Molecular profiling is being explored in pancreatic adenocarcinoma as a tool to assist with early detection, prognosis, and patient selection in targeted therapy clinical trials.
Most cases of pancreatic adenocarcinoma (PDAC) are diagnosed in the metastatic or locally advanced stage. It is the fourth leading cause of cancer death in the United States,1,2 with a 5-year overall survival (OS) around 10% in this country2 despite years of research and therapeutic development. For those patients with unresectable or refractory PDAC, locoregional and systemic chemotherapies remain the main treatment options outside of a clinical trial. In December 2019, the FDA approved the first targeted therapy, olaparib (Lynparza), in the maintenance setting for individuals with metastatic pancreatic adenocarcinoma whose tumors harbored a deleterious or suspected deleterious germline BRCA mutation.3 Since 2010, a large body of transcriptomic and immunophenotypic research done in PDAC has led to the development of numerous targeted- and immunotherapy-based clinical trial options as well, many of which are biomarker-driven and are predicated on the identification of a specific genetic or immunologic signature.
While tissue-based assays remain the mainstay for molecularly characterizing PDAC, 2 limitations for the application of these assays are access to tissue and tumor heterogeneity in metastatic sites. Tumor-based biopsies can lead to ambiguous or inconclusive samples and thus may underestimate the mutational burden of heterogeneous tumors.4 Spatially separated heterogeneous somatic mutations and chromosomal imbalances translating into phenotypic intratumor diversity have been described across tumor types as well.4 Furthermore, there is often neither a safe nor feasible area to biopsy an individual’s tumor due to anatomic factors. Novel blood-based sequencing platforms have emerged and are now FDA approved in malignancies such as non–small cell lung cancer. These assays allow for molecular profiling based on circulating tumor DNA (ctDNA) that is released into the blood via apoptosis, necrosis, and active secretion.
Previous studies have suggested that the presence of ctDNA may be prognostic in both localized and advanced PDAC and that ctDNA concentration may be correlated with recurrence after tumor resection.5 Similarly, high ctDNA concentration has been found to be associated with distant organ metastases.6,7 Whether ctDNA dynamics will aid in treatment decision-making or be associated with response to therapy in PDAC, as has been found in other cancers, remains to be seen.8-12 Few data are currently available that characterize the genomic landscape in individuals with metastatic PDAC using blood-based assays. Herein, we characterize the mutational landscape of patients with metastatic PDAC who received blood-based molecular profiling.
We performed a retrospective review of 77 consecutive patients with refractory, metastatic PDAC who were referred to the Sarah Cannon Research Institute (SCRI) at HealthONE, an oncology drug development unit in Denver, Colorado, between August 2014 and May 2019. This analysis includes all patients with metastatic PDAC who were referred to SCRI; who elected to undergo blood-based molecular profiling, either before or after the initial referral; and had a full report available. Data for those tested prior to referral were not available, and thus it was unknown at what stage those patients were tested. We evaluated patient demographics; prior medical history, including initial stage at diagnosis, comorbidities, and prior cancer-related therapies (surgery, radiation, and systemic chemotherapy or prior clinical trial data); and source of referral (academic vs community). As a proxy for defining academic vs nonacademic practices, we classified each referring oncologist based on whether they are affiliated with a National Cancer Institute–designated cancer center. Blood samples were evaluated at the Clinical Laboratory Improvement Amendments–licensed and College of American Pathologists–accredited clinical laboratories of Guardant Health, Inc for ctDNA analysis.
Clinical outcomes and variables evaluated in our analysis include all alterations identified at the time of testing; percent ctDNA (% ctDNA) present in the bloodstream; and potential actionability of alterations. The determination of whether an alteration was potentially actionable or not via a clinical trial is based on the actionable gene list created by the Institute for Personalized Cancer Therapy – Precision Oncology Decision Support team at The University of Texas MD Anderson Cancer Center.13 These standards mandate that a gene is potentially actionable if there is supporting evidence that the gene is a driver for tumorigenesis, where any action on the gene can either determine sensitivity or resistance to drugs, and that it applies to all alteration types. Additionally, there must be a clinically available agent targeting the gene that is in at least preclinical development.
