Long-term analysis of two prospective studies that incorporate mitomycin C into an adjuvant chemoradiation regimen for pancreatic and periampullary cancers
SUMMARY
This is the first prospective study with long-term follow-up that evaluates mitomycin C with standard split-course or continuous 5-fluorouracil-based adjuvant chemoradiation in resected pancreatic or periampullary cancer. Incorporation of mitomycin C appears safe and effective in the adjuvant setting. In an era in which the resected periampullary tumor can be sequenced and analyzed for mutations in DNA repair pathways, it may be worthwhile to integrate mitomycin C into a standard regimen for the patients.
ABSTRACT
Purpose: To report toxicity and long-term survival outcomes of two prospective trials evaluating mitomycin C (MMC) with 5-fluorouracil-based adjuvant chemoradiation in resected periampullary adenocarcinoma. Methods: From 1996 to 2002, 119 patients received an adjuvant 4-drug chemotherapy regimen of 5- fluorouracil, leucovorin, MMC, and dipyridamole with chemoradiation on two consecutive trials (Trial A and B). Trial A patients received upfront chemoradiation (50 Gy split-course, 2.5 Gy/fraction) followed by four cycles of the 4-drug chemotherapy with bolus 5-fluorouracil. Trial B patients received one cycle of the 4-drug chemotherapy with continuous infusion 5-fluorouracil followed by continuous chemoradiation (45-54 Gy, 1.8 Gy/fraction) and 2 additional cycles of chemotherapy. Cox proportional hazards models were performed to identify prognostic factors for overall survival (OS).
Results: Of the 62 Trial A patients, 61% had pancreatic and 39% non-pancreatic periampullary carcinomas. Trial B (n=57) consisted of 68% pancreatic and 32% non-pancreatic periampullary carcinomas. Resection margin and lymph node status were similar for both trials. Median follow-up was longer for Trial A than Trial B (197.5 vs. 107.0 months), with median OS of 32.2 and 24.2 months, respectively. Rates of 3-, 5-, and 10-year OS were 48%, 31%, and 26% in Trial A and 32%, 23%, and 9% in Trial B. On multivariate analysis, lymph node-positive resection was the strongest prognostic factor for OS. A pancreatic primary and positive margin status were also associated with inferior survival (p<0.05). Rates of grade ≥3 treatment-related toxicity in Trials A and B were 2% and 7%, respectively. Conclusions: This is the first study to report long-term outcomes of MMC with 5-fluorouracil-based adjuvant chemoradiation in periampullary cancers. As MMC may be considered in DNA repair-deficient carcinomas, randomized trials are needed to determine the true benefit of adjuvant MMC.
INTRODUCTION
Periampullary adenocarcinomas originate in one of four anatomical locations—the pancreatic head or uncinate process (PDAC), distal common bile duct (DCBD), ampulla of Vater, or duodenum. Their incidence has been increasing and the associated prognosis is generally poor. Projected 5-year post- resection survival rates range from 7% to 29% for patients with PDAC and 23% to 69% for patients with non-PDAC periampullary adenocarcinoma.1-8No standard adjuvant therapy regimen has been established for these patients and management is typically extrapolated from PDAC paradigms as treatment strategies continue to evolve.4,9 Historic studies used split-course radiotherapy,10-12 while modern studies incorporate continuous radiotherapy.13-15 Previously, Isacoff and colleagues reported improved outcomes with 5-fluorouracil (5-FU), leucovorin (LV), mitomycin C (MMC), and dipyridamole (DPM) in patients with locally advanced PDAC.16,17 Studies have suggested that MMC may be most effective in PDAC patients with certain mutations, in particular those who have a family history of PDAC and/or harbor mutations in a gene coding for DNA repair proteins (such as BRCA2 or PALB2).18-23 With the current rise in next-generation sequencing and precision medicine, the impact of DNA-intercalating agents such as MMC on outcomes in patients with such dismal prognoses as periampullary cancer is brought into question.In an effort to further investigate the efficacy of MMC integration with 5-FU-based chemoradiotherapy (CRT) for these malignancies, we enrolled patients with resected periampullary adenocarcinoma on a pair of prospective trials. In 1996 and 1999, respectively, we initiated two consecutive clinical trials incorporating adjuvant 5-FU, leucovorin, MMC and DPM with CRT for patients with resected periampullary adenocarcinoma. Patients in Trial A received upfront CRT with bolus 5-FU and split- course of CRT as outlined previously,1 with timing similar to the Gastrointestinal Tumor Study Group (GITSG) trial, followed by four cycles of the same chemotherapy.
