Fax: (713) 563-2366
Prognostic factors in patients with unresectable locally advanced pancreatic adenocarcinoma treated with chemoradiation
Article first published online: 2 NOV 2006
Copyright © 2006 American Cancer Society
Volume 107, Issue 11, pages 2589–2596, 1 December 2006
How to Cite
Krishnan, S., Rana, V., Janjan, N. A., Abbruzzese, J. L., Gould, M. S., Das, P., Delclos, M. E., Palla, S., Guha, S., Varadhachary, G., Evans, D. B., Wolff, R. A. and Crane, C. H. (2006), Prognostic factors in patients with unresectable locally advanced pancreatic adenocarcinoma treated with chemoradiation. Cancer, 107: 2589–2596. doi: 10.1002/cncr.22328
- Issue published online: 17 NOV 2006
- Article first published online: 2 NOV 2006
- Manuscript Accepted: 2 OCT 2006
- Manuscript Revised: 13 SEP 2006
- Manuscript Received: 7 AUG 2006
- locally advanced;
- pancreatic cancer;
- prognostic factors;
Although patients with locally advanced pancreatic cancer (LAPC) have an extremely poor prognosis, they are a heterogeneous group. Prognostic factors are inadequately defined for disease-free survival and overall survival in patients with LAPC who are receiving chemoradiation, so more definitive prognostic factors would be very useful for designing clinical trials.
Between December 1993 and July 2005, 247 patients with nonmetastatic LAPC were treated at M. D. Anderson Cancer Center (Houston, Tex) with concurrent chemoradiation (CRT). Median radiation dose was 30 Gy (range, 15–52.2 Gy). Radiosensitizers included 5-fluorouracil (54%), gemcitabine (33%), and capecitabine (13%). Actuarial univariate and multivariate statistical methods were used to determine significant prognostic factors for disease-free survival and overall survival.
Median follow-up was 4.3 months (range, 1–63 months). Median disease-free survival and overall survival were 4.2 months and 8.5 months, respectively. On univariate analysis, prognostic factors for improved disease-free survival were a Karnofsky performance scale (KPS) status of >80 (P < .01) and a hemoglobin (Hgb)level at presentation of ≥12 (P = .03). On multivariate analysis, KPS was the only independent prognostic factor for disease-free survival. Median disease-free survival was 4.9 months among patients with a KPS score of >80 and was 3.9 months among those with a KPS score of ≤80. On univariate analysis, prognostic factors for improved overall survival were an Hgb level of ≥12 (P = .02), KPS>80 (P < .001), and <5% weight loss (P = .03). On multivariate analysis, Hgb and KPS were independent prognostic factors for overall survival.
In the current study, KPS score was an independent prognostic factor for disease-free and overall survival among patients treated with chemoradiation for LAPC. The pretreatment Hgb level was an additional independent prognostic factor for overall survival. Cancer 2006. © 2006 American Cancer Society.
Pancreatic cancer is the third most common gastrointestinal malignancy and the fifth leading cause of cancer-related death in western countries.1 In 2006, estimates are that 33,730 new cases of pancreatic cancer will be diagnosed and a roughly equal number of deaths will be attributed to pancreatic cancer in the United States.2 Unfortunately, <10% of patients are eligible for a margin-negative surgical resection, the only potentially curative treatment for pancreatic cancer, which confers a 15% to 25% rate of 5-year overall survival.3 About two-thirds of all pancreatic cancer patients have radiographically detectable metastatic disease at presentation, and the remaining patients have locally advanced unresectable disease.4 Typically in the United States, standard treatment for locally advanced pancreatic cancer consists of a combination of chemotherapy and radiation therapy, but the integration of these modalities and their respective dose schedules vary considerably.
