Adjuvant therapy for pancreatic cancer

A review

Abstract

Pancreatic cancer is uniformly fatal unless it can be surgically resected. Survival rates for the 15% to 20% of patients who have resectable disease, however, are a disappointing 10% to 30%, depending on the status of margins and surrounding lymph nodes. In the mid-1980s, a landmark study by the Gastrointestinal Tumor Study Group was the first to demonstrate a survival benefit from adjuvant therapy in the form of chemoradiation. Since then, several studies in both North America and Europe have tested the role of adjuvant chemotherapy or chemoradiation in pancreatic cancer, and the results have stirred great controversy. For this review, the evidence for adjuvant therapy in pancreatic cancer was examined, and the significant practice differences that exist between North American and European oncologists were highlighted. The authors investigated the results from the European Study Group for Pancreatic Cancer-1 trial and the reasons why that study has served to reinforce rather than resolve these trans-Atlantic differences. They also reviewed preliminary data from more recent adjuvant trials and explored the possible benefits of a neoadjuvant approach. Cancer 2008. © 2007 American Cancer Society.

More than 33,000 individuals develop pancreatic adenocarcinoma each year in the United States, and >95% ultimately will die from the disease.1 The only meaningful chance for a cure is surgical resection, but only 15% to 20% of patients have potentially resectable disease at presentation. Furthermore, the prognosis is poor even for those who undergo complete (R0) resection. Five-year overall survival rates after pancreaticoduodenectomy for lymph node-negative and lymph node-positive disease are 25% to 30% and 10%, respectively.2–6

Changes in Management: 1980 to 2007

There is some evidence that outcomes may be improving over time. A recent population-based analysis of 396 Medicare patients residing in 1 of 11 Surveillance, Epidemiology, and End Results (SEER) Registry areas who underwent surgery with curative intent for pancreatic cancer between 1991 and 1996 reported a 3-year survival rate of 34%.7 This trend may be related to several major changes that have occurred over the last 20 years.

First, computed tomography (CT) imaging has improved markedly and currently allows for the recognition of early metastatic disease. In addition, the increasing use of positron emission tomography (PET) scans also may increase the earlier detection of metastatic disease. In a study of 42 patients with pancreatic cancer, PET changed the clinical staging over CT scanning in 5 patients (11.9%).8 In another study, 16% of 59 patients who were deemed resectable by CT scanning had distant metastatic disease identified by PET scanning. Many of those patients would have undergone resection of the primary tumor in previous years that clearly would have had a negative impact on survival statistics.

Second, the pathologic staging of pancreatic cancer has changed markedly. Before the 1990s, the traditional pathologic margins of a pancreaticoduodenectomy included the stomach and duodenum, the bile duct, and the pancreas. The peripancreatic region, including the retroperitoneal margin, was not identified routinely in pathology reports of patients undergoing pancreaticoduodenectomy. This area often represents the technical resection limits of a pancreaticoduodenectomy because of extension of the tumor to the major vessels. In the late 1980s, it became increasingly clear that the retroperitoneal margin was the most common positive margin, and patients with a positive retroperitoneal margin had a cure rate ≤10%. Focus on this margin has intensified, and many surgeons consider a resection that leaves a positive retroperitoneal margin as palliative and will not offer a pancreaticoduodenectomy to patients who are at high risk for a positive retroperitoneal margin.

Finally, surgical technique has improved dramatically over the last 25 years. Before 1990, the mortality rate for pancreaticoduodenectomy ranged from 10% to 20%. It is very clear that the experience of the surgeon and surgical team dramatically affect the mortality of a pancreaticoduodenctomy.9 Birkmeyer and colleagues evaluated the Medicare database and divided surgeons into 3 groups based on the relative number of resections per year. Surgeons who performed <2 resections each year had a mortality rate of 14.7% compared with 4.6% for surgeons who performed ≥4 resections per year.10 Recently, Hollenbeck and colleagues evaluated 9153 patients who underwent pancreatic resection between 1993 and 2003 and observed that hospitals in the top decile for volume had a surgical mortality rate of 1.7% compared with 11.7% mortality for hospitals in the lowest decile for volume.11

These factors may have influenced the results of the adjuvant studies discussed in this review and make it very difficult to compare studies across decades. Systemic chemotherapy, radiation therapy (RT), or combination chemoradiotherapy have been used after surgical resection (adjuvant therapy) and even before resection (neoadjuvant therapy) in an effort to improve cure rates. Although the benefit of adjuvant therapy has become more apparent in recent years, the optimal choice of treatment modality (chemotherapy with or without RT) remains intensely controversial.

