Adjuvant Chemotherapy With Fluorouracil Plus Folinic Acid vs Gemcitabine Following Pancreatic Cancer Resection: A Randomized Controlled Trial

Context Adjuvant fluorouracil has been shown to be of benefit for patients with resected pancreatic cancer. Gemcitabine is known to be the most effective agent in advanced disease as well as an effective agent in patients with resected pancreatic cancer.

Objective To determine whether fluorouracil or gemcitabine is superior in terms of overall survival as adjuvant treatment following resection of pancreatic cancer.

Design, Setting, and Patients The European Study Group for Pancreatic Cancer (ESPAC)-3 trial, an open-label, phase 3, randomized controlled trial conducted in 159 pancreatic cancer centers in Europe, Australasia, Japan, and Canada. Included in ESPAC-3 version 2 were 1088 patients with pancreatic ductal adenocarcinoma who had undergone cancer resection; patients were randomized between July 2000 and January 2007 and underwent at least 2 years of follow-up.

Interventions Patients received either fluorouracil plus folinic acid (folinic acid, 20 mg/m², intravenous bolus injection, followed by fluorouracil, 425 mg/m² intravenous bolus injection given 1-5 days every 28 days) (n = 551) or gemcitabine (1000 mg/m² intravenous infusion once a week for 3 of every 4 weeks) (n = 537) for 6 months.

Main Outcome Measures Primary outcome measure was overall survival; secondary measures were toxicity, progression-free survival, and quality of life.

Results Final analysis was carried out on an intention-to-treat basis after a median of 34.2 (interquartile range, 27.1-43.4) months' follow-up after 753 deaths (69%). Median survival was 23.0 (95% confidence interval [CI], 21.1-25.0) months for patients treated with fluorouracil plus folinic acid and 23.6 (95% CI, 21.4-26.4) months for those treated with gemcitabine (χ²₁ = 0.7; P = .39; hazard ratio, 0.94 [95% CI, 0.81-1.08]). Seventy-seven patients (14%) receiving fluorouracil plus folinic acid had 97 treatment-related serious adverse events, compared with 40 patients (7.5%) receiving gemcitabine, who had 52 events (P < .001). There were no significant differences in either progression-free survival or global quality-of-life scores between the treatment groups.

Conclusion Compared with the use of fluorouracil plus folinic acid, gemcitabine did not result in improved overall survival in patients with completely resected pancreatic cancer.

Trial Registration clinicaltrials.gov Identifier: NCT00058201

Pancreatic cancer is one of the major causes of cancer death globally, with a 5-year survival rate of less than 5%.^1,2 The outlook for those patients who can undergo surgical resection is better, and in specialized centers, resection rates greater than 15% can be achieved.³ Although surgery cannot guarantee a cure, the 5-year survival does improve to around 10% following resection.³ There is a clear need to improve long-term survival in these patients. While the added survival benefit of adjuvant chemoradiotherapy with or without maintenance chemotherapy^4-7 remains unclear,⁸ a more certain survival benefit has been demonstrated from adjuvant chemotherapy.^6,9-14

The European Study Group for Pancreatic Cancer (ESPAC)-3 trial was designed to compare the survival benefit of adjuvant fluorouracil plus folinic acid vs gemcitabine, which during the conduct of the ESPAC-1 trial had become established as the standard care for advanced pancreatic cancer.¹⁵ Initially this was a 3-group study that included an observation group based on the survival uncertainty of adjuvant chemotherapy⁶; however, the observation group was removed from the design following the definitive results of ESPAC-1.¹² In 2007, the Charité Onkologie Clinical Studies in GI Cancer (CONKO)-001 trial reported improved disease-free survival in patients randomized to receive adjuvant gemcitabine compared with those randomized to receive surgery alone.¹³ With 1088 patients randomized, the ESPAC-3 trial represents the largest-ever adjuvant trial conducted in pancreatic cancer, to our knowledge, and results are presented herein.