Treatment with radiation therapy, with or without chemotherapy, has not been prospectively compared with best supportive care in a randomized study. Early randomized trials by the Gastrointestinal Trials Study Group (GITSG) have demonstrated that combining chemotherapy and radiation therapy is more effective than chemotherapy or radiotherapy alone.5, 6 These trials used bolus 5-fluorouracil (5FU) chemotherapy and split-course radiation therapy, which could be considered suboptimal. Combined modality regimens nearly double the median survival over radiation therapy alone and marginally improve median survival over chemotherapy alone. In contrast to these studies, an Eastern Cooperative Oncology Group randomized study that compared 5FU chemotherapy with combined modality therapy showed no difference in median survival between the 2 groups, although toxicities were significantly higher in the chemoradiation group.7 Other chemotherapy regimens used during this period included streptozocin, mitomycin, methyl lomustine, and doxorubicin.6, 8 None of these agents was shown to be superior to 5FU. Since then, the promising antitumor activity of gemcitabine in locally advanced and metastatic pancreatic cancer and its potential for radiosensitization have led to renewed interest in combining newer agents with radiation therapy.9–14 In all of these studies, no clear stratification of the cohort of treated patients into low-risk and high-risk groups based on pretreatment patient-specific and tumor-specific variables was performed.
The prognostic factors for pancreatic cancer described in literature include performance status, carbohydrate antigen 19-9 (Ca 19-9) level, C-reactive protein levels, and the degree of anaplasia.7, 15–17 There are insufficient data on prognostic factors in patients with unresectable pancreatic cancer who receive chemoradiation as definitive therapy. To our knowledge, there are only 2 small studies that discuss prognostic factors in this population.7, 15 The current study was undertaken to identify potential prognostic factors and their relative contribution to survival in a large cohort of locally advanced pancreatic cancer patients who were treated with chemoradiation therapy. The current study represents the largest cohort of unresectable pancreatic cancer patients treated with concurrent chemoradiation therapy with mature follow-up.
MATERIALS AND METHODS
Patient Identification and Selection
Between December 1993 and July 2005, 370 consecutive patients received chemoradiation therapy for locally advanced pancreatic cancer initially deemed unresectable for cure at M. D. Anderson Cancer Center. Of these patients, 76 received systemic gemcitabine-based chemotherapy before they received chemoradiation therapy. Forty-seven patients received concurrent chemotherapy on protocol with bevacizumab. The remaining 247 patients received chemoradiation with either fluoropyrimidine-based or gemcitabine-based chemotherapy as their initial and primary treatment modality. To avoid the potential confounding impact of either bevacizumab or the use of chemotherapy before chemoradiation, this group of patients was selected for analysis. The University of Texas M. D. Anderson Cancer Center Institutional Review Board approved this retrospective study. All rules and regulation of the Health Insurance Portability and Accountability Act (HIPAA) were strictly followed during the execution of this study.
Pretreatment evaluation included a complete history and physical examination, biopsy, endoscopic retrograde cholangiopancreaticography (ERCP), dedicated pancreatic cancer protocol abdominal and pelvic computed tomography (CT 3-phase contrast-enhanced thin-slice helical scan), and chest x-ray. Complete laboratory tests included a full blood count, blood electrolytes, creatinine, urea, liver transaminases, alkaline phosphatase, and total bilirubin. Routine laparoscopy was not performed. The treating physician(s) prospectively recorded Karnofsky performance scale (KPS) status and weight loss before therapy. A dedicated multidisciplinary team, including a medical oncologist and a radiation oncologist, evaluated all patients; selected patients were seen by surgical oncologists. Patients were discussed at a weekly multidisciplinary conference. Tumors that extended to the celiac axis or the superior mesenteric artery or tumors that occluded the superior mesenteric (SMV)-portal venous confluence were deemed locally advanced and unresectable based on review of CT images. Tumors involving only the SMV were deemed potentially resectable at our institution and, therefore, excluded from the current study. Biopsy or cytology confirmation of diagnosis was required for treatment in all patients. Patients who had diagnostic pathology performed outside of M. D. Anderson Cancer Center had their pathology findings reviewed at our institution.
