Outcome of Pancreatic Cancer Surveillance Among High-Risk Individuals Tested for Germline Mutations in BRCA1 and BRCA2

Pancreatic ductal adenocarcinoma (PDAC) remains a highly lethal disease and is the third leading cause of cancer-related death in the United States (1). SEER data placed the age-adjusted incidence rate at 12.5 per 100,000 in 2010–2014, with the mortality rate nearing that at 10.9 per 100,000 (2). This poor outcome is attributed to the advanced stage of this disease that is almost always present when patients present with symptoms of PDAC such as abdominal pain or jaundice. Screening for some other gastrointestinal cancers can be effective in improving cancer mortality, as has been shown for colorectal cancer. While the approximately 30% reduction in colorectal cancer mortality over the last several decades is attributable to several factors, there is little doubt that screening programs have made a significant impact (3).

There is some evidence that when PDAC is found incidentally, before symptoms develop, survival can be dramatically improved. Patients with small (<2 cm) pancreatic cancers or those without nodal disease have demonstrated improved 5-year survival rates following resection of PDAC (4). Recent studies on long-term surveillance programs show higher resectability rates for asymptomatic screen-detected PDAC compared with symptomatic PDAC and improved 3-year and 5-year survival rates (5, 6).

The detection of PDAC precursor lesions is an area of interest in preventing the development of pancreatic cancer in high risk individuals (HRI). These lesions include intraductal papillary mucinous neoplasms (IPMN) and pancreatic intraepithelial neoplasia (Pan-IN). IPMNs can be observed with cross-sectional imaging such as magnetic resonance cholangiopancreatography (MRCP), and endoscopy such as endoscopic ultrasound, while Pan-IN lesions cannot (7). Family history is an established risk factor for PDAC (8), as are smoking (9), obesity (10, 11), diabetes (12), chronic pancreatitis (13), and pathogenic mutations in several genes, including BRCA1/2 (14–16).

Mutations in BRCA2 are the most common germline mutations influencing risk of PDAC with prevalence estimates ranging from approximately 2% to 19% (17, 18). BRCA1 mutations also contribute, albeit less, to overall prevalence of inherited predisposition to PDAC (18, 19). The relative risk of pancreatic cancer for carriers of BRCA1 mutations has ranged from 0.8 to 4.7 (14, 18, 20–26). In studies of carriers of BRCA2 mutations, the reported range is somewhat higher, from 2.0 to 21, with most results around 3–6 (14, 18, 20–24, 27, 28).

Hoping to identify precursor lesions such as IPMN or early PDAC in individuals at higher risk for development of PDAC, we began a surveillance program within the framework of our larger familial pancreatic tumor registry study at Memorial Sloan Kettering Cancer Center (MSKCC, New York, NY) in 2002. The MSKCC Pancreatic Tumor Registry (29) includes both patients with PDAC and at-risk family members, including those with germline mutations known to increase PDAC risk.

The question leading to this study was whether, among healthy individuals with positive family history of PDAC, those with known pathogenic mutations in BRCA1/2 would have more pancreatic abnormalities compared with individuals who have undergone germline genetic testing with no mutations in BRCA1/2 identified.

The MSKCC Pancreatic Tumor Registry opened enrollment in May 2003. As of June 30, 2017, the Registry had enrolled 554 relatives of patients with pancreatic cancer. Of these, 243 HRIs agreed to take part in surveillance with MRCP, with some participants undergoing CT in select circumstances. Regarding germline genetic test results, 190 of the 554 HRIs have known BRCA1/2 mutation status. This analysis is based on individuals who have both undergone surveillance and have known BRCA1/2 mutation status (Fig. 1).

