Locally Advanced Pancreatic Cancer: Association Between Prolonged Preoperative Treatment and Lymph-Node Negativity and Overall Survival

Importance  Treatment of patients with locally advanced/borderline resectable (LA/BR) pancreatic ductal adenocarcinoma (PDAC) is not standardized.

Objective  To (1) perform a detailed survival analysis of our institution’s experience with patients with LA/BR PDAC who were downstaged and underwent surgical resection and (2) identify prognostic biomarkers that may help to guide a decision for the use of adjuvant therapy in this patient subgroup.

Design, Setting, and Participants  Retrospective observational study of 49 consecutive patients from a single institution during 1992-2011 with American Joint Committee on Cancer stage III LA/BR PDAC who were initially unresectable, as determined by staging computed tomography and/or surgical exploration, and who were treated and then surgically resected.

Main Outcomes and Measures  Clinicopathologic variables and prognostic biomarkers SMAD4, S100A2, and microRNA-21 were correlated with survival by univariate and multivariate Cox proportional hazard modeling.

Results  All 49 patients were deemed initially unresectable owing to vascular involvement. After completing preoperative chemotherapy for a median of 7.1 months (range, 5.4-9.6 months), most (75.5%) underwent a pylorus-preserving Whipple operation; 3 patients (6.1%) had a vascular resection. Strikingly, 37 of 49 patients were lymph-node (LN) negative (75.5%) and 42 (85.7%) had negative margins; 45.8% of evaluable patients achieved a complete histopathologic (HP) response. The median overall survival (OS) was 40.1 months (range, 22.7-65.9 months). A univariate analysis of HP prognostic biomarkers revealed that perineural invasion (hazard ratio, 5.5; P = .007) and HP treatment response (hazard ratio, 9.0; P = .009) were most significant. Lymph-node involvement, as a marker of systemic disease, was also significant on univariate analysis (P = .05). Patients with no LN involvement had longer OS (44.4 vs 23.2 months, P = .04) than LN-positive patients. The candidate prognostic biomarkers, SMAD4 protein loss (P = .01) in tumor cells and microRNA-21 expression in the stroma (P = .05), also correlated with OS. On multivariate Cox proportional hazard modeling of HP and prognostic biomarkers, only SMAD4 protein loss was significant (hazard ratio, 9.3; P = .004).

Conclusions and Relevance  Our approach to patients with LA/BR PDAC, which includes prolonged preoperative chemotherapy, is associated with a high incidence of LN-negative disease and excellent OS. After surgical resection, HP treatment response, perineural invasion, and SMAD4 status should help determine who should receive adjuvant therapy in this select subset of patients.

The 5-year overall survival (OS) of patients with pancreatic ductal adenocarcinoma (PDAC) is 5.8% and has not improved during the past 10 years.^1,2 One of the main reasons for this poor prognosis is that more than 80% of patients present with either metastatic or locally advanced/borderline resectable (LA/BR) disease.¹ Those who have metastases at diagnosis are treated with definitive chemotherapy. Those with LA/BR disease have tumors that involve the local blood vessels³; these comprise about 40% of all patients.⁴ For more than 25 years,⁵ one approach has been to treat patients with LA/BR disease with preoperative downstaging therapy, with an overall goal of shrinking the tumor to enable complete surgical removal. In the most recent series, this approach has yielded survival rates for patients with LA/BR tumors that come close to those with early-stage resectable disease.⁶

Because of this recent change in management, downstaging treatment of patients with LA/BR disease is not standardized. It varies widely with respect to the duration and type of therapy.⁷ The most common practice has been to give 3 or 4 cycles of chemotherapy with or without radiation therapy during the course of 4 months and then reassess the tumor with repeat imaging.³ Patients whose tumors do not locally enlarge or develop metastatic disease are offered an attempt at surgical resection. Often times, these operations require vascular resections to achieve negative margins. This common approach has resulted in widely variable survival benefits from 17 to 33 months for resected patients with LA and BR disease, respectively.^8,9 This wide range suggests that a different and, potentially, uniform strategy may help to better select patients who would benefit from an operation and improve survival outcomes.

