Adjuvant Therapy for Pancreatic Cancer: Title and subTitle BreakOne Small Step Forward

The great tragedy facing the majority of patients with newly diagnosed adenocarcinoma of the pancreas is the persistent high rate of lethality:most newly diagnosed individuals will die within a year.¹ Approximately 20% of patients are considered for surgical therapy; however, only about half of these individuals undergo successful resections.^{2 - 4} Surgery remains the only opportunity for cure and can be performed with significant reduction in rates of operative morbidity and mortality, particularly at experienced high-volume centers.⁵ Use of neoadjuvant strategies in the preoperative setting to improve surgical resectability remain experimental. Adjuvant therapy for the postoperative pancreatic cancer patient has been a highly controversial topic with some strong differences in opinions about the use of adjuvant radiation, chemotherapy, or combined chemoradiation, particularly between investigators in the United States and Europe.^{6 - 7}

In this issue of JAMA, Oettle et al present the results of CONKO-001, a well-designed randomized clinical trial, one of the largest of its kind, to determine the benefits of gemcitabine for patients with resected pancreatic cancer.⁸ This trial is of considerable importance because of the significant numbers of patients, the appropriateness of the chosen agent, and the validity of the stratification factors. The trial met its primary end point showing a statistically significant advantage in estimated median disease-free survival for those individuals who received postoperative gemcitabine compared with observation (13.4 months vs 6.9 months, respectively; P<.001). Furthermore, gemcitabine appeared to confer benefit at years 1, 2, 3, and 5 as evaluated by estimated disease-free survival rates (eg, year 5, 23.5% vs 5.5%, respectively).

Although there was no significant difference in median overall survival (22.1 months for gemcitabine vs 20.2 months for observation, by intent-to-treat analysis [ITT]), estimates of overall survival rates suggest that differences may emerge at years 3 (34% vs 20.5%, respectively) and 5 (22.5% vs 11.5%, respectively). Subgroup analyses must be interpreted with caution given the relatively small numbers of patients. There are credible results, however, favoring patients who have had an R0 resection (histologically tumor-free surgical margins) vs an R1 resection (only approximately 17% of the study population), and for those with smaller tumors and those with negative lymph nodes. In the ITT population, the beneficial effects of gemcitabine on disease-free survival were seen across subsets, including those with R0 and R1 resections, large and small primary tumors, and presence or absence of positive lymph nodes, suggesting that even subgroups of patients with a poor prognosis experienced a treatment advantage. Quality of life was similar for both groups. The authors clearly describe some of the hazards of their post hoc subgroup analyses; however, their multicenter patient population should be viewed as a “real life” population of patients. The report of significant improvement in disease-free survival across the entire population of patients and subgroups is valid. The lack of overall survival advantage is most likely a result of insufficient numbers of patients; however, the observation of survival differences as a trend favoring patients who received gemcitabine at years 3 and 5 appears important and is most likely related to the chemotherapy effect.

Other more recent randomized adjuvant pancreatic cancer trials provide some insight as to the significance of the results. The ESPAC-1 trial was a 2 × 2 factorial design of 289 patients evaluating observation, chemoradiation alone, chemotherapy alone (with fluorouracil), and both chemoradiotherapy and chemotherapy.⁹ Although this trial had several limitations, including limited statistical power and lack of separate analysis for each of the 4 groups in the 2 × 2 factorial design, the study did demonstrate an improved median survival favoring the 147 patients who received chemotherapy compared with the 142 patients who did not (20.1 months vs 15.5 months, P = .009). This is comparable to median overall survival in the ITT group as currently reported in the CONKO-001 trial (22.1 vs 20.2 months). Although disease-free survival is not specifically addressed in ESPAC-1, there was an estimated 12-month recurrence-free survival rate of 58% vs 43% (chemotherapy vs non-chemotherapy, respectively). This compares with an estimated 1-year disease-free survival rate of 58% vs 31%, favoring gemcitabine in the CONKO-001 trial.

The United States Gastrointestinal Intergroup recently reported the preliminary results of RTOG 9704, which included 451 analyzable patients assigned to either adjuvant gemcitabine plus infusional fluorouracil and radiation or infusional fluorouracil plus infusional fluorouracil and radiation.¹⁰ RTOG 9704 included slightly more lymph node–negative patients compared with the CONKO-001 trial and fewer than in ESPAC-1. Although the data are incomplete, there appear to be almost twice the number of patients with positive margins in RTOG 9704 compared with both CONKO-001 and ESPAC-1. More patients in the CONKO-001 trial had T3-4 lesions compared with RTOG 9704, although there was an imbalance in the T3-4 lesions between the 2 groups in the RTOG trial. The median overall survival rates in RTOG 9704 (pancreatic head patients: 20.5 months [gemcitabine] vs 16.9 months [fluorouracil]) were less than reported in CONKO-001 but similar to ESPAC-1. The 3-year overall survival was comparable for the gemcitabine groups in both CONKO-001 and RTOG 9704.

