Vascular Resection for Pancreatic Cancer: 2019 French Recommendations Based on a Literature Review From 2008 to 6-2019

Introduction: Vascular resection remains a subject of debate in the management of Pancreatic Ductal Adenocarcinoma (PDAC). These French recommendations were drafted on behalf of the French National Institute of Cancer (INCA-2019). Material and Methods: A systematic literature search, with PubMed, Medline® (OvidSP), EMBASE, the Cochrane Library, was performed for abstracts published in English from January 2008 to June 2019, and identified systematic reviews/metaanalyses, retrospective analyses and case series dedicated to vascular resections in the setting of PDAC. All selected articles were graded for level of evidence and strength of recommendation was given according to the GRADE system. Results: Neoadjuvant treatment should be performed rather than direct surgery in borderline and locally advanced non-metastatic PDAC with venous and/or arterial infiltration (T4 stage). Patients who respond or those with stable disease and good performance status should undergo surgical exploration to assess resectability because cross-sectional imaging often fails to identify the extent of the remaining viable tumor. Combining vascular resection with pancreatectomy in these cases increases the feasibility of curative resection which is still the only option to improve long-term survival. Venous resection (VR) is recommended if resection is possible in the presence of limited lateral or circumferential involvement but without venous occlusion and in the absence of arterial contact with the celiac axis (CA; cephalic tumors) or the superior mesenteric artery (SMA; all tumor locations) (Grade B). The patients should be in good general condition because mortality and morbidity are higher than following pancreatectomy without VR (Grade B). In case of planned VR, neoadjuvant treatment is recommended since it improves both rate of R0 resections and survival compared to upfront surgery (Grade B). Due to their complexity and specificities, arterial resection (AR; mainly the hepatic artery (HA) or the CA) must be discussed in selected patients, in multidisciplinary team meetings in tertiary referral centers, according to the tumor location and the type of arterial extension. In case of invasion of a short segment of the common HA, resection with arterial reconstruction may be proposed after neoadjuvant therapy. In case of SMA invasion, neoadjuvant therapy may be followed by laparotomy with dissection and biopsy of peri-arterial tissues. A pancreaticoduodenectomy (PD) with SMA-resection is not recommended if the frozen section examination is positive (Grade C). In case of distal PDAC with invasion of the CA, a distal pancreatectomy with CA-resection without arterial reconstruction may be proposed after neoadjuvant therapy and radiologic embolization of the CA branches (expert opinion). Conclusion: For PDAC with vascular involvement, neoadjuvant treatment followed by pancreatectomy with venous resection or even arterial resection can be proposed as a curative option in selected patients with selected vascular involvement.


Material and Methods:
A systematic literature search, with PubMed, Medline ® (OvidSP), EMBASE, the Cochrane Library, was performed for abstracts published in English from January 2008 to June 2019, and identified systematic reviews/metaanalyses, retrospective analyses and case series dedicated to vascular resections in the setting of PDAC. All selected articles were graded for level of evidence and strength of recommendation was given according to the GRADE system.
Results: Neoadjuvant treatment should be performed rather than direct surgery in borderline and locally advanced non-metastatic PDAC with venous and/or arterial infiltration (T4 stage). Patients who respond or those with stable disease and good performance status should undergo surgical exploration to assess resectability because cross-sectional imaging often fails to identify the extent of the remaining viable tumor. Combining vascular resection with pancreatectomy in these cases increases the feasibility of curative resection which is still the only option to improve long-term survival. Venous resection (VR) is recommended if resection is possible in the presence of limited lateral or circumferential involvement but without venous occlusion and in the absence of arterial contact with the celiac axis (CA; cephalic tumors) or the superior mesenteric artery (SMA; all tumor locations) (Grade B). The patients should be in good general condition because mortality and morbidity are higher than following pancreatectomy without VR (Grade B). In case of planned VR, neoadjuvant treatment is recommended since it improves both rate of R0 resections and survival compared to upfront surgery (Grade B). Due to their complexity and specificities, arterial resection (AR; mainly the hepatic artery (HA) or the CA) must be discussed in selected patients, in multidisciplinary team meetings in tertiary referral centers, according to the tumor location and the type of arterial extension. In case of invasion of a short segment of the common HA, resection with arterial reconstruction may be proposed after neoadjuvant therapy. In case of SMA invasion, neoadjuvant therapy may be followed by laparotomy with dissection and biopsy of peri-arterial tissues. A pancreaticoduodenectomy (PD) with SMA-resection is not recommended if the frozen section examination is positive (Grade C). In case of distal PDAC with invasion of the CA, a distal pancreatectomy with CA-resection without arterial reconstruction may be proposed after neoadjuvant therapy and radiologic embolization of the CA branches (expert opinion).
Conclusion: For PDAC with vascular involvement, neoadjuvant treatment followed by pancreatectomy with venous resection or even arterial resection can be proposed as a curative option in selected patients with selected vascular involvement.

