Introduction

Breast cancer is the most common cancer in women worldwide. Although less frequent in black women, breast cancer mortality is higher than in whites. Black women, particularly those who are premenopausal, more commonly present with estrogen receptor negative (ERneg) tumors and ERneg subtypes, including “triple negative” (TN) breast cancers, which are also negative for the progesterone receptor (PR) and HER2. Because TN tumors frequently resemble “basal-like” breast cancers pathologically, the two terms are often used interchangeably; despite extensive overlap, however, they are not synonymous. The ERneg, TN, and basal-like categories are more aggressive and carry worse prognoses than ER-positive (ERpos) tumors and lack obvious molecular targets, including HER2 and ER, for known therapies such as trastuzumab, tamoxifen, and aromatase inhibitors. These ERneg, poor-prognosis tumors, combined with adverse socioeconomic factors, contribute to the higher breast cancer mortality in women of African descent [1].

Epidemiology of breast cancer health disparities in women of African ancestry

Although breast cancer incidence is lower among women of African ancestry (AA, black women) than whites at ages >40 years, a crossover in age-adjusted incidence is evident in the higher rates in black women <40 years [2]. From 1992 to 2004 breast cancer incidence among US women <40 years was 16.8 versus 15.1 per 100,000 women-years for black versus white women, with even higher rates for AA women younger than 30 years [3]. Stage-for-stage breast cancer morbidity and mortality are higher for black than white women, with black women being far more likely to be diagnosed with aggressive breast cancer at a later-stage, less likely to receive stage-appropriate treatment, and more likely to have lower stage-for-stage survival rates [4, 5]. These observations suggest a composite of socioeconomic and biological disparities as potential causes for the worse outcomes. Socioeconomic status (SES) contributes to breast cancer disparities [6], seen in lower levels of mammographic screening together with higher overall mortality and 5-year case-specific probability of death among AA compared to white women [7]. SES is also reflected in residential segregation, with black women who are living in segregated areas being less likely to receive adequate breast cancer care, in turn contributing to disparity in treatment, although not mortality [8].

Yet, endogenous biological differences also contribute to breast cancer disparities. In two large adjuvant trials, AA and white women with comparable treatment-related variables differed in progression-free and overall survival [9], suggesting a biological role in the differential outcomes. In fact, tumors in the AA women had higher mitotic indices, more cyclin D1 over-expression, and higher grades [10–12]. Poor-prognosis breast cancers among premenopausal black women have high proliferation rates, poor differentiation, frequent P53 mutations, greater lymph node involvement, and larger tumor size [13–15]. In young AA women, ER/PR-negative tumors showed a significantly higher frequency of hypermethylation in four of five designated genes; cyclin D2 methylation was significantly associated with shorter survival [16].

Importantly, ERneg breast cancer and its pathological subtypes (see below) occur more frequently in AA women, particularly premenopausal AAs [9, 15]. These clinically aggressive ERneg subtypes are also more common in women of AA living in Africa, Great Britain, and the Caribbean [17–19].

Biology of ERneg subtypes of breast cancer

Breast cancer is a heterogeneous disease for which progressive clarification of subtypes has been based on evolving technology. Immunohistochemistry (IHC), a mainstay of classification, distinguishes ERpos from ERneg breast cancers, the latter being associated with greater aggressiveness and poorer prognosis. Subsets of ERneg cancers based on staining characteristics for three key breast cancer cell antigens, ER, PR, and HER2, include the much-discussed TN [20, 21]. IHC staining of cytokeratins (CKs) further sub-classifies tumors into clinically meaningful pathological subtypes [22–24]. “Basal-like” refers to the 8–20% of breast tumors that stain positively for high molecular weight CKs (CK5/6, CK14, and CK17) since these CKs are expressed in the basal/myoepithelial cells of normal breast ducts [21, 22, 25]. These are high-grade cancers with short overall and disease-free survival [24, 26–30] and are usually, but not always, TN [20, 28, 31]. Basal-like cancers often express HER1/EGFR [24, 28]. Variable expression of known basal-like markers makes their classification according to IHC staining characteristics somewhat inconsistent [24, 25, 30, 32–34]. Sporadic basal-like tumors share key features with BRCA1-mutation-associated hereditary breast cancers, including morphology, a frequent TN status, and expression of basal CKs, p53, P-cadherin, and HER1/EGFR. Conversely, breast cancers in women carrying BRCA1 mutations are frequently basal-like as well as TN [35–37].