The second trial (Trial B) incorporated one cycle of the 4-drug chemotherapy with continuous infusion (CI) 5-FU followed by acontinuous course of CRT and two additional cycles of the same chemotherapy. Herein, we present the long-term clinical and therapeutic outcomes of the first two prospective clinical trials with long-term follow-up to evaluate the integration of MMC with standard 5-fluorouracil-based adjuvant CRT in resected PDAC or non-PDAC periampullary cancer.Both Trials A and B were approved by our Institutional Review Board (IRB). The study populations consisted of patients with PDAC or non-PDAC periampullary adenocarcinoma who underwent curative resection. Patients with negative (R0), microscopic (R1), or macroscopic residual disease (R2) at the time of resection were eligible. Restaging after surgery included complete history and physical examination, computed tomography (CT) scan of the chest/abdomen/pelvis, and laboratory studies. The final cohort included 62 patients in Trial A and 57 patients in Trial B.Eligibility criteria included: (1) age ≥18 years; (2) Karnofsky Performance Status ≥70%; (3) adequate bone marrow function; (4) adequate hepatic function; and (5) adequate renal function. Patients were excluded for: (1) prior malignancy within 5 years; (2) prior abdominal irradiation; (3) distant metastatic disease; and (4) poorly controlled medical condition(s) that could be exacerbated by the treatment.All patients underwent a pancreaticoduodenectomy. Resection margins were positive (R1) if the carcinoma was close (within 1 mm) or present at the final soft tissue margin. Gross residual disease (R2) was based on the surgical report and/or residual disease seen on first postoperative imaging.Postoperatively, patients were evaluated by radiation and medical oncologists to discuss treatment options and determine eligibility.
Therapy schemas are outlined in Supplemental Table 1. Radiation for Trials A and B consisted of 15-MV photons. All patients were simulated on the Picker AcQ Sim-CT simulator (Picker, Inc. Cleveland, OH) and treatment planning was completed using computerized dosimetry. Isodose curves were generated on axial slices at the isocenter and at least two additional levels. Critical normal organs at risk (OARs) and tumor target volumes were electronically contoured by one radiation oncologist. The treatment volume was designed to include the preoperative tumor volume, primary lymph node drainage stations, and the para-aortic lymph nodes with a 1.5- to 2-cm margin. The encompassed vertebral body levels were T11– L4 inclusive. For both trials, the dose range was 50.4-54 Gy.Patients were treated on Trial A from April 1996 to July 1999.1 Chemotherapy details can be found in Supplemental Table 1. Radiation was delivered as a split course of 50 Gy with a 2-week break after the initial delivery of 25 Gy. Irradiation was delivered using 3- or 4-field technique, custom alloy blocking,2.5 Gy per fraction, 1 fraction per day, and 5 fractions per week. The daily spinal cord dose did not exceed 1.9 Gy per fraction. Portions of the kidney, specifically the right kidney, received a full dose although the treatment plan ensured that 50% of the functioning renal parenchyma received no more than 35% of the daily dose or a total of 17.5 Gy. Radiation and chemotherapy began concurrently on Day 1, 41-86 days after surgery. Two cycles of chemotherapy were administered during radiation followed by four additional cycles of identical chemotherapy.Patients were treated on Trial B from August 1999 to April 2002.
