10.1245/s10434-006-9104-y
Annals of Surgical Oncology 13:1485-1492 (2006)
© 2006 Society of Surgical Oncology
Evaluation of Preoperative Therapy for Pancreatic Cancer Using a Prognostic Nomogram
Rebekah R. White, MD1,
Michael W. Kattan, PhD2,
John C. Haney, MD3,
Bryan M. Clary, MD3,
Theodore N. Pappas, MD3,
Douglas S. Tyler, MD3 and
Murray F. Brennan, MD1
1 Department of Surgical Oncology, Memorial Sloan-Kettering Cancer Center, New York, USA
2 Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, USA
3 Department of Surgery, Duke University Medical Center, Durham, USA
Correspondence: Address correspondence and reprint requests to: Murray F. Brennan, MD, Department of Surgical Oncology, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021, USA; E-mail: m-brennan{at}mskcc.org
 |
ABSTRACT
|
|---|
Background: Theoretical benefits of preoperative chemoradiation therapy (preop CRT) for pancreatic cancer include improved efficacy, resectability, and patient selection. The goal of this study was to evaluate the applicability of a nomogram, which was developed for patients undergoing resection without preop CRT and which incorporates several post-resection pathological factors, to a population of patients who received preop CRT prior to resection.
Methods: From 1994 to 2004, 82 patients with biopsy-proven, radiographically localized adenocarcinoma of the pancreatic head underwent preop CRT followed by pancreaticoduodenectomy (PD); 50 concurrent patients underwent PD without preop CRT. Mean nomogram-predicted disease-specific survival (DSS) rates were compared with observed DSS rates from the time of resection.
Results: Despite having more locally advanced tumors on initial staging (21 vs. 8%; P < .05), patients who received preop CRT had smaller resected tumors (mean 2.3 vs. 3.1 cm; P < .01), were less likely to have T3 tumors (54 vs. 80%, P < .01), were less likely to have positive lymph nodes (29 vs. 58%, P < .01), and had fewer positive lymph nodes (mean .4 vs. 1.9, P < .01), all factors that imply treatment effect and favorably impact on nomogram-predicted DSS. Observed DSS was similar to predicted DSS in both groups.
Conclusions: The similarity in observed and predicted DSS following resection in patients who received preop CRT suggests that the effects of preop CRT—whether treatment, selection, or no effect—are reflected by the nomogram. The ability of the nomogram to evaluate the effects of preop CRT on survival is limited by the potential effects of preop CRT on factors within the nomogram.
Key Words: Pancreatic cancer • Prognosis • Nomogram • Preoperative therapy • Neoadjuvant therapy • Chemoradiation therapy
|
INTRODUCTION
|
|---|
Randomized trials have suggested that the benefit of postoperative (adjuvant) chemoradiation therapy in patients with resected pancreatic cancer—if it exists—is small.1–4 In these trials, up to one third of patients did not receive planned postoperative therapy due to complications or delayed recovery.2,4 One of several potential advantages of preoperative (neoadjuvant) chemoradiation therapy (preop CRT) is that all resected patients receive chemoradiation in a timely fashion. Other theoretical advantages include the delivery of chemoradiation to well-oxygenated tissues, the avoidance of radiation to fixed loops of bowel, and the potential to improve the resectability of marginally unresectable tumors. One of the most compelling rationales for preop CRT, however, is that occult metastatic disease present at the time of diagnosis is given the opportunity to manifest itself, thereby avoiding the morbidity of patient resection. Data from non-randomized series have suggested that patients who receive preop CRT prior to resection may have better survival than patients who do not receive preop CRT prior to resection.5–7 In these series, 10 to 30% of patients manifest distant metastatic disease during preoperative therapy and are not resected. How much of the apparent improvement in survival is attributable to treatment effect versus selection effect is unclear.