Among a total of 247 patients, there were 144 (58%) men and 103 (42%) women. Median age was 64 years (range, 39–85 years). Patient characteristics are summarized in Table 1.
|African American||21 (9)|
|Abdominal pain||166 (67)|
|Back pain||81 (33)|
|Change in bowel pattern||91 (36)|
|≤12 g/dL||60 (25)|
|>12 g/dL||177 (75)|
|% Weight loss|
Treatment characteristics are summarized in Table 2. Radiation therapy was administered by using mega voltage x-rays. Most (82%) patients received radiation therapy that used a 4-field conformal technique. Other techniques that were used included 2-field (15%), 3-field (1%), and intensity-modulated radiation therapy (IMRT) (2%). Treatment volumes were nodal in most patients (87%), and local fields (gross tumor volume with margin) were used for the remaining patients. Most (89%) patients received 30 Gy of radiation therapy in 2 weeks (range, 15–52.2 Gy). Concurrent chemotherapy consisted of 5FU (n = 133, 54%), gemcitabine (n = 81, 33%), and capecitabine (n = 33, 13%). 5FU was administered as a protracted venous infusion at a dose of 300 mg/m2 daily Monday through Friday. Gemcitabine was administered at a dose of 350–400 mg/m2 infused over 30 minutes weekly for 7 weeks starting 24 hours to 48 hours after the first radiation dose. Capecitabine was given at a dose of 800–900 mg/m2 in divided doses twice daily during the days when radiation therapy was administered.
|30 Gy in 10 fractions||220 (89)|
|50.4 Gy in 28 fractions||27 (11)|
|Tumor size (area axial CT scan)|
|Moderately differentiated||31 (36)|
|Poorly differentiated||38 (44)|
|Signet ring||1 (<1)|
|Giant cell||1 (<1)|
Chemotherapy dose adjustments were based on previously published criteria.18 All patients were evaluated every week during treatment, and all adverse effects were monitored and recorded. Complete blood counts and blood chemistries were obtained weekly during treatment.
Patients were scheduled for follow-up visits with a medical or radiation oncologist every 3–4 months. Abdominopelvic CT scans and chest x-rays were performed at these visits to monitor diseases status. Therapy was individualized at the time of progression.
Various factors selected for investigation were chosen on the basis of previously published reports.15, 19–21 In addition, routine demographic, patient-specific, tumor-specific, and treatment-specific variables were analyzed. Demographic variables included age, sex, and race. Patient-specific variables analyzed included Karnofsky performance scale score (KPS > 80 versus KPS ≤ 80), pretreatment hemoglobin and bilirubin levels, and weight loss >5% in 3 months before presentation. Hemoglobin was analyzed as a continuous variable and by dichotomizing at the median value of 12 g/dL. Bilirubin was analyzed as a continuous variable and by dichotomizing at the traditional normal level of 1 g/dL. We had insufficient data on tumor size, tumor differentiation, and tumor markers including Ca 19-9, CEA, and Ca125. We believed that the retrospective analysis of the presence or absence of various presenting symptoms (including jaundice, anorexia, nausea, abdominal and/or back pain, and fatigue) was not sufficiently robust or necessarily biologically plausible enough to warrant inclusion in our study of potential prognostic factors. We did analyze treatment-specific factors and included radiation dose and the type of chemotherapy given concurrently with radiation therapy.
