Expansion of Cancer Risk Profile for BRCA1 and BRCA2 Pathogenic Variants

Key Points

Question  Which cancer types and their clinical characteristics are associated with pathogenic variants in BRCA1 and BRCA2 in addition to breast, ovarian, prostate, and pancreatic cancers?

Findings  In this case-control study of 63 828 patients with 14 common cancer types and 37 086 controls, pathogenic variants in BRCA1 were associated with biliary tract cancer, in BRCA2 with esophageal cancer, and in BRCA1/2 with gastric cancer.

Meaning  The study results suggest that the range of cancer types associated with pathogenic variants in BRCA1 and BRCA2 is broader than that determined from previous analyses, potentially indicating the broader clinical relevance of BRCA1/2 genetic testing.

Importance  The clinical importance of genetic testing of BRCA1 and BRCA2 in breast, ovarian, prostate, and pancreatic cancers is widely recognized. However, there is insufficient evidence to include other cancer types that are potentially associated with BRCA1 and BRCA2 in clinical management guidelines.

Objective  To evaluate the association of BRCA1 and BRCA2 pathogenic variants with additional cancer types and their clinical characteristics in 100 914 individuals across 14 cancer types.

Design, Setting, and Participants  This case-control analysis to identify cancer types and clinical characteristics associated with pathogenic variants in BRCA1 and BRCA2 included DNA samples and clinical information from 63 828 patients with 14 common cancer types and 37 086 controls that were sourced from a multi-institutional hospital-based registry, BioBank Japan, between April 2003 and March 2018. The data were analyzed between August 2019 and October 2021.

Main Outcomes and Measures  Germline pathogenic variants in coding regions and 2 bp flanking intronic sequences in BRCA1 and BRCA2 were identified by a multiplex polymerase chain reaction–based target sequence method. Associations of (likely) pathogenic variants with each cancer type were assessed by comparing pathogenic variant carrier frequency between patients in each cancer type and controls.

Results  A total of 65 108 patients (mean [SD] age at diagnosis, 64.1 [11.6] years; 27 531 [42.3%] female) and 38 153 controls (mean [SD] age at registration, 61.8 [14.6] years; 17 911 [46.9%] female) were included in this study. A total of 315 unique pathogenic variants were identified. Pathogenic variants were associated with P < 1 × 10⁻⁴ with an odds ratio (OR) of greater than 4.0 in biliary tract cancer (OR, 17.4; 95% CI, 5.8-51.9) in BRCA1, esophageal cancer (OR, 5.6; 95% CI, 2.9-11.0) in BRCA2, and gastric cancer (OR, 5.2; 95% CI, 2.6-10.5) in BRCA1, and (OR, 4.7; 95% CI, 3.1-7.1) in BRCA2 in addition to the 4 established cancer types. We also observed an association with 2 and 4 other cancer types in BRCA1 and BRCA2, respectively. Biliary tract, female breast, ovarian, and prostate cancers showed enrichment of carrier patients according to the increased number of reported cancer types in relatives.

Conclusions and Relevance  The results of this large-scale registry-based case-control study suggest that pathogenic variants in BRCA1 and BRCA2 were associated with the risk of 7 cancer types. These results indicate broader clinical relevance of BRCA1 and BRCA2 genetic testing.

BRCA1 and BRCA2 were identified in the 1990s as the causative genes underlying hereditary breast and ovarian cancer syndrome.¹ Genetic testing began during the same decade for treatment of patients and their relatives. In addition, polyadenosine diphosphate-ribose polymerase (PARP) inhibitors were developed based on the mechanism of the homologous recombination repair defects associated with pathogenic variants in these genes.² The target BRCA1 and BRCA2 cancer types have expanded to prostate³ and pancreatic cancers⁴ because pathogenic variants were enriched in these patients, and the therapeutic efficacy of PARP inhibitors in these cancers has also been shown.^5,6

Risk for additional cancer types, such as biliary tract cancer,⁷ cervical cancer,^8,9 colorectal cancer,⁹ endometrial cancer,⁹ esophageal cancer,^10,11 and stomach cancer,^7,8,11-13 has been reported by analyzing family members for the presence of pathogenic variants and performing case-control analyses. However, the evidence for an association with these cancer types has not been considered sufficient to be adopted into clinical management guidelines, probably because of the small sample size for each cancer type, weak statistical evidence, or a singular focus on family members with pathogenic variants. In addition, evidence to date for different cancer types has been derived from various studies of different design. Robust evidence for additional cancer types in a single population is necessary to design and implement clinical trials that assess efficacy of PARP inhibitors.

