Article Text
PDF
Abstract
Background Previous publications and utilisation of risk models for BRCA1 and BRCA2 mutation identification suggests that multiple primary disease in an individual is a strong predictor of a BRCA1/2 mutation and that this is more predictive than the same cancers occurring in close relatives.
Methods This study assessed the pathological mutation detection rates for BRCA1, BRCA2 and the CHEK2c.1100 delC mutation in 2022 women with breast cancer, including 100 with breast/ovary double primary and 255 with bilateral breast cancer.
Results and discussion Although detection rates for mutations in BRCA1/2 are high at 49% for breast/ovarian double primary and 34% for bilateral breast cancer, the differential effect of multiple primaries in an individual appears to have been overestimated, particularly in those families with only a few malignancies. Nonetheless, bilateral breast cancer does differentially enhance detection rates in strong familial aggregations. CHEK2 1100 DelC mutation rates were lower in bilateral than for unilateral cases at 0.8% compared to 2%. The detected mutation rates for isolated double primary breast and ovarian cancer was 14% (3/22) compared to 17% (17/99) for the same two primaries in two close relatives in families with no other cases of breast/ovarian cancer. Risk models may need to be adjusted if further studies corroborate these findings.
- Diagnostics tests
- obstetrics and gynaecology
- cancer: breast
- genetic epidemiology
Statistics from Altmetric.com
Introduction
Since the identification of the BRCA11 and BRCA22 genes, there has been widespread testing of families and isolated cases with breast or ovarian cancer or a combination of the two. While the chances of identifying a BRCA1/2 mutation increases with the increasing number and decreasing age of cancers in the family, there has been a suggestion that families containing individuals with breast and ovarian double primaries in particular have very high rates of mutation.3–5 In contrast double primary/bilateral breast cancer has not been thought to be such a strong predictor6 and may be over scored in current risk models.7 More recently a third susceptibility gene, CHEK2, was associated with increased risk of breast cancer in families without BRCA mutations.3 In contrast to BRCA1/2, the CHEK2c.1100 delC mutation has been reported to be strongly associated with bilateral breast cancer.8–10
In our population we did not feel that the predictive value of double primary disease was as great as the literature suggested or that the existing models predicted. We therefore aimed to assess the predictive value of double primary disease in identifying mutations in BRCA1, BRCA2 and CHEK2. This study reviewed the screening for BRCA1/2 mutations in North West England among non-Jewish kindreds and assessed the impact of the presence of multiple primary disease on the identification of BRCA1/2 mutations and the presence of CHEK2c.1100delC.
Methods
Index cases and relatives
Breast and ovarian cancer families from North-West England have been screened for BRCA1/2 mutations since 1996. The greater Manchester area is below average for socioeconomic status and life expectancy is shortened and cancer incidence generally higher, although breast cancer rates are similar to UK averages. Both confirmation of cancer diagnoses and mutation testing is undertaken with informed consent. The study was carried out with local ethical committee approval. Women who attend the specialist genetic clinics in this region with a family history of breast/ovarian cancer have a detailed family tree elicited with all first, second and if possible third degree relatives recorded. All cases of breast or abdominal cancers are confirmed by means of hospital/pathology records, from the Regional Cancer Registries (data available from 1960), or from death certification, with particular attention to confirming a diagnosis of ovarian cancer and contralateral breast cancer to accurately identify cases of double primaries.
