Tamoxifen and risk of contralateral breast cancer in BRCA1 and BRCA2 mutation carriers: a case-control study

Transcription

1 Tamoxifen and risk of contralateral breast cancer in BRCA1 and BRCA2 mutation carriers: a case-control study Steven A Narod, Jean-Sébastien Brunet, Parviz Ghadirian, Mark Robson, Ketil Heimdal, Susan L Neuhausen, Dominique Stoppa-Lyonnet, Caryn Lerman, Barbara Pasini, Patricia de los Rios, Barbara Weber, Henry Lynch, for the Hereditary Breast Cancer Clinical Study Group* Summary Background Women with a mutation in BRCA1 or BRCA2 have a high risk of developing breast cancer and of contralateral cancer after the initial diagnosis of breast cancer. Tamoxifen protects against contralateral breast cancer in the general population, but whether it protects against contralateral breast cancer in BRCA1 or BRCA2 mutation carriers is not known. Methods We compared 209 women with bilateral breast cancer and BRCA1 or BRCA2 mutation (bilateral-disease cases), with 384 women with unilateral disease and BRCA1 or BRCA2 mutation (controls) in a matched case-control study. Age and age at diagnosis of breast cancer (range years) were much the same in bilateral-disease cases and controls, and both groups had been followed up for the same time for a second primary breast cancer. History of tamoxifen use for first breast cancer was obtained by interview, or by self-administered questionnaire. Findings The multivariate odds ratio for contralateral breast cancer associated with tamoxifen use was 0 50 (95% CI ). Tamoxifen protected against contralateral breast cancer for carriers of BRCA1 mutations (odds ratio 0 38, 95% CI ) and for those with BRCA2 mutations (0 63, ). In women who used tamoxifen for 2 4 years, the risk of contralateral breast cancer was reduced by 75%. A reduction in risk of contralateral cancer was also seen with oophorectomy (0 42, ) and with chemotherapy (0 40, ). Interpretation Tamoxifen use reduces the risk of contralateral breast cancer in women with pathogenic mutations in the BRCA1 or BRCA2 gene. The protective effect of tamoxifen seems independent of that of oophorectomy. Lancet 2000; 356: *Members are listed on page 1880 Centre for Research on Women s Health, Women s College Hospital, University of Toronto, Toronto, Canada (S A Narod MD, P de los Rios MSc); ClinTrials BioResearch, Senneville, Canada (J-S Brunet MSc); Epidemiology Research Unit, Centre Hopital de Université de Montreal, Hotel-Dieu, Montreal, Canada (P Ghadirian PhD); Department of Human Genetics and Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY, USA (M Robson MD); Unit of Medical Genetics, Norwegian Radium Hospital, Oslo, Norway (K Heimdal MD); Department of Medical Informatics, University of Utah, Salt Lake City, UT, USA (S L Neuhausen PhD); Institut Curie, Paris, France (D Stoppa-Lyonnet MD); Lombardi Cancer Center, Georgetown University Medical Center, Washington DC, USA (C Lerman PhD); Istituto Nazionale Tumori, Milan, Italy (B Pasini MD); Departments of Medicine and Genetics, University of Pennsylvania, Philadelphia, PA, USA (B Weber MD); and Department of Preventive Medicine and Public Health, Creighton University School of Medicine, Omaha, NE, USA (Henry Lynch MD) Correspondence to: Dr Steven Narod, Centre for Research on Women s Health, Women s College Hospital, 790 Bay Street, Room 750a, Toronto, Ontario, Canada M5G 1N8 ( steven.narod@swchsc.on.ca) Introduction Many familial breast cancers are hereditary, and might be associated with germline mutations in the BRCA1 or BRCA2 gene. 1,2 Breast and ovarian cancers in families with several affected members are likely to be hereditary. 1,2 In such families in women who carry deleterious BRCA1 or BRCA2 mutations the lifetime risk of breast cancer is estimated to be as high as 80%, 2 but estimates based on unselected cases of breast cancer are lower than 80%. 3,4 In the 10 years after diagnosis of breast cancer in a BRCA1 or BRCA2 mutation carrier, the risk of contralateral breast cancer is about 35%. 5,6 Strategies for reducing the risk of primary and contralateral breast cancer include prophylactic mastectomy and chemoprevention. 7,8 Tamoxifen reduces the risk of primary invasive and premalignant breast cancer in women at high-risk 9 and of contralateral breast cancer in unselected women. 10 However, the use of tamoxifen has not yet been assessed in known carriers of BRCA1 or BRCA2 mutations. To investigate the potential benefit of tamoxifen in the reduction of risk of contralateral breast cancer among BRCA1 or BRCA2 mutation carriers, we studied women with bilateral hereditary breast cancer and age-matched controls with unilateral hereditary breast cancer. Methods Patients and controls Information about patients with hereditary breast cancer was submitted to the study centre by investigators at each of 34 contributing centres in eight countries. These centres were requested to complete forms for all known cases of female breast cancer carrying a verified BRCA1 or BRCA2 mutation. Most of these patients were identified through genetic counselling and risk-assessment programmes offered to women from high-risk families. The data centre received information for 1243 cases of invasive breast cancer in carriers of pathogenic BRCA1 or BRCA2 mutations. Of the 1243 cases, 296 (24%) had bilateral breast cancer. 