|Home | About | Journals | Submit | Contact Us | Français|
Despite recent completion of several trials of adjuvant therapy after resection for pancreatic adenocarcinoma, the absolute impact on survival and the identification of appropriate patients for treatment remains controversial.
We sought to identify the impact of adjuvant therapy and factors associated with any improvement in survival after resection of pancreatic cancer.
Through the California Cancer Registry, we identified all California residents diagnosed with pancreatic cancer between 1994 and 2002. Factors potentially impacting survival were analyzed, including: patient demographics, tumor characteristics, and treatment provided. Univariate and multivariate survival analyses were performed by Kaplan-Meier and Cox regression methods.
26,518 patients were identified; 3,196 (12.1%) underwent resection as their primary treatment. The median overall survival was 16 months for resected patients. Prognostic factors associated with better survival included: negative lymph node status, well differentiated tumors, younger age, female gender, and receipt of any adjuvant therapy. On multivariate analysis, adjuvant therapy demonstrated a statistically significant, though modest, impact on survival with a hazard ratio of 0.79 (95% CI 0.72 – 0.87, p<0.001). The benefit of adjuvant therapy was only apparent in those patients with lymph node positive or poorly differentiated tumors.
Adjuvant therapy provides for a modest improvement in overall survival following surgical resection of pancreatic cancer. The absolute effect is most pronounced in those with poor prognostic indicators. In order to identify effective systemic therapy for this deadly cancer, future clinical trials of adjuvant therapy should focus on these groups of patients.
Pancreatic cancer is the fourth leading cause of cancer death in the United States with an overall 5 year survival of 5% for all patients.1,2 The only reasonable chance for long term survival is curative surgical resection, though this is appropriate for only a small minority of patients as most present with advanced disease. Even smaller still is the number of patients who receive adjuvant therapy after resection. The reasons for this are multifactorial, including: the occurrence of postoperative complications that limits timely receipt of therapy, the decline of performance status following surgery that precludes delivery of therapy, and the perception among both patients and oncologists of the limited benefit of adjuvant therapy. Despite the completion of several Phase III trials of adjuvant therapy in resected adenocarcinoma of the pancreas, the benefit of adjuvant treatment has remained controversial, and there is limited information on which subgroups of patients may benefit more or less from postoperative therapy.
The current standard adjuvant therapy remains controversial given the various regimens examined in the Phase III trials, as well as concerns about the conduct of the various trials.3 In general, however, there has been a consistent observation of some modest benefit following the delivery of adjuvant therapy. Each of these trials has typically included less than 150 patients per treatment arm, and thus lacks statistical power to identify specific subgroups of patients with pancreatic cancer that may, or may not, benefit. There are several population-based studies that have been recently reported that identify practice patterns and outcomes of therapy in pancreatic cancer.4-9 The benefit of large dataset analyses lies in the size of the databases to allow sufficient statistical power to examine questions that cannot be addressed in randomized trials. California has created a cancer-reporting system, which registers and collects treatment and follow-up data on all patients within the state diagnosed with cancer.10 From this, we have created a database of patients diagnosed and treated for pancreatic cancer, which is the largest and most diverse series to date.11
The goal of our study was to evaluate the impact of adjuvant treatment (any modality of chemotherapy, radiation, or both) on survival in a large, diverse population of patients who underwent curative-intent pancreatic resection. Furthermore, we hypothesized that specific subgroups could be identified that achieve more (or less) benefit from adjuvant therapy. This study represents the largest series to date of patients with resected pancreatic cancer, and therefore has the statistical power to further refine our understanding the use of postoperative adjuvant therapy.