The major endpoints of this study were overall survival and disease-free survival. We also analyzed local recurrence (defined as any recurrence at or adjacent to the initial primary site as determined by CT scans), lymph node recurrence (any recurrence in the regional lymph nodes on CT imaging), and metastatic recurrence. Severe toxicity was defined as treatment-related toxicity that resulted in hospitalization for supportive care for >5 days; symptomatic gastrointestinal bleeding with endoscopic evidence of gastric or duodenal ulceration or with resultant transfusion; more than 3-dose deletions during the planned 7 weekly gemcitabine infusions; or treatment-related toxicity that required surgical intervention or that resulted in death.18
Duration of overall survival was calculated in months from the first day of radiation treatment to the time of death or censorship. Survival time was censored at the time of the last follow-up on record if death was not observed. Disease-free survival time was censored at the date of the last follow-up on record if no recurrences were observed, and death was not observed. Dates of local and distant failure were determined by using imaging studies, primarily abdominopelvic CT scans. The significance of differences in proportions was calculated with a chi-square test, and the differences in means with a Student t test.
Survival probabilities were estimated nonparametrically by using the Kaplan-Meier product-limit method.22 Prospectively selected prognostic factors were investigated by univariate and multivariate analyses. Each variable identified as statistically significant on univariate analysis was used in the multivariate model. Comparisons between groups for overall survival and disease-free survival were performed with a log-rank test.23 Cox proportional hazards modeling was used to examine the effect of various prognostic factors on overall survival and time to local progression depending upon results of the univariate analysis.24 All tests were 2-sided and P ≤ .05 was considered statistically significant.
Median time from diagnosis to start of treatment was 0.6 months. The median clinical and radiographic follow-up time was 4.3 months (range, 1–63 months). Most (227 of 247, 92%) patients were deceased at the time of this analysis. Median follow-up was 5.3 months among the 20 survivors (range, 1–63 months).
Tumor and treatment characteristics are summarized in Table 2. The median actuarial overall survival was 8.5 months (range, 1–78 months; Fig. 1). Estimated rates of overall survival at the end of 1 year and 2 years were 25% and 8%, respectively. Median disease-free survival was 4.2 months (range, 1–63 months) from the start of radiation therapy (Fig. 1). Median time to local failure was 6.0 months (range, 1–63 months), and median time to distant failure was 5.7 months (range, 1–63 months). At the time of analysis, 227 patients had died; 223 deaths were due to primary cancer, and 4 deaths were due to other causes. Seven (3%) patients were able to undergo margin-negative resection after chemoradiation therapy for pancreatic cancer that was initially considered unresectable. These patients had a median overall survival of 29.4 months (range, 5.6–63 months) and a median disease-free survival of 24.2 months (range, 5.6–63 months). At the time of analysis, 4 patients were alive.
Twenty-two (27%) of 81 patients treated with gemcitabine and 10 (8%) of 133 patients treated with 5FU had severe acute toxicity (P ≤ .001). Only 1 patient treated with capecitabine (Xeloda) had severe toxicity and required total parenteral nutrition for intractable nausea and vomiting. Among 10 patients treated with 5FU, 8 had to be admitted for supportive care, and 2 had gastrointestinal bleeding. The patients with severe toxicity due to gemcitabine included 4 patients with gastrointestinal bleeding and 14 patients admitted for management of intractable nausea and vomiting, 3 of whom also had grade 3 hematologic toxicity. Eight patients treated with gemcitabine had grade 3 hematologic toxicity.
Table 3 provides results of the univariate analysis of all factors considered to be prognostic for survival outcomes. On univariate analysis, the prognostic factors for improved overall survival were Hgb ≥ 12 (P = .02; Fig. 2), KPS > 80 (P < .001; Fig. 3), and <5% weight loss in the preceding 3 months (P = .03). The significant prognostic factors for disease-free survival were KPS > 80 (P < .01; Fig. 4) and Hgb ≥ 12 (P = .03). In particular, the concurrent chemotherapy regimen had no impact on survival outcomes.