83 (28%) cases with bilateral disease were excluded from our study: 28 (10%) who had their first cancer diagnosed before Jan 1, 1970 (ie, before tamoxifen was in use); 39 (13%) who had contralateral cancer within 1 year of the initial diagnosis; 14 (5%) who had ovarian cancer diagnosed before contralateral breast cancer; and two (0 6%) who had a contralateral mastectomy before the development of contralateral breast cancer. Synchronous bilateral cases were excluded because the patient would not be able to take tamoxifen before the onset of the contralateral cancer. A control was not identified for four cases with bilateral disease who were therefore excluded from the analyses. Thus, 209 (71%) eligible patients with bilateral breast cancer were included. Controls were selected from women with unilateral breast cancer in the registry database. They were born within 5 years of the birth date of the bilateral-disease case, and diagnosed with breast cancer at an age within 5 years of age of first diagnosis of breast cancer of the case. Cases and controls were carriers of mutations in the same gene (BRCA1 or BRCA2) and were matched for residence (ie, 1876 THE LANCET Vol 356 December 2, 2000

2 Cases with bilateral Controls disease (n=209) (n=384) Year of birth* (10 5) (10 4) Year of diagnosis* (6 4) (6 3) Year of interview* (1 3) (1 8) Age at 1st breast cancer (years)* 39 4 (8 9) 40 6 (8 8) Age at 1st breast cancer (years) <30 31 (15%) 40 (10%) (42%) 152 (40%) (32%) 142 (37%) (12%) 50 (13%) Age at 2nd breast cancer (years)* 45 5 (9 8).. Mutation BRCA1 168 (80%) 308 (80%) BRCA2 41 (20%) 76 (20%) Parity 0 36 (17%) 60 (16%) 1 29 (14%) 65 (17%) (70%) 259 (66%) Place of residence Canada 47 (23%) 81 (21%) USA 98 (47%) 194 (51%) Europe 64 (31%) 109 (25%) Race or ethnic group Ashkenazi Jewish 61 (29%) 128 (33%) French-Canadian 15 (7%) 28 (7%) Other white 123 (59%) 221 (57%) Non-white 10 (5%) 7 (2%) Smoking history Never 106 (56%) 198 (58%) Before 1st breast cancer 39 (21%) 76 (22%) After 1st breast cancer 44 (23%) 65 (19%) *Values are mean (SD) or number (%). Smoking information was missing for 20 bilateral-disease cases and 45 controls. Table 1: Characteristics of cases with bilateral disease and of controls Europe or North America) at time of diagnosis. Each control was followed up for a similar length of time as the matched case. Women were ineligible to serve as controls if they were diagnosed before 1970, if they had a contralateral mastectomy, or if they had a diagnosis of ovarian cancer before or during the follow-up. For each case with bilateral cancer, we attempted to identify two controls with unilateral disease, and were able to do so for 175 of the 213 cases of bilateral breast cancer: one control was found for an additional 34 bilateral-disease cases. Cases with bilateral disease and controls had not received tamoxifen before diagnosis of initial breast cancer. All the exposures to tamoxifen in cases and controls are defined for the period equivalent to that before development of the contralateral cancer of the matched case. Because of the requirement for genetic testing our patients represent prevalent cases; the median year of diagnosis of the first breast cancer was 1986 ( ). Our study was restricted to living patients because mutation analysis can be done only in living women, and because risk factor information was obtained by questionnaire. The questionnaires were completed between 1980 and 2000 (on average 11 8 years after initial diagnosis). Mutation analysis Mutations were assayed by several established detection techniques, and all mutations were confirmed by direct sequencing of DNA samples. For women in Ontario, DNA samples were screened by the protein truncation test for mutations in exon 11 of BRCA1 and exons 10 and 11 of BRCA2. 11 The Ashkenazi Jewish women were screened for three founder mutations, two in BRCA1 (185delAG and 5382insC) and one in BRCA2 (6174delT). 12 French Cases with bilateral Controls p disease (n=209) (n=384) Tamoxifen use* 22 (11%) 81 (21%) Tamoxifen duration (years)* (90%) 303 (79%) (6%) 50 (13%) (1%) 19 (5%) > 4 6 (3%) 12 (3%) Oophorectomy* Ever 13 (6%) 57 (15%) Before cancer 5 (2%) 19 (5%) After cancer 8 (4%) 38 (10%) Surgery* Lumpectomy 73 (36%) 169 (45%).. Mastectomy 131 (64%) 209 (55%) Radiotherapy* 101 (48%) 212 (55%) Chemotherapy* 88 (42%) 234 (61%) Information on type of surgery was missing for five cases with bilateral disease and six controls. All p values were calculated with a univariate conditional logistic regression. *Values are number (%). Table 2: Factors associated with treatment of first breast cancer Canadian women were screened for six founder mutations. 