eAppendix 1. Pancreatic Cancer Patient Research Registry
eAppendix 2. Reference Controls
eAppendix 3. Custom Sequencing of Mayo Clinic Pancreatic Cancer Cases
eAppendix 4. Bioinformatics Analysis of Sequencing Data
eAppendix 5. Cleaning and Filtering of Reference Control Sequencing Data
eAppendix 6. Statistical Analysis
eTable 1. Genes in the Custom QIAseq Panel Evaluated for Presence of Mutations in Patients With Pancreatic Cancer
eTable 2. Mutations in Individual Genes Among Cases With Pancreatic Cancer and gnomAD Controls of Different Racial and Ethnic Populations
eTable 3. Allele Count (AC) of Mutations in Mayo Clinic Cases With Pancreatic Cancer and Two Samples of Reference Controls
eTable 4. Frequency of Mutation Carriers by Gene Among Pancreatic Cancer Cases With Personal and Family History of Cancers
eTable 5. Frequency of Mutations in Individual Genes Among Cases With Pancreatic Cancer and ExAC_non_TCGA Reference Controls From Different Racial and Ethnic Populations
eTable 6. Associations Between Panel Gene Mutations and Pancreatic Cancer Using ExAC_non_TCGA Reference Controls
eTable 7. Frequency of Mutation Frequencies for Individual Genes Among Non-Hispanic White Cases With Pancreatic Cancer by Personal and Family History of Cancers
eTable 8. Comparisons of Mutation Frequencies Between Non-Hispanic White Cases With Pancreatic Cancer and gnomAD_NFE Controls
eTable 9. Associations of Panel Gene Mutations With Pancreatic Cancer Based on Non-Hispanic White Patients With Pancreatic Cancer and ExAC_NFE_non-TCGA Reference Controls
eTable 10. Comparisons of Mutation Frequencies Between Patients With Pancreatic Cancer as the Initial Cancer and gnomAD Controls
eTable 11. Comparisons of Mutation Frequencies Between Cases With Pancreatic Cancer With Family History of Pancreatic, Breast, Ovarian, Colorectal, or Gynecologic Cancer and gnomAD Controls
eTable 12. Comparisons of Mutation Frequencies Between Cases With Pancreatic Cancer Without Family History of Pancreatic, Breast, Ovarian, Colorectal, or Gynecologic Cancer and gnomAD Controls
eTable 13. Comparisons of Mutation Frequencies Between Cases With Pancreatic Cancer, Excluding Those With Family History of Pancreatic Cancer, and gnomAD Controls
eTable 14. Comparisons of Mutation Frequencies Between Cases With Pancreatic Cancer, Excluding Those With Family History of Breast Cancer, and gnomAD Controls
eTable 15. Comparisons of Mutation Frequencies Between Cases With Pancreatic Cancer, Excluding Those With Family History of Ovarian Cancer, and gnomAD Controls
eTable 16. Comparisons of Mutation Frequencies Between Cases With Pancreatic Cancer, Excluding Those With Family History of colorectal cancer, and gnomAD Controls
eTable 17. Comparisons of Mutation Frequencies Between Cases With Pancreatic Cancer, Excluding Those With Family History of Gynecologic Cancer, and gnomAD Controls
eTable 18. Associations of Mutations in Six Pancreatic Cancer Genes With Personal and Family History of Cancers
eTable 19. Associations Between Mutations in Each Panel Gene and Age of Diagnosis of Pancreatic Cancer Cases
eTable 20. Demographic and Clinical Characteristics of Patients With Pancreatic Cancer Included in the Survival Analysis
eFigure. Kaplan-Meier Analysis of Overall Survival for Pancreatic Cancer Cases With Mutations (Carrier) in the 6 Genes Associated With Pancreatic Cancer and Cases Without Mutations (Non-Carrier)
Customize your JAMA Network experience by selecting one or more topics from the list below.
Hu C, Hart SN, Polley EC, et al. Association Between Inherited Germline Mutations in Cancer Predisposition Genes and Risk of Pancreatic Cancer. JAMA. 2018;319(23):2401–2409. doi:10.1001/jama.2018.6228
Are there germline mutations in cancer predisposition that are associated with pancreatic cancer?
In a case-control study that included 3030 patients with pancreatic cancer and 123 136 reference controls, 6 genes were independently associated with pancreatic cancer, with odds ratios between 2.58 and 12.33 after correction for multiple comparisons. In aggregate, these genes were observed in 5.5% of patients with pancreatic cancer.