INTRODUCTION-RESECTABILITY OF PANCREATIC DUCTAL ADENOCARCINOMAS (PDAC)
Selection of patients for vascular resection is based on the probability of obtaining complete surgical resection (R0), because unlike R1 resection, this can result in prolonged survival or even be curative [Level of Evidence (LE) 3] (1-7). The presence and extent of vascular involvement are determined on high-quality thin-section images, with an anatomical basis for the classification of tumors as "borderline resectable" or "locally advanced" but not metastatic (8,9). Many classifications have been used to define the extent of PDAC, which is based on the relationship between the tumor and the venous or arterial axes (10, 11) (LE 3). The most common system is the National Comprehensive Cancer Network's (NCCN) classification, updated in November 2018 (12) (LE 2) (Supplemental Material 1). The notion of a "borderline" tumor has recently changed to take into account the anatomical classification, the probability of a histologically incomplete resection (R1), the patient's clinical status (general condition, co-morbidities, performance-status, "fragility syndrome") and the "biological" status of the disease (LE 2) (13)(14)(15)(16). The International Consensus on the definition of "borderline" tumors recommends to use a threshold CA 19-9 rate ≥ 500 units/mL for the latter (14)

(LE 3) (Supplemental Materials 2, 3).
A recent study (17) has shown that a standardized pathological protocol R0-resection based on 1 mm clearance was rarely achieved after upfront venous resection due to microscopic involvement of the SMV-groove (LE 4). It is important to note that patients considered to be at high risk of R1 resection and/or those with an unfavorable clinical and/or "biological" status are now candidates for neoadjuvant therapy (18)(19)(20)(21)(22)(23)(24)(25)(26) (LE 3). In one North American study (27) the benefits of neoadjuvant therapy were found to be significant in the presence of "unilateral" venous involvement (Ishikawa II-III) (LE 3). The PV patency ratio and its improvement under treatment are new prognostic indicators for PDAC treated with preoperative chemo-radiotherapy (28) (LE 4).
For borderline resectable PDAC, several more recent studies including two meta-analyses (29, 30) (LE 3), one phase II trial (31) and one randomized controlled trial (32) (LE 2), have confirmed that survival was improved after neo-adjuvant therapy followed by surgery than after upfront surgery followed by adjuvant therapy, even in an intent-to-treat analysis. The NCCN recommendations version 1.2019 (November 8, 2018) state that: "Immediate" resection of borderline tumors is no longer recommended (unlike 2016 recommendations), despite the absence of a randomized trial (neoadjuvant therapy vs. "immediate" surgery) and the definition of the best therapeutic protocol to use" (12,15) The purpose of neoadjuvant therapy is to increase the rate of patients candidates for potentially curative secondary resection. A systematic review published in 2017 (33) compared the pathological data in patients who underwent "upfront" surgery to those who underwent surgery after "neoadjuvant treatment." A significant reduction in the relative risk (RR) of R1 resection (RR = 0.66) and other negative predictive factors (tumor size, lymph node metastases, perineural extension, and lymphatic emboli) were observed after neoadjuvant treatment (LE 3) (Supplemental Material 4).
Due to the high prevalence of "borderline resectable" and "locally advanced" PDAC (around 15 and 25% respectively) and the lack of consensus about the treatment of theses entities, our aim was to establish recommendations regarding the treatment of PDAC with vascular involvement based on the existing literature.

