Robotic and open distal pancreatectomy with celiac axis resection for locally advanced pancreatic body tumors: a single institutional assessment of perioperative outcomes and survival

Abstract

Background

Distal pancreatectomy with celiac axis resection (DP-CAR) is an option for T4 tumors of the pancreatic body. We examined the perioperative and oncologic outcomes of open and robotic DP-CAR at a high-volume pancreatic center.

Methods

Retrospective review of all consecutive DP-CARs. Patient demographics, 90-day perioperative outcomes, and disease specific survival were collected.

Results

30 DP-CARs were performed (11 Robotic, 19 Open). Both groups had similar preoperative/tumor characteristics, and 27 of 28 PDA patients received neoadjuvant chemotherapy. Robotic DP-CAR was associated with decreased OT (316 vs. 476 min), reduced EBL (393 vs. 1736 ml) and lower rates of blood transfusion (0% vs. 54%) (all p < 0.05). No robotic DP-CAR required conversion. Both groups had similar rates of 90-day mortality, major morbidity, LOS, readmission, and receipt of adjuvant therapy. Similarly, both approaches were associated with high R0 resection rates (82% vs. 79%). At a median follow-up of 33 months, median overall survival for the PDA cohort was 35 months, with no difference in the robotic and open approach (33 and 40 months, p = 0.310).

Conclusions

With a median survival approaching 3 years, DP-CAR represents an effective treatment for select patients with locally advanced pancreatic body cancer, regardless of approach.

Introduction

Surgery remains the mainstay for cure in pancreatic ductal adenocarcinoma (PDA). Since only 9% of patients present with localized resectable disease,^{1, 2} efforts have focused on down-staging advanced disease with neoadjuvant therapy in hopes of offering an increasing proportion of patients the potential benefit of resection.^{3, 4} Neoadjuvant chemotherapy, with or without radiation, can serve to increase the R0 resection potential, sterilize regional lymph node basins, and treat micrometastatic disease.^{5, 6, 7, 8, 9} Such multimodality therapy in borderline resectable head PDA, when followed by margin-negative resection and adjuvant therapy, allows select patients to achieve significant improvements in survival, and spares those with occult metastatic disease and aggressive tumor biology the morbidity of a large resection.¹⁰

The anatomical and operative considerations for pancreatic body tumors are different from lesions of the pancreatic head, but the objectives of multimodality therapy are similar. Accordingly, an aggressive approach should be undertaken with locally advanced T4 tumors of the pancreatic body, provided the benefit of increased survival is not outweighed by substantial increases in perioperative morbidity and mortality. In 1953, Appleby described the en bloc resection of the celiac axis with a total gastrectomy and distal pancreatectomy (DP) for the treatment of locally advanced gastric cancer.¹¹ A modified version of this procedure for the treatment of locally advanced PDA omits the gastrectomy, and consists of Distal Pancreatectomy with en bloc Celiac Axis Resection (DP-CAR). Since then, sporadic reports on the safety and feasibility of DP-CAR have emerged, albeit in small numbers and in the absence of long-term data on survival.^{12, 13, 14, 15, 16, 17}

Recent reports have highlighted a potential increase in morbidity and mortality following DP-CAR.¹⁸ In an attempt to curb the postoperative morbidity of pancreatic resections, we have expanded the application of minimally invasive approaches to include resections of locally advances pancreatic body/tail tumors.^{19, 20} Since minimally invasive DP harbors potential benefits to its open counterpart,^{21, 22} we postulated that minimally invasive DP-CAR might be associated with improved outcomes compared to the open approach. We therefore report here a large experience of robotic-assisted DP-CAR, and compare its perioperative and oncologic outcomes to the open approach. Additionally, this report represents one of the few longitudinal assessments of survival for resected locally advanced pancreatic body tumors after multimodality therapy.

Materials and methods

After obtaining approval from the Institutional Review Board all patients who underwent DP-CAR at the University of Pittsburgh Medical Center between July 1, 2007 and August 31, 2015 were identified from a prospective database. Health Insurance Portability and Accountability Act compliance was assured.