The eligibility criteria for HRIs changed over the course of the Registry study, reflecting what we and others learned about conducting surveillance in this population. In general, changes resulted in requiring stronger family history and older age at beginning surveillance. Other changes reflected the growing numbers of HRIs undergoing genetic testing for mutations in BRCA1/2 and genes involved in other rare cancer syndromes. When enrollment began in May 2003, we included relatives with ≥1 first-degree relative (FDR) and ≥ 1 other affected relative. In 2011, we changed eligibility requirements to require HRIs to have ≥2 affected FDRs, while individuals with known genetic syndromes were required to have ≥1 FDR or ≥1 SDR (second-degree relative) with PDAC. The main analysis in this article is based on those individuals who had ≥1 FDR with pancreatic cancer, had surveillance at age ≥45 years, and had testing for pathogenic mutations in BRCA1/2 (n = 83). We also report separately on a group of 18 HRIs with known BRCA1/2 mutations and second-degree, but not first-degree, relatives with pancreatic cancer who underwent surveillance at age ≥45.

HRIs in this article were not tested for all PDAC susceptibility genes. However, any HRIs with known pathogenic mutations in other PDAC susceptibility genes were excluded from this analysis.

HRIs are identified in several ways: by study or clinical staff if they are related to a patient with pancreatic cancer; by referral from the MSKCC Clinical Genetics Service; or by self-referral after finding our Registry on the internet. HRIs identified by study and clinical staff are referred to the Clinical Genetics Service for genetic counseling if they have not been seen there previously.

All aspects of the study were performed after approval by the MSKCC Institutional Review Board and conducted in accordance with the U.S. Common Rule. After providing informed written consent, all participants in the Registry are personally interviewed by a trained research study assistant. The interview covers established and potential risk factors for PDAC. Respondents also complete a detailed questionnaire to ascertain personal and family history of cancer. The family history questionnaire includes questions on the birthplace and religion of each grandparent to assess genetically high risk populations such as individuals with Ashkenazi Jewish ancestry. Alternatively, if a participant was already seen by the MSKCC Clinical Genetics Service, the pedigree created during that consultation by a board-certified genetic counselor is used. Follow-up questionnaires are administered approximately every 2 years to update information on lifestyle exposures [such as smoking, body mass index (BMI), and diabetes] and family history of pancreatic and other cancers. For this analysis, we used data from the baseline questionnaire, except for personal and family history of cancer, which was from updated family history questionnaires.

HRIs are offered surveillance with MRCP/CT at enrollment and at 1-year intervals. Those with cysts and/or pancreatic lesions identified on screening may undergo follow-up imaging every 6–9 months at the discretion of the treating physician. We utilize MRCP as the primary surveillance tool because of its sensitivity and because there is no associated radiation exposure. By choice, per treating physician's recommendation, or as follow-up for other medical issues, some participants also had abdominal and pelvic CT scans during the period of surveillance (n = 13/83 in main analysis; 5/18 in secondary analysis). Endoscopic ultrasound and/or surgery are recommended if the MRCP indicates an abnormal finding in the pancreas, such as a large or enlarging pancreatic cyst size ≥ 3 cm, cyst size that increases over surveillance intervals, a dilated main pancreatic duct, or the presence of a solid component in the pancreas on imaging. In December 2012, the surveillance protocol was updated to recommend that HRIs with pathogenic germline mutations in BRCA1/2 undergo annual endoscopic ultrasound evaluations scheduled approximately 6 months after MRCP.

We compared the characteristics of HRIs with and without pathogenic mutations in the BRCA1/2 genes, separately and together; Table 1 summarizes these results with P values from χ² tests. We determined the proportion with an abnormality on MRCP/CT within those groups and whether abnormalities noted were prevalent or incident. We further compared the characteristics of those HRIs with and without findings of IPMN, the most common lesions observed. We described the lesions found and the outcome of follow-up for those with screen-detected IPMNs. We used logistic regression to estimate ORs to evaluate the independent influence of BRCA1/2 mutation status and characteristics associated with presence of IPMNs.

As shown in Table 1, the median age of HRIs with at least one FDR was 54 (range 45–69) and 75% (62/83) of participants were women. Nearly all were white (77/83, 93%) and non-Hispanic (80/83, 96%), and over half (45/83, 54%) had 2 or more Ashkenazi Jewish grandparents. Most participants were college educated (79/83, 95%). Most were of normal weight or were underweight (57/83, 69%), 35% (29/83) had ever smoked cigarettes (only 4% were current smokers), and 6% (5/83) had been diagnosed with diabetes. Nearly half (39/83, 47%) had a personal history of cancer. Most of the reported cancers were breast cancer in women (n = 25). Overall, 20% (17/83) had ≥2 FDRs with pancreatic cancer and 36% (30/83) had 1 FDR and ≥1 SDR. The median length of total time in surveillance for the 83 HRIs was 48 months (range 0–149). Most participants (77%) had three or more screening events (i.e., MRCP, CT, or endoscopic ultrasound) over the course of surveillance.

