Extended vs standard lymph node dissection in robot-assisted radical prostatectomy for intermediate- or high-risk prostate cancer: a propensity-score-matching analysis

Authors


Correspondence: Koon Ho Rha, Department of Urology, Urological Science Institute, Yonsei University College of Medicine, 50 Yonsei-ro, Seodaemun-gu, Seoul 120-752, Korea

e-mail: khrha@yuhs.ac

Abstract

What's known on the subject? and What does the study add?

  • Although lymph node dissection (LND) is known as the most accurate method of nodal staging, the therapeutic role of LND remains undetermined. This is mainly because of the lack of randomized prospective studies and the fact that retrospective analyses often result in bias and misinterpretation.
  • To overcome the limitation of retrospective analysis, we matched preoperative variables using propensity scores and compared the outcomes between patients treated with robot-assisted eLND and sLND. In the matched cohort, robot-asssited eLND achieved an increased detection rate of lymph node metastases; however, the therapeutic benefit was not statistically significant between the two groups on short-term follow-up.

Objective

  • To compare the pathological and biochemical outcomes between extended lymph node dissection (eLND) and standard lymph node dissection (sLND) in patients undergoing robot-assisted radical prostatectomy for intermediate- or high-risk prostate cancer.

Patients and Methods

  • A total of 905 patients underwent robot-assisted radical prostatectomy and lymph node dissection (LND) by a single surgeon between June 2006 and January 2011. Of these, 170 patients who underwent robot-assisted eLND and 294 patients who underwent robot-assisted sLND for intermediate- or high-risk prostate cancer were included in the study.
  • Propensity-score matching was performed using the preoperative variables which included age, body mass index, prostate-specific antigen, clinical stage, biopsy Gleason score 1 and 2, total number of biopsied cores, number of positive cores and prostate volumes.
  • Pathological and biochemical outcomes were assessed according to the extent of LND.

Results

  • The median (range) follow-up period was 36 (12–77) months and the median number of lymph nodes removed was 21 and 12 in the eLND and sLND groups, respectively.
  • Propensity-score matching resulted in 141 patients in each group. Although patients who underwent eLND had a higher clinical stage, biopsy Gleason score and number of positive cores than those treated with sLND in the entire cohort, there were no preoperative between-group differences in the matched cohort.
  • In the matched cohort, lymph node metastases were detected at a significantly higher rate in the eLND than in the sLND group (12.1 vs. 5.0%, P = 0.033).
  • In the matched cohort, the 3-year biochemical recurrence-free survival rates were 77.8 and 73.5% in the eLND and sLND groups, respectively, which was not significant (hazard ratio 0.85, P = 0.497).

Conclusion

  • Robot-assisted eLND achieved an increased lymph node yield and higher detection rate of lymph node metastases; however, robotic eLND did not alter biochemical outcomes in a short-term follow-up.
Abbreviations
LND

lymph node dissection

eLND

extended LND

sLND

extended LND

BCR

biochemical recurrence

IQR

interquartile range

HR

hazard ratio

Introduction

Lymph node dissection (LND) is the most accurate method of determining nodal staging in prostate cancer, but the indication and optimum extent of LND remains a matter of debate. The extent of LND is associated with the detection rates of lymph node metastases [1-4], and current guidelines recommend extended LND (eLND) in patients at risk of lymph node metastases [5-7].

With the increasing application of robot-assisted techniques used in radical prostatectomy, a significant decline in the use of LND has been reported [8]. Meanwhile, several studies have described the technical feasibility and safety of robot-assisted eLND [9-11]. Specifically, lymph node yield appears to be similar to those in open series, with robotic techniques not being a prohibitive factor for performing eLND; however, there is a lack of data regarding accurate comparison of robot-assisted eLND with standard LND (sLND), especially with respect to biochemical outcomes.

The therapeutic benefit of LND remains debatable, and no consistent conclusion has been reached. Some authors have demonstrated that the number of lymph nodes removed or the extent of LND is associated with oncological outcomes, supporting the possible therapeutic benefits of LND [12-14]. By contrast, several other studies have been unable to find any evidence that eLND has a beneficial impact on prostate cancer outcomes [15-17]. In the analysis of LND, it should be acknowledged that the extent of LND affects nodal staging, and a statistical artifact known as the ‘Will Rogers phenomenon’ can easily lead to a bias in results [18]; therefore, analysis exclusively of patients who have lymph-node-negative or lymph-node-positive disease may result in improved outcomes in patients who undergo eLND, and controlling pathological nodal staging may lead to similar results.