Six genes were identified that were associated with pancreatic cancer; further research is needed for replication in other populations.
Individuals genetically predisposed to pancreatic cancer may benefit from early detection. Genes that predispose to pancreatic cancer and the risks of pancreatic cancer associated with mutations in these genes are not well defined.
To determine whether inherited germline mutations in cancer predisposition genes are associated with increased risks of pancreatic cancer.
Design, Setting, and Participants
Case-control analysis to identify pancreatic cancer predisposition genes; longitudinal analysis of patients with pancreatic cancer for prognosis. The study included 3030 adults diagnosed as having pancreatic cancer and enrolled in a Mayo Clinic registry between October 12, 2000, and March 31, 2016, with last follow-up on June 22, 2017. Reference controls were 123 136 individuals with exome sequence data in the public Genome Aggregation Database and 53 105 in the Exome Aggregation Consortium database.
Individuals were classified based on carrying a deleterious mutation in cancer predisposition genes and having a personal or family history of cancer.
Main Outcomes and Measures
Germline mutations in coding regions of 21 cancer predisposition genes were identified by sequencing of products from a custom multiplex polymerase chain reaction–based panel; associations of genes with pancreatic cancer were assessed by comparing frequency of mutations in genes of pancreatic cancer patients with those of reference controls.
Comparing 3030 case patients with pancreatic cancer (43.2% female; 95.6% non-Hispanic white; mean age at diagnosis, 65.3 [SD, 10.7] years) with reference controls, significant associations were observed between pancreatic cancer and mutations in CDKN2A (0.3% of cases and 0.02% of controls; odds ratio [OR], 12.33; 95% CI, 5.43-25.61); TP53 (0.2% of cases and 0.02% of controls; OR, 6.70; 95% CI, 2.52-14.95); MLH1 (0.13% of cases and 0.02% of controls; OR, 6.66; 95% CI, 1.94-17.53); BRCA2 (1.9% of cases and 0.3% of controls; OR, 6.20; 95% CI, 4.62-8.17); ATM (2.3% of cases and 0.37% of controls; OR, 5.71; 95% CI, 4.38-7.33); and BRCA1 (0.6% of cases and 0.2% of controls; OR, 2.58; 95% CI, 1.54-4.05).
Conclusions and Relevance
In this case-control study, mutations in 6 genes associated with pancreatic cancer were found in 5.5% of all pancreatic cancer patients, including 7.9% of patients with a family history of pancreatic cancer and 5.2% of patients without a family history of pancreatic cancer. Further research is needed for replication in other populations.
Quiz Ref IDCancer predisposition gene testing is useful for identifying individuals who may benefit from screening, prevention, and early detection of breast, ovarian, and colorectal cancer1,2 and may be beneficial for individuals at risk of pancreatic cancer.3,4 Family members of those with germline predisposition gene mutations may also benefit from enhanced cancer screening and prevention strategies. For instance, screening of CDKN2A (RefSeq NM_000077 .4) mutation carriers has been associated with early detection of resectable pancreatic tumors.3
Epidemiologic studies have shown that 10% to 20% of pancreatic cancers are associated with an inherited component.5 Deleterious mutations in BRCA2 (RefSeq NM_000059 .3), PALB2 (RefSeq NM_024675 .3), and CDKN2A cancer predisposition genes have been detected in families of patients with pancreatic cancer.6-8 Germline mutations in BRCA1 (RefSeq NM_007294 .3) have also been associated with an increased risk of pancreatic cancer in families (relative risk, 2.26; 95% CI, 1.26-4.06),9 similar to mutations in mismatch repair genes in families of patients with Lynch syndrome (hazard ratio of cumulative increased risk, 8.6; 95% CI, 4.7-15.7).10 Germline mutations have also been observed in 7% of patients unselected for family history of pancreatic cancer11 and in 3.9% of 854 patients with pancreatic adenocarcinoma.12
Quiz Ref IDThis study used a custom cancer predisposition gene panel developed for hereditary cancer genetic testing to assess the prevalence of deleterious germline mutations among patients with pancreatic cancer in 21 predisposition genes implicated in susceptibility to solid tumors (eTable 1 in the Supplement).1,13,14 DNA for panel testing was obtained from a series of 3030 patients with pancreatic cancer from a Mayo Clinic pancreas cancer registry, and DNA sequence data for the same predisposition genes were obtained from publicly available Genome Aggregation Database (gnomAD) and Exome Aggregation Consortium (ExAC) reference control groups.15,16 Associations between mutations in each gene and pancreatic cancer were evaluated to establish a defined subset of genes that confer susceptibility to pancreatic cancer.
