Overall survival of the 158 patients with pancreatic adenocarcinoma who underwent pancreaticoduodenectomy with separate pathologic analysis of second-echelon lymph nodes. At a median follow-up of 65.1 months, the median overall survival was 26.5 months. The 5-year actuarial overall survival rate was 27.3%.
Overall survival of 158 patients with pancreatic adenocarcinoma who underwent pancreaticoduodenectomy with separate pathologic analysis of second-echelon lymph nodes, stratified by lymph node status. Lymph node status was the sole predictor of long-term outcome. The 82 patients with positive first- or second-echelon lymph nodes had a median survival of 23.1 months compared with 35.7 months for the 76 patients with negative lymph nodes (P = .04).
Overall survival of 82 patients with positive first- and/or second-echelon lymph nodes. The 65 patients with positive first-echelon lymph nodes had a median survival of 23.9 months compared with 23.1 months for the 17 patients with positive second-echelon lymph nodes (P = .57).
Timothy M. Pawlik, Eddie K. Abdalla, Carlton C. Barnett, Syed A. Ahmad, Karen R. Cleary, Jean-Nicolas Vauthey, Jeffrey E. Lee, Douglas B. Evans, Peter W. T. Pisters. Feasibility of a Randomized Trial of Extended Lymphadenectomy for Pancreatic Cancer. Arch Surg. 2005;140(6):584–591. doi:10.1001/archsurg.140.6.584
The required sample size of a prospective randomized trial comparing standard pancreaticoduodenectomy with pancreaticoduodenectomy plus extended lymphadenectomy for pancreatic adenocarcinoma is prohibitively large, making such a trial infeasible.
Retrospective cohort study.
Comprehensive cancer center.
We identified 158 patients who underwent pancreaticoduodenectomy for pancreatic adenocarcinoma with separate pathologic analysis of second-echelon lymph nodes, defined as lymph nodes along the proximal hepatic artery and/or the great vessels.
Main Outcome Measures
To estimate the sample size required for a randomized trial, we devised a biostatistical model with the following assumptions: extended lymphadenectomy can benefit only patients who (1) actually have disease removed from second-echelon nodes, (2) have microscopically negative (R0) primary tumor resection margins, and (3) do not have visceral metastatic (M0) disease.
Seventy-six patients (48.1%) had negative first- and second-echelon lymph nodes, 65 (41.1%) had positive first-echelon and negative second-echelon lymph nodes, and 17 (10.8%) had positive first- and second-echelon lymph nodes. Patients with positive second-echelon lymph nodes had an R0 resection rate of 47.1%. At a median follow-up of 65.1 months, 4 patients with positive second-echelon lymph nodes were alive, but 3 had recurrent disease. This implies that only 1 patient (5.9%) with positive second-echelon lymph nodes may have had true M0 disease. Therefore, only 0.3% of patients (10.8% with positive second-echelon lymph nodes × 47.1% with R0 resection × 5.9% with M0 disease) may achieve a survival benefit from extended lymphadenectomy. A randomized trial of standard pancreaticoduodenectomy vs pancreaticoduodenectomy with extended lymphadenectomy would require 202 000 patients in each study arm to detect such a small difference.
Definitive evaluation of the potential benefits of extended lymphadenectomy would require a prohibitively large sample size. Adequately powered randomized trials to address the potential benefit of extended lymphadenectomy are infeasible.
In Japan and in some Western treatment centers, there has been a general belief that more extensive surgery may improve outcome for patients with localized, operable pancreatic adenocarcinoma.1- 5 Specifically, extended lymphadenectomy has been used in addition to pancreaticoduodenectomy in an effort to improve pathologic staging and increase survival in patients undergoing surgical resection. Initial retrospective reports from centers in Japan1,2 suggested that 5-year overall survival rates in patients treated with pancreaticoduodenectomy plus extended lymphadenectomy were higher than those in patients treated by pancreaticoduodenectomy with standard lymphadenectomy. Subsequent prospective randomized trials performed in Europe3 and the United States4 failed to confirm a survival benefit from extended lymphadenectomy. Nevertheless, several investigators5- 9 have continued to advocate further randomized trials to better define the role of extended lymphadenectomy for operable pancreatic cancer.