The highlights from the results of these 3 most recent randomized clinical trials and other smaller trials do not resolve the controversy as to the optimal adjuvant therapy strategy for pancreatic cancer patients using currently available chemotherapy agents with or without radiation. It is also difficult to make comparisons among trials, as suggested by the CONKO-001 investigators, because there are not only treatment variations but also differences among the various subsets of patients within the trials that can potentially influence survival statistics (ie, the proportion of R0 vs R1, N0 vs N+, and T1-2 vs T3-4). CONKO-001 and the preliminary results of RTOG 9704 do suggest, however, that a gemcitabine-containing platform rather than single-agent fluorouracil represents an appropriate choice for current patients with resected pancreatic cancer and a potential building block for future clinical trials. It is also appropriate to discuss the choice between gemcitabine alone and gemcitabine plus chemoradiation with the individual patient, a decision that should be driven in part by toxicity concerns and perhaps risk of local recurrence, particularly for those with an R1 resection. Given the high percentage of individuals who develop distant metastases after surgical resection, as noted for example in CONKO-001 and ESPAC-1, and the suggestion in CONKO-001 that even patients with R1 resections appear to benefit from gemcitabine, adjuvant gemcitabine alone should be included routinely in discussion with the individual patient. It is equally critical to inform each patient that disease-free survival is important as a measure of quality of life; however, the impact on overall survival is less certain regardless of the trends demonstrated in recent trials. Although it is argued that overall survival should be the optimal statistical end point for pancreatic cancer patients, improved disease-free survival can offer patients with a notoriously lethal disease an opportunity to enjoy longer periods of time relatively free of symptoms. It is possible, as suggested by the CONKO-001 authors, that with additional time, overall survival may become significant, particularly because it is estimated that at 5 years, nearly twice the number of patients in the gemcitabine group will be alive compared with those who received no adjuvant therapy. A cautionary note is the small number of patients. In colon cancer, with a far larger data set than available for pancreatic cancer, disease-free survival has been shown to correlate well with overall survival.¹¹

Clearly, oncologists are no closer to solving the mysteries that encompass this most deadly disease. Currently, there are no definitive leads from clinical trials of advanced pancreatic cancer to give guidance as to the forthcoming strategy for adjuvant therapy. Combination chemotherapy trials linked to gemcitabine, including cisplatin, irinotecan, oxaliplatin, and fluoropyrimidines, did not meet their statistical end points, with the exception of an initial report of a large randomized trial comparing gemcitabine with or without capecitabine.¹² Two recent meta-analyses (12 trials with 3687 patients, 16 trials with 5561 patients) comparing gemcitabine combination therapy regimens suggested that the platinum combinations resulted in significantly prolonged progression-free or overall survival and overall response, particularly for those patients with the best performance status.^{13 - 14} The interest in biologic or targeted therapies has generated recent clinical trials exploring biologic combinations with gemcitabine. Thus far, only one trial combining the epidermal growth factor receptor inhibitor erlotinib with gemcitabine has shown a small but statistically significant survival advantage compared with gemcitabine alone.¹⁵ Whether any of the limited select advantages in response, progression-free survival, or overall survival from combination therapy will translate to a survival benefit for pancreatic cancer patients undergoing resection is unknown. Furthermore, the paucity of biological data to link a tumor biological profile with intervention efficacy, for example, is a major obstacle confronting the design of future pancreatic cancer clinical trials, both for advanced disease and for those with resected tumors.

Empirical clinical research design has certainly resulted in the introduction of new agents to improve the outcome for cancer patients. Empiricism, however, often results in failure to detect statistical significance, leaving many patients with a poorly understood lack of benefit from an intervention. As new agents emerge, particularly the biologics, integration of research disciplines (eg, molecular biology, experimental imaging, and clinical trials) is imperative. Worldwide, there is a need to move away from the question of current chemotherapy agents vs chemoradiation to a research focus based on enhancing understanding of biologic principles. Given the rapid lethality of pancreatic cancer, this is no easy task. Whether it is possible to build on the small steps taken with CONKO-001 results remains to be seen. It is unlikely, however, that these small steps alone will provide the necessary enhancement of benefit beyond the improvements in surgery to profoundly alter the course of this most challenging of cancers.