- To investigate the learning curve for performing extended pelvic lymphadenectomy (ePLND) during laparoscopic radical prostatectomy (LRP) in patients with intermediate- and high-risk prostate cancer.
the cumulative sum
deep vein thrombosis
European Association of Urology
high-intensity focused ultrasound
lymph node (involvement)
(extended) (limited) pelvic LN dissection
(laparoscopic) (robot-assisted) radical prostatectomy
Pelvic lymph node dissection (PLND) at the time of radical prostatectomy (RP) is the most reliable method of accurately determining lymph node (LN) status in a patient with prostate cancer . Furthermore, the removal of LNs may improve biochemical and survival outcomes [2, 3] and lead to the early initiation of androgen-deprivation therapy (ADT) with proven cancer-specific survival benefit . Although PLND has been advocated for decades, its exact place in the management of localised prostate cancer continues to be debated . The latest (2007) guidelines from the AUA recommend that a PLND should be performed during RP in patients that have a ‘higher risk’ of LN involvement (LNI) . The most recently published (2012) guidelines of the European Association of Urology (EAU) recommend that extended PLND (ePLND) should be performed in intermediate- and high-risk prostate cancer, if the estimated risk for positive LNs exceeds 5% . Briganti et al.  have reported that in these patients the estimated risk of LN positivity is 15–40%. Limited PLND (lPLND) should no longer be performed, because it misses at least half the LNs involved . As the most commonly used prostate cancer nomograms, i.e. those attributed to Partin  and the Memorial Sloan-Kettering Hospital , are based on data from patients who have had lPLND, they will tend to under-estimate the risk of LNI. Conversely, nomograms specifically based on ePLND indicate that the risk to patients in the intermediate- and high-risk groups  of LNI is significantly greater than previously appreciated .
In addition to its greater accuracy in staging prostate cancer, and in contrast to its imaging counterparts, PLND also allows for the removal of involved LNs. Masterson et al.  showed an inverse correlation between the number of histologically normal LNs removed and the risk of biochemical recurrence, confirming the fact that LN micrometastases may evade detection during pathological examination. This is further supported by the observation of Heidenreich et al.  that ePLND improved cancer survival in men with involved LNs by 23% and in men with no LNI by 15%. Two studies have shown a correlation between the total number of LNs removed and the probability of detecting LNI [14, 15], and several studies have shown a good prognosis even in LN-positive patients who have undergone PLND, with and without the use of adjuvant ADT, as long as the number of affected LNs removed is no greater than two [16-19].
It is logical to conclude that one should try to remove as many LNs as possible as: i) histological evaluation of LNs is the best test of LNI; ii) prostate cancer is known to ‘skip’ LNs to involve more proximal nodes; and iii) the removal of LNs in patients with ≤2 LNs may still be associated with a good prognosis. However, the literature shows that a more extensive PLND leads not only to a lengthier operation but may also be associated with a significantly higher complication rate , although there is some disagreement about this [20, 21]. The question therefore arises as to how best to comply with existing guidance to give patients the best oncological result from RP whilst at the same time managing patient expectations about safety during the surgical learning curve. Knowledge of the length and gradient of the learning curve for ePLND, currently absent from the literature, would help to plan the allocation of personnel within a Urology Department and to plan operating lists. It would also be of potential use in identifying scenarios in which the introduction of ePLND might be undesirable, i.e. in low-volume practices.
The present study reports the initial results of one surgeon with experience of open RP, laparoscopic RP (LRP) and robot-assisted RP (RARP) in performing laparoscopic (the approach with which he has the most experience) ePLND in patients with intermediate- or high-risk prostate cancer.
As a result of the increasing evidence in favour of performing ePLND for patients with intermediate- and high-risk prostate cancer the authors adopted this as standard practice in March 2008. Before this date the lead author had performed or supervised 868 cases of LRP during a 96-month period, which had included ePLND in eight high-risk patients and lPLND in 311 intermediate-risk cases. During the 48 months since March 2008 the authors performed a further 882 cases of LRP (Table 1), of which 500 had an ePLND (56.7%) for intermediate- or high-risk prostate cancer. The series also included 14 patients who had salvage surgery after failed brachytherapy (six patients), external beam radiotherapy (seven) and two sessions of high-intensity focused ultrasound (HIFU, one). All PLNDs were done before dissection of the prostate and the nodal tissue was sent separately from each side of the pelvis, but not from each location on each side. The series was consecutive. All patients had a bone scan and CT/MRI to exclude metastases.