More refined taxonomic categories of breast cancer are based on distinctions in gene expression at the messenger RNA (mRNA) level as assayed on gene-dense microarray chips. An early study used hierarchical clustering analysis to group breast tumor samples based on common expression of specific, “intrinsic” genes. Co-expressed genes often belonged to one of eight cellular pathways previously implicated in carcinogenesis [38]. Key IHC-based biomarkers, including ER-α and HER2, were represented on the mRNA array. In fact, clustering of breast cancer samples based on common expression profiles has revealed the most important discriminator of subtypes to be the ER status of the tumor [28]. Categorization of each sample in terms of its profile of eight gene expression clusters pointed to the correlation of a given expression profile with an established pathological type of breast cancer, as based on IHC staining [32, 38]: ERpos/luminal-like (subsequently divided into luminal A, B, and C [39]); basal-like (ERneg, CKs5/6, and 17-positive); HER2/Erb-B2-positive; and normal breast. ERneg samples comprise at least two subtypes, basal-like and HER2-positive.

Despite the detailed information embedded in microarray analysis, this technology is not yet suited to large population studies or routine clinical application, in part due to the expense of arrays and the difficulty of utilizing clinically available formalin-fixed, paraffin-embedded tissues for this purpose [29]. Furthermore, the equivalence of array-based and clinical pathologic criteria in distinguishing breast cancer subtypes has not been rigorously validated [30, 40–42], in part due to inherent heterogeneity of sub-types such as basal-like. The resulting inconsistencies are compounded by the interchangeable use of the terms “basal-like” and TN [21, 43, 44]. Although they frequently share histopathological, clinical, and demographic features [21, 29], including absence of expression of ER, PR, and HER2, the two terms are not synonymous. Like basal-like tumors, the 10–17% of breast cancers that are TN are high-grade invasive ductal of no special type, metaplastic carcinomas, or medullary cancers. Both types express HER1/EGFR in up to 66% of cases [21], both behave aggressively, often presenting as interval cancers, both types are ERneg and HER2-negative and hence, lack tailored therapies, and both are more prevalent in AA and young (<50 years) women. However, differences between the two types are evident in the 15–54% of basal-like cancers that express at least one of the markers, ER or HER2 (“double negative”) [21, 29, 45]; conversely, only 85% of tumors that are ERneg and Her2-negative are basal-like by expression arrays. In addition, the “basal” CKs and HER1/EGFR, key to defining a tumor as basal-like by IHC, are expressed in only 46–90% of TN tumors [21, 44, 46, 47].

Connecting ERneg breast cancer biology to racial disparities

The evidence for a higher incidence of ERneg [48], particularly TN, breast cancer, among AA women dovetails nicely with the potential of expression arrays to delineate breast cancer subtypes in an increasingly precise manner. To surmount the technical challenges to reliable application of microarray analysis to large-scale studies, the complex mRNA expression clusters have been distilled into equivalent, or “surrogate”, IHC markers [28, 30, 44, 49]. A panel of four antibodies specific for ER, HER1, HER2, and CK5/6, has yielded a useful set of IHC subtype definitions (IHC surrogates for microarray-based definitions): basal-like (ERneg, PR-negative, HER2-negative, CK5/6-positive, and/or HER1-positive); HER2-positive/ERneg (HER2-positive, ERneg, PR-negative); luminal A (ERpos and/or PR-positive, HER2-negative); and luminal B (ERpos and/or PR-positive, HER2-positive). A surrogate IHC antibody panel accurately identified 21 tumors determined as basal-like based on their gene expression profiles and was successfully applied in large-scale fashion to tumors from 930 patients to show association between low survival and basal-like CK expression [28]. In the Carolina Breast Cancer Study (CBCS) population [13, 50], the antibody panel revealed that patients with the basal-like tumor subtype were more likely to be AA. These data are consistent with the previously discussed evidence for excess ERneg breast cancers among AA women [48]. Although the overall study population had a 20% prevalence of basal-like tumors, the AA patients had a 26% prevalence compared to 16% in non-AAs. The excess of basal-like tumors in AAs was restricted to premenopausal women (basal-like comprising 27.2% of AA premenopausal versus 16% of AA postmenopausal breast cancers) [50].