Case selection/patient eligibility

The following criteria were used for determining eligibility for DP-CAR at our institution: (i) presence of a T4 pancreatic body/tail tumor invading any of the branches of the celiac artery without involvement of the celiac trunk; (ii) no involvement of the GDA; and (iii) in the case of PDA, receipt of neoadjuvant therapy. Neoadjuvant therapy is not given to downstage the tumor per se, but to allow for the treatment – and potential declaration – of occult micrometastatic disease prior to committing patients to a DP-CAR. Since all of these cases were treated outside the context of clinical trials, decisions regarding the nature of the neoadjuvant chemotherapy regimen used, its duration, and the timing of surgery were made at a weekly multidisciplinary pancreatic cancer meeting. The operative approach (open vs. robotic) is left to the discretion of the surgeon, since some surgeons in our hospital system are more familiar with the robotic platform than others. Notably, in an effort to ensure patient safety, our first robotic DP-CAR was performed in 2011, nearly 3 years after implementation of the robotic hepatopancreaticobiliary surgery program at UPMC. This ensured that the learning curve for complex robotic pancreatic resections – such as the pancreaticoduodenectomy and distal pancreatectomy – was overcome prior to performing the DP-CAR.²⁰

Patient variables and definitions

Cases were assessed for patient, perioperative, and oncologic variables. Patient variables included age, sex, Charlson Comorbidity Index (CCI) score, Body Mass Index (BMI), prior abdominal surgery, receipt of neoadjuvant therapy, and preoperative tumor size (in cm) based on cross-sectional imaging and/or endoscopic ultrasound. Perioperative data examined included operative time, estimated blood loss and rate of blood transfusion. Complications were defined and graded according to the Clavien-Dindo classification,²³ and rate and severity of postoperative pancreatic fistulae (POPF) was determined according to the criteria established by the International Study Group for Pancreatic Fistula (ISGPF).²⁴ All postoperative outcomes were followed to 90 days. Pathologic variables included tumor histology and size, the presence of lymphovascular or perineural invasion, lymph node harvest and metastases, and margin status and clearance (in mm). A margin-negative resection (R0) was identified as no tumor at the inked margin regardless of distance. Adjuvant therapy receipt and timing was recorded. Overall survival was defined from the date of diagnosis to date of death or last censored visit.

Operative technique

Our approach to open DP-CAR has been previously described.¹⁷ Typical port placement for the robotic DP-CAR is depicted in Fig. 1a. The procedure begins with standard diagnostic laparoscopy. Using conventional laparoscopic instruments, the lesser sac is opened (taking care to preserve the right gastroepiploic arcade) up to the angle of His and the left diaphragmatic crus. The robot is docked for the remainder of the case. The robotic DP-CAR dissection has four phases that allow safe circumferential dissection of the celiac trunk (see Supplemental Video):

• 1)Medial dissection: The common hepatic artery (CHA) is exposed at the superior border of the pancreas and traced distally to expose the takeoff of the gastroduodenal artery (GDA). The CHA is test-occluded with a laparoscopic bulldog clamp. Under ultrasound guidance, adequate collateral arterial blood flow to the proper hepatic artery via the GDA is confirmed. The retropancreatic neck tunnel is created and the pancreatic neck is transected using a linear stapler (Fig. 1b). The CHA is then transected using a linear vascular stapler, ensuring that the takeoff of the GDA is not compromised. The splenic vein is then transected using a stapler (Fig. 1c). Dissection continues posteriorly until the anterior surface of the superior mesenteric artery (SMA) is identified. The SMA is traced proximally to the aorta, invariably allowing for the identification of the celiac trunk superiorly. Frequent use of the robotic ultrasound probe is necessary here to confirm the origins of both the SMA and celiac axis (see “bunny ear” configuration in Fig. 1d).
• 2)Anterior dissection: The CHA is traced back to the celiac axis along the superior border of the pancreas. The left gastric artery and vein are transected distal to their takeoff from the celiac trunk using linear staplers, and a complete lymphadenectomy is performed on the superior and right side of the celiac trunk.
• 3)Lateral dissection: The splenic flexure is lowered by dividing the spleno-colic attachments, and the entire spleno-pancreatic complex is lifted off the retroperitoneum after division of the splenorenal and spleno-diaphragmatic ligaments. The dissection proceeds from lateral to medial taking the retroperitoneal fascia en-block until the lateral (left) wall of the celiac trunk is reached. Here a complete lymphadenectomy of the left side of the celiac trunk is performed.
• 4)Transection of celiac trunk: With the celiac artery completely isolated, it is transected at its origin using the endovascular linear stapler (Fig. 1e). The specimen is retrieved using a specimen retrieval bag, and a 19-French round, fluted closed suction drain is left in the pancreatic resection bed.

