Does Initial Learning Curve Compromise Outcomes for Robot-Assisted Radical Cystectomy? A Critical Evaluation of the First 60 Cases While Establishing a Robotics Program

Introduction

O pen radical cystectomy (ORC) with bilateral pelvic lymph node dissection is the gold standard surgical treatment for patients with muscle invasive bladder cancer and bacillus Calmette-Guérin-refractory high-grade superficial bladder cancer. In recent years, robot-assisted laparoscopic radical cystectomy (RARC) with bilateral pelvic lymph node dissection (PLND) has gained popularity as a minimally invasive alternative. The concept of a learning curve was derived in 1936 when Wright proposed a mathematical model for the aircraft industry. ¹ There is a well-recognized period when surgeon inexperience makes an operation more difficult or lengthy that has been coined by some as the learning curve of any given operation. ² There is no widely accepted standard way to define or measure this well recognized phenomenon, however.

Learning curve metrics, including lymph node (LN) yield and surgical soft tissue margins that correlate with oncologic efficacy, are paramount in the evaluation of the learning curve of RARC. It is critical to maintain the oncologic efficacy of RARC during the learning curve because positive surgical margins and low LN yield have significant impact on patient outcomes. ^3,4 Robot-assisted pelvic lymphadenectomy is technically demanding and the extent of dissection, open or robotic, is not yet clearly established in the literature. ^5,6 Therefore, early critics of RARC have questioned whether equivalent oncologic outcomes are possible with the new technique, especially during the learning curve.

Proponents of RARC have suggested that the learning curve is lessened compared with pure laparoscopic radical cystectomy, with one study showing that RARC proficiency can be achieved by the 20th case by surgeons with previous robotic experience at a high-volume center. ⁷ Other studies, however, evaluating robotic surgery have shown a highly variable learning curve largely based on individual surgeon factors. ^2,8 Experience with other robotic pelvic operations, such as radical prostatectomy, as well as familiarity with open cystectomy may lesson the learning curve. Also, delineating what defines the learning curve for any given operation is highly variable.

Because RARC is gaining popularity across the globe, the goal of this analysis is to ensure that oncologic principles are not compromised during the learning phase of this operation, especially because bladder cancer is a lethal disease process where cure is largely dependent on sound oncologic principles. We established our robot-assisted cystectomy program in January 2008.

We sought to evaluate the effect of the initial learning curve on operative, postoperative, and pathologic outcomes of the first 60 RARC performed at our newly established robotics program using a modified mentorship program.

Patients and Methods

Before initiating this analysis, we obtained approval from our Institutional Review Board in recognition of and compliance with the United States Health Insurance Portability and Accountability Act of 1996 guidelines. From January 2008 through March 2010, a total of 60 patients with clinically localized bladder cancer underwent RARC with bilateral PLND at a single medical center and were included in this analysis. Data were prospectively compiled in our bladder cancer database, including patient demographics, preoperative disease characteristics, intraoperative variables, postoperative variables, and pathologic outcomes. ORC is still performed at our institution in selected patients who have had multiple previous abdominal surgeries or severe medical comorbidities, such as chronic obstructive pulmonary disease. The surgeries were performed with da Vinci^® standard, S, or Si (Intuitive Surgical,^® Sunnyvale, CA) robotic assistance and extracorporeal urinary diversion. Patients were counseled on different options for extracorporeal urinary diversion that included ileal conduit, continent cutaneous, or orthotopic neobladder. The orthotopic neobladder was constructed extracorporeally and then placed into the pelvis for robotic neobladder-urethral anastomosis.

We elected to analyze the first 60 cases because it represented our initial experience with RARC, and the senior author served as a mentor and team leader during these cases. The junior faculty (AKK and JAP) had no previous experience with RARC, so during their learning curve, AKH served as a mentor and preceptor. We have now performed more than 100 RARC at our institution. We noted that after the 60th case, the junior faculty had developed a comfort level with the operation and no longer needed active mentorship. We divided the patients into three distinct groups or tertiles (first 20, middle 20, and last 20) to analyze the effect of the learning curve on the LN yield, margin status, hospital stay, complications, operative time (inclusive of RARC, PLND, repositioning of patient for diversion, extracorporeal diversion, and closure), and estimated blood loss. The senior author served as a preceptor for 15, 8, and 7 cases in the 1st, 2nd, and 3rd tertiles, respectively, helping to guide the junior faculty through the steep part of their learning curve.