The participants in this study were recruited into the Mayo Clinic Biospecimen Resource for Pancreas Research, a prospective patient registry focused on pancreatic cancer.17 Patients were identified and invited to participate at the time of diagnosis. Detailed information about the process of recruitment, biospecimen collection, and maintenance of the registry is provided in eAppendix 1 in the Supplement. All participants diagnosed as having pancreatic ductal adenocarcinoma, who were recruited from October 12, 2000, through March 31, 2016, with available genomic DNA extracted from peripheral blood lymphocyte samples were included in the study. Patients completed questionnaires on demographic and clinical characteristics and family history of cancer. Race was self identified as American Indian/Alaskan Native, Asian/Asian American, black/African American, Native Hawaiian/other Pacific Islander, white, and multiracial. Ethnicity was self identified as Hispanic/Latino or non-Hispanic/non-Latino. The study was approved by the Mayo Clinic Institutional Review Board. All patients provided written informed consent for research genetic testing. Results have not been systematically disclosed to participants.
Quiz Ref IDReference control data were obtained from the public gnomAD (http://gnomad.broadinstitute.org/),15 which contains exome sequencing data from 123 136 unrelated individuals sequenced as part of various disease-specific and population genetic studies (eAppendix 2 in the Supplement). The gnomAD data set was generated using multiple exome capture methods and sequencing chemistries and was subset to racial and ethnic groups including African/African American, Hispanic, Asian, and non-Finnish European for this study. A second reference control data set of 53 105 germline exomes from ExAC (http://exac.broadinstitute.org),15,16 excluding samples from cancer cases from The Cancer Genome Atlas (TCGA) project (ExAC non-TCGA) was used to assess consistency in results (eAppendix 2). The ExAC non-TCGA data set was generated using multiple exome sequencing methods and was also subset to racial and ethnic groups including African/African American, Hispanic, Asian, and non-Finnish European for this study. All reference control groups may have included a small number of pancreatic cancer cases because individuals with cancer were not excluded.
Genomic DNA samples were subjected to multiplex polymerase chain reaction using a QIAseq (Qiagen Inc)18 custom panel of target regions covering all coding regions and consensus splice sites from 21 cancer predisposition genes: ATM (RefSeq NM_000051 .3), BARD1 (RefSeq NM_000465 .3), BRCA1, BRCA2, BRIP1 (RefSeq NM_032043 .2), CDH1 (RefSeq NM_004360 .4), CDKN2A, CHEK2 (RefSeq NM_007194 .3), FANCC (RefSeq NM_000136 .2), MLH1 (RefSeq NM_000249 .3), MRE11A (RefSeq NM_005591 .3), MSH2 (RefSeq NM_000251 .2), MSH6 (RefSeq NM_000179 .2), NBN (RefSeq NM_002485 .4), NF1 (RefSeq NM_001042492 .2), PALB2, PMS2 (RefSeq NM_000535 .6), PTEN (RefSeq NM_000314 .6), RAD51C (RefSeq NM_058216 .2), RAD51D (RefSeq NM_001142571), and TP53 (RefSeq NM_000546 .5) (eAppendix 3 in the Supplement). Libraries derived from each DNA sample were individually bar coded by dual indexing. Sequencing was performed on a HiSeq 4000 with 150-bp paired-end reads of 768 pooled libraries per lane. Median sequence read depth was 200×. These genes were selected based on inclusion in commercial hereditary cancer genetic testing panels as known or candidate predisposition genes for several solid tumors including breast, ovarian, endometrial, colorectal, or pancreatic cancers.1,13,14,19 Results from 19 genes are presented because no mutations were identified in RAD51D or PTEN.