Figure 1.

(a) Standard port placement for robotic DP-CAR. (b) Vascular control and stapled transection of the common hepatic artery. (c) Vascular control and stapled transection of the splenic vein just proximal to the junction of the superior mesenteric vein. (d) Robotic-assisted intraoperative ultrasound demonstrating the junction of the aorta, celiac axis, and superior mesenteric artery. (e) Stapled transection of the celiac artery

Supplemental Video related to this article can be found online at http://dx.doi.org/10.1016/j.hpb.2016.05.003

The following are the Supplemental Video related to this article:

Supplemental Video

Robotic distal pancreatectomy with celiac axis resection

Statistical analysis

IBM SPSS 21.0 software for Macintosh was used for all data analysis (IBM, Armonk, NY). Variables were assessed for normal distribution using the Shapiro–Wilk test. Continuous variables were compared using the independent samples t-test or Mann–Whitney U test as appropriate. Categorical data were compared using the Pearson Chi-squared test or Fisher's exact test as appropriate. Median overall survival was determined using the Kaplan–Meier method and Log-Rank test. A p value of <0.05 was considered statistically significant.

Results

Patient characteristics

A total of 30 patients underwent DP-CAR during the 8-year study period (open = 19, robotic = 11). All 11 robotic DP-CARs were performed after 2011, whereas 11 of the 19 open cases were performed prior to 2011 and 8 were performed from 2011 to 2015 (p = 0.002). Baseline patient demographics are presented in Table 1. The groups were similar with respect to age, sex, BMI, age-adjusted CCI score, and preoperative tumor size on CT scan or EUS, and had similar rates of prior abdominal surgery. All but one patient undergoing DP-CAR for PDA (n = 28) received neoadjuvant chemotherapy (21 = gemcitabine-based regimen, 5 = FOLFIRINOX, 1 = both gemcitabine and FOLFIRINOX sequentially, 1 = refused neoadjuvant chemotherapy). Overall, median time from PDA diagnosis to DP-CAR was 4.6 months, and was similar between the robotic and open cohorts (5.1 vs. 4.5 months, p = 0.92).

Table 1.

Patient demographics of robotic-assisted and open DP-CAR

Variable	Overall	R-DP-CAR	O-DP-CAR	P
Total patients, n (%)	30 (100)	11 (37)	19 (63)
Age (years)	61.9 ± 11.3	63.5 ± 15.0	62.2 ± 9.6	0.783
Sex, n (%)
Male	17 (57)	6 (55)	11 (58)	0.858
Female	13 (43)	5 (45)	8 (42)
BMI (kg/m2)	26.1 (24.4–30.0)	28.1 (24.5–30.3)	26.0 (23.6–28.4)	0.328
Age-adjusted CCI score	4.8 ± 1.6	4.6 ± 2.2	5.0 ± 1.3	0.480
Tumor size (cm)
CT	4.0 ± 1.8	3.9 ± 2.3	4.1 ± 2.1	0.768
EUS	3.3 ± 0.9	3.1 ± 1.1	3.5 ± 1.0	0.287
Prior abdominal surgery, n (%)	18 (60)	7 (64)	11 (58)	1.000
Neoadjuvant chemotherapya, n (%)	27 (96)	11 (100)	16 (94)	0.413
Time from Diagnosis to DP-CARa, months	4.6 (3.6–6.1)	5.1 (3.1–7.0)	4.5 (3.8–5.8)	1.000

Normally distributed variables are expressed as mean ± SD; otherwise median (25th percentile–75th percentile) as IQR. R-DP-CAR, robotic DP-CAR; O-DP-CAR, open DP-CAR; BMI, Body Mass Index; CCI, Charlson Comorbidity Index.