We performed a large-scale sequencing study across 14 common cancer types in 63 828 patients and 37 086 controls whose data were drawn from a Japanese nationwide biobank. We used these data to estimate the risk of each cancer type and clinical characteristics associated with pathogenic variant carrier status. These data provide a broad view of cancer risks associated with pathogenic variants in BRCA1 and BRCA2.

An overall procedure is shown in eFigure 1 in the Supplement. We obtained samples from 65 108 patients with 14 cancer different types (biliary tract, breast, cervical, colorectal, endometrial, esophageal, gastric, liver, lung, lymphoma, ovarian, pancreatic, prostate, and kidney) from BioBank Japan, a multi-institutional, hospital-based registry that collected DNA and clinical information from across Japan between April 2003 and March 2018.^14,15 Family history of cancer refers to reported cancer in first-degree and/or second-degree relatives. Among them, 4128 patients (6.3%) had 2 to 5 cancer types. We also enrolled 38 153 controls 20 years or older with no history or family history of cancers. Compared with our previous publications for breast,¹⁶ colorectal,¹⁷ pancreatic,¹⁸ and prostate¹⁹ cancers, the analyses presented included 14 448 additional controls and 8247 additional cancer cases (2984 breast [36.2%], 3722 colorectal [45.1%], 1535 pancreatic [18.6%], and 6 prostate [0.1%]).

All participants provided written informed consent. The study was approved by the ethical committees of the Institute of Medical Sciences, University of Tokyo, and RIKEN Center for Integrative Medical Sciences.

For germline sequencing, we analyzed all coding regions and 2 bp flanking intronic sequences (16 111 bp) of all transcripts of BRCA1 (CCDS11453-6, 9) and BRCA2 (CCDS9344) that were registered in the Consensus CDS, release 15,²⁰ by a multiplex polymerase chain reaction–based target sequence method.²¹ After sequencing the pooled DNA libraries using 2 × 150-bp paired-end reads on a HiSeq2500 (Illumina), the genetic variants were identified using the GATK (version 3.7-0; Broad Institute).²² We deposited custom scripts at https://github.com/Laboratory-for-Genotyping-Development/TargetSequence.git.

We determined the association of genetic variants with the amino acid sequence using the SnpEff, version 4.3t.²³ Protein position was reported according to CCDS11456 (the longest one) for BRCA1 and CCDS9344 for BRCA2. We assigned clinical significance for all variants using BRCA1 and BRCA2 variant classification criteria that were developed by members of the Evidence-based Network for the Interpretation of Germline Mutant Alleles Consortium.^24,25 Pathogenic and likely pathogenic variants were collectively referred to as pathogenic variants.

Because sex differences in the contribution of pathogenic variants to breast cancer are well known,²⁶ we analyzed females and males separately for breast cancer. We conducted 2 separate association analyses. A logistic regression analysis under a dominant model with age at diagnosis for cases and age at registration for controls as a covariate was used. We eliminated samples without age at diagnosis or registration from this calculation. The first analysis examined all patients and controls for each cancer type. Including the control participants without family history would improve the power to detect association but lead to biased risk estimates. Therefore, we reported the results of association testing without risk estimates. For the second analysis, we calculated cancer risk using selected patients with cancer without a family history for comparison with controls without a family history to minimize overestimation of odds ratios (ORs). The cumulative risk and its 95% CIs of each cancer to age 85 years were estimated for carriers and noncarriers of pathogenic variants in BRCA1 and BRCA2 using the method described by previous studies^27-30 (eAppendix in the Supplement).

We also investigated how family history is useful in detecting patients with pathogenic variants. Family history of a given cancer type was denoted if at least 1 relative reported that cancer type. In each of 7 associated cancer types identified in this study, we calculated the proportion of patients with pathogenic variants according to reported family history of the 7 associated cancer types.

Statistics methods are described where results are shown. All statistical tests were 2-sided, and statistical significance was set at P < .05. The Bonferroni correction was applied to adjust for multiple comparisons. We set the threshold of significance to 1 × 10⁻⁴ for the burden test based on the justified recommendation from a review article that focused on clinical validity of gene panel sequence tests.³¹ Analyses were performed using R, version 3.5.2 (R Foundation), or Stata, version 16.0 (StataCorp).

Table 1 shows the characteristics of 65 108 patients with cancer (representing 69 550 case diagnoses) and 38 153 controls. The mean (SD) age at diagnosis varied between cancer types from 49.7 (13.2) years for cervical cancer to 70.2 (7.3) years for prostate cancer. The proportion of cases reporting a family history of the same cancer type was lowest for endometrial cancer (46 [2.4%]) and highest for gastric cancer (3010 [28.1%]).

