Overall survival according to resection margin status. CI indicates confidence interval; HR, hazard ratio.
Overall survival rates according to resection margin status and whether or not chemoradiotherapy (CRT) was included in the treatment.
Overall survival rates according to resection margin status and whether or not chemotherapy (CT) was included in the treatment.
Overall survival according to different combinations of randomized treatment in patients who had clear resection margins (R0). CRT indicates chemoradiotherapy; CT, chemotherapy; and OBS, observation.
Overall survival according to different combinations of randomized treatment in patients who had microscopically involved resection margins (R1). CRT indicates chemoradiotherapy; CT, chemotherapy; and OBS, observation.
Butturini G, Stocken DD, Wente MN, Jeekel H, Klinkenbijl JHG, Bakkevold KE, Takada T, Amano H, Dervenis C, Bassi C, Büchler MW, Neoptolemos JP, Pancreatic Cancer Meta-analysis Group. Influence of Resection Margins and Treatment on Survival in Patients With Pancreatic CancerMeta-analysis of Randomized Controlled Trials. Arch Surg. 2008;143(1):75-83. doi:10.1001/archsurg.2007.17
To assess the influence of resection margins and adjuvant chemoradiotherapy or chemotherapy on survival for patients with pancreatic cancer by meta-analysis of individual data from randomized controlled trials.
Structured MEDLINE search for published studies.
A meta-analysis of published randomized controlled trials and individual data.
Individual data were obtained from 4 recently published trials (875 patients: 278 [32%] with R1 and 591 [68%] with R0 resections).
Kaplan-Meier estimates of survival were compared using log-rank analyses. Pooled hazard ratios of the effects of chemoradiotherapy and chemotherapy treatments on the risk of death were calculated separately and across groups according to resection margins status. Six hundred ninety-eight patients (80%) had died, with a median follow-up of 44 months in the surviving patients. Resection margin involvement was not a significant factor for survival (hazard ratio [HR], 1.10; 95% confidence interval [CI], 0.94-1.29; log-rank χ2 = 1.4; P = .24). The 2- and 5-year survival rates, respectively, were 33% and 16% for R0 patients and 29% and 15% for R1 patients. Chemoradiotherapy in R1 patients resulted in a 28% reduction in the risk of death (HR, 0.72; 95% CI, 0.47-1.10) compared with a 19% increased risk in R0 patients (HR, 1.19; 95% CI, 0.95-1.49). Chemotherapy in R1 patients had a 4% increased risk of death (HR, 1.04; 95% CI, 0.78-1.40) compared with a 35% reduction in risk in the R0 subgroup (HR, 0.65; 95% CI, 0.53-0.80).
Adjuvant chemotherapy but not chemoradiotherapy should be the standard of care for patients with either R0 or R1 resections for pancreatic cancer.
Pancreatic cancer remains a leading cause of cancer death in the Western world with about 31 000 and 65 000 deaths per year in the United States and Europe, respectively.1- 3 Curative resection of pancreatic cancer can be performed only in 10% to 15% of patients owing to advanced stage and an overall 5-year survival rate of about 10%.1,2,4 Improvements in surgical technique have led to a significant reduction of perioperative morbidity and mortality in experienced centers in the subgroup of patients with resectable disease.5- 7 The role of adjuvant treatment, either as radiotherapy or chemotherapy, might further improve survival. The European Study Group for Pancreatic Cancer 1 (ESPAC-1) trial was the first study powered to evaluate the effects of both adjuvant radiotherapy and chemotherapy, concluding no survival benefit for chemoradiotherapy but a significant survival benefit for chemotherapy.8,9
Several reports have shown resection margins to be an important factor for the long-term prognosis of patients after pancreatic cancer resection.5- 12 Resection margin classification is the widely accepted method to define the extent of resection in patients with pancreatic cancer. An R0 resection is one that has no microscopic or macroscopic residual tumor in the specimen; an R1 has microscopic residual tumor on the surgical edges of the specimen; and an R2 defines the macroscopic residual cancer after a surgical procedure initially intended to be radical. Patients were divided into 3 different subgroups according to these definitions; each of them has been treated differentially in the past. Early prospective studies excluded R1 resections from adjuvant treatment and the R2 subgroup. That patients with R1 resections have a worse survival rate than R0 resection implies the need to stratify patients in controlled randomized trials of adjuvant treatment. The leading question is whether the