13 Other women in the study were screened for mutations by techniques done routinely in the participating laboratories. These techniques included direct sequencing of DNA, protein truncation test, heteroduplex analysis, single-strand conformation analysis, and allele-specific oligonucleotide hybridisation. In every woman, the actual sequence variant in BRCA1 or BRCA2 was established by direct sequencing of DNA. Study protocol Bilateral-disease cases and controls completed a questionnaire that asked about medical and surgical treatment of their initial breast cancer. In some centres the questionnaire was administered by telephone interview. The years of diagnosis of the initial and contralateral breast cancer were recorded. Women were asked if they had received tamoxifen for first breast cancer, and if so, the dose, dates, and duration of treatment: tamoxifen use was defined as that given as treatment for the initial breast cancer, and not for the recurrent cancer. Women were also asked to report on the type of surgical treatment they received (lumpectomy, unilateral mastectomy, or bilateral mastectomy). They were asked if they had undergone bilateral oophorectomy, before cancer diagnosis or thereafter, and if they had had chemotherapy or radiotherapy for the initial breast cancer. Whether patients smoked, their reproductive history, and ethnic origin were also recorded. Statistical analysis A matched case-control analysis was done. The frequency and mean duration of tamoxifen use were compared between bilateral-disease cases and controls. The duration of tamoxifen use was compared by Wilcoxon s two-sample test. The univariate odds ratios and p values for contralateral breast cancer associated with tamoxifen use were calculated with conditional logistic regression. Cases with bilateral disease and controls were also compared for parity, smoking history, oophorectomy, and other medical treatments with Wilcoxon s two-sample test for continuous variables and univariate conditional logistic regression for categorical variables. To estimate the odds ratio for contralateral breast cancer associated with tamoxifen use, after adjustment for the other covariates, including other treatments received, we used a multivariate conditional logistic regression model. THE LANCET Vol 356 December 2,

3 To control for possible confounding, ethnic group, smoking, and parity were also included in the multivariate analysis. To estimate the protective effect of tamoxifen separately for BRCA1 and BRCA2 carriers, odds ratio estimates were generated for these subgroups with the matched-pair subsets. All calculations were done with the SAS (version 8.0) statistical package. Results The characteristics of the 209 cases with bilateral disease and 384 unilateral disease controls were much the same (table 1). The average time from first cancer to diagnosis of contralateral cancer was 6 1 years for the cases with bilateral disease. The average period of follow-up was 9 7 years for controls, and every control was followed up for as long as the matched case. Table 2 shows tamoxifen use, which was higher in controls than in cases with bilateral disease. In women who had taken tamoxifen the likelihood of developing cancer in the contralateral breast was reduced (table 3). The results of the multivariate analysis, controlling for other treatments, ethnic group, parity, and smoking, were mostly similar to those of univariate analysis (table 3). Among tamoxifen users, the average duration of use for bilateraldisease cases was 3 27 (SD 3 01) years, and was 2 67 (2 14) years for controls (p=0 40). The average age at first use of tamoxifen was 43 7 (12 0) years for the cases with bilateral disease and 45 7 (9 0) years for the controls. There were 476 patients with BRCA1 mutations (168 cases with bilateral disease and 308 controls) and 117 with BRCA2 mutations (41 bilateral-disease cases and 76 controls). BRCA1 mutation carriers were diagnosed with their first cancer at a mean of 38 9 (8 3) years and BRCA2 mutation carriers at 45 2 (9 3) years. Tamoxifen use was reported for 64 (13%) BRCA1 mutation carriers and for 39 (33%) BRCA2 mutation carriers. The protective effect of tamoxifen was apparent in BRCA1 and BRCA2 mutation carriers; of the BRCA1 mutation bilateral-disease cases and matched controls, the univariate odds ratio was 0 38 (95% CI ) and of the BRCA2 mutation carriers the odds ratio was 0 63 ( ). The median year of diagnosis of first breast cancer for the 593 patients was 1986 (range ). Tamoxifen use was reported by 23 (4%) women diagnosed before 1980, by 91 (15%) diagnosed from 1980 to 1989 and by 158 (26%) diagnosed since The odds ratio for second primary cancer associated with tamoxifen use was 