Local radicality and survival outcome of pancreatic cancer surgery

Abstract Pancreatic cancer remains a therapeutic challenge. Surgical resection in combination with systemic chemotherapy is the only option promising long‐term survival and potential cure. However, only about 20% of patients are diagnosed with tumors that are still in a resectable stage. Even after potentially curative resection and modern regimens for adjuvant chemotherapy, the majority of patients develop local and systemic recurrence resulting in median overall survival times of 28‐54 months. The predominance of systemic recurrence and its impact on survival may lead to the assumption that surgical radicality and local control play only minor roles in the treatment of pancreatic cancer. This review provides an overview of the recent literature on surgical radicality and survival outcome in pancreatic cancer. The current evidence on the extent of lymphadenectomy, the prognostic impact of the extent of lymph node involvement, and the impact of the resection margin status on postresection survival are reviewed. Data from recent studies performed in the context of modern surgery and adjuvant therapy provide good evidence of a considerable impact of local radicality on survival after pancreatic cancer surgery. Surgical techniques that have been developed to refine oncological resections and to increase local control as well as resectability are highlighted. These techniques include artery‐first approaches, level‐3 dissection with removal of the periarterial nerve plexus, the triangle operation, and extended resections. Local radicality and quality of surgical resection remain among the most important parameters that determine the chances for survival in patients with non‐metastatic pancreatic cancer.

28-54 months after potentially curative resection and the best available regimens for adjuvant therapy. [1][2][3][4][5] The recurrence patterns point to problems with both local surgical radicality and early systemic spread with presence of micrometastatic disease at the time of surgical resection as two mechanisms resulting in poor survival outcomes. While the problem of early systemic spread can only be overcome by development of more effective systemic therapies, surgeons may be able to impact on local control by the radicality and quality of surgical resection. However, some surgeons and many oncologists believe that with early systemic recurrence seen in most patients with pancreatic cancer, surgical radicality and local control play only minor roles in this disease.
This review provides an overview of the recent literature on surgical radicality and survival outcome in pancreatic cancer with a focus on the extent of lymphadenectomy and the resection margin status as two surrogate markers of local radicality. Surgical techniques developed to refine oncological resections and to increase local control are highlighted. While strategies of neoadjuvant therapy that are used to achieve resectability in unresectable tumors 6 may also increase local control in resectable and borderline-resectable tumors, this review focuses on studies performed in the setting of upfront resection.

| E X TENT OF LYMPHADENEC TOMY AND OUTCOME IN PAN CRE ATIC C AN CER
Pancreatic cancer is a tumor characterized by early lymphatic invasion and early spread to regional lymph nodes. The rate of lymph node metastases in resectable pancreatic cancer is high at about 70%-80%. [7][8][9][10][11][12] The presence of lymph node metastases impacts on tumor stage and is an important prognostic factor associated with decreased survival. In a recent large study performed with a strategy of upfront surgical resection with a radical regional lymphadenectomy and adjuvant chemotherapy, median overall survival in N0 versus N+ tumors was 33.2 months versus 23.6 months and 5-year survival rates were 31.7% versus 17.4%, respectively. 12 It has been a long-standing topic of research and debate among surgeons if the prognosis of pancreatic cancer with lymph node metastases can be improved by extended lymphadenectomy. [13][14][15][16][17] Five randomized controlled trials published between 1998 and 2014 have compared a "standard" regional versus an "extended" lymphadenectomy for pancreatic head cancer (Table 1). [13][14][15][16][17] Overall, the individual trials as well as a recent meta-analysis based on these trials 18 came to the conclusion that extended lymphadenectomy does not result in improved survival but is associated with increased morbidity and should, therefore, not be recommended as the standard procedure.
