Our results indicate that the risk of breast cancer in parous BRCA1/2 mutation carriers is not statistically significantly different from that in nulliparous mutation carriers. However, in parous BRCA1/2 mutation carriers, the risk of breast cancer decreased by approximately 14% for each additional birth. This risk reduction, however, appeared to be restricted to women older than 40 years. The protective effect associated with number of full-term pregnancies appeared to be similar in BRCA1 and BRCA2 mutation carriers, and was roughly comparable to that seen for breast cancer in the general population ; in the largest dataset analyzed, the estimated risk reduction, was 7.0% (SE 1.0%) with each additional birth in women who had never breastfed (15
) We found some evidence that the association between age at first full-term pregnancy and breast cancer risk differed in BRCA1 and BRCA2 mutation carriers. For BRCA2 mutation carriers, full-term pregnancy before the age of 20 years was associated with a lower risk of breast cancer than later age at first pregnancy, whereas in BRCA1 mutation carriers, later age at first pregnancy appeared to be associated with a lower risk of breast cancer. Neither interrupted pregnancies nor history of breastfeeding were associated with the risk of breast cancer in this cohort. However, considering that the estimated relative risk reduction associated with duration of breastfeeding in the general population is quite modest (4.3% for every 12 months of breastfeeding, according to the collaborative group on Hormonal Factors in Breast Cancer ), our data can not rule out a risk reduction of this magnitude.
Our study has several limitations. First, our results are based on retrospective information obtained from women who opted for BRCA mutation screening and genetic testing. And one assumption that underlies the method of weighting used in our analyses is that the absolute disease risks are well estimated and ascertainment is not dependent on the covariates of interest (26
). This assumption would be violated if having children changed the likelihood of women to opt for genetic testing. This was indeed observed in two studies, including both affected and unaffected women, that the uptake of genetic testing was greater in women with children than in nulliparous women (28
). Unfortunately the magnitude of the potential bias when this assumption is violated has not been investigated. However, if ascertainment was associated with high parity, the difference in the number of children between affected and unaffected women would be reduced and weighting would lead to results that were biased toward the null rather than results that overestimated associations. Moreover, if parity was indeed associated with a reduced risk of breast cancer in BRCA1/2 mutation carriers, tested individuals with higher parity would, in effect, be susceptible to lower breast cancer risks than the breast cancer risks used to compute the weights. Consequently, the weights assigned to women with breast cancer would be too low thus resulting in estimates which are biased toward the null as shown by the simulations in Antoniou et al. (26
). Bias could also arise if unaffected women are more likely to choose to undergo testing when they have children and affected women are more likely to opt for a test because of their diagnosis of breast cancer. We are unaware of any study that has assessed whether uptake of genetic testing in women differs according to her breast cancer affection status as well as the number of children she has borne and we cannot assess this potential bias using available IBCCS data. However, this possible bias will be explored further by using data from the GEO study being conducted in The Netherlands; this study is collecting detailed information from women who opted for genetic testing for BRCA genes and from those who did not. Nevertheless, the influence of reproductive factors on genetic testing might differ by country and this possibility warrants further study in other populations as well.
Second, the difference in birth year between affected and unaffected women could be another source of bias in our study, particularly if the reproductive patterns in our study population have changed substantially across birth cohorts. For example, the unaffected women were born an average of 7 years later than the affected women. To correct for this potential bias, all of our analyses were performed stratified by the year of birth (in decades). The method used for weighting, which was cohort-specific, also contributed to correcting this potential bias. Moreover, we assessed this possible bias by analyzing the data by birth cohort and found similar HRs for childbearing and breastfeeding for all birth cohorts.
Another possible bias that might have influenced our results is survival bias resulting from the inclusion of affected women who survived long enough to participate in this study. Thus, some of the observed associations might, in part, be related to the breast cancer prognosis rather than to the risk of breast cancer. We therefore analyzed data from the pseudo-incident cohort to evaluate the effect of such a bias on our findings. We cannot totally exclude the possibility that a residual survival bias remained in the pseudo-incident cohort because some women with early-onset breast cancer who had a poor prognosis may not have still been alive 5 years later to participate in our study. However, the results based on the pseudo-incident cohort were not substantially different from those based on the entire cohort, suggesting that survival had negligible effect on our results.