FASTQ files of DNA sequences were generated for each sample based on unique dual indexes. Reads were trimmed with Cutadapt version 1.1020 and aligned with BWA-MEM version 0.7.10.21 Sequence realignment, recalibration, haplotype calling, and depth of coverage were conducted using Genome Analysis Toolkit version 3.4-46 (University of Birmingham). A minimum quality threshold22 of Q20 was applied to identify cases eligible for analyses. Annotation of variants from cases with pancreatic cancer and from gnomAD and ExAC non-TCGA reference controls16,23 (eAppendix 4 in the Supplement) was provided through the Biological Reference Repository tool kit,24 leveraging dbNSFP version 3.0,25 ClinVar,26 and CAVA.27 Variants were viewed and filtered with VCF-Miner.28 All loss-of-function variants (nonsense, frameshift, consensus splice sites [±1 or 2]) and any intronic or missense variants defined as pathogenic or likely pathogenic in ClinVar in patients with pancreatic cancer were validated by Sanger sequencing (eAppendix 3 in the Supplement). Variants in pancreatic cancer cases and in both gnomAD and ExAC non-TCGA reference controls were filtered using established approaches (eAppendix 5 in the Supplement).1
The primary outcome was case-control status, where case status was assigned to all individuals with pancreatic cancer in the Mayo Clinic registry. All individuals in the gnomAD and ExAC data sets were controls. A secondary outcome was overall survival after diagnosis of pancreatic cancer. Vital status was ascertained by using personal/family correspondence, a study follow-up questionnaire, medical records, or an external service (LexisNexis Accurint). Duration of overall survival was calculated from the date of diagnosis at a Mayo Clinic location until date of death, date last known alive, or date of censorship of June 22, 2017.
Analyses were based on patients (Table 1 and eTable 2 in the Supplement) with good-quality sequence data. Frequencies of mutations in individual genes were calculated overall and by patient characteristics (personal history of other cancer; family history of breast, colorectal, ovarian, gynecologic, and pancreatic cancer). Associations between mutations in each gene and pancreatic cancer were assessed by logistic regression, comparing combined mutation frequencies by gene in patients with pancreatic cancer with frequencies in gnomAD reference controls after weighting for the relative frequency of racial and ethnic populations. Association analysis included patients with pancreatic cancer after exclusion of patients with missing race information or other race (multiracial, American Indian/Alaskan Native, and Native Hawaiian/other Pacific Islander) (Table 1). Confidence intervals were estimated by the profile likelihood method. Sensitivity analyses using ExAC non-TCGA reference controls, selected for race/ethnicity, as with gnomAD controls, were undertaken to assess consistency in results (eAppendix 6 in the Supplement). Sensitivity analyses were conducted to account for the influence of personal and family history of pancreatic, breast, ovarian, colorectal, and gynecologic cancer on the associations between each gene and pancreatic cancer and to evaluate associations between genes and pancreatic cancer using non-Hispanic white cases and gnomAD non-Finnish European and ExAC non-Finnish European non-TCGA reference controls. For comparisons within individual populations, odds ratios (ORs) and corresponding 95% confidence intervals were estimated by inverting Fisher exact test.29 Significance of associations was adjusted for multiple testing by Bonferroni correction. Associations between mutation status in each predisposition gene and age at diagnosis were tested using the Kolmogorov-Smirnov test,30 and associations with patient characteristics were evaluated using logistic regression adjusted for age at diagnosis (eAppendix 6). All analyses were performed with R software version 3.4.2 (R Project for Statistical Computing). All statistical tests were 2-sided, and an adjusted P<.05 was considered statistically significant.
The patient population for survival analysis was restricted to the subset of 2698 adenocarcinoma cases diagnosed at a Mayo Clinic location within 3 months (≤92 days) of an initial diagnosis. This date of diagnosis was defined as either (1) the date of tissue-based diagnosis for those with pathology-proven disease (97%) or (2) the date of first clinical diagnosis for patients without pathology information (3%) and was used to avoid immortal time bias.31 The association between mutations in pancreatic cancer predisposition genes and overall survival was evaluated using Cox regression models adjusted for age at diagnosis, sex, and disease staging (resectable, locally advanced, or metastatic). The significance of associations with survival was estimated by likelihood ratio test (eAppendix 6 in the Supplement). The proportional hazards assumption was tested using the residuals from the Cox model.32 All analyses were performed with R software version 3.4.2.