Cases were selected for local NHS mutation testing if they were assessed as having at least a 20% chance of a mutation being identified using the Manchester Score. This included all individuals with double primary breast/ovary. Cases below the 20% threshold, but in whom there were at least two breast cancers in the family, were selected for research testing. Only individuals affected with breast or ovarian cancer were selected for testing as per UK guidelines. Index cases were tested with a whole gene approach which since 2000 has consisted of sequencing of all BRCA1 and BRCA2 coding exons, including the flanking splice donor and acceptor sites and multiple ligation dependant probe amplification (MLPA) (MRC-Holland P002 and P045 probe sets). As part of the MRC-Holland P045 MLPA kit a test for the CHEK2c.1100delC mutation is incorporated. At total of 1140 samples not meeting the National Institute for Health and Clinical Excellence (NICE) guidelines for NHS testing have been tested on a research basis at the Institute of Cancer Research in Sutton, employing a similar whole gene approach. The full coding sequence and intron–exon boundaries of both BRCA1 and BRCA2 were screened in 86 fragments by conformation sensitive gel electrophoresis (CSGE) as previously described,11 as well as MLPA and CHEK2 c.1100 delC testing. All families were subject to the same degree of cancer confirmation. All pathogenic mutations for BRCA1 and BRCA2 are noted and placed onto a dedicated Filemaker Pro 7 database. Women with self reported Jewish ancestry were initially multiplex tested for the three common mutations and all such women were excluded, as were families identified through testing in other genetics centres.
All 2022 females (882 NHS plus 1140 research) affected with any case of breast cancer and tested for BRCA1/2 mutations were included in this study along with 36 cases of ovarian malignancy with a family history of a single breast cancer (for comparison with an isolated double primary). Data were collected on the family history, age of cancer diagnosis, the presence of further primary cancer, and whether or not a male breast cancer or ovarian cancer was confirmed in a relative. The Manchester scoring system was used to assess the significance of the breast/ovarian cancer history.11 This system assesses the likelihood of a BRCA1/2 mutation and scores breast and ovarian cancers individually in the family, giving a higher score the younger the age at diagnosis.11 A combined score of 20 reflects a 20% likelihood of identifying a BRCA1/2 mutation, with a score of 15 representing a 10% threshold. All details including the Manchester score were entered onto the Filemaker Pro 7 database. Proportions of each subset of double primary identified with mutations were compared where appropriate with χ2 testing. Carrier probabilities with BRCAPRO12 and Myriad tables13 were calculated with the CancerGene software package from the University of Texas Southwestern Medical Center (CaGene version 4.0). Additional information about CancerGene is available at http://www.utsouthwestern.edu.
Results
In total, 2022 females with breast cancer and 36 with ovarian cancer only were screened for mutations in BRCA1 and BRCA2 with a whole gene approach. A pathogenic BRCA1 or BRCA2 mutation was identified in 322 (16%) women. Table 1 shows the proportion of mutations identified in families containing either a case with breast–ovary double primary or a pair of first degree relatives where one had breast cancer and the other ovarian cancer. All cases of breast–ovary double primary were confirmed mostly from pathology and cancer registry data. Although the overall mutation rate in families containing a case with breast–ovary double primary was high at 49/100 (49%), there was no difference in mutation detection between an isolated (sporadic) breast–ovary double primary 3/22 (14%) and a breast and ovarian first degree relative (FDR) pair without any other family history 17/99 (17%) (p=0.93). Only the FDR pairs with breast and ovarian cancer diagnosed with a Manchester score of 20 or higher breached the 20% UK threshold for testing.14
Detection rates for BRCA1 and BRCA2 mutations in breast ovary double primaries and first degree relative (FDR) pairs with breast (br) and ovarian (ov) cancer
Comparison of the mutation detection rates in women with bilateral breast cancer are shown in table 2. Attempts were made to confirm all double primaries and >90% were confirmed as separate primaries mostly from pathology and cancer registry data. In 21 cases there was clear evidence of a previous contralateral breast cancer treatment, but histology was not obtainable and these were still included. There was a range of 0–36 years between primaries with a mean of 8.2 years. In eight cases the first primary was DCIS (ductal carcinoma in situ) and in three the contralateral was in situ disease. The detection rate was higher in those with bilateral disease (34%, 87/255; 44 BRCA1; 43 BRCA2) versus unilateral disease (14.5%, 255/1767) (table 3). Rates increased to 61% (27% BRCA1 and 34% BRCA2) and 34% (25% BRCA1 and 9% BRCA2) in families with an ovarian cancer case and 82% (9% BRCA1 and 73% BRCA2) and 28% (3% BRCA1 and 25% BRCA2), respectively, in families with a male breast cancer (MBC) case. The rates decreased to 23% in bilateral cases and 9.5% in unilateral breast cancer cases without a family history of either ovarian or male breast cancer. Selection of women with their first primary aged <45 years also increases the likelihood of detecting a BRCA1/2 mutation, but this effect in bilateral cases is largely accounted for by the age and number of breast cancer cases in the family, as there is not any differential increase in detection rates in families with similar Manchester scores. The distribution of unilateral and bilateral cases also showed a clear dependence (data not shown).