0 30 ( ) for women treated before 1980, 0 29 ( ) for those treated from 1980 to 1989, and 0 68 ( ) for those treated since Tamoxifen use was associated with odds ratios of 0 36 ( ) and 0 71 ( ) for women diagnosed in North America and Europe, respectively. However, the two odds ratios did not differ significantly (p value for interaction=0 21). The protective effect of tamoxifen was greater in North America than in Europe (data not shown). The European patients differed from American patients in that they had a higher proportion of BRCA1 mutations (56 [88%] vs 75 [77%]), a younger age at first cancer diagnosis (38 5 years [SD 7 4] vs 40 8 years [9 3]), and a more recent diagnosis of first primary breast cancer (March/April, 1987 vs May/June, 1985). The protective effect of tamoxifen was slightly greater for women who were diagnosed with the second breast cancer at or after age 50 years (0 39, ) than for those who were diagnosed with contralateral breast cancer before age 50 (0 47, ). The protective effect of tamoxifen increased with duration of tamoxifen use up to 4 years (table 3). We recorded a small increase in risk of contralateral breast cancer in the multivariate analysis for Univariate analysis Multivariate analysis odds ratio (95% CI) odds ratio (95% CI) Tamoxifen use Never Ever 0 45 ( ) 0 50 ( ) Tamoxifen duration (years) < ( ) 0 47 ( ) ( ) 0 25 ( ) > ( ) 1 53 ( ) Bilateral oophorectomy Yes 0 39 ( ) 0 42 ( ) Timing of bilateral oophorectomy Before cancer 0 42 ( ) 0 43 ( ) After cancer 0 37 ( ) 0 42 ( ) Chemotherapy Yes 0 42 ( ) 0 40 ( ) Radiotherapy Yes 0 76 ( ) 0 93 ( ) We adjusted for other treatments, ethnic group, parity, and smoking habits, in multivariate analysis. Overall multivariate p value for tamoxifen duration was Table 3: Association between tamoxifen, oophorectomy, chemotherapy, and radiotherapy and risk of contralateral breast cancer those who had used tamoxifen for more than 4 years, but there were only six cases with bilateral disease and 12 controls in this subgroup. Oestrogen-receptor status was available for only a few patients. In such patients with bilateral disease 16 of 56 (29%) who had first primary cancers and 22 of 74 (30%) who had contralateral breast cancers were oestrogenreceptor positive. Of the women who did not receive tamoxifen, 17 of 64 (27%) contralateral tumours were oestrogen-receptor positive, and of those who received tamoxifen, five of ten (50%) contralateral tumours were oestrogen-receptor positive (odds ratio 2 77 [95% CI ]; p=0 15). We also noted a difference in the frequency of oophorectomy between cases with bilateral disease and controls (table 2). Oophorectomies done before menopause might affect the subsequent breast cancer risk. Among 185 women diagnosed with first breast cancer before age 50 years and their matched controls, the odds ratio associated with oophorectomy was 0 31 ( ). The protective effect of oophorectomy was much less for those who had their first breast cancer at or after age 50 years (0 85, ). However, the interaction between oophorectomy and age of breast cancer diagnosis was not significant (p=0 23). The protective effect of tamoxifen was present in the subgroups of women with oophorectomy (0 36, ) and without oophorectomy (0 49, ). The odds ratio for tamoxifen use combined with oophorectomy, compared with no tamoxifen or oophorectomy was 0 16 ( ). We showed a protective effect in women who were initially treated with chemotherapy for the first breast cancer. These women had a 60% reduction in risk of contralateral cancer (table 3). The protective effect of chemotherapy did not seem to be due to its effect on ovarian function; it was present in women without ovaries (0 43, ) and in those with their ovaries intact (0 45, ). For tamoxifen use, oophorectomy, and chemotherapy the protective effect on the risk of contralateral breast cancer was analysed according to time elapsed since initial diagnosis (table 4). Chemotherapy greatly reduced the risk of diagnosis 1878 THE LANCET Vol 356 December 2, 2000

4 Years since diagnosis of primary breast cancer >10 (58 cases; (57 cases; (51 cases; (43 cases; 111 controls) 111 controls) 93 controls) 69 controls) Tamoxifen use Never Ever ( ) ( ) ( ) ( ) Oophorectomy Never Ever ( ) ( ) ( ) ( ) Chemotherapy Never Ever ( ) ( ) ( ) ( ) Table 4: Odds ratio (95% CI) for contralateral breast cancer, associated with treatment, by time since primary breast cancer of contralateral breast cancers within 2 years of treatment but the risk rose after 10 years. The main effect of chemotherapy therefore seemed to be the eradication of prevalent, subclinical contralateral breast cancers. The protective effect of tamoxifen did not persist longer than 10 years (table 4); by contrast, the protective effect of oophorectomy was longlasting. This difference is consistent with the notion that oophorectomy permanently reduces endogenous oestrogen