R1 patients should be treated as part of the R0 subgroup in the adjuvant setting. The involvement of resection margins was shown to be an influential prognostic factor in the interim analysis of ESPAC-1,8 whereas earlier prospective studies were too small to include this factor.13,14 A detailed analysis of prognostic factors based on the interim analysis of the ESPAC-1 trial concluded that resection margin–positive pancreatic tumors represented a more biologically aggressive cancer.12 These patients benefit from resection and adjuvant chemotherapy but not chemoradiotherapy. The magnitude of benefit for chemotherapy treatment, however, was reduced for patients with R1 margins vs those with R0 margins.12
In the final report of the ESPAC-1 trial, which included patients randomized in the original 2 × 2 design, the effect of resection margin status on survival was again not significant (P = .10).9 The overall survival benefit for adjuvant chemotherapy, but not chemoradiotherapy, was confirmed in a recently published meta-analysis using individual patient data.15 Subgroup analyses, however, estimated that chemotherapy was less effective and chemoradiotherapy was more effective in patients with cancerous resection margins. Our aim was to further investigate the influence of resection margins through a meta-analysis of randomized controlled trials, providing more detail for the planning of subsequent trials to specifically investigate resection margin status.8,9,13- 17
A protocol for the inclusion of trials into the meta-analysis was followed and has been previously reported.15 All included trials randomized patients with pancreatic ductal adenocarcinoma to adjuvant therapy following resection for pancreatic cancer. Patients with tumors other than pancreatic ductal adenocarcinoma were excluded. Patients were grouped according to the type of randomized adjuvant treatment (chemoradiotherapy or chemotherapy) and resection margin status at randomization (R0 or R1 resection).
The main outcome measure for analysis was overall survival measured from date of operation to the date of death (from all causes) or censor date. Kaplan-Meier18 estimates of survival were compared using standard and stratified log-rank tests.19 The log-rank expected numbers of deaths and variance were used to calculate individual and pooled hazard ratios (HRs) and confidence intervals (CIs). Pooled HRs indicating the effects of chemoradiotherapy and chemotherapy treatments on the risk of death and χ2 tests for statistical heterogeneity (interaction) were calculated separately and across groups according to resection margin status.
Four trials of adjuvant treatment for patients with resected pancreatic ductal adenocarcinoma were identified with individual data available on 875 patients.8,9,13- 17 Important tumor and clinical characteristics, including resection margin status, were included for the studies as reported previously.15 The ESPAC-1 trial used a 2 × 2 factorial design and therefore randomized patients for both chemoradiotherapy and chemotherapy treatments. The meta-analysis included 477 patients who were randomized for chemoradiotherapy treatment (241 to receive chemoradiotherapy; 236 not to receive chemoradiotherapy) and 681 patients who were randomized for chemotherapy treatment (345 to receive chemotherapy;
336 patients not to receive chemotherapy).
Tumor characteristics across resection margin groups are presented in table 1. Two hundred and seventy-eight (32%) patients had R1 resections and 591 (68%) had R0 resections at the time of randomization. Clinical and tumor characteristics generally appeared balanced between the R0 and R1 groups except for an increased proportion of node-positive patients in the R1 group (63% vs 49%).
Overall, 698 of the 875 patients (80%) had died, with a median follow-up of at least 24 months in surviving patients within individual trials and 44 months overall (interquartile range [IQR], 24.9-63.8 months). Two hundred twenty-six (81%) R1 patients and 467 (79%) R0 patients had died. Median follow-up of the 52 R1 patients who were still alive was 50 months (range, 0-105; IQR, 29-67 months) compared with 40 months (range, 0-106; IQR, 25-63 months) for the 124 living R0 patients.
Resection margin involvement was not a statistically significant prognostic factor (HR, 1.10; 95% CI, 0.94-1.29), log-rank χ2 = 1.4; P = .24) (Figure 1), though there was a trend for R1 patients to do worse, with a median survival of 14.1 months (95% CI, 11.9-16.4) compared with 15.9 months for patients with R0 resections (95% CI, 14.6-17.4)
(Table 2). The 2-year and 5-year survival rates, respectively, were 33% and 16% for R0 patients and 29% and 15% for R1 patients.