Chemotherapy was administered similarly to Trial A, with the exception that patients received CI 5-FU and the dose of MMC was 8 mg/m2/day (2 mg less). Radiation was delivered to patients in Trial B according to the technical aspects as outlined above for Trial A. However, daily fractions of 1.8 Gy were delivered continuously (with noplanned break) for 25-30 fractions. Furthermore, one cycle of chemotherapy was administered prior to CRT and 2 additional cycles were delivered following CRT.All patients were evaluated for toxicity weekly during therapy. Toxicity was graded using the National Cancer Institute Common Toxicity Criteria version 1.0 (CTCAE 1994). Treatment was held for ANC<1,000/mm3 and platelet count <75,000/mm3. For any non-hematologic toxicity of grade ≥3 attributable to radiation, further treatments were held until toxicity resolved to grade ≤1 toxicity. Radiation dose escalation was not allowed. If treatment delayed radiation for >12 weeks, the patient was removed from the study.Modifications of 5-FU, LV, DPM, and MMC were based on nadir ANC and platelet counts or worst- grade non-hematologic toxicity attributable to these chemotherapies in the immediately preceding cycle. The 5-FU/LV/DPM was delayed until toxicity had resolved to grade 0 or 1. The dose of DPM was decreased by 25% for DPM-related toxicities such as headache, whereas 5-FU and LV doses were not modified.Hematologic toxicities were measured at different time points on the two protocols. For the split-course treatment in Trial A, toxicity was assessed twice, once during CRT and once after CRT. Toxicities were assessed once after CRT in Trial B.Paper and electronic charts as well as the National Familial Pancreatic Tumor Registry (NFPTR)25 were reviewed to examine possible correlations of family history of cancer with OS.
A family history of PDAC was defined as having at least one first degree relative with a diagnosis of PDAC. A family history ofbreast or ovarian cancer was defined as having at least one first or second degree relative with either one of these carcinomas.The primary statistical endpoints are toxicity and OS of patients treated on Trials A and B. Toxicity is reported descriptively. Follow-up information was obtained from medical records, with restaging occurring every three months for Year 1, every four months for Year 2, and every six months thereafter. Outcomes between the two trials were not directly compared because the study designs were single-arm and treatment was not randomized. Event time distributions were estimated using the Kaplan-Meier method26 and comparisons within each trial were made using the log-rank statistic27 or the proportional hazards regression model.28 All factors in Table 1 in addition to family history of PDAC, gastrointestinal cancers, and breast or ovarian cancer were tested for an association with OS. Median follow-up was calculated with the reverse Kaplan-Meier method. Binomial probabilities were compared with Chi-square tests and reported with exact binomial 95% confidence intervals. The Cochran-Mantel-Haenszel (CMH) test was used to evaluate the association of PDAC with positive margins stratifying for protocol. The Breslow-Day test for homogeneity of odds ratios was used to confirm assumptions of stratified analyses. Two-sided significance testing was used throughout the analysis, and a p-value of 0.05 was considered statistically significant.
RESULTS
Demographic, tumor, and treatment characteristics for both trials are given in Table 1. The patients on these protocols had very similar demographics and disease characteristics. The median (interquartile range, IQR) age of patients on Trials A and B was 60 (IQR: 56, 67) and 59 (IQR: 54, 67), respectively. Both trials had approximately 60% of patients with PDAC, included over 92% Caucasian patients, and were approximately evenly split on gender. Tumor characteristics were mostly similar (Table 1). One notable difference was radical lymph node dissection, which was more common on Trial A (31%, 95%CI: 19.56, 43.65%) compared to 12% (95% CI: 5.08, 23.68%) on Trial B likely due to a separate overlapping surgical clinical trial.29-31Median follow-up was 197.5 months for Trial A and 107.7 months for Trial B. Median OS in Trial A was32.2 months and 24.2 months in Trial B (Figure 1). Rates of 3-, 5-, and 10-year OS were 48%, 31%, and 26% in Trial A and 32%, 23%, and 9% in Trial B. Univariate OS analyses for both trials are shown in Table 2.On Trial A, a PDAC diagnosis, margin-positive resection, lymph node-positive resection, and T stage 3-4 were associated with decreased OS (Table 2). Although PDAC patients in Trial A had a significantly inferior OS than non-PDAC patients (median: 16.9 vs. 49.9 months, p=0.016), long-term OS rates in PDAC were impressive (34% 3-year, 21% 5-year, and 21% 10-year; Figure 2A). Patients on Trial A with a family history of PDAC had a trend towards improved OS (median: 164.4 vs. 28.7 months, p=0.058), while those with a family history of breast or ovarian cancer had decreased OS (p=0.039; Supplemental Figure 1).A PDAC diagnosis was also associated with decreased OS in Trial B (Table 2). Additional risk factors for worse OS included female gender, tumor size ≥3 cm, PNI, and poor differentiation (Table 2).