Detection rates for BRCA1 and BRCA2 mutations in breast double primaries and predictions from models
Rates of BRCA1, BRCA2 and CHEK2 c.1100 delC mutations in bilateral and unilateral breast cancer
CHEK2 c.1100delC mutations were found slightly less frequently among the bilateral breast cancer group with 2/240 (0.8%) compared to 33/1660 (2%) in the unilateral group (p=0.3). This difference narrowed when the individuals with BRCA1 or BRCA2 mutations were excluded (table 3).
BRCAPRO appears to overestimate the likelihood of a BRCA1 mutation, particularly in individuals with multiple primaries (p=0.0002 for bilateral breast cancer; p=0.08 for breast/ovary). For a breast–ovary double primary this is by a factor of 33% (49 expected, 36 found; table 1) and for bilateral breast cancer by close to 100% (table 2). Myriad tables are particularly poor at prediction in bilateral cases, significantly underestimating the number of mutations (p=0.0001). Table 4 shows the predictions for various likelihood levels for Manchester scoring, BRCAPRO and Myriad. The poor discrimination of the Myriad tables is demonstrated by the fact that 23/125 (18%) of samples with a Myriad score of <10% have mutations. The sensitivity of the models and scoring system are shown in table 5 at the 10% and 20% threshold.
Rates of BRCA1/2 mutations in bilateral and unilateral breast cancer cases with no personal or family history of ovarian cancer or mbc and in breast–ovary double primaries
Detection rates and sensitivity at different thresholds
There was a consistently higher detection rate for bilateral cases than unilateral cases for Manchester scores above the 10% 16 point threshold (table 6). At the 16–19 point and 20–24 point levels this was borderline significant (p<0.052 and p=0.048 respectively), and at the 30–39 points this was highly significant (p<0.0001). Detection rates at <16 points were virtually identical with only 1/62 bilateral cases having a mutation. By adjusting the bilateral scores upwards by adding the combined BRCA2 scores for the bilateral breast cancers a second time (equivalent to a 50% upward adjustment, table 6), but only for combined scores above the 15 point threshold, this equalised the detection rates between unilateral and bilateral cases at similar Manchester scores.
Rates of BRCA1/2 mutations in bilateral and unilateral breast cancer cases with no ovarian cancer in proband
Discussion
This report has documented the involvement of BRCA1, BRCA2 and CHEK2c.1100 delC mutations in breast cancer patients with double primaries of the breasts and/or ovaries. It has previously been reported that the mutation rates in BRCA1 and BRCA2 are higher with double primaries than in series of unilateral breast cancer. However, the literature is unclear as to whether identification of an individual with two separate primaries is more predictive of a BRCA1 or BRCA2 mutation than the same two primaries in a first degree pair. The results of our analysis do not suggest that any mutations are more likely with a breast and ovarian double primary than with the same two primaries in two close relatives. Although the overall rate of mutations is high at 49% (49/100), the rate in a case with breast–ovary double primary without a family history is below 20%. This is in contrast to the high rates suggested from previous reports.3 4 However, this rate is similar to that for isolated FDR pairs, one with breast cancer and one with ovarian cancer.