The ESPAC-3 trial was initially introduced as a 3-group study designed to compare the survival benefit of resection alone (observation) with either adjuvant fluorouracil plus folinic acid or gemcitabine. The first patient was entered on July 7, 2000. Following the definitive results from ESPAC-1,¹² the recommendation of the independent data and safety monitoring committee to cease randomization into the control group was adopted on June 20, 2003. The trial design of ESPAC-3 (version 2) therefore necessitated removal of the control group from the original ESPAC-3 (version 1) trial design. ESPAC-3 (version 2) is thus a 2-group, international, open-label, phase 3, randomized controlled study of adjuvant chemotherapies comparing fluorouracil plus folinic acid with gemcitabine.

The trial was approved by ethics committees at the national and local level according to the requirements of each participating country. All patients entered into the study provided written informed consent following a full explanation of the study and reading of the patient information sheet. There were 159 centers in 17 countries: Australia and New Zealand (26), Canada (15), Czech Republic (1), Finland (1), France (15), Germany (13), Greece (3), Hungary (2), Ireland (2), Italy (3), Japan (7), Poland (1), Serbia (1), Sweden (8), Switzerland (1), and the United Kingdom (60).

Patients were eligible if they had undergone complete macroscopic (R0 or R1) resection for ductal adenocarcinoma of the pancreas with histological confirmation and with no evidence of malignant ascites, peritoneal metastasis, or spread to the liver or other distant abdominal or extra-abdominal organs. The type and extent of resection was determined using an established international classification.¹⁶ Patients had to be fully recovered from the operation, with a World Health Organization performance score of 2 or lower and a life expectancy of more than 3 months. Patients with previous use of neoadjuvant chemotherapy or other concomitant chemotherapy and with pancreatic lymphoma, macroscopically remaining tumor (R2 resection), or TNM stage IVb disease were excluded.

Patients were randomly assigned to each treatment group on a 1:1 basis according to a computer-generated variable-size blocked randomization method. Patients were stratified at randomization by country and resection margin status (R0 vs R1).

Folinic acid (20 mg/m²) was given as an intravenous bolus followed by intravenous bolus fluorouracil (425 mg/m²) given on 5 consecutive days every 28 days for 6 cycles (24 weeks). Gemcitabine (lyophilized powder diluted in normal saline) was given as an intravenous infusion over 30 minutes (1000 mg/m²), administered once a week for 3 out of every 4 weeks (1 cycle) for 6 cycles (24 weeks). Toxicity was assessed using the National Cancer Institute Common Toxicity Criteria for Adverse Events (version 2), with a clearly defined protocol for modifications and delays.

Quality of life was assessed using the European Organization for Research and Treatment of Cancer (EORTC) QLQ-C30 (version 3) and ESPAC-32 patient questionnaires at baseline and at 3 and 6 months and yearly until 5 years.¹⁷

The trial was designed to test the primary hypothesis, ie, that overall length of survival does not differ between that achieved with adjuvant fluorouracil plus folinic acid and that achieved with gemcitabine. Secondary end points were progression-free survival, toxicity, and quality of life. Power calculations were based on expected 2-year survival rates. The ESPAC-1 trial had shown that 2-year survival with fluorouracil plus folinic acid was in the order of 40% to 45%.^6,12 ESPAC-3 was powered to detect a clinically meaningful increase in survival of 10% with gemcitabine. Recruiting 515 patients (275 deaths) in each treatment group would allow 10% differences in 2-year survival to be detected using a 2-sided α = .05 level of significance with at least 90% power.

Overall survival was measured from the date of resection to date of death from any cause. Patients remaining alive were censored at the date last seen alive. Progression-free survival was measured from date of resection to date of death from any cause or date of local tumor recurrence or metastases. Patients remaining alive and progression-free were censored at the date last seen alive. Survival estimates were calculated using the Kaplan-Meier method¹⁸ and compared using the unweighted Mantel-Haenszel version of the log-rank test.¹⁹ Median, 12-month, and 24-month survival estimates are presented with 95% confidence intervals (CIs).