In women of AA, reproductive factors, including multiparity, younger age at menarche and early age at first pregnancy have been shown to protect against ER/PR-positive breast cancer but may be positively associated with ER/PR-negative tumors [50–52]. A meta-analysis of epidemiologic studies reported decreased risk of both receptor subtypes in association with breastfeeding and late age at menarche, the latter association being more significant for ER/PR-positive breast cancer [53]. Breastfeeding ≥6 months versus never breastfeeding were also protective for TN disease [54]. In their application of IHC surrogate markers for breast cancer subtypes to the CBCS, Millikan et al. [50] showed that breastfeeding was protective for basal-like breast cancer, with significant trends for lifetime duration of lactation, number of children breastfed, and average number of months’ breastfeeding per child. Both basal-like breast cancer and risk factors for this subtype were more prevalent among premenopausal AA than white women [50]. The authors estimated that encouraging breastfeeding and reducing abdominal adiposity could prevent approximately 68% of basal-like breast cancers.

Low SES was associated with basal-like cancers, late-stage at diagnosis and poor survival in a number of studies. Lower income correlated with increased risk of late-stage breast cancer, whereas higher educational attainment tracked with higher rates of breast cancer [55]. Regarding ERneg tumors, a few studies revealed that low SES and experiencing social deprivation placed women at increased risk for ERneg breast cancer risk [56, 57]. TN breast cancer was associated not only with race/ethnicity of African descent but also with having a lower SES [58]. For example, data from the California Cancer Registry showed that women with TN tumors were significantly more likely to be of African and Hispanic descent as well as to live in socioeconomically deprived areas [45, 58, 59].

SES may actually operate through lifestyle risk factors in influencing ERneg breast cancer. Dietary and physical activity behaviors (Tables 1, 2), although associated with breast cancer susceptibility, have not been extensively studied in relation to ER status. Healthy diets [60–62], high phytoestrogen intake [63–65] high folate intake [66, 67], fiber intake [68, 69], and possible calcium intake [70] may protect against ERneg tumors. The protective effect of folate occurs specifically in women consuming alcohol [71]. Also, postmenopausal women consuming ≥1,272 dietary folate equivalents of total folate over 10 years appear to derive a greater benefit for ERneg than ERpos breast cancers [67]. Although vitamin D had a protective effect on ERneg breast cancers in some studies, this is not always significant; in other studies the benefit existed only for ERpos cancers [72, 73]. However, a small case study of 91 women with breast cancer found that low serum vitamin D level and vitamin D deficiency were associated with increased risk for TN tumors [74]. Strenuous to moderate levels of physical activity may lower ERneg breast cancer risk [75, 76], with exercising during adolescence and within the last 10 years being associated with decreased risk of ER/PR-negative breast cancer in both premenopausal and postmenopausal women [77]. However, Bardia et al. [78] did not find a significant association between physical activity and ERneg breast cancer risk; rather, this study showed an association with ERpos breast cancer risk. In one study abdominal adiposity was associated with increased risk of basal-like breast cancer [50], while others showed that obese women with ERneg disease had reduced disease-free survival and increased mortality [79, 80]. Although prevalent among pre- and postmenopausal AA women, obesity has not been implicated as a risk factor for ERneg tumors in this population.

