Background BRCA1 or BRCA2 mutations confer increased risks of breast and ovarian cancer, but risks have been found to vary across studies and populations.
Methods We ascertained pedigree data of 582 BRCA1 and 176 BRCA2 families and studied the variation in breast and ovarian cancer risks using a modified segregation analysis model.
Results The average cumulative breast cancer risk by age 70 years was estimated to be 45% (95% CI 36 to 52%) for BRCA1 and 27% (95% CI 14 to 38%) for BRCA2 mutation carriers. The corresponding cumulative risks for ovarian cancer were 31% (95% CI 17 to 43%) for BRCA1 and 6% (95% CI 2 to 11%) for BRCA2 mutation carriers. In BRCA1 families, breast cancer relative risk (RR) increased with more recent birth cohort (pheterogeneity = 0.0006) and stronger family histories of breast cancer (pheterogeneity<0.001). For BRCA1, our data suggest a significant association between the location of the mutation and the ratio of breast to ovarian cancer (p<0.001). By contrast, in BRCA2 families, no evidence was found for risk heterogeneity by birth cohort, family history or mutation location.
Conclusions BRCA1 mutation carriers conferred lower overall breast and ovarian cancer risks than reported so far, while the estimates of BRCA2 mutations were among the lowest. The low estimates for BRCA1 might be due to older birth cohorts, a moderate family history, or founder mutations located within specific regions of the gene. These results are important for a more accurate counselling of BRCA1/2 mutation carriers.
- Cancer: Breast
- Clinical Genetics
- Genetic Epidemiology
Statistics from Altmetric.com
Germline mutations in the breast cancer genes BRCA1 and BRCA2 confer increased lifetime risks of developing breast and ovarian cancer.1 ,2 Initial risk estimates were based on highly selected families ascertained through the Breast Cancer Linkage Consortium (BCLC) to identify disease loci. Risk of breast cancer at age 70 years for women carrying BRCA1 or BRCA2 mutations appeared to be as high as 85%, whereas ovarian cancer risk was 64% for BRCA1 and 27% for BRCA2.3–,5 Since these studies were based on multiple-case families, the reported risks may overestimate the risks in carriers with a more modest family history which are nowadays mostly seen for genetic testing. Indeed, subsequent population-based studies provided lower risk estimates using data from breast cancer cases less selected for strong family history.6–,14 In a meta-analysis using data from 22 population-based studies including 280 BRCA1 and 218 BRCA2 families, the mean cumulative risk of breast cancer by age 70 years was estimated to be 65% for BRCA1 and 45% for BRCA2. The corresponding risks of ovarian cancer were 39% for BRCA1 and 11% for BRCA2, respectively.15 This study and others reported that breast and ovarian cancer risks in BRCA1/2 families may vary according to the location of the mutation, as well as the origin of he mutation, and birth cohorts.14–19 Additionally, there was evidence that risks vary between families, suggesting that genetic and non-genetic factors influence breast cancer risk among BRCA1/2 mutation carriers15 ,20 ,21 consistent with recent findings that common breast and ovarian cancer susceptibility alleles modify cancer risks in mutation carriers.22–,25
The aim of this study was to investigate breast and ovarian cancer risks in the largest series of Dutch BRCA1/2 typed families taking into account the breast cancer incidence in the Dutch population. This is the highest in Western countries and has been steadily increasing over time.26 Families were ascertained according to clinical referral guidelines in The Netherlands. In this population, BRCA1/2 mutations were found in approximately 16% of the tested families.27 Moreover, strong founder effects for a substantial proportion of mutations in BRCA1/2 have been identified in The Netherlands.28–,30 We further investigated whether breast and ovarian cancer risks in these families differ according to birth cohort, degree of family history and mutation position.
