- Email: [email protected]
Risk of ovarian cancer in breast-cancer patients with a family history of breast or ovarian cancer: a population-based cohort study
ARTICLES
Articles
Risk of ovarian cancer in breast-cancer patients with a family history of breast or ovarian cancer: a population-based cohort study Kjell Bergfeldt, Bosse Rydh, Fredrik Granath, Henrik Grönberg, Lukman Thalib, Hans-Olov Adami, Per Hall
Summary Background Patients with breast cancer who have mutations in the high penetrance genes BRCA1 and BRCA2, have an increased risk of ovarian cancer. Because these mutations are rare, easily obtained information such as age and family history of breast or ovarian cancer might be preferable for assessment of ovarian cancer risk in clinical practice. Methods We linked data from the Swedish Cancer Register to the Swedish Generation Register and generated a cohort of 30 552 breast-cancer patients born after 1931, with information on breast and ovarian cancer diagnosis from 146 117 first-degree relatives. Standardised incidence ratios (SIRs) with 95% CIs were calculated with nationwide rates of ovarian cancer, adjusted for age and calendar year. Findings During a mean follow-up of 6 years, 122 incident ovarian cancers were identified in the cohort, yielding an overall SIR of 2·0 (95% CI 1·6–2·4). The risk was higher in breast-cancer patients diagnosed before the age of 40 years, with a family history of breast cancer (5·6; 1·8–13·1) or ovarian cancer (17·0; 3·5–50·0). A consistently increased risk was noted in patients with a relative who was diagnosed before the age of 50 years, with either breast or ovarian cancer. Women with a family history of ovarian cancer have an almost 10% risk of developing ovarian cancer before the age of 70. Interpretation In young women with breast cancer, the risk of ovarian cancer is greatly raised when a family history of breast or ovarian cancer is present. Close medical surveillance, and perhaps even prophylactic oophorectomy, might be justified in high-risk groups. Lancet 2002; 360: 891–94. Published online August 28, 2002. http://image.thelancet.com/extras/01art11091web.pdf
Department of Medical Epidemiology, Karolinska Institutet, S-171 77 Stockholm, Sweden (K Bergfeldt MD, B Rydh BSc, F Granath PhD, H Grönberg MD, L Thalib PhD, Prof H O Adami MD, P Hall MD); Department of Oncology and Pathology, Karolinska Institutet, Stockholm (K Bergfeldt); Department of Oncology, Umeå University Hospital, Umeå, Sweden (H Grönberg); Faculty of Medicine, Kuwait University, Kuwait (L Thalib) Correspondence to: Dr Kjell Bergfeldt (e-mail: [email protected])
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
Introduction Because survival rates for women with breast cancer have improved, the risk of developing another primary malignant disease is an issue of increased importance. Results from epidemiological studies show that women with breast cancer have a small excess risk overall of primary ovarian cancer,1–4 and this risk seems to be highest among—or even confined to—women who are younger than 50 years at diagnosis of breast cancer.1,5 A particularly high rate of primary ovarian cancer is found in breast-cancer patients with mutations in the high penetrance genes BRCA1 or BRCA2, which are associated with hereditary breast cancer and ovarian cancer. Such women experience an almost 50% cumulative risk of developing ovarian cancer by the age of 70 years.6,7 To estimate the risk of subsequent ovarian cancer in a clinical setting, use of information such as age at onset of breast cancer, and number and age of affected relatives, would be more practical than mutation screening. Moreover, mutations in BRCA1 or BRCA2 are present in small numbers in all patients with breast cancer,8–12 and they seem to account for only a limited fraction of all breast cancers with a genetic component.13 However, the role of family history of breast cancer or ovarian cancer as a useful predictor of risk for a primary ovarian cancer has not yet been investigated in any population-based study. Therefore, our aim was to quantify the risk of ovarian cancer in a large population-based cohort with long-term follow-up and information on cancer in relatives obtained from a nationwide database in Sweden.
Methods Register data All Swedish citizens are assigned a ten-digit national registration number that is used in all official registers, including the nationwide databases used for this study. The national registration number provides an opportunity to collect data by linkage of several registers. This number is based on six digits that show year, month, and day of birth, together with four digits added to create a unique personal identifying number. The Swedish Cancer Register, established in 1958, contains information on incident cases of cancer in Sweden. Clinicians, pathologists, and cytologists are obliged to report cases of cancer, which makes the register close to 100% complete.14 The Register of Population and Population Changes contains the official Swedish census data from 1960 onwards. The names, addresses, and national registration numbers of all Swedish residents are included. The register has been expanded to include data on immigration and emigration. Data on causes of death have been gathered for more than 200 years in Sweden, initially in parish offices and, since the middle of the 20th century, in a centralised and computerised form as the Cause of Death Register. This register contains date of death, and the main cause of death. Present completeness of the register is estimated to be 99%.15 The Swedish Generation Register, established in the early 1990s, presently provides information on all Swedish
891
For personal use. Only reproduce with permission from The Lancet Publishing Group.