As shown in Table 1, HRIs who tested positive for pathogenic mutations in BRCA1/2 were more likely to be of Ashkenazi Jewish descent (65% vs. 40%; P = 0.03). They were also more likely to be male (33% vs. 14%; P = 0.05) and to be non-Hispanic (100% vs. 91%; P = 0.04). Those who were BRCA1/2 negative had stronger family history, with 88% having at least 1 FDR and another first- or second-degree relative affected, compared with 34% of those who were BRCA1/2 positive (P < 0.0001); this reflects differences in eligibility requirements as described above. BRCA2 mutations were more common (n = 30) than BRCA1 mutations (n = 18; Table 1). Most mutations (29/48) were the Ashkenazi Jewish founder mutations: c.68_69delAG (n = 6) and c.5266dupC (n = 5) in BRCA1 and c.5946delT (n = 18) in BRCA2. Median length in surveillance was longer for BRCA1/2-negative HRIs (median 60 months; range 0–149) than for BRCA1/2 positive (median 36 months; range 0–127), again reflecting changes in eligibility over time.

Our article focuses on the findings and follow-up of HRIs with presumed IPMN. For the remaining 54 HRIs in the main cohort without identified pancreatic lesions, the median months of follow up was 36 months (35 for the BRCA1/2 positive and 66.5 for the BRCA1/2 negative) and the median number of screening events was 5 (3.5 for the BRCA1/2 positive and 5.5 for the BRCA1/2 negative).

Overall, about one-third (29/83, 35%) of HRIs with known BRCA1/2 mutation status were found to have an abnormality on MRCP/CT (Table 2); nearly all of these (27/83, 33%) were interpreted as branch-duct IPMN. The proportion with IPMN was similar for those with (31%) and without (34%) any pathogenic BRCA mutation and for those with mutations in BRCA1 (33%) and in BRCA2 (30%). Most IPMN (overall, 21/27, 78%) were prevalent lesions, found on initial surveillance, and this did not vary by presence or absence of BRCA1/2 mutations.

For the remaining 2 HRIs with abnormalities found on surveillance, 1, with a BRCA2 mutation, was identified as having chronic pancreatitis (confirmed by endoscopic ultrasound); the other, with no mutations in BRCA1/2, was found to have a pancreatic mass that was confirmed to be metastases of high-grade serous carcinoma of mullerian origin.

The 18 HRIs in this additional group had at least one affected SDR, but no affected FDR; most (14/18, 78%) had only one SDR. Overall, these HRIs were similar demographically to the BRCA1/2-positive HRIs with stronger family history, described above, except that a nonstatistically significant higher proportion were aged ≥60 (45% vs. 25%; P = 0.28). The proportion of this subgroup who were found to have abnormalities was 39%; all were considered to be IPMNs. Median length in surveillance was longer for those with negative findings (median 39 months, range 21–110) than those with positive (median 1 month, range 0–102). Five of the 11 HRIs with negative findings were recently recruited and had only one screen.

Our follow-up of the 27 HRIs with presumed branch-duct IPMN in the main analysis is summarized in Supplementary Table S1. Endoscopic ultrasound was recommended and undertaken for 19 (12 BRCA1/2 positive and 7 BRCA1/2 negative) and confirmed pancreatic lesions in 12 of these cases (7 in BRCA1/2-positive and 5 in BRCA1/2-negative individuals; Fig. 2). Endoscopic ultrasound did not visualize the pancreatic lesions initially identified on imaging for 7 HRIs, likely due to the small size of these pancreatic cysts, which ranged from 0.3 to 1.4 cm on imaging. These small cysts were again noted on follow-up MRCP for all 7 patients, and they continue to be monitored with MRCP for any changes.