Subjects and methods
Families and mutation testing
Families were recruited through departments of clinical genetics in eight (of the nine) academic centres in The Netherlands during the period 1995–2005. Mutation screening was performed if an index case or family fulfilled one of the following criteria: (1) families with three or more first-degree relatives with breast and/or ovarian cancer in two successive generations, (2) bilateral breast cancer, (3) two first-degree relatives diagnosed with breast cancer including at least one case diagnosed before the age of 50, (4) two affected first-degree relatives consisting of one case diagnosed with premenopausal breast cancer under the age of 50 years, and another with ovarian cancer, (5) one woman with ovarian cancer and breast cancer diagnosed before the age of 60 years, (6) breast cancer before the age of 36 years, (7) a male individual diagnosed with breast cancer, or (8) two first-degree relatives diagnosed with ovarian cancer regardless of age. During the 10-year-period of genetic testing, the proportion of families with a pathogenic BRCA1/2 mutation decreased from 30% to 16%. reflecting that the most severely affected families were referred in the initial phase of genetic testing. The proband was the first family member who was counselled at the family cancer clinic, while the index carrier was the first mutation carrier identified in the family, mostly the youngest breast cancer or ovarian cancer patient. However, other scenarios occurred as well, for example, if there was no breast or ovarian cancer case available for testing, presymptomatic mutation screening was offered to the proband. Once the index carrier was identified, mutation testing was offered to other family members. This resulted in the identification of carriers and non-carriers of the disease-causing pathogenic mutation in the family.
Each laboratory used a variety of techniques to screen the complete coding regions of BRCA1 and BRCA2, including a combination of PTT, DGGE, DHPLC, direct sequencing, mutation-specific PCR and MLPA (multiplex ligation-dependent probe amplification) to detect large genomic deletions and amplifications in the BRCA1 gene.27 ,31–,33 With these techniques, the probability of finding a disease-causing mutation was more than 90%. Families with double heterozygosity for BRCA1 or BRCA2, and families in which unclassified variants were detected, were excluded. All mutations were checked and standardised on nomenclature according to the Breast cancer Information Core (BIC). The study was approved by the medical ethical committee of each participating centre.
At all centres, a pedigree of each family was drawn and full pedigree information was electronically transferred to a local MsAccess database. For each family member, information was collected on dates of birth and dates of death when applicable, date of genetic counselling, dates of cancer diagnoses and types of cancer, confirmation by pathology report, dates and types of prophylactic surgery, genetic status (carrier, non-carrier, not tested), type of mutation and date of testing. After pooling the local databases into a nationwide database, overlap of family data within and between centres was examined using date of birth, type of mutation, pedigree structure and, if necessary, first four letters of the last name as known in the local registry. Among the 816 mutation-positive families, overlaps were identified between 58 families (7%), resulting in 758 unrelated unique families available for the risk analyses. All pedigrees were checked for inconsistencies in the pedigree structure using the PedCheck program.34 Missing dates of birth (54%) were imputed from the known birth dates of the sibs or were estimated from average age differences between generations within each family as was described before.35 Missing dates of diagnosis for deceased individuals with known cancer sites and date of death (53%) were imputed from sex-specific, site-specific and calendar-year-specific median survival rates in the general population derived from The Netherlands Cancer Registry and the Comprehensive Cancer Centre South. Missing dates at diagnosis for living family members with known cancer sites were imputed by using mean sex-specific, site-specific and calendar-year-specific ages at diagnosis derived from the Netherlands Cancer Registry, including 31% female breast and 27% ovarian cancer cases.
The average age-specific breast and ovarian cancer risks up to age 70 years for female family members were estimated simultaneously using modified segregation analysis implemented in the pedigree analysis software MENDEL as described elsewhere.4 ,5 ,15 ,36 Female family members (n=20 405) were considered to be at risk from birth until the first of the following events: first breast cancer diagnosis (n=2789); ovarian cancer diagnosis (n=700), other cancer diagnosis (n=1056), prophylactic mastectomy (n=120) or oophorectomy (n=190), or both interventions simultaneously (n=80), death (n=1970), last contact with the study centre, last DNA test of a family member and 70th birthday (n=13 500). Thus, only the first breast or ovarian cancer was considered as an event in the analysis and, therefore, the risk estimates should be interpreted as the risk of breast or ovarian cancer, whichever occurred first. In total, 51% of the breast cancer cases and 59% of the ovarian cancers of the analytic cohort were confirmed by pathology reports.
A single autosomal-dominant model and a mutation frequency of 0.001 for both genes were assumed. The incidences of breast and ovarian cancer at age λ(t) were assumed to follow a Cox proportional hazards model: λ(t)=λ0 (t) exp(G(t)). The function λ0(t) is the baseline age-specific incidence rates for non-carriers and exp(G(t)) the relative hazard at age t (HRs) for carriers compared to non-carriers. Non-carriers were assumed to develop breast and ovarian cancer according to the age-specific, and cohort-specific incidence rates in The Netherlands. The latter were derived from the Eindhoven Cancer Registry from the period 1978 till 1990 and Netherlands Cancer Registry from 1990 onwards.