Chemoradiotherapy

The rationale for combined-modality therapy is driven by the failure pattern after surgical resection; surgical series from the 1990s reported that >50% of patients developed locoregional recurrence without evidence of distant metastases.12–15 The Gastrointestinal Tumor Study Group (GITSG) initially demonstrated improved survival in the late 1960s and early 1970s when external beam RT (EBRT) was combined with concurrent 5-fluorouracil (5-FU) chemotherapy compared with RT alone in patients with locally advanced, unresectable pancreatic cancer.16, 17 This combined-modality regimen subsequently was moved into the adjuvant setting in an attempt to improve the surgical cure rate. Despite the publication of several randomized trials evaluating the role of adjuvant chemoradiotherapy in patients with resected pancreatic cancer, its true benefit remains controversial. A review of the 3 most important adjuvant chemoradiation pancreatic trials below highlights the reasons for this controversy and reveals the historic origins of the difference that has arisen between North American and European practices (see Table 1).

Table 1. Landmark Trials of Adjuvant Therapy for Resectable Pancreatic Cancer

Study: treatment arm (no. of patients)	Median OS, mo	P	2-Year survival, %	P
OS indicates overall survival; GITSG, Gastrointestinal Tumor Study Group; CRT, chemoradiation therapy; 5-FU, 5-fluorouracil; EORTC, European Organization for Research and Treatment of Cancer; ESPAC, European Study Group for Pancreatic Cancer. * CRT was delivered at a dose of 40 grays (Gy) as split course (20 Gy over 2 weeks with 2 weeks off between courses), plus 5-FU bolus 500 mg/m^2 on Days 1–3 during each 20-Gy course of RT, followed by weekly bolus 5-FU 500 mg/m^2 for 2 years until toxicity or progression. † The data presented exclude the 104 patients who had ampullary tumors. ‡ CRT was delivered at a dose of 40 Gy as a split course (given as in GITSG) plus 5-FU 25 mg/kg as a continuous infusion during radiotherapy; no post-CRT 5-FU was given. § Not powered to compare the 4 groups directly. ‖ CRT was delivered the same as in the GITSG study; chemotherapy was delivered as 5-FU 425 mg/m^2 plus leucovorin 20 mg/m^2 on Days 1–5 every 28 days × 6.
GITSG
A: Surgery alone (n = 22)	11	.035	15	–
B: Surgery + CRT + 5-FU (n = 21)*	20		42
EORTC
A: Surgery alone (n = 54)†	12.6	–	23	.099
B: Surgery + CRT (n = 60)‡	17.1		37
ESPAC-1 (2 × 2 analysis)
A: Surgery alone (n = 69)	16.9	–§
B: Surgery + CRT (n = 73)‖	13.9
C: Surgery + 5-FU (n = 75)	21.6
D: Surgery + CRT + 5-FU (n = 72)	19.9
CRT: B + D (n = 145)	15.9	.05	29	–
No CRT: A + C (n = 144)	17.9		41
Chemotherapy: C + D (n = 147)	20.1	.009	40	–
No chemotherapy: A + B (n = 142)	15.5		30

GITSG

The GITSG study, which was conducted in the late 1970s and early 1980s, was the first randomized trial to demonstrate that adjuvant therapy confers a survival benefit in patients with resected pancreatic cancer and set the stage for all subsequent adjuvant trials. Patients were assigned randomly to either observation or 40 grays (Gy) of EBRT split into 2 courses of 20 Gy with a 2-week rest interval between courses. Concurrent bolus 5-FU 500 mg/m² per day on the first 3 days of each course of RT was given. After the completion of chemoradiotherapy, patients received maintenance bolus 5-FU 500 mg/m² per day on a weekly basis for 2 years or until they developed disease progression.18 After taking 8 years to accrue only 43 patients, the study was closed. The GITSG study was flawed in several respects: 1) the lengthy time for accrual, 2) the small number of patients, 3) 25% of patients on the treatment arm did not begin postoperative treatment until >10 weeks after resection, 4) the use of bolus 5-FU without leucovorin, and 5) the use of split-course RT at a relatively low dose of 40 Gy. Nevertheless, patients who received postoperative chemoradiotherapy had significantly longer median overall survival (20 months vs 11 months) and better 2-year survival rate (42% vs 15%). After the study was closed, an additional 32 patients were registered on the combined-modality arm, and a subsequent report that included these and the original 43 patients demonstrated that the survival benefit in patients who received chemoradiotherapy was consistent.19

European Organization for Research and Treatment of Cancer

Recognizing some of the limitations of the GITSG study, European investigators in a study sponsored by the European Organization for Research and Treatment of Cancer (EORTC) randomly assigned 114 patients with resected pancreatic adenocarcinoma either to receive postoperative, concurrent 5-FU 25 mg/kg per day as a continuous infusion plus split-course EBRT to 40 Gy or to undergo observation.20 Although there was a trend toward improved outcomes in the group that received adjuvant therapy, the improvements were not statistically significant (2-year survival rate, 23% vs 37% for the control and treated patients, respectively; P = .099). In addition to the suboptimal dose and delivery of RT as well as the use of bolus 5-FU, this EORTC study was criticized, because 20% of patients who were randomized to receive treatment never received it. Proponents of chemoradiation view this study as a fatally flawed and underpowered study.

