High Prevalence of Hereditary Cancer Syndromes and Outcomes in Adults with Early-Onset Pancreatic Cancer

Pancreatic ductal adenocarcinoma (PDAC) is currently the eleventh most common cancer in incidence and the third leading cause of cancer-related deaths in the United States (1). The majority of PDACs are sporadic; however, 5% to 10% may have a hereditary cause (2). PDAC is considered a disease of the elderly, and most diagnoses are made in individuals over the age of 65, with a median age at diagnosis of 70 years (3). Individuals diagnosed with PDAC under the age of 60 are considered to be young onset and potentially at high risk for a genetic predisposition. Although pancreatic cancer is typically associated with environmental and lifestyle factors such as smoking, obesity, and diabetes (4–7), inherited germline mutations confer a significantly elevated lifetime risk for PDAC. Germline mutations in a growing number of genes have been associated with increased risk of PDAC, including ATM, APC, BRCA1, BRCA2, CDKN2A, MLH1, MSH2, MSH6, PMS2, PALB2, STK11, and PRSS1 (2, 8–11).

Familial pancreatic cancer (FPC) is defined as a family with at least two first-degree relatives (FDR) with pancreatic cancer without an identifiable syndrome or genetic mutation in the family. Relatives meeting FPC criteria have an empirically increased risk to develop PDAC over the general population; these individuals can be even further stratified depending on their degree of relationship with the affected relative(s) (12). The standard incidence rates to develop PDAC in individuals with one FDR, two FDRs, or three or more FDRs are 3 to 4, 5 to 7, and 17 to 32, respectively (3, 13).

Germline mutations confer an elevated lifetime risk for PDAC (2, 8–11). Typically, germline mutations associated with higher risk for PDAC have been suggested by kindreds with multiple generations of pancreatic or related cancers (breast, ovarian, colon, etc.), early cancer diagnoses, individuals with multiple primary tumors, and/or Ashkenazi Jewish ethnicity (13).

Recent studies have found that an appreciable fraction of patients with apparent sporadic pancreatic cancers have germline mutations (14, 15). In search of predictive factors for germline mutations, previous studies have looked at age of presentation, but results have been ambiguous. Grant and colleagues (16) and Holter and colleagues (17) reported 3.8% and 4.5% prevalence of germline mutations in PDAC patients, respectively, while Hu and colleagues (18) reported 9.4% prevalence of mutations in established PDAC genes and up to 13.5% total pathogenic mutations. These three studies found no differences in age with respect to mutations, but the numbers of young-onset PDAC included in the analysis were rather low. Very recently, a major study from Shindo and colleagues (15) has found slightly higher prevalence of germline mutations in younger patients. Therefore, we aimed to definitively determine if patients with germline mutations present earlier than those without mutations. Given that only 30% of individuals with PDAC are diagnosed under the age of 60 (17), we examined germline mutation prevalence in a high-risk cohort (HRC) which includes a large young-onset PDAC subpopulation during an 11-year period at the University of Texas MD Anderson Cancer Center (MDACC). As validation, we examined a general cohort (GC) of PDAC patients also seen at MDACC.

Included patients were referred to genetic counseling, based on established criteria (Supplementary Table S1), seen at the MDACC from 2005 to 2016. All patients diagnosed with PDAC diagnosed under the age of 60 (young-onset PDAC) met referral criteria regardless of family history. From a total of 584 patients seen in the HRC, 261 were patients with young-onset PDAC. The standard genetic counseling consultation included obtaining a pedigree of at least three generations, risk assessment, and review of the patient's personal medical history and risk factors. Genetic testing was recommended based on formal risk assessment made by a board-certified genetic counselor (S.A. Bannon and M.E. Mork). All genetic testing was performed at Clinical Laboratory Improvement Amendments-certified laboratories. The landscape of genetic testing options has dramatically evolved over the time during which these patients were seen for genetic counseling. Single-gene analysis based on personal or family history or multigene panel testing were performed. In the panels, up to 13 mutations were tested: ATM, APC, BRCA1, BRCA2, CDKN2A, MLH1, MSH2, MSH6, PMS2, PALB2, STK11, TP53, and EPCAM.