The implementation of phase 3 trials requires substantial patient, institutional, and financial resources. Before embarking on additional randomized trials, the feasibility of performing a definitive randomized trial of extended lymph node dissection should be evaluated. In particular, attention needs to be focused on the frequency of involvement of second-echelon lymph nodes (ie, those nodes removed as part of the additional lymphadenectomy specimen), the macroscopically and microscopically complete resection rate, and the true frequency with which patients who undergo pancreaticoduodenectomy have no visceral micrometastatic disease at the time of laparotomy. Using surgical pathologic findings and outcomes data from a large cohort of patients treated at a single institution, we sought to estimate the required sample size and thereby evaluate the feasibility of a prospective randomized trial comparing standard pancreaticoduodenectomy with pancreaticoduodenectomy plus extended lymphadenectomy for pancreatic adenocarcinoma.
Between December 7, 1988, and December 16, 2002, three hundred four patients with pancreatic adenocarcinoma were identified from the Department of Surgical Oncology database as having undergone pancreaticoduodenectomy at the University of Texas M. D. Anderson Cancer Center, Houston. In 158 patients (52.0%), second-echelon lymph nodes—those nodes typically included as part of an extended lymphadenectomy as performed in previously reported randomized trials3,4—were dissected and removed as a distinct specimen from the primary tumor (ie, separate from the duodenum, pancreas, and first-echelon nodes) at the discretion of the operating surgeon. These patients are the subject of the current article. In the remaining 146 patients (48.0%), second-echelon lymph nodes were either not removed or not identified as such by the operating surgeon or pathologist. This study was approved by the University of Texas M. D. Anderson Cancer Center institutional review board.
The surgical technique used for standard pancreaticoduodenectomy has been described elsewhere.10 Standard pancreaticoduodenectomy included clearance of all soft tissues and lymphatics immediately to the right of the superior mesenteric artery, as well as removal of the lymphoareolar tissue along the proximal hepatic artery. First-echelon lymph nodes were defined as the peripancreatic and pancreatoduodenal lymph nodes that were included as part of the main pancreaticoduodenal specimen. Second-echelon lymph nodes were defined as nodes adjacent to the proximal hepatic artery and/or nodes within the retroperitoneum anterior to the inferior vena cava or aorta. The lymph node(s) anterior to the hepatic artery are commonly removed to allow identification of this vessel during the portal dissection; the aortocaval nodes can be separately excised after performing Kocher and Cattell-Braasch maneuvers to expose the great vessels and their anterior lymphoareolar tissues. These 2 lymph node groups (common hepatic artery and aortocaval) may be removed as a separate specimens allowing pathologic analysis.
The resected specimens were processed according to a standard protocol.11 or, more recently, according to the procedures and recommendations of the sixth edition of the American Joint Committee on Cancer Staging Manual (AJCC).12 The retroperitoneal (RP) margin was defined as the soft tissue margin directly adjacent to the proximal 3 to 4 cm of the superior mesenteric artery. Early in our experience, the RP margin was evaluated by microscopic examination of an en face section measuring approximately 2 mm in thickness; the margin was recorded as positive (R1, microscopically positive following a gross complete resection) or negative (R0, microscopically negative) for carcinoma. For the past 5 years, the RP margin was evaluated following recommended American Joint Committee on Cancer pancreatic cancer staging principles.12 The RP margin was recorded as positive (R1) or negative (R0) for tumor; and if negative, the distance in millimeters from the tumor to the inked margin was recorded. An R0 resection was defined as a macroscopically complete resection with microscopically negative surgical margins, an R1 resection was defined as a macroscopically complete resection with microscopically positive surgical margins, and an R2 resection was defined as a macroscopically incomplete resection.
A clinicopathologic biostatistical model was devised to estimate the required sample size of a prospective randomized trial of pancreaticoduodenectomy with and without extended lymphadenectomy for pancreatic adenocarcinoma.13
The biostatistical model used the following assumptions:
R0 resection is required for extended lymphadenectomy to confer a possible survival benefit. That is, as for other gastrointestinal tract cancers, an R0 resection is required for potential long-term survival.
Only patients who have pathologically involved second-echelon lymph nodes can benefit from extended lymphadenectomy. This premise is based on the fact that removal of pathologically negative lymph nodes is unlikely to confer a therapeutic effect.
