Pancreatic cancer palliation: Using tumor stage to select appropriate operation

Pancreatic Cancer Palliation: Using Tumor Stage to Select Appropriate Operation Terence P. Wade, MD, Todd J. Neuberger, MD, Thomas J. Swope, BS, Katherine S. Virgo, PhD, Frank E. Johnson, MD, st. Louis, Missouri

To +sess the effect of tumor stage on the surgical palliation of pancreatic cancer, 350 cancers from 74 U.S. Department of Veteraps Affairs (DVA) hospitals from 1987 to 1991 were staged from pathologic and operative data, then grouped by initial surgery: biliary bypass only (BO) , gastric bypass only (CO), or combined biliary and gastric bypass (BC). Re-operations were recorded as later gastric and/or biliary bypass: Stages I-II (local disease) : BO (n = 52) -6 later gastric ( 12%)) 3 later billater gastric (5%); 3 iary (6%); BG (n = 60)-3 later biliary (5%). Stage III (positive nodes) : BO later gastric (4%); BC (n = (n = 26)-l 35) - 1 later gastrobiliary bypass (3%). Stage IV (metastases) : BO (n = 7 1) -3 later gastric (4%), 3 later biliary (4%); BG (n = 70)-2 later gastrobiliary bypass (3%). CO (all stages) : (n = 41) -1 later gastric (2%)) 4 later biliary ( 10%). Using a two-factor ANOVA comparing survival by stage and original surgery, we found that stage had a significant effect on survival (p = 0.0001) , but the type of initial bypass operation had no effect. Re-operation after palliative pancreatic cancer surgery was necessary in less than 5% of patients with BG. Initial BG reduced the incidence of re-operation for patients with jaundice and without metastatic disease, and may also benefit patients with gastric obstruction alone. Patients with jaundice who have peritoneal or liver metastases can be treated effectively with BO if they have no symptoms of gastric outlet obstruction.

P

ancreatic cancer is a virulent disease that is increasing in incidence in the United States [I]. It frequently does not cause symptoms until late in the course of the disease, and its presentation is most commonly characterized by jaundice, gastric outlet obstruction, and pain, caused by the local invasion of tumor into the bile duct, duodenum, and neural plexi, respectively. Pancreatic resection can be curative, but most patients (80% to 95%) from reported series [2-41 have unresectable disease at diagnosis and require palliative bypass of their biliary and/or gastric obstructions. In some reported series, bypass procedures have a 10% to 14% operative mortality [2,5] and a mean survival of 5.4 to 6.6 months. Thus, the goal of therapy is most often to palliate the jaundice and/or gastric outlet obstruction of incurable pancreatic cancer, using operative and/or endoscopicpercutaneous methods. When operative exploration is performed for unresectable cancer, appropriate procedures involve bypass of the obstructed bile duct or stomach to small bowel, and may also include prophylactic bypass of organs that are not yet obstructed. Our group has previously reported the results of bypass operations for pancreatic cancer in 1,180 patients from the US. Department of Veterans Affairs (DVA) hospitals using computerized records from 1987 to 1991 [q. In this review, we noted that gastrojejunostomy can be added to biliary bypass without increasing operative morbidity or mortality, and that the performance of combined biliary and gastric bypass decreased the frequency of re-operation from that required after either type of bypass performed alone. Although the treatment of pain and cachexia is an important therapeutic goal, surgical therapy has been less effective for these problems and was not a part of the present study. The present review includes patients from the earlier series in whom we could obtain sufficient operative and pathologic records from local DVA tumor registrars to allow pathologic staging of these patients’ cancers, in order to examine the effect of tumor stage on the selection of appropriate palliative operations. PATIENTS AND METHODS

From the Departments of Surgery, John Cochran Department of Veterans Affairs Medical Center, and St. Louis University, St. Louis, Missouri. Requests for reprints should be addressed to Terence P. Wade, MD, St. Louis University School of Medicine, Department of Surgery, 3635 Vista Avenue, St. Louis, Missouri 63 110-0250. Presented at the 34th Annual Meeting of The Society for Surgery of the Alimentary Tract, Boston, Massachusetts, May 17-19, 1993.

