Ann Surg Oncol DOI 10.1245/s10434-014-3839-7

ORIGINAL ARTICLE – PANCREATIC TUMORS

Defining the Learning Curve for Team-Based Laparoscopic Pancreaticoduodenectomy Paul J. Speicher, MD, Daniel P. Nussbaum, MD, Rebekah R. White, MD, Sabino Zani, MD, Paul J. Mosca, MD, Dan G. Blazer III, MD, Bryan M. Clary, MD, Theodore N. Pappas, MD, Douglas S. Tyler, MD, and Alexander Perez, MD Department of Surgery, Duke University Medical Center, Durham, NC

ABSTRACT Background. The purpose of this study was to define the learning curves for laparoscopic pancreaticoduodenectomy (LPD) with and without laparoscopic reconstruction, using paired surgical teams consisting of advanced laparoscopictrained surgeons and advanced oncologic-trained surgeons. Methods. All patients undergoing PD without vein resection at a single institution were retrospectively analyzed. LPD was introduced by initially focusing on laparoscopic resection followed by open reconstruction (hybrid) for 18 months prior to attempting a totally LPD (TLPD) approach. Cases were compared with Chi square, Fisher’s exact test, and Kruskal–Wallis analysis of variance (ANOVA). Results. Between March 2010 and June 2013, 140 PDs were completed at our institution, of which 56 (40 %) were attempted laparoscopically. In 31/56 procedures we planned to perform only the resection laparoscopically (hybrid), of which 7 (23 %) required premature conversion before completion of resection. Following the first 23 of these hybrid cases, a total of 25 TLPDs have been performed, of which there were no conversions to open. For all LPD, a significant reduction in operative times was identified following the first 10 patients (median 478.5 vs. 430.5 min;

Presented at the 67th Annual Meeting of the Society of Surgical Oncology, Phoenix, AZ, USA, 12–15 March 2014. A video of the presentation of the data in this article presented at the 67th Annual Society of Surgical Oncology Cancer Symposium is available at www.surgonc.org/vm. Ó Society of Surgical Oncology 2014 First Received: 10 February 2014 A. Perez, MD e-mail: [email protected]

p = 0.01), approaching open PD levels. After approximately 50 cases, operative times and estimated blood loss were consistently lower than those for open PD. Conclusions. In our experience of building an LPD program, the initial ten cases represent the biggest hurdle with respect to operative times. For an experienced teaching center using a staged and team-based approach, LPD appears to offer meaningful reductions in operative time and blood loss within the first 50 cases. Pancreaticoduodenectomy (PD) remains one of the most technically demanding operations in surgical oncology. Minimally invasive PD was first described as early as 1994, but two decades later the laparoscopic approach has yet to achieve the growth in popularity that has occurred with the implementation of other advanced laparoscopic procedures.1–6 However, over the last few years a growing body of evidence has emerged suggesting that laparoscopic PD is a feasible alternative to the traditional open approach in appropriately selected patients. Recent small studies have demonstrated that laparoscopic PD can be performed as safely as the traditional open approach, with possibly less blood loss and better lymph node harvest; however, this is often at the expense of longer operative times.7–12 Despite these findings, almost all of these studies have reported on the experience of a single surgeon’s experience. Despite the potential advantages of laparoscopic PD, there has been considerable variability in experience and outcomes among early reports, and very little attention has been focused on the learning curve for this challenging operation.13 The purpose of this study was to define the institutional learning curves for laparoscopic PD with and without laparoscopic reconstruction, using paired surgical teams consisting of advanced laparoscopic-trained surgeons and advanced oncologic-trained surgeons. Our primary objective was to characterize our experience in

P. J. Speicher et al. TABLE 1 Baseline characteristics of all patients undergoing pancreaticoduodenectomy Characteristics

Overall (n = 140) Totally laparoscopic (n = 25) Hybrid approach (n = 31) Open approach (n = 84) p value

Age (years)

65 (58–72)

61 (57–69)

69 (62–72)