Based on the aforementioned criteria, the following genes were considered actionable based on their sensitivity to respective targeted agents: AKT1, ALK, ARAF, ARID1A, ATM, BRAF, BRCA1, BRCA2, CCND1, CCND2, CDK4, CDK6, CDKN2A, CDKN2B, EGFR, ERBB2, FGFR1, FGFR2, FGFR3, HRAS, IDH1, IDH2, JAK2, JAK3, KIT, KRAS, MAP2K1, MAP2K2, MET, MPL, MYC, NF1, NOTCH1, NRAS, NTRK1, PDGFRA, PIK3CA, PTEN, PTPN11, RET, ROS1, SMO, STK11, and TSC1. The following genes were deemed actionable based on context-specific criteria: CCNE1 (sensitivity to CDK2 inhibitors; resistance to CDK4/6 inhibitors); ESR1 (presence is sensitizing to hormone therapy; mutations cause resistance to antihormone therapy); NRAS (sensitivity to MEK inhibitors; resistance to cetuximab [Erbitux] and panitumumab [Victibix]); RAF1 (activating alterations cause sensitivity to MEK inhibitors and resistance to RAF inhibitors; inactivating alterations cause resistance to dasatinib). Only AR and RB1 were deemed actionable solely because of their resistance to antihormone therapy and to CDK4/6 inhibitors, respectively. The following were considered nonactionable: TP53, CTNNB1, APC, GNAS, NFE2L2, MLH1, RIT1, SMAD4, HNF1A, CDH1, GATA3, VHL, FBXW7, and RhoA. Co-occurring alterations were also analyzed using descriptive methods. Data analysis was performed using R Studio.14 We used the Mann-Whitney U test to compare the mean number of comorbidities and mutations between individuals who died and those who were still alive. To calculate median survival, we calculated the number of months between initial diagnosis and either death or last date of contact. This research was approved by the HealthOne HCA institutional review board.
As seen in Table 1, all patients had refractory disease after pancreaticoduodenectomy, chemotherapy, and/or radiation therapy. Of 77 patients, 41 (55%) were men. Median (SD) age was 66 (9.3) years (range, 44-83). Patients had between 0 and 7 medical comorbidities (median, 2). Functional status was determined by ECOG performance score (PS), ranging from ECOG PS 0 to 3, with 69% (53/77 patients) scoring ECOG PS 1. Additionally, 28 of 77 patients (39%) were smokers whereas 41 (57%) were not, and 3 had unknown status (4%). Initial staging varied from stage I to stage IV, but the largest proportion of patients, 37 of 77 (48%), had stage IV disease at diagnosis and had de novo disease. As seen in Table 1, surgical intervention prior to referral occurred in 18 of 77 patients (23%), including 13 pancreaticoduodenectomies, 1 distal pancreatectomy, 1 segmental pancreatectomy, and 1 total pancreatectomy. Ninety-seven percent of patients (75/77) underwent chemotherapy prior to referral, which consisted of either gemcitabine and nab-paclitaxel (Abraxane), FOLFIRINOX (folinic acid, fluorouracil, irinotecan, oxaliplatin), fluorouracil, or a combination of these. Eleven patients (14%) had enrolled in another clinical trial prior to referral and 22 received radiation.
Blood from 34 of 77 patients (44%) was found to harbor 1 or more genetic alterations, 7 patients (9%) had 1 genetic alteration, and 5 patients (6%) had no alterations. Forty-
seven patients (61%) underwent Guardant-360 testing; however, the stage in which they were tested was not available in those who were tested prior to referral. Thirty-seven patients (48%) were referred to our clinic with stage IV disease, however, and 20 of these patients with stage IV disease (26%) were tested after referral. Thirty patients (39%) did not undergo Guardant-360 testing. Seventeen of these 30 patients underwent FoundationOne testing, 2 underwent Caris MiProfile testing, and 11 underwent no testing.