We identified all patients diagnosed with cancer of the pancreas in the state of California between 1 January 1994 and 31 December 2002 through the California Cancer Registry (CCR). Follow-up data was available through 31 December 2003. The CCR is a population-based registry that has been collecting cancer incidence and mortality data for the entire population of California since 1988 through a system of eight regional registries; health care providers are required by state law to report all cancer cases to the registry. Registry data is extracted from patient medical records and collected in a prospective fashion. Inclusion criteria were: a diagnosis of pancreatic cancer (coded by International Classification of Diseases for Oncology) and receipt of curative-intent resection (pancreaticoduodenectomy, distal pancreatectomy, total pancreatectomy, partial pancreatectomy). We assumed that patients undergoing surgical resection would not have advanced disease (ie hepatic metastases, peritoneal metastases, locally advanced/unresectable disease) that were apparent preoperatively or intra-operatively, which constitute traditional and accepted contraindications to curative-intent surgical resection. This study was approved by the Institutional Review Board of UC Davis.
Exclusion criteria were: unknown age or sex, diagnosis made at the time of autopsy or receipt of surgical exploration without resection (e.g. palliative bypass). This dataset, in various iterations of inclusion and exclusion criteria, has been used to analyze factors related to outcomes of pancreatic cancer.11-12 Factors previously reported to impact survival include: age, tumor size, location of tumor, lymph node status, histologic grade, extent of resection, and receipt of adjuvant therapy.11-14 Patients were staged according to the AJCC 5th edition criteria, which was in effect throughout the study timeframe. Treatment information is limited to that planned or given within six months after diagnosis. Information on duration of treatment or completion of planned treatment is not available, nor is information on type of chemotherapeutic agent received, patient co-morbidities, postoperative complications, nor complete information on extent of resection/positive margins. Adjuvant therapy may be incompletely reported to population based registries, and thus potentially under-reported.15
Vital status and follow up is obtained through both passive and active means within the registry, and both cancer-related and all-cause death are recorded. The registry is cross-referenced annually with the California State Death File, National Death Index, Social Security Death Master File, Medicare, Medicaid, Department of Motor Vehicles, Voter Registration, and National Change of Address Databases; active follow-up occurs through direct contact with hospitals, physician offices, patients, and their families. This has led to the development of a survival dataset in which 2% of data-points are censored in survival studies. Socioeconomic status is broken into quintiles based on a previously reported index of socioeconomic status derived from 2000 US census block data (including: median income, median rent, median house value, portion of residents with a blue-collar job, education index, proportion below 200% poverty level, and proportion older than 16 without a job).16
As the results of the various clinical trials evaluating the adjuvant therapy of pancreatic cancer were reported, the standard regimen for the adjuvant treatment of pancreatic cancer may have changed during the course the study. Therefore, additional analyses were adjusted by year of diagnosis grouped into three blocks: 1994 – 1997; 1998 – 2000; 2001-2002. The first report demonstrating efficacy of adjuvant therapy for pancreatic cancer is the GITSG trial published in 198517, 18; therefore 5-FU based chemoradiation was considered the standard of care until the late 1990s. In 1998, gemcitabine was approved by the FDA for use in the treatment of patients with advanced/metastatic pancreatic cancer19, 20, and many physicians began using gemcitabine in the adjuvant setting. We therefore considered 1998 through 2000 as an intermediate group in which there may have been a transition from 5-FU-based therapy to gemcitabine-based regimens.
Patients who underwent curative-intent resection alone were compared with patients who underwent curative-intent resection plus any adjuvant treatment (chemotherapy, radiation therapy, or both). Continuous data (age) are reported as the mean (+/- standard deviation) and were compared by unmatched t-tests. Categorical variables (sex, T stage, TNM stage, grade/differentiation, lymph node status, etc.) were compared using chi square tests. Univariate time comparisons with Kaplan-Meier survival estimates were used to compare survival between groups, and tested by log-rank test. Mortality is reported as overall mortality, rather than disease-specific mortality (although these curves are almost identical in pancreatic cancer). A Cox proportional hazards model was created to compare overall survival between patients who received surgery alone vs. surgery with adjuvant therapy, and adjusted for confounding variables. The survival estimates were adjusted to account for treatment changes (i.e., change in adjuvant therapy) across the three time periods. A p-value of 0.05 was considered to be significant. Analyses were performed with STATA 9.2 (Statacorp, College Station, TX).