208

THE AMERICAN JOURNAL OF SURGERY

We previously described in detail [7] the methods used to compile this series of patients with pancreatic cancer. In brief, admission data from the national DVA hospital system for fiscal years 1987 to 1991 were abstracted at hospital discharge for patients with a diagnosis of pancreatic cancer. Days of hospitalization, diagnoses, complications, and mortality data on all patients with suspected or known pancreatic cancer were compiled from the DVA Patient Treatment Files, using the

VOLUME 167 JANUARY 1994

PANCREATIC CANCER PALLIATION

International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnostic codes of 157.0 to 3, and 157.5 to 9, which identify patients with pancreatic cancer. These codes specifically excluded 27 patients with diagnoses of pancreatic islet cell tumors or lymphoma (ICD-9-CM code 157.4) from the study, allowing us to confine the analysis to patients coded with pancreatic adenocarcinomas. Patients who had a diagnostic code for distal bile duct malignancy (ICD-9-CM = 156.1) or ampullary adenocarcinoma (ICD-9-CM = 156.2), in addition to a pancreatic cancer code, were not excluded from the analysis because of the apparent difficulty in separating these tumors both clinically and histologically from pancreatic adenocarcinoma. Bypass procedures for gastric and/or biliary obstruction were defined as gastric and/or biliary anastomosis to small intestine without pancreatic resection. Other patients who had a diagnosis of pancreatic cancer but did not undergo operative bypass were excluded at this point in the analysis, resulting in 1,180 patients with ICD-9-CM coding data indicating that palliative operative bypass procedures had been performed for pancreatic cancer. The Patient Treatment Files also identified the DVA hospital where the bypass was performed. Information regarding each patient was then requested from the tumor registrar of the treating DVA institution, to include either a copy of their American College of Surgeons/Commission on Cancer registry sheet, or cop ies of operative notes, pathology reports, and discharge summaries that would allow tumor staging to be performed using the American Joint Committee on Cancer (AJCC) TNM method [8] (Table I). Although in-hospital mortality was available from Patient Treatment Files data, mortality after discharge was not. The Beneficiary Identification and Records Location System (BIRLS) contains records of all veterans whose beneficiaries applied for death benefits at the time of the veterans’ deaths, whether the death occurred outside the hospital or within DVA or other hospitals. The date of death is recorded for each veteran in whose name benefits are requested, and BIRLS was searched to the compilation date of “15 Dee 92.” A recent comparison of mortality recording systems estimated that 80% to 89% of all veterans’ deaths are recorded by BIRLS [9], but we have found that recording was more efficient for patients recently treated for pancreatic cancer in the DVA hospital system [7]. The BIRLS mortality information agreed with Patient Treatment Files records when both were available, and there was no suggestion that requests for benefits varied significantly across treatment groups. The records of the U.S. Social Security Administration were then reviewed for a death record in those patients without a BIRLS death record. Patients without a record of death in any of these three data systems were assumed to be alive. Patients with gastric or biliary diversion without mention of enteric anastomosis (i.e., gastric or biliary intubation) were excluded, as were any patients who underwent pancreatic resection. Operations were categorized by initial type of bypass performed (GO = gastroenterostomy THE AMERICAN

TABLE I Pancreatic Cancer Staging, Using the 1992 TNM Classlflcatlon [8] Tumor(T) Tl No direct extension of the primary tumor beyond the pancreas T2 Limited direct extension to duodenum, bile duct, or stomach T3 Advanced direct extension (“fixation”) Nodal involvement (N) NO Regional lymph nodes not involved Nl Regional lymph nodes are involved Metastases beyond regional nodes (M) MO No distant metastases Ml Distant metastases present TNM staging system Stage I Tl-2, NO, MO Stage II

T3,NO, MO

Stage III

Tl-3, Nl, MO

Stage IV

Tl-3. NO-I, Ml

Cancer is confined to the pancreas Direct extension into adjacent tissues Regional nodal involvement present Metastatic disease present