exposure, whereas the effect of tamoxifen on reducing hormone exposure is temporary. However, the number of patients in each of these subcategories was fairly small, and the confidence limits of the estimates for the subgroups overlap. We examined whether tamoxifen use, oophorectomy, and chemotherapy were independent protective factors, using a conditional multivariate logistic regression analysis. All three treatments were independently protective (table 3), and the effects were present after adjustment for ethnic group, smoking, and parity. Discussion Our results suggest that the risk of contralateral breast cancer is reduced by 50% in carriers of BRCA1 and BRCA2 mutations when tamoxifen is used for the treatment of the initial breast cancer. Women and their physicians were not aware of the BRCA1 or BRCA2 carrier status at the time of initial diagnosis. Tamoxifen use was reported more often by BRCA2 mutation carriers than by BRCA1 mutation carriers. This difference is to be expected, in view of the older age of onset of breast cancer in the BRCA2 mutation carriers than in the BRCA1 mutation carriers. Furthermore, most breast tumours that arise in BRCA1 mutation carriers are oestrogen-receptor negative, 14,15 whereas BRCA2-mutation-associated tumours are typically oestrogen-receptor positive. 16 The effect of tamoxifen in the National Surgical Adjuvant Breast and Bowel (NSABP) P1 trial seemed to be limited to the prevention of oestrogen-receptor positive tumours, 9 and some workers 17 have predicted that anti-oestrogen therapy will therefore be ineffective as chemoprevention for women with BRCA1 mutations. We have shown that tamoxifen is effective in both BRCA1 and BRCA2 mutation carriers. Furthermore, in a subset of the data we noted an association between oestrogen-receptor status of the contralateral breast cancer and past tamoxifen use. A greater proportion of the contralateral breast cancers were oestrogen-receptor positive in women with a history of tamoxifen use than in those with no past use, but the samples sizes were small. The NSABP P1 trial did not provide data on the subset of BRCA1 and BRCA2 mutation carriers, and perhaps only a small number of participants in that study were mutation carriers. There are several reasons to believe that breast cancer risk in women with BRCA1 or BRCA2 mutations might be reduced by blocking the activity of endogenous oestrogens. The incidence of breast cancer in BRCA1 mutation carriers declines after age ,19 Premenopausal oophorectomy in BRCA1 mutation carriers substantially reduces the risk of subsequent breast cancer. 20 Circulating oestrogen concentrations increase during pregnancy; in BRCA1 and BRCA2 mutation carriers, pregnancy increases the risk of early-onset breast cancer (by contrast with non-carriers for whom the effect of pregnancy is protective). 21 Brunet and colleagues 22 reported that smoking protected against breast cancer in BRCA1 and BRCA2 mutation carriers. This effect is probably an antioestrogenic mechanism. One of the functions of the normal BRCA1 protein is proposed to be the reduction of the proliferative response of the breast epithelium during periods of oestrogen exposure. 23 Further, tamoxifen might reduce breast cancer risk through mechanisms other than receptor-mediated oestrogen blockade. This hypothesis is consistent with our finding that tamoxifen seems to be protective in women who have undergone an oophorectomy and the protective effects of oophorectomy and tamoxifen seem to be independent and additive. We believe that our study design is the best way to address this issue. We could not analyse this dataset by a historical cohort approach, because genetic testing was offered only to living women, and, in many instances, was offered a considerable time after diagnosis. Additionally, women with bilateral cancer were more likely to have been offered genetic testing than were unilateral cases. Therefore a cohort study would be biased towards the inclusion of both long-term survivors and bilateral cases. We have kept to a minimum the possible effect of survivor bias, because our design required the controls were followed up for at least as long as the matched case. If genetic testing becomes widespread our findings might be confirmed by a cohort of incident cases, but these data will not be available for many years. Because of the difficulty in establishing carrier status for deceased women, our case-control study was restricted to living patients with breast cancer. Generally, the women completed the questionnaires 10 years after their initial diagnosis of breast cancer, and because of this time difference our findings, which are based on breast cancer survivors, might not represent the population of patients in general. For example, if women who develop contralateral breast cancer after using tamoxifen have a poor survival relative to those with bilateral breast cancer who