METHODOLOGY
The National Institute of Cancer (INCa) commissioned these Guidelines in January 2017 and appointed a guideline leader (chair A.S.) who invited selected authors, all involved in the management of PDAC, to participate in the project development (May 2017). The key questions were prepared by the coordinating team and then approved by the other members. The coordinating team formed task-force subgroups, each with its own leader (J.R.D. for surgery), and divided the key topics among these task forces (October 2017). Process and steps taken to reach the final recommendations were illustrated in Table 1.
The INCa team independently performed systematic literature searches, with PubMed, Medline R (OvidSP), EMBASE, the Cochrane Library, and the internet for abstracts published from January 2008 to December 2017. Each task force also performed a systematic literature search. The literature search was restricted to abstracts published in English. Searches were updated every 3 months until June 2019. The search focused on fully published randomized controlled trials (RCTs), metaanalyses, prospective series and national and international guidelines and consensus. However, the literature search concerning vascular resections identified no RCT, only 4 systematic reviews/metaanalyses on venous resection and 2 systematic reviews/meta-analysis on arterial resection; thus, retrospective analyses and case series were also included. Conversely all case reports were excluded.
Manuscripts from abstracts containing relevant data were included. A summary of each reviewed manuscript was completed and summarized in literature tables for each key topic to prepare evidence-based and well-balanced statements on the assigned key questions for each task force. All selected articles were graded by the level of evidence and strength of recommendation according to The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system (Supplemental Material Methodology).
Each task force developed a draft and proposed statements on their assigned key questions, which were discussed on 4 plenary meetings (from November 2017 to September 2018). Recommendations were formulated based on the available evidence. All recommendations included a Grade rating based on the quality of evidence and strength of recommendation (Supplemental Material Methodology).
A first synthesis of the work from different groups was completed in January 2019. Thereafter, a combined document with all recommendations was created, which was reviewed and approved by all the group leaders, finalized and submitted to a national external review by 70 physicians, oncologists and surgeons (out of 126 solicited by the INCa), selected by regional cancer networks and 17 scientific societies collaborating in this project. The final manuscript was drafted after taking into account all comments and answering questions from the external validation group (April 2019) (Supplemental Material Methodology). All authors agreed on the final draft of the manuscript containing the recommendations. After agreement of all group members during a final plenary session held in Paris in May 2019, the guidelines was published online in November 2019 (https://www.e-cancer.fr).
These Guidelines will be considered for review every year or sooner if new and relevant evidence becomes available. An update will be done every 3 years after the publication of the recommendations. Any updates to the Guidelines in the interim will be noted on the INCa website.
These Guidelines are an official statement of the French Association of Hepatobiliary and Pancreatic Surgery and Liver Transplantation (ACHBT). It provides practical advice on how to manage pancreatic adenocarcinoma.

VASCULAR RESECTION FOR PANCREATIC CANCER Venous Resection
"What Are the Indications of Venous Resection?" Recommendations Venous resection associated with pancreatectomy is recommended if resection is possible in the presence of limited lateral or circumferential involvement but without venous occlusion and in the absence of arterial contact with the celiac trunk (cephalic tumors) or the superior mesenteric artery (all tumor locations) (Grade B).
The patients should be in good general condition because mortality and morbidity are higher than in pancreatectomy without venous resection (Grade B).
In the case of a planned venous resection, neoadjuvant treatment is recommended since it improves the rate of R0 resections and survival (Grade B).

