Abstract
Background: We report the results of a phase II trial of adding the anti-ascular endothelial growth factor (VEGF) bevacizumab to gemcitabine neoadjuvant chemotherapy for patients with borderline and unresectable non-metastatic pancreatic cancer. Patients and Methods: Patients were assigned to one of the two treatment arms. Both groups received 1,000 mg/m2 gemcitabine on days 1, 8, and 15 of a 4-week cycle for a total of four cycles. Group 1 received 5 mg/kg bevacizumab for six weeks (three doses), every second week, starting at week 6 of gemcitabine therapy. Group 2 received 5 mg/kg bevacizumab for 12 weeks (six doses), every second week, starting at week 1 of gemcitabine therapy. The objective of the present study was to assess the rate of complete radical resection and overall survival. Results: A total of 30 patients were enrolled: 19 patients had unresectable and 11 patients had borderline-resectable pancreatic cancer. Eleven patients (37%) underwent resection. The median overall survival of patients who underwent tumor resection was 13 months (95% confidence interval=11-15 months). Conclusion: In general, adding bevacizumab to neoadjuvant gemcitabine does not improve outcomes for patients with locally advanced pancreatic cancer. However, in individual cases, surgery is consequently possible and prolonged survival may be observed.
Pancreatic cancer is a lethal disease that causes 32,500 cancer deaths per year in Europe, with an incidence rate approximately equal to its long-term mortality rate (1). Despite recent advances in drug treatment, such as erlotinib and protein-bound paclitaxel (2-4), palliative chemotherapy has not yielded satisfying results, and complete surgical resection remains the only treatment offering at least the potential of prolonged survival, occasionally enabling long-term survival. While the majority of patients present with disease already metastatic at the time of diagnosis, there is still a sizeable number of patients with non-metastatic but borderline-resectable or locally advanced disease who would be candidates for surgical resection after down-sizing of their primary tumor. This has led to multiple research efforts and clinical trials actively exploring neoadjuvant multimodality strategies. The optimal neoadjuvant chemotherapeutic regimen has not yet been found, and the value of neoadjuvant radiotherapy remains controversial (5-7).
Gemcitabine, a desoxycytidine analog that competes with cytidine for incorporation into DNA and therefore inhibiting DNA synthesis, has been a standard-of-care for the treatment of advanced pancreatic cancer for more than a decade. The superiority of gemcitabine combination therapy over gemcitabine mono-therapy has been well documented (8, 9). In addition to new cytotoxic combinations, biologically-targeting approaches are currently explored in pancreatic cancer (10, 11). At the time this study was initiated, the combination of established chemotherapy (i.e. gemcitabine) with anti-angiogenic agents was a new and promising treatment modality in oncology, including palliative treatment of pancreatic cancer.
Vascular endothelial growth factor (VEGF) plays a key role in the growth and metastasis of many tumor types including pancreatic cancer (12). Autocrine and paracrine growth stimulation by VEGF of pancreatic cancer cells and microvascular endothelial cells in a dose-dependent manner have been observed (13, 14). Bevacizumab (Avastin®, Genentech, South San Francisco, CA, USA), a recombinant humanized monoclonal antibody to VEGF, showed significant improvement in response and eventually also in survival outcomes in patients with advanced colorectal cancer (15). Moreover, a phase II study also indicated significant activity of bevacizumab in combination with gemcitabine in advanced pancreatic cancer, with a response rate of 21%, a median overall survival (OS) time of 8.8 months, and a 1-year survival rate of 29% (16). Treatment regimens using anti-bevacizumab were investigated in many fields of clinical oncology, and eventually reached a standard-of-care in some (15, 17-19). At the time this study was initiated, it was completely unclear for how long anti-VEGF therapy had to be administered in order to achieve minimal/optimal response, particularly in the time-sensitive preoperative setting. This was the rationale for including two different treatment durations in the present study.
The objective of this phase II study was to assess the rate of complete radical resection (R0) and overall survival after addition of two different durations of bevacizumab therapy to neoadjuvant gemcitabine therapy in patients with non-metastatic borderline-resectable and locally advanced ductal pancreatic adenocarcinoma.
