Association of Clinical Factors With a Major Pathologic Response Following Preoperative Therapy for Pancreatic Ductal Adenocarcinoma

Key Points

Question  What clinical factors are associated with a major pathologic response following preoperative therapy for pancreatic ductal adenocarcinoma?

Findings  In this study of 583 patients with histopathologically confirmed pancreatic ductal adenocarcinoma who were treated with preoperative therapy, those with a pathologic complete response or less than 5% viable cancer cells had a significantly longer median survival duration (73.4 months vs 32.2 months). On multivariable logistic regression analysis, young age, low baseline cancer antigen 19-9 level, and the use of gemcitabine as a radiosensitizer were associated with a major pathologic response.

Meaning  The patient, treatment, and tumor-related factors identified in this study may define a group of patients most likely to experience a significant response to preoperative therapy.

Importance  We previously demonstrated that a major pathologic response to preoperative therapy, defined histopathologically by the presence of less than 5% viable cancer cells in the surgical specimen, is an important prognostic factor for patients with pancreatic ductal adenocarcinoma. However, to our knowledge, the patients most likely to experience a significant response to therapy are undefined.

Objective  To identify clinical factors associated with major pathologic response in a large cohort of patients who underwent preoperative therapy and pancreatectomy for pancreatic ductal adenocarcinoma.

Design, Setting, and Participants  Retrospective review of a prospectively maintained database at University of Texas MD Anderson Cancer Center. The study included 583 patients with histopathologically confirmed pancreatic ductal adenocarcinoma who received preoperative therapy prior to pancreatectomy between 1990 and 2015.

Exposures  Preoperative therapy consisted of systemic chemotherapy alone (n = 38; 6.5%), chemoradiation alone (n = 261; 44.8%), or both (n = 284; 48.7%) prior to pancreatoduodenectomy (n = 514; 88.2%), distal pancreatectomy (n = 62; 10.6%), or total pancreatectomy (n = 7; 1.2%).

Main Outcomes and Measures  Clinical variables associated with a major pathologic response (pathologic complete response or <5% residual cancer cells) were evaluated using logistic regression.

Results  Among all patients, the mean (SD) age was 63.7 (9.2) years, and 53.0% were men. A major pathologic response was seen in 77 patients (13.2%) including 23 (3.9%) who had a complete pathologic response. The median overall survival duration was significantly longer for patients who had a major response than for those who did not (73.4 months vs 32.2 months, P < .001). On multivariate logistic regression, only age younger than 50 years, baseline serum cancer antigen 19-9 level less than 200 U/mL, and gemcitabine as a radiosensitizer were associated with a major response. The number of these positive factors was associated with the likelihood of a major response in a stepwise fashion (0, 7.5%; 1, 12.7%; 2, 16.9%; 3, 35.7%; P = .009).

Conclusions and Relevance  Although a major pathologic response occurs infrequently following preoperative therapy for pancreatic ductal adenocarcinoma, it is associated with a significantly improved prognosis. Of the patient- and treatment-related factors we analyzed, only young age, low baseline cancer antigen 19-9, and gemcitabine as a radiosensitizer were associated with a major pathologic response. Given its association with long-term survival, better predictors of response and more effective preoperative regimens should be aggressively sought.

Recent practice guidelines have recognized the administration of chemotherapy and/or radiation therapy prior to pancreatectomy for localized pancreatic ductal adenocarcinoma (PDAC) as the preferred treatment strategy for patients with borderline resectable cancer¹ and an acceptable treatment option for patients with potentially resectable cancer.² Purported benefits of this approach include the selection of patients with favorable tumor biology and a physiologic profile appropriate for major surgery, early treatment of micrometastatic disease, facilitation of a margin-negative resection, and guaranteed delivery of all components of multimodality therapy. During the past 3 decades, refinement of the therapeutic regimens administered in the preoperative setting has permitted the rational use of potentially curative surgery in patients with anatomically advanced cancer and has contributed to a significant increase in overall survival duration following pancreatectomy.³