only, BO = anastomosis between biliary tract and small intestine only, BG = gastroenterostomy plus bilio-enterostomy). The timing of later re-operations (including percutaneous or endoscopic biliary intubations) was tabulated in days after the initial operation, recorded as a later gastric and/or biliary procedure, and categorized by initial tumor stage. The factors of m-operation rate, days to re-operation, complications, and survival were also stratified by initial surgery and stage. x2 analysis was used to compare demographic characteristics and the presence of complications across treatment categories. Analysis of variance (ANOVA) was used to compare continuous data such as age and survival. Date of treatment was used as a proxy for the date of diagnosis in calculating survival. RESULTS From the list of 1,180 patients whose files carried a diagnosis code for pancreatic cancer and a procedure code for biliary and/or gastric bypass, staging information was sought from institutional tumor registrars, and 466 replies were received from 74 hospitals. Patient and tumor records were sufficient to allow 356 cancers to be staged. A death record was conlirmed in 91% of these patients whose cancers were staged, thus making followup nearly complete; 90% of staged patients (n = 320) had a pathologic confirmation of pancreatic adenocarcinoma, while three patients had a diagnosis of bile duct or ampullary cancer, two had cystic tumors, and one was of duodenal origin. These six tumors with more unusual pathologies were excluded from the following analysis. Mean survival was calculated from the date of the operation; then results were categorized by tumor stage and by the type of bypass performed. In general, mean survival exceeded 200 days for patients without systemic

JOURNAL

OF SURGERY

VOLUME

167

JANUARY

1994

209

TABLE II Mean Days of Survival, (No. of Patients), and 30-Day Mortality (%) of Patients With Pancreatic Cancer After Palliative Bypass, Grouped by Stage* Stage Bypass Biliaty only Gastric only Both bypassed

I

326 (20) 10% 648 (3) 0% 177 (13) 30%

Wean survival rates after onginal bypass operation are presented, pancreatic

cancer stage at operation.

Number

Stage II

Stage Ill

Stage IV

260 (31) 10% 22 (4) 75% 265 (47) 13%

248 (28) 8% 52 (3) 0% 260 (34) 9%

121 (89) 21% 75 (31) 16% 124 (68) 20%

and the percentage

TABLE III Re-Operations After Palliative Surgery for Pancreatic Cancer, Grouped by Stage and Initial Bypass Performed* Later Gastric (%)

initial Bypass

Later Biliary (%)

Stages I-II BO (n = 51) BG (n = 60)

6 (12) 3 (5)

3 (6) 3 (5)

Stage Ill BO (n = 26) BG (n = 34)

1 (4) 0

0

Stage IV BO (n = 69) BG (n = 69)

3 (4) 2 (3)

3 (4) 0

1 (2)

4 (10)

GO bypass [all stages]? (n = 41) 80 = biliary only bypass;

BG = combined

1 (3)

biliary and gastric bypasses;

GO =

gastric only bypass.

lRe-operations

include

both operative

bined biliary and gastric procedures

and percutaneous

are recorded

procedures.

Com-

here as both gastric and biliary

re-operations. Too

few patients

with GO bypass were present

to allow a proper

analysis

by

stage.

TABLE IV Days from Initial Bypass to Re-Operation, Grouped by Initial Bypass and Stage Initial Bypass

Stages I-II

Stage Ill

Stave IV

Biliaty onty Biliary and gastric

264 (9)* 72 (4)

50 (1) 146 (1)

130 (5) 150 (2)

29 (1)

21 (2)

59 (2)

Gastric only

of patients

who died within 30 days of operation

of patients in each group is listed in parentheses.

is also listed, each grouped

by

,

The highest risk of re-operation was incurred by patients who had cancers ranging in involvement from stage I to stage II and who underwent BO bypass as their initial operation, as 12% (6 of 51) required subsequent gastric bypass and 6% (3 of 51) underwent a second biliary bypass procedure (Table III). Patients with stage III or IV cancers causing biliary obstruction had a 3% to 4% incidence of any later bypass procedure, regardless of the initial bypass procedure chosen. Biliary bypass procedures were equally distributed between cholecystoenteric and choledochoenteric anastomoses, both as the initial choice of biliary bypass and also for those bilioenteric anastomoses that required re-operation. Choice of bile duct or gall bladder for bilioenteric anastomosis did not influence survival, complication, or re-operation rate. Reoperation after initial gastric bypass alone was not analyzed by stage, because only 5 re-operations were required in the 41 GO patients. Despite the short mean survival after initial GO, 4 of these 41 patients (10%) underwent a re-operative biliary bypass procedure before death (Tables II and III). Patients who underwent re-operation after an initial BO bypass did not have a reduction in overall survival when compared by stage to those patients who underwent an initial BG bypass. The mean time to re-operative gastric bypass after initial BO bypass of stage I and stage II cancers was 264 days, more than 100 days longer than the mean time to re-operation after any other type of initial bypass or stage (Table IV). Complication rates did not differ significantly by initial bypass type. Patient survival after re-operation averaged 136 days and was not affected significantly by stage or by the initial or re-operative procedure employed.