Comments
Performing venous resection (VR) followed by reconstruction of the mesenteric-portal venous axis during pancreatic resections for PDAC may allow "en-bloc" resection facilitated by the superior mesenteric artery (SMA) "first" approach (34,35). Pancreaticoduodenectomies (PDs) are associated with VR in up to 25% of cases in France and Europe (less frequently in the US and more frequently in Japan). Distal pancreatic resections are associated with VR in 5-35% of cases (36)(37)(38)(39) (56)], a segmental resection followed by direct end-to-end anastomotic reconstruction with a "growth factor" (type 3 VR) or a "long" resection. In the latter setting (type 4 VR), if mesenteric root mobilization and lowering of the right liver are insufficient to compensate for the length of the VR (57), interposed graft reconstruction may be used, including an autologous venous or peritoneal (58), a cryopreserved homologous (59), a heterologous (60), or a prosthetic (61-65) graft. A recent study (66) reported the prognosis after reconstruction in 229 VRs (LE 4) and the median benefit to survival with segmental VR followed by a end-to-end anastomosis (usually planned). In this study, 129 patients underwent lateral VR followed by a direct suture (Group 1: 56%), 64 underwent a segmental VR followed by end-to-end anastomosis (Group 2: 28%) and 36 underwent VR followed by interposed graft reconstruction (Group 3: 16%). The surgical morbidity and mortality were comparable in all 3 groups. However, median survival was significantly different in the three groups: 27.6 months, 18.8 months and 13 months in groups 2, 1, and 3, respectively (66). If the venous splenoportal confluence is resected, the splenic vein territory is at risk of: (a) segmental portal hypertension (SPH) with gastric congestion; (b) varicose veins at the gastrojejunal anastomosis and pancreaticojejunal sites and esophageal varices with a late risk of upper GI bleeding; and (c) splenomegaly and thrombocytopenia, in case of prolonged survival (67, 68) (LE 4). In case of gastric congestion, reimplantation of the splenic vein is possible in the inferior mesenteric vein (IMV) (69) or the left renal vein (44, 54) (LE 4). However, reimplantation is not necessary if the resection has preserved the confluence between the splenic and left gastric veins and/or the IMV (70, 71) (LE 4). 5. Impact of the reconstruction technique on the long-term permeability of venous reconstructions. A recent meta-analysis has shown that reconstruction with interposition grafts (IG) influences the long-term permeability of venous reconstructions but not survival (65). This meta-analysis of 14 studies including 257 VRs with IG and 570 VRs without, showed that when venous reconstruction was performed with an IG, postoperative morbidity, mortality, and survival at 1, 3, and 5 years were comparable to those observed with other reconstruction techniques.
Finally, acute thrombosis is very rare in the immediate post-operative period. In a multicenter study of 406 VRs, only 7 patients developed acute thrombosis (1.7%) (38) (LE 4) and in a Japanese study of 197 VRs, only 3 patients developed acute thrombosis (1.5%) (75) (LE 4). Overall, the 1-year permeability rates ranged from 82 to 93% (61-63, 75) ( Table 3) (LE 4). Conversely, late thrombosis is frequent, often associated with recurrence (75% after a median of 15 months), and accompanied by portal hypertension and ascites in 75% of cases (61,72,75,76)   Finally, it is difficult to intraoperatively distinguish V+ from adventitious fibrotic adhesions secondary to peritumoural inflammation (96), particularly after neoadjuvant treatment. Although a desmoplastic reaction will result in negative pathologic examination of the resected vein, the benefit of a neoadjuvant strategy exceeds the risk of incomplete resection (15) (LE 2). In all studies, the survival of V-patients is comparable to that of patients with "standard" resection (97) (LE 4). However, a matched comparative study on small samples (98) (19 PD+VR with V-: 10 "upfront" VR/9 after chemoradiotherapy vs. 19 patients in the control group: 11 "upfront" VR/8 after chemoradiotherapy) reported that survival was better in patients who underwent VR and whose vein was V-than in patients who underwent standard PD, suggesting the benefit of systematic venous resection in the absence of any venous contact (LE 3).  (77). Moreover, the N+ rate was higher (80 vs. 65%) and the prognosis was poorer than that in patients with "clearly" resectable tumors (median cancer-specific survival: 14.4 vs. 24.4 months and median recurrence-free survival: 12 vs. 16.5 months; p = 0.0038). Survival was correlated with the severity of venous involvement observed on preoperative CT scan (Nakao Types B, C, or D: median specific survival 26, 12, and 16 months, respectively). Post-operative chemotherapy had a positive impact on cancer-specific survival regardless of the type of venous extension (Nakao Type B: 26 vs. 13 months; Type C: 27 vs. 8.6 months, p < 0.0001; Type D: 20 vs. 9.6 months, p < 0.0052), but compliance to treatment at 3 and 6 months was lower in case of venous involvement (57 and 45% vs. 73     2. In case of SMA invasion, neoadjuvant therapy is recommended, followed by laparotomy with dissection and biopsy of peri-arterial tissues in case of tumor stability or tumor response. If the frozen section examination is positive, a PD with arterial resection is not recommended (Grade C). 3. In case of distal PDAC with invasion of the celiac axis, neoadjuvant therapy is recommended. In case of stabilization or tumor response, a distal pancreatectomy with celiac axis resection without arterial reconstruction may be proposed after radiologic embolization of the CA branches (expert opinion).