The response of the primary tumor to preoperative therapy, measured histologically by the extent of residual carcinoma in the resected surgical specimen, represents an important prognostic factor for patients with PDAC,^4-7 just as it does for patients with cancers of the breast, rectum, and esophagus.^8-10 However, a pathologic complete response (pCR) occurs infrequently in PDAC relative to these other cancers. We previously identified only 6 of 223 patients (2.7%) treated with preoperative therapy and pancreatoduodenectomy who experienced a pCR, while another 36 patients (16.1%) had minimal residual tumor (ie, <5% viable cells) in the surgical specimen.⁵ Patients who had either a pCR or less than 5% residual cancer cells had a significantly longer survival duration than patients in whom the response to therapy was less robust (ie, 5%-100% residual cancer cells).

A complete understanding of the factors associated with a significant response to preoperative therapy would have profound clinical importance. These factors might suggest the underlying mechanisms by which treatment effect occurs. More practically, the presence of these factors might identify a group of patients for whom a multimodality strategy using preoperative therapy would be particularly appropriate. Several clinical features have been found to be predictive of pCR in patients with breast, rectal, and esophageal cancer.^11-13 However, to our knowledge, factors associated with a significant treatment response have not been defined in PDAC.

In this study, we sought to identify clinical factors associated with a major pathologic response (pCR or <5% residual viable tumor cells) in a large cohort of patients who underwent preoperative therapy and pancreatectomy for PDAC at the University of Texas MD Anderson Cancer Center.

The institutional review board of the University of Texas MD Anderson Cancer Center, Houston, approved this retrospective study. Individual informed consent was waived. We used a prospectively maintained institutional pancreatic tumor database¹⁴ to identify all patients who received preoperative chemotherapy and/or chemoradiation (CRT) prior to pancreatectomy for PDAC between January 1990 and December 2015. Patients who received preoperative therapy prior to referral (n = 76) and those in whom histopathologic treatment effect was not recorded (n = 111) were excluded.

Prior to the initiation of preoperative therapy, all patients underwent comprehensive clinical and radiographic staging, including cross-sectional imaging of the abdomen and pelvis, a chest radiograph or chest computed tomography/magnetic resonance imaging scan, and laboratory studies, including serum cancer antigen (CA) 19-9 measurement. The most recent CA 19-9 level prior to initiation of preoperative therapy was used for this analysis. Tumor anatomy was classified as potentially resectable, borderline resectable, or locally advanced according to standardized radiographic criteria.¹⁵ A histopathologic diagnosis of PDAC was required for initiation of preoperative therapy.

Decisions regarding preoperative therapy were made as part of a multidisciplinary program, and treatments were administered either on or off protocol. Several preoperative treatment regimens were used during the study period.³ External-beam radiation therapy with concurrent 5-fluorouracil, capecitabine, or gemcitabine was generally delivered to a total of 50.4 Gy over 6 weeks or to a total of 30 Gy over 2 weeks. When both systemic chemotherapy and CRT were used, systemic chemotherapy was administered prior to CRT.

Within 4 to 8 weeks of completing all intended preoperative therapy, patients underwent repeated clinical and radiographic staging. Those without evidence of disease progression and with adequate performance status were considered for surgical resection. Pancreatoduodenectomy, distal pancreatectomy, or total pancreatectomy was performed according to standardized techniques.¹⁶

All surgical specimens were evaluated by dedicated gastrointestinal pathologists using a standardized protocol.¹⁷ R1 margin status was defined as evidence of cancer cells present at the inked bile duct, pancreatic neck, or superior mesenteric artery margin. The histopathologic response to preoperative therapy was measured as the percentage of residual viable cancer cells within the tumor.^5,18 Based on our prior work,⁵ a pCR (no residual viable cancer cells) or less than 5% residual viable cancer cells were defined as evidence of a major pathologic response.

