Clinical trials of postoperative therapy for patients with resectable pancreatic cancer to date have been constrained by methodological limitations that make decisive conclusions difficult to reach. The Gastrointestinal Tumour Study Group (GITSG) study included few patients, and the European Organization for Research and Treatment of Cancer (EORTC) study did not stratify patients by resection margin status and lacked sufficient statistical power to detect a survival difference between groups for patients with pancreatic cancer. The European Study Group for Pancreatic Cancer (ESPAC-1) trial introduced considerable selection bias by allowing clinicians to choose the randomization scheme to which patients were entered; however, the authors published the results of the ESPAC 2x2 factorial design separately, which were free of data contamination and represented a clean methodological design. Patients in the ESPAC-1-plus trials were allowed to receive background therapy outside of the randomly assigned regimen, according to patient or physician preference, thus confounding the results of the comparisons. The ESPAC-1 trial reported that a considerable number of patients did not receive treatment according to protocol and variations in radiotherapy quality control were allowed between study centres. Of the patients for whom treatment details were available, 21% who were randomized to receive chemoradiotherapy (CRT) were given more or less than 40 Grays (Gy), and 9% received no CRT, while 33% who were randomized to receive chemotherapy (CT) were given less than six cycles, and 17% received no CT. Similarly, a significant number of patients randomized to the treatment arm of the Norwegian postoperative CT trial were not treated (20%) or did not complete therapy (37%). The Japanese study by Takada et al did not use an intention-to-treat analysis. Those limitations make the interpretation of some study results problematic and underline the importance of sufficiently powered trials with clean methodological designs to better clarify the role of postoperative therapy in this patient group.
The initial positive result of the small GITSG study that led to a conventional recommendation for postoperative CRT has been refuted by the larger ESPAC-1 trial. It now appears more probable that the GITSG study was positive not because of the CRT but rather the subsequent two years of postoperative CT. Postoperative CRT with split-course radiotherapy (RT) can no longer be routinely recommended for patients after resection of pancreatic cancer. However, it is possible that CRT could still be beneficial if given with superior modern treatment planning techniques, with the elimination of split-course RT regimens and when given in combination with newer CT agents such as infusional 5-fluorouracil (5FU) or gemcitabine. Additionally, the role of postoperative CRT in margin-positive patients requires clarification, as only a small minority of patients in those studies were margin positive. The individual patient data (IPD) meta-analysis suggested improved outcomes with CRT in margin-positive patients compared to margin-negative patients; however, there was insufficient statistical power to make comparisons between those subgroups. These are topics of relevance for future trials.
Because of the complicated design of the ESPAC-1 study, and the differences in the results depending on randomization group, the ESPAC-1 investigators felt that a larger, more specific confirmatory trial would be appropriate (ESPAC-3). As that study, at interim analysis, has dropped the observation arm due to inferiority, there is now a clear role for postoperative CT for patients with resected pancreatic cancer. That trial continues to investigate the role of gemcitabine as postoperative therapy compared to 5FU/leucovorin (LV).
At present, there is more evidence available for the overall survival advantages seen with postoperative 5FU/LV than for gemcitabine in the postoperative setting. Most CT regimens used in the reported trials were 5FU-based for a period of at least four months. Given the extensive experience with the Mayo regimen in the colorectal cancer postoperative setting, and the use of this regimen in the largest trial (ESPAC-1), that would seem a reasonable choice for postoperative therapy. Although in the metastatic setting gemcitabine has been compared to 5FU/LV and found to be associated with better quality of life, studies comparing those two regimens in the postoperative setting are ongoing. The Radiation Therapy Oncology Group (RTOG) 9704 study evaluated the addition of gemcitabine to postoperative adjuvant 5FU CRT. All patients received 5FU CRT and either 5FU or gemcitabine before and after CRT. In this study, 42% of patients randomized to the 5FU CRT plus 5FU crossed over to receive gemcitabine. The addition of gemcitabine to 5FU CRT improved survival in patients with pancreatic head cancer but not in the analysis of all eligible patients. Emerging data from the ESPAC-3 trial will determine if six months of postoperative gemcitabine is equivalent or superior to 5FU/LV. The higher drug acquisition cost of gemcitabine and longer administration time should be considered prior to the widespread adoption of gemcitabine as standard postoperative therapy over the more studied 5FU/LV regimen. There are currently insufficient data to support the routine use of preoperative therapy for patients with potentially resectable pancreatic cancer.
The Norwegian study by Bakkevold et al demonstrated superior outcomes with combination chemotherapy using 5FU, doxorubicin, and mitomycin-C (MMC) compared to observation alone. Although that study provides further evidence for the role of chemotherapy as postoperative treatment, it is not possible to determine the independent effect of the doxorubicin or the mitomycin from the trial, and there is an absence of supporting data for those agents. In addition, significant toxicity was observed in patients who received the combined chemotherapy regimen. Therefore, the routine use of doxorubicin or MMC in the postoperative setting cannot be recommended.