We previously reviewed our institution’s experience of patients with LA/BR PDAC.¹⁰ In this study, we focused on (1) the accuracy of computed tomography (CT)/magnetic resonance imaging in determining residual vessel involvement by tumor after downstaging treatment and (2) our selection criteria used to identify candidates who would benefit from surgical resection. We found that our approach involved a longer duration of downstaging therapy than is commonly used and that the median survival was excellent. Since that study, we have added another 15 patients with downstaged LA/BR disease who were resected. The objectives of this study in patients with downstaged LA/BR PDAC who underwent surgical resection were to (1) perform a detailed survival analysis of our series and (2) identify novel prognostic biomarkers to help identify patients with poor survival. That group might reasonably be expected to benefit from additional therapy given in an adjuvant setting.

Informed written consent was obtained from all patients in this study approved by the University of California, Los Angeles institutional review board. From 1992-2011, all patients at the University of California, Los Angeles Medical Center with a diagnosis of stage III biopsy-confirmed PDAC, who were initially deemed surgically unresectable either by cross-sectional imaging or exploratory laparotomy and who subsequently underwent a surgical resection after downstaging therapy, were included in this analysis. Fourty-nine patients met inclusion criteria. All available medical records (electronic and paper), including clinical, pathology, laboratory, and imaging reports, as well as daily treatment logs, were reviewed. For OS analyses, patients were censored at the date of their last confirmed clinic appointment; if they had no visits in the past year, the social security death index was queried to update their survival record. Vascular involvement prior to treatment was defined by contrast-enhanced CT (1 patient had a magnetic resonance image) and/or operative exploration. After treatment, vascular invasion was assessed intraoperatively by submitting thick tissue adherent to the superior mesenteric vein/portal vein for frozen-section analysis, which if positive for cancer was followed by a vascular resection. Margins were analyzed, and histopathologic (HP) grade and treatment response were scored using College of American Pathologists guidelines.¹¹ Recurrence was defined by imaging, including in some cases positron-emission tomography/CT and/or increasing CA19-9 levels. Local recurrence was defined as expanding soft-tissue thickening in the resection bed. Distant recurrence was defined as a new lesion not in the previous operative field and/or new lymphadenopathy.

Archival formalin-fixed, paraffin-embedded tumor samples were incubated at 60°C for 1 hour, deparaffinized in xylene, and rehydrated with graded alcohol washes. Antigen retrieval was performed by boiling in 0.01M sodium citrate buffer for 15 minutes, followed by quenching of endogenous peroxidase with 3% hydrogen peroxide. After blocking for 1 hour with 5% donkey serum in phosphate-buffered saline at room temperature, primary antibodies were added (SMAD4, 1:100; Santa Cruz Biotechnology Inc and S100A2, 1:50; DSHB) and incubated overnight at 4°C. Also, 1:250 biotin-conjugated antimouse secondary antibody (Jackson ImmunoResearch Laboratories Inc) was subsequently added and developed using Elite Vectastain ABC kit (Vector Laboratories). Staining of the tumor cells was then scored as positive or negative by 2 blinded reviewers. Any discrepancies were then reviewed by a third independent reviewer to establish a consensus.

On rehydration as just described, formalin-fixed, paraffin-embedded slides were washed 3 times with diethyl pyrocarbonate–treated phosphate-buffered saline, digested with 5-µg/mL proteinase K at 37°C for 30 minutes, washed, and then dehydrated in graded alcohol. Slides were hybridized at 55°C for 2 hours with 50 nmol–L alpha-linolenic acid–modified digoxigenin-labeled probes for microRNA-21 (miR-21, Exiqon). After stringency washes (5x, 1x, 0.2x saline-sodium citrate), slides were placed in blocking solution for 1 hour at room temperature, followed by overnight incubation at 4°C in alkaline phosphatase–conjugated antidigoxigenin Fab fragment solution. Antibody signal was amplified with nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl-phosphate (Roche) and then tissue was counterstained with Nuclear Fast Red (Vector). Scoring of the tumor cells and stroma were then completed as previously described.

SPSS version 20.0.0.1 (IBM) was used for statistical analysis and data management. Unless otherwise specified, continuous variables are reported as medians with interquartile ranges. The t test was used for comparison of means. Kaplan-Meier analysis was informed by the log-rank test. Univariate Cox proportional hazard models were used to calculate hazard ratios (HRs) of clinicopathologic factors with 95% CIs. Multivariate Cox regression analyses were performed in a stepwise fashion with backward selection of statistically significant univariate parameters using P < .10 as the initial entry criterion. Statistical significance was defined as P < .05.