The hazard ratio (HR) of the treatment effect is presented for gemcitabine compared with that for fluorouracil plus folinic acid. Hazard ratios of the treatment effect within stratification subgroups at randomization are estimated (without significance testing) with tests of heterogeneity to determine if treatment effects differ across subgroups. The treatment effect was adjusted by stratification factors at randomization (country and resection margin status) and other identified prognostic factors in the multivariate setting using Cox proportional hazards modeling²⁰ incorporating a random effect into the hazard function for country effect. Factors with a log-rank significance of P < .10 were explored further in the multivariate setting using backward selection techniques. Classification variables were used for ordinal variables with more than 2 categories. The functional form of the relationship between continuous factors and log-hazard (specifically age, tumor size, and postoperative carbohydrate antigen 19-9 [CA19-9] level) was assessed, and factors were included in the multivariate models with a nonlinear transformation if appropriate.²¹ The assumption of proportional hazards was assessed and confirmed by including a time-dependent covariate.

The number of patients receiving treatment and the percentage of protocol dose of chemotherapy and the range of total doses received was calculated. The number of patients experiencing at least 1 high-grade toxic episode (grade 3/4) of each toxicity type or serious adverse event is reported as a percentage of the total number of patients randomized within each treatment group. Proportions were compared using the Fisher exact test with the significance level set at P < .005 and with Bonferroni adjustment to account for multiple testing.

Quality-of-life domain scores were calculated according to the EORTC QLQ-C30 scoring manual and linearly transformed to produce a standardized score ranging from 0 to 100. Higher scores for the functional and global health scales indicated better quality of life, whereas higher scores for the symptom scales and items indicated poorer quality of life. Standardized area under the curve (AUC) scores¹⁷ are average observed symptomatic and functional quality-of-life scores per month within a 12-month duration from surgery, calculated from the linearly transformed scores and compared across treatments using the Mann-Whitney nonparametric test.

All statistical analyses were carried out using SAS version 9.1 (SAS Institute Inc, Cary, North Carolina) and R version 2.7.2 (R Project for Statistical Computing; http://www.r-project.org) on an intention-to-treat basis, retaining patients in their randomized treatment groups and including protocol violators and ineligible patients. A 2-sided significance level of P < .05 was used throughout.

The last of the 1088 patients recruited was randomized on January 8, 2007. The database was locked on March 18, 2009.

Five hundred fifty-one patients were randomized to receive fluorouracil plus folinic acid, and 537 were randomized to receive gemcitabine (Figure 1). Four ineligible patients were reported (2 in each group) and have been included in the analysis on an intention-to-treat basis. The clinical characteristics of patients and surgical and pathological details are shown in Table 1.

Four hundred eighty-six patients (88%) received 2326 cycles of fluorouracil plus folinic acid and 478 (89%) received 2464 cycles of gemcitabine. Sixty-five patients (12%) in the fluorouracil plus folinic acid group and 59 (11%) in the gemcitabine group did not start treatment. Three hundred one patients (55%) in the fluorouracil plus folinic acid group and 323 (60%) in the gemcitabine group received all 6 cycles of treatment. Median time from randomization to the start of chemotherapy was 10 (interquartile range [IQR], 5-18) days for the fluorouracil plus folinic acid group and 8 (IQR, 5-14) days for the gemcitabine group. Median time receiving chemotherapy was 4.7 (IQR, 3.1-5.0) months for the fluorouracil plus folinic acid group and 5.1 (IQR, 4.0-5.3) months for the gemcitabine group. Median dose intensity was 79% (range, 3%-141%) of the planned protocol for the fluorouracil plus folinic acid group and 89% (range, 6%-122%) for the gemcitabine group.

Seven hundred fifty-three patients (69%) had died at the time of analysis (388 [70%] in the fluorouracil plus folinic acid group and 365 [68%] in the gemcitabine group). Median length of follow-up of 335 living patients was 34.2 (IQR, 27.1-43.4; range, 0.4-86.3) months, equal across treatment groups. Overall, 282 of patients remaining alive (84%) had undergone follow-up for more than 2 years. Median survival was estimated as 23.2 months (95% CI, 21.7-24.9), with 12-month and 24-month rates estimated as 79.3% (95% CI, 76.9%-81.8%) and 48.6% (95% CI, 45.6%-51.6%), respectively. Median survival for patients treated with fluorouracil plus folinic acid was 23.0 (95% CI, 21.1-25.0) months and for patients treated with gemcitabine was 23.6 (95% CI, 21.4-26.4) months (Figure 2).