Univariate log-rank analyses confirmed that tumor grade (log-rank χ2 = 34.96, P < .001), nodal involvement (log-rank χ2 = 34.48, P < .001), and tumor size (log-rank χ2 = 6.07, P = .01) were statistically significant prognostic factors of survival, though grade was missing from the Norwegian trial13 data and tumor size was missing from the Japanese17 trial data. Hazard ratio estimates (Table 2) were similar across resection margin groups for the effect of age and nodal involvement. Patients with moderately or poorly differentiated tumors had increased risk of death in both the R0 and R1 groups (less so in the R1 group). Increased tumor size appeared to be less important in the R1 group compared with the R0 group.
The overall pooled estimate of the HR indicated no significant difference in the risk of death with chemoradiotherapy (HR, 1.09;
95% CI, 0.89-1.32; stratified P = .43) with estimated median survivals of 15.8 months with chemoradiotherapy (95% CI, 13.9-18.1) and 15.2 months without (95% CI, 13.1-18.2).15 Two- and 5-year survival rates, respectively, were estimated at 30% and 12% with chemoradiotherapy and 34% and 17% without.15
The 188 R0 patients allocated to receive chemoradiotherapy had a median survival of 15.9 months (95% CI, 14.0-18.5) compared with 15.8 months in the 183 patients not allocated to chemoradiotherapy (95% CI, 13.4-20.1) (Figure 2). Two- and 5-year survival rates, respectively, were estimated at 30% and 10% with chemoradiotherapy and 38% and 20% without in R0 patients (Table 3). In the R1 group, the 53 patients allocated to chemoradiotherapy had a median survival of 14.7 months (95% CI, 11.5-20.5) compared with 11.2 months for the 53 patients not randomized to chemoradiotherapy (95% CI, 9.4-16.7). Two- and 5-year survival rates, respectively, were estimated at 30% and 18% with chemoradiotherapy and 19% and 8% without in R1 patients (Table 3).
Overall, there was a nonsignificant increase in the risk of death estimated at 9% with chemoradiotherapy (95% CI, − 11% to 32%).15 There was significant heterogeneity for the difference in the effect of chemoradiotherapy dependent on resection margin status (χ² = 4.2, P = .04) in which chemoradiotherapy was estimated to be more effective in R1 patients: an estimated 28% reduction in the risk of death (HR, 0.72; 95% CI, 0.47-1.10) compared with an estimated 19% increased risk of death in R0 patients (HR, 1.19; 95% CI, 0.95-1.49).15
The overall pooled estimate of the HR indicated a 25% significant reduction in the risk of death with chemotherapy (HR, 0.75; 95% CI, 0.64-0.90; stratified P = .001), with median survival estimated at 19.0 months with chemotherapy (95% CI, 16.4-21.1) and 13.5 months without (95% CI, 12.2-15.8).15 Two- and 5-year survival rates, respectively, were estimated at 38% and 19% with chemotherapy and 28% and 12% without.15
In the R0 group, the 236 patients allocated to chemotherapy had a 7-month survival advantage over the 222 patients randomized to no chemotherapy: 20.8 months (95% CI, 17.7-23.2) compared with 13.8 months (95% CI, 12.2-16.4) (Figure 3). Two- and 5-year survival rates, respectively, were estimated at 42% and 22% with chemotherapy and 27% and 10% without in R0 patients (Table 3). The effect was less apparent in the smaller subgroup of R1 patients: the median survival of the 109 patients allocated to chemotherapy was 15.0 months (95% CI, 11.7-18.1) compared with 13.2 months in the 114 patients randomized to no chemotherapy (95% CI, 10.5-17.6). Two- and 5-year survival rates in R1 patients, respectively, were estimated at 29% and 14% with chemotherapy and 31% and 17% without (Table 3).
Overall, there was a reduction in the risk of death, estimated at 25%, in favor of chemotherapy (95% CI, 10%-36%).15 The beneficial effect of chemotherapy was apparent in the R0 subgroups but was not confirmed in the R1 subgroup. There was significant heterogeneity for the difference in the effect of chemotherapy dependent on resection margin status (χ2 = 7.3, P = .007) in which chemotherapy was estimated to be less effective in R1 patients: a 4% increased risk of death (HR, 1.04; 95% CI, 0.78-1.40) compared with 35% reduction in the risk of death in R0 patients (HR, 0.65; 95% CI, 0.53-0.80).15 The overall median survival according to randomized treatment and resection margins status is reported in Table 4 and shown in Figure 4 and Figure 5.