64 (58–72)

0.141

Female sex

78 (55.7)

16 (64)

14 (45.2)

48 (57.1)

0.339

Body mass index

25.3 (22.9–29.2)

24 (24–29)

26 (24–29)

25 (22–29)

0.392

Cancer case

107 (77)

20 (83.3)

25 (80.6)

62 (73.8)

0.609

16 (64)

18 (58.1)

38 (45.2)

0.181

9 (36)

14 (45.2)

26 (31)

0.364

Preoperative biliary stent

72 (51.4)

Neoadjuvant therapy

49 (35)

Tumor size (cm)

2.6 (1.8–3.8)

Positive nodal disease

49 (35)

2 (1–4) 11 (44)

2 (2–4)

3 (2–4)

0.613

14 (45.2)

24 (28.6)

0.148

Continuous variables are summarized as median (interquartile range) and compared with one-way ANOVA. Categorical variables reported as proportions (%) and compared with Pearson’s Chi square test ANOVA analysis of variance

developing a laparoscopic PD program, followed by incorporating laparoscopic PD as a teachable procedure into an established surgical training program. METHODS This study was approved by the Duke University Institutional Review Board, and patients were consented for laparoscopic and possible open PD. All patients undergoing PD at a single institution were retrospectively analyzed, and any patient requiring vein resection was excluded from study inclusion. Laparoscopic PD was introduced in March 2010, with the first case performed as laparoscopic resection followed by open reconstruction (hybrid approach). We continued to develop our hybrid experience for 18 months prior to attempting a totally laparoscopic PD (TLPD) in October 2011. Following the introduction of the totally laparoscopic approach, cases were performed as either hybrid or TLPD, based on specific surgeon preferences and patient factors. Only eight procedures were performed as hybrid cases following the introduction of TLPD, and all were planned as hybrid procedures prior to the operation. All procedures were performed by surgical teams consisting of a fellowship-trained minimally invasive surgeon and one of four surgical oncologists experienced in open pancreatic surgery. A total of five surgeons participated in the initial phases of this pilot program and are included in this study. For comparison of baseline characteristics, cases were stratified into three groups based on surgical technique (i.e. open, hybrid, or TLPD) using an intent-to-treat approach in which intended surgical approach was established prior to the start of each procedure. The primary outcome measures for this study were operative time and estimated blood loss (EBL). To examine the laparoscopic learning curve, patients were grouped into cohorts of ten patients, with the exception of the most recent cohort, which consisted of only six patients due to sample size. Consistent with previous

learning curve studies, individual cases were defined as having ‘long’ operative times if the case duration was longer than the longest mean operative time of any 10-patient cohort. Similarly, ‘short’ cases were defined as any procedure with an operative time shorter than the shortest mean operative time for any of the 10-patient cohorts. To assess the independent learning curves for laparoscopic dissection and reconstruction separately, cases were first stratified by hybrid versus TLPD, and cases were then similarly grouped into 10-patient cohorts before assessing operative times. Secondary outcomes included surgical site infection (SSI), readmission, reoperation, pancreatic fistula formation, lymph node retrieval, and surgical margin status. Comparisons of baseline characteristics across groups, as well as postoperative secondary outcomes, were performed using Pearson’s Chi squared test or Fisher’s exact test for categorical variables and one-way analysis of variance (ANOVA) for continuous endpoints. For the primary endpoints of operative time and EBL, means were calculated across successive patient cohorts and compared using Kruskal–Wallis ANOVA. An affirmative decision was made to control for type I error at the level of the comparison, and a p value of \0.05 was considered statistically significant. All analyses were performed using R version 3.0.1 (The R Foundation for Statistical Computing, Vienna, Austria). RESULTS Between March 2010 and June 2013, 140 patients undergoing PD at our institution met study inclusion criteria and were included for analysis, of which 56 (40 %) were attempted laparoscopically. In 31 of these 56 procedures, we planned to perform only the resection portion laparoscopically followed by open reconstruction (hybrid approach). Of these hybrid cases, seven (23 %) required premature conversion to open prior to completion of the