Survival estimates at 12 and 24 months were 78.5% (95% CI, 75.0%-82.0%) and 48.1% (95% CI, 43.8%-52.4%), respectively, for the fluorouracil plus folinic acid group and 80.1% (95% CI, 76.7%-83.6%) and 49.1% (95% CI, 44.8%-53.4%) for the gemcitabine group. Log-rank analysis revealed no statistically significant difference in survival estimates between the treatment groups (χ²₁ = 0.7; P = .39; HR, 0.94 [95% CI, 0.81-1.08]).

Six hundred eighty-eight patients (63%) developed local recurrence, metastases, or both; of these, 597 had died. Two hundred forty-four patients (22%) were alive and progression free. Progression-free survival analysis was based on all patients, of whom 844 (78%) had either progressive disease or died. The median progression-free survival was 14.3 (95% CI, 13.5-15.1) months, with 12-month and 24-month rates of 58.7% (95% CI, 55.7%-61.6%) and 30.1% (95% CI, 27.3%-32.9%), respectively. The median progression-free survival for patients treated with fluorouracil plus folinic acid was 14.1 (95% CI, 12.5-15.3) months and 14.3 (95% CI, 13.5-15.6) months for patients treated with gemcitabine (Figure 2).

Survival estimates at 12 and 24 months were 56.1% (95% CI, 51.8%-60.3%) and 30.7% (95% CI, 26.7%-34.6%), respectively, for the fluorouracil plus folinic acid group and 61.3% (95% CI, 57.1%-65.5%) and 29.6% (95% CI, 25.6%-33.5%) for the gemcitabine group. Log-rank analysis revealed no statistically significant difference in progression-free survival estimates between the treatment groups (χ²₁ = 0.40; P = .53; HR, 0.96 [95% CI, 0.84-1.10]).

Patients receiving fluorouracil plus folinic acid had significantly increased grade 3/4 stomatitis (P < .001) and diarrhea (P < .001), whereas patients receiving gemcitabine reported significantly increased grade 3/4 hematologic toxicity (P = .003) (Table 2). One hundred seventeen patients (11%) reported 149 treatment-related serious adverse events, the majority attributable to inpatient hospitalization. Seventy-seven patients (14%) receiving fluorouracil plus folinic acid reported 97 treatment-related serious adverse events, compared with 40 (7.5%) receiving gemcitabine, who reported 52 events (P < .001).

Univariate survival analysis of categorical variables revealed that not smoking, World Health Organization performance status 0, negative resection margins, negative lymph node status, well-differentiated tumors, stage I disease, and tumors with no local invasion were associated with improved survival (Table 3 and eFigure 1 and eFigure 2). The increased risk of death in patients with positive margins compared with patients with negative margins was 35% (log-rank χ²₁ = 16.3; P < .001; HR, 1.35 [95% CI, 1.17-1.56]). There was no significant difference in the effect of treatment across subgroups according to R status (test of heterogeneity, χ²₁ = 0.3, P = .56). The continuous covariates of tumor diameter (Wald χ²₁ = 10.1, P = .001) and postoperative CA19-9 level (Wald χ²₂ = 126.6, P < .001) were also each significantly associated with survival at univariate analysis but not age (Wald χ²₁ = 0.7, P = .40).

Factors with a log-rank significance of P < .10 were considered for inclusion in the Cox proportional hazards frailty modeling: sex, smoking, performance status, grade of disease, lymph node status, stage (I/II vs III/IV), and local invasion. The continuous covariates tumor size and postoperative CA19-9 level were included under nonlinear transformations. Stratification factors (country [random effect] and resection margin status) and treatment group were included in all models.