|Age, years)||64.0 (43–78)|
|Body mass index, kg/m2||27.0 (18–39)|
|PSA level, ng/mL||8.0 (1–62.5)|
|Biopsy Gleason||7 (6–10)|
|Failed primary intervention:||14 (2.8)|
|external beam radiotherapy||7 (1.4)|
ePLND was performed via a transperitoneal approach to minimise the risk of postoperative lymphocele formation and consisted of excision of all nodal tissue within the following anatomical boundaries: distal to where the ureter crosses the common iliac artery, proximal to the pubic bone, medial to the mid-diameter of the external iliac artery and lateral to the bladder, as previously described by Mattei et al. . The remainder of the LRP was performed as previously described .
The operating time (time from first incision to closure of the last wound) was recorded. A second-generation cephalosporin was administered as antibiotic prophylaxis for the first 48 h. Deep vein thrombosis (DVT) prophylaxis consisted of pneumatic intermittent calf compression during surgery, s.c. low molecular-weight heparin 5000 units once daily or enoxaparin 40 mg once daily until discharge and calf compression stockings. Pelvic drains were removed when the volume draining was <100 mL/12 h. Patients were discharged home when comfortable. Patients whose drains were still active at the time of their expected departure were discharged home with the drain still in situ, to be removed by their District Nurse once the volume draining was <100 mL/12 h. Catheters were removed at 8–14 days without a prior cystogram, except after salvage surgery when a cystogram was performed after 3 weeks.
All surgical specimens were marked according to side and fixed in formalin. LNs were identified by palpation, described macroscopically according to their size and consistency, and then processed. LNs of >0.3 cm diameter were bisected before embedding in paraffin wax and all LNs were cut and stained with haematoxylin and eosin before microscopic examination by one of three Consultant Uro-Pathologists.
The parameters chosen as being the key markers of the learning curve for ePLND were operating time, LN count and complications. LN parameters were compared with the authors' historical controls, i.e. their preceding 311 cases of lPLND. Data was patient-reported using a questionnaire and was prospectively recorded using Microsoft Excel software (Microsoft Corporation, Redmond, USA) and analysed using Analyse-it version 2.26 software (Analyse-it Software Ltd, Leeds, England). Continuous variables were compared using the independent samples t-test and rates using Fisher's exact test.
The cumulative sum (CUSUM) method was used to calculate the learning curves [23-25]. The simple CUSUM  was used for the binomial data (complications) and was based on the formula: Sn = Σ(Xo – Xi), where Sn is the cumulative sum, Xo is the target competence level (both 0.95 and 0.90 levels were examined) and Xi = 1 for a successful outcome and 0 for a failure. Thus, a negative trend of the CUSUM line indicates success, whereas a positive trend indicates failure for the procedure under analysis. When the curve changes its gradient from upwards (positive slope) to downwards (negative slope), the learning curve is considered complete. With procedures that have a high success rate early on, the simple CUSUM plot starts in a negative direction (the direction of success) and maintains its gradient, thus concluding that there is no learning curve. For the operating time and LN yield a cumulative average was used to plot the learning curves, as a plateau or minimum rate was desired. Wright's Model was used to calculate the rate from that procedure onward that showed a linear trend on a logarithmic scale. This was repeated until the rate was <5%, which was where the learning curve was considered to have plateaued.