Only patients who have involved second-echelon lymph nodes without visceral metastatic disease (M0) may derive a potential survival benefit from the removal of additional pathologically positive lymph nodes. Lymphadenectomy is unlikely to provide a survival benefit in patients who have visceral metastatic disease.
Using these assumptions, the fraction of patients undergoing pancreaticoduodenectomy who might derive a survival benefit from extended lymph node dissection can be defined by the following equation:
FDPB = FN2 × FR0 × FTM0,
where FDPB is the fraction of patients who may derive a potential benefit; FN2 is the fraction of patients undergoing second-echelon nodal dissection who have positive second-echelon lymph nodes; FR0 is the fraction of patients undergoing pancreaticoduodenectomy who undergo R0 resection and have positive second-echelon lymph nodes; and FTM0 is the fraction of patients undergoing pancreaticoduodenectomy who have positive second-echelon lymph nodes but truly have M0 disease.
All data are presented as median values or numbers and percentages of patients. A univariate test (χ2) was used to identify variables that might affect survival. Factors that appeared to be significantly associated with survival were entered into a Cox proportional hazards model to test for significant effects, while adjusting for multiple factors simultaneously. Survival was measured from the date of diagnosis. Actuarial survival was calculated using the Kaplan-Meier method. Differences in survival between groups were examined using the log-rank test. P<.05 was considered statistically significant.
Table 1 lists the clinicopathologic features of the 158 patients with pancreatic adenocarcinoma who underwent pancreaticoduodenectomy with separate pathologic analysis of second-echelon lymph nodes. The median age of the patients was 64 years (age range, 39-81 years). Preoperative chemotherapy or chemoradiation was administered to 110 patients (69.6%). Fifty-two patients (32.9%) underwent vascular resection and reconstruction as part of their pancreaticoduodenectomy. The median operative time was 9.1 hours (range, 4.4-15.7 hours), and the median estimated blood loss was 1000 mL (range, 150-5100 mL). Two patients died within 30 days of resection (one of sepsis and the other of gastrointestinal tract hemorrhage) for a perioperative mortality rate of 1.3%.
Final pathologic analysis revealed a median tumor size of 3.0 cm (range, 0.2-6.0 cm), and most tumors were either moderately (n = 83 [52.5%]) or poorly (n = 43 [27.2%]) differentiated. With regard to pathologic nodal status of the 158 patients, 76 patients (48.1%) had negative first-and second-echelon lymph nodes, 65 patients (41.1%) had positive first-echelon and negative second-echelon lymph nodes, and 17 patients (10.8%) had positive first- and second-echelon lymph nodes. No patient had positive second-echelon lymph nodes without positive first-echelon lymph nodes. Most patients (n = 126 [79.7%]) had an R0 resection. If both first- and second-echelon lymph nodes were negative, the R0 resection rate was 89.5%. If first-echelon lymph nodes were positive and second-echelon lymph nodes were negative, the R0 resection rate was 84.6%. If both first- and second-echelon lymph nodes were positive, the R0 resection rate was 47.1% (P<.001).
At a median follow-up of 65.1 months, the median overall survival for the 158 patients who had separate pathologic analysis of second-echelon lymph nodes was 26.5 months, and the 5-year overall survival rate was 27.3% (Figure 1). The median overall survival of the 146 patients who did not have separate analysis of second-echelon lymph nodes was similar (24.7 months; P = .91; data not otherwise presented). Among the 158 patients who underwent pancreaticoduodenectomy with separate pathologic analysis of second-echelon lymph nodes, the longest-living survivor was alive and disease-free at 11.7 years of follow-up. Tumor size, presence of vascular invasion, and a history of vascular resection and reconstruction were not associated with survival (all P>.05). Patients with a negative RP margin tended to have a longer median overall survival (28.1 months) than patients with a positive RP margin (18.4 months), although this difference did not reach statistical significance (P = .13). Lymph node status was the sole significant predictor of long-term survival. Patients with positive first- or second-echelon lymph nodes had a median overall survival of 23.1 months, compared with 35.7 months for patients with negative lymph nodes (P = .04) (Figure 2). The 5-year overall survival rates for patients with node-positive vs node-negative disease were 20.0% and 32.1%, respectively. Survival was similar for patients with positive first-echelon lymph nodes and patients with positive second-echelon lymph nodes (P = .57) (Figure 3). At last follow-up, 4 (23.5%) of the 17 patients with positive second-echelon lymph nodes were alive, but 3 had recurrent disease (Table 2). This implies that only 1 patient (5.9%) with positive second-echelon lymph nodes may have had true M0 disease at the time of pancreaticoduodenectomy.