26,518 patients with pancreatic cancer were identified, of which only 3,196 (12%) underwent a curative-intent pancreatic resection. The percentage of patients undergoing surgery increased with time; from 11% in the early group (1994-1997) to 13% in the later time frame (2001-2002; p=0.001). Of those that were resected, 58% (N= 1,852) received any adjuvant treatment (chemotherapy, radiation, or both). Characteristics of the study groups are presented in Table 1. Patients that received adjuvant therapy tended to be slightly younger, and were more likely to have poorer prognostic features (e.g., greater proportion of tumors > 3cm, poorly differentiated tumors, more likely to harbor positive nodes, and more advanced disease) than patients who did not receive adjuvant therapy. No significant disparity in receipt of adjuvant treatment was noted by socioeconomic status. Interestingly, fewer patients received adjuvant therapy as time progressed (1994-1997: 65%, 1998-2001: 52%; 2002-2003: 50%, p<0.001). Although there were only 162 T4 patients (Stage IVA by AJCC 5th edition criteria), there were 303 patients identified as Stage IV; these additional 141 patients (4.4% of the entire cohort) were grouped as Stage IV due to metastatic (i.e, M1) disease. In this situation, the delivery of surgical and systemic therapy is palliative, though it is unclear whether these patients were noted to have M1 disease intraoperatively or on final histopathology. In addition, 2768 patients (10.4% of the entire study group) were noted to have histology other than that of adenocarcinoma. These patients would not normally be considered for adjuvant chemotherapy and are thus unlikely to impact the analysis of the relative benefit of adjuvant therapy. It remains unclear what perception in the clinical utility of adjuvant therapy occurred in this time frame that decreased the enthusiasm for the administration of adjuvant therapy following surgical resection of pancreatic cancer.
Median overall survival for all patients was 16 months. Factors associated with improved survival included: age less than 60 years, negative lymph nodes, well-differentiated tumors, female gender, and receipt of any adjuvant therapy (Table 2). Interestingly, sex did not affect the survival in the entire sample, though it has been shown in other studies to have a minor impact on outcome.11 A trend towards improved median survival was seen in patients who received adjuvant therapy vs. surgery alone (16 mo. vs. 15 mo., p=0.11), but this did not result in increased proportions of long-term survivors (Figure 1A). Median overall survival was noted to improve between the three-time periods (Figure 1B; 14 mo. vs. 16 mo. vs. 17 mo, respectively; p<0.001). This improvement in survival over time was seen in patients who received surgery and adjuvant therapy, but not in patients who underwent surgery alone (data not shown).
A multivariate Cox analysis was then performed on those patients with completed data to exclude any effect of confounding variables in the patients with unknown histologic grade, tumor size or lymph node status (N=2,433; 76% of the study population). Female gender, negative lymph nodes, well-differentiated tumors, and receipt of adjuvant therapy were associated with improved survival following surgical resection of pancreatic cancer (Table 3). Receipt of any adjuvant treatment was associated with a modest overall survival benefit (HR = 0.79, 95% CI 0.72 – 0.87; Figure 2). We then examined the interaction of receipt of adjuvant therapy with tumor lymph node involvement as well as tumor grade on overall survival. We found that patients with lymph node positive disease achieved a survival benefit from the administration of adjuvant therapy (HR = 0.68, 95% CI 0.58-0.78), while those patients with lymph node negative disease did not (HR = 0.97, 95% CI 0.84-1.1) (Figure 3). It is interesting to note that the univariate analysis suggested a detrimental effect to the administration of adjuvant therapy in the lymph node negative patients (Table 2), though the multivariate analysis accounts for the additional tumor-related factors associated with worse outcome, which were more frequently present in the subgroup of patients receiving adjuvant therapy. When the impact of adjuvant therapy was examined by tumor grade, patients with well-differentiated tumors did not benefit (HR 1.2; 95% CI 0.97-1.7), though patients with moderately differentiated (HR 0.79; 95% CI 0.71-0.91) and poorly differentiated tumors (HR = 0.67, 95% CI 0.50-0.76; Figure 4) achieved a significant survival benefit. Although it would appear that patients suffered a poorer survival following adjuvant therapy for well-differentiated tumors, this was not statistically significant.