The preceptor is the primary surgeon responsible for the patient but supervises and assists the surgeon learner in the acquisition of new skills. The most challenging parts of the operation noted by the junior faculty that necessitated the most preceptorship were the extended lymphadenectomy, posterior dissection, as well as the apical dissection during anterior pelvic exenteration in females.

We used standard descriptive methods to characterize the three tertiles. For comparisons, we used analysis of variance for nonparametric comparisons of continuous data. We used the Fisher exact test to compare categorical data between the three groups. All statistics were performed using JMP^® 8 (SAS Institute Inc, Cary, NC).

Results

The patient's clinical, operative, and pathologic characteristics are shown in Table 1. There was no significant difference between the tertiles in regard to patient sex, age, body mass index (BMI), American Society of Anesthesiologists (ASA) class, and previous abdominal surgery history. Our overall patient population poses great challenges as evidenced by their significant comorbidities, with the majority being ASA class 3 or greater, the increasing trend toward obesity with average BMI of 29 in the 3rd tertile, and 50% having had previous abdominal surgery. A majority of patients underwent extracorporeal ileal conduit urinary diversion in all tertiles. Two patients in the 3rd tertile had previous pelvic external beam radiation for a history of prostate cancer, and a total of three patients underwent neoadjuvant chemotherapy (two in the 2nd tertile and one in the 3rd tertile) based on the practices of our local medical oncologists.

The mean total operative time (skin to skin) trended down from the 1st to 3rd tertile from 525 minutes to 449 minutes, respectively (P=0.059). Mean estimated blood loss for the entire operation was unchanged across tertiles. Mean hospital stay trended down from the 1st to 3rd tertile from 9.2 days to 7.4 days, respectively (P=0.29). We have previously reported that the median hospital stay in our RARC series is shorter compared with our ORC experience at our institution. ⁹

Table 2 lists all complications, and overall complications decreased as the learning curve progressed from 14 (70%) in the 1st tertile to 6 (30%) in each of the 2nd and 3rd tertiles (P<0.013). Complications were considered early if they occurred within 3 months of the date of surgery. Patients were followed per institutional protocol every 3 months in year 1 of surgery and every 6 months in year 2 and 3 to systematically evaluate long-term complications. Complications occurring >3 months postoperatively at median follow-up of 24, 18, and 11 months for the 1st, 2nd, and 3rd tertiles, respectively, were compiled and are listed in Table 2.

The tumor pathologic stage was similar between the three tertiles with 45%, 30%, and 35% of the patients in the 1st, 2nd, and 3rd tertiles, respectively, having pT₃/pT₄ disease. LN positive disease was present in 30%, 20%, and 40% of patients in the 1st, 2nd, and 3rd tertiles, respectively. The mean total LN yield was also similar between the three tertiles, with a mean total LN yield of 17.0, 19.1, and 14.4 from the 1st to 3rd tertile, respectively (P=0.131). There was no difference in the positive soft tissue margins across the three tertiles with one patient in each of the 1st and 2nd tertiles and four in the 3rd tertile with positive margins (P=0.189). Of note, all four positive margins in the 3rd tertile were in patients with pT₄ bladder cancer on final pathologic evaluation, two of whom had undergone previous radiologic therapy for prostate cancer. One of these positive margins was focally positive at the bladder base in a patient who had direct prostatic invasion, and the other patient's final pathologic status was pT₄N₂.

After the 60th consecutive RARC performed at our institution, active mentorship was no longer needed, and complete data are available on the subsequent 30 RARC that were performed. Mean estimated blood loss, total operative duration, and hospital stay all continued to show promising trends of 302 mL, 401 minutes, and 6.6 days, respectively. The mean LN yield increased to 19.4, and four patients with extravesical disease as well as positive LNs sustained positive soft tissue margins. The overall complication rate for this cohort was 23% (7/30) with only one major complication.