ARTICLES
inhabitants born after 1931, and who were alive in 1960. For people born after 1960, data are linked with those of close relatives who include parents, siblings, children, uncles, aunts, and cousins, although data quality decreases with the distance of the relationship. Adoptions and other non-biological relations are flagged. Individuals who immigrated to Sweden after 1960, but died or emigrated before 1992, are not included, nor are individuals who immigrated after 1992. Study population 35 532 women born after 1931 and diagnosed with breast cancer between 1958 and 1998 were identified in the Swedish Cancer Register. Information included date of breast-cancer diagnosis, and, date and site of any previous or subsequent malignant disease until the end of 1998. Additional follow-up data were obtained by linkage to the Cause of Death Register and the Register of Population Changes. We excluded five patients because of insufficient follow-up data, together with 3637 women registered as immigrants and 64 women who had ovarian cancer diagnosed before their breast-cancer diagnosis. After linkage to the generation register we excluded an additional 1338 women who did not have any linked relatives in the registry. Hence, 30 552 patients with breast cancer were included in the study cohort (table 1). Follow-up started at breast-cancer diagnosis and ended on Dec 31, 1998, the date of subsequent ovarian-cancer diagnosis, emigration, or death, whichever occurred first. As a consequence of our study design, women diagnosed with breast cancer when they were older than 60 years had a short follow-up because they were diagnosed during the 1990s. From the generation register, we identified 146 162 individuals as first-degree relatives of the cohort (25 869 mothers, 24 162 fathers, 19 109 sisters, 19 382 brothers, 28 005 daughters, and 29 635 sons). By relinkage to the Swedish Cancer Register, we identified incident breast and ovarian cancer with date of diagnosis among this group of relatives (table 1). Thus, we could construct family backgrounds of breast and ovarian cancer incidence for all members of the study cohort. Analyses We estimated the expected incidence of ovarian cancers using person-years at risk and age-specific and calendarspecific incidence rates from the Swedish Cancer Register. We calculated standardised incidence ratios (SIR) by dividing observed numbers of ovarian cancers by expected numbers. Risk of ovarian cancer in the cohort was calculated in relation to breast or ovarian cancers in relatives at the time of diagnosis of the index cases. We also calculated the cohort’s risk related to breast or ovarian cancer among relatives, irrespective of date of diagnosis. Index cases were grouped by age at diagnosis (<40 years, 40–49 years, and ⭓50 years). Ovarian cancer risk was also related to the age at which family members were diagnosed with breast or ovarian cancer, or both (below or above 50 years). We tested statistical significance using 95% CIs and assumed a Poisson distribution of the observed ovarian cancer cases.16 Number of women
Age at breast cancer diagnosis All ages 30 552 <40 years 5071 40–49 years 12 526 >49 years 12 955
Ovarian cancer cases
122 34 56 32
We did internal analyses within the cohort by applying Cox’s regression modelling to control for confounders and effect modification by age or calendar period of diagnosis. Crude analysis of the relative risk related to family history of breast or ovarian cancer was compared with models in which patients were grouped by age at breast-cancer diagnosis and calendar period (both variables were divided in 5-year intervals). We investigated the main effect of age at breast-cancer diagnosis by including this factor with family history, in a model stratified only for calendar period. We investigated potential effect modification by including interaction terms between family history and age at breast-cancer diagnosis, and time since breast cancer diagnosis (<5 years, 5–10 years, >10 years) in the statistical model. Likelihood ratio tests for homogeneity were applied to assess the significance levels for the interactions. Person-years at risk were used as an offset variable. Lastly, we estimated the absolute risk of ovarian cancer among breast-cancer patients, on the basis of the annual incidence proportion—ie, number of ovarian cancer cases divided by number of person-years. We then estimated the cumulative risk of women developing ovarian cancer within 30 years from breast cancer diagnosis, by multiplying annual incidence by 30. The result was expressed as absolute risk in percent. Role of the funding source The sponsors of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.
Results Mean age at breast-cancer diagnosis was 48 years (range 11–66 years). During a mean follow-up of 6 years, 122 cases of ovarian cancer were recorded in the study cohort of 30 552 women with breast cancer (table 1). Mean time to ovarian cancer diagnosis was 7 years (SD=5·9). Among 146 162 relatives, 3689 cases of breast or ovarian cancer were documented. Overall, we found a two-fold risk of ovarian cancer in the study cohort (SIR 2·0; 95% CI 1·6–2·4; table 2). Patients without any family history of breast or ovarian cancer had a 60% increased risk overall, but the excess risk seemed confined to premenopausal ages, and was three-fold in women younger than 40 years (3·3, 2·2–4·9). A family history of breast or ovarian cancer in a close relative was associated with a four-fold (4·3, 2·9–6·0) increased risk of ovarian cancer, and in women diagnosed before the age of 40 years, the risk was seven-fold (7·3, 3·1–14·3). In patients older than 40 years at diagnosis, the SIRs were smaller but remained raised (table 2). Next we analysed risk of ovarian cancer separately in breast-cancer patients who had a family history of breast cancer and in those with a family history of ovarian cancer. Although the small number of cases made some analyses difficult, the excess risk was two to four times higher at all ages among breast-cancer patients with a family history of ovarian cancer, compared with those who had a family history of breast cancer (table 2). In the
Number of relatives with breast or ovarian cancer
Number of relatives with breast cancer
Number of relatives with ovarian cancer
Number of relatives with both cancers
Total number of female relatives
3689 (12·1%) 604 (2·0%) 1607 (5·2%) 1478 (4·8%)
3111 (10·2%) 490 (1·6%) 1361 (4·5%) 1260 (4·1%)
501 (1·6%) 95 (0·3%) 212 (0·7%) 194 (0·6%)
77 (0·3%) 19 (0·1%) 34 (0·1%) 24 (0·1%)
72 983 11 622 30 666 30 695
Table 1: Characteristics of the cohort of Swedish breast-cancer patients born after 1931
892
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
For personal use. Only reproduce with permission from The Lancet Publishing Group.