|Variable||Value||Position in series, patient number|
|Mean (range) operative time, min||200 (120–420)|
|Mean (range) blood loss, mL||200 (10–1400)|
|Transfusion, N (%)||8 (1.6)|
|Mean (range) prostate weight, g||52.0 (25–178)|
|Mean (range) Hospital stay, nights||3.0 (2–14)|
|All complications, N (%)||36 (7.2)|
|Generic, N (%)||19 (3.8)|
|Small bowel injury||1||81|
|Acute renal failure due to chronic urine retention||1||43|
|Bladder neck stenosis||6||66,103,111,283,310,354|
|Port site hernia||2||196,242|
|PLND-specific, N (%)||17 (3.4)|
|Obturator nerve injury||2||23,137|
|Infected pelvic collection||2||99,298|
|(a) Clinical parameters|
|Gleason grade, n/N (%):|
|4 and 5||0/13||0/0||1.00|
|6||1/144 (0.7)||5/76 (6.6)||0.04|
|all 7||1/123 (0.8)||46/358 (12.8)||<0.001|
|3+4||0/79 (0)||30/271 (11.1)||<0.001|
|4+3||1/44 (2.3)||16/87 (18.4)||0.012|
|8||1/27 (3.7)||8/39 (20.5)||0.10|
|T stage, N (%)|
|T1||1/102 (9.8)||7/177 (4.0)||0.29|
|T2||2/196 (1.0)||40/282 (14.2)||<0.001|
|d'Amico risk category, N (%):|
|intermediate||1/225 (0.4)||27/320 (8.4)||<0.001|
|high||2/79 (2.5)||35/180 (19.4)||<0.001|
|(b) Pathological parameters|
|Median (range) LN count||6 (2–8)||14 (5–46)||<0.001|
|Gleason grade, N (%):|
|4 and 5||0/12||0/0||1.00|
|6||1/140 (0.7)||1/56 (1.8)||0.98|
|all 7||1/120 (0.8)||41/395 (10.4)||<0.001|
|4+3||1/42 (2.4)||21/85 (24.7)||0.002|
|8||1/34 (2.9)||2/14 (14.3)||0.40|
|T stage, N (%):|
|2b/c||1/184 (0.5)||11/270 (4.1)||0.03|
|3b||1/25 (4.0)||29/66 (43.9)||<0.001|
The median (range) preoperative PSA level was 8.0(1–62.5) ng/mL and the Gleason score was 7 (6–10). In all, 64% of patients were in the d'Amico intermediate-risk group and 36% were in the high-risk group. Patients having salvage surgery had a variable distribution of periprostatic fibrosis but no specific difficulty was encountered in performing their ePLND and there were no complications in these patients. In the whole series, there were no intraoperative blood transfusions and no conversions to open surgery. The learning curve for operating time is plotted on a logarithmic scale in Figure 1 to better demonstrate the beginning of the learning curve. The operating time fell at a steady rate of 2.7% after the 15th patient and plateaued after 130 patients. The median blood loss of 200 mL and postoperative transfusion rate of 1.6% for the series were low and compare favourably with other reported laparoscopic and robotic series.
Certain complications were deemed to be PLND-specific: obturator nerve injury, ureteric injury; postoperative bleeding; lymphocele; and infected pelvic collection. Figures 2 and 3 show the learning curves for PLND-specific complications and all complications encountered in the series, respectively. Both graphs had polynomial functions fitted to the data to determine where the gradient of the curve changed. At competence levels of 5% and 10%, the learning curve for all complications ended after 346 and 136 patients, respectively. At a 5% competence level the learning curve for PLND-specific complications was 40 cases and there was no learning curve at a 10% competence level. The small bowel injury was sustained during port insertion in a patient with a multiply-operated abdomen and was not recognised until 24 h after surgery. It was managed by open small bowel resection and anastomosis. One of the rectal injuries was in a post-radiotherapy salvage case and had a temporary defunctioning colostomy. The other two cases were recognised intraoperatively and successfully managed by laparoscopic suture. Both obturator nerve injuries recovered fully. Both ureteric injuries were managed laparoscopically, one by suture repair and stent insertion and the other by reimplantation and stent insertion with no evidence of stenosis at 2 years. Six patients were returned to the operating theatre ≤24 h of LRP for haemodynamic instability and a falling haemoglobin level, all of whom were managed laparoscopically by aspiration of blood, saline lavage and haemostasis. Two patients required CT-guided drainage of infected pelvic fluid collections, one at 4 weeks after surgery and the second 3 months after surgery. Two patients developed symptomatic (pain) lymphoceles (defined as a discrete and cystic mass containing lymphatic fluid), both of whom were managed conservatively with resolution of symptoms by 3 months. The three patients that developed lower limb lymphoedema were managed by compression stockings and leg massage. Prolonged lymphatic drainage was not included as a complication as it did not affect patients' discharge from hospital.