Discussion

Robot-assisted laparoscopic pelvic lymphadenectomy at the time of radical cystectomy has been shown in several reports to be safe, feasible, and efficacious; however, RARC remains, rightfully, under intense scrutiny, because the technique is evolving and no long-term oncologic outcome data exist. We have evaluated our data in the first 60 RARC that were performed at our newly established robotics program using a preceptorship model to ascertain the effect of the learning curve that is associated with this operation on important oncologic parameters, including LN yield and margin status. Our data compare favorably with important operative, perioperative, and pathologic characteristics of four large RARC series listed in Table 3. ^{10 –13}

There are no long-term >5-year oncologic outcome data available for RARC. In the interim, as the data mature, surrogates of oncologic efficacy, including LN yield and surgical margin status, have been closely assessed to evaluate for cancer safety in real time. One series, however, recently reported disease-specific survival rates of 82% and 72% at 12 and 36 months, respectively, in 59 patients who underwent RARC with mean follow-up of 25 months (range 6–49 mos), which was comparable to ORC. ¹⁴

Pelvic lymphadenectomy is an integral part of radical cystectomy. Both the adequacy of lymphadenectomy and the number of nodes removed have been the subject of criticism for ORC, laparoscopic radical cystectomy, and RARC. The surgical extent of the dissection reported in RARC series has been either to the common iliac bifurcation (termed standard) or to the aortic bifurcation (extended). Several of the RARC series noted that the LN yield increased with progression of the learning curve. Guru and associates ¹¹ looked at their RARC series and divided 47 consecutive patients into quintiles and noted that as experience increased, so did the mean LN yield—from 13 to 23 from the 1st to 5th quintile. Pruthi and colleagues, ⁷ however, analyzed a single center's first 50 RARC by dividing this case series into quintiles and noted that the mean LN yield in the 1st quintile (mean LN yield 21) did not change by the 5th quintile (mean LN yield 20), thus concluding that the learning curve for RARC did not have an impact on overall LN yield and may be a surgeon-dependent phenomenon. The mean LN yield in our series is comparable to the average LN yield reported in similar contemporary series and did not change as the learning curve progressed, similar to the findings in the series by Pruthi and coworkers. ⁷

The International Robotic Cystectomy Consortium (IRCC) recently reported the results from their prospective multi-institutional database of 527 patients who underwent RARC, revealing a mean LN yield of 17.8 with a total of 82.9% of patients undergoing lymphadenectomy. ¹⁵ There also appeared to be a learning curve associated with the performance of lymphadenectomy. Therefore, it has become increasingly clear that robot-assisted pelvic LN dissection is safe, feasible, and effective when compared with the gold standard open pelvic LN dissection.

It has been reported that the number of LN retrieved during pelvic lymphadenectomy may be influenced by the way the nodes are retrieved (en bloc or packets), because pathologists counted significantly more nodes when sent in packets vs en bloc. ¹⁶ Some of the RARC LN in our series were sent en bloc with the cystectomy specimen, whereas some were sent in packets based on surgeon preference. Certainly, the role of the pathologist is quite important, because interpretation and processing by the pathology team are vital, because it reflects on the number of nodes in a given patient. In our series, the variability in LN yield across tertiles (although not statistically significant) is likely multifactorial in part because of technical factors (this is a multisurgeon series during the learning curve) but also differences in individual pathologists who also could have played a role. Furthermore, in a RARC series that involved two institutions, the average number of LN retrieved using the bifurcation of the aorta as the proximal extent of pelvic LN dissection and submitting the packets separately was 12.3 in 27 patients, which was lower than the authors expected. ¹⁷ Despite identical anatomic boundaries and sending the nodes in packets, however, the mean LN counts were 11.3 at one institution vs 16 at the second institution, thus highlighting the potential for institutional variability among pathologic processing. Therefore, the extent of dissection is probably more important than the number of LN retrieved by any given institution.

The learning curve in our series did not have a statistically significant impact on positive soft tissue surgical margins across tertiles, perhaps being a reflection of inadequate power to detect a difference, because the positive margin rate was highest in the most recent tertile (20%). A total of seven patients had pT₄ bladder cancer on final pathologic evaluation in the 3rd tertile, four of whom had positive margins. Two of these patients had undergone previous radiation therapy for prostate cancer, thus the increased rate of margin positivity in these patients is perhaps an artifact of our increasing comfort level with complex cases using this technique. One of these positive margins was focally positive at the bladder base in a patient who had direct prostatic invasion,and the other patient's final pathologic evaluation was pT₄N₂, thus highlighting the aggressive nature of the disease in these patients with positive margins in our series.