The number of alterations were 119 (total, nonunique) and 96 (unique), and the median number of alterations per patient was 3. Median mutant allele frequency (% ctDNA) was 0.5% (range, 0.09%-75.2%). TP53 was the most commonly altered gene (29 alterations, 23/77 patients [30%]), followed by KRAS (27 alterations, 26/77 patients [34%], with 1/77 being the potentially actionable G12c [1%] and 12/77 [16%] being G12D, BRCA2 (9 alterations, 7/77 patients [9%]), SMAD4 (4 alterations, 2/77 patients [3%]), and CDKN2A (4 alterations, 4/77 patients [5%]). Of the patients with alterations, 36% (28/77) had 1 or more potentially actionable alterations, most commonly BRCA2 (12%), STK11 (2%), PIK3CA (1%), NF-1 (1%), EGFR (1%), FGFR (1%), NRAS (1%), and KRAS G12C (1%). In these genes, mutations (vs amplifications, deletions, or others) occurred most frequently. Table 2 outlines the data to support actionability of notable mutations given the utility of specific treatments found to act on these genes. Table 3 provides a comprehensive list of the single nucleotide variants, copy number amplifications, fusions, and indels evaluated for, in selected genes.
Given the association between mechanisms of resistance to targeted therapies and co-occurring alterations, we also assessed patterns of co-occurring alterations in this data set, which are detailed in Table 4. We noted that KRAS co-occurred with various mutations in 10 different combinations: KRAS as G12D with TP53 mutations 9 times, with TP53 and SMAD4 five times, as G12V with TP53 four times, as G12D with CDKN2A four times, with BRCA2 four times, and as G12D in combination with CDKN2A and TP53 three times.
A total of 24 patients in our data set were diagnosed with other cancers within their lifetimes, and 6 of the 24 patients were found to harbor mutations in the blood that had therapies associated with them based on the Guardant360 test. Two patients had basal cell skin cancer, and 1 patient had an unspecified skin cancer. The mutations identified in these 24 patients included NF-1, BRCA2, GNAS, and CDK2NA. One patient with prostate cancer had a BRCA2 mutation. One patient with colorectal cancer had both CDKN2A and KRAS mutations.
At the time of retrospective review, 20 individuals were still alive, 30 individuals had died, 23 individuals had an unknown status, and 4 individuals had missing status information. Based on the information in our data set, the median survival was 21 months (range, 3-68 months). It should be noted that 1 patient in the “alive” groups has been alive for 126 months, which is an outlier. This was a healthy man with no other medical problems, aged 72 years, who underwent treatment with FOLFIRINOX as well as with gemcitabine/nab-paclitaxel but had disease progression with both treatments. He completed 2 clinical trials (with entrectinib (Rozlytrek), a Trk/ALK/ROS1 inhibitor, and niraparib (Zejula), a PARP inhibitor), both with no success, so he transitioned to hospice. He did not undergo any molecular profiling testing as he had already transitioned to hospice at the time.
The next highest number of months being alive is 50. This patient, a man aged 78 years, was mostly healthy otherwise, had an ECOG PS of 1, and was diagnosed with pancreatic adenocarcinoma in 2015. He had undergone treatment with FOLFIRINOX, capecitabine, and gemcitabine/nab-paclitaxel, as well 2 clinical trials, but progressed in his disease and was referred once he developed pulmonary and gastric metastases. The 2 clinical trials involved an aurora B kinase inhibitor and niraparib. He had the following blood-based mutations: TP53 A79fs, GNA11 D236D, ERBB2 (HER2) E147E, and APC D1571del. Interestingly, he also underwent FoundationOne tumor molecular profiling, which revealed the following tissue mutations: BRCA1 exon 10 K654fs, EGFR, FRCC1, KRAS exon 2 G12D, MGMT, TLE3, TOP2A, IOPO1, TS, and THBBS.
The members of both the groups with individuals who had died and those who were still alive at the time of retrospective review had a median of 2 comorbidities. This, however, was not statistically significant (P = .07). The mean number of mutations between the 2 groups was almost identical and not statistically significant, and proportions of patients with each cancer stage were similar in progression trend. The data we had on mutations in those who were still alive consisted of 32 patients (unknowns excluded). The median number of mutations for those alive was 2 (range, 0-4) and for those who had died was 3 (range, 0-15). The individual who had 15 is an outlier. The difference between groups in number of mutations was not statistically significant (W = 132.4; P = .25).
Herein, we report the biologic and clinical correlates of genomic alterations among 77 patients with metastatic PDAC using blood-based ctDNA analysis.
