Web site http://www.eusimaging.com
Survival advantage of combined chemoradiotherapy compared with resection as the initial treatment of patients with regional pancreatic carcinoma
An outcomes trial
Article first published online: 13 NOV 2000
Copyright © 2000 American Cancer Society
Volume 89, Issue 2, pages 314–327, 15 July 2000
How to Cite
Snady, H., Bruckner, H., Cooperman, A., Paradiso, J. and Kiefer, L. (2000), Survival advantage of combined chemoradiotherapy compared with resection as the initial treatment of patients with regional pancreatic carcinoma. Cancer, 89: 314–327. doi: 10.1002/1097-0142(20000715)89:2<314::AID-CNCR16>3.0.CO;2-V
- Issue published online: 13 NOV 2000
- Article first published online: 13 NOV 2000
- Manuscript Revised: 10 MAR 2000
- Manuscript Accepted: 10 MAR 2000
- Manuscript Received: 28 JUN 1999
- pancreatic carcinoma;
- endoscopic ultrasonography
Resection of pancreatic carcinoma is resource-intensive with a limited impact on survival. Chemotherapy and/or radiotherapy (RT) have been shown to be effective palliation. To examine whether preoperative chemoradiotherapy as the initial treatment improves survival for patients with a regional pancreatic adenocarcinoma with a minimal chance of being resected successfully, an outcomes trial was conducted.
Patients with radiologically regional tumors were staged by laparotomy and/or computed tomography followed by endoscopic ultrasonography, angiography, and/or laparoscopy. Those with locally invasive, unresectable, regional pancreatic adenocarcinoma initially were treated with simultaneous split-course RT plus 5-fluorouracil, streptozotocin, and cisplatin (RT-FSP) followed by selective surgery (Group 1). Patients determined to have a resectable tumor initially underwent resection without preoperative chemoradiotherapy, with or without postoperative chemoradiotherapy (Group 2).
Over 8 years 159 patients presenting with nonmetastatic pancreatic adenocarcinoma were administered RT-FSP or underwent surgery for resection. Group 1, comprised of 68 patients initially treated with RT-FSP, had a 0% mortality rate within 30 days of entry. In 20 of 30 patients undergoing surgery after RT-FSP, tumors were downstaged and resected. Group 2, comprised of 91 patients who initially underwent successful resection, had a 5% mortality rate within 30 days of entry. Postoperatively, 63 of these patients received chemotherapy with or without RT. The median survival for Group 1 was 23.6 months compared with 14.0 months for Group 2 (P = 0.006) despite more advanced disease cases in Group 1. Survival favored RT-FSP regardless of whether lymph nodes were malignant. The dominant prognostic factor of earlier stage pancreatic carcinoma having an expected survival advantage was reversed by the initial nonoperative treatment.
Based on a reversal of the expected trend that patients with earlier stage resectable carcinoma (T1,2, N0,1, M0) who undergo removal of their tumors survive longer than patients with more advanced regional disease (T3, N0,1, M0), survival was found to improve significantly for patients reliably staged as having locally invasive, unresectable, nonmetastatic pancreatic adenocarcinoma when initially treated with RT-FSP. Cancer 2000;89:314–27. © 2000 American Cancer Society.
Pancreatic carcinoma accounts for 5% of cancer deaths in the United States. Between 1930 and 1980, the incidence of pancreatic carcinoma doubled and stabilized.1 Although partial or total pancreatectomy to remove the tumor is traditionally considered the patient's only chance for cure, less than 10% of patients have resectable lesions. Furthermore, less than 10% of patients undergoing resection are reported to survive 5 years.2–5
Various combinations of radiation therapy and/or radiopotentiator chemotherapy have been shown to have some effect for palliation,6–21 but the role of combined chemoradiotherapy in patient management has not been defined, and practice patterns are inconsistent.9 Whipple pancreatoduodenectomy is a resource-intensive treatment that has not been proven to significantly improve survival. A recent comprehensive review cast additional doubt on the efficacy of surgery both for initial treatment by resection as well as for diagnosis by laparotomy.22 Also, accurate diagnosis and staging can be accomplished for almost all patients without surgery by appropriate combinations of intravenous contrast-enhanced computed tomography (CT), endoscopic ultrasonography (EUS), fine-needle aspiration (FNA), and laparoscopy.2, 23, 24 Nonoperative palliation for obstructive jaundice is also safer and at least as effective as surgery.25
To examine the outcome on survival of a combined chemoradiotherapy regimen6 formulated for regional, nonmetastatic pancreatic adenocarcinoma, we designed a study in which patients with locally invasive, unresectable, regional pancreatic adenocarcinoma (a tumor classified as not resectable) were treated initially with simultaneous split-course radiotherapy plus 5-fluorouracil (5-FU), streptozotocin, and cisplatin (RT-FSP) with or without subsequent resection and compared with patients with resectable disease who underwent surgical resection (all visible tumor could be resected) without preoperative chemoradiotherapy, with or without subsequent postoperative chemoradiotherapy.
All patients included were diagnosed with regional, pancreatic adenocarcinoma. The date of diagnosis was established either when pathologic examination of pancreatic tissue showed carcinoma or when EUS confirmed a mass with features characteristic of adenocarcinoma,26–28 subsequently confirmed pathologically. Prospective groups of patients used for analysis included patients diagnosed with regional, unresectable carcinoma who started the RT-FSP protocol, patients undergoing resection of pancreatic carcinoma at 2 hospitals (MSH and SMH), and patients evaluated at one EUS center and referred to a specialized center for pancreatobiliary carcinoma surgery. Survival was analyzed for consecutive patients treated from January 1, 1989 to January 1, 1997. Informed consent was obtained from all subjects before initiation of treatment. Treatment protocols had been approved by the hospital internal review board.