Surgery was recommended for 4 HRIs. Patient 1, with a mutation in BRCA2, underwent surgery due to increasing size cystic lesions throughout the pancreas. Surgical pathology showed IPMN with low grade dysplasia and chronic pancreatitis in the remaining pancreas. This patient presented approximately 9 months later with jaundice and metastatic cancer in her liver; liver biopsy was consistent with pancreaticobiliary origin. Cross-sectional imaging of the remaining pancreas showed no pancreatic mass to account for this metastatic disease. The patient is now deceased.

Baseline MRCP for Patient 2, BRCA1/2 negative, identified cystic lesions in the pancreatic head and neck. Surgery was later recommended due to increasing size of the cystic lesions and dilatation of the main pancreatic duct. Pathology revealed multifocal IPMN of borderline malignant potential. The patient was followed closely after surgery, with MRCP showing no evidence of suspicious lesion in the pancreatic remnant, until expiring from uterine cancer about 6 years later.

Patient 3, also BRCA1/2 negative, had surgery after MRCP noted a cystic lesion in the tail of the pancreas. Endoscopic ultrasound with FNA (fine needle aspiration) identified mucinous cells consistent with a mucinous neoplasm. The decision to move forward with surgery was primarily patient driven; despite a lack of worrisome features such as increased cyst size and ductal dilatation, the patient opted for surgery given her family history of PDAC. Pathology showed PanIN-2, focal fibrosis, and chronic inflammation. The patient is in good health 9 years after surgery, with follow-up MRCPs showing no new or suspicious lesions in the residual portion of the pancreas.

MRCP identified few cystic lesions in the pancreatic body and uncinate process of patient 4, a BRCA1 mutation carrier. Endoscopic ultrasound confirmed these lesions and the patient was referred to surgery to discuss these findings given her strong family history and BRCA1 mutation; given the options to undergo surgery as a prophylactic measure overall or continue watchful surveillance of her cysts, the patient chose the latter. She continues to be under surveillance 10 years later with no significant change in the identified lesions.

One HRI, patient 6, was scheduled for a repeat endoscopic ultrasound for follow-up on a screen-identified oval mass at the time of writing; subsequent scans for all remaining HRIs with pancreatic abnormalities in this analysis did not indicate significant changes in the identified lesions.

In the secondary analysis of HRIs known to have BRCA1/2 mutations, we identified 7 cases with screen-identified lesions (Fig. 2). Endoscopic ultrasound confirmed the pancreatic lesions in 3 of these cases; all were recommended to continue surveillance to monitor their presumed IPMN (see Supplementary Table S2).

We investigated characteristics of HRIs that might have influenced the presence of IPMNs. Table 3 shows characteristics of those with (n = 27) and those without IPMNs (n = 56); none of these differences reached statistical significance in χ² tests. Logistic regression models, with presence of IPMN as the outcome, found no significant associations either. ORs were close to 1 for presence of BRCA1/2 mutations and borderline significant for increasing age (Table 4).

Our Registry population and close coordination with the Clinical Genetics Service provide an opportunity to evaluate our hypothesis that surveillance in individuals with pathogenic mutations in BRCA1/2 might be beneficial to them by identifying pancreatic lesions that could be characterized as premalignant. Our comparison group is unique: those known not to have mutations in either of these genes. We found no difference in pancreatic abnormalities between those with and without mutations. Results were similar in those with mutations in BRCA1 and BRCA2. In an additional subset of 18 HRIs who were BRCA1/2 positive but had weaker family history, the proportion of abnormalities found on MRCP was also similar. On the basis of our study, HRIs with mutations in BRCA1/2 do not appear to be more likely to have pancreatic abnormalities on surveillance. Because this report is based on a small number of HRIs, research should be continued on outcomes of surveillance programs among the BRCA-positive population as well as the general high risk population. We do not recommend changes in clinical practice at this time.

A recent study (5), which focused on neoplastic progression in HRIs undergoing surveillance, included 41 HRIs with known mutations in BRCA1/2 or PALB2; 2 of the 41 HRIs showed neoplastic progression, a similar proportion to that among all HRIs (n = 354) in that study. Bartsch and colleagues (30) studied a group of individuals at risk for pancreatic cancer based on their family history and BRCA1/2/PALB2 mutation status in three European centers. Their study included 17 individuals with known mutations. In contrast to our study, they observed a higher proportion with potentially significant lesions compared with all others, 18% versus 6%.

