Surgical Resection Versus Palliative Chemoradiotherapy for the Management of Pancreatic Cancer with Local Venous Invasion: A Decision Analysis
- First Online:
- Cite this article as:
- Abramson, M.A., Swanson, E.W. & Whang, E.E. J Gastrointest Surg (2009) 13: 26. doi:10.1007/s11605-008-0648-y
- 116 Views
Benefit from pancreaticoduodenectomy (PD) combined with superior mesenteric-portal vein (SMV-PV) resection in the management of pancreatic adenocarcinoma with local venous invasion remains controversial.
Using formal decision analysis, we compared survival associated with PD plus SMV-PV resection when applied to patients with pancreatic adenocarcinoma with isolated local venous invasion (Group 1) versus that achieved with palliative chemoradiotherapy when applied to patients with locally advanced pancreatic cancer (Group 2). Individual studies were identified using Medline. A total of 1,324 and 709 patients were analyzed for Groups 1 and 2, respectively. Patients with distant metastases were excluded.
Overall decision analysis favored surgical resection (Group 1) over palliative chemoradiotherapy (Group 2). Sensitivity analyses indicated that this decision is sensitive to the perioperative mortality rate and the percentage of surgical resections with microscopic (R1) or macroscopic (R2) residual tumor at the resection margin. In contrast, sensitivity analysis revealed that the decision is not sensitive to the percentage of cases in which true venous invasion by cancer is documented histologically.
Surgical resection may confer a survival advantage over palliative chemoradiotherapy in select patients with pancreatic cancers with presumed local venous invasion.
KeywordsPancreatic adenocarcinomaPancreaticoduodenectomySurvivalDecision analysis
Pancreatic cancer is the fifth leading cause of cancer-related deaths in the USA, causing an estimated 33,370 deaths in 2007.1 Surgical resection offers the only potential for cure for patients with this disease.
The benefit of surgical resection for the subset of patients with locally advanced pancreatic cancers with isolated portal vein and/or superior mesenteric vein (PV-SMV) invasion is controversial. Although it has been suggested that resection of these lesions can be performed safely (with acceptable long-term survival rates), the survival benefit of surgical resection over palliative chemoradiotherapy has not been confirmed in a randomized controlled trial.
As such, in this study, we conducted formal decision analysis in order to compare survival in patients with pancreatic cancer based on two competing treatment strategies; Group 1 patients underwent surgical resection (pancreaticoduodenectomy (PD) with venous resection and reconstruction) for pancreatic adenocarcinoma with isolated local venous invasion and Group 2 patients underwent palliative chemoradiotherapy for locally advanced pancreatic cancer.
Materials and Methods
A systematic Medline search was conducted using the search term “pancreatic cancer AND vein resection” to identify English language publications containing data relevant to Group 1. Despite an extensive search of the literature, we were unable to identify studies reporting survival data specifically for patients with isolated local venous invasion treated solely with chemoradiation. As such, we used survival data for patients with locally advanced cancers treated with palliative chemoradiation as a proxy for survival among Group 2 patients. We believe this is an appropriate strategy, given that nonresected patients with Stage IIA (locally invasive; resectable; T3, N0, M0), Stage IIB (locally invasive; resectable; T1,2, or 3, N1, M0), and Stage III (locally advanced; unresectable; T4, any N, M0) disease are reported to have nearly identical 1-year survival rates (25.0%, 26.9% and 27.0%, respectively).7,8 Upon detailed review of the selected articles for both groups, additional articles were subsequently identified that met inclusion criteria. Critical appraisal of each study was performed by the authors, and studies were selected on the basis of the inclusion criteria used for this analysis.
Inclusion and Exclusion Criteria
Letters, reviews without original data, animal studies, studies without survival data, and overlapping studies were all eliminated from the analysis. In studies that did not explicitly state survival time periods in the text, we have estimated survival using Kaplan–Meier Survival curves.
Perioperative mortality was defined as death within 30 days of the surgery. However, in-hospital or operative mortality was used as a substitute for perioperative mortality when death within 30 days of surgery was not reported.
Studies used for Group 1 included patients having pancreatic cancer with local invasion into the portal vein and/or superior mesenteric veins. All studies meeting these criteria were included regardless of the extent of lymph node dissection and/or venous resection. In some of the studies, survival data included patients undergoing simultaneous vein and arterial resections. Additionally, some studies included in our analysis included patients undergoing additional neoadjuvent and/or adjuvant therapies. Studies used for Group 2 included patients with locally advanced pancreatic cancers who received palliative chemoradiotherapy. Studies were included regardless of the type or duration of the treatment regimen. Some studies in this group included patients with incompletely resected pancreatic cancer with residual disease. Patients were excluded from both Groups 1 and 2 if there was any evidence of metastatic disease.