A model based on 766 patients with complete data (545 deaths) identified grade of disease (Wald χ²₃ = 28.8, P < .001), nodal status (Wald χ²₁ = 19.1, P < .001), and CA19-9 level (Wald χ²₂ = 110.4, P < .001) as significant independent prognostic factors of overall survival (Table 4). To maximize the data for modeling, further analysis excluding CA19-9 level, which was associated with a substantial amount of missing data (321 patients), resulted in a model based on 1030 patients with complete data (715 deaths). This confirmed grade of disease (Wald χ²₃ = 25.2, P < .001), nodal status (Wald χ²₁ = 41.7, P < .001), performance status (Wald χ²₂ = 10.9, P = .004), tumor size (Wald χ²₁ = 8.9, P = .003), and smoking status (Wald χ²₃ = 9.2, P = .03) as significant independent prognostic factors of overall survival.

Tests of heterogeneity within pathological (eFigure 3) or demographic (eFigure 4) subgroups did not reveal any significant findings.

Five hundred sixty-five patients (280 randomized to receive fluorouracil plus folinic acid and 285 to receive gemcitabine) completed quality-of-life questionnaires, including a baseline questionnaire. The subgroups were representative of patients in the main study based on patient characteristics. Of these, 438 completed 3-month questionnaires, 417 completed 6-month questionnaires, and 307 completed 12-month questionnaires. Standardized AUC scores are based on average standardized scores ranging between 0 and 100. There were no significant differences in mean standardized AUC for global quality-of-life scores across treatment groups conditional on patient survival; mean standardized AUC was 43.6 (SD, 20.1) for patients receiving fluorouracil plus folinic acid, compared with 46.6 (SD, 19.7) for those receiving gemcitabine (P = .08).

There have been few large randomized controlled trials of adjuvant treatment following resection in pancreatic cancer. The first of these, the ESPAC-1 trial,^6,12 concluded that chemotherapy with fluorouracil plus folinic acid improved overall survival but chemoradiotherapy did not.^6,12 The failure of adjuvant chemoradiotherapy to enhance survival was also reflected in the results of the EORTC multicenter prospective randomized trial.⁵ The Radiation Therapy Oncology Group (RTOG) 9704 trial randomized 538 patients to receive either prechemoradiation and postchemoradiation gemcitabine or prechemoradiation and postchemoradiation fluorouracil.⁷ The median survival in the 451 eligible patients was 16.7 and 18.8 months, respectively (P = .34), and in the 388 patients with cancer of the pancreatic head was 20.5 months vs 16.9 months, respectively (P = .09).⁷ The primary end point in the CONKO-001 trial was disease-free survival.¹³ This was 13.4 months for gemcitabine and 6.9 months for surgery alone (P < .001), while the median overall survival was 22.1 months and 20.5 months, respectively (P < .06).¹³

The ESPAC-3 trial found a median survival of 23.0 months for patients treated with fluorouracil plus folinic acid and 23.6 months for those treated with gemcitabine and a median progression-free survival of 14.1 months and 14.3 months, respectively. Tumor grade, nodal status, tumor size, postoperative serum CA19-9 levels, performance status, and smoking were all independent prognostic factors of overall survival. Although resection margin status was significant on univariate analysis, this was not so on multivariate analysis, confirming the previous results of ESPAC-1 that primary tumor characteristics dominate outcome.²³

The prognostic significance of CA19-9 level in ESPAC-1 mirrored that in the RTOG trial, with both studies using postresectional values.²⁴ This is important: preoperative levels are artificially elevated in the presence of obstructive jaundice, because CA19-9 is excreted in bile and there is no simple correction factor. In the CONKO-001 trial, patients with CA19-9 levels greater than 2.5 times the upper limit of normal were excluded, indicating that in that study there was a bias toward patients with a more favorable prognosis.¹³ That tobacco smoking affected long-term outcome was a novel finding and should add further weight against the use of tobacco.

The absence of an overall survival difference between postoperative adjuvant fluorouracil plus folinic acid compared with gemcitabine contrasts with the findings of a much smaller study in patients with nonresected advanced pancreatic cancer that showed a survival benefit with gemcitabine as compared with fluorouracil.¹⁵ The fluorouracil regimen used in that trial (600 mg/m² bolus once weekly without folinic acid) was less intensive than that used in ESPAC-3.¹⁵ This fluorouracil regimen may be less efficacious than the Mayo Clinic regimen, but there are no large randomized trials that have directly compared these 2 treatments in pancreatic cancer.

