The Epidemiology of Pregnancy-Related Breast Cancers: Are We Ready to Deliver?

Breast cancer incidence rates in women under age 50 years have been increasing globally for close to 30 years (1). As population-based screening by mammography does not usually start until age 50 in countries where it is available, higher breast cancer incidence cannot be accounted for by increased breast cancer screening. In the U.S., breast cancer rates have been increasing in women under 55; annual percent increases are highest in women under age 40 who also are not routinely screened (1). Coupled with these disturbing trends is the reality that breast cancer in young women is often more aggressive with worse outcomes particularly for women of color. Thus, understanding the burden of early onset breast cancer is of major public health significance to reduce cancer health disparities. Secular trends in risk factors are often investigated to explain secular trends in incidence and reproductive trends have changed over the last decades. Yet, reproductive trends alone cannot account for changing global breast cancer incidence (2). Further, the oldest tumor registry in the U.S. supports that breast cancer incidence has been increasing in women under 40 years for 80 years and long before the baby boom (3). Thus, changing breast cancer incidence patterns cannot be accounted for solely by changes in parity – after all, pregnancy is not a carcinogen. However, pregnancy is a window of susceptibility (WOS) for breast cancer and therefore understanding how the breast tissue changes during this WOS is critical for understanding how carcinogens may have a greater impact during this period.

The WOS hypothesis postulates that the breast is most susceptible to carcinogenic exposures during certain periods across the life course (4). For example, the time between pubertal breast development and first pregnancy has widened translating into a longer period where the breast tissue is less differentiated and more susceptible to carcinogens. Yet, the key to understanding WOS is understanding the opposing effects of pregnancy on breast cancer risk interacting with other environmental and lifestyle exposures. For over half a millennium it has been recognized that pregnancy is related to breast cancer as certain groups of women without children, like nuns, were more likely to have breast cancer. Renowned cancer epidemiologist Joseph Fraumeni quantified this observation in a study of over 30,000 nuns in a 1969 JNCI paper (5). However, not long after, it became much clearer that the effects of pregnancy on breast cancer might be very different in the short-term. This “dual” or “transient” effect of pregnancy has been challenging to measure as overall parity and age at pregnancies are highly correlated. One approach measures the independent effect of age at first birth and the age at last birth. In a 1987 AJE study, as epidemiologists Kvâle and Heuch followed 63,090 Norwegian women for 20 years, they found an increase in breast cancer risk immediately following childbirth as well as among women with widely spaced pregnancies (6). Timing matters.

In the current issue of CEBP, Vohra and colleagues present important data related to postpartum breast cancer (PPBC; ref. 7). PPBC is distinct from pregnancy-associated breast cancer; however, definitions in the literature vary presenting a challenge to studying pregnancy-related breast cancers (see Fig. 1). Vohra and colleagues compared three groups of women diagnosed with breast cancer under age 50 years defined by the timing of the last full-term pregnancy (FTP): (i) last FTP within 10 years (recently postpartum), (ii) last FTP >10 years (remotely postpartum), and (iii) no FTP (nulliparous). Recently postpartum women were younger at the time of diagnosis (median = 39 years) compared with both the remotely postpartum (median = 45 years) and nulliparous (median = 42 years). In addition, the recently postpartum women were mostly of Black race and older at the time of their last FTP. Although there were differences in some clinical and molecular characteristics between ever parous and never parous, Vohra and colleagues did not find that recently postpartum women had more aggressive clinical and molecular tumor characteristics when compared with the remotely postpartum.

Intriguingly, Vohra and colleagues did find that differences based on timing of pregnancy for immunologic markers. Specifically, women who were recently postpartum were more likely to have had an upregulation of adaptive immunity genes. Consistent with the wound-healing theory and the postpartum breast microenvironment (8), Vohra and colleagues found that adaptive immunity and adaptive immune cells were elevated in recently postpartum women compared with nulliparous women, when controlling for breast cancer subtype, race, age, and study phase. Levels of adaptive immunity were attenuated among remotely postpartum women.