Patients and Methods
From August 2006 to August 2009, we recruited patients into a single-Institution phase II study at our the Comprehensive Cancer Center investigating the effects of bevacizumab (in two different durations) in combination with gemcitabine in the neoadjuvant treatment of borderline- and locally advanced non-metastatic pancreatic cancer.
Eligible patients had biopsy-proven borderline- or locally advanced non-metastatic pancreatic cancer and provided written informed consent before enrollment into the trial. The study was performed in accordance with the law and regulations approved by the local Ethics Committee (385/2005). All patients had a computed tomographic (CT) scan or magnetic resonance imaging (MRI) demonstrating borderline or locally advanced pancreatic cancer at the study inclusion. For the definition of resectability, internationally accepted criteria (20-22) were used (see details below). Presence of metastatic disease was excluded by CT of the chest and abdomen. Other inclusion criteria were: an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1, absence of other coexisting malignancy or any invasive cancer diagnosed within the previous five years, age >18 years, life expectancy of >12 weeks, adequate laboratory parameters (absolute neutrophil count >1,500 mm3; bilirubin <1.5 mg/dl; aspartate aminotransferase and alkaline phosphatase levels <3-fold upper limits of normal; and creatinine levels <2 mg/dl). In the case of biliary obstruction, patients underwent biliary drainage/stenting procedures prior to trial therapy. Patients who had undergone major surgery within the previous 28 days or who had a history of bleeding or coagulation disorders were ineligible. Patients with any prior therapy for pancreatic cancer were ineligible. The study protocol was approved by the Institutional Review Board and was registered in the public trial registry EudraCT (no. 2005-004519-32).
Treatment plan and dose adjustments (NeoAvastin). Patients were randomly assigned to one of two treatment arms. Both groups received 1000 mg/m2 gemcitabine at a fixed rate of 10 mg/m2/min on days 1, 8, and 15 of four consecutive 28-day cycles. Treatment group 1 (“short” bevacizumab) received 5 mg/kg bevacizumab intravenously every second week for six weeks (tree doses), starting at week 6 of gemcitabine therapy. Patients in treatment group 2 (“long” bevacizumab) received 5 mg/kg bevacizumab intravenously every second week for 12 weeks (six doses), starting at week 1 of gemcitabine therapy. The fourth chemotherapy cycle did not include bevacizumab, providing a time gap deemed necessary for surgical safety between the last dose of the VEGF antibody and pancreatic surgery.
Adverse effects were graded according to the National Cancer Institute Common Toxicity Criteria for Adverse Events (CTCAE) version 4.0 (23). A 25% gemcitabine dose reduction was performed when the absolute neutrophil count fell to between 1 and 1.5×109/l or the platelet count fell to between 50 and 75×109/l on the planned day of treatment. In the event of any grade 3 to 4 toxicity, gemcitabine and bevacizumab were paused until values returned to baseline. Filgrastim was administered in cases of low neutrophil count. Before chemotherapy infusion, all patients received dexamethasone-21-dihydrogen phosphate and tropisteron as pre-medication.
Evaluation of resectability. Pancreatic cancer stage was evaluated by multidetector thin-slice CT or MRI, which were performed at the time of enrolment and after the last cycle of NeoAvastin therapy. Locally advanced pancreatic cancer was defined as: more than 180 degrees encasement of the superior mesenteric artery, infiltration of the celiac trunk, occlusion of the or portal vein without the possibility of vascular reconstruction or aortic invasion. Borderline resectable pancreatic cancer was defined as abutment (tumor involvement <180° of the circumference vessel) of the superior mesenteric or portal vein and of the superior mesenteric or hepatic artery. Short-segment encasement or occlusion of the superior mesenteric or portal vein amenable to vascular resection and reconstruction was also considered as borderline resectable disease (20-22, 24).
Response criteria and indications for surgical exploration. Radiological tumor assessment was determined according to the Response Evaluation Criteria in Solid Tumors (RECIST 1.0) (25). A team of senior staff radiologists, experienced in pancreatic imaging, evaluated radiological staging and assessment of response. The assessors were unaware of the clinical response or assigned treatment group in the trial. Patients who showed partial response or at least stable disease throughout neoadjuvant therapy were considered for surgery. In individual cases with unclear imaging findings, the indication for explorative laparotomy was seen. The decision-making process was based on standardized multi-disciplinary tumor board assessment, the treating surgeons judgment, and patients' preference.