After quality control, 63 828 patients (68 219 case diagnoses) and 37 086 controls were included, with 99.85% of the target region covered by at least 20 sequence reads. We applied 1810 genetic variants to the interpretation of clinical significance. After standardized review, 315 variants (17.4%) were assigned as pathogenic (eTable 1 in the Supplement). eFigure 2 in the Supplement shows the distribution of pathogenic variants found in patients. Three BRCA1 and 8 BRCA2 pathogenic variants were observed in 10 or more patients.

The proportion of pathogenic variants in cases differed significantly across the 7 regions serviced by hospitals in Japan by χ² test (eFigure 3 in the Supplement). There was a 5.8-fold difference in BRCA1 between 0.15% in Tokai-Hokuriku and 0.85% in Tohoku. BRCA2 showed a 5.0-fold difference between 0.28% in Okinawa and 1.37% in Kinki. These differences could be largely explained by the different proportion of founder pathogenic variants (eFigure 3 in the Supplement). After excluding these founder pathogenic variants, there was no difference in frequency of BRCA1 or BRCA2 pathogenic variants across the 7 regions.

Figure 1 shows the patient carrier frequency for each of the 14 cancer types. Male patients with breast cancer had a very high carrier frequency of pathogenic variants²⁶ in BRCA2 (18.9%), but not BRCA1 (1.89%). Patients with ovarian cancer showed the next highest proportion (BRCA1: 4.86%; BRCA2: 3.42%). Frequency exceeding 1% was seen for several other cancer types (2 cancer types for BRCA1, 4 cancer types for BRCA2). Carrier frequency of pathogenic variants in BRCA1 was 0.44% in 1 cancer type, 0.85% in 2 cancer types, and 0.69% in 3 cancer types. They were 0.97%, 1.40%, and 1.74% in BRCA2 (eFigure 4 in the Supplement). Carrier frequency for females and males in each cancer type was significantly correlated for BRCA1 and BRCA2 (eFigure 5 in the Supplement).

We performed association analyses with all patients and controls for each cancer type (eTable 2 in the Supplement). Among the 30 analyses, 17 yielded statistically significant results. To provide a more conservative estimate that accounted for control selection criteria, we calculated the disease risk for each cancer type (Table 2); association analyses were not possible for some cancer types because of limited numbers of cases in those subsets. Pathogenic variants in BRCA1 were significantly associated with increased risk of 5 cancer types: ovarian, female breast, biliary tract, gastric, and pancreatic cancers. Pathogenic variants in BRCA2 were associated with increased risk of 7 cancer types: female breast, gastric, ovarian, male breast, pancreatic, prostate, and esophageal cancers. These associations were more strongly observed in the analysis with all patients (eTable 2 in the Supplement). We conducted 2 sensitivity analyses in breast cancer. We performed a logistic regression analysis with 7 regions of Japan as a categorical covariate to assess the possibility that population stratification might skew risk estimation. Results for BRCA1 and BRCA2 (eTable 3 in the Supplement) were comparable with estimates from the main analysis shown in Table 2. We also performed an association analysis with patients with breast cancer only because of a potential bias caused by the presence of more than 1 cancer type. The results for BRCA1 and BRCA2 (eTable 3 in the Supplement) were comparable with those from the main analysis. Heterogeneity of ORs between BRCA1 and BRCA2 were shown in ovarian (I², 90.9%) and prostate cancer (I², 77.4%) (Table 2).³² We also observed an association with lymphoma and lung cancer for BRCA1 and endometrial, cervical, kidney, and liver cancers for BRCA2 (Table 2).

The cumulative risk of cancer to age 85 years was estimated for carriers and noncarriers of pathogenic variants in BRCA1 and/or BRCA2 for the 7 significantly associated cancer types (Figures 2 and 3). In BRCA1, breast cancer showed the highest cumulative risk at 72.5% (95% CI, 20.4%-90.5%) followed by ovarian cancer at 65.6% (95% CI, 12.8%-86.4%), gastric cancer at 21.3% (95% CI, 6.9%-33.4%), pancreatic cancer at 16.0% (95% CI, −3.9% to 32.1%), and biliary tract cancer at 11.2% (95% CI, −1.1% to 22.1%). In BRCA2, the highest cumulative risk was also breast cancer at 58.3% (95% CI, 38.3%-71.9%), followed by prostate cancer at 24.5% (95% CI, 6.9%-38.8%), gastric cancer at 19.3% (95% CI, 11.9%-26.0%), ovarian cancer at 14.8% (95% CI, 4.6%-23.9%), pancreatic cancer at 13.7% (95% CI, 3.7%-22.8%), and esophageal cancer at 5.2% (95% CI, 1.7%-8.5%).