The percentage of patients potentially benefiting from extended lymph node dissection was calculated using the following values: FN2, 10.8%; FR0, 47.1%; and FTM0, 5.9%. Thus, FDPB = 10.2% × 47.1% × 5.9% or 0.3%. Therefore, 3 in 1000 patients treated by pancreaticoduodenectomy plus extended lymph node dissection might derive a potential survival benefit from the extended lymphadenectomy.
The sample size needed for a randomized clinical trial to detect a survival benefit (if any) for patients undergoing extended lymphadenectomy was estimated as follows. We assumed a 100% long-term survival rate for patients with positive second-echelon lymph nodes and M0 disease who underwent pancreaticoduodenectomy plus extended lymphadenectomy. With this assumption, the improvement in survival for the entire population would be only 0.22%. This estimate was based on a 5-year overall survival rate of 27.30% for the standard resection group compared with 27.52% for the extended lymphadenectomy group (assuming that 99.7% of patients undergoing extended lymphadenectomy have a 5-year survival rate of 27.30% and the other 0.3% with positive second-echelon lymph nodes and M0 disease have a 5-year survival rate of 100%). To detect such a difference of 0.22% between standard pancreaticoduodenectomy and pancreaticoduodenectomy with extended lymphadenectomy using a log-rank test with a 2-sided significance level of .05 and 80% power would require 202 000 patients in each study arm. A multicenter phase 3 trial with an accrual rate of 200 patients per year would take 1010 years for patient accrual.
The present study is, to our knowledge, the first to use clinicopathologic data to estimate the feasibility of conducting a randomized trial to evaluate extended lymph node dissection for pancreatic adenocarcinoma. Our data demonstrate that in a population of patients with localized pancreatic adenocarcinoma treated with pancreaticoduodenectomy plus extended lymphadenectomy, only 3 in 1000 patients might derive a survival benefit from the extended lymphadenectomy, assuming that such a benefit exists. While the derived trial sample size and accrual duration are instructive, the most important finding of the current study may be that extended lymphadenectomy would result in only a 0.22% potential incremental improvement in survival compared with standard lymphadenectomy. In fact, this marginal survival benefit probably represents an overestimation as it was based on the assumption of 100% long-term survival of patients with positive second-echelon lymph nodes and M0 disease treated with extended lymphadenectomy. The 0.22% difference in potential survival does not remotely approach the more commonly accepted incremental clinical benefit of 10% to 20% used for the design of most clinical trials evaluating new cancer treatments.
Clinical reports (prospective or retrospective) that include data for patient cohorts treated by pancreaticoduodenectomy and extended lymphadenectomy with separate pathologic analysis of the extended lymphadenectomy specimens are extraordinarily helpful in clinical trial planning. Data from 2 specific reports4,5 can be compared with those in the current study. Henne-Bruns et al5 reported results of a prospective nonrandomized study of 46 patients who were treated with pancreaticoduodenectomy and extended radical retroperitoneal lymphadenectomy. The extended lymphadenectomy included removal of all lymphatic, connective, and neural tissue along the left side of the superior mesenteric artery and along the aorta between the inferior mesenteric artery and the diaphragm. Only 4 patients (9%) were found to have metastatic disease in the separately analyzed second-echelon lymph nodes. All 4 patients also had involved first-echelon lymph nodes. In a study by Yeo et al4 a separate dissection with pathologic analysis of the RP lymph nodes was performed as part of a single-institution randomized trial of pancreaticoduodenectomy with (n = 56) or without (n = 58) extended RP lymphadenectomy. Extended lymphadenectomy included antrectomy and removal of all lymphoareolar tissue along the great vessels between the inferior mesenteric artery and the renal vessels. Among the 58 patients who had extended lymph node dissection, 10% had metastatic carcinoma in the resected RP lymph nodes, and none of those 10% had RP nodes as the only site of lymph node involvement.4 Thus, given our findings and those of the other studies that involved separate pathologic analysis of second-echelon lymph nodes, it is reasonable to conclude that approximately 10% of patients presenting with operable pancreatic adenocarcinoma harbor disease in the second-echelon lymph nodes. The corollary of this finding is that 90% of patients who undergo pancreaticoduodenectomy with extended lymphadenectomy cannot derive a therapeutic benefit from the procedure regardless of their surgical margin or metastatic disease status.