European Study Group for Pancreatic Cancer 1 trial

In an attempt to answer the question regarding adjuvant therapy once and for all, in 1994, the European Study Group for Pancreatic Cancer (ESPAC) began a 2 × 2 factorial design in which the relative benefits of adjuvant chemotherapy, chemoradiotherapy, or chemoradiotherapy followed by chemotherapy would be compared with observation alone. In an effort to be more pragmatic and to improve accrual, the ESPAC investigators then allowed clinicians to choose from 2 other randomization schemes: 1) chemoradiotherapy or no chemoradiotherapy and 2) chemotherapy or no chemotherapy. The final results were presented in 2 separate publications, a report in which the results of all 541 patients were pooled from the 3 parallel randomized trials21 and a later report that focused on the 289 patients who were randomized to the 2 × 2 factorial design.22 In the initial report of the pooled analysis, there was no difference in survival when the 175 patients who received postoperative chemoradiotherapy were compared with the 178 patients who did not receive it (median overall survival, 15.5 months vs 16.1 months, respectively). In contrast, there was a significant survival benefit for adjuvant chemotherapy alone when 238 patients who received it were compared with 235 who did not receive it (median survival, 19.7 months vs 14 months, respectively).

Subsequently, a second report was published that detailed the results from the 289 patients who were randomized to the originally intended 2 × 2 factorial design.22 There was no significant benefit for chemoradiotherapy, and the data demonstrated a trend toward worse survival for the group that received chemoradiotherapy (the 2- and 5-year survival rates were 29% vs 41% and 10% vs 20% for the chemoradiotherapy and no chemoradiotherapy groups, respectively). It is noteworthy that local recurrence rates were similar in both groups, and there were more recurrences overall (84 recurrences vs 74 recurrences) in the chemoradiotherapy group as well as a shorter recurrence-free interval (10.7 months vs 15.2 months, respectively).

For patients who received postoperative chemotherapy compared with patients who received no chemotherapy, the 2-year survival rates (40% vs 30%, respectively) and the 5-year survival rates (21% vs 8%, respectively) were significantly greater. It also is worth noting that 33% of the patients who were assigned to received adjuvant chemotherapy did not complete all 6 courses, and 17% received no chemotherapy at all. The median survival for patients who received chemotherapy versus no chemotherapy was 20.1 months versus 15.5 months, respectively.

A recent meta-analysis included 875 patients from GITSG, EORTC, and ESPAC-1 (all 3 parallel trials were included) in addition to patients from Norwegian and Japanese adjuvant chemotherapy trials.23 In that analysis, no significant difference was observed in the risk of death for patients who received chemoradiation (hazard ratio, 1.09; P = .43), and a significant decrease in the risk of death was observed for patients who received chemotherapy (hazard ratio, 0.75; P = .001). Because the ESPAC-1 data dominated (550 of 875 patients analyzed), it is difficult to interpret the findings of that meta-analysis. The ESPAC-1 trial was criticized for many reasons: 1) violation of the randomization design by allowing patients and clinicians to select which trial to enter and to allow “background” chemoradiation or chemotherapy if a patient was assigned to an undesirable arm, 2) comparing patients on treatment actually received rather than conducting an intent-to-treat analysis, 3) the use of a split-course radiation schedule and unequal dosing of radiation among participants, 4) the lack of RT quality assurance, and 5) the inclusion of patients with positive margins.