ARTICLES
Age at breast-cancer diagnosis <40 years Number observed Overall analysis 34 No family history 26 Any family history 8 History of breast cancer 5 History of ovarian cancer 3 History of breast and ovarian cancer 0 Relative diagnosed <50 years of age 5 Family history at diagnosis Relative diagnosed before index case 6 Relative diagnosed before index case and <50 years 4
40–49 years
>49 years
SIR (95% CI)
Number SIR (95% CI) observed
Number observed
3·8 (2·6–5·3) 3·3 (2·2–4·9) 7·3 (3·1–14·3) 5·6 (1·8–13·1) 17·0 (1·3–50·0)
56 42 14 9 4
1·9 (1·4–2·4) 1·6 (1·2–2·2) 3·6 (2·0–6·1) 2·7 (1·3–5·2) 8·7 (2·4–22·5)
32 22 10 7 2
1·4 (0·9–1·9) 1·1 (0·7–1·6) 3·9 (1·9–7·2) 3·2 (1·3–6·6) 6·1 (0·7–22·1)
1
9·8 (0·3–54·6)
1
14·7 (4·8–34·3)
3
3·6 (0·8–10·6)
13·9 (5·1–30·1)
10
28·2 (7·7–72·2)
2
··
All ages SIR (95% CI)
Number observed
SIR (95% CI)
122 90 32 21 9
2·0 (1·6–2·4) 1·6 (1·3–2·0) 4·3 (2·9–6·0) 3·3 (2·0–5·1) 9·4 (4·3–17·8)
24·9 (0·6–139·0)
2
11·6 (1·4–42·0)
4
7·6 (2·1–19·4)
12
7·1 (3·7–12·4)
4·3 (2·0–8·1)
6
2·6 (0·7–6·6)
22
4·7 (3·0–7·1)
4·1 (0·5–14·8)
2
5·4 (0·7–19·4)
8
8·0 (3·4–15·7)
Table 2: Standardised incidence ratios (SIRs) of ovarian cancer in 30 552 women diagnosed with breast cancer
former category, the SIR for patients diagnosed before the age of 40 was 17·0, and seemed to decrease with an increase in age at breast-cancer diagnosis. Cohort patients with both breast and ovarian cancer among their close relatives had particularly high risks, but these estimates were based on a small number of cases (table 2). When we restricted our analysis to patients with a family history at the time of their breast-cancer diagnosis, overall excess risk was similar to that calculated for women with a family history at any time during the follow-up period (SIR 4·7 [3·0–7·1] vs 4·3 [2·9–6·0]), but for the patients younger than 40 years, the risk was almost doubled (SIR 13·9 [5·1–30·1] vs 7·3 [3·1–14·3]). Analysis within the cohort yielded a relative risk of 2·5 (95% CI 1·6–3·7) for patients with a family history of breast or ovarian cancer, with reference to patients without such a family history. This estimate was unchanged after we controlled for both calendar period and age at breast-cancer diagnosis. By contrast with the SIR analysis, the internal analysis yielded no effect of age at diagnosis of breast cancer, which verified the almost constant absolute risk for different age-groups shown in table 3. The relative risk (95% CI) with reference to patients younger than 40 years of age was 0·9 (0·6–1·5) for those between 40 and 49 years, and 1·0 (0·5–1·7) for those older than 50 years. Furthermore, we investigated whether the effect of family history differed with follow-up time, or with age at diagnosis. Neither of these factors showed any significant modification effect on the risk associated with family history (p=0·21 for follow-up and p=0·53 for age at diagnosis).
We found the risk of developing ovarian cancer within 30 years of breast cancer diagnosis was close to 10%, and was consistent over all age groups in women with a relative diagnosed with ovarian cancer. Patients with breast cancer among their relatives experienced a smaller, but similarly age-independent cumulative risk of about 3% (table 3).
Discussion We noted a two-fold increased risk of a primary ovarian cancer in women with a primary breast cancer. In women without a family history of breast or ovarian cancer, this high risk seemed confined to patients diagnosed at young age. Our most interesting finding, however, was the striking excess risk and substantial absolute risk in women with early-onset breast cancer, and a family history of breast cancer, or especially ovarian cancer. Also noteworthy is the finding that patients with a family history, especially if an ovarian cancer is present, have a high risk of ovarian cancer. This risk is present for postmenopausal women, and seems to be constant over time. Strengths of our study include its population-based design, complete follow-up of all patients in the cohort, and identification of exposure (family history) with optimum specificity without any differential misclassification. Some limitations of our study are the possible underidentification of family history of cancer, which was probably attributable to the design of the Swedish Generation Register. To be recorded in the register, a relative must have been alive in 1960. Hence,
Age at breast-cancer diagnosis <40 years AR% (95 % CI) No family history 1·9 (1·2–2·7) Any family history 3·7 (1·7–7·6) History of breast cancer 2·9 (0·9–6·8) History ovarian cancer 8·7 (1·8–25·3) History of breast ·· and ovarian cancer Relative diagnosed 6·9 (2·2–16·0) <50 years of age Family history at diagnosis Relative diagnosed 6·2 (2·3–13·5) before patient Relative diagnosed 10·3 (2·8–26·2) before patient and <50 years
40–49 years Observed/PYR AR% (95 % CI) 26/41 978 8/6205 5/5166 3/1039 0
>49 years Observed/PYR AR% (95 % CI)
1·5 (1·1–2·1) 42/81 621 3·4 (1·9–5·7) 14/12 401 2·5 (1·2–4·8) 9/10 600 8·0 (2·2–20·5) 4/1491 9·7 (0·3–53·9) 1/310
5/2189
3·3 (0·7–9·7)
3/2726
6/2913
3·9 (1·9–7·2)
10/7671
4/1168
3·5 (0·4–12·6)
2/1724
1·4 (0·9–2·1) 5·1 (2·7–10·4) 4·1 (1·7–8·5) 7·9 (0·9–28·5) 32·3 (0·9–179)
All ages Observed/PYR AR% (95 % CI)
Observed/PYR
22/47 268 10/5291 7/5069 2/760 1/93
90/170 867 32/24 724 21/20 834 9/3290 2/600
1·6 (1·3–1·9) 3·9 (2·7–5·5) 3·0 (1·9–4·6) 8·2 (3·8–15·6) 10·0 (1·2–36·1)
9·7 (2·6–24·6) 4/1243
5·8 (3·0–10·2) 12/6158
4·0 (1·5–8·8)
4·4 (2·7–6·6) 22/15067
6/4484
6·8 (0·8–24·5) 2/885
6·4 (2·7–12·5) 8/3777
AR=absolute risk, PYR=person-years at risk of ovarian cancer after breast cancer diagnosis.
Table 3: Estimates of absolute risk of developing ovarian cancer within 30 years of breast-cancer diagnosis
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
893
For personal use. Only reproduce with permission from The Lancet Publishing Group.