Patients excluded from analysis were those with any detectable metastatic (M:1) carcinoma, prior pancreatic resection, or other prior carcinoma within 5 years. Patients with ampullary, bile duct, duodenal carcinoma, or with tumors characterized as cystic, endocrine, sarcoma, lymphoma, pseudotumorous pancreatitis, or benign adenomas were excluded because these lesions have a more favorable prognosis and alter survival curves of pancreatic adenocarcinoma. Before starting the RT-FSP chemoradiotherapy protocol, patients with regional disease were excluded if the carcinoma was so advanced that it could not be encompassed within the RT field, or it was associated with varices or nonregional lymph nodes visible on CT. These patients received a different protocol (methotrexate, leucovorin, 5-FU, and cisplatin [MLFP]). Once eligibility was determined, no patient was excluded from the primary analysis unless adenocarcinoma was not subsequently confirmed pathologically. Survival outcome was analyzed both excluding and including these patients with regional cancer who either did not have a pathologic diagnosis or received MLFP.
Pretreatment Pathologic Tumor Staging
Tumors were staged initially by physical examination and CT. To exclude metastases and confirm local invasion and resection status by standard criteria,28–34 we performed EUS, repeat CT, angiogram, laparoscopy and/or laparotomy. A tumor was defined as unresectable if (1) metastases beyond regional lymph nodes were noted, (2) malignant lymph nodes were noted out of the resection field, (3) invasion of major blood vessels was clearly demonstrated nonoperatively, or (4) at laparotomy significant vessel invasion was noted making resection unlikely without vessel resection and reconstruction.
Staging was according to the TNM classification of the American Joint Committee on Cancer.35 The absence of residual cancer in the resected specimen after chemoradiotherapy was defined as a complete pathologic response (Stage 0). The tumor was Stage I (T1 or T2, N0, M0) if the pancreatic tumor was localized and involved only the pancreas, bile duct, and/or duodenum. If the tumor invaded regional vessels and lymph nodes were free of tumor, it was Stage II (T3, N0, M0). A tumor was Stage III if neoplastic disease was present in regional lymph nodes (T1, T2, or T3, N1, M0). Metastases extending beyond regional lymph nodes defined Stage IV disease (any T, any N, and M1).
Patients eligible for initial treatment with RT-FSP had regional, invasive, unresectable pancreatic carcinoma able to be encompassed within a radiotherapy field extending greater than 3 cm beyond the tumor and greater than 1.5 cm around the porta hepatis and regional lymph nodes. For the radiotherapy regimen, a linear accelerator was used to deliver 5400 centigrays (cGy) in 3 split courses. Initially, 200 cGy/day on Days 1–5 and 8–12 were administered, followed by 200 cGy/day on Days 29–33 and 36–40. On Day 56, an additional 1400 cGy were given in 7 equal fractions, 3 to the initial field and 4 to a smaller field with a newly measured 3-cm margin. The 5-FU was given every 28 days as a continuous 1-g/m2/24-hour infusion over a period of 108 hours beginning 8–16 hours before the first 200 cGy of RT and ending 4–8 hours after the fifth 200 cGy of RT. Streptozotocin (Zanosar; Pharmacia, Peapack, NJ) was given monthly as a morning 4-hour 300-mg/m2 infusion, piggy back, on the first 3 days of treatment without interruption of the 5-FU infusion. Cisplatin (Platinol; Bristol-Meyers Squibb Oncology, Princeton, NJ) 100 mg/m2 in 250 mL of saline with 12.5 g mannitol, was given monthly over 30 minutes immediately after the third dose of streptozotocin, also without interruption of the 5-FU infusion. Before and after cisplatin, patients were hydrated, typically with greater than 1000 mL of saline given over 2 hours with diuretics to induce diuresis, while giving antiemetics. Dosage schedules were modified based on each individual's tolerance to the drugs. After 3 courses of RT-FSP, additional chemotherapy was administered until there was no evidence of disease for more than 1 year after resection. Subsequent treatment initially consisted of leucovorin 200 mg/m2 and 5-FU 500 mg/m2, every 2 weeks with an escalation of dosage to 1000 mg/m2 of 5-FU based on clinical evidence of tolerance or relapse. Because this study was anticipated to take many years, after the initial 3 courses of the RT-FSP protocol, which generally took 3–4 months, subsequent chemotherapy varied in such a way as to maximize clinical outcome according to the current “state-of-the-art.” Various combinations of drugs such as high dose methotrexate, bolus and infusion 5-FU, and hydroxyurea were administered depending on clinical course and drug availability.
Although recommended for primary resection patients, specific postoperative chemotherapy and radiotherapy was not part of the initial stratified, treatment protocol. Because of limited available data demonstrating benefits of postoperative chemotherapy and radiotherapy for all stages of regional disease,4, 8 treatment was given according to each patient's risks and physicians' protocols, followed by analysis of this nonstratified variable.
Monitoring and Assessment of RT-FSP Patients
Group 1 patients initially were monitored and assessed with a complete blood count with platelets done twice weekly during the chemotherapy week of each course and in following weeks as needed. Complete chemistries were obtained on Days 1, 3 (before treatment with cisplatin), 5, and 8. CT and tumor markers (CEA, CA 19-9, and OC-125) were obtained before treatment began, after 10 weeks, and every 3 months thereafter. Additional chemistries and CT were performed if the patient deteriorated clinically. Initial response and restaging was assessed by CT and tumor markers. World Health Organization (WHO) Grade 3 or 4 toxicity events were recorded.