All patients were determined unresectable owing to vascular involvement, predominantly the superior mesenteric vein/portal vein (30 of 46 [65%]) (Table 1). More than half had venous involvement alone (25 of 46 [54.3%]), 12 had arterial involvement alone, and 9 had a combination of arterial/venous involvement. Most tumors were located in the head of the pancreas (45 of 49 [91.8%]). The median duration of downstaging chemotherapy in this cohort was 7.1 months. Most patients (38 of 44 [86.4%]), received a multidrug regimen of gemcitabine- or 5-fluorouracil–based therapy. Most patients (38 of 49 [77.6%]) did not receive radiation. Downstaging therapy was associated with significant treatment effect as evidenced by a decrease in median tumor size (3.1 cm to 1.7 cm, P < .001) (Figure 1) and serum CA19-9 (96.3 U/mL to 20.0 U/mL, P = .006). Furthermore, 15 patients had apparent resolution of vascular involvement by CT after therapy. These data suggest that the patients included in this study were treated for longer than in other reported series,^3,6 and this highly select subgroup achieved a demonstrable clinical and radiographic treatment response.

Approximately half of patients (24 of 49 [49.0%]) initially underwent attempted resection that had to be aborted owing to apparent vascular invasion. Of those 24 patients who were deemed unresectable at this initial operation, 18 (75.0%) had a bypass procedure performed for the purpose of palliation. Importantly, all of these patients were subsequently reexplored and successfully resected after downstaging therapy. For the whole cohort, most underwent a pylorus-preserving Whipple resection (37 of 49 [75.5%]). Only a small number of tumors (3 of 49 [6.1%]) required vascular resection. Interestingly, only 1 of the 3 resected vein walls was involved with invasive carcinoma.

Pathologic review of this highly selected group of stage III patients with LA/BR PDAC revealed high rates of margin-negative resections (85.7%). The median number of lymph nodes (LNs) examined was 12 (interquartile range, 8.5-18) with 37 of 49 patients (75.5%) having LN-negative disease. Of those with positive margins, the uncinate margin was the most frequently involved (4 of 7 [57.1%]). Strikingly, in those patients who had a reported HP treatment response, approximately half (11 of 24 [45.8%]) were complete (pathologic complete response) with no viable cancer found in the excised specimen. On subgroup analysis, LN-negative patients received a longer duration of chemotherapy than LN-positive patients (mean [SD], 8.2 [4.8] months vs 6.2 [2.8] months), although this trend was not statistically significant. While these results can be explained purely on the basis of selection bias, it is also plausible that the longer duration of chemotherapy was eradicating disease in the LNs.

The median OS in this study was 40.1 months, with a disease-free survival of 23.2 months. Approximately half of patients had recurred by the last follow-up (24 of 49 [49.0%]), with distant being more common (13 of 24 [54.2%]) than local (3 of 24 [12.5%]) recurrences. Eight patients were assumed to have recurred, although the site had not been identified given increasing CA19-9 serum tumor markers and/or otherwise unexplained death. The overall 5-year survival was 42.9%. The observation that only 3 patients had evidence of local recurrence at last follow-up suggests that we effectively treated local disease.

Univariate analysis of clinicopathologic factors revealed LN invasion, perineural invasion (PNI), margin status, and HP response to be associated with survival (Table 2). Kaplan-Meier analyses are shown in Figure 2A-D. Decreases in serum CA19-9, tumor size, HP grade, and location of recurrence did not correlate with survival.

We then evaluated prognostic biomarkers that have been previously described in early-stage resectable PDAC to be associated with decreased survival: loss of SMAD4 protein expression,¹² S100A2 protein overexpression,¹³ and miR-21 overexpression in tumor cells¹⁴ and stroma.¹⁵ The percentage of patients who were positive for the poor prognostic markers, S100A2 and miR21 in tumor cells and stroma, was lower than in prior reports (Table 3). SMAD4 and miR-21 stromal expression correlated with survival on univariate analysis (Table 2). Controlling for significant prognostic factors (LN, PNI, margin, and miR-21 stromal expression), multivariate Cox regression analyses revealed SMAD4 to be the most significant factor associated with prognosis (P = .004). Survival curves for SMAD4 protein expression are depicted in Figure 2E. Together these data suggest that HP treatment response, PNI, and SMAD4 status are most closely correlated with survival for patients after downstaging therapy and surgical resection, and they could be used to help determine which of these patients receive optimal benefit from additional adjuvant therapy (Figure 3).