ARTICLES
relatives who died from breast cancer or ovarian cancer before this year would not be recorded, and neither would relatives diagnosed before 1958, when cancer registration started. These limitations indicate that the risk of ovarian cancer might be even higher than suggested. Additionally, although this is a large study, we had limited power for more detailed analyses because of the small number of ovarian cancer cases. Extended follow-up of our cohort will remedy this concern. Lastly, our risk estimates would be biased if the rate of oophorectomy in the cohort differed from that in the female population used for comparison. Because oophorectomy has been a part of treatment in young breast cancer patients,17,18 a somewhat higher prevalence in the cohort seems most likely, and would result in an underestimation of the risk. Further methodological restrictions include the ascertainment of time-at-risk for cohort members who develop a family history during follow-up. For such an age-dependent covariate, time-at-risk up to the date of breast or ovarian cancer diagnosis in a close relative should be recorded as unexposed, and the remaining follow-up as exposed. However, such a splitting of timeat-risk might not be important for genetic traits that obviously do not vary over time. As a corollary, our risk estimates did not change substantially when we used different analytic approaches. Our results expand the knowledge of an association between breast cancer and the risk of subsequent ovarian cancer. Diagnosis at a young age has been noted as a minor risk factor, and patients with breast cancer who have mutations in BRCA1 or BRCA2 have an extremely high risk.6,7 Although such mutations are rare (present in <5% of all breast cancer cases), hereditary factors might account for close to 25% of breast-cancer cases, with similar causation in ovarian cancer.8–13 Therefore, our study lends support to theories of a connection between as yet unknown genes and cancer susceptibility. A reasonable proxy for unknown genetic risk factors might be a family history of breast or ovarian cancer; the development of a Swedish nation-based family register that provides large enough cohorts allowed us to test this hypothesis. The clinical implications of our findings are not obvious, especially because there is no effective screening strategy for ovarian cancer.19–21 However, provided that our results are confirmed by other investigators, they seem to allow identification, based on easily obtained clinical information, of a small subgroup of women with breast cancer who are at particularly high risk of ovarian cancer. In this subgroup, counselling, and perhaps even prophylactic oophorectomy, might be considered.
Conflict of interest statement None declared.
Acknowledgments We thank the Karolinska Institute for providing the resources and facilities that made this report possible.
References 1
2
3
4 5
6 7
8
9
10
11
12
13
14 15 16
17 18 19
Contributors K Bergfeldt, P Hall, F Granath, and H-O Adami developed the idea to use register linkage to assess ovarian cancer risk as a function of family history. B Rydh and K Bergfeldt analysed data in collaboration, and F Granath and L Thalib did statistical analysis. H Grönberg advised on genetic matters. All authors revised the report and agreed to its final form.
894
20 21
Harvey EB, Brinton LA. Second cancer following cancer of the breast in Connecticut, 1935–82. Natl Cancer Inst Monogr 1985; 68: 99–112. Ewertz M, Mouridsen HT. Second cancer following cancer of the female breast in Denmark, 1943–80. Natl Cancer Inst Monogr 1985; 68: 325–29. Brenner H, Siegle S, Stegmaier C, Ziegler H. Second primary neoplasms following breast cancer in Saarland, Germany, 1968–1987. Eur J Cancer 1993; 10: 1410–14. Volk N, Pompe-Kirn V. Second primary cancers in breast cancer patients in Slovenia. Cancer Causes Control 1997; 8: 764–70. Suris-Swartz PJ, Schildkraut JM, Vine MF, Hertz-Picciotto I. Age at diagnosis and multiple primary cancers of the breast and ovary. Breast Cancer Res Treat 1996; 41: 21–29. Ford D, Easton DF, Bishop DT, Narod SA, Goldgar DE. Risks of cancer in BRCA1-mutation carriers. Lancet 1994; 343: 692–95. Frank TS, Manley SA, Olopade OI, et al. Sequence analysis of BRCA1 and BRCA2: correlation of mutations with family history and ovarian cancer risk. J Clin Oncol 1998; 16: 2417–25. Loman N, Johannsson O, Kristoffersson U, Olsson H, Borg A. Family history of breast and ovarian cancers and BRCA1 and BRCA2 mutations in a population-based series of early-onset breast cancer. J Natl Cancer Inst 2001; 93: 1215–23. Syrjakoski K, Vahteristo P, Eerola H, et al. Population-based study of BRCA1 and BRCA2 mutations in 1035 unselected Finnish breast cancer patients. J Natl Cancer Inst 2000; 92: 1529–31. Anglian Breast Cancer Study Group. Prevalence and penetrance of BRCA1 and BRCA2 mutations in a population-based series of breast cancer cases. Br J Cancer 2000; 83: 1301–08. Papelard H, de Bock GH, van Eijk R, et al. Prevalence of BRCA1 in a hospital-based population of Dutch breast cancer patients. Br J Cancer 2000; 83: 719–24. Newman B, Mu H, Butler LM, Millikan RC, Moorman PG, King MC. Frequency of breast cancer attributable to BRCA1 in a population-based series of American women. JAMA 1998; 279: 915–21. Lichtenstein P, Holm NV, Verkasalo PK, et al. Environmental and heritable factors in the causation of cancer—analyses of cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med 2000; 343: 78–85. Anon. Cancer incidence in Sweden 1998. Stockholm: Centre for epidemiology, 2000. Anon. Causes of death 1997. Stockholm: National Board of Health and Welfare, 2000. Breslow NE, Day NE. Statistical methods in cancer research. Volume II—The design and analysis of cohort studies. IARC Sci Publ 1987; 82: 1–406. Ovarian ablation for early breast cancer (Cochrane Review). Cochrane Database Syst Rev 2000; 3: CD000485. Dees EC, Davidson NE. Ovarian ablation as adjuvant therapy for breast cancer. Semin Oncol 2001; 28: 322–31. Hensley ML, Castiel M, Robson ME. Screening for ovarian cancer: what we know, what we need to know. Oncology (Huntingt) 2000; 14: 1601–07. Menon U, Jacobs IJ. Recent developments in ovarian cancer screening. Curr Opin Obstet Gynecol 2000; 12: 39–42. Dorum A, Kristensen GB, Abeler VM, Trope CG, Moller P. Early detection of familial ovarian cancer. Eur J Cancer 1996; 32A: 1645–51.