‘No. of patients in each group is listed in parentheses

COMMENTS but was reduced by about half when these metastases were present (Table II). Using a two-factor ANOVA comparing survival by stage and original surgery, we found that stage had a significant effect on survival (p = O.OOOl),but the type of initial bypass operation had no effect. Thirty-day operative mortality (and complication rate) for BO was 15% (24%), for GO 20% (32%), and for BG 17% (22%). These results did not differ significantly among groups.

metastases,

210

THE AMERICAN

JOURNAL OF SURGERY

Since the selection of palliative operation for incurable pancreatic cancer is a serious measure, many authors have reported their results, and their goal for patients is clear: to achieve life-long relief of biliary and gastric obstruction while minimizing operative complications and in-hospital time. When a patient presents with a biliary obstruction only, the value of gastric bypass in the prevention of later gastric obstruction must be balanced against the possible increase in operative time and complications incurred by creating a gastrointestinal anasto-

VOLUME 167 JANUARY 1994

PANCREATIC CANCER PALLIATION

mosis. Similarly, although patients who present with gastric obstruction and without jaundice are rare, might they benefit from prophylactic biliary bypass or do the risks outweigh the potential benefits? The question of prophylactic gastric bypass in patients who have jaundice has been frequently studied, but no consensus exists, although this topic has attracted ardent advocates [ 10,l I] and opponents [ 12,131. Advocates of prophylactic bypass note a high incidence (over 2%) of later gastric obstruction requiring re-operation, while op ponents cite an increased operative morbidity and mortality rate (25% to 29%) when prophylactic gastric bypass is performed. We believe that this lack of consensus is due, at least in part, to the statistically small numbers of patients in the previous studies and the lack of stratification of patients by tumor stage. Since the DVA data set allows access to such a large number of patients over a period of only 5 years and the tumor registrars’ data allowed the cancers of 350 patients to be staged, we were able to analyze these practical issues with considerable statistical weight. In addition to the large numbers of patients, these results represent the efforts of hundreds of surgeons in hospitals throughout the U.S. who serve a well-described patient population in close cooperation with their primary care physicians. We believe that this report of their results reflects those achieved in hospitals across the nation, and is less likely than prior reports to be influenced by the unavoidable selection bias inherent in academic institutions with sophisticated subspecialist support. Single-institutional series must rely on the compilation of cases over long periods of time or must allow for the limitations of small patient numbers, which may well be as informative about local surgical skills and referral patterns as about the questions studied. The present study compiled enough patients and information to analyze pancreatic cancer palliation by stage of disease at presentation, which, since the first modem description of pancreatic cancer staging [14], has been shown to be the most important influence on survival, regardless of treatment. Unfortunately, neither advocates nor opponents of prophylactic gastric bypass in patients who have jaundice have stratified their outcomes by stage, which makes it impossible to properly compare previous results to those of our present report. However, although our overall reoperation rate of 8% (29 of 350) is low in comparison to recent reviews [2], the present study makes a strong argument for prophylactic gastric bypass (when resection is unsuitable) in patients who are jaundiced and who have pancreatic cancer ranging from stage I to stage II, because the operative mortality and morbidity rates with BO or BG bypass in this large series were similar. Although the overall rate of re-operation after BO bypass (11%; 160f 146)andBGbypass(5%,8of 163)iscomparable to those of prior reports, the gastric re-operative rate of 12% after BO bypass in stage I to stage II cancer is 2 to 3 times that seen herein with BO bypass for stage III to stage IV disease (4%) or the rate with BG at any stage (4% to 5%). THE AMERICAN