In the general population, pregnancy is associated with a reduced risk of breast cancer. However, there is a transient increase in breast cancer risk after each birth, which is then followed by a reduced risk of breast cancer, so that the association between pregnancy and a reduced risk of breast cancer only becomes evident in women who are older than 40 or 50 years (30
). The magnitude of the risk reduction associated with childbirth that we observed in our cohort of BRCA1/2 mutation carriers is consistent with that observed in the general population. However, we did not detect any transient increases in risk in our cohort. The mechanisms responsible for the long-term association between a full-term pregnancy and a reduced risk of breast cancer may differ between BRCA1/2 mutation carriers and the general population. Most previous studies of BRCA1/2 mutation carriers have not found statistically significant associations between childbearing and the risk of breast cancer (17
). However, these studies were small and their estimated relative risks are, therefore, imprecise. Only one study has reported an association between parity and a reduced risk of breast cancer in BRCA1 mutation carriers (23
). Our results are consistent with those of Jernström et al. (19
), who also found that the risk of breast cancer by age 40 years in BRCA1/2 mutation carriers increased with the number of full-term pregnancies. However, in that case–control study, women who were diagnosed with breast cancer after age 40 years were also used as age-matched control subjects for case patients who were younger than 40 years at breast cancer diagnosis. Thus, Jernström et al. (19
) might have overestimated the risk of breast cancer associated with childbearing by age 40 years if childbearing was associated with a reduced risk of breast cancer after age 40 years. More recently, the same group published results of a case–control study that was based on a larger sample of 1,260 matched sets and found that the risk of breast cancer increased with increasing parity, but only among BRCA2 mutation carriers who were 50 years or younger (33
). In this analysis, again, breast cancer case patients were only included if they could be matched to a control subject. Thus, the survival bias might have been even more exaggerated by excluding as control subjects all women who had developed breast cancer at any time. It would be useful to evaluate how much the matching strategy as well as the familial clustering of mutation carriers in that study affected the results.
In the general population, younger age at first birth is associated with decreased risk of breast cancer (34
). The association between high parity and a reduced risk of breast cancer has been found to be particularly strong among women who first gave birth before the age of 20 years (35
). In our study, we observed a reduced risk of breast cancer among BRCA2 mutation carriers who had a full-term pregnancy before age 20. However, among BRCA1 mutation carriers, the risk of breast cancer was inversely related to the woman’s age at first birth. Other studies of BRCA1 and BRCA2 mutation carriers that examined the effect of a woman’s age at first birth found no decreased risk associated with a young age at first childbirth (20
). One of those studies (20
) found that mothers of BRCA1/2 mutation carriers who had their first full-term pregnancy after age 30 years had a lower risk of developing breast cancer than mothers of BRCA1/2 mutation carriers who had their first full-term pregnancy before age 25 years. These findings are compatible with our observations for BRCA1 mutation carriers (18
). Although the difference in the pattern of risk between BRCA1 and BRCA2 mutation carriers that we observed might be due to chance, it might also reflect real differences in the natural history of breast cancer in BRCA1 and BRCA2 mutations carriers. In particular, BRCA1 mutations confer a much greater risk of breast cancer at very young ages than do BRCA2 mutations, and invariably predispose women who carry such mutations to estrogen receptor-negative tumors (36
The BRCA1 gene is thought to play a key role in the normal proliferation and differentiation of cells in the mammary gland. For example, in mice BRCA1 expression is regulated during mammary gland development and increases during puberty and pregnancy, possibly to limit proliferation and promote differentiation (37
). Recent studies in cell lines and in rats have shown that estrogens may increase expression of BRCA1, which decreases the activity of estrogen receptor-α-mediated pathways, thereby suppressing cell proliferation (9
). In some experimental animal models, susceptibility to mammary cancer is strongly related to the proliferation of the mammary epithelial cells and inversely related to the degree of differentiation (11
). Pregnancy furthers the differentiation of the terminal end buds and induces dramatic changes in the parenchyma-stroma ratio of breast tissue, thereby conferring protection against the development of breast cancer (11
). The lobular architecture of the breast tissue in parous women who had a family history of breast cancer was found to resemble that of nulliparous women without a family history rather than that of parous women without a family history (39
). Thus, it was postulated that the breast tissue from women with hereditary breast cancer suffers from a disturbance of cell differentiation following pregnancy and altered interactions between the epithelium and the stroma (39
). Nevertheless, a higher proportion of more differentiated lobular structures in the parous than nulliparous women was also observed in women from hereditary breast cancer families (39
). These observations support the hypothesis that pregnancy is associated with a reduced risk for breast cancer in BRCA1/2 mutation carriers, but also raise the possibility that the extent and pattern of this protective effect may be different from that observed in the general population.
In conclusion, our data provide evidence that multiple full-term pregnancies are associated with a moderate reduction in the risk of breast cancer in BRCA1 and BRCA2 mutation carriers, which is only evident in women older than 40 years. This decrease in breast cancer risk appears to be consistent with that found in the general population. Further studies are needed to understand the mechanisms underlying the observed long term protective effect of full-term pregnancies on breast cancer risk. Nevertheless, the decrease in risk of breast cancer associated with multiple pregnancies might be used to revise the risk estimates given to BRCA1/2 mutation carriers. The association of age at first full-term pregnancy with breast cancer risk, however, were less consistent with that seen in the general population and might differ by gene. These findings may reflect differences in the pathogenesis of cancers associated with the two genes and warrants further investigation. population and might differ by gene. These findings may reflect differences in the pathogenesis of cancers associated with the two genes and warrants further investigation.