All patients with metastatic pancreatic cancer who received first-line chemotherapy at MDACC, between January 2010 and January 2016, and consented for DNA testing, were eligible. Results were considered for research purposes only and not used to make clinical decisions. Exome capture was performed from 500 ng of genomic DNA using the KAPA library (KAPA Biosystems) and sequenced using an Illumina HiSeq 2500 instrument with 200× average coverage. Genetic variants were classified based on available public databases[Exome Aggregation Consortium Database 0.3, COSMIC release 70,72, NHLBI exome sequencing project ESP6500SI-V2, dbSNP129,138, ClinVar v20150330, Exome Variant Server (EVS), 1000 Genomes, dbVar, Human Gene Mutation Database, HGVS, and DECIPHER] accessed between January 2017 and December 2017 and in silico analysis (SIFT, PolyPhen-2, MutationTaster). For the present study, we have focused on analyzing the 13 same genes that were included in the multigene panel testing performed in HRC patients.

For both cohorts, clinical and pathologic data were abstracted from the Electronic Health Record. MDACC Institutional Review Board approved this study. The study was conducted in accordance with the Belmont report.

Patients diagnosed with only a pathogenic germline mutation (in both cohorts) were assigned to the hereditary group and those with a variant of uncertain significance (VUS) or no identifiable mutations were assigned to the sporadic group. Patients harboring a pathogenic gene mutation and an additional VUS were assigned to the hereditary group.

Categorical variables are reported as frequencies and percentages; continuous data are summarized as mean and standard deviation (SD). The χ² and Fisher exact tests were used to evaluate associations between categorical variables and mutation status. The t test was used to compare the distributions of continuous variables (such as age) between mutation statuses. Univariate logistic regression models were used to evaluate the association between the risk factor (age) and mutation status. Odds ratios (OR) and 95% confidence intervals (CI) were estimated to measure the strength of association. The recursive-partitioning method was used to select optimal cutoff point for continuous age based on mutation status using HRC. The identified optimal cutoff point for age was validated using GC. Overall survival (OS) was defined as the time from diagnosis to death from any cause. Living patients were censored at date of the last follow-up. Kaplan–Meier curves were estimated for the survival distributions by mutation status. The log-rank test was used to test the difference in survival distributions between subgroups. Univariate Cox proportional hazard models were used to determine the effects of mutation status on OS. Hazard ratios and 95% CI were provided. All tests are two-sided. P values less than 0.05 are considered statistically significant. All analyses were conducted using SAS 9.4 (SAS) and S-Plus 8.0 (TIBCO Software Inc.) software.

Between 2005 and 2016, 584 patients with PDAC received genetic counseling. The characteristics of this population are described in Table 1. The mean age at diagnosis was 61.44 years, and slightly more than half were female (53.8%). In terms of racial/ethnicity characteristics, the majority of patients were non-Hispanic white (82.7%), while Hispanic, black, and Asian represented 6.8%, 7.2%, and 3%, respectively. A large proportion of patients had no previous history of tobacco use (58.2%) while alcohol consumption was reported by 62.7% of patients. Importantly, 13.6% of patients had personal history of breast cancer and 4.5% had gynecologic cancer. A total of 171 patients were found to have at least one FDR with breast cancer (29.2%). With respect to family history of pancreatic cancer, 127 patients had at least one FDR (21.7%) and 111 had at least one second-degree relative (SDR) affected (19%).

From the 584 patients in the HRC, 277 underwent genetic testing for hereditary pancreatic cancer genes. We compared the clinical and pathologic characteristics of patients who underwent testing with those who did not, and we found that younger patients were more likely to be tested (P = 0.019). We then observed that patients with personal history of chronic pancreatitis were less likely to be tested (P = 0.014) and those with a personal history of breast or gynecologic cancers were more likely to undergo genetic testing (P = 0.0001 and P = 0.0085). Those with at least one FDR with breast cancer were also more likely to undergo genetic testing than those without an FDR with breast cancer (P = 0.0047; Supplementary Table S2). These results were expected as young patients or those with stronger personal or family history of cancer received a stronger recommendation for genetic testing at the time of genetic counseling.