We investigated the association between carrier status and age at diagnosis for the 7 associated cancers. eTable 4 in the Supplement suggests that pathogenic variants in BRCA1 were associated with earlier age at diagnosis of female breast cancer (−5.7 years). Meanwhile, those with BRCA2 were associated with earlier age at diagnosis of female breast cancer (−5.7 years) and prostate cancer (−2.2 years). Carriers with BRCA2 pathogenic variants showed a later diagnosis of ovarian cancer (4.1 years). The proportion of pathogenic variants according to diagnosis age (by 10-year age group) is shown in eFigure 6 in the Supplement.

eFigure 7 in the Supplement describes associations between the carrier status in BRCA1 and BRCA2 for cases with a diagnosis of the 7 associated cancers and reported family history of each of these cancer types. For BRCA1, family history of ovarian cancer was strongly enriched in female breast, ovarian, and pancreatic cancers. However, for BRCA2, family history of breast cancer was broadly enriched in 5 cancer types.

We also investigated whether carrier patients had specific histological subtypes. We observed a different distribution of histological subtypes only for breast and ovarian cancer, which is largely consistent with previous reports³³ (eTable 5 in the Supplement).

We investigated the extent to which carriers with pathogenic variants were enriched for reported family history (none, 1, or 2 or more of these 7 cancer types) with the Cochran-Armitage test. Biliary tract, female breast, ovarian, and prostate cancers showed increasing enrichment of carrier patients according to the increased number of reported cancer types in relatives (eFigure 8 in the Supplement).

This large-scale registry-based case-control study analyzed BRCA1 and BRCA2 in 63 828 patients with 14 cancer types and 37 086 controls. The proportion of pathogenic variants varied across different regions in Japan, mainly because of differences in the proportion of founder pathogenic variants. We observed that biliary tract, esophageal, and gastric cancer were significantly associated with BRCA1 and/or BRCA2 pathogenic variant status in addition to the 4 established cancer types. Six other cancer types showed an association. Patients with pathogenic variants were more likely to report a family history of the 7 associated cancer types.

The results of this large-scale registry-based study suggest that pathogenic variants in BRCA1 and/or BRCA2 are associated with increased risk of biliary tract, gastric, and esophageal cancers. Further studies are needed to reveal the mechanisms linking pathogenic variants and these cancer types for the potential efficacy of PARP inhibitors because homologous recombination repair defects were observed in some patients with biliary and esophageal cancers.³⁴ These cancers are known to have a higher incidence rate in East Asian countries.³⁵ The estimated lifetime cumulative risks of breast and ovarian cancer in BRCA1 and BRCA2 are broadly consistent with the previous estimation³⁶; the present study could not detect that the cumulative risks of ovarian cancer were low up to age 40 years for BRCA1 carriers and age 50 years for BRCA2 carriers because we could not calculate age-specific estimated ORs. Cumulative risk of prostate cancer for BRCA2 carriers was lower than that estimated in the UK and Ireland,³⁷ probably because prostate cancer incidence rate is higher in European countries.³⁵ Conversely, cumulative risk of gastric cancer was estimated at around 20% for both genes, which is likely higher than for European populations because of the higher incidence rate in East Asia countries.³⁵ Taken together, the cumulative risk for each cancer type would be associated with the different incidence rate in each country. We also observed associations with risk of 6 additional cancer types. The results suggest that the range of cancer types associated with pathogenic variants in BRCA1 and BRCA2 is likely broader than that determined from previous analysis of largely European ancestry cohorts.

We observed a large difference in the carrier frequency between the 7 regions. This could be largely explained by the distribution of founder pathogenic variants. Founder pathogenic variants are known to be associated with carrier frequency in a population, which could change the best strategy for genetic tests³⁸; regional differences should also be considered when designing a suitable strategy among genetic tests for selected patients, all patients, or all unaffected individuals. It also suggests that population-matched and region-matched controls would be indispensable for precise risk estimation of cancer predisposition genes rather than ExAC and gnomAD.³⁹ Taken together, more detailed information that accounts for populations and regions would improve precision medicine with genetic testing of BRCA1 and BRCA2.

These risk association findings, together with our analysis of an association with family history of cancer and clinical phenotypes, are relevant for developing and adapting guidelines about genetic testing, treatment options, and treatability with PARP inhibitors for each cancer type. Depending on the cancer type, different guidelines exist to prioritize patients for receiving germline gene testing. The National Comprehensive Cancer Network guidelines³⁸ recommend that all patients with pancreatic cancer should be tested. Meanwhile, patients with breast or ovarian cancer are selected based on several criteria. This study suggests that the family history of the 7 associated cancer types efficiently identified patients with pathogenic variants. Therefore, this information would be useful to expand indications for genetic testing of individuals with family history of these cancer types.

