Current guidelines for BRCA testing of breast cancer patients are insufficient to detect all mutation carriers

Germline mutations in the BRCA1 and BRCA2 genes are associated with a high lifetime risk of breast and ovarian cancer [1, 2]. Knowledge of one’s BRCA status is of importance for healthy women as cancer may be prevented through risk-reducing mastectomy and salpingo-oophorectomy [3‐5]. Identification of a pathogenic BRCA mutation in a woman diagnosed with breast cancer (BC) may influence treatment and prognosis of her current cancer but also enable prevention of future cancers [6‐12]. Consequently, surgeons and oncologists more and more frequently want to offer genetic testing at time of diagnosis.

Because of the high costs associated with genetic analyses, BRCA1/2- testing has traditionally been restricted to BC patients having an a priori high risk of being a carrier. These factors include young age at diagnosis (below 45 years), triple-negative breast cancer (TNBC) or a family history of breast- and/or ovarian cancer [13‐22]. The American Society of Clinical Oncology (ASCO), The National Comprehensive Cancer Network (NCCN) in the US and the Norwegian Breast Cancer Group (NBCG) all have guidelines for BRCA testing of BC patients based on these risk factors (Additional file 1: Figure S1), and according to The National Institute for Health and Care Excellence (NICE) in the UK, BRCA testing should be offered to BC patients with a probability of having a mutation is 10% or more [23‐26]. There are also corresponding guidelines for predictive testing of healthy women.

During the recent years, the cost of genetic testing has decreased due to the advent of new and more efficient DNA-sequencing technologies. Consequently, BC patients are now often offered multi gene panel testing. These panels include BRCA1/2 and the other high risk breast cancer genes TP53 and PTEN, but also genes with more moderate cancer risk and genes whose clinical significance is still not resolved [27, 28]. Testing is nevertheless still mostly restricted to patients fulfilling certain high risk criteria for being mutation carriers, and few studies have described BRCA testing of unselected groups of BC patients [29‐35]. To our knowledge, only two studies have performed testing with sequencing and Multi-Ligation Probe Amplification (MLPA) of all patients included [30, 35]. Knowledge of the clinical characteristics of BC caused by BRCA mutations in unselected BC cohorts is therefore limited. Moreover, there is also limited information about the sensitivity and specificity of current guidelines for BRCA testing to identify carriers in cohorts not selected for high risk factors. With the ongoing changes in opportunities for genetic testing we believe it is necessary to assess whether the current strategies for BRCA testing are sufficient to enable mutation positive women to benefit from the potential of both cancer cure and prevention that lies within such testing. Observations from BRCA testing of less selected groups of BC patients are necessary for this evaluation.

The NBCG guidelines used in Norway are regularly revised. Because it became clear that identification of a BRCA mutation could have implications for treatment, a subjective criteria was introduced a few years ago. If the treating physician considered the test result to be of importance for treatment decisions, testing could be offered even in the absence of other high risk factors such as young age or family clustering. As a consequence, testing could be offered also to BC patients with an a priori low risk of being carriers. Due to this change in practice we have been able to compare the sensitivity of previous and present national and international guidelines for BRCA testing in BC patients without the selection bias described.

This report summarizes the results of BRCA testing in South-Eastern Norway according to these revised Norwegian guidelines from 1st of January 2014 to 31st of August 2015. The study had three specific aims: Firstly, it was to calculate the prevalence of BRCA mutations in this cohort of BC patients that as a whole had an a priori low risk of being mutation carriers, describe the spectrum of mutations, and the number of mutation positive female relatives identified. Secondly, we wanted to describe and compare clinicopathological features of BC among carriers and non-carriers. The third aim was to calculate the sensitivity and specificity of different guidelines used for diagnostic testing [23‐26], and also to evaluate how many mutation carriers that had a family history of cancer that qualified for predictive testing before they were diagnosed with BC [26].

We identified a mutation in 3.1% of BC patients. In a recent study from the Western region of Norway, 405 BC patients were tested for 30 specific BRCA1/2 mutations and with MLPA [32]. Sequencing was performed on 94 of these. A mutation was found in only 1.7% of participants. Both studies are small and consequently they do have limitations. However, the observed difference may at least partly be explained by the fact that all patients in our study were tested with sequencing and MLPA and not for selected mutations only. In our study, 16 out of 29 (55%) women with Norwegian ancestry did not have any of the 10 most common Norwegian founder mutations [39], and five (17%) had a mutation that had not been previously observed in our population. In comparison, in 2007 the 10 founder mutations accounted for about two-thirds of all detected mutation carriers at our department [39]. This reflects that in 2007 most patients were tested for a limited number of mutations, whereas today sequencing and MLPA is offered to all who qualify for testing in our health region. Our findings also illustrate that there are mutations within our population that are and may remain rare. By testing only for frequently observed mutations in the Norwegian population, a substantial number of mutation positive women with a pathogenic BRCA mutation will not be found.