The majority of our patient population had metastatic disease prior to referral but had good performance status. Based on the information in our data set, the median survival was 21 months (range, 3-68). All patients had refractory disease after pancreaticoduodenectomy, chemotherapy, and/or radiation therapy. Patients had between 0 and 7 (median, 2) medical comorbidities. Functional status ranged from ECOG PS 0 to 3, with 69% (53/77) patients having ECOG PS 1. Additionally, 28 of 77 patients (39%) were smokers, whereas 41 (57%) were not, and 3 had unknown status (4%). Initial staging varied from stage I to stage IV, but the largest proportion of patients, 37 of 77 (48%), had stage IV disease at diagnosis and had de novo disease. As seen in Table 1, surgical intervention prior to referral occurred in 18 of 77 patients (23%). Ninety-seven percent of patients (75/77) underwent chemotherapy prior to referral, which consisted of either gemcitabine and nab-paclitaxel, FOLFIRINOX, fluorouracil, or a combination of these. Eleven patients (14%) had enrolled onto another clinical trial prior to referral and 22 received radiation.
Thus, this is consistent with the outcomes described in patients with PDAC referred to other phase 1 centers.15 Data from other institutions that do not enroll patients on early-phase trials would need to be analyzed to determine whether our data reflect the general population of those individuals with metastatic PDAC, but our data are consistent with those of the population that is generally referred for early-phase trial consideration at other institutions.
Our analysis confirmed earlier reports that blood-based ctDNA detection is feasible in patients with metastatic PDAC. As data regarding blood-based molecular testing prior to referral were not available, it is unclear what percent of patients had testing at what stage. Thirty-seven patients (48%) were referred to our clinic with stage IV disease, however, and 20 of these patients with stage IV disease (26%) were tested after referral. Percent ctDNA was calculated; however, we did not have tissue from each patient to confirm concordance. Forty-five percent of patients had at least 1 alteration, with the most frequently characterized alterations being TP53, KRAS, BRCA2, SMAD4, and CDKN2A.
Botrus and colleagues collected ctDNA in 282 patients with locally advanced or metastatic PDAC and demonstrated that 90% of ctDNA samples harbored at least 1 genomic alteration; each sample harbored an mean of 2.7 alterations, with a median allelic fraction of 0.40%.16 These results contrast with those of our study, in which only 45% of patients had at least 1 genomic alteration, but they are similar in that the patients in our study had an mean of 3 alterations and a median allelic frequency (% ctDNA) of 0.5% (range, 0.09%-75.2%). Consistent with our analyses, Botrus et al found that the most commonly identified alterations were in TP53, KRAS, SMAD4, CDKN2A, and EGFR, and the most common potentially actionable alterations were KRAS, PIK3CA, ATM, EGFR, and MYC.16 Interestingly, the presence of KRAS, SMAD, and BRCA1 mutations were identified more frequently in patients at baseline and with disease progression. TP53 and BRCA2 mutations, on the other hand, appeared to be acquired alterations in their data set. The consistency in these 2 data sets suggest that our results may be generalizable outside of these institutions.
Patel and colleagues analyzed ctDNA in 112 patients with PDAC, both resectable and metastatic. Interestingly, despite previous data showing that tissue-derived DNA and ctDNA are comparable, with high concordance rates (greater than 95%),17 concordance rates between ctDNA and tissue DNA alterations were 61% for TP53 and 52% for KRAS in this particular study.18 These results highlight the need for multiple institutions to ask these same questions in order to better understand these complex questions involving the utility of assessing ctDNA in PDAC. In our study, we were unable to evaluate concordance, because there were patients who did not undergo surgery or tissue biopsy.
Thirty-six percent of patients in our analysis had potentially actionable alterations—most frequently BRCA2, followed by STK11, KRAS, PIK3CA, NF-1, EGFR, and FGFR. KRAS and TP53 most often occurred together. In our data set, the majority of the patients with a BRCA2 alteration that was identified upon blood-based testing did not have prior tissue-based molecular profiling performed. This underscores a potential use for blood-based molecular profiling. In 2020, the National Comprehensive Cancer Network Guidelines for Genetic/Familial High-Risk Assessment, version 1.2020, was expanded to provide information about genetic consultation and BRCA2 testing in patients with pancreatic cancer. Blood-based germline testing is already being utilized, currently; however, whether blood-based ctDNA testing may facilitate or supplement these guidelines remains to be seen.