Nomograms are tools, designed from established databases, to predict outcome. They are capable of utilizing individual patient data not included in conventional staging systems to predict individual outcome more accurately. Most nomograms rely heavily on pathological factors and do not include treatment factors. Theoretically, a nomogram derived from a patient dataset that included a single treatment modality (i.e., resection) could be used to test the impact of an additional therapy. If an additional treatment improved outcome, then the observed outcome should be better than the predicted outcome. A nomogram has been developed at Memorial Sloan-Kettering Cancer Center (MSKCC) that predicts survival for patients with resected pancreatic cancer based on a combination of known prognostic factors (Fig. 1
).8 This nomogram was constructed using an internal population of patients who underwent resection without pre-operative therapy and has been validated on an external population of patients who were also treated with this standard approach.9
View larger version (15K):
[in this window]
[in a new window]
|
FIG. 1. A prognostic nomogram for predicting 12-, 24-, and 36-month disease-specific survival probabilities following resection of pancreatic cancer. Reproduced with permission from Brennan et al.8
|
|
The goal of the current study was to evaluate the applicability of a nomogram that incorporates several post-resection pathological tumor variables to a population of patients who received preop CRT prior to resection. At Duke University Medical Center (DUMC), most patients with either potentially resectable or locally advanced pancreatic cancer receive preop CRT. We hypothesized that, if the predominant effect of preop CRT is either to treat distant metastatic disease or to select out patients with occult distant metastatic disease, observed survival in patients undergoing preop CRT followed by resection would be better than the survival predicted by the nomogram for this group of patients. If the predominant effect of preop CRT is to improve locoregional tumor variables such as size, T-stage, and nodal status, factors that are included within the nomogram, then observed survival might be accurately predicted by the nomogram.
|
METHODS
|
|---|
Between 1994 and 2004, 196 patients with biopsy-proven, radiographically localized adenocarcinoma of the pancreatic head received predominantly 5-flourouracil (5FU)-based CRT with neoadjuvant intent at DUMC. The current study focuses on the 82 patients who subsequently underwent pancreaticoduodenectomy, including 64 of 117 patients with potentially resectable tumors on initial staging CT (55% resected) and 18 of 79 patients with locally advanced tumors on initial staging CT (23% resected), as defined by circumferential venous involvement and/or any arterial involvement. Non-circumferential venous involvement was considered potentially resectable, and vascular resection was performed when deemed necessary by the surgeon. Preop CRT patients were identified and maintained in a prospective database; additional data for the nomogram (weight loss, back pain, number of positive and negative lymph nodes) were collected by retrospective review. Fifty concurrent patients with pancreatic adenocarcinoma who underwent pancreaticoduodenectomy at DUMC without preop CRT due either to lack of preoperative tissue diagnosis, inadequate biliary drainage, or patient/surgeon preference were retrospectively identified.
Preop CRT regimens were extremely varied, and over a third of patients received CRT at institutions close to their homes. A planned total dose of 4500 cGy of external-beam radiation therapy (EBRT) was delivered in 180 cGy fractions, with or without a 540 cGy boost to the tumor bed. The median total dose of EBRT received by patients who underwent resection was 4500 cGy (range 900–5400 cGy). The vast majority received infusional 5-FU, either alone (63 patients), with bolus Mitomycin-C (four patients), or in combination with Mitomycin-C and cisplatin (seven patients). Five patients received an oral formulation of 5FU, and three patients received Gemcitabine as part of phase I clinical trials. Following resection, 30% of patients in the preop CRT group received additional adjuvant chemotherapy (most commonly Gemcitabine), while 75% of patients in the no preop CRT group received adjuvant 5FU-based CRT.
Disease-specific survival (DSS) was calculated using the date of death due to disease or last follow-up. Observed DSS was estimated using the Kaplan-Meier method and compared using the log-rank test. Comparisons between the patients who did and did not receive preop CRT were performed using the Chi-squared test for dichotomous variables, the Student’s T test for continuous variables, and the Mann-Whitney U test for ranked categorical variables. Predicted DSS probabilities were calculated using the corrected group-prognosis method10 applied to the regression model underlying the MSKCC nomogram (Fig. 1
). 8 Using this method, the probabilities at each failure time were then averaged across all patients, and the predicted survival curve was graphed as the mean survival versus time. In addition, a new multivariate model was specifically constructed for the patients who received preop CRT with Cox proportional hazards regression using the variables included within the MSKCC nomogram, excluding splenectomy and tumor location, since patients with tumors in the body or tail were not included in this study.
|
RESULTS
|
|---|
The characteristics of the two groups are listed in Table 1. Despite having a higher proportion of locally advanced tumors on initial staging CT (22 vs. 8%; P < .05), patients who received preop CRT had smaller resected tumors (mean 2.3 vs. 3.1 cm; P < .01), were less likely to have T3 tumors at resection (54% vs. 80%, P < .01), were less likely to have positive lymph nodes (29 vs. 58%, P < .01), and had fewer positive lymph nodes (mean 0.4 vs. 1.9, P < .01). Three patients who received preop CRT (4% of resected) had complete pathologic responses but were considered T1 for application of the nomogram. The groups were similar with respect to age, gender, differentiation, number of negative lymph nodes, need for portal vein resection, margin status, back pain, and weight loss.