Table 1 Selected epidemiological studies: protective effect of dietary factors and ERneg breast cancer
Full size table
Table 2 Selected epidemiological studies: protective effect of physical activity and ERneg breast cancer
Full size table

Therapeutic interventions in ERneg breast cancer

A need exists for therapies targeted to the ERneg subtypes, especially in women of African descent who are at elevated risk of developing these aggressive cancers. Animal models that mimic attributes of ERneg breast cancer have been used to identify agents or regimens to treat and prevent these cancers. In the MMTV-Neu (mouse HER2) mouse model for ERneg, PRneg, Neu-positive breast cancer, RXR agonists and EGFR1/2 inhibitors showed preventive activity [83, 84]. Human basal-like/TN signatures are approximated in tumors in (1) SV40-T-antigen-expressing and (2) BRCA1-knockout mouse models [85]. The latter association is expected, given the prevalence of basal-like/TN features among human BRCA1-mutation-associated breast cancers [35–37], although only 40% of mouse BRCA1-knockout tumors resemble human TN tumors by microarray analysis. Preventive protocols in this ERneg model have shown that: (1) early ovariectomy inhibits later tumor formation; (2) tamoxifen is ineffective; (3) early treatment with a HER1/EGFR inhibitor decreases tumor formation; and (4) a high affinity poly(adenosine diphosphate [ADP]-ribose) polymerase (PARP) inhibitor is effective and specific in preventing BRCA1-associated lesions [86–88]. T-antigen mouse model tumors resemble human basal-like breast cancers, with alterations in P53 and RB [89]. The ornithine decarboxylase inhibitor difluoromethyl ornithine (DFMO) and certain RXR agonists (Targretin) prevent T-antigen-driven tumor development [90, 91]. Also, certain cytotoxic chemotherapy recapitulates activity in human basal-like/TN tumors, where two-thirds respond, with 40% giving a long-term response (>10 years) (Perou, personal communication). This application of animal models to the search for therapy targeting ERneg breast cancer subtypes provides important leads to drug interventions in clinical trials in humans.

The juncture where ERneg breast cancer biology meets health disparities presents a major public health challenge. Given the disproportionate burden of ERneg and basal-like/TN tumors among AA women, the lack of directed interventions against ERneg breast cancers is especially problematic. Basal-like/TN breast cancers lack established targets like ER or HER2 that can serve as a focus for treatment; therefore, the main goal in developing treatment is to move away from empirical chemotherapy toward drugs with highly specific activity against these aggressive ERneg subtypes. Traditionally, anthracycline-based and/or cisplatin-based regimens have been selected for first-line therapy, although with mixed results. Gene expression studies are refining our understanding of molecular pathways that are clinically important in subsets of breast cancers without established targets [39, 92]. Such biomolecular characterization of ERneg tumors provides the basis for clinical trials designed to characterize drug effects on specific targets and correlate such effects with clinical and tumor biomarker responses. These trials offer an invaluable resource for improving ERneg breast cancer treatment and prevention. The inclusion of minority women in ERneg target identification trials is critical.

Querying ClinicalTrials.gov with two search terms, “breast cancer” and “ERneg”, identifies about 20 trials that are studying a handful of promising molecular targets for ERneg breast cancer. These targets (Table 3) include EGFR (erlotinib, cetuximab), defective DNA repair (platinum), c-Kit (dasatinib), VEGF (bevacizumab, sunitinib), Chk1 (gemcitabine), and PARP (BSI-201, AZD2281). PARP inhibitors are likely to be useful, particularly in BRCA-mutation carriers [93]. However, a theoretical rationale suggests that PARP inhibitors should work more broadly in ERneg breast cancer even in women who are not mutation carriers [94]. Specifically with regard to breast cancer prevention in women at high risk for ERneg breast cancer, several trials are currently investigating agents such as green tea, lapatinib, PARP inhibitors, polyphenon E, targretin, 9cUAB30, and atorvastatin.

Table 3 Current clinical treatment trials of hormone receptor negative and triple negative tumors
Full size table

Discussion and future directions

The intertwined mix of factors that contribute to cancer health disparities seen in complex diseases such as cancer has been the subject of much discussion [95]. An especially controversial component of this discussion relates to the relative contribution of endogenous/biological variables that characterize a given group, as opposed to exogenous factors, including SES, cultural and behavioral characteristics. Among endogenous contributors, genetic/genomic modifications, such as specific rare mutations in established cancer-associated genes as well as the clustering of common genomic polymorphisms in selected populations, are being investigated for their role in cancer risk in these groups [96–99].