Pathological grading and staging was performed by the Institutional team of pathologists according to the American Joint Committee on Cancer (AJCC) staging manual (sixth edition, 2002) (26). A resection was classified as R0 if all pancreatic margins and the common bile duct transection margin were free of tumor. The histological tumor response was assessed by a score proposed by Le Scodan et al. (27).
Serum carbohydrate antigen (CA) 19-9 levels were measured at study enrolment and after the last cycle of chemotherapy. CA19-9 was considered to be in the normal range when lower than 37 U/ml. CA19-9 samples drawn from patients with elevated total bilirubin (>2.0 mg/dl) were excluded from the analysis.
In addition, as part of a linked translational research project, a cohort of 20 study-group patients, with borderline resectable and locally advanced pancreatic cancer, was monitored for blood levels of proangiogenic cell populations and angiogenesis factors at 2-week intervals to evaluate their predictive marker potential for treatment response and overall survival (28, 29). Briefly, peripheral blood was analyzed for the detection of circulating endothelial cells (CECs: CD45−, CD31+, CD146+) and proangiogenic endothelial progenitor cells (EPCs: CD3−, CD19−, CD33−, CD34+, CD133+). Plasma samples were analyzed by a commercially available enzyme-linked immunosorbent assay (ELISA) for VEGFA, thrombospondin-1 (THBS1) and a comparable sandwich ELISA system for platelet derived-endothelial cell growth factor (TYMP).
Follow-up. Patient had follow-up investigations by clinical evaluation, CT scans and assessment of CA19-9 levels at least every six months after resection. Patients not undergoing surgical treatment were offered standard-of-care palliative and best supportive care treatment, and were followed-up for survival status via primary physician telephone contact or regular re-evaluations at our Institution.
Statistical evaluation. Overall (OS) and disease-free survival times were estimated according to the Kaplan–Meier method. OS to the time of local progression or to metastatic disease were calculated starting from the day of diagnosis of pancreatic cancer. The OS time of a patient still alive was censored at the time of the last available follow-up. The time to local progression or metastatic disease was determined from pancreatic cancer diagnosis until a patient experienced recurrence documented at the relevant site, or was otherwise censored at an earlier date of death or last follow-up. Variables were compared with χ2 or Student's t-test using SPSS 17.0.1 software for Mac OsX (SPSS Inc., Chicago, IL, USA).
Results
A total of 32 patients with documented borderline or locally advanced but non-metastasized pancreatic cancer were screened for study inclusion. One patient was ineligible because of previous neoadjuvant radiochemotherapy, a second patient could not be included in the trial because of unmanageable recurrent cholangitis. Thus, 30 patients were enrolled into the study: 11 patients were assigned to treatment group 1 (short bevacizumab) and 19 into treatment group 2 (long bevacizumab).
Patient's characteristics are listed in Table I. 57% of patients were women and the median age at study inclusion was 62 years (range=43-80 years). A total of 11 (37%) patients had borderline resectable pancreatic cancer at the time of diagnosis and enrolment into the clinical trial; locally advanced unresectable pancreatic cancer was present in 19 (63%). The median time from diagnosis to start of treatment was 34 days (range=21-105 days). The majority of patients (N=27, 90%) completed all four 28-day cycles of NeoAvastin. Three patients received fewer than four cycles of chemotherapy due to disease progression (n=1) or unmanageable adverse events (n=2). Chemotherapy was interrupted for a week because of grade III toxicity in six patients (in total 12 delayed cycles; two patients of group I and four of group II). No other protocol deviation due to patient non-compliance or other reason was recorded.
Patient demographics grouped as responders to NeoAvastin (with subsequent tumor resection) compared to non-responders (without resection).