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
For personal use. Only reproduce with permission from The Lancet Publishing Group.
Articles
Risk of ovarian cancer in breast-cancer patients with a family history of breast or ovarian cancer: a population-based cohort study Kjell Bergfeldt, Bosse Rydh, Fredrik Granath, Henrik Grönberg, Lukman Thalib, Hans-Olov Adami, Per Hall
Summary Background Patients with breast cancer who have mutations in the high penetrance genes BRCA1 and BRCA2, have an increased risk of ovarian cancer. Because these mutations are rare, easily obtained information such as age and family history of breast or ovarian cancer might be preferable for assessment of ovarian cancer risk in clinical practice. Methods We linked data from the Swedish Cancer Register to the Swedish Generation Register and generated a cohort of 30 552 breast-cancer patients born after 1931, with information on breast and ovarian cancer diagnosis from 146 117 first-degree relatives. Standardised incidence ratios (SIRs) with 95% CIs were calculated with nationwide rates of ovarian cancer, adjusted for age and calendar year. Findings During a mean follow-up of 6 years, 122 incident ovarian cancers were identified in the cohort, yielding an overall SIR of 2·0 (95% CI 1·6–2·4). The risk was higher in breast-cancer patients diagnosed before the age of 40 years, with a family history of breast cancer (5·6; 1·8–13·1) or ovarian cancer (17·0; 3·5–50·0). A consistently increased risk was noted in patients with a relative who was diagnosed before the age of 50 years, with either breast or ovarian cancer. Women with a family history of ovarian cancer have an almost 10% risk of developing ovarian cancer before the age of 70. Interpretation In young women with breast cancer, the risk of ovarian cancer is greatly raised when a family history of breast or ovarian cancer is present. Close medical surveillance, and perhaps even prophylactic oophorectomy, might be justified in high-risk groups. Lancet 2002; 360: 891–94. Published online August 28, 2002. http://image.thelancet.com/extras/01art11091web.pdf
Department of Medical Epidemiology, Karolinska Institutet, S-171 77 Stockholm, Sweden (K Bergfeldt MD, B Rydh BSc, F Granath PhD, H Grönberg MD, L Thalib PhD, Prof H O Adami MD, P Hall MD); Department of Oncology and Pathology, Karolinska Institutet, Stockholm (K Bergfeldt); Department of Oncology, Umeå University Hospital, Umeå, Sweden (H Grönberg); Faculty of Medicine, Kuwait University, Kuwait (L Thalib) Correspondence to: Dr Kjell Bergfeldt (e-mail: [email protected])
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
Introduction Because survival rates for women with breast cancer have improved, the risk of developing another primary malignant disease is an issue of increased importance. Results from epidemiological studies show that women with breast cancer have a small excess risk overall of primary ovarian cancer,1–4 and this risk seems to be highest among—or even confined to—women who are younger than 50 years at diagnosis of breast cancer.1,5 A particularly high rate of primary ovarian cancer is found in breast-cancer patients with mutations in the high penetrance genes BRCA1 or BRCA2, which are associated with hereditary breast cancer and ovarian cancer. Such women experience an almost 50% cumulative risk of developing ovarian cancer by the age of 70 years.6,7 To estimate the risk of subsequent ovarian cancer in a clinical setting, use of information such as age at onset of breast cancer, and number and age of affected relatives, would be more practical than mutation screening. Moreover, mutations in BRCA1 or BRCA2 are present in small numbers in all patients with breast cancer,8–12 and they seem to account for only a limited fraction of all breast cancers with a genetic component.13 However, the role of family history of breast cancer or ovarian cancer as a useful predictor of risk for a primary ovarian cancer has not yet been investigated in any population-based study. Therefore, our aim was to quantify the risk of ovarian cancer in a large population-based cohort with long-term follow-up and information on cancer in relatives obtained from a nationwide database in Sweden.
Methods Register data All Swedish citizens are assigned a ten-digit national registration number that is used in all official registers, including the nationwide databases used for this study. The national registration number provides an opportunity to collect data by linkage of several registers. This number is based on six digits that show year, month, and day of birth, together with four digits added to create a unique personal identifying number. The Swedish Cancer Register, established in 1958, contains information on incident cases of cancer in Sweden. Clinicians, pathologists, and cytologists are obliged to report cases of cancer, which makes the register close to 100% complete.14 The Register of Population and Population Changes contains the official Swedish census data from 1960 onwards. The names, addresses, and national registration numbers of all Swedish residents are included. The register has been expanded to include data on immigration and emigration. Data on causes of death have been gathered for more than 200 years in Sweden, initially in parish offices and, since the middle of the 20th century, in a centralised and computerised form as the Cause of Death Register. This register contains date of death, and the main cause of death. Present completeness of the register is estimated to be 99%.15 The Swedish Generation Register, established in the early 1990s, presently provides information on all Swedish
891
For personal use. Only reproduce with permission from The Lancet Publishing Group.