Apparently, patients with regional or systemic metastases who underwent BO bypass did not survive long enough to require later re-operations at a statistical rate greater than that after BG bypass (4% to 5%), since these patients’ mean survival of 156 days (stage III = 248 days, stage IV = 121 days) was generally less than the mean time to a later gastric bypass after initial BO (205 days). Similarly, prophylactic biliary bypass should be considered for patients presenting with gastric obstruction only from pancreatic cancer, since the incidence of patients requiring later biliary procedures was 10%. Mowever, even this large study did not have a sufficient number of patients to determine the effect of tumor stage on this choice. This study also offers insight on another topic of controversy in pancreatic cancer palliation: the use of percutaneous stents to replace operative biliary-intestinal anastomosis for the palliation of jaundice. Since 12% of patients without regional or systemic metastases who had an initial BO bypass required a later gastrointestinal anastomosis, these patients should undergo a prophylactic gastric bypass if resection of their cancer is not performed. Since nonoperative biliary intubations offer no opportunity to perform gastrointestinal anastomosis, and thus avoid a later operation, these less invasive methods should be considered in patients without gastric obstructive symptoms who have documented metastatic cancer and are thus unlikely to survive long enough to require a later gastrointestinal bypass. Patients with only local or regional diise will likely survive a long time (200 to 400 days); our analysis indicates that they are best served with a permanent operative bypass of both their bile duct and stomach if pancreatic resection is not performed. In summary, BG bypass is the most effective operation for the palliation of pancreatic cancer and should be performed in patients who present without systemic metastases. In contrast, patients with systemic metastases can be treated effectively with biliary bypass only if they have no symptoms of gastric outlet obstruction, since these patients rarely survive long enough to require later re-operation. The choice of bile duct or gall bladder for biliary anastomosis has no effect on complication, mortality, or reoperation rates. The recent avaiIability of large data sets can answer questions with statistical significance that have been puzzling surgeons for years, and we welcome comparisons of other large-group outcome studies to the present report. As data entry criteria become more stringent and widely used and further information (laboratory results, drug administration, nursing care requirements, and so forth) can be extracted from computers, we expect that detailed treatment recommendations will be supported by data from these broad, national experiences. REFERENCES 1. Haddock G, Carter DC. Aetiology of pancreatic cancer. Br J Surg 1990; 77: 1159-66. 2. Watanapa P, Williamson RCN. Surgical palliation for pancreatic cancer: developments during the past two decades. Br J Surg 1992; 79: 8-20.

JOURNAL

OF SURGERY

VOLUME

167

JANUARY

1994

211

3. Sarr MG, Cameron JL. Surgical palliation of unresectable carcinoma of the pancreas. World J Surg 1984; 8: 906-18. 4. Warshaw AL, Fernandez-de1 Castillo C. Pancreatic carcinoma. N Engl J Med 1992; 326: 455-65. 5. Sarr MG, Cameron JL. Surgical management of unresectable carcinoma of the pancreas. Surgery 1982; 91: 123-33. 6. Neuberger TJ, Wade TP, Swope TJ, Virgo KS, Johnson FE. Palliative operations for pancreatic cancer in the hospitals of the U.S. Department of Veterans Affairs: 1987-1991. Am J Surg. In press. 7. Wade TP, Virgo KS, Radford DM, Johnson FE. Treatment of pancreatic cancer in the U.S. veteran: complications and outcomes. J Am Co11 Surg. In press. 8. Beahrs OH, Henson DE, Hutter RVP, Kennedy BJ, editors. American Joint Committee on Cancer. Manual for staging of cancer. 4th ed. Philadelphia: JB Lippincott, 1992: 109-13. 9. Boyle CA, Decoufle P. National sources of vital status information: extent of coverage and possible selectivity in reporting. Am J Epidemiol 1990; 131: 160-8. 10. Sarr MG, Gladen HI, Beart RW Jr., van Heerden JA. Role of gastrocnterostomy in patients with unresectable pancreatic carcincma. Surg Gynecol Obstet 1981; 152: 597-600. 11. Singh SM, Reber HA. Surgical palliation for pancreatic cancer. Surg Clin North Am 1989; 69: 599-611. 12. Weaver DW, Wiencek RG, Bouwman DL, Walt AJ. Gastrojejunostomy: is it helpful for patients with pancreatic cancer? Surgery 1987; 102: 608-13. 13. Holbrook AG, Chester JF, Britton DC. Surgical palliation for pancreatic cancer: will biliary bypass alone suffice? J Royal Sot Med 1990; 83: 12-4. 14. Hermreck AS, Thomas CY, Friesen SR. Importance of pathologic staging in the surgical management of adenocarcinoma of the exocrine pancreas. Am J Surg 1974; 127: 653-7.