We then examined potential barriers to the administration of adjuvant therapy. Significant disparities in the delivery of cancer care have been noted based on age and ethnic status.11 Indeed, patients less than 60 years old were more likely than patients older than 60 to receive adjuvant therapy (64% vs 55%, p<0.001). In patients under 60 years, the adjuvant therapy group tended to have larger tumors (by T stage), positive lymph nodes, and fewer patients with stage I disease (Table 4). No significant disparity in receipt of adjuvant treatment was noted by ethnic status for the entire group. However, there did appear to be an interaction between age and ethnic status; in those patients <60 years, non-Hispanic white patients were slightly more likely to receive adjuvant therapy, and Hispanic patients slightly less likely (Table 4). The same decline in the delivery of adjuvant therapy by time period was also noted; 70% of patients in the first time period (375 of 537 in 1994-1997) compared with 58% in the latter time period (147 of 252 in 2001-2002).
Using a population-based dataset, we observed a modest improvement in overall survival with adjuvant therapy for surgically-resected pancreatic cancer that persisted after controlling for other factors influencing outcomes. While the current status of adjuvant therapy following resection of pancreatic cancer lacks a consensus, the five previous studies have demonstrated a consistent, modest survival benefit.3 A 7 month prolongation of median survival was observed in the GITSG trial of 5-FU-based chemoradiation17; while a more modest 2 month prolongation was observed in the CONKO trial of adjuvant gemcitabine.21 Despite over twenty years of clinical trials in the adjuvant therapy of pancreatic adenocarcinoma, only 1256 patients have been enrolled in these randomized trials which has been insufficient to allow the identification of patients subgroups who are more or less likely to benefit from adjuvant treatment strategies.18, 21-24 It is interesting to note that despite these trial data, up to one-half of patients in the latter time period did not receive therapy; whether this is due to referral patterns, changing reimbursement, or emerging pessimism of the disease is unclear.
Unlike the majority of adjuvant trials in other common malignancies (i.e. lung, colon and breast), the clinical trials of adjuvant therapy in pancreatic cancer have grouped together patients with Stage I-III disease with variable stratification of factors known to affect prognosis (e.g. lymph node status and histologic grade). Picozzi et al. have reported careful analysis of the reported trials of adjuvant therapy and noted variations among the trials that have confounded the interpretation of data.3 The GITSG trial had no stratification criteria, and did not report node positivity or tumor grade in the patient demographics.18 More recent trials such as the CONKO trial included some stratification variables (resection margin, T-stage and N-stage), though histologic grade has not been a stratification criteria in any of the recent major trials despite the impact on outcome that we and others have reported.11,13 The effect of tumor grade is more powerful on survival that tumor size, and may be more significant than lymph node status; we observed that patients with a well-differentiated tumor had a mean survival of 32 months while those with poorly differentiated tumors had a mean survival of just 7 months (Table 2). In the EORTC study, there was a 9% difference in the incidence of poorly differentiated tumors between the treatment arms21 while the ESPAC-1 study had an 8% difference in the incidence of poorly differentiated tumors among the treatment arms.23 Therefore, differences in established prognostic variables among treatment arms may have confounded the outcomes of past trials. Our data suggests that tumor grade has a greater impact in accounting for the differences in outcome from adjuvant therapy than tumor size or lymph node status, and should be considered as a stratification variable in future clinical trials of adjuvant therapy.