For refinement of staging, EUS evaluation with the Olympus GF-UM echoendoscope system (Olympus America, Lake Success, NY) was performed by the same endosonographer (H.S.) in most RT-FSP treated patients demonstrated by CT to have a downstaged tumor based on a decrease in bidimensional tumor mass by greater than 50%, or to less than 2 cm. Standard EUS scanning images and criteria for the pancreas and surrounding structures were used.36–38 Angiography was performed in some patients to further assess or confirm resection status.33 If downstaging to resectability was demonstrated,33, 39 patients underwent laparoscopy, laparotomy, and, if possible, resection. Early in the study, surgery was performed on patients with unresectable disease according to EUS, to confirm the nonoperative diagnosis and/or stage.
Definition of Variables for Assessment of Surgery
Perioperative mortality was defined as death after surgery and before hospital discharge or in less than or equal to 30 days. Major complications were defined as death, reoperation, anastomotic leak, fistula, wound dehiscence, hemorrhage, infections, primary cardiopulmonary events, gastrointestinal obstruction, ischemic events, and hematologic events. Delayed gastric emptying also was noted. Length of hospital stay after tumor resection was calculated using the day of surgery as Day 1.
Survival was measured from the date of diagnosis to the date of death or July 1, 1997, the censor date. Survival was examined using the Kaplan–Meier method. Statistical comparisons were conducted by the log rank and Wilcoxon methods. For a two-sided alpha error of 5% and a power of 80% (a 1.7-hazard rate ratio), the required number of patients observed until death was 48 in one group and 72 in the other group. Univarient analysis and Cox regression multivarient analysis were performed. The Cox regression analysis method was used to determine the significance of the group effect after adjustment for lymph node status.
Subjects were allocated into one of the two treatment groups according to the stage of the pancreatic carcinoma assessed by the pretreatment pathologic tumor staging protocol. An analysis of the intention-to-treat population was applied to the data from subjects that consented to treatment, demonstrated by receiving at least the initial treatment for each respective group, i.e., the first course of the RT-FSP protocol (Group 1) or a tumor resection (Group 2). Interim survival analysis was performed and trial enrollment stopped when the required number to detect a 1.7-hazard rate ratio was met. Thirty-day mortality was defined as death within the first 30 days after initiation of the initial treatment.
From January 1, 1989 to January 1, 1997, pathologically confirmed regional pancreatic adenocarcinoma was diagnosed for 159 consecutive patients who were treated initially either with RT-FSP or with surgical resection. Group 1 consisted of 68 patients who were not candidates for tumor resection; the tumor was staged as locally invasive and unresectable.27–33 They were treated initially with RT-FSP. Group 2 consisted of 91 patients with tumors staged as resectable that were initially treated by pancreatoduodenectomy (standard Whipple procedure) in 76 patients, distal pancreatectomy in 10, or total pancreatectomy in 5. Patient characteristics are shown in Tables 1 and 2. No patient was lost to follow-up through the censor date of July 1, 1997.
|Characteristic||RT-FSP Group 1||Resected Group 2|
|No. of patients||68||91|
|Site of carcinoma (%)|
|Presenting symptoms (%)|
|Median age (yrs/range)||64/40–86||66/28–85|
|Median duration of onset of symptoms to the time of diagnosis (mos)||2.0||1.5|
|Stage/class||RT-FSP Group 1 % (no. of patients)||Resected Group 2 % (no. of patients)|
|I||0 (0)||40 (36)|
|II||49 (33)||15 (14)|
|III||51 (35)||45 (41)|
|T:1||0 (0)||31 (28)|
|T:2||0 (0)||39 (36)|
|T:3||100 (68)||30 (27)|
|N:0||49 (33)||55 (50)|
|N:1||51 (35)||45 (41)|
Ten additional patients with regional, unresectable disease who were possible candidates for RT-FSP were excluded from Group 1. Six had extensive advanced regional disease and therefore were treated with MLFP, which initially emphasizes systemic treatment more than the RT-FSP protocol. Four had no pathologic confirmation of adenocarcinoma; attempted biopsies did not confirm carcinoma. Diagnosis could only be confirmed by tumor markers, imaging studies, and clinical course progressing to metastatic carcinoma in three of the four.
RT-FSP Group 1
Tumors were located in the head of the pancreas in 53 patients, in the body in 10, and in the tail in 5. Before RT-FSP, adenocarcinoma was staged as an unresectable, locally invasive tumor without distant metastases in 38 patients (56%) by CT, EUS, angiogram and/or laparoscopy; the other 30 (44%) also were staged by laparotomy. Patients with an unresectable tumor without evidence of metastatic disease on both CT and EUS evaluation were included without laparoscopy or laparotomy, even though laparoscopy can find small metastatic lesions missed by nonoperative modalities.
After RT-FSP was initiated, all patients completed initial phases of the first 56 days of the protocol, which included full course radiotherapy in all but 7 patients. A full course of radiotherapy was not received by these patients because of the following: poor tolerance/age (2 patients), refused complete course of RT (3 patients), radiation seeds placed at prior laparotomy (1 patients), and progression to metastatic disease in less than 1 month (1 patient). Six other patients progressed to metastatic disease within 2–3 months for which chemotherapy was modified to emphasize systemic treatment. One other patient refused to complete the third cycle of chemotherapy, electing to undergo a Whipple procedure.