Following surgical resection, adjuvant therapy was administered selectively based on individual patient and tumor characteristics. Patients were typically evaluated initially every 3 to 4 months and then later every 6 months, with cross-sectional imaging, physical examination, and CA 19-9 analysis according to a standardized protocol.¹⁹ Locoregional recurrence was defined as the development of a new low-density mass in the region of the resected pancreas or new lymphadenopathy localized to the root of the mesentery occurring as a component of first failure following surgery.

Clinical, demographic, and pathologic variables were compared among patients with less than 5% and at least 5% residual cancer cells. Categorical variables were compared using Pearson χ², while continuous variables were compared using the Mann-Whitney U test. Next, demographic and clinical factors were investigated for their association with clinical response using univariate and multivariate logistic regression modeling. Factors included in the final model were determined using backward conditional regression. Finally, overall survival was calculated from the date of diagnosis to the date of death or last follow-up using the Kaplan-Meier method and compared between the 2 groups using the Mantel-Cox log-rank test. Curves were created using GraphPad Prism 6.0 (GraphPad Software Inc), and statistical analyses were performed using SPSS, version 24.0 (SPSS Inc), with statistical significance established at P < .05. The P value was 2-sided.

A total of 583 patients with localized PDAC received preoperative chemotherapy (n = 38; 6.5%), CRT (n = 261; 44.8%), or both (n = 284; 48.7%) prior to pancreatectomy between 1990 and 2015 and had their tumor specimens graded for treatment effect. Most patients (n = 514; 88.2%) underwent pancreatoduodenectomy.

A major pathologic response to preoperative therapy was observed in the surgical specimens of 77 patients (13.2%); among them, 23 (3.9%) had no viable cells, and 54 (9.3%) had less than 5% viable cells. The specimens of the remaining 506 patients (86.8%) had at least 5% viable cells (Table 1). Patients who had a major pathologic response were younger (mean [SD] age, 61.5 [10.6] years vs 64.1 [9.0] years; P = .02) and more likely to present with a CA 19-9 level less than 200 U/mL (70.1% vs 56.9%, P = .03) than patients who did not, but the groups were similar in terms of the regimens they had received and their mean duration of therapy. Patients who had a major response were less likely to have undergone vascular resection than patients who did not, and their tumors were smaller and less likely to have positive margins, positive lymph nodes, lymphovascular invasion, or perineural invasion (Table 1). Patients with a major response were also less likely to have been treated with postoperative chemotherapy (18.2% vs 37.0%, P = .001).

On univariate logistic regression, younger age, lower baseline CA 19-9 level, and the use of gemcitabine as a radiosensitizer were associated with a significant treatment effect (Table 2). No other treatment-related variables, including type of chemotherapy, dose of radiation, or duration of preoperative therapy, were associated with treatment effect. On multivariate logistic regression, age younger than 50 years, CA 19-9 level less than 200 U/mL, and gemcitabine as a radiosensitizer were associated with major response (Table 2). The number of predictive factors was associated with the rate of a major pathologic response in a stepwise fashion (0, 7.5%; 1, 12.7%; 2, 16.9%; 3, 35.7%; P = .009, Figure 1).

The median overall survival duration and 5-year overall survival rate of patients who had a major pathologic response were both significantly better (73.4 months and 58.2%, respectively) than those of patients who did not have a major response (32.2 months and 28.8%, respectively) (P < .001; Figure 2). In contrast, there was no significant difference in locoregional recurrence between the groups (11.7% vs 19.6%, P = .08).

Table 3 reports the clinical profile of the 23 patients who had a pCR (0% viable cancer cells). None of these patients developed locoregional recurrence, but 4 of them (17.4%) developed distant recurrence: 2 in the lungs and 2 in the liver. In contrast, 9 of the 54 patients (16.7%) who had a major pathologic response but still had up to 5% residual cancer cells experienced locoregional recurrence as a component of first failure.