A VUS was identified in 4.9% of the tested patients. Our numbers are comparable to what others have revealed [43]. Studies have reported that physicians, with limited formal training in genetics, may misinterpret VUS results [44‐46]. This was dealt with in the current study as all patients with a VUS were referred to genetic counseling. There is a worry that information about a VUS may have a negative psychological impact on the patient [47]. However, studies have also demonstrated that it is interpreted as more similar to a test result where no pathogenic variant has been detected than to a result with an identified pathogenic variant [46]. Addressing the issue of patients’ interpretation of risk and possible psychological impact was beyond the scope of this study, but should be closer evaluated in future studies. By offering testing only for a set of already known and described mutations one would avoid the challenges associated with identifying VUS. We have however described that a substantial number of mutation carriers will be missed by testing only for known mutations. It is our opinion that the benefits associated with identifying all carriers (and the corresponding risk associated with not identifying a mutation carrier) outweigh the current challenges associated with identifying VUS. One may also hypothesize that the frequency of VUS may decrease in the future as more people are undergoing testing.

By comparing carriers and non-carriers tested at OUH-U we observed that even though testing was offered broadly, mutation positive women still differed from mutation negative in terms of the known high risk aspects for being carriers: age of onset, triple negativity and family history. We found no difference in HER2-status between the two groups, and these findings are in accordance with a recent study where HER2-status was not found to be a reliable predictor of BRCA-status [48]. Mutation carriers had a higher score for Ki67 than mutation negatives, and this has also been described in a few studies [49, 50]. The observed differences between the two groups are also illustrated by the fact that each of the test criteria has a high specificity (see Table 4).

Whereas the mutation positive differed as a group from mutation negative, selecting patients for testing based on the known high risk factors will identify carriers with varying sensitivity (see Table 3). Testing only those with BC below 40 years or TNBC identified 31.6% and 34.2% of carriers respectively, and less than 50% of carriers qualified for testing according to the NICE guidelines. By use of the comprehensive ASCO, NCCN and old NBCG criteria, where the different single characteristics are combined in order to increase sensitivity, between 84.2 and 89.5% would be identified. NBCG has recently suggested that testing should be offered to women with TNBC under the age of 60 [26]. By adding this aspect to the original stringent criteria, 34/38 (89.5%) would have been identified. In a recent study where 488 women with BC were tested for mutations in 25 cancer susceptibility genes, Tung et al. found that all BRCA-mutation carriers fulfilled the NCCN guidelines [35]. We do not know whether the difference in observed sensitivity is due to chance or systematic differences between the two cohorts.

The ASCO, NCCN and NBCG criteria include an assessment of the patient’s family history of cancer. The family histories of the mutation positive BC patients identified in our study were thoroughly investigated by genetic counselors and medical geneticists following the identification of the mutation, resulting in the sensitivity estimates presented. The observed estimates may therefore be higher than what is realistic in the clinical setting when family history is taken by the admitting physician at time of diagnosis. It may be difficult for the patient to know or recall detailed information about their family history of cancer when asked in a possibly stressful diagnostic setting. In line with this, Høberg-Vetti et al. found in their study from the Western part of Norway that 2 out of 26 (7.7%) mutation carriers reported a negative family history of cancer at time of diagnosis and testing, but closer evaluation revealed that they did have a family history of breast and/or ovarian cancer [32]. We also worry that the complexity of the NCCN, ASCO and NBCG criteria make them difficult to use and implement systematically in a busy clinical setting. Both these aspects could lead to fewer patients being offered testing, even those fulfilling the criteria. This is illustrated in several studies. Febbraro and colleagues observed that only 34% of breast cancer patients fulfilling NCCN guidelines were referred to genetic counseling and testing [51]. In a recent Swedish study where all BC patients were tested retrospectively, it was found that 65% of the mutation carriers fulfilled Swedish criteria for testing, but only 18% had been identified in regular clinical routine [52]. Moreover, even though all mutation carriers fulfilled the NCCN criteria in the study by Tung et al., 13.3% of the carriers identified through this research project had not been tested clinically [35].