In addition to genetic predisposition to cancer, biologic factors include somatic features of cancer, such as subsets of common cancers that are disproportionately represented in specific racial groups. Our review addresses one such example, ERneg breast cancers and their TN/basal-like subsets, both distinguished by definitive gene expression profiles. These occur with higher frequency in women of AA who have breast cancer, particularly those who are premenopausal. Because of the worse prognosis of these subtypes, they undoubtedly contribute to the worse outcomes observed in this population. The question remains, however, whether and to what extent the increased propensity toward these aggressive cancers is biologically innate, i.e., genetically determined, and to what extent these cancers are byproducts of exogenous factors, including components of SES, culture, and behavior. The argument has been made that the excessive concentration of aggressive cancer subtypes in particular populations, such as the ERneg breast cancers in AA women, can be explained by epigenetic, as opposed to genetic, factors [95]. Epigenetic factors include molecular mechanisms that regulate gene expression, offering a venue through which the effect of the environment is mediated. Thus, the malleability of epigenetic changes as they interact with factors such as diet [100, 101] and physical activity [102] might explain differential cancer presentations and outcomes in specific racial groups. Yet, epigenetic modifications cannot be completely extricated from genetic influences. Apart from the specialized case of the <1% of genes that are imprinted, recent evidence points to the heritability of more generalized epigenetic modifications, both DNA methylation and histone modifications [103]. Therefore, future research must address the interplay between SES, health behaviors and both genetic and epigenetic factors, as they jointly contribute to the worse cancer presentations and outcomes in individual racial groups.

More in-depth research into the SES and health behavioral components of this interplay will be necessary in order to disentangle socioeconomic disparities from race/ethnicity and from genetic factors with regard to their contributions to breast tumor characteristics. This effort requires prospective cohort studies involving large numbers of racially/ethnically diverse women from whom carefully annotated data regarding tumor subtypes are collected. The heterogeneity of breast cancer, especially as regards racially diverse populations, will feed into cancer prevention strategies. Similarly, deeper understanding of gene-environment interactions as they relate to breast cancer risk and prognosis will allow clinicians to address the worse outcomes evident in patients from specific populations.

The National Healthcare Disparities Report by the Agency for Healthcare Research and Quality (AHRQ) detailed progress that has been made in eliminating health disparities and identified gaps in quality and access to care [104]. This detailed portrait of cancer health disparities indicates that the cause is multifactorial, especially in women of AA. Experts have debated what factor(s) should be targeted to achieve the greatest impact in eliminating disparities: should the target be socioeconomic, race, biology, access, substandard care, lack of insurance, higher stage of disease at diagnosis, or other not so well defined factors? A panel of experts convened by the American Society of Clinical Oncology’s (ASCO) Health Disparities Advisory Group debated these issues [105], and ASCO issued a policy statement pointing to low income, lack of insurance, and access to care as playing a major role in health disparities [106]. Importantly, the Institute of Medicine Report, Unequal Treatment: Confronting Racial and Ethnic Disparities in Healthcare, concluded that racial and ethnic disparities continue to persist even after controlling for income and insurance status [107]. Clinical studies have shown that even in equal access systems or where identical care is given, minorities continue to have worse outcomes in organ specific cancers that are hormonally influenced, e.g., breast, prostate, and ovarian cancers [9, 108, 109]. Why these hormonally sensitive cancers, and not others, add another component to the worse outcomes is not obvious and presents an area of health disparities that is ripe for investigation.

The ERneg breast cancers and their TN/basal-like subsets constitute a concrete biological class that is overrepresented in AA women. Therefore, a major part of the agenda to address health disparities as they relate to breast cancer in AA women must include studies of this aggressive subset of breast cancers. In addition to the epidemiologic investigations just discussed, confirmation of targeted drug activity in the treatment setting via appropriately designed clinical trials will spawn a new generation of prevention trials where the putative targets of drug action can be confirmed and monitored. The knowledge from this effort should not only increase the opportunity to intervene against ERneg breast cancer for prevention and treatment, but also provide the personalized approach that will be beneficial in addressing the deficit in breast cancer health care outcomes that are experienced by women of AA and other minority groups.

In summary, more than one factor needs to be considered in the effort to eliminate breast cancer health disparities. Additional research to address interactions among known factors as well as interactions with other undiscovered factors that influence cancer disparities, both intrinsic/endogenous and extrinsic/exogenous, is necessary.