ARTICLES
inhabitants born after 1931, and who were alive in 1960. For people born after 1960, data are linked with those of close relatives who include parents, siblings, children, uncles, aunts, and cousins, although data quality decreases with the distance of the relationship. Adoptions and other non-biological relations are flagged. Individuals who immigrated to Sweden after 1960, but died or emigrated before 1992, are not included, nor are individuals who immigrated after 1992. Study population 35 532 women born after 1931 and diagnosed with breast cancer between 1958 and 1998 were identified in the Swedish Cancer Register. Information included date of breast-cancer diagnosis, and, date and site of any previous or subsequent malignant disease until the end of 1998. Additional follow-up data were obtained by linkage to the Cause of Death Register and the Register of Population Changes. We excluded five patients because of insufficient follow-up data, together with 3637 women registered as immigrants and 64 women who had ovarian cancer diagnosed before their breast-cancer diagnosis. After linkage to the generation register we excluded an additional 1338 women who did not have any linked relatives in the registry. Hence, 30 552 patients with breast cancer were included in the study cohort (table 1). Follow-up started at breast-cancer diagnosis and ended on Dec 31, 1998, the date of subsequent ovarian-cancer diagnosis, emigration, or death, whichever occurred first. As a consequence of our study design, women diagnosed with breast cancer when they were older than 60 years had a short follow-up because they were diagnosed during the 1990s. From the generation register, we identified 146 162 individuals as first-degree relatives of the cohort (25 869 mothers, 24 162 fathers, 19 109 sisters, 19 382 brothers, 28 005 daughters, and 29 635 sons). By relinkage to the Swedish Cancer Register, we identified incident breast and ovarian cancer with date of diagnosis among this group of relatives (table 1). Thus, we could construct family backgrounds of breast and ovarian cancer incidence for all members of the study cohort. Analyses We estimated the expected incidence of ovarian cancers using person-years at risk and age-specific and calendarspecific incidence rates from the Swedish Cancer Register. We calculated standardised incidence ratios (SIR) by dividing observed numbers of ovarian cancers by expected numbers. Risk of ovarian cancer in the cohort was calculated in relation to breast or ovarian cancers in relatives at the time of diagnosis of the index cases. We also calculated the cohort’s risk related to breast or ovarian cancer among relatives, irrespective of date of diagnosis. Index cases were grouped by age at diagnosis (<40 years, 40–49 years, and ⭓50 years). Ovarian cancer risk was also related to the age at which family members were diagnosed with breast or ovarian cancer, or both (below or above 50 years). We tested statistical significance using 95% CIs and assumed a Poisson distribution of the observed ovarian cancer cases.16 Number of women
Age at breast cancer diagnosis All ages 30 552 <40 years 5071 40–49 years 12 526 >49 years 12 955
Ovarian cancer cases
122 34 56 32
We did internal analyses within the cohort by applying Cox’s regression modelling to control for confounders and effect modification by age or calendar period of diagnosis. Crude analysis of the relative risk related to family history of breast or ovarian cancer was compared with models in which patients were grouped by age at breast-cancer diagnosis and calendar period (both variables were divided in 5-year intervals). We investigated the main effect of age at breast-cancer diagnosis by including this factor with family history, in a model stratified only for calendar period. We investigated potential effect modification by including interaction terms between family history and age at breast-cancer diagnosis, and time since breast cancer diagnosis (<5 years, 5–10 years, >10 years) in the statistical model. Likelihood ratio tests for homogeneity were applied to assess the significance levels for the interactions. Person-years at risk were used as an offset variable. Lastly, we estimated the absolute risk of ovarian cancer among breast-cancer patients, on the basis of the annual incidence proportion—ie, number of ovarian cancer cases divided by number of person-years. We then estimated the cumulative risk of women developing ovarian cancer within 30 years from breast cancer diagnosis, by multiplying annual incidence by 30. The result was expressed as absolute risk in percent. Role of the funding source The sponsors of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.
Results Mean age at breast-cancer diagnosis was 48 years (range 11–66 years). During a mean follow-up of 6 years, 122 cases of ovarian cancer were recorded in the study cohort of 30 552 women with breast cancer (table 1). Mean time to ovarian cancer diagnosis was 7 years (SD=5·9). Among 146 162 relatives, 3689 cases of breast or ovarian cancer were documented. Overall, we found a two-fold risk of ovarian cancer in the study cohort (SIR 2·0; 95% CI 1·6–2·4; table 2). Patients without any family history of breast or ovarian cancer had a 60% increased risk overall, but the excess risk seemed confined to premenopausal ages, and was three-fold in women younger than 40 years (3·3, 2·2–4·9). A family history of breast or ovarian cancer in a close relative was associated with a four-fold (4·3, 2·9–6·0) increased risk of ovarian cancer, and in women diagnosed before the age of 40 years, the risk was seven-fold (7·3, 3·1–14·3). In patients older than 40 years at diagnosis, the SIRs were smaller but remained raised (table 2). Next we analysed risk of ovarian cancer separately in breast-cancer patients who had a family history of breast cancer and in those with a family history of ovarian cancer. Although the small number of cases made some analyses difficult, the excess risk was two to four times higher at all ages among breast-cancer patients with a family history of ovarian cancer, compared with those who had a family history of breast cancer (table 2). In the
Number of relatives with breast or ovarian cancer
Number of relatives with breast cancer
Number of relatives with ovarian cancer
Number of relatives with both cancers
Total number of female relatives
3689 (12·1%) 604 (2·0%) 1607 (5·2%) 1478 (4·8%)
3111 (10·2%) 490 (1·6%) 1361 (4·5%) 1260 (4·1%)
501 (1·6%) 95 (0·3%) 212 (0·7%) 194 (0·6%)
77 (0·3%) 19 (0·1%) 34 (0·1%) 24 (0·1%)
72 983 11 622 30 666 30 695
Table 1: Characteristics of the cohort of Swedish breast-cancer patients born after 1931
892
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
For personal use. Only reproduce with permission from The Lancet Publishing Group.