|Outcomes||P||Hazard ratio (CI)|
|KPS ≤80||<.001||1.59 (1.21–2.09)|
|Hgb <12||.02||1.44 (1.06–1.95)|
|<5% Weight loss||.03||0.72 (0.53–0.97)|
|KPS ≤80||<.01||1.42 (1.09–1.86)|
|Hgb <12||.03||1.42 (1.05–1.91)|
On multivariate analysis, Hgb<12 (hazard ratio [HR] = 1.48, P = .03) and KPS ≤ 80 (HR = 1.55, P < .01) were independent prognostic factors for overall survival. Patients with a pretreatment Hgb of <12 had a median overall survival of 7.3 months compared with 9.0 months for those with Hgb of ≥12 (Fig. 2). Median overall survival was 10.3 months among patients with KPS > 80 and 7.6 months among those with KPS ≤ 80 (Fig. 3). On multivariate analysis (Table 4), KPS was the only independent prognostic factor for disease-free survival. Median disease-free survival was 4.9 months among patients with KPS > 80 and 3.9 months among those with KPS ≤ 80 (Fig. 4).
|Outcome||P||Hazard Ratio (CI)|
|KPS <80||<0.01||1.55 (1.13–2.14)|
|Hgb <12||.03||1.48 (1.05–2.06)|
|KPS <80||.03||1.35 (1.02–1.78)|
Pretreatment patient-related and tumor-related characteristics that are prognostic for overall survival and disease-free survival can be identified in patients with locally advanced pancreatic cancer who are treated with chemoradiation therapy. Our analysis suggests that the KPS score and the pretreatment hemoglobin level are independent prognostic factors for overall survival, while KPS alone is an independent prognostic factor for disease-free survival.
A handful of small previous reports7, 15, 16, 20, 25–27 have evaluated prognostic factors that determine outcome in patients with unresectable pancreatic cancer (Table 5). Many of these studies included patients with metastatic pancreatic cancer in addition to unresectable nonmetastatic pancreatic cancer. Only 1 of these studies identified prognostic factors for disease-free survival.7 Performance status is the only prognostic factor that is consistently shown to influence overall survival in definitively treated patients. No definitive prognostic factors for disease-free survival have been consistently identified. Compared with the 2 studies that evaluated prognostic factors in unresectable patients receiving definitive chemoradiation as treatment,7, 15 our study additionally showed that the pretreatment Hgb level was prognostic for overall survival. The precise significance of the pretreatment Hgb level is not immediately apparent. The fact that it merely influenced overall survival and not disease-free survival would suggest that patients died of other causes. However, only 4 patients died of causes other than their primary cancer. It is possible that pretreatment Hgb level is a partial surrogate for KPS. All the same, these results would suggest that pretreatment Hgb level may be factored into the determination of additional statistical endpoints (such as disease-free survival) for comparison in future prospective cancer clinical trials. In addition, it may be worthwhile to analyze differences in pretreatment Hgb level between trials when comparing outcomes of treatment groups across clinical trials. Admittedly, this approach cannot be used as a surrogate for prospective randomized trials.
|Authors||Patients||Treatments||Significant prognostic factors|
|Klaassen DJ et al7||91 (unresectable only)||Definitive chemoradiation, chemotherapy||Performance status, degree of anaplasia, reduced appetite|
|Ikeda M et al15||55 (unresectable only)||Definitive chemoradiation||Performance status, Ca19-9 level >1000 and regional lymph node swelling|
|Terwee CB et al26||1020 (pooled from 5 studies, metastatic)||Palliative surgery||Presence of jaundice predicted better prognosis|
|Engelken FJ et al27||325 (unresectable & metastatic)||Palliative surgery||Absence of therapeutic intervention, CRP > 5 mg/dL, leukocytosis, GGT > 165 U/L|
|Ueno H et al25||103 (metastatic only)||Chemotherapy||Performance status, CRP > 5 mg/dL, Ca19-9 level >10,000 U/mL|
|Sezgin C et al16||67 (Unresectable & metastatic)||Chemotherapy only (gemcitabine)||Performance status|
|Ishii H et al20||65 (unresectable & metastatic)||Systemic chemotherapy||Performance status, CEA level <10 ng/mL, absence of distant metastases|
KPS score has been a recognized prognostic factor for treatment outcomes in multiple cancers treated with chemoradiation therapy with definitive intent.28–30 This may be due to the inherent aggressiveness of tumors in patients with poorer KPS scores, increased disease burden among these patients, or inability and/or unwillingness of these patients to receive all prescribed treatment. Similarly, Hgb level is a known prognostic factor for treatment outcomes in other cancer, such as cervical cancer, breast cancer, and head and neck cancer.31–33 Anemia related to malignancy is a poorly understood phenomenon, and it is not immediately evident that correction of this anemia by using growth factors reverses the poor prognosis conferred by anemia. A few recent trials have shown that correction of anemia by using recombinant erythropoietin has led to poorer outcomes.34, 35 Nevertheless in our analysis, identification of anemia as a predictive factor for overall survival further corroborates the hypothesis that hemoglobin level has profound influences on patient and tumor outcomes.
