Survival in prospectively ascertained familial breast cancer: Analysis of a series stratified by tumour characteristics, BRCA mutations and oophorectomy
Version of Record online: 21 AUG 2002
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 101, Issue 6, pages 555–559, 20 October 2002
How to Cite
Møller, P., Borg, Å., Evans, D. G., Haites, N., Reis, M. M., Vasen, H., Anderson, E., Steel, C. M., Apold, J., Goudie, D., Howell, A., Lalloo, F., Mæhle, L., Gregory, H. and Heimdal, K. (2002), Survival in prospectively ascertained familial breast cancer: Analysis of a series stratified by tumour characteristics, BRCA mutations and oophorectomy. Int. J. Cancer, 101: 555–559. doi: 10.1002/ijc.10641
- Issue online: 17 SEP 2002
- Version of Record online: 21 AUG 2002
- Manuscript Accepted: 21 JUN 2002
- Manuscript Revised: 11 JUN 2002
- Manuscript Received: 19 MAR 2002
- breast cancer;
Dedicated clinics have been established for the early diagnosis and treatment of women at risk for inherited breast cancer, but the effects of such interventions are currently unproven. This second report on prospectively diagnosed inherited breast cancer from the European collaborating centres supports the previous conclusions and adds information on genetic heterogeneity and the effect of oophorectomy. Of 249 patients, 20% had carcinoma in situ (CIS), 54% had infiltrating cancer without spread (CaN0) and 26% had cancer with spread (CaN+). Five-year survival was 100% for CIS, 94% for CaN0 and 72% for CaN+ (p = 0.007). Thirty-six patients had BRCA1 mutations, and 8 had BRCA2 mutations. Presence of BRCA1 mutation was associated with infiltrating cancer, high grade and lack of oestrogen receptor (p < 0.05 for all 3 characteristics). For BRCA1 mutation carriers, 5-year survival was 63% vs. 91% for noncarriers (p = 0.04). For CaN0 patients, mutation carriers had 75% 5-year disease-free survival vs. 96% for noncarriers (p = 0.01). Twenty-one of the mutation carriers had undergone prophylactic oophorectomy, prior to or within 6 months of diagnosis in 13 cases. All but 1 relapse occurred in the 15 who had kept their ovaries, (p < 0.01); no relapse occurred in those who had removed the ovaries within 6 months (p = 0.04) Contralateral cancer was more frequently observed in mutation noncarriers, but this finding did not reach statistical significance. Our findings support the concept that BRCA1 cancer is biologically different from other inherited breast cancers. While current screening protocols appear satisfactory for the majority of women at risk of familial breast cancer, this may not be the case for BRCA1 mutation carriers. The observed effect of oophorectomy was striking. © 2002 Wiley-Liss, Inc.
Classified by aetiology, breast cancer is not a single disease. A minor fraction of cases are dominantly inherited. Inherited breast cancer is genetically heterogenous and attributable to mutations in several genes. The majority of cases in northern Europe are caused by genetic factors, so far not identified.1, 2 The prognosis of inherited breast cancer is a subject of debate.3 Dedicated clinics have been established for the early diagnosis and treatment of high-risk groups,4 but the effects of such interventions are currently unproven. One methodologic problem is that it is both ethically and practically impossible to randomise high-risk groups to assess the effects of interventions.
Eleven collaborating European clinical genetic centres joined forces through a Biomed2 demonstration programme on the management of inherited breast cancer. We published a preliminary report on the efficacy of early diagnosis and treatment, judged by stage at diagnosis and survival.5 The activity continues as an open international research collaboration. The present report is a more comprehensive update on our series of prospectively identified breast cancers in high-risk groups, stratified by tumour characteristics, BRCA1/2 mutation carrier status and the effect of prophylactic oophorectomy.
MATERIAL AND METHODS
Healthy women judged to be at risk for inherited breast cancer, according to preset criteria, were enrolled in follow-up programmes. For details of the inclusion criteria and follow-up programmes, see previous reports.4, 5 In brief, risk estimations were based on family history but without genetic testing prior to inclusion. Documented family history verified lifetime risk for breast cancer of 20% or more, implying that most women had a number of affected relatives. Genetic counselling was performed at the various clinical genetics centres. Follow-up surveillance included mammographic and clinical examination at least every year, in some centres every second year after the menopause. Informed consent and blood samples for diagnostic mutation analyses were obtained according to national legislation. All information was kept in medical files or approved research registers. All data were anonymous before export to the combined data set. The present report describes those patients in whom breast cancer was diagnosed for the first time at some point after enrolment in a surveillance programme.
