The long road to unbiased estimates of pancreatic cancer incidence in the hereditary breast and ovarian cancer syndrome

It may surprise the readers of Cancer to find that 30 years after the cloning of the BRCA1 and BRCA2 genes,^{1, 2} articles are still being published on the cancer risks associated with the genes responsible for hereditary breast and ovarian cancer syndrome (HBOC). The article by Katona and colleagues on the incidence of pancreatic cancer in women with BRCA mutations published in this issue of Cancer ³ 3 decades after it was first recognized that BRCA2 contributes to pancreatic cancer susceptibility² attests to the inherent challenges of determining unbiased estimates of lifetime cancer risk in HBOC.

The association between pancreatic ductal adenocarcinoma (PDAC) and HBOC was firmly established for individuals with germline mutations in BRCA2 by the Breast Cancer Linkage Consortium more than 25 years ago.⁴ By using complex segregation analysis of pedigrees to estimate relative risk (RR) and cumulative lifetime risk of pancreatic cancer among 3728 individuals from 173 families transmitting a BRCA2 mutation, this international consortium of investigators from North America and Europe reported an RR of 3.51 (95% confidence interval [CI], 1.87–6.58), with cumulative lifetime risks of 2.3% (95% CI, 1.1%–3.5%) for female BRCA2 carriers and 3.2% (95% CI, 1.6%–4.9%) for male carriers. The same group subsequently found a lower level of risk among 699 families segregating a BRCA1 mutation (RR, 2.26; 95% CI, 1.26–4.06).⁵ Other investigators found a somewhat higher pancreatic cancer risk in BRCA2 families than the Breast Cancer Linkage Consortium, with RRs ranging from 4.1 (95% CI, 1.9–7.8)⁶ to 5.9 (95% CI, 3.2–10.0),⁷ although the CIs overlap with the earlier study.⁴ Most recently, Li et al. reported the largest study of cancer risk in HBOC to date on behalf of the Consortium of Investigators of Modifiers of BRCA1/2 (CIMBA) with an analysis that included 3184 families transmitting a BRCA1 mutation and 2157 families transmitting a BRCA2 mutation.⁸ The CIMBA investigators reported pancreatic cancer RRs of 4.27 (95% CI, 2.01–9.05) and 4.34 (95% CI, 2.19–8.62) for female BRCA1 and BRCA2 carriers, respectively. This was modestly higher than for male mutation carriers, and translated to a cumulative lifetime risk of 2.3% for female carriers of both BRCA1 and BRCA2 mutations.

The studies reported to date on PDAC risk among BRCA mutation carriers are retrospective analyses of pedigrees from families ascertained because of the presence of multiple family members with HBOC-associated cancers or population-based studies of patients with cancer. In those analyses, a retrospective cohort was usually reconstructed, with follow-up commencing on the date of birth of probands and their relatives. The analyses required statistical adjustment to account for ascertainment bias that arises from nonrandom sampling of families and individuals with respect to cancer status. It is important to note that the analytic technique chosen to correct for selection bias in retrospective cohort analyses can affect risk estimates quite substantially.⁹ Furthermore, retrospective studies are most susceptible to information bias as a consequence of incomplete or inaccurate reporting of cancer diagnoses in family members. For these reasons, prospective cohort studies produce the most reliable risk estimates,^{9, 10} although they are extremely challenging to conduct given the large number of mutation carriers who need to be enrolled and the long follow-up period required. Against this backdrop, the study by Katona and colleagues,³ which aimed to update estimates of the cumulative incidence of pancreatic cancer for women carrying BRCA mutations via a prospective cohort design, is a welcome addition to the body of literature on cancer incidence in HBOC.

This report from the Hereditary Breast Cancer Clinical Study Group included 8295 women with germline pathogenic or likely pathogenic variants in the BRCA1 or BRCA2 genes who were enrolled in a longitudinal cohort study at 66 high-risk clinics in 12 countries.³ Genetic testing criteria varied across enrollment sites, with most women having a personal or family history of HBOC-associated cancer. In general, participants were not selected for genetic testing or study participation specifically because of a personal or family history of pancreatic cancer. Seventy percent of participants carried a BRCA1 mutation, and 29% carried a BRCA2 mutation; 1.2% reported a first-degree relative (FDR) affected with pancreatic cancer, and 0.6% reported an affected second-degree relative (SDR). The cumulative risk of pancreatic cancer from age 40 to 80 years was calculated via the Kaplan–Meier method. Cox regression was performed to identify the risk factors for pancreatic cancer in the cohort, and a survival analysis was conducted for those diagnosed with pancreatic cancer.