^aFor PDA patients only (n = 28).

Operative variables and postoperative complications

Perioperative data and morbidity are presented in Table 2. Robotic DP-CAR was associated with shorter operative time (316 vs. 476 min; p = 0.005), less blood loss (393 vs. 1736 ml; p = 0.010), and lower rates of blood transfusion (0% vs. 54%; p = 0.006). There was no difference between the groups in rate of overall complications, major (Clavien-Dindo grade 3–4) complications, or POPF (overall and ISGPF Grade B/C). Major complications in the open group consisted of pneumothorax (n = 1), respiratory failure requiring tracheostomy (n = 1), and intra-abdominal abscess/sepsis (n = 4). Major complications in the robotic group included gastrointestinal bleed (n = 1), gastric ischemia/perforation (n = 2), and fluid collection requiring drainage (n = 1). There were no robotic conversions to open. The postoperative mean length of stay and 90-day readmission rate was similar between the two groups.

Table 2.

Intraoperative and perioperative outcomes for patients undergoing DP-CAR

Variable	Overall	R-DP-CAR	O-DP-CAR	P
Operative time (min)	430.8 ± 229.9	315.6 ± 74.4	476.2 ± 163.1	0.005
Concomitant vein resection, n (%)	14 (47)	4 (36)	10 (53)	0.466
Tangential, n (%)	8 (57)	4 (100)	4 (40)
Segmental, n (%)	6 (43)	0 (0)	6 (60)
Estimated blood loss (ml)	1552.5 ± 1565.6	392.7 ± 277.3	1735.9 ± 1561.5	0.010
Estimated blood loss >500 ml, n (%)	15 (56)	2 (18)	13 (81)	0.002
Blood transfusion, n (%)	7 (29)	0 (0)	7 (54)	0.006
Reoperation, n (%)	2 (7)	0 (0)	2 (11)	0.520
Overall complications, n (%)	22 (73)	8 (73)	14 (73)	1.000
Major complications, n (%)	10 (35)	4 (40)	6 (32)	0.698
Overall POPF, n (%)	13 (43)	4 (36)	9 (47)	0.708
Grade B/C POPF, n (%)	6 (20)	2 (18)	4 (21)	0.626
90-day mortality, n (%)	4 (14)	0 (0)	4 (22)	0.265
Length of stay (days)	10.7 ± 6.5	10.4 ± 5.9	10.8 ± 7.0	0.851
90-day readmission, n (%)	16 (57)	6 (67)	10 (53)	0.687

Normally distributed variables are expressed as mean ± SD; otherwise median (25th percentile–75th percentile) as IQR. POPF, Postoperative pancreatic fistula.

Overall, there were 4 mortalities within 90 days of resection. These were (i) postoperative day (POD) 54 from bleeding (secondary to anticoagulation for a pulmonary embolus) in the setting of a Grade C POPF; (ii) POD 73 from a small bowel obstruction managed at an outside hospital; (iii) POD 52 from gastric necrosis and multi-system organ failure; (iv) POD 24 from Gram-negative sepsis and multi-system failure related to an infected intra-abdominal fluid collection, likely a Grade C POPF.

Oncologic outcomes and survival for PDA patients

Oncologic data for the PDA cohort are presented in Table 3. Although both groups had similarly high rates of R0 resection, robotic DP-CAR was associated with a greater total lymph node yield. Adjuvant therapy was administered to the majority of patients in both groups at a similar frequency and in a similar postoperative time frame. Follow-up time was significantly longer in patients who underwent open DP-CAR (36 vs. 20 months; p = 0.046) (Table 3). Median overall survival for the entire cohort was 35 months (Fig. 2a), with no difference in survival between the groups (robotic = 33 vs. open = 40 months; p = 0.310) (Fig. 2b). Regardless of approach, margin-negative resection was associated with significantly longer overall survival compared to margin-positive resection (40 vs. 15 months; p = 0.011) (Fig. 3).