In the present study, to overcome the limitations of retrospective analyses, we used propensity-score matching to adjust for preoperative variables that are associated with lymph node metastases or the performance of LND. All surgeries were performed by a single surgeon using a robot-assisted technique, and we compared the pathological and biochemical outcomes between eLND and sLND in patients with intermediate- or high-risk prostate cancer.

Materials and Methods

Between June 2006 and January 2011, 905 patients with prostate cancer underwent robot-assisted radical prostatectomy by a single surgeon. All data were collected prospectively into an electronic database, and analysed after institutional review board approval (4-2012-0487). The inclusion criteria for the study included intermediate- or high-risk localized prostate cancer and patients who underwent eLND or sLND. Risk stratification was based on the D'Amico criteria [19]. Patients with neoadjuvant treatment or incomplete clinicopathological data, including information regarding the extent of LND or follow-up <12 months, were excluded from the analysis. A total of 464 patients, 170 from the eLND group and 294 from the sLND group, were included in the final analysis.

From 2006 LND was performed in all patients using a standard template that included the external iliac and obturator fossa area; eLND was used starting in 2008. All surgeries were performed via a transperitoneal approach and LND was performed before radical prostatectomy. The decision to perform eLND was based on the risk of lymph node metastases and consent of patients. Patient selection was also influenced by the surgeon's learning curve during the study period, and was eventually determined according to the surgeon's discretion. The boundaries of eLND included the intrapelvic area (internal iliac, obturator, external iliac) and common iliac area up to the ureteric crossing. Peritoneal incisions commenced from the lateral to medial umbilical ligament. After the bladder was dropped down, the peritoneal incision proceeded proximally up to the ureteric crossing over the iliac vessel, after which the lymphatic tissue covering the common iliac artery was removed. Dissection of the external iliac packet was limited by the lateral border of the external iliac artery and inferiorly by the node of Cloquet. After dissecting the tissues around the common iliac artery and its bifurcation area, the internal iliac artery was identified and nodal tissues around the internal iliac were removed. Lymphatic tissues within the obturator fossa were also removed, sparing the obturator nerve.

Clinicopathological variables including age, body mass index, preoperative PSA, clinical stage, prostate volume on TRUS, biopsy data and pathological information were investigated. After surgery, PSA levels were assessed every 3 months for the first year and semiannually thereafter. Biochemical recurrence (BCR) was defined as detection of serum PSA > 0.2 ng/mL with a secondary confirmatory increase at least 6 weeks after surgery. Subsequent treatment (i.e. radiotherapy or hormonal therapy) was not performed in the adjuvant setting and was delayed until biochemical failure.

For the development of the propensity-score-matched pair, the five-to-one-digit matching algorithm was used. Patients were matched by a propensity score of up to five digits and if a match was not available, attempts were made for a four-, three-, two-, and one-digit match. The propensity score of each patient was calculated using multivariate logistic regression for eLND with the preoperative variables which included age, body mass index, PSA, clinical stage, biopsy Gleason score 1 and 2, total number of biopsied cores, number of positive cores and prostate volumes on TRUS. The discrimination and calibration abilities of the propensity score model were assessed by means of C statistics and the Hosmer–Lemeshow test. The model had a C statistic of 0.727 and a Hosmer–Lemeshow goodness-of-fit P value of 0.911, indicating that this model was reasonably calibrated with strong discrimination.

The clinicopathological features were compared between the two groups. Complications that were presumably associated with LND were also recorded and lymphocele was regarded as a complication when it presented with clinical symptoms or required any type of invasive treatment. Qualitative variables were compared using a chi-squared test while quantitative variables were compared using Student's t-test. BCR-free survival was calculated by Kaplan–Meier analysis. The effect of eLND on BCR was assessed using Cox's proportional hazard regression by calculating the hazard ratio (HR). All statistical procedures were performed using SAS software version 9.2 (SAS Institute). A P value <0.05 was considered to indicate statistical significance, and all P values were two-sided

Results

Characteristics of the study cohort are shown in Table 1. In the entire cohort, preoperative clinicopathological features differed significantly between the two groups. Briefly, patients in the eLND group had significantly higher clinical stage, biopsy Gleason score, and number of positive biopsied cores. PSA levels in patients in the eLND group were higher than those in sLND group, with a trend toward significance (P = 0.074). Propensity-score matching resulted in a cohort of 141 patients in each group. In the matched cohorts, there were no between-group differences with respect to preoperative clinicopathological variables.