However, a closer look to the data reveals that although the definitions used for the "standard" and the "extended" lymphadenectomy were quite similar among these trials, there is a considerable heterogeneity and a lack of comparability of data available from these studies, as evidenced by the number of examined lymph nodes. The median numbers of examined lymph nodes are a surrogate marker of the actual extent of lymphadenectomy and vary between 13 and 17 lymph nodes in the "standard" and between 20 and 40 lymph nodes in the "extended" lymphadenectomy groups among the trials. Recent studies from Europe and Japan report median numbers around 23-26 examined lymph nodes for "regional" lymphadenectomy 4,12 and, therefore, numbers range between the numbers reported for "standard" and "extended" lymphadenectomy in the available randomized controlled trials on the topic. The Japanese Pancreas Society (JPS) has established a comprehensive nomenclature of the different lymph node stations that are relevant for pancreatic cancer surgery 19,20 which was not yet consistently used in some of the above-mentioned randomized trials. This nomenclature has meanwhile been internationally adopted and allowed to set international standards for the extent of lymphadenectomy in pancreatic cancer. Based on this nomenclature the International Study Group on Pancreatic Surgery (ISGPS) released consensus recommendations of a standard regional lymphadenectomy to be performed during pancreatoduodenectomy for pancreatic head cancer and during distal pancreatectomy for pancreatic body and tail cancers in 2014. 21 Briefly, the principles of a standard lymphadenectomy are defined by a radical removal of all regional lymph nodes including all lymph nodes on the tumor-oriented side of the celiac axis and the superior mesenteric artery.
The term "extended" lymphadenectomy should be reserved for retroperitoneal lymph nodes and such extended lymphadenectomies are not recommended as a standard. Lymph nodes in this area are internationally considered extraregional and, therefore, distant metastases (M1). While some surgeons consider extraregional lymph node metastases to be a contraindication for surgery, small observational studies suggest that the prognosis after their resection is much better than in the setting of distant organ metastases. 22 In a study focused on the prognostic impact of the extent of lymph node involvement, survival was quite similar after resection of tumors with ≥8 positive regional lymph nodes and resection of additional inter-aortocaval lymph node metastases (median survival: 18.3 vs 13.6 months; identical 5-year survival of 9.9% vs 9.9%). 12 These data suggest that retroperitoneal lymph node metastases are just a more advanced extent of lymph node involvement but do not have the same biological and prognostic implications as distant organ metastases. With a 5-year survival of around 10%, a significant proportion of patients with inter-aortocaval lymph node metastases appear to benefit from surgical resection. We, therefore, recommend taking frozen section biopsies of retroperitoneal lymph nodes whenever their involvement is suspected based on imaging or surgical exploration and to perform an extended retroperitoneal lymphadenectomy if metastases are confirmed. A sensible alternative may be to abandon upfront resection and to administer chemotherapy with a neoadjuvant intention in these selected patients.
Recent observational studies performed in the context of a "radical" regional lymphadenectomy have renewed the international discussion on the prognostic significance of the extent of regional lymph node involvement in pancreatic cancer. Based on the JPS nomenclature of lymph nodes and data available in Japan due to a historically more meticulous approach toward lymphadenectomy in pancreatic cancer, two categories for lymph node-positive tumors dependent on anatomical groups of regional lymph nodes involved had been already used for prognostic staging in Japan. 23 In contrast, only one category (N1) for all tumors with regional lymph node metastases irrespective of the extent of lymph node involvement was still used in the seventh edition of the TNM staging manual. 24 Following up on a study from Japan showing that the number of positive lymph nodes was a powerful predictor of prognosis in pancreatic cancer if a thorough lymphadenectomy is performed 25 12 In this study median overall survival of patients with regional lymph node metastases ranged from 31.1 months with one PLN to 18.3 months with ≥8 PLN. The differences in 5-year survival rates were even more pronounced ranging from 31.4% to 9.9% with one and ≥8 PLN, respectively. The extent of lymph node involvement was confirmed as an independent predictor of overall survival by multivariable analyses with a cut-off at four PLN. 12 Together with two smaller confirmatory studies 26,27 and a population-based study using the Surveillance, Epidemiology, and End Results Program While the above-mentioned studies clearly demonstrate the importance of lymphadenectomy for prognostic staging, direct evidence on a therapeutic effect of lymphadenectomy in pancreatic cancer remains limited. An analysis of the SEER database including 7685 patients with stage I and II pancreatic cancer found that retrieval of 20 or more regional lymph nodes was associated with increased survival in node-negative as well as node-positive cancers after adjustment for other prognostic factors. 31 While the improved survival with ≥20 retrieved lymph nodes in node-negative cancers may be explained by effects of stage migration, the improved survival with ≥20 retrieved nodes in node-positive cancers points to a possible therapeutic effect of radical lymphadenectomy. 31 In contrast to overall survival, the extent of lymphadenectomy may more directly affect local recurrence and recurrence-free survival. However, good evidence on the effect of lymphadenectomy on the pattern of recurrence after resection for pancreatic cancer is lacking. While early systemic recurrence is more relevant in limiting survival in the majority of patients with pancreatic cancer, about 24% of patients are first affected by isolated local recurrence. 2 A significant proportion of these isolated local recurrences may originate from regional lymph node metastases. In a radiological study of computed tomography scans performed for surveillance in 99 patients after pancreatic cancer resection, 17% of patients developed isolated local recurrence, including six patients with isolated lymph node recurrences and 11 patients with lymph node and additional perivascular recurrences. 32 In a study on re-resection for isolated local recurrence of pancreatic cancer, 41 (72%) of 57 patients with isolated local recurrence proven histologically upon surgical exploration, underwent successful re-resection associated with a median survival of 26 months after re-resection. 33 The majority of these recurrences was perivascular or located in locoregional or retroperitoneal lymph nodes.