(LE 3). A study in Japan
Since this meta-analysis, several retrospective monocentric studies including small numbers of patients have reported AR results (mainly common HA or celiac axis, more rarely SMA) (44, 117-119) (LE 4). A recent systematic review (115) (2000-2016) of 13 studies including 70 patients undergoing pancreatectomy with SMA resection, which is rarely performed (out of 10,726 undergoing pancreatectomy) concluded that there was no evidence to support SMA resection. Indeed, in the 25 patients with available individual patient-level outcome data, perioperative morbidity ranged from 39 to 91%, the mortality rate was 25% and median survival was only 11 months (LE 4).
However, the increasing use of neoadjuvant therapy protocols has increased the pool of selected patients who are candidates for "secondary" resection despite an initial suspected arterial invasion (116).
On the other hand, the Heidelberg group reported that they described as a more conservative approach to the major arterial axes, in particular the SMA: -a first study published in 2016 (122) (123) described radical tumor removal by sharp dissection along the CA and the SMA with complete dissection of all soft tissue between both arteries and superior mesenteric/portal vein ("TRIANGLE operation"). In case of positive frozen section(s) of the arterial sheaths, "nonresection" and palliative treatment were indicated. This study included a consecutive series of 15 patients. The R0 resection rate (1 mm) was 40% (6/15) in patients who had pancreatectomy with "arterial sparing" resection (LE 4).
2. Three additional situations can be distinguished and in each of these 3, AR must be planned: 2.1 First, anatomical variants of HA: "Right" HA arising from the SMA during a planned PD for a "clearly" resectable tumor: a) The HA may be "accessory": recent data suggest that preoperative embolization by interventional radiology followed several days later by "en bloc" resection may be performed with no significant risks of liver/biliary ischemia due to development of intrahepatic arterial shunts (124) (LE 4). A systematic review has shown the feasibility and lack of morbidity and mortality of this strategy (125) (LE 4). This strategy avoids opening the accessory tumor/HA interface, which can be exposed with the risk of tumor spillage in case of RHA preservation (126) (LE 4).
b) The HA can perfuse the "total liver" (Michell type 9: 1-5%): in this rare setting, it requires reconstruction of any type [direct anastomosis, by an interposed "reversed" saphenous vein graft (44,53,117) (LE 4) or reversal of the splenic artery (127) (LE 4)] to ensure vascularization of the biliary tree and the hepatico-jejunostomy following PD. HA should be reconstructed before continuing pancreatic resection to avoid any liver ischemia, particularly when an associated venous resection is needed. This procedure is contraindicated when the CA is invaded at its origin on the aorta or if the GDA is invaded (132-134) (LE 4). Indeed, DP-CAR requires a tumor-free and patent GDA to ensure "reverse flow" vascularization of the liver and bile ducts from the SMA through the PDA and GDA (12) (LE 2) and, for many authors, the use of preoperative occlusion of CHA or, at the best, the 3 branches of the CA to favor development of arterial collaterals thus reducing the risk of bile ducts and gastric ischemia (135,136) (LE 4). Embolization should be performed 1-2 weeks before resection (137,138) (LE 4). This procedure, which avoids any arterial reconstruction, remains controversial. Some authors prefer reconstruction in case of insufficient flow during an intraoperative "Doppler" control (139,140) (LE 4). Embolization is not effective in the case of "total liver" HA arising from the SMA, which requires reconstruction (137,138) (LE 4).