Median DSS from the time of diagnosis in patients who received preop CRT prior was 26 vs. 20 months in patients who did not receive preop CRT (P < .05; Fig. 2). However, since the nomogram incorporates several factors obtainable only at the time of resection (e.g., lymph nodes, margins, need for portal vein resection) and was designed to predict survival from the time of resection, we utilized DSS from time of resection for all subsequent analyses. The median time interval from diagnosis to resection in patients who received preop CRT was 3.6 months (range 2.5 to 50 months). The difference in post-resection DSS between the two groups was not statistically significant (24 vs. 20 months; Fig. 3). Patients who had locally advanced tumors prior to CRT had significantly worse survival than did patients who had potentially resectable tumors prior to CRT (35 vs. 16 months, P = .05). Even after exclusion of patients who had locally advanced tumors, the difference in post-resection DSS between patients who did and did not receive preop CRT prior to resection was not statistically significant (35 vs. 20 months). Median post-resection follow-up in surviving patients was 31 months in patients who received preop CRT group but only 14 months in patients who did not receive preop CRT. Actual five-year survival in the 41 patients who received preop CRT prior to resection in 2001 was 10%, similar to the MSKCC population. 11
View larger version (7K):
[in this window]
[in a new window]
|
FIG. 2. Disease-specific survival from the time of diagnosis for patients who received preop CRT (dashed line) versus patients who did not receive preop CRT (solid line) prior to resection.
|
|
View larger version (6K):
[in this window]
[in a new window]
|
FIG. 3. Disease-specific survival from the time of resection for patients who received preop CRT (dashed line) versus patients who did not receive preop CRT (solid line) prior to resection.
|
|
Predicted DSS was compared with observed post-resection DSS by plotting mean nomogram-predicted DSS for each group at 1, 2, and 3 years against observed DSS at these time points (Fig. 4). The diagonal line represents the performance of an ideal nomogram. Observed DSS was similar to predicted DSS for patients who did not receive preop CRT prior to resection, confirming that the nomogram was applicable to similar patients from another institution. Observed DSS was also very similar to predicted DSS for patients who did receive preop CRT prior to resection, suggesting that either preop CRT has little effect on post-resection DSS from the time of resection or that any effects of preop CRT on DSS are accounted for by locoregional tumor factors included within the nomogram.
View larger version (11K):
[in this window]
[in a new window]
|
FIG. 4. Observed disease-specific survival (DSS) versus mean predicted DSS at 1, 2, and 3 years. The lines with error bars represent the performance of the MSKCC nomogram applied to the DUMC patients. The diagonal line represents the performance of an ideal nomogram.
|
|
In order to depict this same comparison in a more familiar way, a nomogram-predicted survival curve was generated as described in Methods and plotted against the Kaplan-Meier estimated survival curve for patients who received preop CRT (Fig. 5). Although the confidence intervals become wider by two years, the curves are nearly superimposable, suggesting that it is unlikely that a difference will emerge with longer follow-up.
View larger version (7K):
[in this window]
[in a new window]
|
FIG. 5. Observed disease-specific survival (solid line) with 95% confidence intervals (dotted lines) versus predicted disease-specific survival (dashed line) from the time of resection for patients who received preop CRT.
|
|
The nomogram calculates predicted survival based on several prognostic factors, which are weighted according to their effects in the multivariate model. To evaluate the significance of these individual prognostic factors specifically in patients who received preop CRT prior to resection, a multivariate model was constructed for this group using the same factors as the model on which the MSKCC nomogram was based (Table 2). Two of the most influential prognostic factors in the MSKCC population, tumor location and need for splenectomy, were not included, since patients with tumors in the body or tail were not included in this study. Another of the most influential prognostic factors in the MSKCC population, number of positive lymph nodes, was also highly predictive in the preop CRT group, but two of the other most influential prognostic factors, maximum path axis and tumor differentiation, were not predictive in the preop CRT group.