We have previously shown that patients with PDAC who receive preoperative chemotherapy and/or CRT and have less than 5% residual viable cancer cells have a favorable prognosis relative to patients who have at least 5% cells in their pancreatectomy specimens.⁵ In this study, we specifically sought to identify predictors of this response to therapy. Using a large data set of patients who received preoperative therapy during a 25-year period, we calculated a major pathologic response rate of only 13.2% and a pCR rate of only 3.9%. The median overall survival duration of patients who had a major response was 73.4 months. Importantly, we identified a patient-related factor (age), tumor-related factor (CA 19-9 level), and treatment-related factor (radiosensitizer) that were each independently associated with development of a major pathologic response.

The results of this study highlight the relative rarity of a pCR among patients with PDAC treated with existing preoperative therapy regimens. Indeed, the pCR rate of 3.9% reported here, remarkably similar to the 3.6% published in a 2010 meta-analysis of patients with PDAC,²⁰ is striking when compared with rates in other cancers. For example, preoperative chemotherapy for breast cancer results in pCR rates ranging from 15% to 60% depending on the cancer subtype and the type of chemotherapy administered.^8,21-23 Pathologic complete response is observed following preoperative CRT for rectal cancer in 15% to 27% of patients,⁹ and several clinical features and biomarkers are known to be predictive.^12,24 Preoperative CRT for esophageal cancer led to an overall pCR rate of 29% in the CROSS trial,²⁵ and the rate was higher in squamous cell carcinoma than in adenocarcinoma. Why pCR is substantially rarer in patients with PDAC is unknown, but at a minimum, this discrepancy points to the need for more effective therapeutics against this disease.

Whether a patient’s response to therapy is related to the underlying biology of the tumor, the cytotoxic efficacy of the treatment used against it, or an immunomodulating ability of its host remains unknown. Our findings suggest that response may be determined by all 3 of these factors. The association between clinical response and young age hints at the possibility of a genetic influence because both somatic and germline mutation status have been associated with pCR in other cancers. For example, breast cancers in patients with germline BRCA mutations demonstrate greater chemosensitivity^26,27 and are associated with higher rates of pCR following neoadjuvant therapy,²⁸ especially when platinum-based therapies are used.²⁹ In PDAC, BRCA carriers have been found to be more sensitive to chemotherapies that lead to DNA damage, which is inefficiently repaired by cells lacking functional BRCA1 or BRCA2.^30,31 Further research into the genetic and molecular influences on clinical and pathologic response to chemotherapy in PDAC is clearly needed.

Equally notable here was the limited number of treatment-related factors that were associated with clinical response. No chemotherapy regimen was found to be more effective than others in this regard nor was a longer treatment interval. Although the data set we used is likely limited in its power to analyze the association between CRT and response because almost all patients received CRT, the trend toward a higher rate of response in patients treated with radiation suggests its possible importance. Indeed, in this study, only 1 major pathologic response was observed in patients who received preoperative chemotherapy alone. Furthermore, only a handful of case reports describing pCR in response to chemotherapy exist in the literature,^32-35 and other studies have also reported higher response rates after CRT than after chemotherapy.^36,37 The possible importance of the dose-related ablative effects of radiation on the primary tumor is also suggested by the relatively high rates of response (9%-16%) reported with stereotactic body radiation, which enables higher doses of radiation to be delivered.³⁸ Finally, our results also suggest that the chemotherapeutic agent used to sensitize the tumor to radiation may also be important. We had previously found that the radiosensitizer gemcitabine was associated with improved survival³⁶; this study suggests that an enhanced pathologic response may be the mechanism through which the improved outcome occurs.

That 5-fluorouracil, leucovorin, oxaliplatin, irinotecan (FOLFIRINOX) was not associated with a higher incidence of a major pathologic response might be surprising given its apparent association with improved resectability rates³⁹ and longer survival durations.⁴⁰ However, a 2016 systematic review of studies evaluating FOLFIRINOX for locally advanced PDAC reported a pooled pCR rate of only 7.1% in patients who underwent pancreatectomy,⁴¹ a rate identical to that reported here (4 of 56 patients who received FOLFIRINOX). Furthermore, it is worth noting that prior single-institution reports of patients who underwent pancreatectomy for locally advanced PDAC were likely enriched with patients who experienced a favorable response to preoperative therapy; in contrast, most patients herein were technically resectable at baseline and did not require a major response to become resectable.