Table 3.

Oncologic outcomes for patients undergoing DP-CAR

Variable	Overall	R-DP-CAR	O-DP-CAR	P
Tumor histology, n (%)
Adenocarcinoma	28 (93)	11 (100)	17 (89)	0.520
Neuroendocrine	2 (7)	0 (0)	2 (11)
Tumor size (cm)	4.2 ± 2.0	3.9 ± 1.9	4.4 ± 2.1	0.521
Lymphovascular invasion, n (%)a	16 (62)	7 (64)	9 (60)	1.000
Perineural invasion, n (%)a	24 (86)	10 (91)	14 (82)	1.000
Lymph node positive, n (%)a	14 (50)	7 (64)	7 (41)	0.440
Total lymph node harvest	23.9 ± 18.0	33.9 ± 21.2	18.1 ± 13.3	0.018
R0 resection margin, n (%)	24 (80)	9 (82)	15 (79)	1.000
Margin clearance >1 mm, n (%)b	14 (52)	4 (40)	10 (59)	0.440
Adjuvant therapy, n (%)a	18 (82)	7 (78)	11 (85)	1.000
Weeks from surgery	9.5 (8.0–15.0)	10.0 (6.4–12.0)	9.0 (8.0–16.0)	0.487
Follow-up (months)a	33.4 ± 18.4	20.2 ± 10.5	35.8 ± 20.1	0.046

Normally distributed variables are expressed as mean ± SD; otherwise median (25th percentile–75th percentile) as IQR.

^aNeuroendocrine excluded.

^bWhen Margin distance was reported.

Figure 2.

Kaplan–Meier estimates for overall survival following DP-CAR. (a) Overall survival following DP-CAR for the entire cohort. (b) Overall survival following robotic or open DP-CAR

Figure 3.

Kaplan–Meier estimates by margin status (R0/R1) for overall survival following DP-CAR. R0 margin status is defined as no tumor at the inked margin, regardless of distance

Discussion

The purpose of this study was to examine the perioperative and oncologic outcomes following robotic and open DP-CAR, and to evaluate the survival of this carefully selected surgical cohort within the context of multimodality therapy. The robotic approach was associated with decreased operative time, blood loss, and blood transfusions. There was no difference in postoperative complications, pancreatic fistulae, 90-day mortality, hospital readmission, or length of stay. Importantly, in patients with PDA, both approaches were associated with an overall survival of nearly 3 years.

Currently, the role of minimally invasive approaches for borderline resectable and locally advanced pancreatic cancer remains controversial. Data are limited to a handful of single-institutional retrospective studies.^{25, 26, 27, 28} This report represents the first comparison of open vs. minimally invasive DP-CAR, and indicates that in the hands of experienced pancreatic surgeons, both approaches are feasible, with acceptable morbidity that can be justified by the substantial improvement in survival for this subset of patients. Typically, median survival for locally advanced PDA is 9–11 months.^{1, 29} The use of effective modern day chemotherapeutic regimens as definitive therapy – although not formally tested in this disease subgroup – may marginally increase survival. This report however, demonstrates that the combination of neoadjuvant therapy, margin-negative resection, and adjuvant therapy can result in prolonged survival in carefully selected patients with T4 pancreatic body adenocarcinoma.

It is important to emphasize that careful patient selection is paramount in this subset of advanced disease. DP-CAR for T4 tumors should likely only be performed in the context of neoadjuvant therapy that serves to spare patients with early metastatic progression from the potential morbidity of the procedure. Although CA19-9 response can guide decision making after neoadjuvant therapy in borderline resectable head PDA,³⁰ we did not use it as an exclusion criterion for DP-CAR. Interestingly, 13 patients (43%) had normal pre-neoadjuvant CA19-9 levels, and of the remaining patients with elevated pre-neoadjuvant CA19-9, all demonstrated a reduction of at least 57% (data not shown). Since this series only includes resected patients, the larger denominator of T4 body/tail PDA patients treated with ‘intent to resect’ is not known. It remains to be seen whether CA19-9 response (or lack thereof) can be more reliably used to select patients for DP-CAR.