We studied 249 patients from 4 European countries: Norway (n = 87), Scotland (n = 79), England (n = 64) and Holland (n = 19). Information on pTNM tumour stage, histopathologic grade (1, 2 or 3) and presence/absence of oestrogen receptors in infiltrating tumours was recorded. Our population included all patients contracting breast cancer within the observation period, irrespective of diagnostic methods. Most were demonstrated on planned examinations, but some were interval cancers (see previous report for detailed discussion5). Tumours were staged as carcinoma in situ (CIS, ductal or lobular), infiltrating cancer without nodal spread (CaN0) or cancer with spread (CaN+). All of the above data were collected from the medical files and not re-evaluated. For patients with bilateral cancer, tumour characteristics refers to the more advanced cancer for synchronous cases and to the first cancer for metachronous cases. More detailed subclassification is, of course, possible and will be appropriate when the data set has grown. Similarly, the material was stratified by country, rather than by participating centre, to ensure sufficient numbers in each group.
Blood samples for mutation analyses were obtained at diagnosis. BRCA1 and BRCA2 mutations were sought in all exons and splicing sites, using a number of methods and including specific search for all known local mutations, as previously described.6, 7
Oophorectomy was recorded. Contralateral breast cancer, relapse and death were recorded as events. Any recurrence or cancer-related death was scored as relapse for CIS and CaN0. Death was scored as relapse for patients initially diagnosed with CaN+. Deaths not related to cancer were censored as unaffected. One non-breast cancer death was observed (oesophageal squamous cell carcinoma) and censored as unaffected. For calculation of contralateral cancer incidence, patients were censored at contralateral prophylactic mastectomy.
Associations were tested by χ2 or Fisher's exact test. Differences between groups were estimated by t-test. Multiple regression was performed and survival estimated with the Kaplan-Meier algorithm using the Systat software package (SPSS, Inc., Chicago, IL). Point estimates for 5-year survival functions are given in the text; the corresponding Taron-Ware p values refer to total distributions.
Mean age at diagnosis for the 249 patients was 49.0 years (range 28–77, SD = 9.5). At diagnosis, 50/249 (20%) had CIS, 134/249 (54%) had CaN0 and 65/249 (26%) had CaN+ (including 1 with distant metastases). Mean follow-up time was 3.1 years (SD = 2.6, range 0–16). Fifteen patients had died, and 5 diagnosed with CaN0 had relapsed (Table I). Twenty-two patients had contracted bilateral breast cancer.
|BRCA1 mutation+ (n = 36)||BRCA1/2 mutation− (n = 205)|
Thirty-six patients had truncating BRCA1 mutations, and 8 had BRCA2 mutations. Mutations were unevenly distributed according to country (Table II). The high prevalence of BRCA1 mutation carriers in Norway reflects the local founder mutations.7 There is no indication that cases attributable to founder mutations behave differently, in terms of the findings recorded in our study, compared to those having less frequent mutations. The number of BRCA2 mutations was considered insufficient for statistical analyses, and the 8 BRCA2 carriers were excluded from calculations on subgroups.
|BRCA1 mutation+||BRCA2 mutation+||BRCA1/2 mutation−|
Of the remaining 241 patients, 36 (15%) carried BRCA1 truncating mutations; but only 1 of 49 (2%) CIS patients carried a mutation (p < 0.01). Details are given in Table I. BRCA1 mutation carrier status was associated with high histopathologic grade and absence of oestrogen receptor in tumour tissue. For patients with infiltrating cancers, the mean age at diagnosis for BRCA1 mutation carriers was 45.2 years compared to 50.3 years for noncarriers (p < 0.01).
Overall 5-year survival was 89% (Fig. 1). Five-year relapse-free survival was 87% (Fig. 2). Five-year survival without contralateral cancer was 90%. One (2.8%) of the BRCA1 mutation carriers contracted bilateral cancer vs. 18 (8.8%) of the noncarriers (p = 0.18). There was no difference between cancers detected at first screen (prevalence round) and cancers detected at later follow-up.