Toxicity and adverse events. The most common toxicities encountered were: leukopenia 20% (group I: 18%; group II: 21%), thrombocytopenia 13% (group I: 9%; group II: 15%) and neutropenia 10% (group I: 18%; group II: 5%). The most common non-hematological toxicities included fatigue, nausea, and anorexia. One patient (group II) developed axillary vein thrombosis during the third cycle of NeoAvastin therapy. A lethal gastrointestinal bleeding occurred in an 80-year-old woman (group II) with tumor infiltration into the duodenum after two cycles of NeoAvastin (four doses of bevacizumab) and another patient suffered lethal bleeding from tumor erosion of the splenic artery after one cycle of gemcitabine therapy only.
Radiologic tumor response. Twenty-eight patients out of 30 (93%) underwent imaging studies before and after NeoAvastin treatment; two patients who died during neoadjuvant therapy had only primary staging. According to the RECIST criteria, stable disease was recorded in 5 patients with borderline-resectable pancreatic cancer (55.6%) and in 14 patients with locally advanced pancreatic cancer (73.7%; in total n=19/28, 67.9%). One patient with borderline resectable disease showed partial response. Progressive disease was recorded in three patients with borderline resectability (33.3%) and in five with unresectable pancreatic cancer (26.3%; in total n=8/28, 28.6%). Half of these patients experienced local disease progression, while the other half developed distant metastases.
Resectability and response rates of group I and II.
Correlation of response with changes in CA19-9 levels. Serum CA19-9 levels were available from 26 patients (87%) both before and after chemotherapy. Four of these patients (15%) had normal CA19-9 levels (<37 kU/l) before neoadjuvant therapy and also demonstrated normal levels during and after completing therapy. The overall median baseline CA19-9 level at study inclusion was 355 kU/l (±2138 SD) and decreased to 48 kU/l (±918 SD) after neoadjuvant bevacizumab/gemcitabine therapy (p=0.005). However, there was no statistically significant difference in the median change of CA19-9 level between patients who demonstrated progressive disease (62 kU/l ±1974 SD) and patients who had stable disease or partial response (233 kU/l ±2114 SD; p=0.940). Likewise, there was no statistically significant difference in the median change of CA19-9 level between patients who underwent resection and those who did not (88 kU/l ±244 vs. 235 kU/l ±2474; p=0.4), nor there was a significant difference between the two treatment groups (I: 210 kU/l ±180 vs. II: 139 kU/l ±2407; p=1.0).
Resectability and treatment outcome. Sixteen patients out of 30 (53%) underwent surgical exploration after neoadjuvant therapy, and definite resection was performed in 11 of them (overall: 37%). When comparing the final resectability between patients deemed to have unresectable vs. borderline-resectable disease before neoadjuvant treatment, no statistically significant difference was noted, but there was a trend for better resectability rates in patients with disease of borderline resectability before therapy (26% vs. 55%; p=0.122). There was no difference in resectability between the two treatment groups (Table II).
Tumor and histological characteristics according to the AJCC criteria 2002 (21).
Pancreaticoduodenectomy was performed in nine patients, one patient underwent total pancreatectomy, and distal pancreatectomy was performed in one case. Partial portal vein resection was necessary in three patients. The median operative time was 350 min (range=175-570 min) and the median postoperative length of hospital stay was 23 days (range=9-146 days).
Operative morbidity and mortality. Overall surgical morbidity was 72% (n=8/11) and most (55%) were technically major complications (Clavien-Dindo-classification ≥3). There were three (n=3/11, 27%) fatalities related to postoperative complications on postoperative days 19, 71 and 85, respectively. Morbidity and mortality were caused by necrotizing pancreatitis of the pancreatic remnant and necrosis of the gastric wall after distal pancreatectomy; recurrent cholangitis with multiple liver abscesses; necrosis of the hepaticojejunostomy due to arterial liver infarction; common bile duct necrosis and biliary leakage leading to overwhelming sepsis. Other observed complications included postoperative hematoma (N=1) and subphrenic fluid collection requiring interventional drainage (N=1).
Overall survival by resection vs. no resection after neoadjuvant NeoAvastin therapy: 13 months (95% confidence interval=11 to 15 months) vs. 13 months (95% confidence interval=11 to 14 months, respectively); p=0.131.
Pathological grading and staging. Tumor staging according to the AJCC sixth edition is listed in Table III. Meticulous pathological assessment confirmed R0 resection in 10 out of 11 patients who underwent. Most patients had T3 node-positive pancreatic cancer (n=7/11, 64%) and demonstrated only minor pathological response after neoadjuvant treatment (n=9/11, 82%).