There are many treatment approaches used for patients with LA/BR PDAC, but they have in common several goals. Those goals are to (1) shrink the primary tumor to achieve a microscopic margin-negative resection, ideally without the need for a vascular resection, and (2) ensure that systemic metastases do not develop shortly after surgery that would nullify the potential benefit of a local resection. At the University of California, Los Angeles, our approach differs from most other centers that administer 3 to 4 cycles of chemotherapy or chemoradiotherapy over a course of 4 months prior to surgery. Because we believe that this is too short a time to meet the goals of downstaging treatment, we began in the early 1990s to treat our patients for a longer period. Therefore, we thought it would be useful to review our unique experience and examine the survival outcomes of the 49 patients with LA/BR PDAC who were downstaged and surgically resected. This analysis could further inform the evolving treatment approach for patients with this stage of disease.

The 49 patients in this study with initially unresectable disease owing to vascular involvement were treated with chemotherapy or chemoradiotherapy for a median duration of 7.1 months. Patients were considered appropriate for surgical resection if they maintained a good functional status, did not develop distant disease or an increasing CA19-9 level, and their primary tumor did not increase in size.¹⁰ These selection criteria yielded favorable outcomes, as the median OS was 40.1 months. Local disease control was successfully achieved in most patients. Excluding the 8 patients in whom the site of recurrence could not be identified, 38 of 41 (92.7%) had no evidence of local recurrence over the course of the study with a median follow-up of 4 years, and pathologic analysis revealed that 11 of 49 patients (22.4%) had no residual tumor after downstaging treatment.

At the time of surgical resection, 42 of 49 patients (85.7%) had their tumors resected with negative margins. Importantly, only 3 vascular resections (6%) were performed, and only 1 of these was found to have the vein wall involved with invasive cancer (2%). The number of patients in our cohort who required vein resection was lower than other series with similar disease-stage patients.⁹ This is an important distinction because histologic involvement of the vein wall is associated with a poor prognosis.¹⁶ These results suggest that using our prolonged preoperative downstaging treatment and selection criteria for surgery, a vein resection is not often required to achieve negative margins and adequate local control.

On pathologic analysis, 37 of 49 patients (75.5%) had no LNs involved with cancer. This rate is among the lowest reported for patients who underwent downstaging treatment.³ It is certainly lower than for patients with early-stage disease who undergo surgical resection without preoperative treatment. The largest reported series of 1000 patients at a single institution cites an incidence of 82% nodal involvement in resection specimens.¹⁷ While it is difficult to separate selection bias from treatment effect, we hypothesized that the chemotherapy is, in many cases, eradicating disease within the LNs. This hypothesis is supported by the observation that 2 patients in this series were initially diagnosed as having cancer via a peripancreatic LN biopsy. However, on pathologic analysis of the Whipple specimen, no LNs were found to be involved with cancer. Nodal positivity is particularly important because it is a marker of systemic disease and a strong predictor of survival.¹⁸ We observed a similar trend in our patients, as those with positive LNs died a median of 21.2 months sooner than patients with LN-negative disease.

There is a lack of information to guide which downstaged patients should be given adjuvant therapy after surgical resection. This decision was made more difficult as patients in our series already completed a median of 7.1 months of therapy prior to surgery. However, the decision could be aided by the development of clinicopathologic and molecular prognostic biomarkers for this specific subgroup. Prognostic biomarkers have been useful clinically to identify the minority of early-stage patients who have a favorable survival and are thus occasionally spared adjuvant treatment.¹⁹

Histopathologic treatment response and PNI were the 2 most significant clinicopathologic variables associated with survival on univariate analysis (HR, 9.0; P = .009 and HR, 5.5; P = .007, respectively). These findings are consistent with prior studies of patients with PDAC who have received neoadjuvant therapy.^20,21 While HP response was only reported in half of the cases, we can assume given the presence of tumor cells that others did not achieve a pathologic complete response. That gives a total pathologic complete response of 11 of 49 (22.4%), which compares favorably with a recently published large series of patients (6 of 223 [2.7%]).²⁰

To enhance the ability to predict survival, in addition to the clinicopathologic variables, we also investigated the prognostic significance of molecular biomarkers that have been examined in early-stage disease. We focused on SMAD4, S100A2, and miR-21 because each has been associated with a metastatic phenotype and therefore may provide a rationale for additional systemic therapy.^22-24 S100A2 has been validated as a predictor of longer survival after surgical resection.¹³ Both S100A2 and miR-21 in tumor cells were expressed in a lower percentage of patients in our series than in those with early-stage disease.^13,14,25 Furthermore, miR-21 stromal expression was only found in 12 of 32 of these patients (37.5%) in comparison with our tissue microarray of 153 patients with early-stage PDAC,²⁶ where expression was positive in 78.4% of patients.¹⁵ On univariate analysis, SMAD4 loss and miR-21 stromal expression correlated with survival (HR, 4.9; P = .01 and HR, 3.2; P = .04, respectively).