Table 1. Preoperative clinicopathological features of patients, stratified by the extent of LND.
 Entire cohortPropensity-score-matched cohort
sLND (n = 294)eLND (n = 170)PsLND (n = 141)eLND (n = 141)P
Age, years  0.188  0.899
Mean (sd)64.3 (7.4)65.2 (6.3) 65.2 (6.5)65.1 (6.6) 
Median (IQR)65 (60–69)66 (62–70) 66 (62–70)66 (61–70) 
BMI, kg/m2  0.156  0.990
Mean (sd)23.8 (2.4)24.1 (2.4) 24.1 (2.6)24.1 (2.5) 
Median (IQR)23.7 (22.2–25.6)24.2 (22.4–25.7) 23.9 (22.2–26.0)24.0 (22.3–25.7) 
PSA level, ng/mL,  0.074  0.396
Mean (sd)11.6 (12.4)13.7 (12.1) 13.7 (16.4)12.4 (8.7) 
Median (IQR)8.4 (5.3–13.7)10.4 (6.6–16.1) 9.1 (5.9–15.1)10.1 (6.6–14.7) 
Prostate volume, mL  0.740  0.946
Mean (sd)37.1 (17.1)37.7 (15.9) 38.0 (18.9)37.9 (15.4) 
Median (IQR)32.4 (25.8–44.0)34.1 (28.4–42.5) 33.2 (25.9–44.4)34.2 (28.6–43.7) 
Clinical stage, n (%)  <0.001  0.867
T1195 (66.3)78 (45.9) 75 (53.2)73 (51.8) 
T263 (21.4)70 (41.2) 47 (33.3)49 (34.8) 
T336 (12.3)22 (12.9) 19 (13.5)19 (13.5) 
Biopsy Gleason score, n (%)  <0.001  0.679
698 (33.4)30 (17.7) 28 (19.9)27 (19.1) 
7142 (48.3)67 (39.4) 67 (47.5)64 (45.4) 
8–1054 (18.3)73 (42.9) 46 (32.6)50 (35.5) 
Risk group, n (%)  <0.001  0.529
Intermediate163 (55.4)47 (27.6) 50 (35.5)45 (31.9) 
High131 (44.6)123 (72.4) 91 (64.5)96 (68.1) 
No. of total biopsied cores  0.540  0.802
Mean (sd)11.5 (3.5)11.3 (2.7) 11.4 (3.3)11.5 (2.9) 
Median (IQR)12 (10–12)12 (10–12) 12 (10–12)12 (12–12) 
No. of positive cores  0.034  0.786
Mean (sd)3.4 (2.5)4.0 (2.9) 3.9 (2.9)3.8 (2.7) 
Median (IQR)3 (2–5)3 (2–5) 3 (2–5)3 (2–5) 

Table 2 shows the pathological outcomes according to the extent of LND. In the matched cohort, the median numbers of lymph nodes removed were 21 and 12 in the eLND and sLND groups, respectively (P < 0.001). Although there was no difference in surgical Gleason score and pathological stage between the two groups, lymph node metastases were detected at a significantly higher rate in the eLND group (12.1 vs. 5.0%, P = 0.033). eLND was associated with a higher rate of lymph node metastases (odds ratio 2.64, 95% CI 1.05–6.54, P = 0.038).

Table 2. Pathological outcomes stratified by the extent of LND.
 Entire cohortPropensity-score-matched cohort
sLND (n = 294)eLND (n = 170)PsLND (n = 141)eLND (n = 141)P
Gleason score, n (%)  <0.001  0.642
675 (25.5)23 (13.5) 21 (14.9)22 (15.6) 
7173 (58.8)86 (50.6) 80 (56.7)83 (58.9) 
8–1046 (15.6)61 (35.9) 40 (28.4)36 (25.5) 
Pathological stage, n (%)  0.089  0.396
T2183 (62.2)96 (56.5) 81 (57.4)89 (63.1) 
T3a84 (28.6)49 (28.8) 42 (29.8)36 (25.5) 
T3b27 (9.2)25 (14.7) 18 (12.8)16 (11.4) 
No. lymph nodes removed  <0.001  <0.001
Mean (sd)11.9 (4.6)21.3 (6.7) 12.2 (4.7)21.1 (6.5) 
Median (IQR)12 ( 9–16)21 (16–25) 12 (9–16)21 (16–25) 
Lymph node metastases, n (%)10 (3.4)23 (13.5)<0.0017 (5.0)17 (12.1)0.033

