Outcomes of resections for pancreatic adenocarcinoma with suspected venous involvement: a single center experience

Background Pancreatic ductal adenocarcinoma (PDAC) patients frequently present with borderline resectable disease, which can be due to invasion of the portal/superior mesenteric vein (PV/SMV). Here, we analyzed this group of patients, with emphasis on short and long-term outcomes. Methods 156 patients who underwent a resection for PDAC were included in the analysis and sub-stratified into a cohort of patients with PV/SMV resection (n = 54) versus those with standard surgeries (n = 102). Results While venous resections could be performed safely, there was a trend towards shorter median survival in the PV/SMV resection group (22.7 vs. 15.8 months, p = 0.157). These tumors were significantly larger (3.5 vs 4.3 cm; p = 0.026) and margin-positivity was more frequent (30.4 % vs 44.4 %, p = 0.046). Conclusion Venous resection was associated with a higher rate of margin positivity and a trend towards shorter survival. However, compared to non-surgical treatment, resection offers the best chance for long term survival.


Background
The extension of the resectability criteria for pancreatic adenocarcinoma (PDAC) has been intensively discussed in the last decade. It has been shown that arterial resections significantly increase mortality [1] but that resections for presumed invasion of the superior mesenteric/portal vein axis (PV/SMV, "venous resections") can be carried out safely without increased morbidity/mortality. In addition, taken all patients together, venous resections are not associated with an inferior survival than standard resections [2]. However, some groups have argued that true (in contrast to suspected) venous infiltration was associated with significantly shorter survival [3][4][5][6] while others have not reproduced these results [7][8][9]. Nonetheless the data on venous resections, especially those on perioperative outcome, have changed clinical practice in that these resections have become the standard for such tumors. Despite these technical improvements and the wide-spread introduction of adjuvant (and neoadjuvant) chemotherapy, prognosis has not much changed [10,11].
At the same time, outcome prediction remains challenging because a considerable subgroup of patients probably does not benefit from surgery. This is because it is difficult to predict pre-/intraoperative undetectable ("subclinical") metastatic disease and in line with this it is currently impossible to predict the biological behaviour of the tumors. Thus, there is currently a gap between our knowledge of the biology of PDAC and the technical/surgical advances. This is also underscored by a recent hypothesis that tumor size (and thus likelihood of vascular infiltration) and the time of metastasis are closely linked. Here, an exponential probability for metastatic spread was calculated depending on the size of the primary tumor [12]. Thus, many tumors deemed resectable will probably already have metastasized, though clinically not detectable [12]. In this study, we analysed our cohort of resected PDAC patients with special emphasis on venous resection/involvement and tumor biology/outcome.

Methods
The Institutional Review Board (Ethikkommission der Medizinischen Fakultät der Technischen Universität München, Munich, Germany) approved prospective and retrospective data collection as well as tissue collection (1926/07 and 5893/13). Written informed consent was obtained from all patients. Analysis was conducted on an anonymized data set.

Database and recorded parameters
In July 2007, we established a prospective database for the assessment of patients who underwent surgery for pancreatic diseases. The following parameters were recorded: tumor entity, age, gender, pre-operative weight loss, the following preoperative blood parameters: GPT/ALAT, GOT/ ASAT, bilirubin, AP, gamma-GT and CA19-9, comorbidities (pre-operative presence of diabetes mellitus, previous or concomitant cancer diagnoses other than PDAC, a history of (acute/chronic) pancreatitis, medical co-morbidities as reflected by the ASA score), presence of preoperative jaundice and/or ERCP/bile duct/pancreatic duct stenting, treatment (e.g. surgical technique, (neo-)adjuvant therapy), intraoperative blood transfusion(s), duration of operation, resection (and reconstruction) of the portal/superior mesenteric vein, histologically confirmed presence of venous invasion, size of the tumor (T), lymph node status (N), resection margin (R), grading (G), UICC-classification, , cl-Whipple "classical" Whipple, TP total pancreatectomy, DP distal pancreatectomy, DPH DP plus partial pancreatic head resection, STD standard deviation postoperative complications (according to the Clavien-Dindo classification [13]). The characteristics of the patients in the different cohorts are depicted in Table 1. Resectability criteria were defined according to the NCCN [14]. Locally advanced unresectable tumors were treated with neoadjuvant treatment protocols as decided in a multidisciplinary tumor board. Occasionally, patients were referred to our hospital after neoadjuvant or palliative intended therapy for borderline or locally advanced tumors. Lymph node dissection was standardized for all cases as described in the recent ISGP definition [15]. Venous resection was not considered a contraindication for surgery. The described approach did not change during the study period.

Inclusion and exclusion criteria
We retrospectively identified 209 patients who underwent an elective pancreatic resection with a final histopathological diagnosis of PDAC between 07/2007 and 07/2011 ( Fig. 1). The following patients were excluded from the analysis: UICC stage IV disease (n = 19), patients who were resected for local tumor recurrence (n = 3), patients on whom arterial resections were performed (n = 6), patients with a history of another cancer disease (n = 17) and patients who died because of surgery-related complications within 4 months from surgery (extension as suggested by Strasberg et al. [16], n = 6); three patients were lost to follow-up. Due to the relatively small patient cohort and considering the 2:1 ratio of the non-vein resected to the vein resected group, we opted for the inclusion of all patients and against a matched case analysis.  whether or not a resection of the PV/SMV (wedge resection, complete resection with end-to-end anastomosis or interposition of a goretex™ graft) had been performed. In this cohort of patients, preoperative imaging (computed tomography of the abdomen or magnetic resonance imaging/magnetic resonance cholangiopancreaticography (MRI/MRCP)) was evaluated to estimate sensitivity and specificity in regard to defining the presence of true venous invasion (versus an inflammatory/desmoplastic reaction at the PV/SMV).

