Parity, Oral Contraceptives, and the Risk of Ovarian Cancer among Carriers and Noncarriers of a BRCA1 or BRCA2 Mutation

Subjects

We identified all Jewish women with pathologically confirmed cancer of the ovary (code 183.0 of the International Classification of Diseases, 9th Revision, Clinical Modification) or primary peritoneal carcinoma, possibly of ovarian origin (code 158), diagnosed in Israel between March 1, 1994, and June 30, 1999. To ensure that no patients with newly diagnosed cancer were overlooked, all the departments of gynecology in the country were monitored continually throughout the study and pathology and oncology departments were checked monthly. For each patient, two control women who were matched for age (within two years), area of birth, and place and length of residence in Israel (according to defined categories) were selected from the Central Population Registry. All living subjects gave written informed consent. The study protocol was approved by ethics panels in Israel and the United States.

The patients were interviewed in the hospital, typically four to six days after gynecologic surgery. We attempted to collect a blood sample to test for BRCA1 and BRCA2 mutations. Blocks of paraffin-embedded tumor samples were obtained routinely. Midway through the study, we began collecting buccal cells from controls for DNA analysis. The controls were interviewed at home. Interviews were conducted by a group of experienced, multilingual, trained interviewers, and when needed, the interview was conducted in the native language of the respondent.

The interviewers were informed of the goals of the study and taught how to administer the questionnaire and conduct an interview by watching practice interviews. The accuracy and thoroughness of each interviewer were periodically checked to help ensure that the method of data collection was standardized. Family information was validated by reinterviewing a random sample of 7 percent of subjects. To improve the respondents' recall with regard to contraceptive history and to establish the patterns of use, interviewers were asked to relate pill intake to life events.

Laboratory Methods

Subjects were tested for the two common founder mutations in BRCA1 (185delAG and 5382insC) and the single founder mutation in BRCA2 (6174delT) as described previously.¹⁷ Briefly, a multiplex polymerase chain reaction was designed to amplify the exons containing the three mutations with the use of fluorescence-labeled primers in a single reaction. Since each mutation is a small insertion or deletion, it can be detected as a length polymorphism with the use of a genetic analyzer (model 310, Applied Biosystems) and Genescan software (Applied Biosystems). Samples known to have mutations were included with each run as controls. Samples available for testing included peripheral blood, paraffin-embedded tissue sections, and buccal cells. DNA was extracted from tissue sections as described previously.¹⁷ Both blood and tissue sections were available for some subjects; the two subjects for whom the results were inconsistent were excluded from the analysis.

Statistical Analysis

We used logistic regression to estimate the effects on the risk of ovarian cancer of having each or any of the three mutations in BRCA1 and BRCA2. We estimated the effects of family history, parity, and oral-contraceptive use in analyses that included all patients, as would be done in a case–control study in which information on genotype was not available. We assessed the effects of parity and oral-contraceptive use further in analyses that included all controls, whether or not genotyping had been performed, but only a subgroup of patients, either patients with a BRCA1 or BRCA2 mutation or patients without a BRCA1 or BRCA2 mutation. Our approach assumed that carrier status was independent of parity and the use of oral contraceptives in the study population. Accordingly, the best estimates of the distributions of the use of oral contraceptives and parity in subgroups defined according to mutation status among the controls are their distributions among the control subjects as a whole. Restriction of logistic-regression analyses to patients who were carriers and controls who were carriers, the ideal method of assessing effects among carriers, would have left only 13 controls in this study, too few to allow us to estimate effects of parity or the use of oral contraceptives among carriers. A personal history of breast cancer and a family history of breast or ovarian cancer cannot be assumed to be independent of carrier status, because among control subjects a personal history of breast cancer and a history of having first-degree relatives with breast or ovarian cancer should be more frequent among carriers of a BRCA1 or BRCA2 mutation than among noncarriers.

All analyses were adjusted for age (in decades); ethnic background (those born in Europe, North or South America, South Africa, or Israel with two parents from these areas are referred to as Ashkenazi; those born in Israel with one parent from the Ashkenazi areas as having mixed ancestry; and all others as non-Ashkenazi); and presence or absence of a personal history of breast cancer (a possible marker for an increased risk of ovarian cancer or a decreased risk as a result of anovulation due to chemical or hormonal treatment), a family history of breast or ovarian cancer (women with a single first-degree relative with breast cancer were considered to be at intermediate risk, and those with one first-degree relative with ovarian cancer or two or more with breast cancer were considered to be at high risk), and a history of gynecologic surgery (tubal ligation, hysterectomy, or unilateral oophorectomy). We also examined the effects of oral-contraceptive use and parity according to mutation status in subgroups categorized according to age (<50 years and ≥50 years) and ethnic background (Ashkenazi and non-Ashkenazi) and to the presence or absence of a family history of breast or ovarian cancer and a personal history of breast cancer.