Pathogenic germline mutations were identified in 48 patients (17.32%; Fig. 1A), VUS were found on 14 patients, and 215 patients had uninformative results. Of the 277 patients who underwent clinical genetic testing in the HRC, 240 (86.7%) were tested by single-gene analysis and 37 (23.3%) by the multigene panel. Of the 130 young-onset (<60) PDAC patients, 28 (21.6%) underwent panel testing and 102 (78.4%) had single-gene testing. In the young-onset cohort, we compared the mutation detection of panel versus single-gene testing. We found no significant difference in the yield of pathogenic mutations detected with panel versus single-gene testing, 21.43% versus 22.5%, respectively (P = 0.9). However, panel testing identified a significantly higher number of VUS than single-gene testing, as expected (25% vs. 1.96%; P = 0.0003; Supplementary Fig. S1). There was significant heterogeneity in the number of patients tested for each gene, as single-gene(s) genetic testing was determined by a genetic counselor based on personal and family history risk assessment; therefore, not all patients in the HRC were tested for all genes. The most frequently mutated genes were BRCA1 and BRCA2 (29/128 patients with BRCA2 mutations and 5/127 with BRCA1, respectively), which is consistent with previous reports (17). The second most common group of mutations detected were in the MMR genes, with 2 of 33 in MLH1, 3 of 34 in MSH2, and 2 of 33 in MSH6. Seven patients had germline mutations in other genes: 2 of 35 in TP53 and ATM (30 tested) and 1 each in APC (31 tested), CDKN2A (37 tested) and STK11 (33 tested; Supplementary Fig. S2A). VUS were considered as negative due to inconclusive association with PDAC predisposition. There were three patients who carried both, one pathogenic mutation and one VUS. For analysis purposes, these patients were considered in the group of patients with pathogenic mutations. Details on the pathogenic mutations found in HRC young-onset patients are listed in Supplementary Table S3.

We then looked at the influence of age in the mutation status of tested patients within the HRC, which was the main goal of the study. We found that the mean age at time of PDAC diagnosis in patients with germline mutations was significantly younger than in patients without mutations [56.6 (SD ± 10.95) vs. 61.07 (SD ± 10.95); P = 0.010; Table 2]. We then compared mutation prevalence in early-onset PDAC patients (<60 years old) with older patients (≥60 years old) and found that early-onset patients had significantly higher odds of testing positive compared with older patients (OR, 1.93; 95% CI, 1.03–3.70, P = 0.039; Fig. 1A and B).

Using the recursive-partitioning method, we identified 42 years of age as the optimal cutoff that best separates age-stratified groups based on mutation status in the HRC. Patients younger than 42 years had remarkably higher odds of testing positive for a mutation than patients older than 42 years (OR, 4.17; 95% CI, 1.42–11.84, P = 0.011; Fig. 1B).

Importantly, we found that approximately half (50.1%) of the patients in the HRC met National Comprehensive Cancer Network (NCCN) 2018 BRCA1/BRCA2 genetic testing criteria (Supplementary Table S4), and 44.6% of patients younger than 60 years met criteria, indicating that 55.4% of young-onset patients were tested solely based on indication of age. Finally, in the mutation-positive HRC, 52.1% of patients met criteria, while in the mutated younger-onset group 41.4% did (Supplementary Table S5).

We then assessed for the association between other potential predictive factors and mutational status, and a significant difference was seen by ethnicity, with Hispanic individuals more likely to test positive (P = 0.026; Table 2). With respect to family history, patients with at least one or more FDR with breast, gynecologic, and colon cancers were significantly more likely to harbor mutations (P = 0.021, P = 0.031, P = 0.022, respectively; Table 2). Also, patients with at least one SDR with gynecologic cancers were more likely to have germline mutations. Patients with family history contributory for pancreatic cancer in FDRs and/or SDRs were not more likely to have germline mutations than those without it (Table 2). There were no other significant differences between the two groups (with mutations vs. without mutations) regarding sex, grade of differentiation, stage, personal history of cancer, and other risk factors.