DISCUSSION In our experience of building a laparoscopic PD program, the initial 10 cases appear to represent the biggest

A

550 500

Hybrid PD learning curve

Totally laparoscopic PD learning curve

450 400 350 300

1-10

11-20

21-31

1-10

11-20

21-25

Case strata Hybrid

B

TLPD

Open control (pre-lap)

Open control (current)

600

Operative time (mins)

550 500 450 400 350 300 250 200 1-10

11-20

21-30

31-40

41-50

51-56

Case strata Laparoscopic

Open control (pre-lap)

Open control (current)

C 16000 Estimated blood loss (mL)

laparoscopic resection. Following completion of the first 23 hybrid cases, we transitioned to attempting the totally laparoscopic approach, and a total of 25 TLPDs were performed along with eight additional hybrid cases wherein the hybrid approach was planned a priori due to surgeon preference. All cases that were planned for TLPD were successfully performed totally laparoscopically, and among these cases there were no conversions to open. Baseline characteristics between the open, hybrid, and TLPD cases were similar (Table 1). For laparoscopic PD cases, a significant reduction in operative times (median 478.5 vs. 430.5 min; p = 0.01) was identified after the first 10 patients, with operative times approaching open PD levels following this period. After approximately 50 cases, operative times and EBL were consistently lower than those for open PD (Fig. 1b, c). When stratified into two separate learning curves (laparoscopic dissection and laparoscopic reconstruction), there was a significant decrease following the first 10 hybrid cases, with an even further decrease following the first 10–20 TLPD cases, specifically (Fig. 1a). Comparison across successive patient cohorts revealed an increasing proportion of TLPD cases versus hybrid cases over time, with a total of only three patients in the most recent two cohorts being approached as hybrids rather than TLPD (Table 2). Complication rates were not significantly different across chronologic patient cohorts, although there appeared to be a trend toward decreasing complications for all endpoints over time (Fig. 2). There was one death (1.8 %) within 30 days among the entire laparoscopic experience, and two deaths (3.6 %) within 90 days. The full breakdown of postoperative endpoints and complications are shown in Table 3. In specifically comparing our TLPD experience with open PD, the totally laparoscopic approach was associated with less EBL (200 vs. 500 ml; p \ 0.001), shorter length of stay (8.5 vs. 12 days; p = 0.04), and increased lymph node retrieval (14.5 vs. 9; p = 0.008). There were no significant differences in margin status, SSI, reoperation, readmission, and 30- or 90-day mortality rates between the three surgical approaches. There were differences between the three groups with respect to ISGPF Grade C fistulae and SSIs, with significantly higher rates of both complications among the hybrid group. However, when excluding the hybrid cases that represented the steepest portion of our early learning curve and comparing only the TLPD cohort to our open controls, there were no significant differences between the two groups for either endpoint (p = 0.13 and 0.99, respectively).

Operative time (mins)

Defining the Learning Curve

14000 12000 10000 800 600 400 200 0 1-10

11-20

21-30

31-40

41-50

51-56

Case strata Laparoscopic

Open control (pre-lap)

Open control (current)

FIG. 1 a Operative times for patients undergoing laparoscopic PD shown in successive 10-patient cohorts, stratified by hybrid versus TLPD. b Operative times for patients undergoing laparoscopic PD shown in successive 10-patient cohorts, regardless of TLPD or hybrid. c Estimated blood loss for patients undergoing laparoscopic PD shown in successive 10-patient cohorts. TLPD totally laparoscopic PD, PD pancreaticoduodenectomy

hurdle to overcome with respect to operative times, after which operative times appear similar to historical open cases. For an experienced teaching center using a staged and team-based method, our results suggest that laparoscopic PD may offer meaningful reductions in operative time and blood loss within 50 cases compared with the traditional open approach. The use of a hybrid approach in

P. J. Speicher et al. TABLE 2 Operative time and other procedure-specific characteristics for patients undergoing LPD in successive 10-patient cohorts Patient cohort

