on BCR rate.^2,3,10,9

To further assess the potential interaction among tumor characteristics, SM status, and CR rate, we performed the following analyses. First, we stratified our patients into risk groups according to their CR risk by using a regression tree analysis. In this step, all available patient characteristics, except SM status, were included as potential predictors of CR. The regression tree analysis selected the independent predictors of CR, as well as their most appropriate cutoffs, to maximize the survival difference between the newly created risk groups. Second, we examined the relationship between positive SM and CR rate in each risk group separately. This allowed us to test the role of positive SM in each risk group. We found that positive SM has a detrimental impact on CR only in patients classified as intermediate-risk group (≤pT3a disease and pathologic Gleason score 8–10) or high-risk group (pT3b–pT4 disease) according to our novel classification. It is noteworthy that patients in the latter two categories accounted for only 15% of the overall population. These results were confirmed by multivariable analyses, where patients with positive SM had 5.8-fold (p < 0.003) and 1.8-fold (p < 0.04) higher CR risk compared to those with negative SM, when they were categorized as intermediate- and high-risk patients, respectively. Conversely, SM status was not an independent predictor of CR risk in any other risk group (all p ≥ 0.5).

Our findings have several clinical implications. First, although data regarding long-term cancer control outcomes of patients treated with RALP are scarce, our study addresses this aspect using a multi-institutional cohort, which gives our findings more generalizability. Second, our findings highlight the importance of considering other tumor characteristics, when evaluating the impact of SM status on patients' prognosis. Despite the current controversy in literature, most experts consider the presence of positive SM as an adverse pathologic event, regardless of other tumor characteristics. Interestingly, our results show that this is true only in a minority of patients. Specifically, only patients with pT3b or higher disease and/or those with pathologic Gleason score 8–10 will have less favorable outcomes, when harboring positive SM. In all other patients, the presence of positive SM should not raise an alarm or “ring a bell.” This is of utmost importance from a clinical perspective, when counseling patients about their prognosis. Likewise, it is of utmost importance from a scientific perspective, as it might explain, at least partially, the controversy of previous reports, which addressed the impact of SM status on cancer control outcomes.^{3,4,5,6,10,7,11} Virtually, all previous studies^{3,4,5,6,10,7,11} based their analyses on the traditional logistic regression analysis, which represents a valid methodology. However, the potential interaction between SM status and other tumor characteristics was rarely accounted for in these analyses. This excludes the possibility of examining the role of SM status in patients' subgroups. Our study circumvents this limitation by using regression tree analysis. The latter allowed us to examine the interaction among SM status, tumor characteristics, and cancer control outcomes. Moreover, it allowed us to present our findings in a simple, easy-to-remember, and clinical-friendly way. Finally, it is necessary to highlight the fact that our novel risk stratification is not meant to replace previous risk stratification (e.g., D'Amico risk classification or CAPRA score). Instead, our novel risk stratification was specifically designed to evaluate the impact of SM status on outcomes, after accounting for other tumor characteristics. In this context, our results can be of help in guiding adjuvant treatment decision. For example, patients in the intermediate- and high-risk groups might benefit from maximizing local control through aRT. However, this latter point warrants further investigation before applying in clinical practice.

Our study has some limitations. First, our results are derived from retrospective observational data. Thus, our findings must be interpreted within the possible limitations of such data. In context, preoperative local and distant staging of patients could be different on the basis of the center or physician preferences. However, the multi-institutional nature of our cohort gives it more generalizability and excludes the limitation of single institutional data. Moreover, we tried to account for potential biases by adjusting to potential confounders using multivariable analyses. Second, a central pathologic review was not performed in our cohort. Nonetheless, the pathologic specimens in our study were examined by an expert and a dedicated genitourinary pathologist in each institution, which guarantees high-quality data. Third, our follow-up might be considered relatively short (mean follow-up 48 months). However, the short follow-up is shared by all previous literature about this topic.^3,4,5,7 Finally, no data about the length of the SM or the extension of the cancer presences were recorded or analyzed.

Conclusions

Our findings show that positive SM has a detrimental impact on CR only in a minority of patients (15%), specifically in those with advanced pathologic stage and/or pathologically, poorly differentiated tumor. For all the remaining patients (85%), positive SM by itself did not increase the risk of CR. This should be considered when counseling patients about their prognosis and when deciding the necessity of adjuvant treatment.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

Abbreviations Used

References

Wieder

, Soloway

. Incidence, etiology, location, prevention and treatment of positive surgical margins after radical prostatectomy for prostate cancer. J Urol, 1998; 160:299–315.

Boorjian

, Tollefson

, Rangel

, Bergstralh

, Karnes

. Clinicopathological predictors of systemic progression and prostate cancer mortality in patients with a positive surgical margin at radical prostatectomy. Prostate Cancer Prostatic Dis, 2012; 15:56–62.

Stephenson

, Wood

, Kattan

, Klein

, Scardino

, Eastham

, et al. Location, extent and number of positive surgical margins do not improve accuracy of predicting prostate cancer recurrence after radical prostatectomy. J Urol, 2009; 182:1357–1363.