have not taken this drug, our estimate of the protective effect of tamoxifen would be too large. However, if tamoxifen use is associated with improved survival in this population (eg, if it is preferentially given to women with less aggressive tumours, or if the drug confers a survival advantage), we would have underestimated the true positive effect of tamoxifen. We believe that the observed reduction in contralateral breast cancer risk is attributable to tamoxifen use and not to systemic differences in the risk profiles between the cases with bilateral disease and controls. These two groups were similar in terms of year of birth and age of cancer onset. Women with primary breast cancer who were oestrogen-receptor positive were more likely to have been prescribed tamoxifen than those who were oestrogen-receptor negative. However, the prognostic features of the first cancer should not affect the risk of contralateral breast cancer. In a large meta-analysis, tamoxifen was equally effective in prevention of contralateral breast cancer in women with THE LANCET Vol 356 December 2,

5 oestrogen-receptor positive and oestrogen-receptor negative first primary tumours. 10 Whether a patient survives or not might depend on the characteristics of the first cancer, but in our study controls were followed up for at least as long as bilateral disease cases. Why was the protective effect of tamoxifen greater in North America than in Europe? A similar difference has been reported in the prevention trials. 9,24,25 There were fewer European women than North American women and the difference in the odds ratios did not differ significantly between the two groups. European patients had a higher proportion of BRCA1 mutations, a younger age at first diagnosis of cancer, and a more recent diagnosis of primary breast cancer than North American patients. However, these minor differences do not explain the recorded differences in protection associated with tamoxifen. Side-effects of tamoxifen include endometrial cancer. 9 This cancer is not a proven feature of BRCA1 or BRCA2 mutation. 26,27 However, in one report, risk of endometrial cancer was higher in first-degree relatives of Jewish women with ovarian cancer who carried a BRCA1 mutation than in relatives of women who did not. 28 Tamoxifen does not increase risk of ovarian cancer in BRCA1 or BRCA2 mutation carriers or the general population. Selective oestrogen-receptor modulators might be an effective alternative to tamoxifen in genetically predisposed women, and might have fewer side-effects. Raloxifene is thought to protect against breast cancer and is not associated with an increased risk of endometrial cancer. 29 However, studies of raloxifene have been done only in postmenopausal women. The large randomised STAR trial is designed to compare tamoxifen and raloxifene in high-risk postmenopausal women for whom mutation status is not known. Because most breast cancers in BRCA1 mutation carriers appear before age 50 years, it is difficult to extrapolate the results of the STAR trial to BRCA1 and BRCA2 mutation carriers, who begin surveillance as early as age 25 years. In our study population very few (12%) patients had their first breast cancer diagnosed before age 30 years whereas (88%) were diagnosed before age 50 years. We have shown a strong protective effect of chemotherapy for risk of contralateral breast cancer. Chemotherapy might be effective in eradicating small breast tumours before they become noticeable. Previous studies have also noted a reduction in risk of contralateral breast cancer with chemotherapy in the general population. 30 BRCA1 mutation-associated breast tumours are commonly high grade and p53 mutation positive, 31,32 and these characteristics might enhance chemosensitivity. A strong protective effect on contralateral breast cancer was also seen with oophorectomy, similar in size to that recorded with tamoxifen and chemotherapy. This finding is in keeping with a report showing that oophorectomy reduces the risk of primary breast cancer in BRCA1 mutation carriers. 20 Ovarian ablation also prevents contralateral breast cancer in the general population. 33 Our study does not allow us to recommend the optimum duration of tamoxifen for chemoprevention, because of its design and the small size of our subgroups. However, we did not find that tamoxifen use for a period longer than 4 years was associated with a further decrease in risk of contralateral breast cancer. Tamoxifen use for 3 years on average provided protection for up to 10 years, after which the risk ratio was unity. In a large meta-analysis of tamoxifen treatment trials, the protective effect on risk of contralateral breast cancer did not decline with longer duration of tamoxifen use, but few of the analysed studies used tamoxifen for longer than 5 years. 