Introduction Poly(ADP-ribose) polymerase inhibitors significantly improve progression-free survival in platinum-sensitive high-grade serous and endometrioid ovarian carcinoma, with greatest benefits observed in women with a pathogenic BRCA1/2 variant. Consequently, the demand for germline BRCA1/2 testing in ovarian cancer has increased substantially, leading to the screening of unselected populations of patients. We aimed to determine the prevalence of pathogenic germline BRCA1/2 variants in women diagnosed with epithelial ovarian cancer, categorised according to the established risk factors for hereditary breast and ovarian cancer syndrome and the Manchester BRCA Score, to inform risk stratification.
Methods A cohort of sequential epithelial ovarian cancer cases recruited between June 2013 and September 2018 underwent germline BRCA1/2 testing by next-generation sequencing and multiplex ligation-dependent probe amplification.
Results Five hundred and fifty-seven patients were screened. Of these, 18% had inherited a pathogenic BRCA1/2 variant. The prevalence of pathogenic BRCA1/2 variants was >10% in women diagnosed with ovarian cancer earlier than 60 years of age (21%) and those diagnosed later than 60 years of age with a family history of breast and/or ovarian cancer (17%) or a medical history of breast cancer (34%). The prevalence of pathogenic BRCA1/2 variants was also >10% in women with a Manchester BRCA Score of ≥15 points (14%).
Discussion Our study suggests that age at diagnosis, family history of breast and/or ovarian cancer, medical history of breast cancer or a Manchester BRCA Score of ≥15 points are associated with a >10% prevalence of germline pathogenic BRCA1/2 variants in epithelial ovarian cancer.
- ovarian cancer
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Ovarian cancer is the eighth most common cancer occurring in women and the second most common cause of gynaecological-related cancer death worldwide.1 Standard of care treatments include cytoreductive surgery and platinum-based and taxane-based chemotherapy.2 3 Molecularly targeted agents offer the promise of anticancer treatments that specifically target biological vulnerabilities within tumour cells, thereby offering alternative therapies to traditional cytotoxic agents. To date, pathogenic BRCA1/2 variants are the only predictive biomarkers validated in ovarian cancer.4 Several phase II/III trials have shown that poly(ADP-ribose) polymerase (PARP) inhibitors significantly improve progression-free survival (PFS) in platinum-sensitive high-grade serous and endometrioid ovarian cancer, with the greatest benefit achieved in women with a pathogenic BRCA1/2 variant.5–10 Indeed, a recently reported randomised, double-blinded, placebo-controlled, phase III trial, SOLO1, demonstrated that 24 months of olaparib maintenance therapy following a partial/complete response to cytoreductive surgery and platinum-based chemotherapy in FIGO stage 3/4 BRCA-mutant high-grade serous or endometrioid ovarian carcinoma reduced the risk of disease progression or death at 3 years with a HR 0.28 (95% CI 0.20 to 0.39, p<0.001).9
The prevalence of germline pathogenic BRCA1/2 variants in ovarian cancer is estimated to be between 10% and 15%, with the majority of heterozygotes diagnosed with high-grade serous ovarian carcinoma.11–15 High-grade serous carcinoma is the most common histological subtype, accounting for approximately 70% of all cases of ovarian cancer.16 17 At present, access to PARP inhibitors as maintenance therapy in Europe and North America is restricted by morphological subtype (serous or endometrioid), BRCA1/2 status (germline or somatic) and/or platinum sensitivity (complete or partial response to the latest platinum-based therapy). It is not surprising therefore that clinical demand for BRCA1/2 testing has increased significantly as oncologists and patients seek to access these drugs.18–23 As a result, germline BRCA1/2 testing is increasingly prevalent in unselected populations of women with ovarian cancer, resembling routine tumour testing for somatic mutations in other tumour types, for example, BRAF (melanoma), RAS (colorectal cancer), EGFR (lung cancer) and PDGFRA and KIT (gastrointestinal tumours). Unlike routine tumour testing for somatic variants, testing for germline BRCA1/2 variants could be stratified according to risk factors associated with hereditary breast and ovarian cancer syndrome.24
In this study, we report the prevalence of germline pathogenic BRCA1/2 variants in a large cohort of women diagnosed with epithelial ovarian cancer in the North West of England, correlating the prevalence of germline pathogenic BRCA1/2 variants with risk factors associated with hereditary breast and ovarian cancer syndrome. Our aim is to inform risk stratification for germline BRCA1/2 testing in epithelial ovarian cancer when conducted in an oncology clinic rather than a specialised genetics department.