The median LN count after ePLND was more than double that of lPLND (14 vs 6, P < 0.001) and increased with experience up to the end of the series. Figure 4 shows the cumulative average of the LN counts over the 500 procedures. The curve plateaus after 150 procedures with a modest rate of increase thereafter of 7.3%. Interestingly, there appeared also to be a second, but less marked, increase in the rate of LN yield of 24% between patients 390–500, suggesting an on-going improvement of technique. The likelihood of LNI correlated with biopsy Gleason grade, with the exception of Gleason 9 cases, possibly because of the few patients in each group (two and 24). ePLND was significantly better at identifying LNI in patients with biopsy Gleason 6 and 7 disease (P = 0.04 and P < 0.001) but this did not apply to patients with Gleason 8–10 prostate cancer, presumably because in this range the under-sampling of LNs by lPLND was adequately compensated by the higher rate of LNI. ePLND was also more effective than lPLND at identifying LNI in patients with clinical stage T2 (P < 0.001) and T3 (P = 0.015) disease and in patients in the intermediate- and high-risk (both P < 0.001) prostate cancer groups. For pathological grade and stage, ePLND identified more positive LNs than did lPLND in patients with Gleason 7 (P < 0.001) and stage T2b–T3b (P = 0.03 to P < 0.001) prostate cancer.
The present large series of laparoscopic ePLND suggests the number of cases needed to be done to overcome the separate learning curves for operative time (130 cases), all complications (136 cases), PLND-specific complications (40 cases) and LN yield (150) during LRP. Although not yet tested in large numbers of patients having RARP the authors feel that similar figures are likely to apply for RARP based on the robot-assisted cases that they have performed to date that have included ePLND.
The initial cases of ePLND added a significant amount of time (1.5–2 h was not unusual) to the total operating time taken to complete the case. However, with standardisation of the steps of the procedure, removal of the nodal tissue in one packet and with greater experience the time taken to perform ePLND fell sequentially. The trend in operating time is not likely to have been significantly influenced by the salvage prostatectomy cases as they accounted for only 2.8% of cases, took only 6 min more to perform (median 211.5 vs 205.5 min; P = 0.73) than non-salvage cases and were evenly distributed in the last three 100-patient cohorts: four, four, and six cases.
The overall complication rate of 7.2% was higher than the 4.8% reported by the authors for their first 1000 cases  but is much less than the three-fold increase in complications noted by Briganti et al.  for open ePLND. At first glance this would tend to suggest that the authors had still not reached the summit of their learning curve by the time they had started to incorporate ePLND into their practice, despite having already done 868 cases of LRP. However, as the series progressed a greater number of cases were used for training and trainees performed a greater proportion of the case, which may explain why the generic complication rate rose as the series progressed. Conversely, trainees performed very few ePLNDs in view of novelty of the procedure and the greater technical difficulty involved. Both ureteric and both obturator nerve injuries occurred during the first 273 cases, which led to the progressive decline in the PLND-specific complication rate seen in Figure 2. Conversely, most bleeding complications occurred in the latter half of the series, which is likely to reflect the more extensive pelvic dissection that led to the higher LN yield as the series progressed. The results seen in the present series compare favourably with the 0.4% incidence of ureteric injury, 0.1% risk of obturator nerve injury, 1.7% risk of DVT/pulmonary embolism and overall complication rate of 6.5% reported by Musch et al.  in 1380 patients having open RP in which 69% of cases had a lPLND and 31% had ePLND. The re-intervention rate in the Musch et al. series was 6.3%, half of which was for lymphocele formation. Comparison of complication rates between series is hampered by the fact that whilst most report all complications [1, 28], some report only complications thought to be directly related to PLND (ureteric, vascular and nerve injuries of the pelvic side-wall, significant lymphoceles and thromboembolic events) .
Although statistical analysis suggests learning curves for all-cause and PLND-specific complications of 136 and 40 cases, respectively, scrutiny of Table 2 shows that 22 of the 136 complications (61%) encountered in this series occurred beyond the learning curve of 136 cases. Additionally, 14 of the 17 (82%) PLND-specific complications occurred beyond the learning curve of 40 cases, including all of the ureteric and half of the obturator nerve injuries. These facts emphasise that that serious complications can occur when performing ePLND beyond the point at which statistical analysis suggests that the steepest part of the learning curve has been overcome and that surgeons need to be aware that experience in performing ePLND does not confer on them immunity from complications.