10 By contrast with tamoxifen, the risk reduction associated with oophorectomy persisted beyond 10 years. We showed the effects of tamoxifen and oophorectomy to be independent in reducing risk of contralateral breast cancer, and, ultimately, the greatest protection may be achieved through the combination of both therapies. We believe that tamoxifen can reasonably be offered to women with BRCA1 or BRCA2 mutations and breast cancer for the prevention of contralateral breast cancer. The effect of tamoxifen on reduction of cancer-associated mortality has not yet been established, and our data do not allow us to determine whether tamoxifen slows the growth of existing tumours or reduces the development of new ones. We believe that tamoxifen will also reduce the occurrence of primary cancers in BRCA1 and BRCA2 mutation carriers, and that tamoxifen chemoprevention and other options should be discussed with healthy women who carry BRCA1 or BRCA2 mutations. Risk factors for second cancers differ from those for the primary cancer, but it is notable that the size of the risk reduction for primary cancer seen in the NSABP P1 prevention trial 9 was similar to that predicted on the basis of the contralateral breast cancers in previous trials. 34 Hereditary Breast Cancer Clinical Study Group W Foulkes, P Tonin (McGill University), C Perret, A-M Mes-Masson, D Provencher (University of Montreal), B Rosen, E Warner, R Nedelcu, E Hoodfar, D Hannah, A Finch, H Jernström, J McLaughlin, K Metcalfe, A Liede (University of Toronto), K Panabaker, C Kim-Sing, D Horsman, C Trevisan, I Olivotto (British Columbia Cancer Agency), D Gilchrist (University of Alberta), P Ainsworth, G Sheridan (University of Western Ontario/LRCC), T Rebbeck, A Eisen (University of Pennsylvania), M Daly, A Godwin, J Wagner-Costalos (Fox Chase Cancer Center), C Lerman, B Peshkin (Lombardi Cancer Centre), O Olopade, S Cummings (University of Chicago), L Cannon-Albright, L Steele (University of Utah), J Garber (Dana Farber Cancer Center), H Lynch, J Lynch, C Snyder, C Durham (Creighton University), N Tung, S Come (Beth Israel Hospital Deaconness Hospital, Boston), B Karlan (Cedars-Sinai Medical Center), M Osborne, N Singer, G Rosenthal, F Gilbert (Strang Cancer Prevention Center), K Huelsman, H Saal (University Cincinnati College of Medicine), D Fishman (Northwestern University), K Offit, Deborah Levin, M D Anderson, G Mills (Memorial-Sloan Kettering Cancer Center), S Merajver (University Michigan Comprehensive Cancer Center), H Risch (Yale University), J Weitzel (City of Hope Hospital), N Lindor (Mayo Clinic), K Hughes (Lahey Clinic), W McKinnon, M Wood (Vermont Familial Cancer Program), G Lenoir, O Serova, S Mazoyer (International Agency for Research on Cancer, Lyon, France), G Evans, F Lalloo (Regional Genetics Service, St Mary s Hospital, Manchester, UK), N Haites (University of Aberdeen), H Olsson, O Johannson, A Borg (Lund University, Lund, Sweden), P Moller (Norwegian Radium Hospital, Oslo, Norway), P Radice, G Spatti, A Conti, V Pensotti, S Manoukian, M Pierotti (Istituto Tumori Nazionale, Milan, Italy), and J Lubinski (Pomeranian Medical University, Szczecin, Poland). Contributors Steven Narod was the principal investigator and study coordinator, and prepared the manuscript. Jean-Sebastien Bunet did the statistical analysis. Study activities were supervised by: Parvis Ghadirian at the University of Montreal hospitals; Mark Robson a Memorial Sloan-Kettering; Ketil Heimdal at the Norwegian Radium Hospital; Susan Neuhausen at the University of Utah; Dominique Stoppa-Lyonnet at the Institut Curie; Caryn Lerman at the Lombardi Cancer Center; Barbra Pasini at Istituto Nazionale Tumori; Barbara Weber at the University of Pennsulvania; and Henry Lynch at Creighton University. Patricia de los Rios organised data submission from the individual centers and ensured completeness and accuracy of the submitted data. Acknowledgments This study was supported by the US National Institutes of Health grants 1R01 CA63682, CA63678, CA57601, CA61231, and CA74415, the National Cancer Institute of Canada Breast Cancer Initiative, the Department of the Army grants DAMD17-94-J-4299, DAMD J- 4260, and DAMD J 4340, the Canadian Genetics Diseases Network, the Canadian Breast Cancer Foundation (Ontario Chapter), the Fonds de La Recherche en Sante du Quebec, the Ontario Cancer Genetics Network, the Women s Cancer Program of the Dana Farber Cancer Institute, the Norwegian Cancer Society, the Italian Association and Foundation for Cancer Research (Special Project: Hereditary Tumours), the Society of Memorial Sloan-Kettering Cancer Center, the Utah Cancer 1880 THE LANCET Vol 356 December 2, 2000