The fact that 37% of the women had a family history of cancer that according to the Norwegian guidelines qualified for referral to predictive genetic testing before their own disease, may be another illustration of the challenges with using assessment of family history as a criteria for genetic testing or referral to genetic counseling. The low number leads us to conclude that the current system of referring healthy women to genetic testing based on their family history is suffering from lack of compliance. These women contracted cancers that could have been prevented had they known about their risk and undergone prophylactic surgery. The reasons for this lack of referral and how it can be improved need to be further explored, but this was not the scope of the current study.

Using age of onset as a criteria for testing will likely lead to increased adherence by surgeons and oncologists compared to guidelines requiring a detailed and complicated assessment of the patient’s family history of cancer. Testing all BC patients below 60 years identified as many or more carriers than all guidelines assessed (see Table 3). Due to the lowered cost of testing and the clinical impact of detecting a BRCA mutation, Finch et al. [53] have recently argued that the threshold for testing should be lowered from a 10% prior probability of being a carrier to 5%. Testing all under 60 in the OUH-U cohort gave a mutation detection rate of 5.5% (see Table 5), i.e. within this threshold. By using this criteria one would have to test 18 BC patients to identify one carrier. As of August 2016, testing these 18 patients had also led to the identification of one female relative per index patient. In Cohort 1 from OUH-U, 235 out of the 440 tested (53.4%) were younger than 60 and 132/440 (30%) fulfilled the old NBCG criteria (see Table 2). In 2014, 3324 Norwegian women contracted BC [54]. Using the calculations from the OUH-U cohort indicate that testing all below 60 years will involve 800 more analyses annually compared to testing only those fulfilling the old NBCG criteria.

One year after the last BC patient in our cohort was tested, 1.1 female relative per identified carrier had tested positive for the mutation and were given the opportunity of cancer prevention. It is likely that this number will increase as more relatives are informed and tested. According to Finch et al., “the value of a cancer genetic testing program comes from the number of cancers prevented” [53]. Even though testing all below 60 years may be feasible and effective, we observed that 10% will still be missed by this strategy. Two mutation carriers were older than 70 years. One may argue that the identification of a mutation in a woman who is 70 years or older may not influence treatment decisions, life expectancy or lead to a significant gain in quality adjusted life years (QALY) for this woman. However, it is likely that women over 70 have adult female relatives that may be at high risk of cancer due to the mutation.

We observed that more than half of the mutation carriers did not have a family history of breast and/or ovarian cancer before they were diagnosed with breast cancer themselves. These findings are in line with other studies reporting that family history has limited value in predicting carrier status [33, 38, 55], and our findings illustrate the difficulties with finding these women prior to disease development. Today, these women cannot obtain genetic testing while still healthy, as a population-based screening protocol is not accessible. Mary Claire King and colleagues consider that the identification of “a woman as a carrier only after she develops cancer is a failure of cancer prevention” [56] and based on their finding that BRCA mutation carriers have a high risk of cancer regardless of their family history [57], argue for population based screening to all women aged 30 years [56].

In a cost analysis of the cancer genetic services in the UK, Slade et al. have demonstrated that the most cost efficient genetic service model is to identify unaffected mutation carriers through an affected mutation positive index person [58], and argue for more comprehensive testing of all cancer patients fulfilling the NICE criteria. Patients fulfilling these criteria have an a priori 10% risk of being carriers. We identified a mutation in 3.1% of carriers, and one may argue that this is too low to warrant testing of all BC patients. We have however, recently shown that the practice of BRCA testing at OUH-U is cost-effective within the frequently used thresholds in Norway [59]. The cost-effectiveness was mainly due to the prevented breast-and ovarian cancers in their female relatives who tested positive for the mutation. Possible life years gained (LYG) due to prophylactic surgery among the BC patients was not included in the calculations in this study. The calculations may therefore be considered a conservative estimate. In addition, the cost of testing is constantly dropping, making the cost-effectiveness of a broad application of BRCA testing to BC patients even larger in the coming years.