DISCUSSION BernardLanger (Toronto, Ontario, Canada): One of

the problems with retrospective reviews is that of selection bias for treatment, and if one selects for treatment those patients who one expects to do better, then the outcome is biased by that selection process. Although the data are interesting, a prospective study (from a large number of institutions) is needed to prove the value of your treatment algorithm. Palliation, which, in the medical context, means “affording relief’ to the patient, was measured by survival and re-operation rate. This is an area where quality-oflife studies are going to be far more important than mortality and survival figures, or re-operation rates. Can you tell us if you have attempted to assess patients’ quality of life? Terence P. Wade: I agree completely that prospective data would be much preferable to the retrospective data we were able to compile. The reason that we have performed this study is to first try to find the questions that we can later answer more appropriately in a prospective manner. The endoscopists palliate a large number of patients with biliary obstruction. Patients with gastric outlet obstruction and pain require palliative measures. This study tried to determine options for treatment based on the stage of the disease. Stage IV disease should compare endoscopic versus operative palliation. Stage I and stage II disease should compare resection versus palliative operation. 212

THE AMERICAN

JOURNAL OF SURGERY

The quality-of-life issues can only be addressed in ; prospective fashion. Larry C. Carey (Tampa, FL): Since there’s som controversy about using the common duct or the gallblad der, did you look at the type of biliary bypass to see if i made any difference? Was there any increased hospital stay in patients whc had gastric bypass? Terence P. Wade: Comparing choledochojejunos tomy with cholecystojejunostomy (each performed il half the patients), the complication and re-operation rate were not significantly different. There was no difference in hospital stay comparing those who had biliary and gastric bypasses performer with those who only had a biliary bypass performed. WI haven’t yet analyzed it by stage or compared gastric resection with gastrojejunostomy. John L. Cameron (Baltimore, MD): Was there an; difference in episodes of cholangitis between your biliar: bypass procedures that used the gallbladder and those that used the common hepatic duct? Terence P. Wade: No differences were reported as 2 complication in these patients’ hospital admissions or ir the number of hospital admissions that followed theil initial operation, but this was a computer-generates study. Richard H. Bell, Jr. (Cincinnati, OH): For patient! with stage I and stage II disease, the operative mortalit] was 15% to 20%. Were the patients with stage I and stage II disease who underwent bypass at a higher risk thar were the patients with stage I and stage II disease wh< underwent resection? Why did patients in both of thest disease stages undergo bypass instead of resection? Dc you have any insight into the causes of mortality in thal relatively good risk group? Terence P. Wade: Patients with stage I and stage 11 disease who underwent a resection had an operative mor tality rate of about 5%. The mortality was higher foi those who underwent a bypass; the surgeons involved in the care of these patients selected them for bypass because it was believed that these patients were not safe candidates for resection. They were at a higher operative risk. Andrew L. Warshaw(Boston, MA): Your conclusion that patients should perhaps undergo both the biliary and gastric bypasses is based on the fact that a significant number of patients return for a second operation for gastric bypasses if you did not perform both the first time. Can you defend your last conclusion, which is that patients who need a gastric bypass only should also undergo a biliary bypass? If the bile duct is not dilated, there is a higher operative risk and, later, postoperative risk of stenosis. If the biliary bypass is not performed but later the patient becomes jaundiced, palliation by nonoperative means is always possible. TerenceP. Wade: The last conclusion was more of a suggestion than a conclusion. These patients do have a higher incidence of re-operation. Your point that they can also receive palliation nonoperatively is well taken. Cholecystojejunostomy, which has been shown in other studies to be equal in efficacy to choledochojejunostomy, can