Women diagnosed with epithelial cancer of the ovary, fallopian tube or peritoneum (FIGO stage 1 to 425) who underwent germline BRCA1 and BRCA2 testing between 1 June 2013 and 1 September 2018 were included. Germline BRCA1/2 testing took place in the oncology clinics at the Christie NHS Foundation Trust, Manchester or the genetics clinics at St Mary’s Hospital, Manchester. Only women treated for ovarian cancer at The Christie Hospital or St Mary’s Hospital were included in the study. Pathogenic (class 5) or likely pathogenic (class 4) BRCA1 and BRCA2 variants were included and will be referred to collectively as ‘pathogenic BRCA1/2 variants’ throughout this manuscript, while variants of unknown clinical significance (class 3) were excluded.26 Cases of non-epithelial ovarian cancer were excluded. Women from a Jewish ancestry were excluded because across the North West of England this group undergo founder mutation testing first, and the Manchester BRCA Scoring System is not designed to assess risk in this population.
A family history was defined as any index case of epithelial ovarian cancer and a first-degree or second-degree relative with breast and/or ovarian cancer. An index case was diagnosed with sporadic ovarian cancer if she had no first-degree or second-degree relative with breast and/or ovarian cancer. All demographic data were extracted from case notes and/or electronic patient records.
In order to account for survival bias, we performed a subgroup analysis according to the year the index case was diagnosed with ovarian cancer (pre vs post 2012). This strategy was adopted because the prevalence of pathogenic BRCA1/2 variants detected in women diagnosed with ovarian cancer before 2012 may have been biased by long-term survivors.27 28 In women diagnosed with ovarian cancer before 2012, the minimum time from the diagnosis of ovarian cancer to subsequent germline BRCA1/2 testing was 18 months (January 2012 to June 2013), an interval that approximates to half the median overall survival for ovarian cancer.17
Germline BRCA1/2 testing
Germline BRCA1 and BRCA2 variants were detected by testing DNA extracted from peripheral circulating lymphocytes. Next-generation sequencing (NGS) was used to detect variants throughout the whole coding sequence of BRCA1 and BRCA2, including at least 15 base pairs beyond each exon–intron junction. Enrichment occurred using a custom-designed long-range PCR-based approach followed by a normalisation step using SequalPrep normalisation plates and library preparation using the Illumina Nextera DNA Library Preparation Kit. NGS analysis was on an Illumina MiSeq using v2 2×150 base pair sequencing chemistry. Single nucleotide variants and small deletions, duplications, insertions and insertion/deletions (<40 base pairs) were called using a bioinformatic pipeline validated to detect heterozygous and mosaic variants in NGS data to an allele fraction of ≥4%. The bioinformatic pipeline was developed for use across a broad range of inherited cancer syndromes, some of which have a high frequency of somatic mosaicism, for example, neurofibromatosis type 2. An allele fraction cut-off of ≥4% for variant detection was determined following clinical validation, as this was the lowest allele fraction limit of detection where both sensitivity and specificity remained high.
Testing for large genomic rearrangements/copy number variation (eg, whole exon or whole gene deletions/duplications) in BRCA1 and BRCA2 was performed by multiplex ligation-dependent probe amplification (MLPA).29 The MLPA MRC Holland probe kits P002-D1 (BRCA1) and P045-C1 (BRCA2) were used to analyse germline DNA. Amplified ligation products were subject to fragment analysis using an ABI 3130xl Genetic Analyser and size called using GeneMapper v2.0 (Applied Biosystems). Copy number status calling was performed using data exported from GeneMapper using custom-developed MLPA spreadsheets that report relative dosage quotient for each probe compared with five reference control samples. All MLPA analysis assays were performed in duplicate for confirmation of results.