OS and disease recurrence. With a median follow-up of over 36 months, 28 patients died and 2 are still alive. In an intention-to-treat analysis, the median OS was 13 months (95% CI=11.9-14.1 months) for the whole cohort. Median OS for patients undergoing curative resection was 13 months (95% CI=10.8-15.2 months), and 13 months (95% CI=11.6-14.4 months) for those patients without resection (p=0.131, Figure 1). Median OS for patients undergoing curative resection, excluding all patients who suffered lethal post-surgical complications, was 17 months (95% confidence interval=9.3-24.7 months).
The median time-to-treatment failure was eight months (95% CI=0-20 months) in the group of resected patients. Three patients had distant metastasis and three patients showed local recurrence as their first relapse, the latter including the patient whose resection was eventually classified as R1. Two patients (both treatment group II, R0 resection), are currently still alive without any signs of recurrence (at 23 and 59 months).
Discussion
Due to the unfavorable prognosis of locally unresectable pancreatic cancer, neoadjuvant strategies have been intensively studied for the past decade. While the search for the most effective neoadjuvant treatment modality has yet to be completed, various trials and meta-analyses have shown that curative resection after down-sizing of locally advanced pancreatic cancer is feasible in a subset of patients [median 33.2%; range=25.8%-41.1% (6)], leading to prolonged survival in individual cases. At the time this study was initiated, clinical trials in patients with colorectal, breast, and non-small cell lung cancer had shown significant benefit from adding bevacizumab to a reference chemotherapy (15, 30, 31). Likewise, a phase II study of bevacizumab and gemcitabine therapy in advanced pancreatic cancer conducted by Kindler et al. initially demonstrated promising results, with a response rate of 21% and a median OS of 8.8 months (16). Unfortunately, these results were not confirmed by a subsequent randomized phase III trial conducted by the same group, published in 2010 (32). Median OS in the definitive study was 5.8 months for gemcitabine plus bevacizumab and 5.9 months for gemcitabine-plus-placebo (p=0.95). Median progression-free survival was 3.8 and 2.9 months, respectively (p=0.07). Overall response rates were 13% and 10%, respectively.
In this phase II study of neoadjuvant bevacizumab and gemcitabine therapy, resection with curative intent could eventually be performed in 6 out 11 initially borderline resectable and in 5 out of 19 patients with locally advanced non-metastatic pancreatic cancer at baseline. In contrast to previous studies of gemcitabine-based neoadjuvant therapy conducted at our Institution (33, 34), the addition of bevacizumab did not demonstrate a significant outcome of survival benefit in patients undergoing resection in an intention to treat analysis. One of the reasons for this, however, could be the unfortunately unusually high perioperative morbidity and mortality we encountered in this series of patients: Those patients undergoing preoperative treatment and surgery without major complications had a median survival of 17 months in this trial, which compares favorably to previously reported series of neoadjuvant treatment in locally advanced pancreatic cancer (6, 33, 34).
With respect to anti-VEGF treatment duration, we did not observe any differences in response rates, resectability, or OS between patients receiving a total of three doses of bevacizumab versus patients receiving six doses, which in consistent with a recent analysis of bevacizumab treatment in patients with cancer, demonstrating that patients treated with higher dose of bevacizumab did not have a more favorable outcome (17). Recorded toxicity was comparable to other studies investigating gemcitabine–bevacizumab regimes (16, 32). In patients ineligible for pancreatic surgery after neoadjuvant bevacizumab, a median survival of 13 months was observed, which is comparable to reported survival rates of non-resected patients in two recent meta-analyses of neoadjuvant therapy in locally advanced pancreatic cancer (17, 18).
Of note, even though we observed radiologically-stable disease within the majority of patients (68%), and only one patient with partial response, the eventual overall resection rate was 37% (N=11/30) and even disease in one patient that did not respond at all radiologically was later successfully resected. This discrepancy between even higher levels of academic radiology and surgical reality is mainly caused by currently unsurpassable difficulties in differentiating pancreatic fibrosis (desmoplasia) in pancreatic cancer, particularly in patients with somewhat limited response to neoadjuvant therapy. Similar discrepancies between imaging studies and the surgically-determined actual local tumor spread has been observed in prior studies conducted at our Institution, and was also described by others before (35, 36).