The overall proportion of HRIs in our study with positive findings on MRCP was 35%; this was higher than the 17% reported in our previous study of a more broadly defined group of at-risk relatives (31) but consistent in other studies, where prevalence of lesions has ranged from 32% to 53% (30, 32, 33).

It is not clear why in this high risk cohort of patients with pathogenic BRCA1/2 mutations, our imaging findings are no different from those individuals with BRCA-negative results. It may be that BRCA1/2 mutations are more strongly associated with Pan-INs, which cannot be detected with the imaging modalities used in our study. Although most of the lesions found in this study were prevalent (found at initial surveillance), it is possible that more incident lesions or changes in existing lesions would be found with longer follow-up (34, 35).

This study has several limitations. Because the study was conducted at a tertiary referral hospital, the HRIs included here do not come from a well-defined population and results may not be generalizable to the overall population. As in other studies (5, 18, 36, 37), our population is not diverse in terms of race; in contrast to other studies, our population includes a relatively large proportion of Ashkenazi Jews. These characteristics reflect both our local population and the self-selection of HRIs for this program. We do not have information on reasons for undergoing testing for mutations in BRCA1/2 and our data do not allow us to disentangle the roles of family history, personal history, and Ashkenazi Jewish background in their decisions. HRIs included in this analysis were not tested for all possible PDAC susceptibility genes. Some of the HRIs had testing only for the three Ashkenazi Jewish founder mutations. However, among the 13 HRIs of Ashkenazi background who tested negative for BRCA1/2, 10 provided documentation that they had more thorough testing with sequencing and/or large rearrangement analyses for mutations, making it unlikely that misclassification influenced the results. Another limitation is that our eligibility criteria changed over the study period. Although inclusion guidelines for the Registry have been consistent with past consensus and present recommendations for who should be screened (38), criteria changes have resulted in differences between those with and without BRCA1/2 mutations in family history and length of follow-up. However, because family history was not related to presence of IPMNs, and most lesions were found on the initial scan and did not change in subsequent scans, these factors are unlikely to have influenced results.

This study includes only a subset of HRIs included in our Registry, and a larger analysis of findings in all surveillance participants is underway. The contribution of this analysis is the direct comparison of those with mutations in BRCA1/2 to those known not to have mutations, with no differences noted on the basis of surveillance with dedicated MRCP/CT and endoscopic ultrasound. What is needed is a biomarker for PDAC that can reproducibly stratify PDAC risk and aide in identifying individuals who are in the early stage of developing a pancreatic neoplasm from those who are not. Work in the identification of such biomarkers is currently ongoing, and it is hoped that this will make an impact on the early identification of individuals at high risk for this disease.

No potential conflicts of interest were disclosed.

Conception and design: E. Salo-Mullen, P.J. Allen, K. Offit, R.C. Kurtz

Development of methodology: A. Saldia, P.J. Allen, R.C. Kurtz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Saldia, P. Nunes, M.L. Samson, E. Salo-Mullen, V. Marcell, Z.K. Stadler, P.J. Allen, K. Offit

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Saldia, X. Liang, E. Salo-Mullen, V. Marcell, Z.K. Stadler, K. Offit, R.C. Kurtz

Writing, review, and/or revision of the manuscript: A. Saldia, P. Nunes, E. Salo-Mullen, V. Marcell, Z.K. Stadler, P.J. Allen, R.C. Kurtz

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Saldia, M.L. Samson, K. Offit, R.C. Kurtz

Study supervision: A. Saldia, R.C. Kurtz

This research was supported in part by the NIH/NCI Cancer Center Support Grant P30 CA008748. The following authors were directly supported by the grant: A. Saldia, S.H. Olson, P. Nunes, X. Liang, M.L. Samson, E. Salo-Mullen, V. Marcell, Z.K. Stadler, P.J. Allen, K. Offit, and R.C. Kurtz.

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.