Ductal adenocarcinoma of the pancreas is considered a systemic disease at time of diagnosis with almost identical incidence and mortality rates. Surgical resection might be beneficial in prolonging survival; only in very few patients will the disease be cured radically. In a large population of well-selected, controlled patients with resections, the ESPAC-1 trial found that 18% of patients had R1 resections.9 R1 patients were reported to have only a marginally worse survival (P = .10), despite reports of a statistically strong effect on overall survival.5- 7,11 Furthermore, overall almost half of the resected patients, irrespective of adjuvant treatment administered, experienced disease relapse only 12 months after operation, grossly more than the rate of R1 resections. In this regard, it is evident that the ductal pancreatic cancer must be considered a systemic disease even in those few cases of apparent R0 resections. This finding is confirmed by the results of the long-term follow-up trial by Takada et al17 in 2002. The 5-year disease-free survival of the 2 R0 subgroups of patients randomized (treated vs observed) was almost the same (13.3% vs 12.8%). The importance of this single tumor parameter and the influence on postoperative therapy is debatable, but for the same reason, it is mandatory to investigate the effect of adjuvant treatment, specifically in R1 patients. All the randomized controlled trials of adjuvant treatment for patients with resected pancreatic cancer were included in this study.8,9,13- 17 A total of 278 R1 patients were identified, 106 of whom were randomized for chemoradiotherapy (53 to receive and 53 not to receive chemoradiotherapy);
223 were randomized for chemotherapy treatment (109 to receive and 114 not to receive chemotherapy). The ESPAC-1 trial used a 2 × 2 design, so that there was a group of R1 patients who were randomized for both chemoradiotherapy and chemotherapy treatments. Furthermore, 32% of all randomized patients within the meta-analysis had R1 resections influenced greatly by the Japanese trial17 in which 83% of the pancreatic ductal patients had resection margin positivity. With an overall median follow-up of 44 months in the surviving 177 patients in the meta-analysis, the mortality rates in R0 and R1 subgroups, respectively, were almost the same (79% vs 81%). The difference between the subgroups was small after 2 years (33% vs 29%) and has almost vanished after 5 years (16% vs 15%). Tumor grade and size and nodal involvement were statistically significant prognostic factors of survival. The importance of resection margins positivity on survival was less evident in the meta-analysis (HR, 1.10; 95% CI, 0.94-1.29; log-rank χ2 = 1.4; P = .24) compared with previous studies.5- 7,11 In particular, 2- and 5-year survival rates were almost equal in the 2 subgroups and the survival curves question whether the problem is related to having pancreatic cancer instead of receiving any adjuvant treatment. Moreover, this meta-analysis confirmed that chemoradiotherapy did not appear to improve overall survival for resected patients; R0 patients, in particular, did not appear to benefit from chemoradiotherapy. The meta-analysis revealed a possible beneficial effect for chemoradiotherapy within the R1 subgroup, with an estimated 28% reduction in the risk of death (HR, 0.72; 95% CI, 0.47-1.10) compared with an estimated 19% increased risk of death in R0 patients (HR, 1.19; 95% CI, 0.95-1.49).15 However, the R1 subgroup was underpowered, with only 106 patients included in this comparison. The meta-analysis demonstrated that chemotherapy did not appear to be as effective in the R1 subgroup, mimicking the effect of chemoradiotherapy in the R0 subgroup. Additional controlled trials are needed to find better treatments in patients with R1 resections for ductal adenocarcinoma of the pancreas.
Future prospective studies should work to precisely define positive resection margins and to better understand the relevance of different resection margins. Adjuvant trials in pancreatic cancer will be guided in part by the results of ongoing trials and, where available, how survival relates to margin involvement as well as the outcome of studies of new agents in advanced pancreatic cancer. The German Lilly phase III trial20 compared adjuvant gemcitabine treatment with observation in 368 pancreatic cancer patients, with disease-free survival as the primary end point. Preliminary results indicated an advantage for gemcitabine treatment,20 but the results are unlikely to demonstrate superiority over fluorouracil and folinic acid, as demonstrated in ESPAC-1, given an overall median survival of 21.6 months and a 5-year survival of 29%.9 The ESPAC-3 trial, an adjuvant trial of nearly 1000 patients, directly compares gemcitabine treatment with fluorouracil and folinic acid (the trial was close to achieving target accrual at the time of manuscript preparation).
