Association of Alterations in Main Driver Genes With Outcomes of Patients With Resected Pancreatic Ductal Adenocarcinoma

Key Points

Question  Are alterations in the 4 main driver genes for pancreatic adenocarcinoma associated with patient outcomes after pancreatic cancer resection?

Findings  In this study involving 356 patients with resected pancreatic adenocarcinoma, immunohistochemistry and next-generation DNA sequencing of formalin-fixed, paraffin-embedded pancreatic adenocarcinoma resection specimens identified alterations in the 4 main driver genes (KRAS, CDKN2A, SMAD4, and TP53). Disease-free survival and overall survival were associated with the presence and pattern of alterations in these 4 genes independent of previously identified prognostic factors.

Meaning  Identifying the pathogenic alterations in the 4 main driver genes of pancreatic adenocarcinoma informs patient outcomes.

Importance  Although patients with resected pancreatic adenocarcinoma are at high risk for disease recurrence, few biomarkers are available to inform patient outcomes.

Objective  To evaluate the alterations of the 4 main driver genes in pancreatic adenocarcinoma and patient outcomes after cancer resection.

Design, Setting, and Participants  This study analyzed protein expression and DNA alterations for the KRAS, CDKN2A, SMAD4, and TP53 genes by immunohistochemistry and next-generation sequencing in formalin-fixed, paraffin-embedded tumors in 356 patients with resected pancreatic adenocarcinoma who were treated at the Dana-Farber/Brigham and Women’s Cancer Center (October 26, 2002, to May 21, 2012), University of Rochester Medical Center (March 1, 2006, to November 1, 2013), or Stanford Cancer Institute (September 26, 1995, to May 22, 2013). Associations of driver gene alterations with disease-free survival (DFS) and overall survival (OS) were evaluated using Cox proportional hazards regression with estimation of hazard ratios (HRs) and 95% CIs and adjustment for age, sex, tumor characteristics, institution, and perioperative treatment. Data were collected September 9, 2012, to June 28, 2016, and analyzed December 17, 2016, to March 14, 2017.

Main Outcomes and Measures  The DFS and OS among patients with resected pancreatic adenocarcinoma.

Results  Of the 356 patients studied, 191 (53.7%) were men and 165 (46.3%) were women, with a median (interquartile range [IQR]) age of 67 (59.0-73.5) years. Patients with KRAS mutant tumors had worse DFS (median [IQR], 12.3 [6.7 -27.2] months) and OS (20.3 [11.3-38.3] months) compared with patients with KRAS wild-type tumors (DFS, 16.2 [8.9-30.5] months; OS, 38.6 [16.6-63.1] months) and had 5-year OS of 13.0% vs 30.2%. Particularly poor outcomes were identified in patients with KRAS G12D-mutant tumors, who had a median (IQR) OS of 15.3 (9.8-32.7) months. Patients whose tumors lacked CDKN2A expression had worse DFS (median, 11.5 [IQR, 6.2-24.5] months) and OS (19.7 [10.9-37.1] months) compared with patients who had intact CDKN2A (DFS, 14.8 [8.2-30.5] months; OS, 24.6 [14.1-44.6] months). The molecular status of SMAD4 was not associated with DFS or OS, whereas TP53 status was associated only with shorter DFS (HR, 1.33; 95% CI, 1.02-1.75; P = .04). Patients had worse DFS and OS if they had a greater number of altered driver genes. Compared with patients with 0 to 2 altered genes, those with 4 altered genes had worse DFS (HR, 1.79 [95% CI, 1.24-2.59; P = .002]) and OS (HR, 1.38 [95% CI, 0.98-1.94; P = .06]). Five-year OS was 18.4% for patients with 0 to 2 gene alterations, 14.1% for those with 3 alterations, and 8.2% for those with 4 alterations.

Conclusions and Relevance  Patient outcomes are associated with alterations of the 4 main driver genes in resected pancreatic adenocarcinoma.

Pancreatic cancer is the third leading cause of cancer-related death in the United States.¹ Large-scale genome sequencing studies have identified multiple molecular pathways involved in pancreatic adenocarcinoma initiation and progression.^2-4 Four main driver genes have been identified—KRAS (NCBI 3845), CDKN2A (NCBI 1029), SMAD4 (NCBI 4089), and TP53 (NCBI 7157)—that are critical for pancreatic cancer growth. The association of these driver gene alterations with patient outcomes has not been clearly established. Therefore, we characterized the status of these 4 driver genes using immunohistochemistry (IHC) and next-generation sequencing (NGS) of DNA in a large, highly annotated, multi-institutional patient population with resected pancreatic adenocarcinoma.