To validate the higher prevalence of germline mutations in early-onset PDAC in a non-HRC, we analyzed a second group of patients with PDAC seen at the same institution referred to as the general cohort (GC). A total of 233 consecutive patients with metastatic PDAC receiving chemotherapy treatment at MD Anderson were enrolled, from which 132 patients had sequencing performed. The selection of patients for sequencing was randomly performed. The characteristics of the tested GC population are described in Supplementary Table S6. The mean age at diagnosis was 59.73 years, and 59.1% of the patients were male. Most of patients were non-Hispanic white (84%), and Hispanic, black, and Asian represented 4.5%, 9%, and 2.2%, respectively. A total of 11 patients were found to have at least one FDR with pancreatic cancer (8.3%), while 9 patients had at least one SDR with pancreatic cancer (6.8%).

From the 132 patients who had sequencing, 9 were found to have a pathogenic germline mutation (6.81%; Fig. 1A). Again in this group, the most common mutations found were in BRCA/2 genes (Supplementary Fig. S2B). The average age of patients with germline mutations was also younger than patients without mutations [48.44 (SD ± 9.65) vs. 60.56 (SD ± 10.57; P = 0.033; Supplementary Table S7]. In this cohort, 59 patients presented with early-onset PDAC (<60) and 7 of them had germline mutations, which is significantly associated with higher odds of testing positive for a mutation compared with older patients (>60; OR, 4.78; 95% CI, 1.10–32.95; P = 0.036; Fig. 1A and B). When using the optimal cutoff of 42 years of age, determined using the HRC data set, again we found significantly higher prevalence of mutations in patients younger than 42 versus those older (OR, 20; 95% CI, 3.16–131.27, P = 0.002; Fig. 1B). No other factors, besides age, were associated with higher prevalence of mutations in the GC (Supplementary Table S7).

We then performed an analysis of clinical outcomes in both cohorts. We first analyzed OS in resectable patients from the HRC, and found that patients with germline mutations had significantly better OS than patients without mutations, with a median survival of 70.4 months versus 32.6 months, respectively (HR, 0.55; 95% CI of HR, 0.33–0.91, P = 0.03; Fig. 2A). Adjuvant chemotherapy regimens in the resectable HRC patients were composed of single-agent gemcitabine and/or platinum-based regimens, including 5-fluorouracil, irinotecan, oxaliplatin (FOLFIRINOX), single-agent oxaliplatin, and cisplatin. Unfortunately, we do not have enough power to detect differences in outcomes by treatment type given the limited number of patients in each chemotherapy type.

We then analyzed survival in PDAC patients from GC (all metastatic). Patients from the GC with germline mutations had significantly improved survival than patients without mutations detected with a median survival of 18.2 versus 9.2 months (HR, 0.44; 95% CI of HR, 0.25–0.76, P = 0.03), regardless of adjuvant chemotherapy treatment regimen (Fig. 2B). When we looked at the metastatic patients from the HRC, we did not find significant differences in survival based on their mutational status (Supplementary Fig. S3A). However, metastatic patients in the HRC, disregarding their mutational status, had significantly longer OS than patients without mutations in the GC (HR, 0.51; 95% CI of HR, 0.39–0.66, P < 0.0001; Supplementary Fig. S3B).

This study analyzes a large HRC of patients diagnosed with pancreatic cancer at a tertiary referral center with the main goal of determining if patients with germline mutations on PDAC susceptibility genes present earlier than those with sporadic PDAC. We determined the prevalence of germline mutations and looked for predictive factors of germline mutations and used a second GC for validation purposes. A total of 17.32% of patients in the HRC were found to have pathogenic germline mutations against 6.81% in the GC, which is consistent with the previously reported prevalence of hereditary pancreatic cancer in the general pancreatic cancer population (15, 16). Interestingly, panel testing did not yield significantly increased detection of pathogenic germline mutations over single-gene(s) testing as guided by genetic counselor risk assessment. However, panel testing did significantly increase the risk of identifying VUS. This finding is significant in the clinical practice context where VUS can pose confusion and add uncertainty to individuals attempting to discern their risk to develop PDAC. Germline mutations for genes associated with increased risk of PDAC were found in 11% to 22% of PDAC patients younger than 60 years old and 44% to 50% of patients younger than 42 years old. Therefore, young age represents a strong predictive factor of germline mutations in PDAC patients. The optimal cutoff point of 42 years old found in the HRC was exploratory and tested in the unselected cohort. However, this cutoff point should be further validated in a larger scaled study in the future.