Of note, the 3-, 5-, and 10-year OS rates for patients with PDAC on Trial B were 21%, 15%, and 0% and 56%, 39%, and 28% for non-PDAC patients (Figure 2B). A family history of cancer, PDAC or breast/ovarian, was not associated with OS in this trial. The few patients with radical lymph node dissections (n=7) had significantly improved OS (p=0.015).In multivariate analysis, lymph node-positive resection was the strongest factor associated with OS (Table 3). Positive margins were more likely in patients with PDAC (31.2% vs. 9.5%, CMH p value = 0.01).This correlation makes it difficult to determine whether the PDAC diagnosis or margin status is a greater contributing factor to inferior OS; however, each factor alone is also significantly associated with decreased OS after adjusting for lymph node status (all p<0.05).Adjusting for other significant factors on multivariate analysis, a diagnosis of PDAC was also marginally associated with decreased OS in Trial B (Table 3). Margin status was not a significant factor for patients on Trial B on univariate or multivariate analyses. Factors significantly associated with worse OS included female gender, PNI, and poor differentiation. A radical lymph node dissection was not significantly associated with OS on multivariate analysis when adjusted for other factors.For strictly descriptive purposes, we also report characteristics of long-term survivors defined as ≥5 years of OS from surgery. Trial A had 19 patients who survived longer than 5 years in comparison with 11 patients in Trial B, for a total of 30 of 119 (25%) long-term survivors.
Long-term survivors were more likely to be male (63%) and had a median age of 60 (IQR: 56-65). The majority of long-term survivors had PDAC (43%) followed by DCBD (30%), ampullary (17%), and duodenal (10%) carcinomas. The majority had a margin-negative resection (87%) and moderately differentiated tumors (47%), with a median tumor size of 3.0 cm (IQR: 2.0-4.0).Non-hematologic and hematologic grade ≥3 toxicities on both trials are itemized in Table 4. Overall, rates of grade ≥3 non-hematologic toxicity in Trials A and B were 14.4% and 23.4%, respectively. For all types of non-hematologic toxicity, the proportion on either protocol was ≤5%. The acute and late grade ≥3 toxicity rates in Trial A were 6.4% and 8.0%, respectively. In Trial B, there were more late versus acutegrade ≥3 toxicities (19.8% vs. 3.6%, respectively). All 9 adverse events on Trial A were stand-alone events, whereas, on Trial B, 1 patient had 3 adverse events, another patient had 2 adverse events, and the remaining 8 patients had 1 adverse event. Two patients on Trial A had a grade 5 toxicity, with one patient dying of gangrenous bowel six months after local disease recurrence and another experiencing a gut infarct 3 years after CRT. No grade 5 toxicities were observed in Trial B. Upon further evaluation, only one (1.6%) adverse event was deemed attributed to CRT in Trial A, while 4 (7.2%) adverse events were considered attributable to CRT in Trial B.