Our results indicate that by testing only for founder mutations in the BC population of the South-Eastern part of Norway, and by testing only those with a family history of cancer, a significant number of mutation carriers will be missed. One may ask whether these results are relevant for screening strategies in other populations. The prevalence of BRCA mutations vary between populations [34, 60, 61], and the indication for genetic screening of all breast and ovarian cancer patients may be stronger in populations with a higher frequency of mutation carriers than in Norway. In populations where there is a stronger founder effect, the number of mutation carriers missed by offering testing for only founder mutations will be lower than what we have observed. However, recent studies have demonstrated that 13% of BRCA1 mutations and 7.2% of BRCA2 mutations in Ashkenazi Jews were non-founders [62]. Similarly, a Polish study found that in families with a family history of breast and/or ovarian cancer having tested negative for Polish BRCA founder mutations, sequencing revealed 31 other BRCA mutations. The detection rate of these mutations was 10% [63]. Sequencing and MLPA may therefore be warranted also in populations with a stronger founder effect than in Norway. We observed that only 40% of mutation carriers had a family history of breast and/or ovarian cancer. There are various reasons for this: Small family size, mutations may be inherited through several generations of men and incomplete penetrance. Family history as a selection tool for testing may have a higher sensitivity in populations with higher birth rates than in Norway. However, most western countries have had a declining birth rate since the 1960s and now have a birth rate between 1.5 and 2 [64]. One may therefore hypothesize that the value of using family history as a selection tool for testing will be even lower in the future.

BC patients are now often offered multi gene panel tests, and this is the direction in which the field of genetic testing is moving rapidly. There are several advantages with this strategy compared to testing only for the BRCA genes. More carriers of pathogenic mutations in other known BC risk genes such TP53, PTEN or PALB2 will be identified. In addition, carriers of mutations in genes that likely would not have been investigated when testing only for one gene at a time will be identified. By testing a sequential series of breast cancer patients for 25 cancer predisposition genes, Tung and colleagues identified carriers of mutations in the MSH6 and PMS2 genes [35]. Conversely, by testing families suspected to have Lynch Syndrome for 112 known or candidate colorectal cancer genes, Hansen and colleagues identified one BRCA1 carrier and two BRCA2 carriers [65]. In sum, through multi gene panel testing more mutation carriers and their mutation positive relatives will be identified and given the opportunity of appropriate cancer surveillance and/or prevention. In the coming years this technology will also likely become more cost effective than traditional Sanger sequencing of one gene at a time. The aim of this study was not to argue against the value of multi gene panel testing, but rather to investigate whether the current strategies for BRCA testing, regardless of technology used, are sufficient to identify all carriers of mutations in these well-known and defined genes.

One limitation to our study is that we have not tested all BC patients. In the OUH-U cohort (Cohort 1) 167/607 = 27.5% of all women diagnosed with BC were not tested. These women were older and fewer filled the NBCG criteria than those who were tested. Unfortunately, we do not have access to the exact number of untested patients in the SERHA series or clinical information about these. If 2400 were treated in SERHA in the study period, about 39% of these (931/2400) were tested. The reason for the lower number of tested in Cohort 2 may be that there was a lower awareness of the possibility of genetic testing at these hospitals, but we cannot exclude that this cohort may be more selected. To assess this, we compared the two cohorts indirectly by comparing the mutation positive BC patients. There was a tendency towards a higher age of onset in Cohort 2 from SERHA, but this difference was not statistically significant. No significant differences were found between mutation carriers in the two cohorts in terms of TNBC and family history (Additional file 3: Table S2). We also observed the same frequency of mutation carries in the two series. The two cohorts may therefore be similar, and it is likely that the untested in Cohort 2 were older and that fewer filled the NBCG criteria than the tested. If there are mutation carriers among the untested in both series, the total frequency of carriers might have been lower, but it is likely that even fewer would have fulfilled the different high-risk criteria.

By offering BRCA testing to a broad group of BC patients we found that 3.1% carried a deleterious mutation, and so far this has led to the identification of 1.1 female mutation positive relative per mutation positive BC patient. Even though mutation carriers differed as a group from mutation negative, criteria for testing based on the high-risk aspects did not detect all BRCA carriers in this BC population. Testing all BC patients below 60 years had a sensitivity matching the commonly used guidelines, and will likely be easier to apply, but 10% of mutation carriers would still be missed. Thirty-seven percent of the women had a family history of cancer prior to their own BC that qualified for predictive genetic testing. They contracted cancers that could have been prevented if the health care system had identified their increased genetic risk. Based on our combined observations, we conclude that the current strategies for BRCA testing are insufficient to detect all carriers. We suggest that it is time to discuss whether BRCA testing should be offered also to BC patients not belonging to a high risk group. If all BC patients are offered BRCA testing, the potential for cancer cure and prevention associated with such testing can be improved even further than what today’s strategies for testing allows. In case of lack of economic resources to fulfill this strategy, at least those aged 60 years or less at time of BC diagnosis should be tested. Our observations also indicate that health services need to be aware of referral possibilities for healthy women with cancer in the family, and the reasons for the low compliance should be explored. Improved strategies both for diagnostic and predictive BRCA testing will identify more mutation positive women prior to cancer development than the current practice.