As a consequence, we believe that selected patients (i.e. those with good performance status) should undergo surgical exploration even when high-resolution imaging shows no particular down-staging after neoadjuvant therapy. Unfortunately, reliable biological markers to accurately monitor true response to neoadjuvant treatment are as yet unavailable. In this study, CA19-9 levels monitored during preoperative gemcitabine–bevacizumab therapy significantly decreased in most patients, but no significant correlation was observed with respect to the change of CA19-9 levels between patients able to undergo resection and those who were not.
There have been concerns that the administration of bevacizumab close to major surgical procedures would raise the risk of postoperative hemorrhage, and potentially disrupt the healing process. For hepatobiliary surgery, it has been demonstrated that the incidence and severity of complications are not influenced by the addition of bevacizumab to drug therapy prior to surgery (37-39). In our study, 4 out of 11 patients undergoing pancreatic resection after neoadjuvant bevacizumab treatment suffered major surgical complications. Although the observed morbidity rate was markedly higher compared to our prior neoadjuvant studies or the usual Institutional standard of primary pancreatic surgery conducted, it remains unclear whether this observation was in fact related to preoperative bevacizumab. It is of note, however, that in two patients of this series, severe surgical complications were caused by (highly uncommon) necrosis of the common bile duct and consequent leakage of the biliodigestive anastomosis. Another patient suffered the complication of postoperative gastric perforation due to partial necrosis of the gastric wall after distal pancreatectomy, also extremely unusual. Intestinal perforations have been reported after bevacizumab use in gastrointestinal and gynecological cancer, occurring in some series in up to 15% of cases (40, 41). Due to the limited number of patients in the present study, we do not consider it appropriate to draw definite conclusions about a bevacizumab-caused increase in perioperative morbidity and mortality.
Due to the major role of angiogenesis in tumor growth and progression, targeting neo-angiogenesis, e.g. VEGF, as a therapeutic means has long been proposed. Unfortunately, the impact of bevacizumab addition to different chemotherapeutic agents is variable and the benefits are often only transitory. Since the first approval of bevacizumab in 2004 for metastatic colon cancer, new insights have been gained to the mechanisms of resistance to antiangiogenic therapies. These include protection of the tumor vasculature either by recruiting proangiogenic inflammatory cells or by protective pericyte coverage, up-regulation of alternative proangiogenic signals and increased invasiveness of tumor cells into local tissue to co-opt normal vasculature (42). In addition, data from our linked translational research project revealed that the myelosuppressive effect of gemcitabine treatment seems to exert unfavorable effects in combination with bevacizumab. Myelosuppression (thrombocytes and monocytes) seems to actually prevent the mobilization of proangiogenic cell populations, which are the cellular target of anti-VEGF therapy (28).
In conclusion, this trial represents a single-Center experience with patients with borderline-resectable and locally advanced pancreatic cancer who were treated with neoadjuvant systemic therapy. The initial hope that adding anti-angiogenic treatment to standard chemotherapy would increase benefits for patients was not fulfilled. This is in line with other reports about the disappointing lack of clinical efficacy of bevacizumab in pancreatic cancer.
Overall, the addition of Bevacizumab did not notably alter response to neoadjuvant chemotherapy or overall patient outcomes. Despite the encouraging observation that selected patients responded well enough to this combination that radical resection became feasible, and even occasional long-term survival was observed, we do not recommend this approach: Adding bevacizumab to neoadjuvant gemcitabine treatment most likely does not offer improved prospects for patients with pancreatic cancer.
Footnotes
-
Grant Support
This work was supported by the Austrian National Bank (OeNB grant no. 12072 issued to S.F. Schoppmann). The study was furthermore supported by an unrestricted grant from Roche Pharmaceuticals.
-
Conflicts of Interest
None of the Authors have any financial and personal relationships with other people or organizations that could inappropriately influence their work.
- Received January 31, 2014.
- Revision received March 10, 2014.
- Accepted March 12, 2014.
- Copyright© 2014 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