Monocentric studies:
Several monocentric, mainly Japanese, studies with small groups of patients, have been published in the past 10 years and were included in two recent systematic reviews, with reported median survivals ranging from 10 to 26 months and 5-year survival rates of 20% (97) (LE 4).
Another recent retrospective study (141) compared the outcomes of patients receiving (n = 11) or not (n = 9) various regimens of neoadjuvant chemotherapy (mainly GEM-nab-PTX). Despite the small number of patients, those who received neoadjuvant chemotherapy had significantly less arterial invasion (p = 0.025), lymphatic invasion (p < 0.0001), and vascular invasion (p = 0.035) with significantly higher recurrence-free and overall survival rates.
-A retrospective study in Japan (142) (144). Twenty patients who underwent DP-CAR ("modified" Appleby; 2.4%) recruited in 16 centers (obviously a limitation for this study) were compared to 172 patients who underwent DP who were matched for age, sex, BMI, albumin blood level, ASA score, pancreatic consistency, main pancreatic duct diameter, and pathology (60% PDAC).
The procedure was longer for DP-CAR (median 276 vs. 207 min; p < 0.01) and the rates of postoperative acute renal failure (10 vs. 1%; p < 0.03) and 30-day mortality ( ARs are therefore very rarely indicated, often associated with venous resection (24, 97, 120, 121) (LE 3), and must be "planned" since they require routine neoadjuvant treatment and frequent preoperative arterial embolization. Surgery should always begin with an "artery first" approach (34,35) to accurately evaluate any persistent arterial involvement confirmed by frozen section examination (34,35,116). Forty (121) to 70% (24, 119) of patients do not receive postoperative chemotherapy (primarily single agent gemcitabine) due to postoperative morbidity and prolonged recovery time and there are few data on patient quality of life. Interestingly, a recent study from the MD Anderson Cancer Center (149) including 127 patients who received neoadjuvant treatment before PD (including vascular resection in 58 (46%); VR = 44, AR = 3, both = 11) reported that: (a) all patients experienced at least a transient skeletal muscle, visceral fat and subcutaneous fat loss; but (b) a relative increase in skeletal muscle (HR = 0.50) and albumin (HR = 0.57) during the first postoperative 12-months were associated with improved overall survival. This suggests that persistent postoperative skeletal muscle loss may represent an early marker of poorer outcomes.

CONCLUSIONS
PD with venous resection improves survival compared to no resection, especially with R0 resection. Mortality and morbidity are higher in PD with venous resection than in PD without vascular resection. PD with upfront venous resection has a poorer oncological results (increased risk of R1 resection, poorer survival) than PD with venous resection after neoadjuvant treatment. PD with arterial resection is associated with increased morbidity and mortality (compared to PD with venous resection) and has not been shown to be beneficial. A distal splenopancreatectomy with celiac axis resection is associated with increased morbidity and mortality and the oncological benefit of this approach has not been clearly demonstrated.
Today, literature provides more support for neoadjuvant therapy in the management of pancreatic cancer. Waiting for RCTs results including clearly resectable tumors, neoadjuvant therapy and a complete R0 resection in all patients who require planned vascular resection with (or without) reconstruction should be the goal. Such patients should be treated by an experienced team in both preoperative/neoadjuvant therapy and vascular resection at the time of pancreatic resection. Such expertise is not available at every centers, which makes another strong case for the regionalization of complex cancer care that involves multiple treatments.

AUTHOR CONTRIBUTIONS
JD and AS contributed equally to the design and implementation of the research, to the analysis of the results, and to the writing of the manuscript.