With an average follow-up time of 8.6 years, the cumulative incidence of pancreatic cancer from age 40 to 80 years was 2.2% for BRCA1 carriers (95% CI, 1.1%–4.3%) and 2.7% for BRCA2 carriers (95% CI, 1.3%–5.4%). These data confirm earlier reports of a slightly higher incidence of PDAC in BRCA2 carriers compared with BRCA1 carriers, with a hazard ratio (HR) of 2.08 (95% CI, 1.05–4.12). Notably, only six of 34 individuals who developed pancreatic cancer reported an FDR or SDR with this disease, although risk was significantly higher in participants who did report an affected family member compared with those who did not (adjusted HR, 5.06; 95% CI, 1.5–17.1). Additional predictors of a pancreatic cancer diagnosis included alcohol consumption of more than nine drinks per week (adjusted HR, 7.37; 95% CI, 2.38–22.8) and a history of diabetes at the time of study enrollment (adjusted HR, 4.63; 95% CI, 1.3–16.1). Four of 34 cases were diagnosed before age 50 years, with an annual incidence rate of 0.018% for age 40–49 years versus 0.053%–0.17% from age 50 to 80+ years. Median overall survival after a diagnosis of pancreatic cancer in this cohort was 0.9 years, with a 5-year survival rate of only 8.8%.

The main strength of this study lies in its prospective design, which eliminates the potential for ascertainment bias and the underreporting of cancer history in relatives—issues commonly encountered in previous retrospective studies. Because pancreatic cancer is relatively rare, the prospective cohort included only 17 incident cases among BRCA1 carriers and 17 cases among BRCA2 carriers. By comparison, the most recent retrospective consortium study reported 95 women with pancreatic cancer among BRCA1 carriers and 101 cases among BRCA2 carriers, although it required statistical adjustments to account for ascertainment bias.⁸ Notably, the prospective and retrospective studies gave consistent findings regarding lifetime risk estimates and age-specific risk patterns of pancreatic cancer in BRCA carriers. Before age 50 years, the risk of pancreatic cancer was low in both BRCA1 and BRCA2 carriers. However, after age 50 years, the risk began to rise more rapidly in BRCA2 carriers compared with BRCA1 carriers. By ages 60 and 70 years, the risk for BRCA2 carriers was approximately double that of BRCA1 carriers. After age 70 years, the risk for BRCA1 carriers increased more rapidly than for BRCA2 carriers, which resulted in similar cumulative risks for both groups by age 80 years. The consistent finding of a later rise in the annual incidence of pancreatic cancer in BRCA1 carriers in both the present study and the study from Li and colleagues⁸ suggests that it may be appropriate to consider beginning pancreatic cancer screening at a later age for BRCA1 carriers compared with BRCA2 carriers.

A methodological issue to note is the attempt to obtain medical records for all participants reporting a cancer diagnosis during the follow-up period. Importantly, the investigators were able to confirm the diagnosis of pancreatic cancer in 94% of incident cases via a review of medical records. A limitation of the study is the large number of women enrolled in the cohort who were excluded from the present analysis for a variety of reasons, including the lack of follow-up data, age younger than 40 years, and a cancer diagnosis that could be confused with pancreatic cancer. The final analytic cohort represented just 46% of the women initially enrolled. However, the most important source of potential bias is the study population itself. The wide range of cumulative lifetime risk estimates reported for HBOC is generally attributed to two factors: the method used to correct for ascertainment bias as already discussed, and differences in the study populations.⁹ The population studied in this analysis was largely similar to the retrospective CIMBA cohort⁸ (primarily individuals ascertained from genetics clinics because of a high-risk personal or family history or cancer cases ascertained from the general population). It is reassuring (and not surprising) that the results of this analysis are very consistent with the earlier study. Estimates of cancer incidence from studies of BRCA mutation carriers who were selected because of a high-risk personal or family history are generally higher than risk estimates from studies of population-based testing for HBOC, where candidates for testing are unselected for personal or family history.^{11, 12} As we enter the era of population genomic screening, it is important to recognize that the incidence of pancreatic cancer in women with HBOC reported by Katona and colleagues is a reliable estimate for individuals selected for genetic testing on the basis of a high-risk personal or family history; however, risk estimates for individuals with BRCA mutations who are identified via population genomic screening have not yet been determined, and conceivably might vary from the estimates reported.