Decision Analysis Models and Calculations
Weighted means were calculated for each variable and used as baseline estimates, taking into account the number of patients contributing to each outcome. As the outcome of interest in this analysis was survival at a given time point, a utility (payoff) of 1 was assigned for a patient surviving to that time point, and a utility of 0 was assigned for a patient surviving less than to that time point.
One-way sensitivity analysis was performed for variables in the decision models to determine the impact of uncertainty in the estimates of probabilities. Threshold values were calculated for variables that would lead to a change in the preferred strategy when traversed. If the decision outcome to select one treatment strategy over the other did not change over the range of the variable being manipulated, the decision was considered to be not sensitive to this variable and, thus, no threshold was identified for that variable. Alternatively, if the decision outcome to select one treatment strategy over the other did change over the range of the variable being manipulated, the decision was considered to be sensitive to this variable, and the value at which the optimal strategy changed was considered to be the “threshold value.”
Analysis of Data Used
Published Studies Included in the Surgical Resection Group (Group 1)
Yekebas et al.10
University Medical Centre Hamburg-Eppendorf (Germany)
Al-Haddad et al.11
Mayo Clinic (Jacksonville)
Shimada et al.12
National Cancer Center (Tokyo, Japan)
Carrere et al.13
Hopital Beaujon, University Paris VII (France)
Nakao et al.14
Nagoya University (Japan)
Riediger et al.15
University Hospitals of Rostock and Freiburg (Germany)
Zhou et al.16
Shanghai Institute of Digestive Surgery (China)
Koniaris et al.17
University of Miami School of Medicine
Poon et al.18
Queen Mary Hospital, University of Hong Kong (China)
Tseng et al.19
University of Texas, M.D. Anderson Cancer Center
Bin Li et al.20
Multiple hospitals in China
Howard et al.21
Indiana University School of Medicine
Nakagohri et al.22
National Cancer Center East (Japan)
Aramaki et al.23
Oita Medical University (Japan)
Kawada et al.24
Hokkaido University (Japan)
Sasson et al.25
Fox Chase Cancer Center and Temple University
Hartel et al.26
University-Hospital Mannheim (Germany)
Shibata et al.27
Sendai City Medical Center and Iwate Medical University (Japan)
van Geenen et al.28
Academic Medical Center (Netherlands)
Bachellier et al.29
Hopital Universitaire de Hautepierre (France)
Park et al.30
Samsung Medical Center (Korea)
Kinoshita et al.31
Kurume University School of Medicine (Japan)
Launois et al.32
Centre Medico Chirurgical Saint Vincent (France)
Civello et al.33
Catholic University School of Medicine (Rome, Italy)
Naganuma et al.34
Mie University School of Medicine (Japan)
Harrison et al..35
Memorial Sloan-Kettering Cancer Center
Roder et al.36
Technische Universitat Munchen (Germany)
Klempnauer et al.37
Hannover Medical School (Germany)
Yeo et al.38
Johns Hopkins Medical Institutions
Allema et al.39
Academic Medical Centre (Netherlands)
Takahashi et al.40
Keio University School of Medicine (Japan)
Ishikawa et al.41
The Center for Adult Diseases (Osaka, Japan)
Published Studies Included in the Chemoradiotherapy Group (Group 2)
Ikeda et al.42
Nat Cancer Center Hosp. East, Japan
Haddock et al.43
Cohen et al.44
Fox Chase Cancer Center
Mishra et al.45
Rich et al.46
University of Virgina
Blackstock et al.47
Martenson et al.48
Technicon-Israel Institute of Tech
Wolff et al.50
Safran et al.51
Talamonti et al.52
Ishii et al.53
Nat. Cancer Center Hosp., Tokyo
Whittington et al.54
University of Pennsylvania
Moertel et al.55
Moertel et al.56
Louvet et al.57
Hospital St. Antoine
Rocha Lima et al.58
University of Miami
Bramhall et al.59
Queen Elizabeth Hospital
McGinn et al.60
University of Michigan
Weighted mean (%)
Range (%) in the literature
1-year survival for treatment with chemoradiation
1-year survival for treatment with surgical resection
Histopathologic proven vein invasion
1-year survival for treatment with surgical resection given true histopathologic vein invasion
1-year survival for treatment with surgical resection given no histopathologic vein invasion
R1 or R2 resections
1-year survival for treatment with surgical resection given an R0 resection
1-year survival for treatment with surgical resection given an R1 or R2 resection
The decision to perform surgical resection versus chemoradiation is sensitive to manipulation of the perioperative mortality rate. At the baseline perioperative mortality of 3.3%, decision analysis favors resection over chemoradiation. However, at a perioperative mortality rate higher than 31%, (intersection of lines in Fig. 4), decision analysis favors treatment with chemoradiation over surgical resection.