WHO Grade 3 and 4 toxicity for the RT-FSP protocol (Table 3) was noted in 19 patients (27%). No patient in Group 1 was hospital bound. The first mortality in the RT-FSP Group 1 occurred more than 4 months after entry into the study.
|Toxic effect||Grade 3 no. of patients (%)||Grade 4 no. (%)|
|Anemia||5 (7)||1 (1)|
|Nausea/vomiting||4 (6)||1 (1)|
|Diarrhea||1 (1)||1 (1)|
After RT-FSP, 22 tumors were determined to have classification downgraded from an unresectable (T3) to a resectable (T2/T1) mass based on satisfying both CT criteria for tumor resection, and EUS tumor resection criteria31 and/or venogram demonstration of a patent superior mesenteric-portal vein confluence (Figs. 1a,b). In only one of these 22 patients eligible for resection did chemoradiotherapy diarrhea toxicity preclude surgical evaluation. In another three patients, neutropenia or thrombocytopenia postponed resection. One patient developed a gangrenous gallbladder that was removed laparoscopically, postponing resection.
Surgery was performed in 30 patients (44%) after RT-FSP. In 10 of these patients, surgery without resection was performed with no perioperative mortality. Findings at laparotomy were isolated peritoneal metastasis in 2 patients, extensive biopsy negative fibrosis that was not resected in 3, and local unresectable neoplastic tissue in 5, of which 1 was debulked and 3 received intraoperative radiotherapy for unresectable disease detected preoperatively on CT in 1 and EUS alone in 2.
Twenty patients (29%) had a standard field tumor resection. In 18 patients, Whipple resection was performed by dissection of adherent tissue/fibrosis to separate the tumor from any major vessels without resection of the vessel; pyloris was preserved in 5 patients. In two patients, major vessels were resected, one with a distal pancreatectomy, and one with a total pancreatectomy and superior mesenteric-portal vein reconstruction. Pathology specimens showed no evidence of disease (NED) in 6 patients (Fig. 2) of which 2 had significant fibrosis in vessel margins (FIV). Malignant cells in fibrosis were found in 13 patients of which 7 had FIV. Malignant tissue at the vessel-parenchymal margin was found in 1 patient. Perioperative mortality occurred in 2 (10%) on Days 30 and 54. Major complications occurred in 4 patients (20%) and included reoperation (1), anastomotic leak (1), fistula (1), hemorrhage (2) [gastrointestinal (1) and intraabdominal (1)], infectious (4)[sepsis (1), abscess (2), and catheter-related fungal infection (1)]. Four patients (20%) had delayed gastric emptying due to gastroparesis/atony or ileus. Median length of hospital stay after tumor resection was 13 days (range, 5–54 days).
Surgical Resection Group 2
Pancreatic tumors were located in the head in 77 patients, in the body in 4, and in the tail in 10. Patients underwent resection without preoperative chemoradiation. There were 5 30-day mortalities after entry into the study. In 27 patients (30%), the tumor was reclassified as “resectable T3 disease” because either resection pathology specimens had malignancy present to the margin in 13 patients, of which 7 required dissection of adherent tissue/fibrosis and 6 required peeling tumor/fibrosis off major vessels, or resection was performed with major vessel resection in 13 patients or with organ resection in 1. The operating surgeon noted probable residual malignant tissue that was not removed in two patients. Of the 13 patients in which major vessel resection was performed, 3 were related to vessel injury, 1 to superior mesenteric artery anomaly, 3 to a small area of invasive adherent tumor noted only after most of the resection had been completed, and 6 to the distal pancreatectomy. Vascular reconstruction was performed on the portal vein in 3 patients, superior mesenteric vein in 2, and superior mesenteric artery in 2.
Perioperative mortality occurred in 6 patients (7%), 5 in less than 30 days, 1 on Day 88. Major complications occurred in 25 patients (27%) and included: reoperation (5), anastomotic leak (2), fistula (1), wound dehiscence (1), hemorrhage (7) [gastrointestinal (5) or intraabdominal (2)], infectious (12) [sepsis (4), intraabdominal abscess (4), pseudomembranous colitis (2), and wound infection (2)], primary cardiopulmonary events (4) [myocardial infarction (1), cardiopulmonary failure (1), pulmonary embolus (1), and acute respiratory distress syndrome (1)], gastrointestinal obstruction (2), ischemic events (3) [small bowel infarct (1), partial liver lobe infarct (1), scrotal gangrene (1)], and hematologic events [disseminated intravascular coagulation (1)]. Four (4%) had delayed gastric emptying due to gastroparesis/atony or ileus. Median length of hospital stay after tumor resection was 12 days (range 5–88 days).
Postoperatively, 63 patients (69%) received chemotherapy with or without radiotherapy. Thirty-nine patients received 5-FU with RT (12 with additional chemotherapy). Twenty-four patients received 5-FU alone or with additional chemotherapeutic agents. Nine received chemotherapy only after evidence of recurrence. Treatments were delayed more than 3 months for 4 patients or limited in 13 by delayed recovery in 7, poor tolerance in 4, or by refusal in 2. Of the 28 patients that did not receive postoperative treatment, 15 were Stage I, 2 Stage II, and 11 Stage III. Fifteen of 36 (42%) Stage I patients did not receive postoperative treatment. Postoperative therapy was not given to 15 patients (15%) due to surgical complications or delayed recovery. An additional 15 patients initially refused postoperative chemotherapy.