There was no difference between countries for the above results, and all associations with mutation carrier status were as expected, except for the low incidence rate of bilateral cancer in BRCA1 mutation carriers. Kaplan-Meier analyses on subgroups therefore treated the data as 1 series without further examination for heterogeneity.
Stratified by stage at diagnosis, 5-year survival was 100% for CIS, 94% for CaN0 and 72% for CaN+ (p = 0.007) (Fig. 3). Five-year disease-free survival was 100% for patients with CIS vs. 92% for CaN0 patients (p = 0.06).
For CaN0 and CaN+ patients grouped together, 5-year survival was 91% for mutation noncarriers and 63% for mutation carriers (p = 0.04) (Fig. 4). Among CaN0 patients, 5-year disease-free survival was 96% for mutation noncarriers and 75% for carriers (p = 0.01) (Fig. 5).
For CaN0 patients, mutation status, grade and oestrogen receptor status were entered into a Cox regression model to assess associations with disease-free survival. Neither mutation status (p = 0.10), grade (p = 0.62) nor oestrogen receptor status (p = 0.20) was significantly associated with survival. Removing grade before repeating the calculations did not make either of the other 2 significant (p = 0.10 and p = 0.18, respectively).
In several of the collaborating centres, prophylactic oophorectomy, at around age 40, was advocated in breast–ovarian cancer kindreds, mainly to reduce ovarian cancer risk. Fifteen non-mutation carriers and 21 BRCA1 mutation carriers had removed their ovaries. Among the mutation carriers, 4 had undergone prophylactic oophorectomy more than 1 year before breast cancer was diagnosed, 3 within 1 year prior to diagnosis, 6 within 6 months after diagnosis and 8 more than 6 months after diagnosis. All oophorectomies had been undertaken before relapse. Fifteen mutation-positive patients had retained their ovaries. One of 21 (4.8%) oophorectomized patients vs. 7/15 (47%) who had kept their ovaries experienced relapse (p = 0.005). Five-year survival for oophorectomized BRCA1 mutation carriers was 67% vs. 44% for nonoophorectomized patients (p = 0.01) (Fig. 6). Five-year disease-free survival for oophorectomized CaN0 BRCA1 mutation carriers was 100% vs. 42% for patients who had kept their ovaries (p = 0.009) (Fig. 7).
The first relapse in any BRCA1 mutation carrier was recorded 1.8 years after the diagnosis of breast cancer. To eliminate the possibility that decision on oophorectomy might have been influenced by outcome of treatment, patients having undergone oophorectomy more than 6 months after diagnosis were excluded. Upon recalculation, the survival difference between the oophorectomized group and the others remained (p = 0.04), while the difference for relapse-free survival in CaN0 patients became marginal (p = 0.06).
Our results confirm that BRCA1 mutation carriers present with biologically different cancers compared to noncarriers. Carriers have high-grade and oestrogen receptor-negative tumours, which are almost invariably infiltrating.8 Both compared to the total remaining group and corrected for stage, survival was worse. All of these findings are in keeping with retrospective studies.9, 10 In addition, more than half of the BRCA1 mutation carriers had had oophorectomy, and all but 1 relapse in mutation carriers occurred in women who had not undergone oophorectomy. A beneficial effect of oophorectomy in breast cancer is not a novel finding. It is also in keeping with previous reports on the preventive effect of oophorectomy in BRCA1 mutation carriers11 and on the preventive effect against contralateral cancer of tamoxifen in BRCA1 mutation carriers.12 In addition, we clearly demonstrate a survival advantage conferred by oophorectomy on BRCA1 mutation carriers with breast cancer. Antioestrogens may prevent oestrogen receptor-positive breast cancer.13 Oophorectomy is often discussed in relation to the effect of oestrogen, but no empirical data on protective effect of oophorectomy related to receptor status are known to the authors.
Because of the low numbers, all figures for survival may be regarded with caution, especially long-time survival. Long-time survivors are, however, an interesting observation; and they are presented, though no more than 38 cases (15%) were observed over more than 5 years. These cases confirm that the long-term survivors in the ascertainment clusters identifying families (data not shown) were not selection artefacts, at least not all of them. This information is important for genetic counselling. It is also of interest that long-term survivors with inherited breast cancer do not die of other cancers.