Combining the significant clinicopathologic and molecular markers on multivariate Cox regression analyses, SMAD4 loss was the most highly correlated with survival. We concluded that if these clinicopathologic and molecular markers were validated in an independent cohort, these observations could provide a rationale for or against adjuvant treatment in 2 seemingly similar patients who were both margin and LN negative. If one had SMAD4 protein loss, PNI, and minimal HP response, treatment would be recommended; while a patient with the opposite profile would be much less likely to benefit.

A significant limitation of this study was that it did not identify all patients who presented to our institution with LA/BR disease and who then underwent efforts to downstage their tumor. Therefore, we cannot precisely determine the fraction of patients successfully downstaged or, of equal importance, determine the clinical course of the nonresponders to treatment. The rates in the literature of patients with LA/BR disease who were successfully downstaged ranges from 10% to 66%.^9,27,28 Using our approach, we estimated that approximately 15% to 25% of our patients were successfully downstaged. Given the early metastatic predisposition of pancreatic cancer,^29,30 it can be reasoned that our longer duration of chemotherapy was more likely to identify patients with initially occult metastases that became clinically apparent during the prolonged treatment period. Thus, these patients did not undergo an operation. In support of this point, other series that included similarly staged patients reported a shorter progression-free survival duration after surgery.⁹

Most of the patients in this series received adjuvant therapy—many in a delayed fashion only once there was evidence of recurrence. Therefore, we were unable to make any assessment on the possible benefits of adjuvant therapy. Our data do suggest that the length of preoperative treatment actually impacts survival, but we cannot say this with certainty. Using standardized operative selection criteria, this could be determined by a randomized clinical trial comparing the duration of downstaging therapy.

Our approach to patients with LA/BR PDAC, which includes prolonged preoperative chemotherapy, is associated with a high incidence of LN-negative disease and excellent OS. Histopathologic treatment response, PNI, and SMAD4 status were significantly associated with OS and may help to determine which patients may benefit from additional adjuvant therapy in this select subgroup of patients.

Corresponding Author: Timothy R. Donahue, MD, Department of Surgery, 72-256 Center for the Health Sciences, University of California, Los Angeles, 10833 LeConte Ave, Los Angeles, CA 90095-6904 (tdonahue@mednet.ucla.edu).

Accepted for Publication: April 18, 2013.

Published Online: December 4, 2013. doi:10.1001/jamasurg.2013.2690.

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

Study concept and design: Kadera, Sunjaya, Isacoff, Tomlinson, Donald, Reber, Donahue.

Acquisition of data: Kadera, Sunjaya, Isacoff, Li, Dawson, Rochefort, Clerkin, Reber, Donahue.

Analysis and interpretation of data: Kadera, Sunjaya, Isacoff, Hines, Tomlinson, Dawson, Donald, Reber, Donahue.

Drafting of the manuscript: Kadera, Sunjaya, Li, Reber, Donahue.

Critical revision of the manuscript for important intellectual content: Kadera, Isacoff, Hines, Tomlinson, Dawson, Rochefort, Donald, Clerkin, Reber, Donahue.

Statistical analysis: Kadera, Sunjaya, Donald, Donahue.

Obtained funding: Kadera.

Administrative, technical, or material support: Kadera, Sunjaya, Isacoff, Li, Dawson, Donald, Clerkin, Reber, Donahue.

Study supervision: Kadera, Isacoff, Hines, Reber, Donahue.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the Gerald S. Levey Surgical Research Award and the David Geffen School of Medicine Short Term Training Program (University of California, Los Angeles). The monoclonal antibody developed by CPTC was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the Eunice Kennedy Shriver National Institute of Child Health and Human Development and maintained by the Department of Biology, University of Iowa, Iowa City.

Role of the Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Previous Presentation: This study was presented at the 84th Annual Meeting of the Pacific Coast Surgical Association; February 19, 2013; Kauai, Hawaii.