Median survival (Fig 3) for Group 1 was 23.6 months compared with 14.0 months for Group 2 (P = 0.006). On the censor date, 28% (19 of 68) in Group 1 were alive, including 9 of 20 who had had resection of a downstaged tumor. In Group 2, 19% (17 of 91) were alive. Thirty-day mortality for Group 1 was 0% compared with 5% for Group 2. Survival was greater for Group 1 at 1, 2, and 3 years. Through the censor date, the only 3 disease free 5-year survivors were in the RT-FSP Group 1. All three patients initially had unresectable adenocarcinoma diagnosed by prior laparotomy in one, and by unequivocal imaging studies in two; all three were downstaged and underwent resection of residual carcinoma.
Analysis of significant treatment and prognostic factors for Groups 1 and 2 are shown in Tables 4 and 5. In the RT-FSP Group 1, three treatment factors were significant by Cox regression analysis, even after adjusting for each other and other possible prognostic factors, in predicting a favorable response (Table 4): regression of demonstrable mass size on CT, eligibility for resection, and resection of tumor. The subgroup of 20 patients with initially unresectable tumors who completed RT-FSP plus resection had a median survival of 32.3 months (Fig. 4) and the greatest 3-year survival of 32%.
|Variable||n||Median (mos)||2 yrs (%)||3 yrs (%)||P value|
|RT-FSP Group 1|
|Regression of CT mass image||CR,a PRb||46||25.5||52||25||0.001||0.001|
|Downgraded and eligible for resection||CR||22||32.9||69||28||0.001||0.003|
|PR, NR, POD||46||18.7||28||15|
|Resected after RT-FSP||Yes||20||32.3||61||32||0.006||0.055|
|Primary resection Group 2|
|Characteristic||n||Median (mos)||2 yr (%)||3 yr (%)||P value|
|RT-FSP Group 1|
|Staged by laparotomy||Yes||30||32.2||58||27||NS||0.030|
|Malignant lymph nodes||Present||35||16.2||29||18||NS||0.028|
|Primary Resection Group 2|
In the resection Group 2, treatment with postoperative chemotherapy with or without RT resulted in an early survival advantage compared with nontreated Group 2 patients that was no longer evident after approximately 3 years. However, the difference between RT-FSP Group 1 and Group 2 patients that received postoperative treatment was still significant (P = 0.01).
For Group 1, two prognostic factors were significant by the Wilcoxon test, lymph node status (N classification), and staging by laparotomy. There is an early survival advantage no longer evident after approximately 2 years for Group 1 patients diagnosed by laparotomy or staged as having no malignant lymph nodes, including those involved by direct extension, similar to Yeo et al.9 For Group 2, there was a trend towards better long term survival in patients staged as having no malignant lymph nodes (Table 5).
The influence on survival of other factors including tumor location also was analyzed. Median survival for patients with carcinoma of the head was greater than for those with tumors of the body/tail for Group 1, 23.8 months and 19.8 months, respectively (P = 0.42), and for Group 2, 14.1 months and 12.5 months, respectively (P < 0.04). Comparisons between Group 1 and Group 2 significantly favored RT-FSP treated patients with tumors of the head (P = 0.02) and body/tail (P < 0.03). There was no significant difference in patient survival when comparing surgeons, either with adequate sample size separately or when grouped in various categories such as according to location. However, there were wide variations for length of hospital stay and complications.
When Group 1 and Group 2 are compared using only laparotomy staged patients according to the Gastrointestinal Tumor Study Group eligibility criteria15–18, 20 and only ideal surgical patients without malignant lymph nodes are analyzed, survival difference was even greater in favor of RT-FSP over initial resection; survival was 29.7 months for Group 1 surgically unresectable Stage II patients compared with 15.7 months for Group 2 surgically resectable Stage I patients (P = 0.02). Whether or not Cox regression analysis is used to adjust for lymph node classification, a significant effect on survival favoring Group 1 is still present, both including (P = 0.004) or excluding (P = 0.011) 30-day mortalities. Exclusion of patients with resected T3 disease in Group 2 did not affect the results. No other factors listed in Tables 1 or 2 were found to have a significant difference in predicting a more favorable outcome. Inclusion of the 10 patients with regional cancer, excluded from Group 1, did not affect result significance.
This study demonstrates that survival of patients with localized, unresectable pancreatic adenocarcinoma initially treated with RT-FSP, followed by tumor resection only for those whose classification was downgraded, was significantly greater than survival of patients with a resectable carcinoma treated initially with segmental or total pancreatectomy. This difference was found even though many factors that introduce staging, selection, pathologic, and clinical bias into survival outcome favor the resection Group 2 over the RT-FSP Group 1. RT-FSP preceded any later surgery without any significant adverse outcome such as early progression of the tumor. Local invasion of blood vessels by pancreatic carcinoma did not preclude disease free 5-year survival.
For patients with pancreatic carcinoma, advancing stage and declining performance status are recognized as the dominant factors that predict a marked decrease in survival.2, 8, 16–21, 35, 40 A statistically significant greater survival for patients with more advanced stage unresectable disease compared with patients with earlier stage resectable disease never has been reported to our knowledge. Pathologic bias factors favoring Group 2 include the following: all T3 classification patients in Group 1 were unresectable; All T3 classification patients in Group 2 were resectable; and more Stage III patients were in Group 1 (> 51%) than Group 2 (45%). The selection bias factor of not including any early Stage I patients in Group 1 also favors Group 2. For the first time to our knowledge, the expected advantage of an earlier stage as the dominant prognostic factor for pancreatic carcinoma has been reversed by initial nonoperative treatment.