In summary, the ESPAC-1 trial is criticized by its detractors for multiple methodological flaws and is celebrated by its proponents for its pragmatic design. It does provide strong support for the role of adjuvant chemotherapy in some form but makes it nearly impossible to draw firm conclusions regarding the benefit of chemoradiotherapy. It also demonstrates the difficulties of performing a multi-institutional study in an illness best cared for by specialists in specialty centers where surgical, pathologic, radiation, and medical quality control is performed. A division between North American and European practice was born when the results of the EORTC, a European trial, did not confirm the benefits of chemoradiation observed in the original GITSG trial, an American study. The results of ESPAC-1, rather than settling the issue of chemoradiation versus chemotherapy alone, have served to deepen the divide across the Atlantic. Most Europeans have accepted the results of ESPAC-1 as demonstrating a clear benefit to chemotherapy alone and no benefit, perhaps even detriment, to chemoradiation. This is reflected in the finding that the ESPAC-3 trial, which is discussed below, was designed with no chemoradiation comparison arm and compares 5-FU with gemcitabine in the adjuvant setting. There has been reluctance on the part of North Americans oncologists to abandon chemoradiation given the flaws of the GITSG, EORTC, and ESPAC-1 studies in the light of such a high rate of local recurrence.

Several other investigators have reported uncontrolled data that suggest a benefit for postoperative chemoradiotherapy in pancreatic cancer.24–32 An evaluation of Medicare patients derived from the SEER database also demonstrated a survival advantage for patients who received adjuvant therapy. The 3-year survival rate in that trial was 45% for patients who received adjuvant chemoradiotherapy versus 30% for patients who did not.7

Post-ESPAC-1 Era

The American and European investigators have embarked on 2 different paths. The American investigators, led by the Radiation Therapy Oncology Group (RTOG), evaluated the role of 5-FU or gemcitabine before and after chemoradiation through the RTOG 97-04 study. The European investigators have abandoned chemoradiation and are evaluating chemotherapy alone after pancreatic cancer resection (see Table 2).

Table 2. Recent Trials of Adjuvant Therapy for Resectable Pancreatic Cancer*

Study: treatment arm	No. of patients	Median DFS, mo	P	Median OS, mo	P
DFS indicates disease-free survival; OS, overall survival; RTOG, Radiation Therapy Oncology Group; 5-FU, 5-fluorouracil; CRT, chemoradiation therapy; GEM, gemcitabine; CONKO, Charité Onkologie (Germany). * A statistically significant benefit was observed for patients with pancreatic head tumors only (the prespecified primary endpoint).
RTOG 97-04*	538
A: Surgery+5-FU+CRT+5-FU		10.5	.10	16.9	.033
B: Surgery+GEM+CRT+GEM		11.5		20.6	11.5
CONKO 001	368
A: Surgery alone		6.9	<.001	20.2	.06
B: Surgery+GEM		13.4		22.1

RTOG 97-04

A United States Intergroup study (RTOG 97-04) was designed to compare 5-FU and gemcitabine before and after 5-FU-based chemoradiotherapy using a “sandwich” design.33 Patients who underwent gross total resection for pancreatic adenocarcinoma who had tumors classified as T1 through T4 and a lymph node status of N0 or N1 were assigned randomly to 1 of 2 treatment arms. One arm consisted of 3 weeks of continuous infusion 5-FU (250 mg/m² per day) followed by chemoradiotherapy (50.4 Gy in 1.8-Gy daily fractions for 5.5 weeks with concurrent infusional 5-FU 250 mg/m² per day) and, starting 3 to 5 weeks later, 2- to 4-week courses of continuous infusion 5-FU (250 mg/m² per day) with a 2-week rest between courses. This control arm is an updated form of the chemoradiation originally used in the GITSG in that: 1) the RT dose is increased from 40 to 50.4 Gy, 2) the RT schedule is continuous, not split course, and 3) the 5-FU is given as continuous infusion rather than as a bolus. The experimental gemcitabine arm was comprised of 3 weekly doses of gemcitabine alone (1000 mg/m² per week), followed by the same chemoradiotherapy protocol as for the conventional chemoradiotherapy arm, and, starting 3 to 5 weeks later, 3 months of single agent gemcitabine (1000 mg/m² weekly for 3 of every 4 weeks). This study is the first Phase III adjuvant trial in pancreatic cancer to require prospective quality control of all RT fields. As there is no chemotherapy alone arm, we cannot distinguish what benefit is attributable to the high-quality, modernized radiation technique.

The RTOG 97-04 study was amended early in its course to allow for the separate analysis of tumors of the pancreatic head because of their better prognosis compared with tumors of the pancreatic body/tail. The initial results were reported at the 2006 Annual Meeting of the American Society of Clinical Oncology. Among patients with pancreatic head tumors (n = 380 patients), those who received gemcitabine had a significantly better median survival (20.6 months vs 16.9 months) and 3-year survival (32% vs 21%) compared with those who received 5-FU. When combining patients who had pancreatic head tumors and patients who had pancreatic body/tail tumors, there was no significant difference between the 2 groups. Whereas nonhematologic toxicity was similar between the groups, patients who received gemcitabine had more grade 4 hematologic toxicity (14% vs 2%). It is likely that the gemcitabine arm of the RTOG 97-04 will become the comparator arm in future adjuvant trials conducted in North America. Preliminary data support the tolerability of regimens that use gemcitabine as a radiation sensitizer,34, 35 but no trials have compared this approach to chemoradiotherapy using 5-FU as the radiation sensitizer, at least in the postoperative setting.