ARTICLES
Age at breast-cancer diagnosis <40 years Number observed Overall analysis 34 No family history 26 Any family history 8 History of breast cancer 5 History of ovarian cancer 3 History of breast and ovarian cancer 0 Relative diagnosed <50 years of age 5 Family history at diagnosis Relative diagnosed before index case 6 Relative diagnosed before index case and <50 years 4
40–49 years
>49 years
SIR (95% CI)
Number SIR (95% CI) observed
Number observed
3·8 (2·6–5·3) 3·3 (2·2–4·9) 7·3 (3·1–14·3) 5·6 (1·8–13·1) 17·0 (1·3–50·0)
56 42 14 9 4
1·9 (1·4–2·4) 1·6 (1·2–2·2) 3·6 (2·0–6·1) 2·7 (1·3–5·2) 8·7 (2·4–22·5)
32 22 10 7 2
1·4 (0·9–1·9) 1·1 (0·7–1·6) 3·9 (1·9–7·2) 3·2 (1·3–6·6) 6·1 (0·7–22·1)
1
9·8 (0·3–54·6)
1
14·7 (4·8–34·3)
3
3·6 (0·8–10·6)
13·9 (5·1–30·1)
10
28·2 (7·7–72·2)
2
··
All ages SIR (95% CI)
Number observed
SIR (95% CI)
122 90 32 21 9
2·0 (1·6–2·4) 1·6 (1·3–2·0) 4·3 (2·9–6·0) 3·3 (2·0–5·1) 9·4 (4·3–17·8)
24·9 (0·6–139·0)
2
11·6 (1·4–42·0)
4
7·6 (2·1–19·4)
12
7·1 (3·7–12·4)
4·3 (2·0–8·1)
6
2·6 (0·7–6·6)
22
4·7 (3·0–7·1)
4·1 (0·5–14·8)
2
5·4 (0·7–19·4)
8
8·0 (3·4–15·7)
Table 2: Standardised incidence ratios (SIRs) of ovarian cancer in 30 552 women diagnosed with breast cancer
former category, the SIR for patients diagnosed before the age of 40 was 17·0, and seemed to decrease with an increase in age at breast-cancer diagnosis. Cohort patients with both breast and ovarian cancer among their close relatives had particularly high risks, but these estimates were based on a small number of cases (table 2). When we restricted our analysis to patients with a family history at the time of their breast-cancer diagnosis, overall excess risk was similar to that calculated for women with a family history at any time during the follow-up period (SIR 4·7 [3·0–7·1] vs 4·3 [2·9–6·0]), but for the patients younger than 40 years, the risk was almost doubled (SIR 13·9 [5·1–30·1] vs 7·3 [3·1–14·3]). Analysis within the cohort yielded a relative risk of 2·5 (95% CI 1·6–3·7) for patients with a family history of breast or ovarian cancer, with reference to patients without such a family history. This estimate was unchanged after we controlled for both calendar period and age at breast-cancer diagnosis. By contrast with the SIR analysis, the internal analysis yielded no effect of age at diagnosis of breast cancer, which verified the almost constant absolute risk for different age-groups shown in table 3. The relative risk (95% CI) with reference to patients younger than 40 years of age was 0·9 (0·6–1·5) for those between 40 and 49 years, and 1·0 (0·5–1·7) for those older than 50 years. Furthermore, we investigated whether the effect of family history differed with follow-up time, or with age at diagnosis. Neither of these factors showed any significant modification effect on the risk associated with family history (p=0·21 for follow-up and p=0·53 for age at diagnosis).
We found the risk of developing ovarian cancer within 30 years of breast cancer diagnosis was close to 10%, and was consistent over all age groups in women with a relative diagnosed with ovarian cancer. Patients with breast cancer among their relatives experienced a smaller, but similarly age-independent cumulative risk of about 3% (table 3).
Discussion We noted a two-fold increased risk of a primary ovarian cancer in women with a primary breast cancer. In women without a family history of breast or ovarian cancer, this high risk seemed confined to patients diagnosed at young age. Our most interesting finding, however, was the striking excess risk and substantial absolute risk in women with early-onset breast cancer, and a family history of breast cancer, or especially ovarian cancer. Also noteworthy is the finding that patients with a family history, especially if an ovarian cancer is present, have a high risk of ovarian cancer. This risk is present for postmenopausal women, and seems to be constant over time. Strengths of our study include its population-based design, complete follow-up of all patients in the cohort, and identification of exposure (family history) with optimum specificity without any differential misclassification. Some limitations of our study are the possible underidentification of family history of cancer, which was probably attributable to the design of the Swedish Generation Register. To be recorded in the register, a relative must have been alive in 1960. Hence,
Age at breast-cancer diagnosis <40 years AR% (95 % CI) No family history 1·9 (1·2–2·7) Any family history 3·7 (1·7–7·6) History of breast cancer 2·9 (0·9–6·8) History ovarian cancer 8·7 (1·8–25·3) History of breast ·· and ovarian cancer Relative diagnosed 6·9 (2·2–16·0) <50 years of age Family history at diagnosis Relative diagnosed 6·2 (2·3–13·5) before patient Relative diagnosed 10·3 (2·8–26·2) before patient and <50 years
40–49 years Observed/PYR AR% (95 % CI) 26/41 978 8/6205 5/5166 3/1039 0
>49 years Observed/PYR AR% (95 % CI)
1·5 (1·1–2·1) 42/81 621 3·4 (1·9–5·7) 14/12 401 2·5 (1·2–4·8) 9/10 600 8·0 (2·2–20·5) 4/1491 9·7 (0·3–53·9) 1/310
5/2189
3·3 (0·7–9·7)
3/2726
6/2913
3·9 (1·9–7·2)
10/7671
4/1168
3·5 (0·4–12·6)
2/1724
1·4 (0·9–2·1) 5·1 (2·7–10·4) 4·1 (1·7–8·5) 7·9 (0·9–28·5) 32·3 (0·9–179)
All ages Observed/PYR AR% (95 % CI)
Observed/PYR
22/47 268 10/5291 7/5069 2/760 1/93
90/170 867 32/24 724 21/20 834 9/3290 2/600
1·6 (1·3–1·9) 3·9 (2·7–5·5) 3·0 (1·9–4·6) 8·2 (3·8–15·6) 10·0 (1·2–36·1)
9·7 (2·6–24·6) 4/1243
5·8 (3·0–10·2) 12/6158
4·0 (1·5–8·8)
4·4 (2·7–6·6) 22/15067
6/4484
6·8 (0·8–24·5) 2/885
6·4 (2·7–12·5) 8/3777
AR=absolute risk, PYR=person-years at risk of ovarian cancer after breast cancer diagnosis.
Table 3: Estimates of absolute risk of developing ovarian cancer within 30 years of breast-cancer diagnosis
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
893
For personal use. Only reproduce with permission from The Lancet Publishing Group.