The IRCC also reported the incidence and predictors of positive surgical margins in their database and noted a 39% positive margin rate in patients with pT₄ disease and that the learning curve was not significantly associated with positive margins at RARC. ¹⁸ The increased incidence of positive margins seen in the IRCC database as well as other RARC single-institution series in patients with extravesical disease needes further investigation. The subtle pathologic differences seen across tertiles in this series may have also been related to selection bias; however, selection bias remains a difficult to measure phenomenon when performing this type of analysis because we continued to perform some open cases during this series.

We did note a statistically significant decrease in the rate of all complications at 90 days as the learning curve progressed because 70% of the patients in the 1st tertile sustained any complication compared with 30% in the 2nd and 3rd tertiles (P=0.013). The complication rate in our series compares favorably with the complication rate at 90 days in a recent large series comparing complications from ORC to RARC. ¹⁹ Some of the complications we experienced are unique to robotic surgery and deserve specific mention. An intraoperative air embolus occurred in a patient who sustained an injury to the external iliac vein during PLND. The vein was repaired after converting to an open operation based on surgeon comfort; the patient sustained no long-term effects and was discharged to home on postoperative day 6. Another patient sustained an intraoperative rectal injury that was repaired primarily in two layers. The patient had slow return of her bowel function but was ultimately discharged home on postoperative day 14. Finally, two patients in the first half of our experience sustained lower extremity compartment syndrome managed by fasciotomy. Since then, we routinely reposition the legs after the robotic portion of the case and have not had any problems since making this modification.

Complications after 90 days were also noted, highlighting the importance of close follow-up because most late complications were urinary diversion related. There was not a statistically significant change in late complications across tertiles. There was a nonstatistically significant trend toward lower EBL and shorter overall operative times (P=0.059) in the 2nd and 3rd tertiles compared with the 1st tertile, which may have had an impact on lowering the complication rate in the 2nd and 3rd tertiles, because other studies have shown that EBL and transfusion requirements are highly correlated with complication rates. ^20,21

The learning curve for any given operation remains difficult to define, because it is likely to vary between individual surgeons based on their previous experience and level of skill. The IRCC analyzed 496 patients in their database who underwent RARC by 21 surgeons at 14 institutions around the world by using predetermined cutoff points for various operative and pathologic parameters in an attempt to define the learning curve for RARC. ¹³ Based on these predetermined cutoff points, these investigators determined that by the 30th case, the individual surgeon had reached an acceptable level of proficiency.

From a technical standpoint, most importantly, we recommend meticulous surgical technique, including a thorough pelvic lymphadenectomy around the anatomic boundaries as we have described with complete clearance of all the nodal, adipose, and loose areolar tissue. Having previous experience with robot-assisted prostatectomy and ORC will certainly aid when establishing a RARC program. When this is performed in a high-volume center with an experienced team and a modified mentorship program, the learning curve encountered in a newly established robotics program does not seem to alter surrogate markers of oncologic efficacy, including LN yield and soft tissue margin status, although critical evaluation of positive margins in patients with pT₄ disease will need to be evaluated. In our experience, complications were higher in the initial 20 cases, but after making modifications and gaining experience, our complication rate is now more acceptable. Certainly, having an experienced member of the team serve as a mentor and preceptor for the junior faculty likely improved our results and shortened the learning curve during the development of our robotic cystectomy program. Therefore, institutions with less experience may not expect to achieve these same results during their learning curves. Future studies with long-term cancer-free survival and critical appraisal of long-term complications should be parameters for evaluation to establish the superiority or inferiority of RARC and its true learning curve.

Conclusions

Performing RARC with extended pelvic LN dissection is safe, feasible, and effective when following strict anatomic boundaries and precise surgical technique. This can be accomplished at a high-volume newly established robotic surgery program with an experienced team without compromising operative, postoperative, or pathologic outcomes during the learning curve. Having an experienced robotic surgeon serve as a mentor and preceptor helps facilitate the learning curve for junior faculty while maintaining patient safety.