Manchester BRCA scoring system
The Manchester Scoring System is a simple-to-use, paper-based model that can be used to determine the combined BRCA1 and BRCA2 carrier probability of an index case with a relevant cancer (table 1).30 The development of the Manchester Scoring System was based on empirical data gathered from the Manchester mutation-screening programme.31 Each individual, from one side of the family, is scored for each gene separately, BRCA1 and BRCA2 (table 1). For index cases of breast cancer or any index case or unaffected relative of an index case of ovarian cancer (<60 years), the BRCA1 and BRCA2 scores are adjusted according to pathology.30 The pathology adjustment takes into account the higher prevalence of germline pathogenic BRCA1/2 variants in triple-negative breast cancer and high-grade serous ovarian carcinoma.32 A Manchester Score of 15–19 points equates to a combined BRCA1 and BRCA2 probability of 10%, and 20 points to a 20% probability.30
Five hundred and fifty-seven women of non-Jewish ancestry underwent germline BRCA1 and BRCA2 testing following a diagnosis of epithelial ovarian cancer (table 2). A total of 103 women (18%) had a pathogenic BRCA1/2 variant (68 BRCA1, 35 BRCA2) (table 2). The mean age at which ovarian cancer was diagnosed differed in patients with pathogenic BRCA1 (51.9 years (range 36–76)) and BRCA2 (59.4 years (range 33–86)) variants. The types of pathogenic BRCA1/2 variants detected are reported in table 3. Twenty-three BRCA1/2 variants of unknown clinical significance (class 3) were detected.
Pathogenic BRCA1/2 variants were most commonly detected in women diagnosed with high-grade serous ovarian cancer, although women diagnosed with this histological subtype were most frequently screened (table 2). All women diagnosed with germline BRCA-mutant endometrioid ovarian cancer had poorly differentiated (high-grade) tumours. No BRCA1/2 heterozygotes had low-grade serous, low-grade endometrioid, undifferentiated or mucinous ovarian cancer (table 2).16 One woman diagnosed with FIGO stage 3C carcinosarcoma of the ovary had inherited a germline BRCA1 variant, although the epithelial histological component of her invasive tumour was high-grade serous. Eighty-four women (15%) had been diagnosed with breast cancer and 268 (48%) had a first-degree or second-degree relative with breast and/or ovarian cancer (table 2).
The prevalence of pathogenic BRCA1/2 variants was >10% in women diagnosed with ovarian cancer under the age of 60 years (21%) (table 4). Also, the prevalence of pathogenic BRCA1/2 variants was >10% in women diagnosed at ≥60 years of age with a family history of breast and/or ovarian cancer (17%) or a medical history of breast cancer (34%) (table 4). In women diagnosed with sporadic ovarian cancer at ≥60 years of age, the prevalence of pathogenic BRCA1/2 variants almost reached 10% (7/76) (table 4). However, in women diagnosed with sporadic ovarian cancer at ≥60 years of age without a medical history of breast cancer, the prevalence of pathogenic BRCA1/2 variants fell below 5% (2/46).
Survival bias may have affected the prevalence of pathogenic BRCA1/2 variants detected in the subgroup of women diagnosed with ovarian cancer at ≥60 years of age with a family history of breast and/or ovarian cancer, prior 2012, although the difference was not statistically significant (24% vs 15%; Fisher’s exact test: p=0.21) (table 5).
The prevalence of pathogenic BRCA1/2 variants was >10% (101/463) in all women with a Manchester BRCA Score of ≥15 points, and there was a stepwise increase in prevalence as the Manchester Score increased (table 6). In contrast, in women with a Manchester Score of <15 points, the prevalence of pathogenic BRCA1/2 variants was substantially <10% (2/94) (table 6).