The models were parameterised in terms of the age-specific log relative hazard for breast and ovarian cancer estimated using maximum likelihood. To adjust for ascertainment we followed the sequential ascertainment rules described by Cannings and Thompson.37 Because family ascertainment was through multiple affected individuals, we maximised the conditional likelihood of observing the phenotypes and genotypes for all individuals in each pedigree given all phenotypic information in the family and the genotypic information of the index carrier. This means that the index carrier was excluded in the analysis. Parameter variances were obtained by inverting the observed information matrix as was described before.15
To evaluate the influence of other factors and to estimate the relative risk (RR) for subgroups, the model was extended to include an additional subgroup parameter: λ(t)=λ0(t) exp((g(t)+ci), where exp(ci) is the RRs of breast and ovarian cancer for individuals belonging in category i compared to the reference category. A likelihood-ratio test was used to test for differences between the extended and the restricted models. We fitted models in which breast and ovarian cancer risks were allowed to vary with birth cohorts (including categories:<1920, 1920–1940,>1940, or born before or after 1940) and family history of breast cancer (<3 breast cancer cases, 3–4 breast cancer cases and 5 or more breast cancer cases). To enable to compare our results with other studies,13 ,38 ,39 the year 1940 was used for stratification. As the majority (83%) of the families had less than three ovarian cancer cases, stratified analysis on family history of ovarian cancer was not performed to avoid imprecise estimates. BRCA1 and BRCA2 mutations were subdivided into three regions, that is, the central ovarian cancer cluster region (OCCR) between nucleotide 2401 and 4190 for BRCA1, and between nucleotide 3059/4075 and 6503/6629 for BRCA2 as were previously defined by Gayther et al and Thompson et al16–,19 and the two surrounding regions 5′ end and 3′ end of the two genes (see figure 1A,B. Genotype-phenotype correlations were analysed by counting the number of breast and ovarian cancer cases within each gene region and by comparing the difference in the ratio of breast to ovarian cancer over the regions. We fitted separate models to estimate cancer risks for mutations located in the three regions and to evaluate mutation-specific risks for two Dutch BRCA1 founder mutations. A Dutch founder mutation was recognised when a common founder of the mutation could be identified, or when the mutation was only reported in The Netherlands. Risk heterogeneity according to mutation location was evaluated without ascertainment correction to maximise power. The distribution of breast and ovarian cancers by mutation location was evaluated using the Kruskal–Wallis test for heterogeneity and Pearson χ2 test. For all statistical tests, results were considered significant at the 0.05 threshold and all p values were two-sided. Statistical analyses were performed using STATA program V.10 (StataCorp, College Station, Texas, USA).
A total of 582 BRCA1 and 176 BRCA2 families were included in this study. The overall distribution of the detected mutations and number of families identified for each mutation are illustrated in figure1A,B (see also online supplementary table S1 for the BIC-HGVS-nomenclature). In total, 89 distinct mutations were distributed throughout each gene. There were no regions of mutation clustering in either genes. The known Dutch founder mutations40 comprised 48% of the BRCA1 mutations and 16% of the BRCA2 mutations. The vast majority of the mutations included frameshift, nonsense and splice-site mutations, large genomic rearrangements categorised as pathogenic by the Breast Cancer Information Core (BIC) database (http://research.nhgri.nih.gov/bic/). However, the Dutch distribution by mutation type for BRCA1 differed significantly from that reported in the BIC registry, demonstrating fewer frameshift and more IVS mutations (p < 0.001). Compared to the BIC data, a higher proportion of mutations in exon11 was observed for BRCA1, but not for BRCA2 (43% vs 36% for BRCA1, and 68% vs 63% for BRCA2).