It could be argued that the excision of second-echelon lymph nodes in the current article more closely approximates lymph node sampling than extended lymph node dissection. In addition, we acknowledge the potential selection bias in identifying our study population that may increase the frequency of positive second-echelon lymph nodes. However, we did not seek to evaluate the potential clinical benefit of extended lymphadenectomy in our patient population but rather sought to determine the frequency of involvement of second-echelon lymph nodes in an effort to quantify the potential therapeutic benefit of extended lymphadenectomy. The fact that our rate of positive second-echelon lymph nodes (10.8%) is virtually identical to the rates observed in the prospective reports by Yeo et al4 and Henne-Bruns et al5 suggest that the lymph node excision approach used in the current study was able to accurately stratify patients with regard to second-echelon nodal status. In aggregate, these data suggest that a 10% frequency of positive second-echelon lymph nodes is a reasonable figure to use for clinical trial planning.
Our clinicopathologic biostatistical model assumes that only patients who undergo R0 resection can benefit from removal of additional pathologically positive lymph nodes. This assumption is largely an extrapolation from general principles of gastrointestinal surgical oncology that espouse R0 resection as necessary for cure. Our data demonstrate that this goal is less frequently achieved in patients with positive second-echelon lymph nodes. In fact, the R0 resection rate in patients with positive second-echelon lymph nodes (47.1%) was almost half the R0 resection rate seen in patients with negative lymph nodes (89.5%) or only positive first-echelon lymph nodes (84.6%) (P<.001). Our finding of an association between which nodal compartment was positive (first echelon vs second echelon) and the R0 resection rate was not reported in the 2 previous trials4,5 that included separate pathologic analysis of second-echelon lymph nodes. This association suggests that pancreatic adenocarcinoma may be a more locally advanced or infiltrative disease in patients with metastasis to the second-echelon lymph nodes.
Another important finding was that virtually all patients who have positive second-echelon lymph nodes also have occult metastatic disease at the time of primary tumor surgery. Indeed, of the 17 patients with positive second-echelon lymph nodes in our study, only 1 remains alive without recurrent disease 49 months following the diagnosis of pancreatic cancer; this patient received chemoradiation (350 mg/m2 of gemcitabine and 300-rad [30-Gy] external beam radiation) prior to pancreaticoduodenectomy. It is, therefore, reasonable to infer that this patient may not have had visceral metastatic disease at the time of surgery. Our M0 rate of approximately 6% is comparable to other estimates in the literature. The single-institution reports of 144 and 132 patients with node-positive pancreatic adenocarcinoma treated with pancreatic resection at The Johns Hopkins University14 and Memorial Sloan-Kettering Cancer Center15 documented 5 (3.5%) and 3 (2.3%) long-term survivors, respectively. It is reasonable to believe that these long-term survivors also had no visceral metastatic disease at the time of primary tumor resection. Thus, for purposes of clinical trial planning, one can expect that only 5% of patients presenting with stage IIB pancreatic cancer (surgically resected, node positive) do not have occult micrometastatic disease.
Although it appears that there may not be a clinically meaningful survival benefit associated with extended lymphadenectomy, it is possible that extended lymphadenectomy improves staging in patients with operable pancreatic adenocarcinoma. However, in our patient population, there were no patients with positive second-echelon lymph nodes who did not also have positive first-echelon lymph nodes. In other words, no patient was identified as having skip metastases, and, thus, no patient was up-staged by the examination of second-echelon lymph nodes. Yeo et al4 and Henne-Bruns et al5 similarly found that the addition of extended lymphadenectomy did not alter the pathologic stage of any patient. Rather continuous lymphatic spread was found in all patients (ie, if second-echelon lymph nodes were positive, first-echelon lymph nodes were also positive), thereby confirming earlier reports that skip metastases to distant lymph nodes are rare in pancreatic adenocarcinoma.16,17 Given the aggregate data, one can make a compelling argument that examination of second-echelon lymph nodes does not result in improved pathologic staging of operable pancreatic cancer. Furthermore, from a clinical treatment perspective, this issue is largely irrelevant because, at present, decisions regarding adjuvant treatment are generally based on the histologic diagnosis of pancreatic adenocarcinoma and not on the presence or absence of positive lymph nodes.