This study has several limitations. We selected controls without family history of cancer because we intended to improve the statistical power for association analysis and limit the effect of family history, including shared genetic and environmental effects.⁴⁰ This would affect the generalizability of the study results. However, the estimated cumulative risks were comparable with those based on prospective cohorts, suggesting that the study design did not greatly affect the results. We analyzed only single-nucleotide variants and small indels, but structural variants are known to be associated with hereditary breast cancer.⁴¹ However, the proportion of pathogenic variants due to structural changes is reported to be very low for the BRCA1 and BRCA2 genes in Japanese populations.⁴² Lastly, we tested for a linear association in several statistical analyses; however, we had the potential to miss other patterns of association.⁴³ In particular, biliary tract cancer in BRCA2 (eFigure 6 in the Supplement) showed an unusual pattern that should be investigated in further studies.

This large-scale registry-based case-control study of 63 828 patients across 14 cancer types and 37 086 population-matched and region-matched controls provided a broad view of carrier frequency, disease risk, family history, and clinical characteristics of pathogenic variant carriers. This information can potentially improve genetic testing of BRCA1 and BRCA2 for various cancer types for Asian countries and encourage similar research in other countries.

Accepted for Publication: February 3, 2022.

Published Online: April 14, 2022. doi:10.1001/jamaoncol.2022.0476

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Momozawa Y et al. JAMA Oncology.

Corresponding Author: Yukihide Momozawa, DVM, PhD, RIKEN Center for Integrative Medical Sciences (IMS), 1-7-22 Suehiro-cho, Tsurumi-ku, Yokohama City, Kanagawa 230-0045, Japan (momozawa@riken.jp).

Author Contributions: Dr Momozawa had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Momozawa, Shiraishi, Kamatani, Kohno, Yoshida, Murakami, Kubo.

Acquisition, analysis, or interpretation of data: Momozawa, Sasai, Usui, Iwasaki, Taniyama, Parsons, Mizukami, Sekine, Hirata, Endo, Inai, Takata, Ito, Matsuda, Nakamura, Sugano, Yoshida, Nakagawa, Matsuo, Spurdle, Kubo.

Drafting of the manuscript: Momozawa, Sasai, Usui, Shiraishi, Iwasaki, Endo, Takata, Matsuo, Spurdle.

Critical revision of the manuscript for important intellectual content: Momozawa, Usui, Taniyama, Parsons, Mizukami, Sekine, Hirata, Kamatani, Inai, Ito, Kohno, Matsuda, Nakamura, Sugano, Yoshida, Nakagawa, Matsuo, Murakami, Spurdle, Kubo.

Statistical analysis: Momozawa, Sasai, Usui, Iwasaki, Matsuo.

Obtained funding: Momozawa, Shiraishi, Kubo.

Administrative, technical, or material support: Momozawa, Shiraishi, Mizukami, Hirata, Kamatani, Endo, Inai, Takata, Ito, Matsuda, Matsuo, Murakami, Kubo.

Supervision: Momozawa, Usui, Kamatani, Endo, Takata, Sugano, Yoshida, Nakagawa, Kubo.

Other - variant classification: Spurdle.

Conflict of Interest Disclosures: Dr Momozawa reported grants from the Japan Agency for Medical Research and Development (AMED) during the conduct of the study as well as grants from Ono and personal fees from LabCorp Japan, GK, AstraZeneca, and Sanofi outside the submitted work. Dr Kamatani reported grants from the Japan Society for the Promotion of Science and AMED outside the submitted work as well as honoraria from Astellas, Chugai, Sandoz, Taisho, and Illumina Japan. Dr Kohno reported grants from AMED during the conduct of the study as well as grants from Sysmex and Chugai and personal fees from Eli Lilly outside the submitted work. Dr Murakami reported grants from the Social Cooperation Research Program outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported by AMED through grants JP19kk0305010, JP20ck0106402, JP19cm0106605, and 20ck0106553. Dr Spurdle and Mr Parsons are supported by Australian National Health and Medical Research funding (grant ID177524).

Role of the Funder/Sponsor: The funding organizations 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; and decision to submit the manuscript for publication.

Additional Contributions: We thank the individuals who participated in this study. We acknowledge the staff of the Laboratory for Genotyping Development in RIKEN, the RIKEN-IMS Genome Platform, and the BioBank Japan project. None of these individuals were compensated for their contributions.