Hybrid casesa

Median OR time (min)

Range (IQR, min)

Long cases

Short cases

EBL (ml)

Total nodes retrieved

Positive margins

1–10

10

478.5

467, 517.2

5

1

400

13

2

11–20

10

430.5

373.2, 440.5

2

4

550

16

1

21–30

4

413.5

389.5, 464.2

2

2

512.5

11.5

1

31–40

4

366.5

311.5, 442

2

6

350

22.5

1

41–50

2

429.5

378, 514

4

2

250

13.5

2

51–56

1

373.5

348.8, 403.5

0

3

250

17

1

LPD laparoscopic pancreatoduodenectomy, OR odds ratio, IQR interquartile range, EBL estimated blood loss, TLPD totally laparoscopic pancreatoduodenectomy a

Hybrid cases defined as laparoscopic resection and open reconstruction; all other cases were TLPDs. No TLPD cases required conversion to open

FIG. 2 Postoperative complication rates across successive cohorts along the learning curve

Complication rate

75%

50%

p-value

Complication

25%

Surgical site infection

0.74

Readmission

0.85

Positive margins

0.99

Grade C fistula

0.76

Reoperation

0.90

0% 11-20

1-10

21-30

31-40

41-50

51-56

Case strata

TABLE 3 Breakdown of standard outcomes by surgical approach Variable

Overall (n = 140)

Operative time (min)

420.5 (360–481)

Estimated blood loss (ml)

400 (250–662.5)

Total nodes retrieved

13 (8–17.5)

Positive margins

26 (18.7)

Totally laparoscopic (n = 25)

Hybrid approach (n = 31)

381 (342–465)

442 (386.5–486.5)

200 (100–425)

600 (312.5–700)

14.5 (9.8–21.5)

15 (10.5–19)

Open approach (n = 84) 425.5 (345.8–478.8) 425 (300–700) 12 (8–16)

0.065

4 (12.9)

18 (21.4)

0.590 0.642

2 (1.4)

0 (0)

1 (3.2)

1 (1.2)

Mortality (90 days)

7 (5)

1 (4)

1 (3.2)

5 (6)

Type C leak

10 (8–14.5) 27 (19.3)

8.5 (7–11.2) 2 (8)

7 (5)

0.234 \0.001

4 (16.7)

Mortality (30 days) Length of stay (days) Type B leak

p value

12 (8.5–18.5) 7 (22.6)

10 (8–14) 18 (21.4)

0.999 0.134 0.285

2 (8)

4 (12.9)

1 (1.2)

0.016

Surgical site infection

58 (41.4)

9 (36)

20 (64.5)

29 (34.5)

0.013

Reoperation

15 (10.9)

2 (8.7)

4 (12.9)

9 (10.7)

0.928

Readmission

51 (37)

7 (30.4)

11 (35.5)

33 (39.3)

0.725

Total costs ($)

21,560

24,590

23,230

19,720

0.189

Direct costs ($)

16,970

19,250

17,660

16,790

0.404

Data are expressed as median (interquartile range) or n (%)