It may also be surprising that only 34.5% of patients (n = 201 of 583) in this series received chemotherapy or chemoradiation following pancreatectomy. However, it must be recognized that the role of postoperative therapy for patients who have already received preoperative therapy remains poorly defined. We have historically administered postoperative therapy selectively in this setting, but in a 2015 review of our data, we found that the administration of systemic chemotherapy following preoperative therapy and pancreatectomy was associated with improved survival relative to observation.⁴² Whether this apparent benefit applies to patients who have had a major pathologic response or a pCR to preoperative therapy is unclear.

One question that remains is whether response of the primary tumor serves as an effective surrogate for response of the micrometastatic disease presumed present in almost all patients with PDAC. Our findings would suggest not entirely: although clinical response measured in the primary tumor was associated with improved survival, it did not guarantee cure. For example, one patient who had a pCR in the primary tumor specimen had viable cancer in a regional lymph node, and many patients who had a major pathologic response, including 17% of patients who had a pCR, developed distant recurrence following pancreatectomy. This phenomenon may be partially explained by differences in treatment effect induced by radiation (which primarily works locally) as opposed to treatment effect induced by chemotherapy (which has a systemic effect). Regardless, given the association between major pathologic response and long-term outcomes, such response may serve as a clinically relevant early end point in clinical trials.

The main limitation of our study is its retrospective and single-institution design. Specifically, the results of this study may be influenced by our bias toward the preoperative administration of chemoradiation, although the role of radiation in the treatment of localized PDAC remains controversial.^43-45 On the other hand, the strengths of the study are its relatively large sample size, the evaluation of a wide range of treatment regimens that allow for comparisons, use of a prospectively maintained database, and the availability of granular data to evaluate treatment-level factors and their association with a novel end point.

In summary, although pCR is a relatively rare outcome following preoperative therapy for PDAC, it is associated with a significantly improved prognosis. However, of the patient- and treatment-related factors we analyzed, only young age (patient), low baseline CA 19-9 level (biology), and gemcitabine as a radiosensitizer (treatment) were associated with a major pathologic response. Future studies should investigate the mechanisms by which pCR develops and focus on the identification of new therapeutic agents with greater cytotoxic potential.

Corresponding Author: Matthew H. G. Katz, MD, Department of Surgical Oncology, The University of Texas MD Anderson Cancer Center, 1400 Pressler St, FCT 17.6058, Houston, TX 77030-4008 (mhgkatz@mdanderson.org).

Accepted for Publication: April 17, 2017.

Published Online: July 12, 2017. doi:10.1001/jamasurg.2017.2227

Author Contributions: Drs Cloyd and Katz had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Cloyd, Maitra, Varadhachary, Javle, Overman, Herman, Vauthey, Aloia, Katz.

Acquisition, analysis, or interpretation of data: Cloyd, Wang, Egger, Tzeng, Prakash, Shroff, Fogelman, Wolff, Overman, Koay, Das, Kim, Aloia, Fleming, Lee, Katz.

Drafting of the manuscript: Cloyd, Tzeng, Javle, Koay, Vauthey, Aloia, Katz.

Critical revision of the manuscript for important intellectual content: Cloyd, Wang, Egger, Tzeng, Prakash, Maitra, Varadhachary, Shroff, Fogelman, Wolff, Overman, Koay, Das, Herman, Kim, Vauthey, Aloia, Fleming, Lee, Katz.

Statistical analysis: Cloyd, Overman, Koay.

Administrative, technical, or material support: Wang, Prakash, Varadhachary, Koay, Kim, Vauthey, Aloia, Fleming, Katz.

Supervision: Tzeng, Maitra, Javle, Fogelman, Overman, Herman, Vauthey, Fleming, Lee, Katz.

Conflict of Interest Disclosures: None reported.

Funding/Support: Supported by grant P30CA016672 from the National Institutes of Health/National Cancer Institute and the Clinical Trials Support Resource.

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