CONKO-001

For the German Charité Onkologie (CONKO) multicenter trial CONKO-001, 368 patients with resected pancreatic carcinoma were assigned randomly to receive either gemcitabine (1000 mg/m² on Days 1, 8, and 15 every 4 weeks for 6 months) or no treatment after surgery.36 Periampullary tumors of nonpancreatic histology were excluded. Patients were stratified by resection margin status, tumor classification, and lymph node status, and the primary endpoint was disease-free survival. After a median of 53 months of follow-up, adjuvant gemcitabine was associated with a significantly longer median disease-free survival (13.4 months vs 6.9 months), which was evident in patients who had negative lymph nodes (24.8 months vs 10.4 months) and positive lymph nodes (12.1 months vs 6.4 months) and in patients with negative margins (13.1 months vs 7.3 months) and positive margins (15.8 months vs 5.5 months). Although survival was not a primary endpoint, there was a clear trend in favor of gemcitabine, because the median survival was 22.1 months in the gemcitabine arm versus 20.2 months in the observation arm (P = .06). The disease-free survival for both arms was less than what was expected from earlier studies in the adjuvant setting, thus raising issues of proper staging techniques. Nevertheless, this study buttresses the conclusion of the RTOG 97-04 in support of gemcitabine in the adjuvant treatment of pancreatic cancer.

ESPAC-3

ESPAC-3 is multicenter European trial that was designed to compare 5-FU/leucovorin and gemcitabine in patients with resected pancreatic carcinoma. Like the German CONKO study, it does not include a chemoradiation arm. The trial is nearing accrual and should report preliminary results by 2008.

Neoadjuvant Therapy

The benefits conferred by adjuvant therapy are modest whether it entails chemotherapy alone or chemoradiotherapy. Furthermore, a substantial proportion of patients who undergo pancreaticoduodenectomy either have a long delay or never receive adjuvant therapy because of insufficient recovery from surgery.37 This has led investigators to evaluate preoperative chemoradiotherapy or a combination of preoperative and postoperative therapies. There are several theoretical advantages of this approach. Neoadjuvant chemoradiotherapy may downstage the primary tumor and improve the R0 resection rate.38 Earlier treatment of micrometastatic disease may be beneficial. Neoadjuvant therapy also biologically stages patients, in that approximately 20% of patients will develop metastatic disease and, thus, will be spared the morbidity of a Whipple operation.

Initial reports of preoperative RT with or without concurrent 5-FU demonstrated no significant increase in perioperative morbidity or mortality, yet they also failed to suggest an obvious improvement in either resectability or overall survival.39–44 Subsequent studies have attempted to overcome the intrinsic chemoresistance and radioresistance of pancreatic cancer by increasing the RT dose, adding intraoperative RT, and/or intensifying the chemotherapy regimen.45–51 These newer approaches also appear to be feasible, but their efficacy remains unproven.

Neoadjuvant chemoradiotherapy can be delivered safely to patients with localized pancreatic cancer without adversely influencing perioperative morbidity or mortality. Whether neoadjuvant therapy is better than adjuvant therapy is uncertain. There are no randomized trials comparing the 2 approaches, and no study has clearly demonstrated improved resectability or survival compared with patients who underwent surgery alone. For patients with resectable pancreatic cancer, the standard of care continues to be surgery followed by adjuvant therapy.

Summary

The majority of patients with resectable pancreatic cancer are destined to develop recurrent disease. Those patients who have pathologically negative regional lymph nodes and resection margins have the best chance for a cure. There is an emerging consensus based on the CONKO study and the RTOG 97-04 trial that adjuvant chemotherapy with gemcitabine improves survival for patients after resection. There is no consensus regarding the use of chemoradiation to prevent local recurrence.

The European approach, emphasizing the results of the EORTC and ESPAC-1 studies, does not include chemoradiation. The North American approach emphasizes the high risk of local failure and the potential benefit from chemoradiotherapy. In addition, the serious design flaws of the EORTC and ESPAC-1 studies cast doubts on their conclusions regarding chemoradiation. Nevertheless, the morbidity of chemoradiation (short and long term) is not trivial, and this question should be ignored no longer. A randomized study with proper attention to surgical, RT, and pathologic quality control is needed to clarify this issue.