Eastham

, Kuroiwa

, Ohori

, Serio

, Gorbonos

, Maru

, et al. Prognostic significance of location of positive margins in radical prostatectomy specimens. Urology, 2007; 70:965–969.

Simon

, Kim

, Soloway

. Prostate specific antigen recurrence rates are low after radical retropubic prostatectomy and positive margins. J Urol, 2006; 175:140–144; discussion 144–145.

Swindle

, Eastham

, Ohori

, Kattan

, Wheeler

, Maru

, et al. Do margins matter? The prognostic significance of positive surgical margins in radical prostatectomy specimens. J Urol, 2005; 174:903–907.

Wright

, Dalkin

, True

, Ellis

, Stanford

, Lange

, et al. Positive surgical margins at radical prostatectomy predict prostate cancer specific mortality. J Urol, 2010; 183:2213–2218.

Yossepowitch

, Briganti

, Eastham

, Epstein

, Graefen

, Montironi

, et al. Positive surgical margins after radical prostatectomy: A systematic review and contemporary update. Eur Urol, 2014; 65:303–313.

Mauermann

, Fradet

, Lacombe

, Dujardin

, Tiguert

, Tetu

, et al. The impact of solitary and multiple positive surgical margins on hard clinical end points in 1712 adjuvant treatment-naive pT2-4 N0 radical prostatectomy patients. Eur Urol, 2013; 64:19–25.

10.

Stephenson

, Eggener

, Hernandez

, Klein

, Kattan

, Wood

, et al. Do margins matter? The influence of positive surgical margins on prostate cancer-specific mortality. Eur Urol, 2014; 65:675–680.

11.

Chalfin

, Dinizo

, Trock

, Feng

, Partin

, Walsh

, et al. Impact of surgical margin status on prostate-cancer-specific mortality. BJU Int, 2012; 110:1684–1689.

12.

Viers

, Sukov

, Gettman

, Rangel

, Bergstralh

, Frank

, et al. Primary Gleason grade 4 at the positive margin is associated with metastasis and death among patients with Gleason 7 prostate cancer undergoing radical prostatectomy. Eur Urol, 2014; 66:1116–1124.

13.

Udo

, Cronin

, Carlino

, Savage

, Maschino

, Al-Ahmadie

, et al. Prognostic impact of subclassification of radical prostatectomy positive margins by linear extent and Gleason grade. J Urol, 2013; 189:1302–1307.

14.

Menon

, Tewari

, Peabody

. Vattikuti Institute prostatectomy: Technique. J Urol, 2003; 169:2289–2292.

15.

Menon

, Shrivastava

, Bhandari

, Satyanarayana

, Siva

, Agarwal

. Vattikuti Institute prostatectomy: Technical modifications in 2009. Eur Urol, 2009; 56:89–96.

16.

Menon

, Tewari

, Baize

, Guillonneau

, Vallancien

. Prospective comparison of radical retropubic prostatectomy and robot-assisted anatomic prostatectomy: The Vattikuti Urology Institute experience. Urology, 2002; 60:864–868.

17.

Freedland

, Humphreys

, Mangold

, Eisenberger

, Dorey

, Walsh

, et al. Risk of prostate cancer-specific mortality following biochemical recurrence after radical prostatectomy. JAMA, 2005; 294:433–439.

18.

Eastham

, Kattan

, Riedel

, Begg

, Wheeler

, Gerigk

, et al. Variations among individual surgeons in the rate of positive surgical margins in radical prostatectomy specimens. J Urol, 2003; 170:2292–2295.

19.

Ginzburg

, Nevers

, Staff

, Tortora

, Champagne

, Kesler

, et al. Prostate cancer biochemical recurrence rates after robotic-assisted laparoscopic radical prostatectomy. JSLS n.d.;16:443–450.

20.

Barocas

, Salem

, Kordan

, Herrell

, Chang

, Clark

, et al. Robotic assisted laparoscopic prostatectomy versus radical retropubic prostatectomy for clinically localized prostate cancer: Comparison of short-term biochemical recurrence-free survival. J Urol, 2010; 183:990–996.

21.

Sooriakumaran

, Haendler

, Nyberg

, Gronberg

, Nilsson

, Carlsson

, et al. Biochemical recurrence after robot-assisted radical prostatectomy in a European single-centre cohort with a minimum follow-up time of 5 years. Eur Urol, 2012; 62:768–774.

22.

Dev

, Wiklund

, Patel

, Parashar

, Palmer

, Nyberg

, et al. Surgical margin length and location affect recurrence rates after robotic prostatectomy. Urol Oncol, 2014.

23.

Choo

, Cho

, Ko

, Jeong

, Lee

, Ku

, et al. Impact of positive surgical margins and their locations after radical prostatectomy: Comparison of biochemical recurrence according to risk stratification and surgical modality. World J Urol, 2014; 32:1401–1409.

24.

Sukumar

, Rogers

, Trinh

, Sammon

, Sood

, Stricker

, et al. Oncological outcomes after robot-assisted radical prostatectomy: Long-term follow-up in 4803 patients. BJU Int, 2014; 114:824–831.