The Radiation Therapy Oncology Group recently reported on the phase III adjuvant trial comparing gemcitabine treatment with fluorouracil-based chemoradiotherapy (with fluorouracil radiosensitization in both arms) in 442 analyzable patients of 538 randomized patients.21 There was no overall difference in survival; however, in further subgroup analysis among 330 patients with tumors in the head of the pancreas, there was better survival in the gemcitabine-based chemoradiotherapy group.21 With a median survival of only 18.8 months,21 however, this result is clearly inferior to the median survival achieved with chemotherapy alone in the ESAPC-1 trial.9 There is further disappointment for chemoradiotherapy in locally advanced pancreatic cancer from a recent phase III French trial,22 which had to be abandoned, because after 119 patients had been randomized, the survival was significantly worse for patients who had had chemoradiotherapy compared with gemcitabine alone.
Two agents, however, have shown great promise when combined with gemcitabine in locally advanced and metastatic pancreatic cancer, resulting in significantly improved survival compared with gemcitabine alone.23,24 Of these, capecitabine,23 an orally active fluoropyrimidine, is particularly attractive in the adjuvant setting because it has a higher tumor response rate compared with erlotinib,24 an orally active tyrosine kinase inhibitor, and has a manageable toxicity. Thus, it is envisaged that the combination of gemcitabine and capecitabine must feature as the next major arm for a large adjuvant trial. With such a recent increase in patients involved in such trials, future meta-analyses should be able to further dissect the specific roles of adjuvant chemoradiotherapy and chemotherapy for patients with R1 tumors in a manner that is, at present, not feasible.
Correspondence: John P. Neoptolemos, MD, Division of Surgery and Oncology, School of Cancer Studies, Royal Liverpool University Hospital, 5th Floor, UCD Bldg, Daulby St, Liverpool L69 3GA, England (firstname.lastname@example.org).
Accepted for Publication: May 9, 2006.
Author Contributions: The principal investigator (Dr Neoptolemos) had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Neoptolemos, Büchler, Dervenis, and Bassi are the lead ESPAC investigators who, along with Ms Stocken (biostatistician), were responsible for initiating the meta-analysis project and developing the protocol. In addition to the data from the ESPAC-1 trial, individual patient data were provided by the lead investigators of the European Organisation for Research and Treatment of Cancer trial (Drs Jeekel and Klinkenbijl), the Norwegian Pancreatic Trials Group (Dr Bakkevold), and the Japanese Pancreatic Group (Drs Takada and Amano). Study concept and design: Butturini, Stocken, Wente, Jeekel, Takada, Dervenis, Bassi, Büchler, and Neoptolemos. Acquisition of data: Butturini, Stocken, Jeekel, Klinkenbijl, Bakkevold, Takada, Dervenis, Bassi, and Neoptolemos. Analysis and interpretation of data: Butturini, Stocken, Wente, Jeekel, Klinkenbijl, Bakkevold, Takada, Amano, Bassi, and Neoptolemos. Drafting of the manuscript: Butturini, Stocken, Wente, Jeekel, Bakkevold, Takada, Amano, Bassi, and Neoptolemos. Critical revision of the manuscript for important intellectual content: Butturini, Stocken, Jeekel, Klinkenbijl, Bakkevold, Dervenis, Büchler, and Neoptolemos. Statistical analysis: Stocken, Wente, Takada, and Amano. Obtained funding: Stocken and Neoptolemos. Administrative, technical, and material support: Stocken, Jeekel, Bakkevold, and Neoptolemos. Study supervision: Butturini, Jeekel, Klinkenbijl, Bakkevold, Bassi, Büchler, and Neoptolemos.
Financial Disclosure: None reported.
Funding/Support: This study was supported by Cancer Research UK, Fondazione Cariverona 2004, and the Ministero dell' Istruzione e della Ricerca (Cofin No. 2005060715), Rome, Italy.
Additional Contributions: We thank all of the patients who took part in these trials and contributed to this research.
Group Information: A list of the Pancreatic Cancer Meta-analysis Group was published in Br J Cancer. 2005;92(8):1372-1381.