The participation rate in the Mayo Clinic Biospecimen Resource for Pancreas Research was 65.6%. High-quality sequencing data were obtained for 3030 of 3046 patients with pancreatic cancer in this case series (eTable 2 in the Supplement). Among the 3030 participants, 2591 (85.5%) consented to registry participation within 30 days of diagnosis. The sample was 95.6% non-Hispanic white and 43.2% female, with 37.2% of patients receiving diagnoses at age 70 years or older. The mean age at diagnosis was 65.3 (SD, 10.7) years, and approximately 5.5% of cases had an additional personal history of breast, ovarian, colorectal, or nonovarian gynecologic cancers (Table 1). While 11.3% of patients had a family history (among first- and second-degree relatives) of pancreatic cancer, 22.3% reported family histories of breast cancer, 16.9% colorectal cancer, and 5.0% ovarian cancer.
In 19 of 21 candidate pancreatic cancer predisposition genes, 253 deleterious germline mutations were identified in 249 patients (8.2%; 95% CI, 7.26%-9.25%) (Table 1 and Table 2; eTables 2 and 3 in the Supplement). While ATM had the highest prevalence of mutations (n = 69) (2.28%; 95% CI, 1.78%-2.87%), mutations were also frequently observed in BRCA2 (1.95%; 95% CI, 1.49%-2.50%), CHEK2 (1.09%; 95% CI, 0.75%-1.53%; excluding the low-risk p.Ile157Thr missense variant), BRCA1 (0.59%; 95% CI, 0.35%-0.94%), PALB2 (0.40%; 95% CI, 0.20%-0.69%), and CDKN2A (0.33%; 95% CI, 0.16%-0.61%). Among the 59 patients with BRCA2 mutations, only 3 carried the Ashkenazi Jewish c.5946delT (6174delT) founder mutation. Smaller numbers of mutations were observed in candidate pancreatic cancer predisposition genes, including FANCC (0.26%; 95% CI, 0.11%-0.52%) and TP53 (0.20%; 95% CI, 0.07%-0.43%). Germline mutations in the MLH1, MSH2, PMS2, and MSH6 mismatch repair genes were detected in aggregate in 0.50% (95% CI, 0.28%-0.82%) of study participants (Table 2).
Patients with pancreatic cancer and mutations were more likely to have personal and family histories of other cancers (Table 1 and eTable 4 in the Supplement). In terms of personal history of cancer, 65 of 513 patients (12.3%; 95% CI, 9.9%-15.9%) with at least 1 other cancer in addition to pancreatic cancer had mutations in the panel genes. Additionally, mutations were detected in 43 of the 343 patients (12.9%; 95% CI, 9.2%-16.5%) with at least 1 first- or second-degree relative with pancreatic cancer. Mutations were also identified in 75 of 675 patients (11.3%; 95% CI, 8.8%-13.7%) with a family history of breast cancer, 21 of 152 patients (13.8%; 95% CI, 8.8%-20.3%) with a family history of ovarian cancer, and 45 of 513 patients (8.9%; 95% CI, 6.5%-11.6%) with a family history of colorectal cancer (Table 1). Overall, 25.7% (95% CI, 20.4%-31.5%) of all mutations (65/253) were associated with multiple primary cancers, 17.8% (95% CI, 13.3%-23.1%) were associated with a family history of pancreatic cancer, and 30.0% (95% CI, 23.5%-36.1%) were associated with a family history of breast cancer (eTable 4). In addition, 124 of 253 mutations (49%; 95% CI, 42.7%-55.3%) were identified in patients with a family history of at least 1 common epithelial cancer (pancreatic, breast, ovarian, gynecologic, or colorectal) (eTable 4).
Six genes were significantly associated with pancreatic cancer compared with gnomAD controls. These included CDKN2A, with mutations in 0.30% of cases and 0.02% of controls (OR, 12.33; 95% CI, 5.43-25.61); TP53, with mutations in 0.20% of cases and 0.02% of controls (OR, 6.70; 95% CI, 2.52-14.95); MLH1, with mutations in 0.13% of cases and 0.02% of controls (OR, 6.66; 95% CI, 1.94-17.53); BRCA2, with mutations in 1.90% of cases and 0.30% of controls (OR, 6.20; 95% CI, 4.62-8.17); ATM, with mutations in 2.30% of cases and 0.37% of controls (OR, 5.71; 95% CI, 4.38-7.33); and BRCA1, with mutations in 0.60% of cases and 0.20% of controls (OR, 2.58; 95% CI, 1.54-4.05) (Table 3). Similar results were obtained using the ExAC non-TCGA reference controls for CDKN2A, ATM, MLH1, BRCA2, and BRCA1 (eTables 5 and 6 in the Supplement), while TP53 exhibited a statistically significant but attenuated association (OR, 3.03; 95% CI, 1.14-6.74).