We found that family history of breast, gynecologic, or colon cancer is actually predictive of germline mutations while family history of PDAC was not predictive, in agreement with previous reports (15, 16) while family history of PDAC was not more prevalent on patients with germline mutations. These data suggest that the germline mutations associated with familial PDAC may not be included in the panel of established PDAC susceptibility genes.

Regarding clinical outcomes, we have found that patients in the HRC without mutations had a shorter OS than those with mutations only in patients older than 60 years. An explanation for this might be that PDAC patients younger than 60 years who tested negative may still harbor a germline mutation in genes for which they were not tested. With respect to metastatic patients, only PDAC patients from the GC with germline mutations were found to have better clinical outcomes compared with those without mutations. Similar to the explanation above, patients from the HRC do not have substantial differences based on their mutational status because those with negative results may also have mutations, which have not been tested for.

The use of a discovery cohort with a large number of young-onset PDAC patients and a validation cohort with patients who were not at high risk for mutations based on personal and/or family history is strength of the study. A major limitation of this study is the use of panels with differing genes analyzed in individual patients of the HRC and not all patients were tested for all genes. The reason for this is that single-gene analysis of one to three genes was standard of care until 2013 when multigene panels became available at our institution. We compared demographic data from patients tested for individual genes or with panels and found no differences between the two groups (Supplementary Table S8), suggesting that this factor should not affect final results. Additionally, panel testing did not yield significantly higher detection of pathogenic mutations, but did significantly increase the likelihood to identify VUS. Future studies would ideally include comprehensive analysis of an entire cohort for the same set of genes to validate these findings. Similarly, variants identified in the GC cohort were subject to classification on the research platform involving public database review and in silico analysis but were not subjected to the stringent clinical variant interpretation as those tested via standard clinical genetic testing in the HRC cohort.

Higher detection rates of germline mutations will potentially affect therapy choices for patients with PDAC. BRCA-associated PDAC have been shown to have higher sensitivity to platinum therapy and poly (ADP-ribose) polymerase (PARP) inhibitors (19) while tumors with genetic defects in MMR genes have greater susceptibility to immune-checkpoint blocking agents (20). Moreover, unaffected family members may be identified and receive genetic counseling and risk assessment to enter pancreatic cancer screening programs and undergo cancer surveillance and prevention (21, 22).

No potential conflicts of interest were disclosed.

Conception and design: S.A. Bannon, M.F. Montiel, J.B. Goldstein, W. Dong, A. Maitra, M.H. Katz, E. Vilar, F. McAllister

Development of methodology: S.A. Bannon, M.F. Montiel, J.B. Goldstein, E. Vilar, F. McAllister

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.A. Bannon, M.F. Montiel, J.B. Goldstein, M.E. Mork, M. Hasanov, G.R. Varadhachary, M.H. Katz, A. Futreal, D.R. Fogelman, F. McAllister

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.A. Bannon, M.F. Montiel, J.B. Goldstein, W. Dong, E. Borras, M. Hasanov, A. Maitra, M.H. Katz, L. Feng, A. Futreal, F. McAllister

Writing, review, and/or revision of the manuscript: S.A. Bannon, M.F. Montiel, J.B. Goldstein, W. Dong, M.E. Mork, E. Borras, A. Maitra, M.H. Katz, A. Futreal, E. Vilar, F. McAllister

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.F. Montiel, J.B. Goldstein, M. Hasanov, F. McAllister

Study supervision: F. McAllister

Dr. F. McAllister is a Paul Calabresi K12 clinical scholar (NCI grant awarded to MDACC K12CA088084-16A1) and V Foundation Scholar. Dr. F. McAllister and Dr. A. Maitra have also received support from philanthropic contributions to the University of Texas MD Anderson Pancreatic Cancer Moon Shots Program.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.