Our findings demonstrate that extended lymphadenectomy is unlikely to offer a clinical benefit beyond that achievable with standard pancreaticoduodenectomy for pancreatic adenocarcinoma. On the basis of data derived in this report, as well as information from other studies that provided separate pathologic analysis of second-echelon lymph nodes, it is clear that reasonably powered clinical trials to evaluate a putative survival benefit of pancreaticoduodenectomy with extended lymphadenectomy are infeasible. These clinicopathologic data and their secondary implications for trial design were unknown at the time of the 2 previously published clinical trials.4,5 Sadly, pancreatic adenocarcinoma is a systemic disease at the time of diagnosis in virtually all patients; therefore, it is unlikely that clinically significant improvements in survival will result from more extensive surgery. From a public health perspective, the limited resources that currently exist for investigation of pancreatic adenocarcinoma should be directed to the regionalization of care, the study of emerging techniques for early diagnosis, and the development of novel multimodality treatment strategies.
Correspondence: Peter W. T. Pisters, MD, Department of Surgical Oncology, Unit 444, University of Texas M. D. Anderson Cancer Center, PO Box 301402, 1515 Holcombe Blvd, Unit 444, Houston, TX 77030 (firstname.lastname@example.org).
Accepted for Publication: February 22, 2005.
Previous Presentation: This paper was presented at the 112th Scientific Session of the Western Surgical Association; November 10, 2004; Las Vegas, Nev; and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript.
Fabrizio Michelassi, MD, New York, NY read by Michael Farnell, MD, Rochester, Minn: Data from 2 retrospective studies conducted in Japan in the late 1980s suggested that extended lymphadenectomy in addition to pancreaticoduodenectomy improves long-term survival in patients with adenocarcinoma of the head of the pancreas. Since then, 4 prospective randomized studies have been performed in an attempt to confirm this observation. These studies were conducted in Europe, Japan, and the United States. Two were carried out at a single institution and 2 were multicenter studies. None has confirmed a survival benefit conferred by extended lymphadenectomy; 2 have indicated that the performance of an extended lymphadenectomy may worsen the quality of life by increasing the chances of postoperative severe diarrhea.
Dr Pisters and colleagues have reviewed their 15-year experience at M. D. Anderson Cancer Center and, applying an elegant mathematical formula, have concluded that extended lymphadenectomy associated with pancreaticoduodenectomy at best benefits only 3 in 1000 patients. This is yet another piece of evidence suggesting that an extended lymphadenectomy is not beneficial in the treatment of pancreatic adenocarcinoma and adding to the body of literature suggesting that extended lymphadenectomy does not confer any additional advantage over adequate lymphadenectomy in the surgical treatment of gastrointestinal tract cancers.
I have 3 questions: The median operative time was 9.1 hours. This is 3 hours longer than the median operative time in any of the 4 prospective randomized studies. How do you explain this? Does preoperative chemoradiation therapy used in 70% of the patients increase the difficulty of the case and, therefore, the length of the procedure?
In 80% of patients the surgical margins were negative. This rate is in line with the experience derived from the prospective randomized studies done to date. Yet, the percentage of patients requiring a vascular resection was much higher at 33%. Why do you think this is the case? Is preoperative chemoradiation therapy responsible for local changes, which may increase the likelihood of a vascular resection?
Median blood loss was 1000 mL. Some authors have suggested that intraoperative blood loss is an important determinant of long-term survival. Did you look at the influence of this variable on outcome?
Dr Pisters: Dr Michelassi’s comments, as usual, are quite insightful. They relate predominately to technical rather than oncologic issues and I will address them systematically.