Defining the Learning Curve

Expert • Pancreatic anastamosis

Advanced

Increasing difficulty

our initial experience was critical in evolving along the learning curve and preparing for the safe completion of TLPD. Using this staged approach allowed us to efficiently and safely implement a laparoscopic PD program, and to examine the independent learning curves for the resection and reconstruction components of the procedure. Overall, our results are consistent with recently published studies, in that the laparoscopic approach was associated with less blood loss, higher numbers of lymph nodes retrieved, and no substantial differences in postoperative morbidity.7–13 Our early experience also confirmed other studies in the literature, revealing initial operative times that exceeded those of our open controls, both historically and concurrent with the study period. However, over time we realized meaningful reductions in operative time, and have found similarities between our laparoscopic PD learning curve and reports for that of other technically demanding operations, such as minimally invasive esophagectomy.14 We believe that the team-based approach utilized in the present study allows for more generalizable expectations regarding the early phases of implementing a laparoscopic PD program as our results are not limited to a single surgeon’s skill set, practice, and individual learning curve. While trainees are now integrally involved in most procedures, we have continued with our strategy of having two surgeons—one minimally-invasive surgeon and one pancreas surgeon— involved in each operation to maximize patient safety in the event of conversion to open. The traditional surgical learning curve is described as having three main portions: a slow and potentially arduous beginning, followed by a steep acceleration phase of rapid learning, and finally a plateau involving slower but continued improvements. With respect to mastering the technical details of laparoscopic PD as a two-surgeon operation, following the first 50 operations we have likely moved past the steep acceleration phase, but acknowledge that we have most likely not reached the true plateau of the learning curve. Following these first 50 cases, we now see operative times and blood loss that are consistently better than our historic and concurrent open controls. Importantly, we also found that once we developed proficiency in the laparoscopic dissection during our initial hybrid experience, and established efficient team dynamics, applying these skills to the reconstruction and anastomoses was not as challenging as we anticipated it would be, likely due to the skill set of the minimally invasive-trained surgeon who was already facile with intracorporeal suturing. Our experience was also influenced by the progressive introduction of trainees into the hybrid approach, and then a similar introduction into the TLPD approach after accumulating further experience, which may explain the shouldering seen following 20 hybrid cases in Fig. 1a.

• Dissection and division of GDA • Dissection and division of CBD • Division of pancreatic body • Dissection of uncinate • Biliary anastamosis • Gastrojejunostomy

Intermediate • Extended Kocher maneuver • Lesser sac dissection • Definition of SMV tunnel • Division at distal duodenum

Beginner • Trocar insertion • Takedown of ligament of Treitz • Division of stomach • Gallbladder dissection • Specimen extraction • Wound closure

GDA: gastroduodenal artery; CBD: common bile duct; SMV: superior mesentric vein

FIG. 3 Stages of learning laparoscopic pancreaticoduodenectomy at our institution, stratified into phases of difficulty—beginner through expert

While we have shown data in support of the feasibility of implementing a laparoscopic PD program, for the approach to be widely accepted and sustainable the technique must not only be mastered but must also be teachable. Because of this, we have transitioned from a learning curve centered on optimizing operative efficiency to one now focused more on trainee education. To implement laparoscopic PD as part of our surgical training program, we have created a phased approach of introducing trainees to the various aspects of the procedure. We have divided the procedure into a series of smaller steps, each of which we have classified into one of four groups ranging from beginner level to expert (Fig. 3). Trainees are then introduced to each of the steps in a progressive, stepwise fashion, starting with some of the less technical aspects of the resection and culminating in the pancreaticojejunostomy anastomosis. Each step must be learned sequentially and, prior to advancing to the next phase, learners must demonstrate proficiency. Doing this allows for a safe and personalized approach to teaching laparoscopic PD. Although we believe our results are likely generalizable to other experienced teaching institutions, our study nonetheless has some limitations that must be acknowledged.

P. J. Speicher et al.

First and foremost, we have described the experience of a single high-volume referral center. Second, as this was a retrospective review, selection bias may have played a role, particularly in our early experience, with less challenging cases being selected for a laparoscopic approach rather than open. While two of the surgeons instituted no selection criteria and planned for all operative candidates to be approached laparoscopically, the remaining surgeons were somewhat more selective initially. However, it is our assertion that such patient selection may be necessary and is not inappropriate in the early stages of implementing a new technique, and that the learning curve will simply reflect this fact. Finally, we make no claims to having reached the pinnacle of laparoscopic PD performance, nor to have fully defined the learning curve, and acknowledge that as we transition to teaching this technique, new challenges and learning opportunities will arise. CONCLUSIONS The results of this study suggest that laparoscopy is a safe and learnable approach to PD when properly implemented. While this technique holds substantial promise, implementing a laparoscopic PD program presents unique challenges, particularly during the first 10 cases. It is our belief that for laparoscopic PD to become an accepted and widely utilized approach to managing pancreatic disease, the technique must be both reproducible and teachable across different levels of training and experience. Disclosure Paul J. Speicher, Daniel P. Nussbaum, Rebekah R. White, Sabino Zani, Paul J. Mosca, Dan G. Blazer III, Bryan M. Clary, Theodore N. Pappas, Douglas S. Tyler, and Alexander Perez report no relevant conflicts of interest.