The study population consisted of 356 patients with resected pancreatic adenocarcinoma, of whom 126 were treated at the Dana-Farber/Brigham and Women’s Cancer Center between October 26, 2002, and May 21, 2012; 90 were treated at the University of Rochester Medical Center between March 1, 2006, and November 1, 2013; and 140 were treated at the Stanford Cancer Institute between September 26, 1995, and May 22, 2013. The institutional review board at each institution granted approval for this study. Patients treated at the Dana-Farber/Brigham and Women's Cancer Center signed written informed consent for participation in this study. Informed consent was waived by the University of Rochester Medical Center and the Stanford Cancer Institute as patients were identified retrospectively, according to institutional review board exempt protocols. Data were collected from September 9, 2012, to June 28, 2016. Data analysis took place from December 17, 2016, to March 14, 2017.

Immunohistochemistry for CDKN2A, SMAD4, and TP53 was performed on formalin-fixed, paraffin-embedded whole-tissue sections (eAppendix and eFigure 1 in the Supplement). After macroscopic dissection, genomic DNA was extracted from tumor tissue and adjacent normal tissue. Pyrosequencing for KRAS hotspot mutations and NGS using a customized, massively parallel sequencing panel were performed to determine the molecular status of KRAS, CDKN2A, SMAD4, and TP53 (eAppendix in the Supplement). For KRAS, tumors were classified as mutant or wild-type on the basis of NGS or pyrosequencing if predefined NGS coverage metrics were not met (eFigure 2 in the Supplement). For CDKN2A and SMAD4, tumors were classified as intact or lost on the basis of IHC results. For TP53, IHC and sequencing data were combined to make an integrated call as wild-type or altered (eAppendix in the Supplement).

Disease-free survival (DFS) was defined as time between surgery and disease recurrence, and overall survival (OS) was defined as time between surgery and death. Disease recurrence was classified as “local” when it occurred in or adjacent to the pancreatic remnant or in the retroperitoneum. Disease recurrence outside these sites was classified as “distant.” Follow-up continued through June 28, 2016, for the Dana-Farber/Brigham and Women’s Cancer Center; March 17, 2016, for the University of Rochester Medical Center; and March 11, 2016, for the Stanford Cancer Institute. A flow diagram of the study population is presented in eFigure 3 in the Supplement.

We evaluated the associations of driver gene alterations with DFS and OS using multivariable-adjusted Cox proportional hazards regression (eAppendix in the Supplement), calculating hazard ratios (HRs) and 95% CIs. We generated Kaplan-Meier curves, from which we calculated median survival, 2-year survival, and 5-year survival. In addition, we analyzed the association of gene alterations with pattern of first recurrence using multivariable-adjusted logistic regression, calculating odds ratios and 95% CIs. All hypothesis tests were 2-sided, and a 2-sided P < .05 indicated statistical significance.

Baseline characteristics of the study population by institution are shown in eTable 1 in the Supplement and by the 4 main driver genes are presented in eTable 2 in the Supplement. Of the 356 patients studied, 191 (53.7%) were men and 165 (46.3%) were women, with a median (interquartile range [IQR]) age of 67 (59.0-73.5) years. The median (IQR) DFS was 13.1 (7.0-27.8) months and OS was 21.0 (11.4-39.6) months, which are comparable to those in randomized trials.^5,6

Activating KRAS mutations were observed in 328 patients (92.1%) (eTable 3 in the Supplement); KRAS mutations affecting 2 separate codons were found in 11 tumors (3.4%). Patients who had KRAS mutant tumors had worse DFS (median [IQR], 12.3 [6.7-27.2] months) and OS (20.3 [11.3-38.3] months) compared with patients who had KRAS wild-type tumors (DFS, 16.2 [8.9-30.5] months; OS, 38.6 [16.6-63.1] months) (Table 1). Five-year OS rates were 13.0% for patients whose tumors were KRAS mutant and 30.2% for those with KRAS wild-type tumors. In addition, patients with KRAS G12D-mutant tumors had particularly poor outcomes, with worse DFS (median [IQR], 9.5 [4.7-17.6] months) and OS (15.3 [9.8-32.7] months) compared with patients with KRAS G12D wild-type tumors (DFS, 14.8 [7.9-32.8]; OS, 24.8 [15.0-46.2]) (Figure and Table 2; eFigure 4 in the Supplement).

By IHC, CDKN2A protein expression was lost in 240 patients (67.4%). Patients who had CDKN2A expression loss by IHC had worse DFS (median [IQR], 11.5 [6.2-24.5] months) and OS (19.7 [10.9-37.1] months) compared with patients with intact CDKN2A (DFS, 14.8 [8.2-30.5] months; OS, 24.6 [14.1-44.6] months) (Table 1). In sensitivity analyses, we classified CDKN2A status using IHC and sequencing data (eTable 4 in the Supplement). Loss of CDKN2A expression by IHC was associated with worse DFS and OS regardless of CDKN2A molecular status (eTable 5 in the Supplement), which likely reflects the inability of NGS to detect the silencing of CDKN2A expression due to promoter methylation and reduced sensitivity for copy number loss in low-cellularity tumors.