Fabrizio’s first question relates to the operative time, which, we agree, is certainly longer than reported by other investigators. Notwithstanding this, our operative mortality rate and morbidity rate are comparable. The increase in operative time seen in our patient population, I believe, is owing to a number of factors. Certainly, it is possible that the frequent use of preoperative radiation treatment may increase preexisting tumor and stent-related fibrosis in the peripancreatic tissue planes complicating the dissection somewhat and thereby adding time to the procedure. A second factor, which undoubtedly increased operative time in our patient population, is the frequent use of vascular resection. As seen in this data set and in other reports from our institution, vascular resection, usually of the SMV [superior mesenteric vein] or portal vein, is performed in 1 of every 3 patients undergoing pancreatic resection at our institution. That adds considerable complexity to the procedure as all of you know and unequivocally adds additional operative time. A third factor I think contributing to the operative times at our institution is the fact that tremendous attention is paid to the RP dissection. This is a point that has been made quite articulately by my partner, Doug Evans, on a number of different occasions and will be the subject of an upcoming publication from our group looking at our entire experience. This portion of the operation, we believe, is the most oncologically significant part of pancreatectomy and although some surgeons try to complete this portion of the procedure expeditiously using a series of quickly applied clamps or even staplers, our approach to this part of the operation is quite difficult and involves sharp dissection along the periadvential plane of the SMA [superior mesenteric artery] with dissection of all lymphoareolar tissue to the right of the SMA. This undoubtedly adds time and when performed in a teaching setting with a fellow certainly adds further time to the procedure. I think in aggregate these 4 factors largely account for the difference in operative time observed in this report compared with that observed in our series.
The second issue Fabrizio brings to light is the vascular resection rate. This is a frequently asked question when we present our data in a number of different forums. The vascular resection rate of about 30% to 33% is certainly higher than that seen in other centers, and I think this is owing to a number of factors. First, referral bias. Because of our publication track record and experience, I know that we are being referred a number of patients from other centers who are specifically referred to our center because of our experience with vascular resection and reconstruction. That, in and of itself, results in a selection bias, which increases the vascular resection rate in our patient population. It is certainly possible, as Fabrizio proposes, that the use of radiation treatment may in some way increase the rate of vascular resection, probably due to obliteration of tissue planes and a lowering of the threshold for vascular resection at the time of surgery simply for patient safety reasons. That is certainly a possibility. When we have looked carefully at the rate of histologic involvement of the vein wall among patients undergoing vascular resection, that rate is about 80% suggesting that the surgeon’s intraoperative assessment of the degree of involvement of the vein wall is really a remarkably accurate estimate of the likelihood of histologic involvement of the adjacent vein. Four of 5 times that we proceed with vascular resection, subsequent pathology reveals histologic evidence of tumor infiltration of the resected vein wall. This would suggest to us that we are probably not performing vascular resection too frequently but that for a variety of reasons discussed, our vein resection rate is higher than that reported by other groups.
The last issue that Fabrizio brings to light is the blood loss and again I believe that the median blood loss of approximately 1 L is also related to the increased frequency of vascular resection and reconstruction. We and others have demonstrated that when vascular resection and reconstruction is performed as part of pancreatectomy, blood loss is certainly increased and transfusion requirements are increased compared with patients undergoing a more conventional standard pancreaticoduodenectomy. We did not analyze this operative factor among the clinical and pathologic factors that we evaluated and, for a variety of reasons, I do not believe that such analysis would alter our conclusions from this article. Indeed, our study did not directly seek to evaluate the role of extended lymph node dissection in our patient population, rather we used our practice of examining second-echelon lymph nodes with separate pathologic analysis to estimate the likelihood of clinical benefit from extended lymph node dissection. The existing literature is limited in this respect. There are very few reports in the literature that separately analyze second-echelon lymph nodes. In our article it was 10%. In the [Johns] Hopkins series, it was also 10%, and in the prospective nonrandomized study from Germany by Henne-Bruns et al, the frequency of N2 (ie, second-echelon lymph node involvement) was 9%. So our conclusion is that our finding of a 10% frequency of involved second-echelon lymph nodes is in line with what has been reported in other reports that involve separate pathologic assessment of lymph nodes. As Tim concluded in his articulate summary of this issue, sadly this disease is one which is systemic at diagnosis in virtually all patients. I think all of us who care for these patients recognize that more surgery is unlikely to result in meaningful clinical benefit for the vast majority of patients with this disease. Surgical resection is certainly an excellent form of palliation for most of these patients but rarely is sufficient for cure.