Thus far, there are no approved agents, with the exception of the PARP inhibitor olaparib, against most mutations identified in PDAC. The prevalence of genetic testing is increasing, but there is a lower consistent prevalence of mutations and a limited number of treatments outside of clinical trials.
Historically, KRAS is known to be mutated in the majority of cases of PDAC and it is believed to be one of the earliest and most critical events of pathogenesis.1 It was initially thought to lack specificity, as it is often elevated in smokers and those with chronic pancreatitis. However, Groot and colleagues compared preoperative and postoperative KRAS ctDNA, with results revealing that an increase in ctDNA postoperatively correlated with tumor recurrence on imaging.5 Furthermore, Park and colleaguesdetermined that ctDNA levels indicated the presence of cancer, and that they correlate well with clinical responses to treatment and progression in patients with PDAC.19 The KRAS mutation has been studied the most, has been associated with worse prognosis,20 and has been observed in 90% of cases of PDAC5,8 and specifically within 4 genetic alleles: G12V, G12Dz, G12R, and Q61H.5
In our study, however, KRAS mutations were found in only 34% of patients (26/77), with only 1% (1/77) possessing the only currently known potentially actionable mutation, KRAS G12C. Whether the presence or absence of KRAS in ctDNA has either prognostic or predictive implications remains to be seen.
The relevance of being able to detect KRAS alterations in the blood is underscored by the recent FDA approval of the first KRAS inhibitor, sotorasib (Lumakras), for KRAS G12C–mutated NSCLC.21 While the current approval is only in NSCLC, 7.1% of individuals with KRAS G12C–mutated colorectal cancer had confirmed response and 73% had disease control. Individuals with PDAC, as well, were noted to have responses to therapy and disease stabilization with sotorasib monotherapy. Other KRAS allele–specific inhibitors are currently in development, as are other MAPK-targeted inhibitors in monotherapy and in combination with sotorasib and other KRAS G12C inhibitors.
Kulemann and colleagues, for example, isolated and genetically characterized circulating tumor cells (CTC) in the blood that are hypothesized to be a means of systemic tumor spread.22 Blood from healthy donors and from otherwise healthy patients with PDAC were evaluated, and KRAS mutations in pancreatic CTCs were compared. It was found that those with more than 3 CTC/ml had a trend for worse median OS compared with patients with fewer or no detectable KRAS mutations.22 Unfortunately, because there is inconsistent detection, isolation, and characterization of CTC in PDAC,22 it is unclear to what extent these mutations shed CTCs into the bloodstream.
The incidence of detecting a KRAS mutation in our data set (27/77 patients; 35%) was vastly different from previous reports of KRAS being identified in tissue biopsies from approximately 92% of patients with PDAC.23 However, Patel and colleagues identified KRAS mutations in only 44% of PDAC18 and Kulemann and colleagues found KRAS in 58% of patients,22 both of which are more consistent with our findings. One potential discrepancy in our results and of those reporting higher KRAS mutation frequency is that many previous reports listed all KRAS mutations, not just potentially actionable ones. Another reason could be that not all KRAS mutations may shed into the bloodstream. These hypotheses require further investigation.
In our analysis, all patients had metastatic PDAC and had varying levels of ctDNA in the blood. It is still unclear what amount of ctDNA is meaningfully significant or what allelic fraction would be associated with an improved likelihood of response to a targeted therapy in PDAC. These questions remain outstanding. It is also worthwhile to note that even precancerous pancreatic lesions may shed ctDNA. The prevalence of KRAS mutations in low-grade precursor pancreatic intraepithelial neoplasia lesions, for example, is listed as greater than 90%.22 ctDNA has also been shown to be present in the bloodstream of patients with noninvasive pancreatic lesions,8 but these may never become carcinomas. The data we present here—as well as those presented by our colleagues who have described ctDNA findings in their advanced PDAC patient data sets in the past—must be examined in context of these collective findings. Moreover, these data raise further questions on whether ctDNA results may play a future role in guiding the workup and management of patients with pancreatic masses.