The Manchester scoring system11 does not alter if an individual has two primaries as each tumour receives a score dependent on age at diagnosis. The data from this study validates this approach for small aggregations, particularly under the 10% threshold. It is not clear from the publications whether other risk models predict an increase in the likelihood of a mutation based on the presence of a double primary of breast and ovary compared to the same cancers in an FDR pair.12 15 However, it is possible to assess this using a nuclear family of an unaffected mother aged 75 and two daughters aged 55 years. An assessment of the extra weight of the individual with two primaries as opposed to a sibling pair can be made. If a breast and ovarian cancer occurred only in one sister with no cancer in the other, the combined likelihood of a BRCA1/2 mutation was 32.3% with BRCAPRO and 31.8% with BOADICEA. This dropped to 16% and 15.1%, respectively, if the two cancers occurred separately in the two sisters. This apparent doubling of likelihood is contradicted by the almost identical rates of BRCA1/2 mutation in the similar nuclear families tested—3/22 (14%) for double primary and (17%) 17/99 for FDR pair. The only group in which the BRCA1/2 mutation rate increased was in the group of patients with both bilateral breast cancer and ovarian cancer without any other significant family history (ie, triple primary). However, this is based on small numbers of cases. BRCAPRO did overpredict the number of mutations among breast–ovary double primaries, but this was not significant (p=0.08) and the numbers are probably not powered to detect this.
In contrast to cases with double primary breast and ovary diagnoses, the detection rates in bilateral breast cancer appeared to be higher than expected based on the two tumours in FDR pairs. Based on the same range of Manchester score the detection rates were consistently higher for bilateral than unilateral cases. In the range of 20–24 points and 30–39 points this reached statistical significance (p=0.048 and p>0.0001, respectively). However, this difference was not apparent for lower Manchester scores. If a similar scenario to the nuclear example for breast and ovary is worked for bilateral breast cancer compared to unilateral breast cancer in siblings aged 55 years (one with diagnosis aged 45 years) the scores for BRCAPRO and BOADICEA are: 8.7% and 9.3% for a BRCA1/2 mutation in the double primary compared to 4.1% and 3.8%, respectively, for two siblings. Again this suggests at least a doubling of the likelihood. The Manchester score in this setting is 10 points (5 for each gene).11 There was not a difference in the detection rate at this Manchester score. There was only a significant difference in detection rates between bilateral and unilateral cases if the rest of the family history was significant. This may be explained by lower phenocopy rates in bilateral cases compared to unilateral cases.16 The rate of phenocopies in our series in non-index cases with breast cancer only is 4% (1/26) for bilateral patients and 17% (32/183) for unilateral cases, although this does not reach statistical significance (p=0.08). These data would suggest that bilateral breast cancer only requires an upward adjustment in mutation probability when there is a concurrent strong family history above the 10% threshold, and this adjustment is probably still too high in existing models. Our findings are in contrast to a number of previous reports that do not indicate that bilateral disease per se is a strong indicator of the presence of a BRCA1/2 mutation.6 17 This would appear to be true in isolated cases or those with minimal family history, but in families with higher Manchester scores, reflecting a strong family history of breast cancer, the rates of detection rise notably. This has been previously suggested by a Danish study of 119 patients with bilateral or multifocal disease aged <46 years.18 One of the previous studies suggested that early age at onset of first primary (<42 years) was a relative predictor of mutation status,17 but data from this study suggest that this is already accounted for in the Manchester score. However, division by this age group is likely to be artificial as it introduces a confounding bias as women diagnosed <45 years who have developed bilateral disease will have longer follow-up than unilateral cases.
A recent study has also questioned the significance of bilateral disease in BRCAPRO.7 Ready et al demonstrated that for bilateral cases with a carrier probability >31%, the proportion of positive tests was significantly lower than predicted by the BRCAPRO model (p<0.05). Taking the same cut-off in our study, 120/168 bilateral breast cancers with BRCAPRO scores above 31% were predicted to have a BRCA1/2 mutation whereas only 75 did (Fisher's exact p<0.0001). It may be that models are overestimating the effect of bilateral breast cancer even in families with a significant family history. The simple formula of doubling the chance of a mutation because presumably two breast cancers have occurred in only two breasts rather than two in four in a sibling pair may be incorrect. Non-genetic factors such as hormonal factors may increase the risk of bilateral disease in one sibling while being discordant in a sister. For example, one sibling may have experienced menarche at 9 years and not had her first full term pregnancy until 35 years, while the unaffected sister experienced menarche at 16 years and first pregnancy at 19 years.