The characteristics of the BRCA1 and BRCA2 families are summarised in table 1. In total, there were 2546 carriers and 2221 non-carriers of the pathogenic mutation found in the 758 BRCA1/2 families. The mean age at diagnosis of all female breast cancer cases in the BRCA1 families was 45 years and in proven BRCA1 mutation carriers 40 years, which was significantly lower than in the BRCA2 families and BRCA2 mutation carriers (48 and 44 years, respectively, both p<0.001). The mean age at ovarian cancer diagnosis in both BRCA1 families and proven BRCA1 mutation carriers was 52 years, which was significantly lower than in BRCA2 families and proven BRCA2 mutation carriers (54 years, p=0.047 and 55 years, p=0.045, respectively). Male breast cancer was found in both BRCA1 (mean age 57 years) and BRCA2 families (mean age 56 years), but was predominantly prevalent in BRCA2 families.
Age-specific risks of breast and ovarian cancer
The estimated age-specific HRs of breast and ovarian cancer for BRCA1 and BRCA2 mutation carriers, respectively, as compared with the general population are shown in table 2. The age-specific breast and ovarian cancer HRs were lower in BRCA2 mutation carriers than in BRCA1 mutation carriers. Among BRCA1 mutation carriers, the estimated HRs for breast cancer decreased with increasing age (p for trend=0.009), whereas the HRs for ovarian cancer increased with age, but this trend was not significant (p=0.189). No evidence was found for an increasing or decreasing trend with age in the breast or ovarian cancer HR estimates for BRCA2 mutation carriers. Among BRCA2 carriers age-specific breast or ovarian cancer risks remained increased up to 70 years, but risks of breast cancer seemed to be constant after age 40 years, which is consistent with other reports.
The average cumulative risks by age 70 years among BRCA1 mutation carriers were estimated at 45% (95% CI 36% to 52%) for breast cancer and 31% (95% CI 17% to 43%) for ovarian cancer (see figure 2A). The corresponding risks for BRCA2 mutation carriers were 27% (95% CI 14% to 38%) for breast cancer and 6% (95% CI 2% to 11%) for ovarian cancer (see figure 2B).
Table 3 summarises the results for models allowing for cohort-specific cancer risks in excess of the birth cohort effects in the general population at large. There was significant evidence that the breast and ovarian cancer risks for BRCA1 mutation carriers varied by birth cohort (likelihood ratio test for heterogeneity comparing the extended with the baseline model (pheterogeneity=0.0006). Compared to mutation carriers born before 1920, BRCA1 mutation carriers born after 1940 had a RR for breast cancer of 2.6 (95% CI 1.4 to 4.9). An increased risk for mutation carriers born after 1940 was also observed for ovarian cancer (RR=3.7, 95% CI 0.9 to 15.8), but this was not statistically significant. To estimate cohort-specific cumulative risks we used an extended model allowing for age-specific HRs for individuals born prior to and after 1940, separately. Figure 2C shows the cumulative breast cancer risk at age 65 years estimated to be 66% (95% CI 46% to 78%) for BRCA1 mutation carriers born after 1940 and 32% (95% CI 22% to 41%) for BRCA1 mutation carriers born before 1940 (see figure 2C). The corresponding estimated cumulative ovarian cancer risks for BRCA1 mutation carriers born after and before 1940 were 45% (95% CI 5% to 71%) and 29% (95% CI 13% to 42%), respectively. We did not find a varying breast and ovarian cancer risk by birth cohort among BRCA2 mutation carriers (pheterogeneity = 0.459), whereas the number of ovarian cancer cases among BRCA2 mutation carriers born after 1940 was too small for a reliable estimation.
Variations in risk by degree of family history were investigated by fitting a model with log HR parameters reflecting the number of known breast cancer cases in the family (<3 breast cancer cases, 3–4 breast cancer cases and 5 or more cases, see table 4). We found evidence for a variation in risk by family history among BRCA1, but not among BRCA2 families (pheterogeneity<0.001 for BRCA1, pheterogeneity=0.601 for BRCA2). In the combined analysis, BRCA1 mutation carriers from families with three or more cases had a statistically significant higher risk of breast cancer as compared with carriers from families with a moderate family history of less than three breast cancer cases (RR=2.1, 95% CI 1.1 to 4.1).