The median (range; interquartile range [IQR]) follow-up period for all patients was 36 (12–77; 24–47) months. Atthe time of analysis, 111 (23.9%) patients experienced BCR and the 3-year BCR-free survival rates were 72.7% in the eLND group and 79.8% in the sLND group. The BCR-free survival rate of the eLND group was significantly lower than that of the sLND group amongst the entire cohort (eLND vs. sLND, HR 1.48, 95% CI 1.00–2.18, P = 0.048 [Fig. 1A]). In the matched cohort, the 3-year BCR-free survival rates were 77.8 and 73.5% in the eLND and sLND groups, respectively. BCR-free survival was higher in the eLND group, and the HRs appeared to trend, in an inverted manner, in favour of eLND, but the difference was not significant (eLND vs. sLND, HR 0.85, 95% CI 0.52–1.36, P = 0.497 [Fig. 1B]).

Figure 1.

Kaplan–Meier curve of BCR-free survival rates according to the extent of lymph node dissection in A, an entire cohort and B, a propensity-score-matched cohort.

Complications relating to LND occurred in 31 patients (6.7%) in the entire cohort; eLND was associated with a higher complication rate (11.8 vs. 2.4%, P < 0.001 [Table 3]). The rate of clinically significant lymphocele was significantly higher in the eLND group (P = 0.043). Fifteen patients in the eLND group had lymphoedema, which resolved in most patients after physical treatment and only three patients (1.8%) had persistent symptoms, all of which were mild.

Table 3. Postoperative complications related to LND stratified by the extent of LND.
n (%)sLND, n = 294eLND, n = 170P
Clinically significant lymphocele1 (0.3)4 (2.4)0.043
Neuropraxia2 (0.7)1 (0.6)0.905
Lymphoedema4 (1.4)15 (8.8)<0.001
Total7 (2.4)20 (11.8)<0.001

Discussion

In the present study, we examined 464 patients with intermediate- or high-risk prostate cancer who underwent robot-assisted radical prostatectomy and LND. The method for selecting patients for eLND was not consistent during the study period, and thus we used propensity-score-matching analysis to compare patients who were treated with robot-assisted eLND with those who underwent robot-assisted sLND with similar clinicopathological features. The analysis focused on pathological and biochemical outcomes.

Several studies have already shown that extensive LND is associated with an increased rate of lymph node metastases, and that LND continues to be the most accurate staging procedure [2, 4, 14, 20, 21]. Heidenreich et al. [2] reported that twice as many lymph node metastases were identified with eLND than with sLND. In another study of 648 men, eLND was associated with a nearly threefold higher number of lymph node metastases compared with sLND, which resulted in an increased odds ratio of 8.31 after controlling for clinicopathological prognostic variables [20]. Many authors have also shown that up to 50% of positive lymph nodes are found along the internal iliac artery [4, 14, 15]. In the matched cohort of the present study, the rates of lymph node metastases were 12.1 and 5.0% in eLND and sLND, respectively. Similarly, we found that the detection rate of lymph node metastases increased approximately twofold after adding the internal iliac and common iliac areas in our analysis.

The therapeutic effect of LND on prostate cancer outcome continues to be a contentious issue. Many researchers did not find improved oncological outcomes with extensive LND, even in patients with higher-risk disease [15, 16, 22, 23]. By contrast, other studies have suggested that LND may improve prostate cancer outcomes by eliminating micrometastatic nodal disease that might otherwise progress to systemic dissemination; however, all these results were from retrospective analyses that used different inclusion criteria, varied definitions of LND extent, and inconsistent dissection techniques from multiple surgeons.

In a retrospective analysis of 4611 patients, Masterson et al. [14] reported that a greater number of lymph nodes removed was significantly associated with BCR-free survival. The data obtained from 13 020 men in the Surveillance, Epidemiology and End Results programme showed that the number of lymph nodes removed was likely to increase with prostate-cancer-specific survival [13]; however, the reported results of improved outcomes were from patients with no lymph node involvement. Indeed, when analysis was restricted to only patients with negative lymph nodes, the ‘Will Rogers phenomenon’ probably leads to a misinterpretation of the data [18]. eLND provides more accurate staging information; thus, patients with lymph node metastases that is diagnosed with eLND may be classified as not having lymph node metastases if sLND were to be performed on them. This would undoubtedly have a negative effect on the prognosis of the group with sLND. Allaf et al. [3] reported that a significant benefit in BCR-free survival for eLND vs sLND was found in patients with lymph-node-positive disease involving <15% of extracted nodes; however, analysis that is restricted to patients with lymph node metastases results in a similar misinterpretation of stage migration. Moreover, the extent of LND differed between the two surgeons who conducted the study and thus their results may have been influenced by surgeon factors.