We used the case-only method¹⁸ to test formally whether there was an interaction between carrier status and the use of oral contraceptives and parity. This method also assumes that carrier status and the exposure of interest in the controls are independent; however, it does not allow the effects of oral-contraceptive use and parity to be adjusted for each other or for other risk factors. For some analyses, we used oral-contraceptive use and parity as continuous variables to present the data more simply and to maximize statistical power; reported parity values of more than 10 were coded as 10. Categorical analyses showed similar patterns of risk with respect to parity and the use of oral contraceptives.

During the five-year study period, 1707 Jewish women were given a diagnosis of ovarian cancer in Israel. Of these women, 1695 (99.3 percent) had pathology reports available; 1226 (71.8 percent of the total) had invasive epithelial carcinoma, 100 (5.9 percent) had invasive peritoneal carcinoma, 263 (15.4 percent) had borderline histologic findings indicating that the lesion had a low malignant potential, and 106 (6.2 percent) had cancers of nonepithelial origin. Of the 1326 women with peritoneal or epithelial cancer, 1124 (84.8 percent) were interviewed, 68 died before we could interview them, 48 were too sick to be interviewed, 86 did not consent to be interviewed, and 9 were subsequently excluded because they reported having undergone a bilateral oophorectomy. The number of cases of ovarian cancer was approximately equal in each year of the study.

Molecular analysis for founder mutations in BRCA1 or BRCA2 was completed successfully in 840 of the 1115 women with peritoneal or epithelial cancer (75.3 percent) who were interviewed.

We interviewed 2397 of the 3567 controls (67.2 percent) whom we contacted. We excluded 128 controls who reported undergoing bilateral oophorectomy. Of the 968 control women from whom we attempted to collect buccal cells, we successfully tested 751 for mutations (77.6 percent).

Table 1. Table 1. Characteristics of the Women with Ovarian Cancer, According to Whether They Underwent Mutation Analysis.

Table 1 shows the characteristics of the patients in whom mutation testing was completed, according to age, ethnic background, and presence or absence of a family history of breast or ovarian cancer. Over half the patients were 60 years of age or older and over 70 percent were classified as Ashkenazi. In the early stages of the study, patients with a family history of breast or ovarian cancer were slightly more likely to have been analyzed for a BRCA1 or BRCA2 mutation.¹⁶ There were no significant differences in the age at diagnosis and ethnic origin between patients who underwent mutation analysis and those who did not undergo testing.

Table 2. Table 2. Effect of a Founder Mutation in BRCA1 or BRCA2 on the Risk of Ovarian Cancer.

Overall, 29.0 percent of patients and 1.7 percent of controls who underwent mutation analysis had a founder mutation in BRCA1 or BRCA2 (Table 2). The prevalence of mutations among patients with invasive epithelial ovarian cancer was very similar to that among those with invasive peritoneal cancer, but it was only 4.3 percent in the group of women with borderline histologic findings (data not shown). Therefore, in further analyses we included only the 840 women with invasive epithelial or peritoneal cancer who underwent mutation analysis.

Table 3. Table 3. Effect of Parity and Use of Oral Contraceptives on the Risk of Ovarian Cancer.

Table 3 shows the effects of parity and oral-contraceptive use on the risk of ovarian cancer among the women who underwent mutation analysis. Similar results were obtained in analyses that included all women (data not shown). There was a significant decrease in risk among women with increasing parity and in those who had used oral contraceptives for five or more years.

Table 4. Table 4. Effect of Parity and Use of Oral Contraceptives on the Risk of Ovarian Cancer, According to Mutation Status.

Table 4 shows the effect of the use of oral contraceptives on the risk of ovarian cancer for patients with a BRCA1 or BRCA2 mutation and for patients with no BRCA1 or BRCA2 mutation, as compared with the entire control group. Although oral-contraceptive use was associated with a significant decrease in risk among patients without a BRCA1 or BRCA2 mutation, it had no protective effect among women with a BRCA1 or BRCA2 mutation. Increasing parity had a protective effect in both groups of women.

In continuous analyses, which may be more powerful and can be more informative in the case of individual analyses, the relative risk among all women was reduced by 3.5 percent (95 percent confidence interval, 0.1 to 6.8 percent) for each year of oral-contraceptive use. The reduction in risk was limited to women who did not have a BRCA1 or BRCA2 mutation (5.8 percent; 95 percent confidence interval, 1.5 to 10 percent); there was no apparent reduction in risk with oral-contraceptive use among the carriers (0.2 percent for each year of use; 95 percent confidence interval, –4.9 to 5.0 percent). By contrast, the reduction in risk for each additional birth was greater in carriers (12 percent; 95 percent confidence interval, 2.3 to 21 percent) than in noncarriers (6.0 percent; 95 percent confidence interval, 1.0 to 11 percent).

Supplementary Appendix 1. Ratio of Effects of Oral-Contraceptive Use and Parity, Expressed as Continuous Variables, According to Mutation Status.