|
DISCUSSION
|
|---|
Non-randomized studies of neoadjuvant therapy for pancreatic cancer have suggested that patients who undergo preop CRT followed by resection may have superior survival to patients who undergo resection without preop CRT.5–7 Estimated five-year survival rates of greater than 30% have been reported for patients with potentially resectable tumors who complete preop CRT followed by resection.12,13 Although survival from the time of diagnosis is the endpoint that matters clinically, the use of this endpoint to compare patients who survived preop CRT before resection to patients who proceeded directly to resection introduces lead-time bias in favor of patients who received preop CRT. Other inherent biases are more like to favor patients who did not receive preop CRT. Patients with very small tumors, for whom it is difficult to establish a preoperative tissue diagnosis, are unlikely to receive neoadjuvant therapy. At most institutions, patients with tumors that are difficult to resect—due either to anatomical or medical reasons—are more likely to receive preoperative therapy than are patients with what appear to be easily resectable tumors. However, the most obvious problem with comparison between these groups of patients is that it is impossible to discern how much of this apparent improvement in survival is due to selection effect versus treatment effect. Patients who manifest distant metastatic disease during preop CRT had occult metastatic disease at the time of diagnosis and would not have benefited from resection. Presumably, patients who do not manifest distant metastatic disease during preop CRT are patients who have either less, or less aggressive, distant disease. If this is the predominant effect of preop CRT, observed survival from the time of resection in these patients should be better than predicted by a prognostic nomogram based on multiple clinical and pathological factors.
The possibility that the treatment being studied affects several of the most important factors in the nomogram limits its ability to evaluate the treatment. If there is minimal treatment effect of preop CRT, post-resection variables included in the nomogram should not be affected by preop CRT. Assuming preop CRT is selecting out at least some proportion of patients with occult metastatic disease, observed survival in resected patients should be better than predicted by the nomogram. However, if there is a significant treatment effect of preop CRT on post-resection variables such as tumor size, margin status, and lymph node status, which would favorably impact on nomogram-predicted survival, observed survival may be similar to, or possibly even worse than, predicted by the nomogram.
In this study, patients who received preop CRT followed by resection had better median DSS from diagnosis than did patients who underwent resection without preop CRT. The difference in post-resection DSS was not statistically significant, although follow-up was relatively immature in the group of patients who did not receive preop CRT. For the reasons described above, it is impossible to draw definite conclusions from these comparisons. At DUMC, neoadjuvant therapy is considered for all patients with localized tumors. As previously reported for this population, over 50% of patients with potentially resectable tumors and approximately 20% of patients with locally advanced tumors on initial staging CT subsequently undergo resection.6 As a consequence, the preop CRT group included 18 patients with locally advanced tumors, as defined by circumferential venous involvement or any arterial involvement. Despite the higher proportion of locally advanced tumors on initial staging, patients who received preop CRT had more favorable post-resection tumor characteristics, including tumor size, T-stage, and lymph node status, all factors that imply treatment effect. These observations alone could conceivably be attributed to preop CRT selecting more favorable tumors, while less favorable tumors progress. However, these observations together with the small percentage of patients in the preop CRT group who had complete pathologic responses seem to imply treatment effect.
The lack of significance of certain factors in the multivariate analysis for patients who received preop CRT may simply reflect the relatively small number of patients, particularly since tumor differentiation could not be determined in 19% of patients. However, given the smaller mean maximum path axis in the resection specimens of patients who received preop CRT, the lack of significance of this typically strong prognostic factor is provocative. It is possible that maximum path axis has been rendered unimportant as a prognostic factor by treatment effect on the primary tumor; it has been observed that irradiated specimens typically demonstrate replacement of tumor by fibrosis rather than shrinkage.7,13 Another explanation may be related to selection. In the MSKCC population, size did not have a monotonic effect on survival; tumor size beyond 4 cm had a beneficial effect.8 It is possible that only tumors with good biology are able to reach a large size without metastasis and that this same selection effect is enhanced by preop CRT.
|
CONCLUSIONS
|
|---|
In addition to providing information with which to counsel patients, nomograms are potentially useful tools for evaluating the effects of a treatment on a group of patients. By combining and weighting multiple prognostic factors, the nomogram more accurately predicts survival than current staging systems. The ideal nomogram would allow accurate prognostication based on factors available preoperatively and noninvasively; such factors are being actively sought. However, several of the most important prognostic factors currently available for pancreatic cancer, such as nodal status and resection margin status, are not accurately assessed prior to resection.
The MSKCC pancreatic cancer nomogram does accurately predict survival for patients who received preop CRT prior to resection and can potentially be used to counsel these patients postoperatively. However, the similarity between observed and predicted DSS allows us to conclude only that the effects of preop CRT—whether treatment, selection, or no effect—are reflected by factors within the nomogram. A nomogram based only on preoperative factors that are not potentially affected by preop CRT would be more helpful in evaluating the effect of preop CRT on survival, but the only way to assess the value of preoperative therapy definitively would be with a randomized controlled trial. Meanwhile, despite the suggestion of locoregional treatment effect, the impact of 5FU-based preop CRT regimens on survival appears to be modest. Given the high rates of distant recurrence seen after completion of resection and CRT in either order, we clearly need more effective systemic agents to realize the theoretical benefits of preoperative therapy.