As with most retrospective analyses, multiple caveats should be applied to interpretation of these findings. First, the results of our study are not universally applicable to patients with obvious metastasis or to those having resectable disease. Second, as opposed to definitive treatment comparisons from prospective trials, retrospective analyses of patient-related and treatment-related variables are exploratory, and, therefore, serve to generate testable hypotheses, not to prove hypotheses. Such an exercise may identify variables worthy of consideration, such as eligibility criteria and/ or randomization stratification variables, for future prospective cancer clinical trials. In addition, it is sometimes possible to identify survival cohorts, which may be helpful in determining more accurate statistical endpoints for comparison in future prospective cancer clinical trials.
As with most retrospective studies, this study has biases that deserve mention. Selection bias could have been introduced by our exclusion of patients who were on protocol therapy that included bevacizumab/biologics. Changes in imaging quality, radiation therapy techniques, and use of additional supportive care measures in later years (such as the more prevalent use of newer generation anti-emetics and colony-stimulating factors) during this period might have introduced unknown confounding factors. Another limitation was the lack of data on tumor size, tumor differentiation, and tumor markers like CEA and Ca19-9. In the absence of adequate data on these tumor markers, the power to detect 1 or more of them as significant predictors is lacking. Last, because most patients (89%) received an unconventional radiation dose and fractionation (30 Gy in 10 fractions), significant conclusions cannot be drawn regarding the prognostic value of dose.
The major strength of this study is its large sample size. It is one of the largest series that has evaluated this selected group of patients with unresectable pancreatic cancer who underwent treatment with definitive chemoradiation therapy. Treatment criteria and the treatment approach were relatively consistent throughout this period. The multidisciplinary management of patients by a dedicated team of clinicians who specialize in pancreatic cancer management is also an asset to such an investigation.
In conclusion, our study suggests that performance status and pretreatment hemoglobin level are independent prognostic factors for patients with unresectable pancreatic cancer being treated with chemoradiation as definitive treatment. These results can be used for designing future clinical trials, for stratifying patients to various treatment strategies, and for predicting life expectancy. The information could then be used by the treating oncology team to define an optimal therapeutic strategy for a particular patient, or to incorporate stratification variables in a study design, or to analyze future clinical trial findings for this group of patients.
- 5Therapy of locally unresectable pancreatic carcinoma: a randomized comparison of high dose (6000 rads) radiation alone, moderate dose radiation (4000 rads + 5-fluorouracil), and high dose radiation + 5-fluorouracil: The Gastrointestinal Tumor Study Group. Cancer. 1981; 48: 1705–1710., , , et al.
- 8Phase II studies of drug combinations in advanced pancreatic carcinoma: fluorouracil plus doxorubicin plus mitomycin C and two regimens of streptozotocin plus mitomycin C plus fluorouracil. The Gastrointestinal Tumor Study Group. J Clin Oncol. 1986; 4: 1794–1798.
- 23Regression models and life tables. [with discussion] J R Stat Soc Ser B. 1972; 34: 187–220.