In the analysis of the interactions between carrier status and the reproductive factors (see Supplementary Appendix 1, available with the complete text of this article at http://www.nejm.org), oral-contraceptive use had less of a protective effect in carriers of a BRCA1 or BRCA2 mutation than in noncarriers, but increasing parity had a greater protective effect. The small number of patients who had a BRCA2 mutation suggests that they are protected by oral-contraceptive use (odds ratio, 0.95 per year of use; 95 percent confidence interval, 0.84 to 1.08), whereas the large number of patients with a BRCA1 mutation suggests that they are not so protected, but the difference could also be due to chance (Supplementary Appendix 1).

When we examined subgroups of carriers, we found some evidence that oral-contraceptive use was protective in older women (odds ratio, 0.97 per year of use; 95 percent confidence interval, 0.90 to 1.04). These women would have been more likely than younger women to have used the high-dose pills common in the 1960s and 1970s.

Our findings show that the use of oral contraceptives and increasing parity protect against ovarian cancer in Israel, as they do in other countries.^19,20 We failed, however, to find clear evidence of a protective effect of oral-contraceptive use among women who had a founder mutation in BRCA1 or BRCA2; by contrast, increasing parity was protective in both carriers and noncarriers.

We identified as carriers 244 of 840 patients with ovarian cancer (29.0 percent). This high prevalence enabled us to investigate whether the factors that have been established as protective in the general population were also protective in carriers. However, the low frequency of oral-contraceptive use and BRCA1 or BRCA2 mutations among the controls precludes us from drawing definitive conclusions, since our study lacked the statistical power to allow us to assess effects in carriers alone or to estimate the interaction between heredity and environmental factors using all the data. This problem forced us to rely on nonstandard statistical techniques.

The precision of our estimates is less than suggested by the confidence intervals if there is, in fact, any uncertainty about the assumption that the use of oral contraceptives and parity are independent of carrier status among Israeli women.²¹ Furthermore, since the case-only analysis¹⁸ does not take demographic or additional reproductive factors into account, distortion of the estimate of interaction is possible. Despite these difficulties, we believe that our study provides substantial evidence that the effects of the use of oral contraceptives differ between women with a BRCA1 or BRCA2 mutation and those without a BRCA1 or BRCA2 mutation.

Contrary to our results, Narod et al. reported that the use of oral contraceptives had a protective effect in women with a BRCA1 or BRCA2 mutation.²² The discrepancy could be due to differences between a population-based and a clinic-based setting, to different methods, or to chance. We could compare the risk in carriers who used oral contraceptives with the risk in noncarriers because we studied all Jewish Israeli women who had ovarian cancer. By contrast, Narod et al.²² studied mainly women from high-risk families, many of whom had undergone prophylactic oophorectomy. Only additional research can resolve the discrepancy.

The reduction in the risk of ovarian cancer associated with multiparity and the use of oral contraceptives has variously been interpreted as a consequence of fewer ovulations,²³ less stimulation of the ovary by gonadotropin,²⁴ or progestin-induced apoptosis.^25,26 There is no obvious reason for a BRCA1 or BRCA2 mutation to influence these effects of oral contraceptives. If, indeed, such mutations do change the effects of oral-contraceptive use and parity, we should look for other differences between carriers and noncarriers in the pathways to ovarian cancer.

There have been reports that oral-contraceptive use has a differential effect on the risk of breast cancer in women with a BRCA1 or BRCA2 mutation²⁷ and in women at high risk because of a family history of the disease.²⁸ Results from a prevention trial of tamoxifen therapy suggest that among the women who were most likely to have a BRCA1 or BRCA2 mutation, the risk of breast cancer was higher among those who were taking the drug than among those who were not taking the drug.²⁹ Together with our data, these results necessitate caution in the use of an approach that bases the need for chemoprevention on factors known to be effective only in noncarriers or in a population that includes both carriers and noncarriers.

Our findings demonstrate the difficulty of assessing the joint effects of a rare genetic factor and environmental factors, even in a large study of a disease that is strongly associated with highly penetrant mutations in a population where such mutations are common. We believe that it is premature to prescribe oral contraceptives for the chemoprevention of ovarian cancer in carriers of a BRCA1 or BRCA2 mutation, particularly in the light of the report of a possible increased risk of breast cancer in such women.²⁹

The members of the National Israel Ovarian Cancer Group are as follows: M. Altaras, S. Anderman, S.U. Anteby, J. Atad, A. Avni, A. Bar-Am, D. Beck, U. Beller, G. Ben-Baruch, M. Ben-Ami, Y. Ben-David, H. Biran, A. Chetrit, S. Cohen, R. Dgani, Y. Fischler, A. Fishman, E. Friedman, R. Gershoni, W. Gotlieb, R. Halperin, G. Hirsh-Yechezkel, D. Idelman, R. Katan, A. Kopmer, Y. Kopolovitz, E. Lahad, L. Lerner-Geva, H. Levavi, T. Levi, B. Lifschitz-Mercer, Z. Liviatan, F. Lubin, J. Markovich (deceased), J. Menczer, B. Modan (chairman), H. Nitzan, M. Oettinger, T. Peretz, B. Piura, S. Riezel, D. Schneider, A. Shani, M. Stark, M. Steiner, Z. Tal, H. Yafe, I. Yanai, S. Zohar, and A. Zoltan.