Margaret Shoup, MD, Chicago, Ill: I think the article confirms many of our biases about proceeding with a clinical trial of this magnitude. You mentioned that you had 110 patients who underwent neoadjuvant therapy. My question to you about this is, did your radiation fields include N2 and if so, did you find an association between nodal status and whether or not they underwent neoadjuvant therapy? My second question is, based on your conclusions that it is unlikely to benefit patients to undergo an extensive lymph node dissection, what is your current policy in patients with pancreatic cancer? Are there any patients on whom you would continue to do an extended lymphadenectomy and if so, how do you decide which patients those would be?
Dr Pisters: Margo, those are excellent questions and let me just address them systematically. First, the use of preoperative chemoradiation. As Tim presented, preoperative radiation therapy was used in approximately 70% of patients in the current study. This certainly could in some way influence the frequency of identification of lymph nodes. It is instructive to look at our experience juxtaposed against other large centers. In this study, we found approximately 50% of patients with node-positive disease. The reports from the Mayo Clinic, Rochester, Minn, and the Hopkins groups show that somewhere closer to 70% of patients have involved lymph nodes at the time of pancreatic resection. It remains open to speculation whether or not this is real difference and whether or not this is in some way due to the use of preoperative radiation therapy. It is certainly possible that the use of radiation treatment decreases the frequency of recovery of positive lymph nodes. The radiation fields used in the era that these patients were treated would certainly include the second-echelon lymph nodes that were excised as part of this study. So these lymph nodes were in the treatment field. As to our current policy for these patients, many observers who are aware of our operative approach to this disease have already pointed out that, in fact, the standard operation that we do involves many of the components of extended lymphadenectomy. The sharp dissection along the periadvential plane of the SMA with exposure of the right lateral surface of the SMA over an 8- to 10-cm length is what would be for many other investigators an extended lymph node dissection. That is the standard operation that we do. We also excise, as most investigators do, the lymph node that extends over the hepatic artery. As most of you know, that is a very large lymph node that is frequently confluent with additional lymph nodes at the celiac axis and along the proximal hepatic artery. We excise that lymph node, not for therapeutic purposes but simply for pragmatic reasons, to facilitate exposure of the hepatic artery and a safe approach to the GDA [gastroduodenal artery]. We do not believe that any form of lymph node dissection beyond that which I described offers therapeutic or staging benefit.
Clive Grant, MD, Rochester, Minn: I share your view of the mechanism of tumor dissemination, but for those surgeons who do consider ELND [elective lymph node dissection] as valuable therapeutically, they might suggest that resection of second-echelon lymph nodes may prevent subsequent systemic dissemination. That is, lymph node metastases metastasize, and in your model the assumption that only the 5% who ultimately survived without distant metastases would benefit actually excludes the very patients that advocates of ELND would suggest might benefit. Could you address the question that, I think, is underlying this entire theory, do lymph node metastases metastasize?
Dr Pisters: Clive, that is an excellent point that you bring up and one which has been discussed in the context of addressing the tumor biology of many different solid tumors. I think if we rely exclusively on our data to answer your question, I believe the patterns of failure observed in these patients are instructive. As Tim presented, among the 17 patients with involved second-echelon lymph nodes, only 1 patient has survived long-term. That patient, I would point out parenthetically, has been followed for 49 months. The Mayo Clinic group, the Hopkins group, and the Sloane-Kettering Memorial group have pointed that many of these patients who are alive and disease free—ostensibly cured of pancreatic adenocarcinoma 60 months following surgery—go on to recur and die more than 5 years following surgery. So I would not say that the book is closed on that specific patient. In the remaining patients, if we look at the pattern of failure in those 16 patients, local recurrence was the first site of failure in only one patient. The remaining 16 patients all recurred in visceral sites or in the peritoneal cavity, undoubtedly from occult peritoneal carcinomatosis present at the time of surgery. So I think that the hypothesis that improved lymph node clearance and excision of pathologically positive lymph nodes may result in improved local control and thereby reduce the risk for subsequent dissemination of tumor to visceral sites is not well supported by our data set or other data sets in the literature.
If we look at our aggregate patient population at M. D. Anderson, considering the fact that our results should be interpreted with the caveat that most patients have received preoperative radiation treatment, local failure as a first site of treatment failure occurs in about 7% of patients. It is a very uncommon event and those of you who care for patients with pancreatic cancer know that most patients fail in distant sites undoubtedly from occult metastatic disease present at the time of diagnosis.