This large and detailed prospective investigation allowed us to investigate the associations between well-known breast cancer risk factors among postmenopausal nulliparous women. Compared with parous women who first gave birth before age 25 years, nulliparous women had a 38% increased risk of breast cancer. The relative increase in breast cancer risk among nulliparous women was only 11% compared with parous women who first gave birth after age 25 years. The relative risks for all established hormone-related breast cancer risk factors were generally consistent between nulliparous and parous women, irrespective of age at first birth. Younger age at natural menopause was associated with a reduced risk of breast cancer, whereas obesity, current MHT use at baseline, family history of breast cancer in a first-degree relative, personal history of benign breast disease, and consumption of 7 or more alcoholic drinks per week were associated with increased breast cancer risk (HRs = 1.3–1.6). The combination of nulliparity and current use of MHT at baseline, obesity, or consumption of 7 or more alcoholic drinks per week was associated with a 1.8- to 2.2-fold increased risk of breast cancer relative to unexposed parous women who first gave birth before 25 years of age.
Our observation that the associations between established hormone-related risk factors and postmenopausal breast cancer are not significantly modified by parity is consistent with observations from previous case-control studies of nulliparous women (15
). The numbers of breast cancer cases in these studies, however, were substantially smaller than those in our study, and only 1 study (17
) directly assessed heterogeneity between nulliparous and parous women. Other large, pooled analyses that conducted subgroup analyses stratified by parity for body mass index (19
), MHT use (18
), family history of breast cancer (20
), and alcohol consumption (21
) similarly concluded that the effects associations between these factors and breast cancer risk did not vary by parity status. The associations between age at natural menopause and benign breast disease have not been previously reported in a large sample of postmenopausal nulliparous women. Few, if any, large studies quantified the joint associations of nulliparity and common modifiable breast cancer risk factors.
On the basis of previously reported structural and molecular differences observed in nulliparous and parous breast tissue (13
), we hypothesized that hormone-related exposures would have differential effects on postmenopausal breast cancer risk in these 2 groups. During pregnancy, the terminal ductal lobule units undergo differentiation, and it is hypothesized that these differentiated cells are less susceptible to hormonal stimulation, DNA damage, and therefore carcinogenesis (8
). Pregnancy may also induce long-term protective molecular changes, and a number of molecular differences such as expression of genes related to DNA repair and apoptosis have been observed between nulliparous and parous breast tissue (10
). In humans, these molecular differences have been reported in postmenopausal women (14
), although structural differences between the groups seem to attenuate with time (13
). The biologic relevance of these molecular differences, however, is not well understood.
Age-related changes to breast tissue may have contributed to our observation that the associations between hormone-related risk factors and postmenopausal breast cancer are not significantly modified by parity. In a small study of breast tissue obtained from women undergoing mammoplasty, well-differentiated lobules (“type 3”) constituted 70%–80% of breast structures among parous women until age 40 years (13
). Among parous women older than age 40 years and nulliparous women of all ages, they mainly observed the least-differentiated lobule type (“type 1”). A larger study of women with benign breast disease examined the prevalence of involution according to age and several breast cancer risk factors (32
). The overall prevalence of complete terminal ductal lobule unit involution was somewhat higher among nulliparous women (27.1%) than among parous women (21.2%), but age seemed to be a much stronger determinant of involution. Irrespective of parity, breast cancer risk appeared to decrease as the degree of terminal ductal lobule unit involution increased. Potentially, age-related involution displaces pregnancy-induced differences, rendering nulliparous and parous breast cells similarly susceptible to the proliferative and/or genotoxic effects of hormone-related exposures.
Parity might also act through a mechanism independent of the hormone-related exposures evaluated in our study. Consistent with this, although not definitive, is the observation by several studies that circulating levels of endogenous estrogen levels in postmenopausal women appear to be similar among nulliparous and parous women (33
We did not examine breast cancer risk according to baseline MHT formulation (i.e., estrogen only or estrogen plus progestin), because the information was not available for all of the cohorts. Although the magnitude of risk of breast cancer differs by formulation (35
), it is unlikely that availability of this information would have changed our inference that parity does not modify the association between baseline MHT and breast cancer. In a recent analysis conducted in the NIH-AARP cohort, there was no evidence of heterogeneity by parity for either estrogen alone or combined estrogen + progestin MHT (38
). All analyses were based on information reported by the women at baseline. As recency of MHT exposure is known to be an important determinant of breast cancer risk (18
), it is possible that we underestimated the association with recent MHT use if women stopped using MHT during follow-up. Additionally, women reporting “never use” at baseline may have used MHT during follow-up, which would also attenuate the relative risk comparing “current” or “former” with “never” users. We also lacked information about the estrogen- and progesterone-receptor status of breast cancers. Previous studies suggest that breast cancer risk factors, including nulliparity, may differ by hormone receptor status (39
). Differences in risk factor profiles (5
) and hormone receptor status patterns of breast cancers (40
) across racial and ethnic groups may limit the generalizability of the magnitude of our results to non-Caucasian women.
A key strength of our aggregated study was the large sample size of more than 32,000 nulliparous women, which allowed us to assess multiple hormone-related breast cancer risk factors with adequate power. In addition, we were able to compare our results with 2 groups of parous women, defined by age at first birth. This is important given that animal (8
) and epidemiologic (2
) studies have clearly established age at first birth as a strong determinant of breast cancer risk. The large sample size also enabled us to estimate the combined associations of parity status and several modifiable breast cancer risk factors with adequate power. Another major strength was the use of prospective data, which ensured that all nulliparous and parous women enrolled were included whether or not they developed cancer.
In summary, the observed risk associated with older age at menopause, obesity, current use of MHT at baseline, first-degree family history of breast cancer, benign breast disease, and regular alcohol consumption was similar among nulliparous and parous women. Although we acknowledge the importance of reducing the prevalence of modifiable breast cancer risk factors in all women, quantification of the joint associations of nulliparity and certain exposures provides risk estimates that are specific to nulliparous women who have a higher baseline risk than parous women.