The study by Katona and colleagues has important implications for cancer control measures in HBOC. A series of screening trials with endoscopic ultrasound or magnetic resonance cholangiopancreatography in a prospectively followed cohort of individuals with hereditary PDAC syndromes found that nearly 60% of PDAC cases diagnosed via screening were stage I compared with only 14% of cases diagnosed outside of screening.¹³ Most importantly, median overall survival was 9.8 years for screening-detected cases versus 1.5 years for cases diagnosed outside of screening (unadjusted HR, 0.13; 95% CI, 0.03–0.50). The decisive question now is: which individuals have a sufficient level of risk to justify the expense and risks associated with screening? Previous versions of the National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Genetic/Familial High-Risk Assessment: Breast, Ovarian, and Pancreatic¹⁴ and other clinical practice guidelines^{15, 16} recommended screening for PDAC for individuals with HBOC only if there was a family history of pancreatic cancer. However, the most recent version of the NCCN Guidelines now recommends pancreatic cancer screening for BRCA2 carriers beginning at age 50 years (or 10 years before the youngest PDAC diagnosis in the family, whichever comes first) regardless of family history, although a positive family history is still required for BRCA1 carriers.¹⁷ This change more closely aligns with updated recommendations from the American Society for Gastrointestinal Endoscopy,¹⁸ which emphasize screening for pancreatic cancer in all patients with a BRCA1 or BRCA2 mutation regardless of family history. The finding by Katona and colleagues that fewer than 20% of individuals with HBOC who developed an incident pancreatic cancer reported a family member with that disease underscores the inadequacy of family history as a criterion for selecting candidates for screening, and supports removal of that requirement from clinical practice guidelines. Moreover, the absolute difference in cumulative risk of pancreatic cancer from age 40 to 80 years between BRCA1 and BRCA2 mutation carriers is minimal in this prospective study (2.2% vs. 2.7%) and in the retrospective study from the CIMBA investigators⁸ (2.3% for female carriers of both BRCA1 and BRCA2), notwithstanding the more significant RR differences and the later rise in annual incidence for BRCA1 carriers. With the publication of this study, it is difficult to find a rationale for requiring a family history of PDAC as a criterion for pancreatic cancer screening in either BRCA1 or BRCA2 carriers.

On the 30th anniversary of the isolation of the BRCA2 gene,² Katona and colleagues from the Hereditary Breast Cancer Clinical Study Group are to be congratulated for their painstaking work over several decades with this large prospective cohort of individuals with HBOC. Their work is generating important data on cancer risks in HBOC that largely confirm findings from earlier retrospective pedigree studies,^{4, 5, 8} and may influence clinical practice guidelines for cancer screening in BRCA mutation carriers. The road to unbiased pancreatic cancer risk estimates in HBOC has been long indeed, and more work remains. There is an urgent need for better predictors of pancreatic cancer risk in HBOC that would provide improved risk stratification and allow a tailored and more cost-effective approach to reducing mortality from pancreatic cancer in families with this important hereditary cancer syndrome. Prediction tools that incorporate the risk factors identified in this study (family history, heavy use of alcohol, and diabetes) would be a step in the right direction; however, it is unlikely that clinical factors alone will provide sufficient discriminatory accuracy for a relatively rare condition. Development of prediction models with clinical utility in HBOC will likely require a combination of clinical and genomic markers of risk such as polygenic risk scores¹⁹ or multicancer early detection tests.²⁰ We hope that this study will accelerate progress toward this goal.

Kent F. Hoskins, reports the following disclosures: nonfinancial research support from Agendia, and research support to the institution from AbbVie, Genentech, Merck, Novartis, Pfizer, and Roche. The other authors declare no conflicts of interest.