In conclusion, gemcitabine did not result in improved overall survival compared with fluorouracil plus folinic acid in patients with resected pancreatic cancer. As a logical progression from these data we have designed the ESPAC-4 trial, currently in progress, to compare combination chemotherapy with gemcitabine plus capecitabine, an orally active fluoropyrimidine,²⁵ with gemcitabine alone.

Corresponding Author: John P. Neoptolemos, MD, Liverpool Cancer Research UK Cancer Trials Unit, Cancer Research UK Centre, University of Liverpool, Fifth Floor, UCD Bldg, Daulby Street, Liverpool, L69 3GA, United Kingdom (j.p.neoptolemos@liverpool.ac.uk).

Author Contributions: Dr Neoptolemos had full access to all the data in the study and takes full responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Neoptolemos, Stocken, Bassi, Ghaneh, Cunningham, Moore, Friess, Dervenis, Büchler.

Acquisition of data: Neoptolemos, Bassi, Cunningham, Goldstein, Padbury, Gallinger, Mariette, Wente, Izbicki, Lerch, Dervenis, Oláh, Butturini, Doi, Lind, Smith, Palmer, Buckels, Thompson, McKay, Rawcliffe, Büchler.

Analysis and interpretation of data: Neoptolemos, Stocken, Bassi, Ghaneh, Goldstein, Moore, Izbicki, Dervenis, Doi, Lind, Valle, Büchler.

Drafting of the manuscript: Neoptolemos, Stocken, Ghaneh, Goldstein, Wente, Dervenis, Oláh, Buckels, Rawcliffe, Büchler.

Critical revision of the manuscript for important intellectual content: Neoptolemos, Stocken, Bassi, Cunningham, Goldstein, Padbury, Moore, Gallinger, Mariette, Wente, Izbicki, Friess, Lerch, Dervenis, Butturini, Doi, Lind, Smith, Valle, Palmer, Thompson, McKay, Büchler.

Statistical analysis: Stocken, Cunningham.

Obtained funding: Neoptolemos, Stocken, Ghaneh, Cunningham, Moore, Gallinger, Lerch, Büchler.

Administrative, technical, or material support: Neoptolemos, Ghaneh, Goldstein, Moore, Izbicki, Lerch, Dervenis, Smith, Valle, Buckels, McKay, Rawcliffe.

Study supervision: Neoptolemos, Bassi, Ghaneh, Padbury, Izbicki, Friess, Dervenis, Oláh, Butturini, Smith, Palmer, Thompson, Büchler.

Ms Rawcliffe was the trial coordinator responsible for central administration ensuring ethical standards for collection and verification of data. The results were interpreted by the ESPAC working party (all of the above). Drs Neoptolemos, Ghaneh, and Stocken prepared the initial draft and were responsible for collating changes proposed by the aforementioned into the final paper before final approval by all participants in the European Study Group for Pancreatic Cancer.

Funding/Support: This study was supported by Cancer Research UK; National Cancer Institute of Canada, Canadian Cancer Society; Fonds de Recherche de la Société Nationale Française de Gastroentérologie; Fondazioone Italiana Malattie del Pancreas; Health and Medical Research Council of Australia, Cancer Councils of New South Wales, Queensland, Victoria, and South Australia. Dr Cunningham's work is funded in part by the National Institute for Health Research Biomedical Research Centre at the Royal Marsden Hospital.

Role of the Sponsors: None of the sponsors or funders had any role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.

The Full List of ESPAC Specialists Who Contributed to the Treatment of Patients in the ESPAC-3 Trial is presented in the eAppendix.

Independent Data and Safety Monitoring Committee: R. P. Ahern, MSc (Institute for Cancer Research, London, United Kingdom), R. C. G. Russell, MD (Middlesex Hospital, London, United Kingdom), P. Clarke, MD (Clatterbridge Centre for Clinical Oncology, Wirral, United Kingdom).

This article was corrected online for typographical errors on 9/7/2010.