In conclusion, the current evidence supports a radical locoregional lymphadenectomy as for example recommended by the ISGPS as a minimum standard lymphadenectomy during pancreatic cancer surgery. 21 Extended retroperitoneal (paracaval/inter-aortocaval/para-aortic) lymphadenectomy should not be performed as a standard procedure because it does not improve overall survival but may increase complications if applied as a standard procedure to unselected patients. In contrast, extended lymphadenectomy may be indicated in selected patients with suspected lymph node metastases in this location during upfront resections or after neoadjuvant therapy. In selected patients with isolated lymph node recurrences that occur during surveillance after pancreatic cancer resection, surgical re-resection can be considered.
Based on the oncological principles of radical en bloc tumor resection, the extent of regional lymphadenectomy is closely connected with the extent and local radicality of resection around the major vessels, especially the celiac axis and the superior mesenteric artery, as discussed in the following section. Surgeons who follow the principles and techniques described below will "automatically" perform an adequate regional lymphadenectomy.

| RE S EC TI ON MARG IN S TATUS AND SURVIVAL IN PAN CRE ATI C C AN CER
In addition to the extent of lymphadenectomy, the resection margin status (R-status) is another important surrogate marker for surgical radicality and another important prognostic factor that can be in- were R1 resections, the medial and posterior margins (located toward the superior mesenteric artery and the celiac axis) were most frequently involved. 34,35 While the need for assessment of circumferential margins was quickly accepted around the world, the strict definition of the R0 status based on the "1-mm rule" was adopted in Europe but not accepted internationally and has not been commonly used in studies from the USA or Asia. 4,36,37 This resulted in a considerable lack of comparability of data on the frequencies and the prognostic impact of R0 and R1 resections. The ISGPS reacted in 2014 by releasing consensus definitions for extended pancreatic resections and borderline-resectable pancreatic cancer that included not only the recommendation to report the resection margin status based on assessment of seven distinct margins but also supported reporting on a 1-mm free margin. 38 3 In a multivariable analysis, R-status was confirmed as an independent predictor of survival in this study. 3 Despite good evidence for the strict R-definition using a "1-  (Table 2). 3,4,[41][42][43][44][45] Of note, the 5-year overall survival rate after R1 resection with direct margin involvement can still be as high as 20%-25% and is, therefore, much better than frequently discussed. While obtaining R0 margins is the main goal of every resection performed for pancreatic cancer, R1 resections (even those with direct margin involvement) are still associated with acceptable outcome and should not be interpreted as failure of surgical therapy.

| SURG I C AL TECHNI QUE S TO IN CRE A S E LO C AL R AD I C ALIT Y IN PAN CRE ATI C C AN CER
In recent years surgical techniques were significantly refined and, together with advances in systemic chemotherapy regimens, resulted in improved outcomes in pancreatic cancer surgery. 1 These advances have led to the possibility to extend the indications for surgical resection from clearly resectable to locally advanced, borderline-resectable and previously unresectable tumors. In the following, we want to highlight selected surgical techniques and strategies that contribute to improved local radicality and improved outcomes in pancreatic cancer surgery.
Consistent with the observation that most R1 resections for pancreatic cancer are located at the posterior and medial margins oriented toward the superior mesenteric vessels and the celiac axis, 34,35 the techniques aiming to increase local radicality are centered on clearance of these vessels as an important and early step during pancreatic cancer resections. With this aim, different techniques have been developed that are today summarized as "arteryfirst approaches". 53 One such technique, the mesenteric approach developed by Nakao et al was already described in 1993 based on a study in 114 patients. 54 Addressing the superior mesenteric artery (SMA) early and even before mobilization of the pancreatic head, was new and contrary to traditional approaches in pancreatic cancer surgery. A detailed review of the mesenteric approach highlights its advantages for locally advanced tumors with potential SMA involvement and tumors located in the uncinate process. 55  It should be noted that at present the available evidence for these advantages of artery-first approaches is relatively low because it is limited to retrospective cohort studies. 56 In order to be effective in increasing R0 rates and radicality, the level of dissection at the arteries should be directly at the adventitial layer of the vessels, resulting in a complete dissection of the nerve plexus (and of lymphatic tissue along with it) corresponding to a level-3 dissection according to Inoue et al. 65 The result of such radical resections at the arteries has also been described as a complete mesopancreas excision. 66 The rationale to always perform a level-3 dissection at the SMA and celiac axis at least semicircumferentially, is based on the assumption that with this technique the tumor-free margin is maximized and tumor cells spreading beyond the tumors along the perivascular nerves (perineural infiltration) are removed.