Separate analyses were performed to evaluate whether the risks of breast and ovarian cancer differed according to the location of the mutation within the two genes. For BRCA1, the breast to ovarian cancer ratio, defined as the number of breast cancer cases divided by the number of ovarian cancer cases, was significantly associated with mutation location (p<0.001, see table 5). The lowest breast to ovarian cancer ratio was observed for mutations within the OCCR, and the highest ratio at the 3′end of the gene (2.6 and 4.5, respectively). There was no evidence that the ratios of breast to ovarian cancer cases differed by mutation position among the BRCA2 families (p=0.665, data not shown). RR of ovarian cancer for BRCA1 mutations in the central region as compared to mutations in the two other regions was 1.8 (95% CI 0.8 to 1.7), the corresponding RR of breast cancer was 0.9 (95% CI 0.5 to 1.7) data not shown. RRs of ovarian cancer for 5′and 3′ BRCA1 mutations, as compared to mutations in the central region, were slightly decreased (RR=0.68 (95% CI 0.53 to 0.87) and RR=0.63 (95% CI 0.50 to 0.79), respectively), while the RRs of breast cancer were slightly increased (RR=1.10 (95% CI 0.94 to 1.28) and 1.26 (95% CI 1.09 to 1.45), respectively).
Mutation-specific cancer risks were estimated for three BRCA1 founder mutations including the 185delAG Ashkenazi Jewish population at the 5′end of the gene, and the Dutch founder mutations 2804delAA in the OCCR and the IVS21-36del510 mutation at the 3′ end of the gene. The cumulative risks of ovarian cancer by age 70 years associated with these mutations were 27%, 56% and 15%, and the corresponding risks for breast cancer were 53%, 34% and 58%, respectively.
To our knowledge, this study comprises the largest consecutive series of clinically ascertained BRCA1/2 mutation-positive families reported so far. The spectrum of mutations in these families is characterised by a high proportion of recurrent Dutch founder mutations. The average cumulative cancer risks both for breast and ovarian cancer by age 70 years in BRCA2 mutation carriers is substantially lower than previously reported. We confirm the lower overall risk estimates for BRCA1 mutation carriers compared to earlier data obtained from highly selected families, although the risks remained within the boundaries of data of the meta-analysis15 and other studies.41 ,42 Our results provide evidence that a more recent birth cohort and a strong breast cancer family history are both associated with higher risks of breast cancer in BRCA1 mutation carriers, but this was not observed in BRCA2 mutation carriers.
Numerous studies have estimated the cumulative cancer risks for BRCA1 and BRCA2 mutation carriers. However, estimating penetrance using retrospective study designs without well-defined family ascertainment criteria is susceptible to various sources of bias, such as ascertainment and testing bias. In this respect, it has become clear that one overall penetrance curve for all mutation carriers may not exist. Risks for a mutation carrier may vary by population and by exposure to other genetic and non-genetic factors. A number of attempts have been made to understand the extent and the nature of the risk variation in more detail.15 ,17 ,19 ,21 ,42 We estimated breast and ovarian cancer risks using the maximum likelihood method. The advantage of the likelihood-based approach is that it incorporates all available genotypic and phenotypic information in the families, taking into account the age-specific and period-specific incidence rates in the Dutch population, and adjusting for ascertainment bias due to sampling families with multiple affected family members. We adjusted for ascertainment bias by conditioning on the index case and all family phenotypes, but this process also reduced the amount of available information in the estimation process. As the mean age of breast cancer diagnosis of the index case was lower than that of other family members with breast or ovarian cancer, the ascertainment adjustment on this case may result in underestimation of the risk of breast and ovarian cancer. Therefore, we conducted a sensitivity analysis in which the index case was randomly selected among all mutation carriers. No difference in cumulative risk of breast cancer at age 70 years in BRCA1 families was found (46% at age 70 years). If we were to perform the analysis without any ascertainment correction, the estimated breast and ovarian cancer risks would be much higher, that is, 76% and 58%, respectively, instead of the ascertainment-adjusted risks of 45% and 31% at age 70 years (data not shown). This could explain the higher estimates found in previous studies that did not adjust for ascertainment and population-specific incidences to account for birth cohort and calendar-specific effects.13 ,38 ,39
Moreover, sensitivity analyses were performed to investigate the impact of the imputations on our results. After excluding relatives with an imputed age at censoring, cumulative risks of breast and ovarian cancer were 42% and 37% for BRCA1, and 27% and 7% for BRCA2 carriers. This demonstrates that our penetrance estimates were not significantly influenced by imputing age at diagnosis and imputing age at death or by potential under-reporting in the older generations. The imputed information proved not to weigh heavily in the analyses, because the family members for whom imputing was necessary were not closely related with the tested individuals.