ARTICLES
relatives who died from breast cancer or ovarian cancer before this year would not be recorded, and neither would relatives diagnosed before 1958, when cancer registration started. These limitations indicate that the risk of ovarian cancer might be even higher than suggested. Additionally, although this is a large study, we had limited power for more detailed analyses because of the small number of ovarian cancer cases. Extended follow-up of our cohort will remedy this concern. Lastly, our risk estimates would be biased if the rate of oophorectomy in the cohort differed from that in the female population used for comparison. Because oophorectomy has been a part of treatment in young breast cancer patients,17,18 a somewhat higher prevalence in the cohort seems most likely, and would result in an underestimation of the risk. Further methodological restrictions include the ascertainment of time-at-risk for cohort members who develop a family history during follow-up. For such an age-dependent covariate, time-at-risk up to the date of breast or ovarian cancer diagnosis in a close relative should be recorded as unexposed, and the remaining follow-up as exposed. However, such a splitting of timeat-risk might not be important for genetic traits that obviously do not vary over time. As a corollary, our risk estimates did not change substantially when we used different analytic approaches. Our results expand the knowledge of an association between breast cancer and the risk of subsequent ovarian cancer. Diagnosis at a young age has been noted as a minor risk factor, and patients with breast cancer who have mutations in BRCA1 or BRCA2 have an extremely high risk.6,7 Although such mutations are rare (present in <5% of all breast cancer cases), hereditary factors might account for close to 25% of breast-cancer cases, with similar causation in ovarian cancer.8–13 Therefore, our study lends support to theories of a connection between as yet unknown genes and cancer susceptibility. A reasonable proxy for unknown genetic risk factors might be a family history of breast or ovarian cancer; the development of a Swedish nation-based family register that provides large enough cohorts allowed us to test this hypothesis. The clinical implications of our findings are not obvious, especially because there is no effective screening strategy for ovarian cancer.19–21 However, provided that our results are confirmed by other investigators, they seem to allow identification, based on easily obtained clinical information, of a small subgroup of women with breast cancer who are at particularly high risk of ovarian cancer. In this subgroup, counselling, and perhaps even prophylactic oophorectomy, might be considered.
Conflict of interest statement None declared.
Acknowledgments We thank the Karolinska Institute for providing the resources and facilities that made this report possible.
References 1
2
3
4 5
6 7
8
9
10
11
12
13
14 15 16
17 18 19
Contributors K Bergfeldt, P Hall, F Granath, and H-O Adami developed the idea to use register linkage to assess ovarian cancer risk as a function of family history. B Rydh and K Bergfeldt analysed data in collaboration, and F Granath and L Thalib did statistical analysis. H Grönberg advised on genetic matters. All authors revised the report and agreed to its final form.
894
20 21
Harvey EB, Brinton LA. Second cancer following cancer of the breast in Connecticut, 1935–82. Natl Cancer Inst Monogr 1985; 68: 99–112. Ewertz M, Mouridsen HT. Second cancer following cancer of the female breast in Denmark, 1943–80. Natl Cancer Inst Monogr 1985; 68: 325–29. Brenner H, Siegle S, Stegmaier C, Ziegler H. Second primary neoplasms following breast cancer in Saarland, Germany, 1968–1987. Eur J Cancer 1993; 10: 1410–14. Volk N, Pompe-Kirn V. Second primary cancers in breast cancer patients in Slovenia. Cancer Causes Control 1997; 8: 764–70. Suris-Swartz PJ, Schildkraut JM, Vine MF, Hertz-Picciotto I. Age at diagnosis and multiple primary cancers of the breast and ovary. Breast Cancer Res Treat 1996; 41: 21–29. Ford D, Easton DF, Bishop DT, Narod SA, Goldgar DE. Risks of cancer in BRCA1-mutation carriers. Lancet 1994; 343: 692–95. Frank TS, Manley SA, Olopade OI, et al. Sequence analysis of BRCA1 and BRCA2: correlation of mutations with family history and ovarian cancer risk. J Clin Oncol 1998; 16: 2417–25. Loman N, Johannsson O, Kristoffersson U, Olsson H, Borg A. Family history of breast and ovarian cancers and BRCA1 and BRCA2 mutations in a population-based series of early-onset breast cancer. J Natl Cancer Inst 2001; 93: 1215–23. Syrjakoski K, Vahteristo P, Eerola H, et al. Population-based study of BRCA1 and BRCA2 mutations in 1035 unselected Finnish breast cancer patients. J Natl Cancer Inst 2000; 92: 1529–31. Anglian Breast Cancer Study Group. Prevalence and penetrance of BRCA1 and BRCA2 mutations in a population-based series of breast cancer cases. Br J Cancer 2000; 83: 1301–08. Papelard H, de Bock GH, van Eijk R, et al. Prevalence of BRCA1 in a hospital-based population of Dutch breast cancer patients. Br J Cancer 2000; 83: 719–24. Newman B, Mu H, Butler LM, Millikan RC, Moorman PG, King MC. Frequency of breast cancer attributable to BRCA1 in a population-based series of American women. JAMA 1998; 279: 915–21. Lichtenstein P, Holm NV, Verkasalo PK, et al. Environmental and heritable factors in the causation of cancer—analyses of cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med 2000; 343: 78–85. Anon. Cancer incidence in Sweden 1998. Stockholm: Centre for epidemiology, 2000. Anon. Causes of death 1997. Stockholm: National Board of Health and Welfare, 2000. Breslow NE, Day NE. Statistical methods in cancer research. Volume II—The design and analysis of cohort studies. IARC Sci Publ 1987; 82: 1–406. Ovarian ablation for early breast cancer (Cochrane Review). Cochrane Database Syst Rev 2000; 3: CD000485. Dees EC, Davidson NE. Ovarian ablation as adjuvant therapy for breast cancer. Semin Oncol 2001; 28: 322–31. Hensley ML, Castiel M, Robson ME. Screening for ovarian cancer: what we know, what we need to know. Oncology (Huntingt) 2000; 14: 1601–07. Menon U, Jacobs IJ. Recent developments in ovarian cancer screening. Curr Opin Obstet Gynecol 2000; 12: 39–42. Dorum A, Kristensen GB, Abeler VM, Trope CG, Moller P. Early detection of familial ovarian cancer. Eur J Cancer 1996; 32A: 1645–51.
THE LANCET • Vol 360 • September 21, 2002 • www.thelancet.com
For personal use. Only reproduce with permission from The Lancet Publishing Group.