As seen in other open and minimal access RP series, the LN count in the present series was significantly higher for ePLND, 14 vs 6 (P < 0.001), most of which are likely to have come from the internal iliac group and others from dissection on the medial side of the external iliac artery. Notwithstanding the limitations of comparing LN counts between series listed above, the LN count in the present series might have been depressed compared with other series, e.g. 28 for ePLND and 11 for lPLND [20, 30] in which nodal tissue was sent as separate packages, as this is known to increase the LN count . In other words, the reported number of LNs does not necessarily reflect the actual number of LNs removed. However, the LN yield certainly compared favourably with other series, for example that of Allaf et al.  in which the counts for lPLND and ePLND were 8.9 and 11.6, respectively. Briganti et al.  has previously reported that LN harvest and positivity at ePLND is related to prior surgical experience but the natural variability of pelvic LN count of 8–56 (mean 20) seen by Weingartner et al.  at post mortem studies precludes setting a limit as being a benchmark for surgical adequacy.
Table 3 shows that ePLND has a significantly greater sensitivity for the detection of LNI compared with lPLND, specifically in those patients whom urologists might not consider to have a particularly high risk of LNI and therefore deserving of ePLND, i.e. those with intermediate-risk prostate cancer and all patients with biopsy Gleason 7 disease, regardless of whether this was 3+4 or 4+3 prostate cancer. This undoubtedly relates to the fact that most existing nomograms used to calculate the probability of LNI do so based on lPLND data and will therefore underestimate this figure.
The results from the present study also confirm that PLND can reasonably be omitted in patients with low-risk prostate cancer (LNI 0% after either lPLND or ePLND) but the 6.6% rate of LNI seen in biopsy Gleason 6 tumors is a concern and probably reflects the 35% upgrade from biopsy Gleason 6 to pathological Gleason 7 seen in most (including the present) RP series. Not surprisingly, patients with both biopsy and pathological Gleason 4+3 disease had a higher rate of LNI than did those with Gleason 3+4 disease (18.4% vs 11.1% and 24.7% vs 6.5%, respectively), which once again calls into question whether it is reasonable to describe all Gleason 7 tumours as being ‘moderately’ aggressive.
It is ironic that although the evidence in favour of performing ePLND in patients with intermediate-risk prostate cancer and above is reflected by internationally respected guidelines, the performance of any (let alone extended) PLND during RP appears to be declining , coincident with a rise in the adoption of minimal access RP. The authors share the concern of Wagner et al.  that in this arena oncological principles might be sacrificed in the interests of surgical expediency. A separate concern relates to the treatment of intermediate-risk prostate cancer by non-surgical means such as brachytherapy, cryotherapy and HIFU, none of which addresses the significant risk of LNI. Anecdotally, the authors have recently seen a number of patients several years after such treatment for intermediate-risk prostate cancer with a rising PSA level, a negative prostate biopsy and a new finding of pelvic lymphadenopathy on cross-sectional imaging representing residual or recurrent cancer in pelvic LNs. The inescapable conclusion, given the 8.4% risk of LNI in patients with intermediate-risk prostate cancer seen in the present series, is that either these treatments should be restricted to low-risk disease or they should be combined with ePLND or radiotherapy to the likely nodal landing sites for prostate cancer.
Limitations of the present study include its consecutive and non-randomised nature, which exposes it to the risk of cohort and selection biases. The total operating times recorded may not have accurately reflected the time taken to perform ePLND in individual cases. Longer follow-up of these will determine the oncological consequences, if any, of performing ePLND.
In conclusion, the present study suggests a learning curve of 130 cases for operating time, 136 cases for all complications, 40 cases for PLND-specific complications and 150 cases for LN yield rate in experienced hands. In demonstrating a risk of LNI for patients with intermediate- and high-risk prostate cancer of 8.4% and 19.4%, respectively it also sets the threshold for the patients being described at ‘higher risk’ of LNI in the 2007 AUA Prostate Cancer guidelines as including all patients with intermediate- and high-risk disease. It also shows that ePLND can be safely incorporated into LRP in a high-volume setting and that this should be done without undue delay to comply with existing guidelines from the AUA and EAU, and benefit patients.
I am greatly indebted to Sabeena Sidhu, Clinical Scientist at St. Luke's Cancer Centre, Guildford, UK for her expert advice in the statistical analysis of the data.
Covidien supplies a salary for a laparoscopic fellow within the department.