By IHC, SMAD4 protein expression was lost in 175 patients (49.2%). Loss of SMAD4 expression was not significantly associated with DFS or OS in our patient population (Table 1). We used our sequencing data to predict whether SMAD4 expression would be present or lost (eTable 4 in the Supplement). The molecular status of SMAD4 was not associated with DFS or OS when SMAD4 expression was lost by IHC (eTable 5 in the Supplement).

For TP53, we used IHC and molecular data to generate an integrated call of TP53 as wild-type or altered (eTable 4 in the Supplement). By this approach, TP53 was altered in 231 patients (64.9%). Altered TP53 was associated with shorter DFS (HR, 1.33; 95% CI, 1.02-1.75; P = .04) but was not associated with OS (HR, 1.18; 95% CI, 0.91-1.53; P = .23) (Table 1).

Twenty-four patients (6.7%) received preoperative therapy, which was not associated with the pattern of driver gene alterations (eTable 6 in the Supplement). In sensitivity analyses excluding these 24 patients, associations between driver gene alterations and patient outcomes were largely unchanged (eTable 7 in the Supplement).

We analyzed the association of combinatorial gene alterations with DFS and OS (eTable 8 in the Supplement). Compared with patients with 0 to 2 gene alterations, patients with 3 or 4 gene alterations had worse DFS (3 alterations HR, 1.37 [95% CI, 1.01-1.86; P = .05]; 4 alterations HR, 1.79 [95% CI, 1.24-2.59; P = .002]) and OS (3 alterations HR, 1.22 [95% CI, 0.91-1.64; P = .18]; 4 alterations HR, 1.38 [95% CI, 0.98-1.94; P = .06]) (Table 2 and Figure; eFigure 4 in the Supplement). The worst outcomes were identified in patients with both KRAS mutant tumors and CDKN2A expression loss. Five-year OS rates were 18.4% for patients with 0 to 2 gene alterations, 14.1% for patients with 3 gene alterations, and 8.2% for patients with 4 gene alterations. In our patient population, alterations in the 4 driver genes were not significantly associated with local recurrence as the first site of disease recurrence (eTable 9 in the Supplement).

In a large, multi-institutional population of patients with resected pancreatic adenocarcinoma, patient outcomes were associated with alterations of the 4 main driver genes. Previous studies have assessed these genes and patient outcomes individually using a variety of methods and patient populations and revealing inconsistent results.^7-10 One study assessed all 4 driver genes among 79 patients who underwent rapid autopsy after death from pancreatic adenocarcinoma.¹¹ Tumors were sequenced by polymerase chain reaction for KRAS, CDKN2A, and TP53, and IHC was performed for CDKN2A, SMAD4, and TP53. Although the sample size was small and included all stages of disease, patients whose tumors had 3 or 4 altered genes had worse DFS and OS than did patients whose tumors had 1 or 2 altered genes in unadjusted analysis. The primary results of the current study are confirmatory in a large, multi-institutional patient population, but multiple-hypothesis testing should be acknowledged when interpreting data across several biomarkers.

Previous studies have suggested that loss of SMAD4 protein expression by IHC was associated with extensive metastatic spread, generating interest in SMAD4 staining as an informative biomarker to guide the use of radiotherapy.¹² However, a subsequent study of 127 patients with resected pancreatic cancer did not replicate these findings.¹³ In our study population, SMAD4 staining was not associated with pattern of disease recurrence after surgical resection.

Adjuvant treatment following surgical resection of pancreatic cancer improves patient survival, but outcomes remain suboptimal.⁵ With the intent of improving cure rates, novel and more aggressive multiagent treatment programs are currently being devised and evaluated in the adjuvant setting.⁶ Furthermore, increasing numbers of patients are receiving chemotherapy and radiation before surgical resection to rapidly initiate therapy against micrometastatic disease and select out those patients with early disease progression unlikely to benefit from surgery.¹⁴ In the future, molecular assessment of pancreatic cancer may help guide the use and components of perioperative treatment programs.¹⁵

This study demonstrates that alterations in the 4 main driver genes are associated with patient outcomes in a large, multi-institutional population of patients with resected pancreatic adenocarcinoma. Understanding the molecular events that determine patient outcomes has the potential to improve treatment approaches for patients with this aggressive malignancy.

Corresponding Author: Brian M. Wolpin, MD, MPH, Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, 450 Brookline Ave, Boston, MA 02215 (bwolpin@partners.org).