NBN and BRIP1 were not significantly associated with pancreatic cancer, but the numbers of mutations in these genes were too low to allow for definitive evaluation of associations with pancreatic cancer. In contrast, CHEK2 was associated with little or no risk of pancreatic cancer (Table 3 and eTable 6 in the Supplement). Similar frequencies of mutations in each gene by phenotypic category and similar associations between 5 of the predisposition genes (other than MLH1) and pancreatic cancer were observed for non-Hispanic white cases (n = 2896), which account for the majority of the study population (eTables 7, 8, and 9 in the Supplement). Sensitivity analyses yielded similar OR estimates for pancreatic cancer for the 6 predisposition genes other than MLH1 when restricting to patients with pancreatic cancer as the first cancer diagnosis (eTable 10 in the Supplement). Similarly, no substantial changes in associations between predisposition gene mutations and pancreatic cancer were observed when restricting analyses to patients with a family history of common epithelial cancers (pancreatic, breast, ovarian, colorectal, and endometrial) (eTable 11 in the Supplement) or to patients without a family history of these cancers (eTables 12, 13, 14, 15, 16, and 17 in the Supplement), except for reduced risk for MLH1 following exclusion of patients with a family history of colorectal cancer (eTables 12 and 16). Similarly, exclusion of patients and reference controls with Ashkenazi Jewish founder mutations in BRCA1, BRCA2, and CHEK2 had little influence on results. Thus, 6 genes significantly associated with pancreatic cancer were designated as pancreatic cancer predisposition genes.
Overall, 167 of 3030 patients (5.5%; 95% CI, 4.7%-6.4%) with pancreatic cancer had deleterious mutations in 1 of the 6 predisposition genes: CDKN2A, TP53, MLH1, BRCA2, ATM, and BRCA1 (Table 4 and eTable 18 in the Supplement). Among all tested patients, 27 of 343 patients (7.9%; 95% CI, 5.3%-11.2%) with a family history of pancreatic cancer and 140 of 2687 patients (5.2%; 95% CI, 4.4%-6.1%) with no family history of pancreatic cancer had a mutation in 1 of the 6 predisposition genes (P = .06) (Table 4; eTable 18). Thus, family history of pancreatic cancer did not inform on the presence of 83.8% of mutations. In addition, 40 of 495 patients (8.1%; 95% CI, 5.8%-10.8%) with another primary cancer diagnosis prior to pancreatic cancer had mutations in these genes (Table 4). Although prior primary cancer was significantly associated with mutation status (OR, 1.67; 95% CI, 1.17-2.48; P = .009), 76% of patients with mutations (127/167) did not exhibit this phenotype. Overall, significant associations were observed between mutations in the 6 predisposition genes combined and advanced stage of disease (resectable: 48/850; locally advanced: 50/1115; and metastatic: 67/1056; P = .04), personal history of other cancers (OR, 1.67; 95% CI, 1.17-2.48; P = .009), family history of breast cancer (OR, 1.58; 95% CI, 1.11-2.23; P = .01), or family history of common epithelial cancers (OR, 1.40; 95% CI, 1.01-1.92; P = .04) (Table 4). Patients with mutations in these 6 genes also had a significantly earlier mean age of diagnosis (62.5 vs 65.5 years; P < .001) (Table 4). In particular, mutations in BRCA2 alone were significantly associated with an earlier age at diagnosis of pancreatic cancer (mean age, 60.5 years vs 63.3 years for noncarriers; P = .01) (eTable 19 in the Supplement).
When comparing characteristics of mutation carriers and noncarriers by individual gene, only patients with deleterious mutations in CDKN2A were more likely to have a family history of pancreatic cancer (OR, 7.91; 95% CI, 2.19-28.57; adjusted P = .005). Similarly, patients with mutations in BRCA2 (OR, 2.07; 95% CI, 1.19-3.50; adjusted P = .04) were more likely to have a family history of breast cancer (eTable 18 in the Supplement).