For tumors with contact of more than 180° of contact to the arteries, and especially for resection after neoadjuvant therapy for primarily unresectable tumors, it is frequently necessary to perform a circumferential level-3 dissection around the SMA and the celiac axis. As these vessels then form a triangle together with the porto- In contrast to venous resections, arterial resections should not be considered for upfront surgery, but for neoadjuvant treatment, which may result in resectability without arterial resections in the majority of cases. 6,67 An exception may be tumors invading the celiac axis resectable by a distal pancreatectomy with celiac axis resection (DP-CAR). This procedure relies on arterial blood supply of liver and stomach by collateralization from the SMA via the gastroduodenal artery after resection of the celiac axis without reconstruction.
Feasibility and acceptable safety with mortality rates of 3%-8% after DP-CAR were demonstrated in several single-center observational studies. 71,72 In a recent retrospective international multicenter study including 191 patients undergoing DP-CAR, the 90-day mortality rate was 5.5% at five high-volume but as high as 18% at 18 low-volume DP-CAR centers, demonstrating the importance of experience with this rare and complex procedure. 73 In a multicenter study performed in 20 European centers in 12 countries, the median survival of 68 patients undergoing DP-CAR was 18 months. 74 A recent single-center study reported a very favorable median survival of 38.6 months with a strategy of neoadjuvant therapy followed by DP-CAR in pancreatic cancer with celiac axis involvement, recommending a neoadjuvant strategy in these patients. 71 The literature on pancreatic cancer resections with arterial resections apart from DP-CAR is restricted to case reports and small series with high risk of bias. A meta-analysis on this topic reported a five-fold increased risk of mortality after arterial resections versus standard resections and poor 1-and 3-year survival rates. 75 However, some patients may benefit from arterial resections, as long-term survival can be observed. 75 Overall, extended resections are associated with shorter survival and may be associated with higher morbidity if compared to standard resections. While venous resections appear to be safe, the need for extended total pancreatectomy and arterial resections are associated with increased morbidity and mortality. Careful patient selection, evaluation of a neoadjuvant strategy and treatment in specialized units are warranted if the need for an extended resection is anticipated. If arterial involvement is anticipated, a strategy of neoadjuvant therapy should usually be preferred.

| CON CLUS I ON AND PER S PEC TIVE
Pancreatic cancer surgery has rapidly evolved in the last decades and along with advances in adjuvant and neoadjuvant therapy, resection rates and survival outcome have significantly improved. However, the prognosis of pancreatic cancer remains poor and most patients eventually develop and die from systemic progression. Therefore, pancreatic cancer has to be considered a systemic disease even in an early clinical tumor stage. In spite of this problem, radical resection with adequate regional lymphadenectomy and radical resection around the large peri-pancreatic vessels is an important prerequisite for good oncological outcomes. There is ample evidence from recent studies performed in the context of high-quality radical surgery and modern adjuvant therapy that local radicality, defined by lymph node variables and by resection margin status data, has a profound impact on survival. In modern pancreatic surgery, radical resections can be facilitated and achieved by several techniques, including artery-first approaches, a level-3 dissection around the arteries, the TRIANGLE operation, and extended resections with resection of additional organs or vessels. Along with surgical radicality, systemic chemotherapy is the second critical cornerstone of long-term survival after pancreatic cancer resection. Currently, the best therapy sequencing of surgery and chemotherapy is one of the most important topics in the field of pancreatic cancer surgery and the results of several randomized controlled trials on the strategies of neoadjuvant therapy or upfront resection in resectable and borderline-resectable pancreatic cancer are eagerly awaited. Independent of the results of these trials, local radicality and quality of surgical resection will remain among the most important parameters determining the chances for survival in patients with non-metastatic pancreatic cancer.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interests for this article.