Studies performed in the late 1980s report an 8–15-month median survival for Stage I and II resectable pancreatic carcinoma.1–7, 22, 25, 41–44 For unresectable carcinoma, prognosis has been even worse. Survival for Stage II and III disease is 4–8 months, and 3–6 months for Stage IV disease. The 23.6-month median survival of RT-FSP treated Group 1 patients is substantially greater than that reported in prior studies for patients with similar stage disease for any conventional type of chemotherapy and/or radiotherapy.1–25, 45, 46 Even the median survival of 18.7 months for Group 1 patients whose classification was not downgraded exceeds that reported for most series of patients undergoing Whipple pancreatoduodenectomy.1–5, 22, 25, 42–44, 47–49 RT-FSP remained superior to primary surgery in our multivariant and specific subset analyses, even excluding 30-day mortalities in the resection Group 2. Patients in Group 1 who developed early metastatic disease after initiation of RT-FSP most likely had undetected metastatic disease at the time RT-FSP was initiated. Therefore, this study also indicates that staging by laparotomy probably improves prognosis by creating stage drift towards less advanced disease when compared with nonoperative staging (Table 5). This stage drift most likely results from staging and clinical bias factors such as: finding at laparotomy undetected more advanced disease in some, not surgically exploring those with radiologically locally advanced tumor, and selection of patients with better overall performance status that can undergo surgical staging.
A study from M.D. Anderson Cancer Center10 to our knowledge is the only study similar to ours; it reports on preoperative and postoperative chemoradiotherapy for a large group of patients with regional disease. However, unlike our study, they exclude patients with cancer that is clearly unresectable because of invasion of regional vessels on high quality CT. Spitz et al.10 report on two groups of patients with potentially resectable disease based on CT. Patients were either treated with preoperative chemoradiotherapy and then operated on or operated on and recommended to have postoperative chemoradiation, of which only 76% received the recommended treatment. Almost half (49%) of the group given preoperative chemoradiotherapy were found to require vascular resection to remove the cancer. Patients with this type of resectable vascular invasion (T3 disease) are similar to RT-FSP patients in our study that initially were found on EUS and/or angiography to have locally invasive cancer not observed on CT. Patients with invasive disease on CT are included into our study in Group 1. Vascular resection was performed in only 1 RT-FSP Group 1 patient (5%). Furthermore, the 26% of their patients, without even a mass on CT that nevertheless required vascular resection to remove the cancer, would again have been classified as T3 preoperatively by EUS and/or angiography in approximately 90%50 and would have been included into our RT-FSP Group 1 rather than a primary resection group. The patients that would have been classified as T3 based only on the findings after resection and pathology evaluation would be similar to the patients with “resectable T3 disease” in our resection Group 2. Their data10 confirm the inaccuracy of computerized tomography in determining unresectability of potentially resectable pancreatic carcinoma due to local invasion of major regional vessels, usually of the portal confluence. For these patients, CT has been reported to have only a 50% accuracy,26 which essentially matches the data of Spitz et al.10 based on the 49% of patients with a “potentially resectable, low density mass in the pancreatic head” on initial CT that completed chemoradiation and surgery, but required vessel resection.
Further complicating initial classification is the finding that a significant number of their patients receiving preoperative chemoradiotherapy must have had T3 disease that was downstaged to a tumor that could be resected without vessel resection, because they report finding malignant lymph nodes less frequently (46%) in the group that they report had “more advanced disease” and received preoperative chemoradiotherapy than the group with “less advanced disease” (63% malignant lymph nodes) that went directly to laparotomy and resection because no mass was observed on CT. The effectiveness of their preoperative chemoradiation must be the reason for this decrease in frequency of malignant lymph nodes in patients with what by all other criteria would appear to be more advanced disease. Because the nonoperative methods used by each study to diagnose local invasion were different, Spitz et al. included many patients with T3 disease into both groups; we included most of these patients in Group 1 receiving RT-FSP.
Survival of the patients treated with preoperative chemoradiotherapy and resection at M.D. Anderson Cancer Center10, 11 was 19–25 months compared with 32 months for our patients receiving RT-FSP and resection. Although overall survival of all patients in their study10 that received preoperative chemoradiotherapy with or without resection is not given, it can be estimated to be approximately 14–16 months based on the data reported (7.2-month median survival for patients not undergoing pancreaticoduodenectomy). This is equivalent to a 15.5-month survival reported by Todd et al.12 for a group of patients with unresectable disease that are similar to our Group 1 patients and who were treated with chemoradiotherapy. They report a 39% response rate with 6 of 15 responders being downstaged, so that resection could be performed in 4.12 These 6 patients had a median survival of 28 months, which is similar to the 32–33 months for patients in our RT-FSP group that also underwent resection or were downstaged and eligible for resection (Table 4). Chemoradiotherapy in Todd et al. consisted of 5-FU, leucovorin, mitomycin, and dipyridamole, whereas Spitz et al. used only 5-FU to limit toxicity. The survival of all RT-FSP treated patients is greater than any other chemoradiation regimens reported for equivalent patients.