Risk stratification by age alone confirmed that women diagnosed with epithelial ovarian cancer under the age of 30 years were unlikely to have a germline pathogenic BRCA1/2 variant (table 7).
By testing germline DNA in women diagnosed with epithelial ovarian cancer across North West England, we found the overall prevalence of pathogenic BRCA1/2 variants exceeded 10% (103/557) (table 2). Furthermore, by separating groups according to established risk factors for hereditary breast and/or ovarian cancer syndrome, we found that the prevalence of pathogenic BRCA1/2 variants was consistently >10% in those women diagnosed with ovarian cancer under the age of 60 years and in those diagnosed over the age of 60 years with either a family history of breast and/or ovarian cancer or a past medical history of breast cancer (table 2).
A number of studies have also assessed the prevalence of germline pathogenic BRCA1/2 variants in ovarian cancer. In an East of England series (GTEOC study), the prevalence of germline pathogenic BRCA1/2 variants among all high-grade serous and endometrioid ovarian cancer cases was 8% (18/232) and increased to 12% (17/146) in women diagnosed under the age of 70 years, but fell to 1% (1/86) in those aged ≥70 years.19 Similarly, in a Scottish series, the prevalence of pathogenic BRCA1/2 variants among unselected non-mucinous epithelial ovarian cancer was 13.1% (31/236), but fell to 8.2% (13/159) in women diagnosed over the age of >70 years.20 In an unselected series from Europe (AGO-TR-1 trial), the prevalence of pathogenic BRCA1/2 variants in epithelial ovarian cancer was 20.8% (109/523) and fell to 10.6% in women diagnosed at ≥60 years of age, but increased to 31.9% (71/109) in women with a family history of breast or ovarian cancer.33 Moreover, in a large Australian study, the prevalence of pathogenic BRCA1/2 variants in non-mucinous ovarian cancer was 14.1% (141/1,001), but fell to 8.3% (38/457) in women diagnosed at ≥61 years of age, 11.2% (103/738) in women without a personal history of breast and 8.3% (62/749) in women without a family history of breast and/or ovarian cancer.11 The data from these series and our study, therefore, suggest that three clinical features could be used to risk stratify for testing for germline BRCA1/2 variants in women diagnosed with ovarian cancer, including age at diagnosis, family history of breast and/or ovarian cancer and medical history of breast cancer. This is important if criteria for selecting which patients to tests are used by funding bodies.
In our study, across the North West of England, selection criterion for germline BRCA1/2 testing was mostly based on an individual’s pathology adjusted Manchester Score of ≥15 points, with 17% (94/557) falling below the 15-point threshold.30 This scoring system provides an alternative method for determining whether an individual’s combined BRCA1 and BRCA2 carrier probability is ≥10% (table 1). In our series, a Manchester Score of ≥15 points was associated with a>10% prevalence of pathogenic BRCA1/2 variants, whereas a Manchester Score of <15 points was associated with a prevalence substantially <10%. Furthermore, one of the BRCA2 heterozygotes with a Manchester Score <15 had a strong family history of prostate cancer with two first-degree relatives diagnosed at <60 years old, giving Manchester Score of 14 (ovarian cancer <60 [5+5], 2 x prostate cancer <60 [+2,+2]). Overall, therefore, the Manchester Score provides a better trade off of sensitivity and specificity than simply excluding women with sporadic ovarian cancer diagnosed after the age of 60 years.
Although this study is unlikely to unduly influence the debate regarding universal germline BRCA1/2 testing in unselected populations of women diagnosed with ovarian cancer versus those at higher risk of inheriting a variant, we consider a number of potential problems with unselected screening beyond the obvious financial burden. First, pathogenic BRCA1/2 variants occur much less frequently in non-high-grade non-serous ovarian carcinoma.11 19 20 Indeed, somatic mutations in other genes are more commonly found in non-high-grade non-serous epithelial subtypes, including PIK3CA, PTEN, KRAS, BRAF, ERBB2 and ARID1A.34–38 Moreover, at present, PARP inhibitors are only licensed in high-grade serous and endometrioid subtypes. Therefore, there does not seem to be a biological rationale or therapeutic incentive for unselected germline BRCA1/2 testing in non-high-grade non-serous/endometrioid subtypes.