While intriguing, there are several methodological issues to consider. First, Vohra and colleagues improved upon some earlier studies with a broad span of age complicating inference, by limiting their analyses to women diagnosed with breast cancer under age 50 years. However even within this more narrow age range, recently postpartum women were more likely to have had a pregnancy and be diagnosed in their 30s, while remotely postpartum women were more likely to have had a pregnancy in their 20s and be diagnosed in their 40s. This contrast highlights the challenge of disentangling the effects of timing. Second, albeit a larger study than most, only 43% to 48% of the recently postpartum, remotely postpartum, and nulliparous women were included for tumor microenvironment (TME) RNA analysis. Those with RNA available were more likely to have larger tumors, higher tumor stage, higher grade, and lymph node positivity. Thus, even with this large study, issues of selection bias and statistical power affect interpretation particularly as many of the tumor markers may be correlated with each other as well as the TME markers. The tumor samples were collected over 20 years during three waves of recruitment, which adds to the complexity of interpretation.

While literature has focused on the innate arm of the immune system within the wound-healing process (9), the lingering presence of adaptive immune surveillance cells raises additional research directives. First, Vohra and colleagues measured expression of many T cell types, given T cells are not differentially equal. However, whether these lingering CD8⁺ T cells are functional or exhausted requires further exploration. Second, cells exist in a heterogenous population where prevalence ratios may be insightful. For example, a prolonged and abnormal CD8:CD4 ratio is associated with impaired epithelialization (or the process by where epithelial cells migrate and repair the wounded area; ref. 10). An impaired or altered cell:cell ratio may provide the needed gateway for cellular exit and the potential for metastasis. Thirdly, the innate and adaptive immunity arms work together; thus, a joint analysis of innate and adaptive immunity is warranted (11).

Breastfeeding likely has a direct role in altering the breast TME. Yet, among the studies that have examined pregnancy-related breast cancer (12–22), few include breastfeeding measures, and if they do analyses are generally limited to ever/never as in the Vohra study (14, 20, 21). More information on breastfeeding constructs (duration and type) will be essential to inform actionable risk-reducing messaging. Tumor-promoting features during postpartum remodeling of the mammary gland is exacerbated in the absence or early cessation of breastfeeding. Therefore, understanding the biological impact of breastfeeding on breast cancer can aid in the identification of alternative breast cancer risk-reducing strategies where an equitable institutional implementation could impact downstream breast cancer health disparities. Black women are most likely to be diagnosed with breast cancer at younger ages, to be diagnosed with aggressive disease, and have lower rates of breastfeeding. Thus, the inclusion of breastfeeding constructs are an important equity issue that should be addressed when studying pregnancy-related breast cancers. Inclusion of breastfeeding constructs can inform the etiology, prognostics, and therapeutics for breast cancer in Black women (23).

We congratulate Vohra and colleagues for their important study, which challenges us to develop more consistent and rigorous approaches to the study of pregnancy-related breast cancer. The existing literature with few exceptions like Vohra and three others (12, 14, 21) all studied fewer than 200 breast cancer cases limiting the ability to examine associations by breast cancer subtypes. The following are a few considerations that warrant broad debate in order to move forward for studying pregnancy-related breast cancer risk and prognosis:

1. Arbitrary definitions but not an arbitrary disease: We need consistent definitions for rigor and reproducibility of findings for studies on pregnancy-related breast cancers (Fig. 1). We agree with Amant and colleagues (24) that definitions have important implications for interpretation. During the postpartum WOS, the duration of time a woman has an increased risk of breast cancer varies by age at first birth, age at last birth, and by breast cancer family history (25).

2. Risk-reducing strategies: We should evaluate the impact of expanding our definition of higher risk women for breast cancer to the immediate window after pregnancy. The postpartum period may be an important window for future risk reduction strategies and early breast cancer screening.

3. Detailed breastfeeding measures: To identify alternative pathway-based strategies for women who cannot or do not choose to breastfeed, we need additional studies to evaluate the impact of breastfeeding duration and the underlying mechanisms on the breast TME.

4. Cellular heterogeneity: We need comprehensive studies to estimate the intracellular interactionswithin theTME,within and between, the innate and the adaptive immunity arms. Such studies could greatly inform the functional features of the wound-healing hypothesis for early onset breast cancer.

No disclosures were reported.