One-way sensitivity analysis revealed that an increase in the percentage of cases in which true venous invasion by cancer is documented histologically resulted in a decrease in the 1-year survival of those patients undergoing surgical resection (Fig. 5). However, the decision analysis favored surgical resection over chemoradiation regardless of the percentage of tumors with histopathologically proven invasion, and thus, the decision is not sensitive to this variable.
Finally, one-way sensitivity analysis demonstrated the decision to treat with surgical resection over chemoradiation is sensitive to the percentage of surgical resections with either microscopic (R1) or macroscopic (R2) residual tumor at the resection margin. At a probability of R1 plus R2 resections greater than 77%, surgical resection is no longer the favored strategy (Fig. 6).
As far as we are aware, this is the first study to utilize decision analysis to assess the potential survival benefit of surgical resection for patients with pancreatic cancer with local venous invasion. Given a lack of randomized controlled trials, decision analysis provides a useful alternative method for comparing these two treatments.
In their recent article, Siriwardana et. al. performed a systematic review of the literature of outcomes associated with synchronous portal-superior mesenteric vein resection during pancreatectomy for cancer.9 The authors pooled data on categories relating to the operation, complications, histopathology, and overall outcome and concluded that even with radical resection, cures are unlikely for patients with tumors involving the portal vein.8
A decision analysis study, such as ours, is not without potential limitations, which we enumerate here. As we have pooled many studies in order to determine baseline probabilities, there is considerable variability within both groups compared in this study, along with some overlap of the two groups in this study. In the surgical group, studies differed in the types of adjuvant and/or neoadjuvant therapies, criteria used for resectability, preoperative imaging studies, curative-resection rates, surgical techniques (e.g., venous reconstructive procedures), extent of resection (e.g., removal of lymph nodes), and tumor location, for example. Similarly, the patient selection and treatment regimens varied widely in the palliative chemoradiotherapy group. To account for this variability in published study data, we completed sensitivity analyses over a wide range of values for the probability of perioperative mortality, probability of histopathologic vein invasion and the probability of an R1 or R2 resection to determine threshold values that would lead to a change in the preferred strategy when traversed.
Decision analysis studies can be limited by paucity of data for specific patient subgroups. Although there are several published reports containing survival data for surgically treated patients based on histopathologically documented invasion status of cancer into the portal and/or superior mesenteric vein, there are fewer reports containing survival data for surgically treated patients (those treated with portal and/or superior mesenteric vein resection) based on documented R-status at the surgical resection margin (Table 3). Furthermore, as far as we are aware, there are no published reports that contain survival data based on both of these variables in the same patient cohort. To overcome this limitation, we elected to analyze survival data using three separate decision trees instead of a single tree. This strategy allowed us to include many more studies than would have been possible for a single decision tree approach.
Finally, as we have noted earlier in our methods and discussion, our study compared two groups of patients that may not be identical with respect to stage distribution, comorbidity profiles, and other variables. For example, although most patients categorized into Group 1 of our study had cancer invasion limited to the portal vein and/or superior mesenteric vein (stage IIA or IIB disease), some of these patients had invasion into the superior mesenteric artery (stage III disease) for which arterial resection was performed.7 For Group 2, the reported data do not permit determination of the distribution of patients among stages IIA, IIB, and III; therefore, it is possible that a greater percentage of Group 2 than Group 1 patients had stage III disease. Nonetheless, patients with nonresected stages IIA, IIB, and III pancreatic adenocarcinoma have virtually indistinguishable 1-year survival rates (25.0%, 26.9%, and 27.0%, for stages IIA, IIB, and III, respectively).8 Given that the primary endpoint of our study was 1-year survival and all patients included in our study had stage IIA, IIB or III disease, we believe that Groups 1 and 2 are comparable for the purposes of this analysis.
In conclusion, we have determined that surgical resection may confer a survival advantage over palliative chemoradiotherapy in select patients with pancreatic cancers with presumed invasion into local veins. We recommend that authors report survival statistics based on histopathologic proven vein invasion and R-status so that a more through evaluation of this data may be completed in the future.