In conflict with retrospective reports,14 we found a low annual incidence rate for contralateral breast cancer in BRCA1 mutation carriers. Since the findings reflect not only the biology of the cancers but also the effects of treatment, we may speculate that the BRCA1 cancers more often received adjuvant chemotherapy, which prevented second cancers. Although speculative, our findings may be interpreted as supporting the continuation of chemoprevention trials.15, 16
No relapse and only 1 contralateral cancer occurred in the 15 patients without proven BRCA1 mutations who had undergone oophorectomy, but the number in this group was insufficient for any statistical analysis. Similarly, because only 1 BRCA1 mutation carrier developed contralateral breast cancer, the effect of oophorectomy on contralateral cancer among mutation carriers could not be addressed.
While it is possible that some BRCA1 and BRCA2 mutations remain to be detected in our patient group, the majority of inherited breast cancers were probably caused by a gene(s) so far not localized.1, 2 We previously reported that the number of prospectively demonstrated cancers in high-risk clinics greatly exceeds the predicted rate for an age-matched, unselected population.17 The interpretation that non-BRCA1/2 mutation carriers had breast cancer attributable to genetic factors was further supported by the high incidence of bilateral cancer. Our findings indicate that, in contrast to BRCA1-associated cancers, other forms of inherited breast cancer can be diagnosed as CIS and, in general, have a favourable prognosis.8 Genetic testing for germline mutations may therefore be seen as a practical procedure to inform the choice of preventive strategies and treatment modalities. Although, in the present series, most of the germline mutations were detected after the diagnosis of breast cancer, advances in molecular technology mean that this is unlikely to be the normal situation in the future.
Oophorectomy has already been suggested as a management option in breast–ovarian kindreds and in BRCA1/2 mutation carriers.18 A major consideration was the risk of ovarian cancer in mutation carriers combined with the poor prognosis.19 Oophorectomy may be considered a modality to prevent11 and, now, to treat breast cancer in BRCA1 mutation carriers. As mentioned above, however, our findings should be confirmed in other series before firm conclusions are drawn.
Our study was initially undertaken to estimate the efficacy of surveillance in high-risk groups and compare it with the reported effect of prophylactic mastectomy.20, 21 The present findings extend and confirm the preliminary report from the same series.5 The results indicate that groups of patients may benefit differently from the surveillance programmes and treatment given so far.
The present results highlight the need to take account of genetic heterogeneity and the putative beneficial effects of oophorectomy in BRCA1 mutation carriers. As a substantial proportion (even the majority) of mutation carriers at some centres now choose oophorectomy, comparisons with historical data on patients treated differently (e.g., prophylactic mastectomy in patients with unknown mutation status and without oophorectomy) may provide only limited information to guide decisions on therapy for future patients.
To evaluate the influence of mutation status on outcome, the usual procedure of selecting controls matched for age and tumour characteristics may be inappropriate because BRCA1 tumours differ for most of the relevant characteristics (grade, stage, oestrogen receptor). When stratification is based on aetiology, breast cancer has a worse prognosis in BRCA1 mutation carriers.10 Matched for prognostic factors, differences between groups reportedly disappear.22 Regression analysis of the CaN0 patients in our series did not support the notion10 that mutation status has an impact on prognosis independent of grade and receptor status.
Among women enrolled in management programmes for familial breast cancer, the majority are well served by current screening protocols. However, a minority with BRCA1 germline mutations fare significantly worse, even though detection of breast cancer at an apparently early stage (CaN0) is achieved for two-thirds of them. For BRCA1 mutation carriers, therefore, it is essential to gather further data that will permit thorough evaluation of alternative management options, such as improved modalities for early diagnosis (magnetic resonance), chemoprevention or prophylactic surgery.
As mentioned above, all studies like ours face the methodologic problem that randomised trials are neither ethically nor practically possible. Some of the flaws of retrospective studies may have been overcome in our prospective series. However, discrepancies between our results and retrospective reports may also reflect time-related changes in treatment. Also, our selection model employing family history may not give results representative of inherited cancer selected differently. Our study is continuing and expanding. The results are recent observations without control groups, and for the present, conclusions should be interpreted with caution.