Second, if unselected germline BRCA1/2 testing becomes the prerogative of oncologists, the additional clinical expertise provided by geneticists may be lost.39–42 No BRCA1/2 test is 100% accurate for all variants, and therefore, accepting a diagnosis of BRCA1/2 wild type or variant of unknown clinical significance in an index case with a strong family history of cancer may be naive. Many NGS-based assays in use will identify variants in the coding regions of BRCA1/2±5–10 base pairs either side of the intron–exon junction, but these assays would not detect rarer pathogenic variants such as deep intronic variants or those located in 5′-untranslated regions.43–45 Furthermore, initially reported variants of unknown significance can be reclassified following further investigations such as segregation analysis, RNA sequencing or additional data from case–control analyses.39 This level of genetic scrutiny only occurs in specialist genetics departments. There is therefore some concern that women diagnosed with epithelial ovarian cancer who have a strong family history of cancer, may evade further necessary diagnostic investigations that would be performed by geneticists, if they are labelled as BRCA1/2 wild type or variant of unknown clinical significance by oncologists alone.
Finally, by only screening for germline BRCA1/2 variants, there is a risk of missing other moderate-to-low penetrance actionable cancer-predisposition genes, such as RAD51C/D, BRIP-1, MLH1, MSH2/6 and PMS2.24 The prevalence of each individual cancer-predisposition gene is too low in ovarian cancer to warrant screening in an unselected population,46–48 however, there is a risk that by focusing testing solely on BRCA1 and BRCA2, other cancer-predisposition genes will remain undetected. In the North West of England, if a woman diagnosed with BRCA1/2 wild-type ovarian cancer has a Manchester Score of ≥20 points, she is offered extended panel testing for alternative germline variants. We would, therefore, recommend that any patient diagnosed with ovarian cancer and having a family history of cancer should be referred to the local genetic department irrespective of their BRCA1/2 status.
There are some limitations of the study. Our study was biased by including mostly women with high-grade serous ovarian cancer and established risk factors for hereditary breast and/or ovarian cancer syndrome. Although we are confident that our series represents an almost comprehensive investigation of patients with high-grade serous ovarian cancer diagnosed under the age of 60 years, we acknowledge that a comparably smaller number of women diagnosed with ovarian cancer later than 60 years of age were tested, especially those without risk factors for hereditary breast and ovarian syndrome. Consequently, the overall prevalence of germline pathogenic BRCA1/2 variants reported in our study should be interpreted in the context of a selected population of women diagnosed with epithelial ovarian cancer.
In conclusion, the findings from our study suggest that if a 10% pretest probability threshold is required prior to germline BRCA1/2 testing in ovarian cancer, then using age at diagnosis, a family history of breast and/or ovarian cancer, a medical history of breast cancer or a Manchester Score of ≥15 should provide appropriate risk prediction.
Contributors Design and conception of study: RDM and DGRE. Data collection and assembly: RDM, MB and DGRE. Data analysis and interpretation: RDM, AJW, GCJ and DGRE. Writing of manuscript: RDM, GJB, NF, EJC, AJW, GCJ and DGRE. Critical review and revision of manuscript: RDM, GJB, NF, MB, ARC, JH, CM, HS, ERW, FL, EJC, RJE, AJW, GCJ and DGRE. Submission of manuscript: RDM
Funding DGRE, EJC and ERW are supported by the all Manchester NIHR Biomedical Research Centre (IS-BRC-1215-20007).
Competing interests None declared.
Patient consent Obtained.
Ethics approval The germline BRCA1/2 database is approved by North Manchester Research Ethics Committee (08/H1006/77).
Provenance and peer review Not commissioned; externally peer reviewed.