DISCUSSION
Through next-generation sequencing, it is now possible to determine whether a patient’s cancer has mutations in DNA repair pathways.32 Studies suggest integration of MMC into the treatment paradigm of patients with these mutations may result in improved outcomes. 33,34 This is the first prospective study with long-term follow-up to evaluate this approach. In summary, these two clinical trials suggest that MMC may be safely incorporated into either split-course or continuous 5-FU-based (CI or bolus) adjuvant CRT regimens in patients with resected PDAC or non-PDAC periampullary adenocarcinoma. Furthermore, these regimens (Trial A and B) resulted in as high as an 18-month OS benefit in comparison with other studies.2,6,7,35-37 Also consistent with the literature, patients with non-PDAC periampullary tumors were found to have a more favorable prognosis and extended OS after surgical resection than patients with PDAC. The rationale behind the novel adjuvant regimens used in this study was based on innovative agents including MMC and DPM under investigation during the development of the study.16 While there was a larger proportion of long-term survivors (>5 years) in Trial A, it is unclear if this difference is attributable to patient selection, the duration or type of chemotherapy, or the split-course radiation treatment (2.5 vs. 1.8 Gy/fraction) which had about 5-8% more biological equivalent dose (BED) for α/β ratios of 10 or 6, respectively, and about 15% more potent for an α/β ratio of 3 (not counting repopulation). In addition, the split-course regimen in Trial A was associated with similar rates of toxicity when compared to the continuous CRT regimen utilized in Trial B, therefore suggesting that MMC can be safely integrated into adjuvant CRT for patients with periampullary cancers, even when combined with bolus 5-FU (in Trial A) which has been shown to be associated with unfavorable toxicity.38
Consistent with previous studies, positive resection margins and lymph nodes appear to be independently associated with worse OS.3,7,39,40 Because neoadjuvant therapy has been associated with a higher likelihood of margin- and node-negative resection in patients with borderline resectable and resectable PDAC,41,42 delivery of chemotherapy and/or CRT prior to surgery may confer improved local control and long-term survival in patients with resectable PDAC and/or non-PDAC periampullary adenocarcinoma.43,44 Earlier detection of these tumors and a better understanding of the genomic profile prior to surgery may be coupled with neoadjuvant therapy to select patients who may benefit from surgery. Positron emission tomography (PET) imaging, in particular, may allow for detection of nodal disease prior to resection. In fact, one study suggests that lymph nodes with maximum standard uptake value (SUVmax) ≥2.8 were an independent factor for prognosis after resection.To our knowledge, this is the largest report prospectively evaluating the role of MMC in the adjuvant management of periampullary cancers.1,47 MMC has been suggested to convert unresectable patients to surgical candidates,43 enhance the therapeutic effects of adjuvant CRT,48 and improve local control49 in periampullary adenocarcinoma. Although it has not been extensively studied in PDAC, MMC has been used in other cancers such as anal cancer50-52 and recent studies suggest its promising role in PDAC patients with a family history of PDAC and/or carcinomas with an inactivating mutation in a Fanconi’s anemia pathway gene (such as BRCA1, BRCA2 or PALB2 ).18-23 Study results from Trial A demonstrated that having a family history of PDAC specifically resulted in a borderline significant improvement in OS (164.4 vs. 28.7 months, p=0.058).
In contrast, however, having a family history of breast or ovariancancer was associated with inferior OS (22.6 vs. 34.4 months, p=0.039). These conclusions are limited by the fact that there were only a small proportion of patients who fit these criteria; therefore, larger trials are needed to further determine whether MMC is beneficial in one subgroup or another.In addition to the relatively small sample size, there are numerous limitations to this study. In order to achieve an analyzable sample size for these rare cancers, all periampullary cancers were included in the study; however, this increases the difficulty with which the results can be interpreted due to the differences in prognoses. To acknowledge this, we separated survival outcomes into PDAC and non- PDAC periampullary cancers. Of note, the AJCC staging system at the time in which patients were enrolled on these clinical trials differs from that in current use today; although it appears that patients with advanced disease were enrolled on the clinical trials, this is not the case and, as such, T stage was not included in the multivariate model. Furthermore, the study design of these historic prospective trials is outdated with the use of bolus 5-FU and split-course CRT. Nonetheless, the purpose of this study is to share the long-term results of these studies and suggest that MMC can indeed be combined with standard adjuvant CRT in periampullary cancers and perhaps be beneficial in patients with mutations in the DNA repair pathway.These prospective studies indicate that an MMC may be safely incorporated into a 5-FU-based CRT regimen in patients with PDAC and non-PDAC periampullary cancers, with promising results of long- term survival. A family history of PDAC correlated with a trend toward improved OS in Trial A. It is unclear if these improvements in OS are due to MMC and will Mitomycin C therefore need to be prospectively evaluated in a larger, randomized series with integrated cancer sequencing for DNA repair defects.