The median overall survival for patients with mutations in the 6 genes associated with pancreatic cancer was 13.6 months (95% CI, 11.5-15.7 months), whereas overall survival for patients without mutations was 11.4 months (95% CI, 10.8-12.1 months). The association between mutation status in these genes and survival was not statistically significant (hazard ratio, 0.86; 95% CI, 0.72-1.02; P = .09) (eFigure in the Supplement). There was no evidence of deviation from proportional hazards for the mutation carrier status (χ2 = 0.03; P = .87).
Quiz Ref IDIn this case-control study, mutations in 6 genes (ATM, BRCA1, BRCA2, CDKN2A, MLH1, and TP53) were found to be associated with pancreatic cancer and were found in 5.5% of patients with pancreatic cancer. Mutations in CDKN2A yielded the highest risks of pancreatic cancer, although the frequency of mutations was low (0.33%). Mutations in ATM, a gene that encodes a multifunctional protein involved in regulating the cellular response to DNA damage,33 were significantly associated with pancreatic cancer. Homozygous ATM mutations cause ataxia-telangiectasia,34 and heterozygous ATM mutations have been associated with moderate risks of breast cancer1 but not pancreatic cancer.35 In the current study, no substantial change in ATM associations were observed when excluding individuals with a personal or family history of breast cancer, suggesting that the association with pancreatic cancer was independent of breast cancer effects. Whether missense mutations, such as in ATM c.7271T>G (p.Val2424Gly), which has been associated with substantially increased risk of breast cancer (OR, 8.0; 95% CI, 2.3-27.4),36 have alternative effects on pancreatic cancer risk remains to be determined. Mutations in TP53 were also significantly associated with pancreatic cancer, but it was not known if the patients carrying these mutations exhibited Li-Fraumeni syndrome phenotypes or had a family history of Li-Fraumeni syndrome.
These results were consistent with a recent study of 854 patients with pancreatic adenocarcinoma that identified mutations in these 6 genes and PALB2 in 3.5% of patients.12 Although mutations in PALB2 are thought to increase risk of pancreatic cancer,7,8 the current study did not find a significant association after Bonferroni correction. CHEK2 mutations were also not significantly associated with pancreatic cancer, even though mutations were frequently observed. It may be that mutations in CHEK2 and other cancer predisposition genes can provide information about risk of other cancers in patients and their relatives.
Given the high frequency of predisposing mutations in this series of patients (>5%) and the absence of effective predictors of mutations, genetic testing of all patients with pancreatic cancer by panel tests may be warranted. In addition, genetic testing and identification of germline mutations may have implications for the relatives of patients with pancreatic cancer because of risks of pancreatic and other cancers. Overall, genetic testing guidelines for patients with pancreatic cancer and for their unaffected relatives must be developed. Currently the National Comprehensive Cancer Network does not provide guidelines for selection of patients with pancreatic cancer for multigene panel testing,37 instead focusing on patients with pancreatic cancer in the context of hereditary breast and ovarian cancer. The best predictors of mutations in patients with pancreatic cancer in the current study were a personal history of another primary cancer, a personal history of breast cancer, and a family history of 1 or more first- or second-degree relatives with epithelial cancers (pancreatic, breast, ovarian, endometrial, or colorectal). However, the specificity for mutations was too low for effective selection of patients for clinical genetic testing.
Although patients with BRCA1 and BRCA2 predisposing mutations may derive therapeutic benefit from testing because tumors may display sensitivity to platinum agents or poly adenosine diphosphate–ribose polymerase (PARP) inhibitors,38,39 it remains to be determined whether patients with germline or somatic mutations in other predisposition genes will benefit from these and other targeted therapies. Benefits of panel testing may also extend to cancer screening and prevention. The International Cancer of the Pancreas Screening (CAPS) Consortium4 and the American College of Gastroenterology (ACG) guidelines,40 based on expert opinion, currently recommend imaging surveillance for individuals with greater than 5% lifetime risk of pancreatic cancer due to mutations in STK11 (RefSeq NM_000455 .4), CDKN2A, and hereditary pancreatitis genes; individuals with mutations in BRCA1, BRCA2, ATM, PALB2, or mismatch repair genes and a first- or second-degree relative with pancreatic cancer; and individuals with a first-degree relative with pancreatic cancer. Thus, the surveillance guidelines already include all of the predisposition genes identified in this study. In addition, the value of surveillance based on germline mutations, but not family history alone, has been empirically demonstrated,3,4,41,42 supporting the potential importance of mutation testing. The genes included in the CAPS and ACG guidelines are consistent with results from the current study except that the moderate risks associated with BRCA1 mutations may not be sufficient to warrant this level of intervention. Given the high case-fatality rate for pancreatic cancer, testing for inherited cancer susceptibility may identify candidates for participation in innovative approaches to screening and prevention.