Since pancreatic cancer carries a dismal prognosis, a careful assessment of the risk-benefit ratio should be considered before embarking on surgical resection for locally advanced disease. A recent publication by Beane et al. used the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database to compare outcomes of DP-CAR to standard distal pancreatectomy (without celiac axis resection) between 2011 and 2012.¹⁸ The authors noted that DP-CAR was associated with a higher rate of acute kidney injury (10% vs. 1%) and 30-day mortality (10% vs. 1%).

With a major complication rate of 35%, our report confirms that DP-CAR remains a morbid operation. The increased mortality however in the report by Beane et al., deserves further scrutiny. Although 46 centers contributed data on the 822 patients included in the study, it took 16 centers to provide data on the 20 patients that underwent DP-CAR, indicating that each center provided only a single case on average. Moreover, 6 of the 43 participating institutions were not considered high-volume centers of pancreatic surgery, and it was not possible to determine if a high volume or low volume institution contributed any of the DP-CAR cases. Since oncologic dissection of the celiac and SMA trunks is a rarely performed procedure, most outcomes within the Beane et al. report potentially reflect surgeons working through their DP-CAR learning curve. This learning curve effect is suggested in our own series. We divided the cohort into an early (2007–2010; n = 11) vs. late (2011–2015; n = 19) experience. DP-CAR in the latter cohort was associated with fewer reoperations (n = 0 vs. n = 2; p = 0.054), and a reduction in postoperative mortality (n = 1 vs. n = 3; p = 0.077). Although not statistically significant likely due to small sample size, the trend suggests that catastrophic outcomes are reduced with increased surgeon experience. There was no difference in the rate of overall complications, major complications, POPF, length of stay, or hospital readmission (data not shown).

The robotic platform may harbor some advantages compared to the open approach. We witnessed a substantial reduction in operative time (>120 min), coupled to reduced blood loss and transfusion rates in the robotic cohort. These parameters have important implications on the rate of postoperative complications and survival, but such benefits may only be realized in larger series. The average robotic operative time of 5 h is only slightly longer than our published operative times for pure robotic DP, and published reports of laparoscopic DP, and is shorter than the largest available open DP-CAR reports [Hirano et al. (532 min) and Tanaka et al. (478 min)].^{12, 13, 19, 21, 22, 31} The reasons for the reduced operative time in the robotic cohort compared to our open cohort are likely multifactorial. The magnified caudal-to-cephalad view of the retroperitoneal field allows for safe and efficient dissection of the structures surrounding the celiac trunk. The ergonomics of the robotic bipolar and unipolar instruments allows for expeditious yet hemostatic dissection of the lymphatics surrounding the aorta, celiac and SMA trunks, which are sources of significant trepidation and bleeding in the open approach. Finally, and most importantly, is the open DP-CAR learning curve (2007–2010), which allowed us to implement robotic DP-CAR between 2011 and 1015 more efficiently.

This report is limited by its retrospective nature and small sample size. The impressive 3-year survival for both cohorts is undoubtedly a surrogate of good case selection, and does not reflect the natural history of all locally advanced pancreatic body/tail PDA tumors. Placing the results of this report within the context of such a large denominator is required to verify the significance of this survival advantage. Additionally, despite the homogeneity in patient and tumor characteristics between both cohorts, the operative advantages witnessed in the robotic group may have been a result of the experience garnered during the earlier open cases, and likely selection bias. Furthermore, we emphasize that the results observed in this report may not be generalizable to other centers and surgeons. We performed robotic DP-CAR only after amassing a 3-year experience of robotic DP and robotic pancreaticoduodenectomy, and would caution attempting this approach by surgeons early in their robotic hepatopancreaticobiliary surgery experience.

In summary, this series demonstrates that the robotic approach to DP-CAR is an acceptable alternative to the open technique in carefully selected locally advanced pancreatic body and tail tumors. Regardless of approach, DP-CAR can achieve reasonable survival in the context of multimodality therapy, however such resections must be carefully weighed against the significant morbidity involved.

Funding

None.

Conflicts of interest/disclosures

None.