In the present study, to eliminate bias caused by stage migration according to the extent of LND, the pathological and biochemical outcomes of the two groups were compared after preoperative variables were matched using propensity-score matching. Multivariate analysis including postoperative pathological variables was not performed. Moreover, all surgeries were performed by a single surgeon with a consistent technique, and the extent of LND was determined by using a definite anatomic template.

A possible explanation for the therapeutic effect of LND lies in improving oncological outcomes by eliminating micrometastatic disease. In the present series, even though the BCR-free survival rate of the eLND group was higher than that of the sLND group in the matched cohort, the difference was not significant. LND performed using a robotic technique may have contributed to these discouraging results; however, the median number of lymph nodes removed in the eLND group was 21, which was comparable to the open series [4, 24, 25]. Meanwhile, a significant proportion of patients with intermediate- or high-risk prostate cancer has extracapsular extension, seminal vesicle invasion or positive surgical margin and these patients would still be at increased risk of BCR even if micrometastatic nodal disease were eliminated by LND. Moreover, biochemical evaluation is a more sensitive tool for detecting recurrence than clinical evaluation, and the number of patients cured biochemically by LND would be smaller than expected. To demonstrate an improvement of biochemical outcome with LND, we believe that a large cohort is necessary. When we calculated the sample size of a prospective trial based on our results to demonstrate a biochemical outcomes with a difference of 4% (76.9 vs. 72.9%) at 3 years after surgery between eLND and sLND groups, ∼1400 patients in each arm would be required for a 5% significance level (two-sided) and 80% power.

In the present study, all procedures were performed using a robot-assisted technique. Recently several studies have reported their experiences in robot-assisted eLND. Table 4 [9, 10, 26-28] shows the results of other robot-assisted eLND series. The number of lymph nodes removed in most series was >15 and the rate of lymph node metastases ranged from 13 to 18%. Even though the inclusion criteria were not uniform among the studies, our findings were similar to those of other robot-assisted eLND series. The incidence of symptomatic lymphocele was <5% in patients who underwent transperitoneal robot-assisted eLND. Orvieto et al. [29] reported that 51% of patients who underwent robot-assisted LND had lymphocele formation on pelvic CT; however, in that study, most lymphoceles were subclinical. Moreover, it seems apparent that robot-assisted eLND via the transperitoneal approach is associated with a lower incidence of symptomatic lymphoceles when compared with the extraperitoneal approach [26].

Table 4. Comparison with other robot-assisted eLND series.
 No. of patientsMedian PSA level, ng/mLMedian no. of lymph nodes removedIncidence of positive lymph nodes, %Lymphocele, n (%)
  1. aFrom January 2010 to May 2010, symptomatic lymphocele was observed in 19% of extraperitoneal robot-assisted radical prostatectomy (n = 30) and 0% in transpertitoneal robot-assisted radical prostatectomy (n = 62).
Feike et al. [9]997.719164 (4.0)
Davis et al. [27]352not available16180% and 19%a
Sagalovich et al. [28]827.41313not available
Silberstein et al. [29]1205.916134 (3)
Yuh et al. [10]1437.820134 (3)
Present study17010.421134 (2.4)

The present study has several limitations. Although we matched the two groups by preoperative variables, the matching process is different from randomization and selection bias may have arisen in the matching process. While all procedures were performed by a single surgeon, the selection of patients for eLND was affected not only by the surgeon's learning curve but also by the surgeon's discretion which was in turn affected by various factors. This may result in selection bias and limit the impact of our results with respect to oncological outcomes. Finally, the follow-up period was relatively short. Thus, even though there were 111 cases of biochemical failure in the present study, which was seen as sufficient for analysis, a larger number of patients with a longer-term follow-up study would be needed to evaluate the difference in oncological outcomes according to the extent of LND.

In conclusion, robot-assisted eLND achieved significantly higher lymph node yield than sLND in patients with intermediate- or high-risk prostate cancer. Analysis of the matched cohorts showed that eLND significantly increased the accuracy of pathological lymph nodal staging, but the extent of LND did not appear to affect biochemical outcomes during a short-term follow-up period.

Acknowledgements

We appreciate our statistical advisor Professor Dae Ryong Kang (PhD) for his excellent review of the article. The Research Foundation of Korea (NRF) grant funded by the Korean government (MEST) (2011-0029348) supported this work.

Conflict of Interest

None declared. No competing financial interests exist.

Ancillary