Received for publication June 2, 2006.
Accepted for publication June 2, 2006.
|
REFERENCES
|
|---|
- Kalser MH, Ellenberg SS. Pancreatic cancer. Adjuvant combined radiation and chemotherapy following curative resection. Arch Surg 1985; 120:899–903.[Abstract/Free Full Text]
- Klinkenbijl JH, Jeekel J, Sahmoud T, et al. Adjuvant radiotherapy and 5-fluorouracil after curative resection of cancer of the pancreas and periampullary region: phase III trial of the EORTC gastrointestinal tract cancer cooperative group. Ann Surg 1999; 230:776–84.[CrossRef][Medline]
- Neoptolemos JP, Stocken DD, Friess H, et al. A randomized trial of chemoradiotherapy and chemotherapy after resection of pancreatic cancer. N Engl J Med 2004; 350:1200–10.[Abstract/Free Full Text]
- Yeo CJ, Abrams RA, Grochow LB, et al. Pancreaticoduodenectomy for pancreatic adenocarcinoma: postoperative adjuvant chemoradiation improves survival. A prospective, single-institution experience. Ann Surg 1997; 225:621–36.[CrossRef][Medline]
- Spitz FR, Abbruzzese JL, Lee JE, et al. Preoperative and postoperative chemoradiation strategies in patients treated with pancreaticoduodenectomy for adenocarcinoma of the pancreas. J Clin Oncol 1997; 15:928–37.[Abstract/Free Full Text]
- White RR, Hurwitz HI, Morse MA, et al. Neoadjuvant chemoradiation for localized adenocarcinoma of the pancreas. Ann Surg Oncol 2001; 8:758–65.[Abstract/Free Full Text]
- Hoffman JP, Lipsitz S, Pisansky T, et al. Benson AB, 3rd. Phase II trial of preoperative radiation therapy and chemotherapy for patients with localized, resectable adenocarcinoma of the pancreas: an Eastern Cooperative Oncology Group Study. J Clin Oncol 1998; 16:317–23.[Abstract/Free Full Text]
- Brennan MF, Kattan MW, Klimstra D, et al. Prognostic nomogram for patients undergoing resection for adenocarcinoma of the pancreas. Ann Surg 2004; 240:293–8.[Medline]
- Ferrone CR, Kattan MW, Tomlinson JS, et al. Validation of a postresection pancreatic adenocarcinoma nomogram for disease-specific survival. J Clin Oncol 2005; 23:7529–35.[Abstract/Free Full Text]
- Ghali WA, Quan H, Brant R, et al. Comparison of 2 methods for calculating adjusted survival curves from proportional hazards models. JAMA 2001; 286:1494–7.[Abstract/Free Full Text]
- Conlon KC, Klimstra DS, Brennan MF. Long-term survival after curative resection for pancreatic ductal adenocarcinoma. Clinicopathologic analysis of 5-year survivors. Ann Surg 1996; 223:273–9.[CrossRef][Medline]
- Breslin TM, Hess KR, Harbison DB, et al. Neoadjuvant chemoradiotherapy for adenocarcinoma of the pancreas: treatment variables and survival duration. Ann Surg Oncol 2001; 8:123–32.[Abstract/Free Full Text]
- White RR, Xie HB, Gottfried MR, et al. Significance of histological response to preoperative chemoradiotherapy for pancreatic cancer. Ann Surg Oncol 2005; 12:214–21.[Abstract/Free Full Text]
This article has been cited by other articles:
|
|
|
|
 
R. Le Scodan, F. Mornex, N. Girard, C. Mercier, P.-J. Valette, M. Ychou, F. Bibeau, P. Roy, J.-Y. Scoazec, and C. Partensky
Preoperative chemoradiation in potentially resectable pancreatic adenocarcinoma: feasibility, treatment effect evaluation and prognostic factors, analysis of the SFRO-FFCD 9704 trial and literature review
Ann. Onc.,
June 5, 2009;
(2009)
mdp015v1.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
F. L. Greene and L. H. Sobin
The Staging of Cancer: A Retrospective and Prospective Appraisal
CA Cancer J Clin,
May 1, 2008;
58(3):
180 - 190.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
A. Iasonos, D. Schrag, G. V. Raj, and K. S. Panageas
How To Build and Interpret a Nomogram for Cancer Prognosis
J. Clin. Oncol.,
March 10, 2008;
26(8):
1364 - 1370.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