Chemoradiotherapy, primary or adjuvant, has been a controversial treatment option. Compared with previously reported chemotherapy and/or radiotherapy regimens, RT-FSP appears more effective than even those utilizing intraoperative treatment.45 The improved results with RT-FSP may be due to the synergism of cisplatin, an established radiosensitizer, with 5-FU given in a 5-fold increased dose, due to the addition of streptozotocin at specific intervals to produce brief tumor regression, and/or due to further radiosensitization with a less toxic, split-course, concurrent radiotherapy regimen.8 We observed complete pathologic response (Fig. 3) in six patients treated with RT-FSP and resection. Of note, there was a focus of atypical, hyperplastic epithelium of a major pancreatic duct possibly representing “residual in situ carcinoma” in a part of the specimen from the patient shown in Figure 3. Although T classification for ampullary tumors can be classified as Tis (carcinoma in-situ) which is an AJCC TNM tumor stage 0, this is not a T classification found for exocrine tumors of the pancreas. The finding of complete pathologic response has not been noted in other studies,11, 14 which indicates that RT-FSP achieves a better local response than other chemoradiotherapy protocols. The RT-FSP regimen possibly could be even better if improvements of rapid fractionation radiation noted by M.D. Anderson Cancer Center11 were used instead of more standard fractionation used in our study.
Other advantages of preoperative compared with postoperative chemoradiotherapy have been discussed in detail.10, 11 The most important of these advantages may be that surgery, the most “toxic” part of the combined modality treatment, can and should be reserved for only those who are clear candidates for resection, and delivered last.
The median 14.0-month survival of our surgical Group 2 patients was representative of the best results reported in most prior studies of similar patients undergoing a Whipple tumor resection. Survival of 16 months for patients in Group 2 who received postoperative chemotherapy is also similar to other reports.9, 18, 19 Factors that contribute to our 16.2-month survival being less than the 18–20 months reported by Gastrointestinal Tumor Study Group (GITSG)18, 19 include 45% of Group 2 patients had malignant lymph nodes (Stage III) compared with only 28% for the GITSG patients; margins free of cancer in all GITSG patients; only 57% of treated Group 2 patients received both chemotherapy and radiotherapy compared with all GITSG patients; 15% of Group 2 patients had tumors of the body or tail compared with only 5% for the GITSG patients, and the GITSG patients were from a small, well controlled study, whereas our outcomes study reflects results in a large group from several institutions. Even with the lowest reported perioperative mortality (< 1%) and morbidity, Yeo et al.9 reported a survival of 20 months for patients offered and, after Day 60, receiving postoperative therapy; this was compared with 12 months for the 28% that did not receive postoperative treatment because of delayed recovery or refusal. For Group 2 patients not receiving postoperative treatment, the survival of 10.5 months, even with a 7% perioperative mortality, was within the range of prior studies. In addition, in our study 42% of patients with Stage 1 disease did not receive postoperative treatment. This was primarily due to patient refusal, although this may have been influenced by the surgery itself as suggested by Pisters et al.11
Similar to other studies9, 10 in which 25–30% of patients do not receive intended postoperative chemoradiotherapy, we found a significant number of patients in Group 2 did not get chemoradiotherapy related to surgical complications. Yeo et al.9 reported a 15-month median survival for patients with complications after pancreaticoduodenectomy. Major complications appear to be significantly higher in our study, even though Yeo et al. report a 30% complication rate after tumor resection for patients that received standard postoperative chemoradiotherapy. Of note, delayed gastric emptying occured in only 4% in our study compared with 20% in their study, which accounted for a majority of their complications. Delayed gastric emptying is related to a pyloris-preserving resection, performed in 84% of their patients, but infrequently in our patients. Excluding this complication, the rest of their complications do not appear to be major ones, whereas in our primary resection Group 2 patients the majority of complications are major. Complication rates among the different surgeons in our study vary widely. However, similar to Yeo et al., we did not find that, at least statistically, complications influenced calculated group survival. For patients receiving postoperative chemoradiation, the difference in survival of 16.2 months for our Group 2 patients compared with 19.5 for patients reported by Yeo et al. could be due to the different types of complications, different chemoradiation regimens, variable inclusion of cancers of the body or tail, and the reported favorable influence on long term survival for a pyloris-preserving resection.
In some studies, actuarial methods have been used for reporting long term survival of up to 55%.22, 25 However, these statistical methods are increasingly inaccurate, as more patients are lost to follow-up. Yeo et al.9 observed a greater actuarial survival for their patients, with a more than 2-fold difference between actuarial and actual 4-year survival. In our study, not one single patient was lost to follow-up. We maximized the survival curve for Group 2 in the manner usually reported in most surgical series2–5, 9, 22, 43, 44 using the selection bias of including only those patients who had a successful partial or total pancreatectomy. Patients undergoing laparotomy for resection without removal of the tumor were excluded. Also excluded were patients with the more favorable prognosic ampullary, cystic, or endocrine tumors. The importance of accurate pathologic confirmation of tumor type and anatomic origin, which can strongly influence survival rates, has been emphasized previously.3
In a recent analysis of the world's pancreatic carcinoma literature over 60 years,22 Gudjonsson calculated that fewer than 300 (0.4%) patients survived more than 5 years. Ten percent of these survivors did not undergo resection, presumably because of unresectable disease. Many 5-year survivors died in the 5th to 7th year.3, 4, 22 Only 14 10-year survivors, all of whom had resection, could be documented.3, 51 Because from 1935 to 1995 only an estimated 30 patients (0.04%) with regional, unresectable tumors were reported to have survived more than 5 years without surgery, 3 long term survivors (4%) in our initial unresectable Group 1 patients is a substantial number.
Patients came from a wide referral base without selection processes other than as discussed in inclusion and exclusion criteria. No patient with unresectable, regional, pathologically confirmed pancreatic adenocarcinoma was excluded from Group 1 once starting the RT-FSP protocol. Primary resection Group 2 consisted of a broad spectrum of patients undergoing pancreatic tumor resection at a major medical center, a community hospital, or a specialty center for pancreatobiliary carcinoma surgery after referral after EUS evaluation. Thus, Groups 1 and 2 are representative of patients with Stage I, II, and III pancreatic carcinoma. Results should be applicable to patient populations at other centers.