Statistics
Continuous variables are reported as means ± SD or median (min.-max./95 %-CI), and were compared using a Student's t-test or a Mann-Whitney-U-test, as appropriate. Categorical variables are summarized as frequency counts and percentages and were compared using Fischer's exact test or Pearsons's chi-square test, as appropriate. Overall survival was defined as time from resection until death or until last follow-up.
Survival analysis was performed using the Kaplan-Meier method; differences were evaluated with the log rank test. A two-sided P value of <0.05 was considered as significant. All statistical analyses were performed using IBM SPSS, v20 for Windows (IBM Inc., USA).

Analysis of venous resection
The cohort of patients on whom venous resections have been performed was compared with the cohort of patients on whom standard pancreatic resections have been performed (Table 1). Except for late portal vein thrombosis (7 cases in the venous resection cohort versus one case in standard resection group), postoperative complications and (early and late) mortality were comparable in both groups ( Table 2). These findings are in line with published data [2,7,17,18], demonstrating that venous resections can be performed safely. Interestingly however, significantly more resections were marginpositive (i.e. R1; 44.4 % vs. 30.4 %, p = 0.046) in the venous resection cohort of patients. This might be a reflection of   [3][4][5][6]19], others have denied such an association [7][8][9]. We thus reassessed the histo-pathology reports on invasion/noinvasion of the vein (Table 1). In those cases were data on vein invasion was available, 30 % were tumor cellnegative whereas in 70 %, tumor cell invasion of the vein was seen. We then analyzed the patients with confirmed invasion of the vein vs. no-invasion; here, no significant differences in regard to preoperative markers, operative time or surgical technique were found (  Table 3). In line with these findings that implicated more advanced tumors in the true invasion group, significantly more patients had lymph node metastases when the vein was truly infiltrated (Table 3). CT or MRI scans were available in our digital imaging system from 23 of 27 patients. Radiological assessment of these scans revealed that CT or MRI predicted true invasion of the vein only at a sensitivity of 69 % and a specificity of 29 % (Table 3). Though the patient number is low, these data demonstrate the difficulties in preoperative prediction of the presence/extent of PV/ SMV invasion.

Survival following venous resections
To determine the extent of selection bias in the venous resection cohort of patients, we stratified these according to the UICC classification and plotted the subgroups following Kaplan-Meier survival analysis (Fig. 2a). This analysis demonstrates that the general prognosis of this cohort of patients is accurately predicted by the UICC. Thus, we assumed that selection bias was not overly strong in our cohort of patients and considered that the following analyses were statistically justified. Though there was a trend towards shorter survival of patients in whom venous resections had to be performed, this was not statistically significant (median survival 22.7 vs. 15.8 months, Fig. 2b; p = 0.157). Pathologically confirmed invasion of the vein was not relevant for prognosis. In addition, patients on whom a total pancreatectomy had to be performed lived shorter than those with pancreaticoduodenectomy or distal pancreatectomy (9.1 months versus 19.4 -28 months, Fig. 2c).
Next, we analyzed tumor grade as a surrogate marker for tumor biology. Patients with well-and moderately differentiated PDACs lived significantly longer than those with poorly differentiated tumors (overall p < 0.001, median survival 39.9 vs. 15.5 months, p < 0.001; Fig. 2d), and this stratification was also significant in the venous resection cohort (median survival 24.5 vs. 12.5 months, p = 0.023).

Discussion
The data presented in this paper confirm previously published results on portal vein resections during surgery for pancreatic adenocarcinoma, i.e. that it can be performed safely without increased morbidity and (early and late) mortality. There was a trend towards shorter survival and more frequent margin-positivity in the group of patients in whom venous resections had been performed. However, this was mainly a result of the outcomes after combined total pancreatectomy and portal vein resection. Such operations were only performed if a pancreaticoduodenectomy or a distal pancreatectomy were technically impossible or not sufficient to achieve macroscopic tumor clearance. This is also reflected in the finding that outcomes were worse in larger tumors. For partial pancreaticoduodenectomy and distal pancreatectomy with venous resections, there were no considerable differences compared to the respective standard resections. Thus, our data are in line with previous publications and support the concept that partial resection of the PV/SMV is justified to achieve negative margins and thereby better local control [20,21]. However, as patients undergoing total pancreatectomy plus PV/SMV resection had worse outcomes, it might be argued that this group of patients with centrally located, large tumors, for whom total pancreatectomy is anticipated/likely, should not be operated upfront, but might be better candidates for neoadjuvant treatment.
Our study also highlights the prognostic importance of tumor biology, using grading as a surrogate marker. Obviously, tumor biology rather than tumor location (and vascular abutment/invasion) is a stronger prognostic factor. This is also supported by recent data on long term survival after PDAC resections. Here, tumor biology was identified as the most important factor, with a significant number of long term survivors having a positive margin status and lymph node metastasis [11]. Further studies are required to better predict and understand tumor biology and stratify patients that benefit most from extensive surgical procedures.
There are some drawbacks of the present study. Although standardized histopathological reporting with special emphasis on resection margins had been introduced in our center in 2007 [22][23][24], venous involvement had not been assessed in all cases. Thus, data regarding preoperative diagnostic accuracy and true venous infiltration have to be interpreted cautiously.

Conclusion
In summary, our data support the current clinical practice of operating on patients with suspected involvement of the PV/SMV, and to perform (partial) resection of the PV/SMV if necessary to achieve tumor-free margins.