A direct comparison of the discrimination between BRCAPRO and the Manchester score is not appropriate as a small part of the dataset was used to develop the Manchester score (48 breast–ovary double primaries and 52 bilateral breast cancers in the 472 developmental set). Nonetheless, it can be seen that the Manchester score continues to discriminate well at both the 10% and 20% threshold even in the specific sets of multiple primary tumours. While Myriad tables work reasonably well for breast–ovary double primaries they should not be used for bilateral breast cancer as the tables did not include this parameter. An intriguing component of the BRCAPRO overestimation can be seen from the fact that in 37 patients predicted to have >98% chance of a mutation, only 27 (73%, p=0.007) had one. Even allowing for a drop to a 90% mutation detection sensitivity this is still a large overestimate. Both the Manchester score and Myriad tables can identify a subset with at least an 80% chance of a mutation (table 4). The various scoring systems and models are being improved with pathology now incorporated into BRCAPRO and the Manchester score.19 Using the pathology adjusted score of 40+ the Manchester score identifies 35/39 (90%) of cases with ovarian cancer as having a BRCA1/2 mutation. There are possible further explanations for why BRCAPRO may overestimate in bilateral sample sets, and these include misdiagnosis of metastatic disease rather than a second primary, selection of less aggressive breast cancers, and lower population frequency of BRCA1/2 than in the BRCAPRO algorithm. As genetic testing cannot be offered in families where all the affected patients have died, this could create a selection bias against aggressive disease.
Our results for CHEK2c.1100delC are in contrast to a number of previous studies that suggest that this mutation is strongly associated with bilateral disease.8 10 20 21 The OR for second primary for CHEK2c.1100delC carriers was 6.438 and 6.510 in two large studies of bilateral disease and significantly higher than for unilateral cases. The high contralateral rate was also suggested by the Peto group to have implications for relatives of CHEK2c.1100delC carriers with bilateral disease, although the 12-fold RR was based on only eight breast cancers.20 It is not clear why this analysis did not detect any difference between unilateral and bilateral cases although the selection criteria for testing meant that nearly all unilateral cases had a family history of breast cancer. The 2% frequency for CHEK2 c.1100delC is at least four times our local population rate.22 The largest series by Fletcher et al 20098 was ascertained from multiple sources including a UK series selected on family history. CHEK2 1100 delC was found in 10/321 (3.1%) compared to 6/1027 (0.6%) controls (OR 5.47, 95% CI 1.78 to 18.43). The 3.1% rate is not statistically significantly greater than either our unilateral or bilateral rates. We feel it is still premature to use CHEK2 testing to guide patient management.
In summary, this study of over 2000 breast cancer cases tested for BRCA1/2 and CHEK2c.1100delC mutations has identified a number of differences with previous reports. The effects of a case with double primary diagnoses do not appear to be as great in mutation prediction as currently utilised in prediction models, although bilateral breast cancer is still a strong predictor in large familial aggregations. CHEK2 does not appear to be as strongly linked to bilateral breast cancer as previously suggested in individuals who have a strong family history.
Acknowledgments
DGE and FL are supported by the Manchester NIHR Biomedical Research centre. We would also like to thank Professor Nazneen Rahman and her team at the Institute of Cancer Research, Margaret Warren-Perry, Sheila Seal, Anthony Renwick, Clare Turnbull, Deborah Hughes and Sarah Hines for screening samples for BRCA1/2 and CHEK2 as part of the Familial Breast Cancer Study, which is funded by Cancer Research UK. We would also like to thank Linda Ashcroft Biostatistician at Christie Hospital for her input.
References
Footnotes
Competing interests None.
Ethics approval This study was conducted with the approval of the Central Manchester University Hospitals Foundation Trust.
Provenance and peer review Not commissioned; externally peer reviewed.