REFERENCES 1. Gagner M, Pomp A. Laparoscopic pylorus-preserving pancreatoduodenectomy. Surg Endosc. 1994;8:408–10.

2. Fleshman J, Sargent DJ, Green E, Anvari M, Stryker SJ, Beart RW, et al. Laparoscopic colectomy for cancer is not inferior to open surgery based on 5-year data from the COST Study Group trial. Ann Surg. 2007;246:655–62; discussion 62-4. 3. Frazee RC, Roberts JW, Symmonds RE, Snyder SK, Hendricks JC, Smith RW, et al. A prospective randomized trial comparing open versus laparoscopic appendectomy. Ann Surg. 1994;219: 725–8; discussion 8-31. 4. Clinical Outcomes of Surgical Therapy Study Group. A comparison of laparoscopically assisted and open colectomy for colon cancer. N Engl J Med. 2004;350:2050–9. 5. Lacy AM, Delgado S, Castells A, Prins Ha, Arroyo V, Ibarzabal A, et al. (2008) The long-term results of a randomized clinical trial of laparoscopy-assisted versus open surgery for colon cancer. Ann Surg. 248:1–7. 6. Stoker DL, Spiegelhalter DJ, Singh R, Wellwood JM. Laparoscopic versus open inguinal hernia repair: randomised prospective trial. Lancet. 1994;343:1243–5. 7. Asbun HJ, Stauffer JA. Laparoscopic vs open pancreaticoduodenectomy: overall outcomes and severity of complications using the accordion severity grading system. J Am Coll Surg. 2012; 215:810–9. 8. Buchs NC, Addeo P, Bianco FM, Ayloo S, Benedetti E, Giulianotti PC. Robotic versus open pancreaticoduodenectomy: a comparative study at a single institution. World J Surg. 2011; 35:2739–46. 9. Chalikonda S, Aguilar-Saavedra JR, Walsh RM. Laparoscopic robotic-assisted pancreaticoduodenectomy: a case-matched comparison with open resection. Surg Endosc. 2012;26:2397– 402. 10. Lai ECH, Yang GPC, Tang CN. Robot-assisted laparoscopic pancreaticoduodenectomy versus open pancreaticoduodenectomy: a comparative study. Int J Surg. 2012;10:475–9. 11. Zhou N-X, Chen J-Z, Liu Q, Zhang X, Wang Z, Ren S, et al. Outcomes of pancreatoduodenectomy with robotic surgery versus open surgery. Int J Med Robot. 2011;7:131–7. 12. Zureikat AH, Breaux JA, Steel JL, Hughes SJ. Can laparoscopic pancreaticoduodenectomy be safely implemented? J Gastrointest Surg. 2011;15:1151–7. 13. Correa-Gallego C, Dinkelspiel HE, Sulimanoff I, Fisher S, Vin˜uela EF, Kingham TP, et al. Minimally-invasive vs open pancreaticoduodenectomy: systematic review and meta-analysis. J Am Coll Surg. 2014;218:129–39. 14. Hernandez JM, Dimou F, Weber J, Almhanna K, Hoffe S, Shridhar R, et al. Defining the learning curve for robotic-assisted esophagogastrectomy. J Gastrointest Surg. 2013;17:1346–51.

Defining the learning curve for team-based laparoscopic pancreaticoduodenectomy.

The purpose of this study was to define the learning curves for laparoscopic pancreaticoduodenectomy (LPD) with and without laparoscopic reconstructio...
296KB Sizes 0 Downloads 3 Views