Published Online: November 2, 2017. doi:10.1001/jamaoncol.2017.3420

Correction: This article was corrected on March 7, 2019, to correct omissions in the Conflict of Interest Disclosures.

Accepted for Publication: August 1, 2017.

Author Contributions: Drs Qian, Rubinson, Nowak, and Morales-Oyarvide share first coauthorship. Drs Hezel, Koong, and Wolpin share last coauthorship. Drs Morales-Oyarvide and Wolpin had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Qian, Rubinson, Nowak, Hahn, Meyerson, Ogino, Koong, Wolpin.

Acquisition, analysis, or interpretation of data: Qian, Rubinson, Nowak, Morales-Oyarvide, Dunne, Kozak, Welch, Brais, Da Silva, T. Li, W. Li, Masuda, Yang, Shi, Gu, Masugi, Bui, Zellers, Yuan, Babic, Khalaf, Aguirre, Ng, Miksad, Bullock, Chang, Tseng, Clancy, Linehan, Findeis-Hosey, Doyle, Thorner, Ducar, Wollison, Laing, Fuchs, Ogino, Hornick, Hezel, Koong, Wolpin.

Drafting of the manuscript: Qian, Rubinson, Nowak, Morales-Oyarvide, Welch, T. Li, Shi, Gu, Bui, Zellers, Doyle, Thorner, Laing, Wolpin.

Critical revision of the manuscript for important intellectual content: Qian, Rubinson, Nowak, Morales-Oyarvide, Dunne, Kozak, Brais, Da Silva, W. Li, Masuda, Yang, Shi, Gu, Masugi, Yuan, Babic, Khalaf, Aguirre, Ng, Miksad, Bullock, Chang, Tseng, Clancy, Linehan, Findeis-Hosey, Doyle, Thorner, Ducar, Wollison, Hahn, Meyerson, Fuchs, Ogino, Hornick, Hezel, Koong, Wolpin.

Statistical analysis: Morales-Oyarvide, T. Li, Yuan, Ng, Wollison, Wolpin.

Obtained funding: Rubinson, Fuchs, Ogino, Koong, Wolpin.

Administrative, technical, or material support: Nowak, Morales-Oyarvide, Dunne, Welch, Brais, Da Silva, W. Li, Masuda, Yang, Gu, Masugi, Bui, Zellers, Babic, Miksad, Bullock, Chang, Clancy, Linehan, Doyle, Thorner, Ducar, Hahn, Fuchs, Hornick, Hezel, Koong.

Study supervision: Qian, Rubinson, Nowak, Tseng, Doyle, Thorner, Ducar, Hahn, Meyerson, Fuchs, Ogino, Hezel, Koong, Wolpin.

Conflict of Interest Disclosures: Dr Fuchs has been a consultant and/or a scientific advisor for Eli Lilly, Entrinsic Health, Pfizer, Merck, Sanofi, Roche, Genentech, Merrimack Pharmaceuticals, Dicerna, Bayer, Celgene, Agios, Gilead Sciences, Five Prime Therapeutics, Taiho, and KEW. No other disclosures were reported.

Funding/Support: This study was funded in part by a grant from the Department of Medical Oncology translational research grant program of the Dana-Farber Cancer Institute (Dr Qian); grant K07 CA148894 from the National Institutes of Health (Dr Ng); grant R35 CA197735 from the National Cancer Institute (Dr Ogino); grants from the Hale Center for Pancreatic Cancer Research, the Perry S. Levy Fund for Gastrointestinal Cancer Research, and the Pappas Family Research Fund for Pancreatic Cancer, as well as grants R01 CA124908 and P50 CA127003 from the National Institutes of Health (Dr Fuchs); funding from MyBlueDots (Dr Koong); and grants from the Hale Center for Pancreatic Cancer Research, the Lustgarten Foundation for Pancreatic Cancer Research, the Pancreatic Cancer Action Network, the Noble Effort Fund, the Peter R. Leavitt Family Fund, the Wexler Family Fund, and Promises for Purple, as well as grant CA130288 from the US Department of Defense and grant U01 CA210171 from the National Institutes of Health/National Cancer Institute (Dr Wolpin).

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.

Additional Contributions: These individuals assisted with designing the targeted sequencing panel: Nabeel Bardeesy, PhD, Massachusetts General Hospital and Harvard Medical School; Alec Kimmelman, MD, PhD, NYU Langone Medical Center; R. Coleman Lindsley, MD, PhD, and Mandar Muzumdar, MD, Dana-Farber Cancer Institute and Harvard Medical School; Eliezer Van Allen, MD, Dana-Farber Cancer Institute and Broad Institute of Massachusetts Institute of Technology and Harvard; and Matthew Vander Heiden, MD, PhD, Koch Institute for Integrative Cancer Research of the Massachusetts Institute of Technology. These individuals were not compensated for their contribution.