6 Registry, and the Utah Department of Health. MR is the recipient of an American Cancer Society Physicians Training Award PRTA-38. References 1 Narod SA, Ford D, Devilee P, et al. An evaluation of genetic heterogeneity in 145 breast-ovarian cancer families. Am J Hum Genet 1995; 56: Ford D, Easton DF, Stratton M, et al. Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. Am J Hum Genet 1998; 62: Struewing JP, Hartge P, Wacholder S, et al. The risk of cancer associated with specific mutations of BRCA1 and BRCA2 among Ashkenazi Jews. N Engl J Med 1997; 336: Thorlacius S, Struewing JP, Hartge P, et al. Population-based study of risk of breast cancer in carriers of BRCA2 mutation. Lancet 1998; 352: Verhoog LC, Brekelmans CTM, Seynaeve C, et al. Survival and tumour characteristics of breast-cancer patients with germline mutations of BRCA1. Lancet 1998; 251: Robson M, Gilewki T, Haas B, et al. BRCA-associated breast cancer in young women. J Clin Oncol 1998; 16: Hartmann LC, Schaid DJ, Woods JE, et al. Efficacy of bilateral prophylactic mastectomy in women with a family history of breast cancer. N Engl J Med 1998; 340: Burke W, Daly M, Garber J, et al. Recommendations for follow-up care of individuals with an inherited predisposition to cancer: BRCA1 and BRCA2. JAMA 1997; 277: Fisher B, Constantino JP, Wickerham DL, et al. Tamoxifen for prevention of breast cancer; report of the National Surgical Adjuvant Breast and Bowel Project PI study. J Natl Cancer Inst 1998; 90: Early Breast Cancer Trialists Collaborative Group. Tamoxifen for early breast cancer: an overview of the randomised trials. Lancet 1998; 351: Hogervorst FB, Cornelis RS, Bout M, et al. Rapid detection of BRCA1 mutations by the protein truncation test. Nat Genet 1995; 10: Tonin P, Weber B, Offit K, et al. Frequency of recurrent BRCA1 and BRCA2 mutations in Ashkenazi Jewish breast cancer families. Nat Med 1996; 11: Tonin P, Mes-Masson AM, Futreal A, et al. Founder BRCA1 and BRCA2 mutations in French-Canadian breast and ovarian cancer families. Am J Hum Genet 1998; 63: Karp SE, Tonin PN, Begin LR, et al. Influence of BRCA1 mutations on nuclear grade and estrogen receptor status in breast cancers in Ashkenazi Jewish women. Cancer 1997; 80: Johannsson OT, Idvall I, Anderson C, et al. Tumour biological features of BRCA1-induced breast and ovarian cancer. Eur J Cancer 1997; 33: Loman N, Johannsson O, Bendahl P-O, Borg A, Ferno M, Olsson AH. Steroid receptors in hereditary breast carcinomas associated with BRCA1 or BRCA2 mutations or unknown susceptibility genes. Cancer 1998; 83: Eisenger F, Jacquemier J, Nogues C, Birnbaum D, Sobol H. Steroid receptors in hereditary breast carcinomas associated with BRCA1 or BRCA2 mutations or unknown susceptibility genes. Cancer 1999; 85: Narod SA, Goldgar D, Cannon-Albright, et al. Risk modifiers in carriers of BRCA1 mutations. Int J Cancer 1995; 64: Easton DF, Ford D, Bishop DT, et al. Breast and ovarian cancer incidence in BRCA1 mutation carriers. Am J Hum Genet 1995; 56: Rebbeck TR, Levin AM, Eisen A, et al. Reduction in breast cancer risk after bilateral prophylactic oophorectomy in BRCA1 mutation carriers. J Natl Cancer Inst 1999; 91: Jernström H, Lerman C, Ghadirian P, et al. Pregnancy and risk of early onset breast cancer in carriers of BRCA1 and BRCA2. Lancet 1999; 354: Brunet JS, Ghadirian P, Rebbeck TR, et al. The effect of smoking on breast cancer incidence in BRCA1 and BRCA2 carriers. J Natl Cancer Inst 1998; 90: Fan S, Wang J, Yuan R, et al. BRCA1 inhibition of estrogen receptor signalling in transfected cells. Science 1999; 284: Veronesi U, Maisonneuve P, Costa A, et al. Prevention of breast cancer with tamoxifen: preliminary findings from the Italian randomized trial among hysterectomised women. Lancet 1998; 352: Powles T, Eeles R, Ashley S, et al. Interim analysis of the incidence of breast cancer in the Royal Marsden Hospital tamoxifen randomised chemoprevention trial. Lancet 1998; 352: Ford D, Easton DF, Bishop DT, Narod SA, Goldgar DE. Risks of cancer in BRCA1-mutation carriers. Lancet 1994; 343: Breast Cancer Linkage Consortium. Cancer risks in BRCA2 carriers. J Natl Cancer Inst 1999; 91: Moslehi R, Chu W, Karlan B, et al. BRCA1 and BRCA2 mutation analysis of 208 Ashkenazi Jewish women with ovarian cancer. Am J Hum Genet 2000; 66: Cummings SR, Eckert S, Krueger KA, et al. The effect of raloxifene on risk of breast cancer in postmenopausal women: results from the MORE randomized trial. JAMA 1999; 281: Chen Y, Thompson W, Semenciw R, Mao Y. Epidemiology of contralateral breast cancer. Cancer Epidemiol Biomarkers Prev 1999; 8: Breast Cancer Linkage Consortium. Pathology of familial breast cancer: differences between breast cancers in carriers of BRCA1 or BRCA2 mutations and sporadic cases. Lancet 1997; 349: Crook T, Crossland S, Crompton MR, Osin P, Gusterson BA. p53 mutations in BRCA1-associated familial breast cancer. Lancet 1997; 350: Meakin JW, Hayward JL, Panzarella T, et al. Ovarian irradiation and prednisone following surgery and radiotherapy for carcinoma of the breast. Breast Cancer Res Treat 1996; 37: O Reagan R, Jordan VC, Gradishar WJ. Tamoxifen and contralateral breast cancer. J Am Coll Surg 1999; 188: THE LANCET Vol 356 December 2,