Surgical resection for regional pancreatic carcinoma has been considered both the only chance for long term survival and the best palliation.22 But, there has been no controlled study to support or disprove this belief.47–49 Despite more advanced disease in Group 1, only 1% of RT-FSP treated patients died in the first 6 months compared with 23% of Group 2 patients. There appears to be no risk in delaying surgery to first undertake combined modality therapy with RT-FSP. In addition, there is no significant difference in operative mortality and morbidity when comparing Group 1 and Group 2. In fact, similar to the data from other centers,10, 11, 14 there appears to be a trend towards less operative complications when a Whipple procedure is performed after preoperative chemoradiotherapy.
These findings support preoperative treatment with chemoradiotherapy, such as RT-FSP, not only for a pancreatic carcinoma greater than 4 cm or beyond Stage I, based on nonoperative staging, but possibly even for Stage I tumors. Studies have shown that less than 10% of tumors larger than 4.5 cm are resectable.13, 25, 52 The significant advantage in survival for Group 1 patients over Group 2 patients receiving postoperative chemotherapy with or without radiotherapy further supports investigation of preoperative chemoradiotherapy for all stages of regional pancreatic carcinoma. Surgery does not appear to be optimal primary treatment, even for palliation, except where the diagnosis cannot be established nonoperatively or significant intestinal obstruction exists. After chemoradiation, patients can be restaged with CT and EUS to determine further treatment, including surgery.53 EUS guided FNA can be used to obtain a tissue diagnosis for tumors not observed even on high quality CT.
Reporting on 978 institutions in the United States in 1996, Janes et al.4 found that fewer than 20% of patients with pancreatic carcinoma received chemotherapy or radiotherapy, and only 4.0% received combined surgery, radiation, and chemotherapy. Most eligible patients thus do not receive chemoradiotherapy for various reasons, which may include the perception that benefits are limited and outweighed by risks of toxicity. The traditional hypothesis1–5, 22, 25, 41–43 that “for highly selected patients, cancer-directed surgery offers the best and even only chance for cure (< 1%) or for long term survival, with a reasonable (3–5%) operative mortality and complication rate in certain institutions”3, 4 may warrant revision. Compared with initial surgical resection for patients with earlier stage tumors, RT-FSP followed by resection when possible is associated with less early mortality and increases survival for patients with tumors reliably staged as unresectable. Preoperative RT-FSP for earlier resectable stages of pancreatic carcinoma requires further investigation.
- 2Cancer of the pancreas—surgery. In: RustgiA, editor. Gastrointestinal cancer: biology, diagnostics, and therapy. Philadelphia: Lippincott-Raven Publishers, 1995: 341–50., .
- 8Cancer of the pancreas—chemotherapy and radiation therapy. In: RustgiA, editor. Gastrointestinal cancer: biology, diagnostics, and therapy. Philadelphia: Lippincott-Raven Publishers, 1995: 325–40., .
- 17The Gastrointestinal Tumor Study Group. A multi-institutional comparative trial of radiation therapy alone and in combination with 5-fluorouracil for locally unresectable pancreatic carcinoma. Ann Surg 1979; 189: 205–8.
- 21Ten-year follow-up of first generation surgical adjuvant studies of the Gastrointestinal Tumor Study Group. In: JonesS, SalmonS, editors. Adjuvant therapy of cancer. VI. Philadelphia: WB Saunders, 1990: 405–15., .
- 23Can endoscopic ultrasonography influence management and outcome of pancreatic cancer? Acta Endoscopica 1995; 25: 519–29., , , .
- 29Assessment of vascular involvement by pancreatic disease—a comparison of endoscopic ultrasonography to computerized tomography and angiography. Gastrointest Endosc 1990; 36: A197., , .
- 34Pancreatic and periampullary carcinoma. In: SchwartzS, EllisH, Cowles-HusserW, editors. Maingot's abdominal operations. Norwalk, CT: Appleton & Lange, 1989: 1633–69..
- 35BeahrsO, HensonD, HutterR, MyersR, editors. Manual for staging of cancer. 3rd ed. American Joint Committee on Cancer. Philadelphia: J.B. Lippincott, 1988.
- 36Endoscopic ultrasonography images of the normal retroperitoneum In: Lightdale C, editor. Endoscopic ultrasonography. Gastrointest Endosc Clin N Am 1992; 2: 637–55..
- 37Identification of major retroperitoneal vascular anatomy with endoscopic ultrasonography. Acta Endoscopia 1995; 25: 491–8..
- 38Artifacts and techniques of endoscopic ultrasonography in investigating gastrointestinal pathologies and therapeutic options. In: PreedyV, WatsonR, editors. Methods in disease: investigating the gastrointestinal tract. London: Greenwich Med. Media, 1998: 141–55..
- 39Correlation of endoscopic ultrasonography images and surgical findings of pancreatic tumors after split course radiotherapy plus simultaneous radiopotentiator chemotherapy. J Gastroenterol Hepatol 1997; 12(Suppl): A229., , , , .
- 51Survival 13 years after pancreatectomy for ductal adenocarcinoma of the head of the pancreas. Wis Med J 1992; 93: 266–9..
- 53Influence of combined chemotherapy, radiotherapy and surgery on survival of patients with unresectable pancreatic carcinoma using computed tomography and endoscopic ultrasound to select treatment. Gastroenterology 1994; 106: A28., , , , , .