This study has several limitations. First, public reference controls were used to estimate the prevalence of each of the 21 cancer predisposition genes in race/ethnicity–matched general populations. However, extensive data cleaning and filtering were used in an effort to normalize the pancreatic cancer cases and the control data. These large reference control data sets were needed because study-matched control data sets are generally not of sufficient size for association studies because of the rarity of individual deleterious mutations in the general population. In this study, both the ExAC non-TCGA and gnomAD reference data sets resulted in very similar findings. Despite partial overlap in these data sets, this consistency strongly suggests that the pancreatic cancer predisposition genes identified in this study are drivers of pancreatic cancer risk in the general population. Quiz Ref IDSecond, the custom panel of 21 genes used in this study did not account for all possible cancer predisposition genes, and the possibility remains that other untested genes may contribute to risk of pancreatic cancer. Third, there are a number of variants of uncertain significance in genes with insufficient data for classification as deleterious or neutral. Fourth, because cases were identified from Mayo Clinic populations in Minnesota, Arizona, and Florida and were younger and less likely to be black or Hispanic than pancreatic cancer patients included in the Surveillance, Epidemiology, and End Results registry, study results may lack generalizability. Further analyses in more racially and ethnically diverse populations are necessary to identify other potential pancreatic cancer susceptibility genes. Fifth, the study did not have sufficient information to estimate lifetime probability of cancer (penetrance) in carriers of the predisposition gene mutations.
In this case-control study, mutations in 6 genes associated with pancreatic cancer were found in 5.5% of all pancreatic cancer patients, including 7.9% of patients with a family history of pancreatic cancer and 5.2% of patients without a family history of pancreatic cancer. Further research is needed for replication in other populations.
Corresponding Author: Fergus J. Couch, PhD, Department of Laboratory Medicine and Pathology, Mayo Clinic, Stabile 2-42, 200 First St SW, Rochester, MN 55905 (email@example.com).
Accepted for Publication: May 16, 2018.
Author Contributions: Drs Petersen and Couch 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. Drs Petersen and Couch contributed equally to this study.
Concept and design: Polley, Bamlet, Samara, McWilliams, Petersen, Couch.
Acquisition, analysis, or interpretation of data: Hu, Hart, Polley, Gnanaolivu, Shimelis, Lee, Lilyquist, Na, Moore, Antwi, Bamlet, Chaffee, DiCarlo, Wu, Kasi, McWilliams, Petersen, Couch.
Drafting of the manuscript: Hu, Hart, Shimelis, Na, Moore, Bamlet, Chaffee, Kasi, Petersen, Couch.
Critical revision of the manuscript for important intellectual content: Hu, Hart, Polley, Gnanaolivu, Shimelis, Lee, Lilyquist, Antwi, Bamlet, DiCarlo, Wu, Samara, Kasi, McWilliams, Petersen.
Statistical analysis: Hu, Hart, Polley, Gnanaolivu, Shimelis, Na, Moore, Bamlet, Couch.
Obtained funding: McWilliams, Petersen, Couch.
Administrative, technical, or material support: Lee, Chaffee, DiCarlo, Wu, Samara, Petersen, Couch.
Supervision: Petersen, Couch.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Drs Wu, DiCarlo, and Samara are employees of Qiagen Inc. Dr McWilliams reports advisory board membership for Bristol-Myers Squibb, Ipsen, and Merrimack. No other disclosures were reported.
Funding/Support: The study was funded by National Institutes of Health grants R01CA192393 and R01CA97075, the Mayo Clinic Specialized Program of Research Excellence in Pancreatic Cancer (grant P50CA102701), the Mayo Clinic Department of Laboratory Medicine and Pathology Research Funds, the Mayo Clinic Center for Individualized Medicine, the Rolfe Pancreatic Cancer Foundation, and the Vernon F. and Mae E. Thompson Charitable Fund.
Role of the Funder/Sponsor: The funding agencies 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; or decision to submit the manuscript for publication.
Previous Presentation: This study was presented in part at the American Society for Human Genetics annual meeting; October 17-21, 2017; Orlando, Florida.