Our method was also used by Milne et al42 to estimate population-specific penetrance of Spanish BRCA1/2 mutation carriers. Also, essentially, the same method was used as in the meta-analysis by Antoniou et al,15 with the exception that family ascertainment in the meta-analysis was through an otherwise unselected breast cancer case diagnosed below age 50 years and required conditioning only on this proband's genotype and phenotype rather than all family phenotypes as in the present and Milne's study. Additionally, only first-degree relatives were used in the meta-analysis. Although the estimated cumulative breast cancer risks for BRCA1 and BRCA2 mutation carriers are lower than previously reported,15 ,41–,43 our BRCA1 point estimate is within the 95% confidence limits of the meta-analysis15 and earlier studies,41 ,42 whereas the BRCA2 point estimates are outside these limits.
The main reason for our lower estimates as compared with the literature might be the inclusion of older birth cohorts and families with a relatively weak family history of breast cancer and potential founder effects. Our study confirms earlier findings that breast cancer risks are higher for BRCA1 mutation carriers from more recent birth cohorts.14 ,15 ,42 By using age-specific, calendar-specific, and cohort-specific population incidence rates as references, our study suggests that the risk of breast cancer in BRCA1 carriers is increased in recent birth cohorts as compared to older ones, and to a greater extent than that in the general population. Breast and ovarian cancer cumulative risks to age 65 years for BRCA1 mutation carriers born after 1940 were estimated to be 66% (95% CI 46% to 78%) and 45% (95% CI 5 to 71%) similar to the average breast and ovarian cancer risks of 65% and 39% reported in the meta-analysis.15 Thus, the difference in breast and ovarian cancer risks found in our study and the meta-analysis might be explained by a difference in birth cohorts. The index carrier selected for the meta-analysis were women alive and recently diagnosed with breast cancer before the age of 50 years. Therefore, it is likely that the meta-analysis represents a younger birth cohort than the present study. This may explain the higher risk estimates in the meta-analysis.
By contrast with other studies, we found no significant evidence for a birth cohort effect in excess to the general population among BRCA2 mutation carriers.15 ,42 One consideration in our analysis is that the families were selected in the period 1995–2004 and the majority of the family members included in our analysis born after 1940 would be too young to have experienced the risks for the entire age range (20–70 years). This might have a larger impact for BRCA2 than BRCA1 given the age-specific incidence patterns in BRCA2 mutation carriers (stable with increasing age) and the later age at onset of cancer in BRCA2 mutation carriers, larger numbers and a longer follow-up time would be needed to obtain more precise estimates for the younger birth cohorts.
The risk estimates for breast cancer in the younger BRCA1 cohort of our study are consistent with a recently published prospective cohort study among unaffected BRCA1 carriers by Mavaddat et al.44 However, the ovarian cancer risk was substantially lower in our study. This may be explained by differences in uptake of prophylactic oophorectomy and potential genetic risk modifiers that cluster in families. Because of limited power and follow-up time, it was not possible to adequately compare the results for BRCA2 carriers.
We found evidence among BRCA1 mutation carriers that breast cancer risk increases with increasing number of family members affected with breast cancer among BRCA1 mutation carriers. The variation of risk by degree of family history is consistent with recent findings that common breast cancer susceptibility alleles also modify cancer risks in carriers.22–,25 The stratification for family history of breast cancer was independent of the estimated HR, because the ascertainment correction includes the phenotype of the family and the number of relatives. The estimated breast cancer risk by age 70 years in BRCA1 mutation families with less than three breast cancer cases was 29%, which was substantially lower than the overall estimated risk of 45%. This supports the idea that BRCA1 mutation carriers from families with a weak or moderate family history of breast cancer should be counselled accordingly. However, for ovarian cancer, this effect remains unclear, possibly due to small numbers.