VOLUME 167 JANUARY 1994

be performed regardless of biliary dilatation and should be considered as a prophylactic bypass. I don’t feel as strongly about biliary bypass in the face of gastric obstruction as I do about gastric obstruction or gastric prophylaxis in the face of biliary obstruction in patients with early-stage disease. John F. Stremple (Pittsburgh, PA): In the March issue of Annuls of Surgery, we published an overview based on 830,000 patients in the public (VA) and private sector. We have another 550,000 patients being reviewed. The Annals of Surgery article presents an overview of 309 operations. We need to look at specifics as you have, Dr. Wade. To the Association of VA Surgeons, Dr. Longo presented 400 cases of anal carcinoma, with details on 207 of the patients. The Veterans Affairs’ data base is extremely powerful and should be tapped into for similar studies that Dr. Wade has just presented, Terence P. Wade: The data base has important strengths, as we’ve noticed. It also has some significant weaknesses. We have very little information between the time of patients’ hospitalization and the time of death, upon which to make any conclusions, and that is why looking at pancreatic cancer was a good place to start. It is more difficult to decide the efficacy of the treatment for anal cancer, because most of the patients survive for a long time. John L. Cameron: How were the gastrojejunostomies performed? Dr. Lillemoe has recently reported 118 patients with unresectable pancreatic cancer. Virtually all of them had posterior gastrojejunostomy, an operative procedure that, in the past, has been thought inappropriate for treatment of pancreatic cancer because, theoretically, the cancer could obstruct the anastomosis. There was virtually no delay in gastric emptying when it was performed posteriorly, and none of them became ob structed. TerenceP. Wade: From the computerized data base, we cannot distinguish, for example, anterior from posterior gastrojejunostomy. L. William Traverse (Seattle, WA): What percent of your patients had the tumor in the head of the pancreas? Neoplastic gastropathy may account for the problems with gastric bypass in some of your patients who were operated on late. When we analyzed our results from endoscopic stenting versus operative biliary bypass with or without gastric bypass, we found an intriguing result. The patients with an operative bypass, with or without gastric bypass, developed gastric outlet obstruction more commonly than those patients who had not had an operation but an endoscopic bypass. Would you tell us the incidence of gastric outlet obstruction after operative procedures versus solely endoscopic procedures? Terence P. Wade: Again, the way we structured the data base does not allow us to extract that information. Murray F. BIT+MZUI (New York, NY): We’ve reviewed and published a similar number of cases from a

single institution, and any form of prophylactic bypass, whether biliary or gastric, was associated with increased hospital morbidity and prolonged stay. The choice of choledochojejunostomy versus cholecystojejunostomy made little consequential difference to the patient. You stated that the incidence of pancreatic adenocarcinema is increasing. What is your evidence? The tail of your survival curve seems high. Since some patients may have an error in the pathologic diagnosis, have you been able to review the pathology in those patients who postoperatively survived longer than 3 years? TerenceP. Wade: The increasing incidence is cited in other papers, but the overall mortality and incident curves are fairly flat. It is increasing as a cause of death in American society because we’re having more success with the other cancers. Proportionately, pancreatic cancer is a growing problem. We have some error in our data that is seen in the tail of the curve, in that not all patients who died had a death benefit requested either from the Social Security Administration or from the Veterans Administration. I think that that’s a more readily acceptable explanation for the survivors in the long-tail segment than some of the other reasons you have quoted. John L. Cameron: The incidence of pancreatic cancer has increased for three or four decades, up until about 1980. Since 1980, the incidence has remained level. Lawrence W. Way (San Francisco, CA): Having worked in Veterans Affairs hospitals myself for many years, I’m concerned about the accuracy of the information that is input into the computer. Have you made any attempt to independently validate the accuracy of the information? Did the patients receive the operation that is logged into the computer? TerenceP. Wade: Requiring staging data allowed us to validate the information. We obtained information from tumor registrar records, from operative notes, or from hospital discharge summaries. We found that all of the patients whose conditions were coded as “pancreatic cancer” and who had a pancreatic resection had pathologic information confirming that they had cancer. Kenneth Warren (Boston, MA): In patients with inoperable carcinoma of the pancreas or periampullary area, Dr. Kate11and I reported on a very small series of patients in whom a biliary, gastric, and pancreatic duct bypass was performed when the pancreatic duct was very large. It did appear to improve the digestion and comfort of these patients if they were having pancreatic pain after eating, but it did not add to their survival. Did any of your patients have a pancreatic anastomosis? Terence P. Wade: There were some physicians who had obviously read your paper, and there were some patients who did have pancreatic duct anastomosis and the other two bypasses performed, as well. We don’t have any information about the quality of these patients’ survival; therefore, we did not include them in the analysis.

THE AMERICAN JOURNAL OF SURGERY

VOLUME 167 JANUARY 1994

213