We have also observed that many of our patients with KRAS alterations harbored at least 1 co-alteration, including mutations in TP53, CDKN2A, BRCA2, and SMAD4. This was similar to the findings of Botrus and colleagues, who found that the most common co-occurring mutations were TP53 and KRAS (n = 74), KRAS and SMAD4 (n = 25), KRAS and CDKN2A (n = 12), and TP53 and SMAD4 (n = 20).16 Previous literature supports the concept that inactivating mutations in tumor suppressor genes such as CDKN2A/p16, TP53, and SMAD4 cooperate with KRAS mutations to cause aggressive PDAC tumor growth.22 These observations suggest that co-targeting mutations may be an area of future research, and they highlight the need to better understand whether these genomic signatures simply play a prognostic role or whether they may also be predictive of response to therapies. Such ongoing analyses will be critical to understand the optimal mechanism by which to target KRAS in PDAC in the ongoing KRAS inhibitor combination studies.
In our data set, the median survival was 21 months (range, 3-68). It should be noted that 1 patient in the “alive” groups has been alive for 126 months, which is an outlier. The next highest number of months being alive is 50. Although the latter patient’s median survival is still higher than that of the typical patient with metastatic PDAC, it is unclear whether he had less aggressive disease, given that he had survived long enough to undergo 3 chemotherapy regimens (despite recurrent disease). This is likely due to a selection bias in patients who received blood-based molecular profiling in our analysis (who may have more indolent tumors), and it may not correlate well with the entire population of individuals with metastatic PDAC, who should also be explored. It is unclear what relevance this may have in terms of what mutations may be present in this subset of patients, or if ctDNA shed in the bloodstream varies among mutations.
The goal of identifying mutations is ultimately to determine if targeting them will lead to improved survival in patients with PDAC. Pishvaian and colleagues performed a retrospective analysis of patients with advanced PDAC who were treated with targeted agents specific to a mutation compared with unmatched therapies; the results suggested a survival improvement with the targeted agents.24 Of 1082 patients, 189 had actionable molecular alterations; 46 and 143 underwent matched and unmatched therapies, respectively. Those who underwent matched therapy had significantly longer median survival.24 While this type of retrospective study can only be hypothesis generating, we think it identifies the potential of targeting alterations in PDAC. Our findings suggest that a blood-based approach to profiling PDAC may remain feasible and may help clinicians identify molecularly tailored therapy options and clinical trial recommendations for a subset of patients with PDAC.
Our study has numerous limitations. First, the study was retrospective, with a small sample size of 77 patients. Second, it was a single-center study in which most patients were residents of a single county (Denver), which may limit the generalizability of our analysis. Third, these patients had advanced disease and thus short life expectancy, limiting our ability to study this population. At the same time, our survival data identify that the patients in this data set may also represent only a subset of patients with advanced PDAC who have a favorable prognosis; however, this data is consistent with those seen at other tertiary early-phase clinical trial centers. Additionally, as all of our patients had metastatic disease refractory to standard treatments, our results may or may not apply to individuals with early-stage PDAC. We did not look at the relationship among metastatic, refractory, and pancreatic adenocarcinoma that responded to standard treatment, to determine if each of the 3 groups had the same mutations. With our study’s relatively small number of patients, we were unable to calculate conclusions on OS in relation to genetic mutations or % ctDNA in a statistically meaningful manner.
We are hopeful that further research will allow the use of ctDNA to predict prognosis and estimate survival for patients diagnosed with PDAC. Further research will also be required to determine if these data are applicable to every patient with metastatic PDAC, if these data are being utilized for trial enrollment, whether FDA-approved therapies are effective in PDAC, and whether these markers are associated with response to therapy on trials.
Our study results show that ctDNA is seen in patients with metastatic PDAC; this fact may help select patients for clinical trials and may potentially help the future development of targeted therapies. More research is required to determine which patients should be tested, and when, and to define the implications for treatment.
1. Department of Surgery, Swedish Medical Center, Englewood, Colorado, United States.
2. Department of Clinical Oncology Research, Sarah Cannon Research Institute at HealthONE, Denver, Colorado, United States.
3. Graduate Medical Education, HCA HealthCare, Denver, Colorado, United States