For BRCA1, the ratio of breast to ovarian cancers appeared to be significantly associated with location of the mutation. Our findings support previous reports suggesting that mutations located toward the 3′end of the BRCA1 gene are associated with lower ovarian cancer risks.16 ,19 ,45 However, in contrast with others, our results indicate a lower ovarian cancer risk for both 5′ and 3′ mutations compared to mutations in the central region, whereas breast cancer risk was increased for mutations at the 3′ end of the BRCA1 gene. Furthermore, there was some evidence for variability in risk with mutation position. The large genomic rearrangement of exon22 in BRCA1 (IVS21-36del510) and one of the common Dutch founder mutations27 conferred lower ovarian cancer and higher breast cancer risks. By contrast, higher ovarian and lower breast cancer RRs were observed for the Dutch founder mutation 2804delAA in exon11 of the BRCA1 gene. As 63% of the person-years comes from recurrent BRCA1 mutations (see figure 1A), mutation-specific risks may affect the overall breast and ovarian cancer risks. However, because of limited power it is too early to conclude that increased frequencies of founder mutations could explain the low breast and ovarian cancer risks in our study. The relation between mutation position and cancer risks in BRCA1 and BRCA2 families remains an important issue for future studies.
In conclusion, when estimating the risk of breast and ovarian cancer in BRCA1/2 mutation carriers, it is important to take the heterogeneity of risk into account. Our results suggest that BRCA2 mutations may confer lower average risks of breast and ovarian cancer than previously reported, while only for BRCA1 mutation carriers, cancer risks varied with birth cohort, family history and mutation position. These findings deserve incorporation in genetic counselling on cancer risks and, consequently, in risk management strategies of BRCA1/2 mutation carriers in The Netherlands. We provided further evidence that a single penetrance estimate for mutation carriers does not exist. Ongoing efforts to identify genetic and non-genetic modifying factors of the risks will enable to provide more personalised cancer risk predictions, which may lead to a more accurate counselling advice for BRCA1/2 mutation carriers in the future.
The authors gratefully acknowledge The Netherlands Cancer Registry and the Comprehensive Cancer Center South for their help in obtaining cancer incidence and survival rates in The Netherlands. For their help with the collection of the family data we would like to thank Aafke Buitelaar, Manita van Baalen, Bart Maertzdorf, Nellie Braakman, Monica Legdeur, Esther Janssen, Nandy Hofland, Mireille Wolfers, Sandra van der Belt-Dusebout, Ellen Crepin and Anja de Snoo. Antonis Antoniou is a UK Senior Cancer Research Fellow at the Centre for Cancer Genetic Epidemiology, Department of Public Health and Primary Care, University of Cambridge in the UK. The HEBON study was supported by the Dutch Cancer Society (NKI 1998–1854).
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
MAR and ACA contributed equally.
Collaborators The Netherlands Collaborative Group on Hereditary Breast Cancer (HEBON): consists of the following Collaborating Centers: Coordinating centre: Netherlands Cancer Institute, Amsterdam, NL: MA Rookus, RM Brohet, FBL Hogervorst, FE van Leeuwen, S Verhoef, MK Schmidt, JL de Lange; Erasmus Medical Center, Rotterdam, NL: JM Collée, AMW van den Ouweland, MJ Hooning, C Seynaeve, CHM van Deurzen; Leiden University Medical Center, NL: CJ van Asperen, JT Wijnen, RAEM Tollenaar, P Devilee, TCTEF van Cronenburg; Radboud University Nijmegen Medical Center, NL: CM Kets, AR Mensenkamp; University Medical Center Utrecht, NL: MGEM Ausems, RB van der Luijt; Amsterdam Medical Center, NL: CM Aalfs, TAM van Os; VU University Medical Center, Amsterdam, NL: JJP Gille, Q Waisfisz, HEJ Meijers-Heijboer; University Hospital Maastricht, NL: EB Gómez-Garcia, MJ Blok; University Medical Center Groningen, NL: JC Oosterwijk, AH van der Hout, MJ Mourits, GH de Bock. The Netherlands Foundation for the detection of hereditary tumours, Leiden, NL: HFA Vasen.
Contributors All authors of this manuscript contributed substantially to: conception and design, or analysis and interpretation of data, drafting the article or revising it critically for important intellectual content, and final approval of the version to be published.
Funding The HEBON study was supported by the Dutch Cancer Society (NKI 1998–1854).
Competing interests None.
Ethics approval The study was approved by the medical ethical committee of each participating centre.
Provenance and peer review Not commissioned; externally peer reviewed.
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.