Some of the improvements in survival in the adjuvant therapy group may be accounted for by selection bias. Those patients who received adjuvant therapy were slightly younger and thus may have had fewer comorbidities as well as recovered from the procedure with a greater functional reserve allowing them to undergo adjuvant therapy. Similarly, patients receiving adjuvant therapy generally had more advanced/aggressive disease manifest by larger tumors, higher rate of node-positivity, and lower proportion of well-differentiated tumors. Our data does not allow us to investigate specific chemotherapy regimens, although there is an improvement in survival by diagnosis year, which may suggest that newer chemotherapeutic agents (e.g., gemcitabine) may have had a positive impact on disease control and survival. We do not believe that this improvement in survival over time is a consequence of stage migration, as there was no change in the survival over time in patients who received surgery alone.
Additional variables we examined that may impact outcome from surgical treatment as well as adjuvant therapy included age, sex, and ethnicity. Although younger patients who received adjuvant therapy appeared to do poorly, this likely represents a selection bias; a more aggressive approach is generally taken with younger patients as demonstrated by a higher proportion of younger patients receiving adjuvant therapy. Age was therefore not included in our multivariate analysis as opposite effects of adjuvant therapy were seen in older and younger patients and proportional hazards assumptions were not met. Women initially did not appear to have longer survival compared to men on univariate analysis (Table 2); however, on multivariate analysis, after accounting for other prognostic features, women overall did slightly better than men (Table 3). This may suggest an impact of hormonal factors or other clinicopathologic differences between men and women in this disease.
We did not include race or socioeconomic status in our multivariate analysis as we have previously noted that these are not predictors of survival in our cohort; rather, that they predict stage of disease at diagnosis and receipt of treatment, which are better predictors of survival.11, 25 Furthermore, there was no difference in receipt of adjuvant therapy by socioeconomic groups (Table 1). We did note that non-Hispanic white patients under 60 years had markedly worse survival with receipt of adjuvant therapy. Previous studies have suggested that non-Hispanic white patients are more likely to present with earlier stage disease, and more likely to receive treatment (surgical and non-surgical) than African-American or Asian patients11, 25-28, and our analysis did show that a higher proportion of non-Hispanic white patients under 60 years did receive adjuvant therapy. Whether the decrement in survival seen in the adjuvant treatment group is a product of a selection bias due to earlier presentation of disease in non-Hispanic white patients, increased referrals for surgery, or related to morbidities of the adjuvant therapy itself is unclear. Additionally, there is the possibility of different biologic characteristics of tumors in different racial/ethnic groups which may affect both natural history of disease, and responsiveness to chemotherapeutics.28
In summary, our analysis of a large population-based database demonstrates that adjuvant therapy after curative resection for pancreas cancer benefits specific subgroups of patients - specifically, patients with positive lymph nodes or poorly differentiated tumors. The lack of benefit in “favorable” subgroups (i.e. well differentiated tumors and node-negative disease) is more likely a consequence of the relative lack of efficacy of current treatment regimens. Therefore, future trials of adjuvant treatment regimens should either stratify patients for entry based on known prognostic variables (e.g. tumor grade or lymph node status) or focus on patients with poor prognostic features to allow the identification of therapeutic regimens with activity.
The authors are indebted to Dr. Garth Utter for assistance with statistical analyses. The collection of cancer incidence data used in this study was supported by the California Department of Public Health as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute’s Surveillance, Epidemiology and End Results Program under contract N01-PC-35136 awarded to the Northern California Cancer Center, contract N01-PC-35139 awarded to the University of Southern California, and contract N02-PC-15105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention’s National Program of Cancer Registries, under agreement #U55/CCR921930-02 awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the author(s) and endorsement by the State of California, Department of Public Health, the National Cancer Institute, and the Centers for Disease Control and Prevention or their contractors and subcontractors is not intended nor should be inferred.