Use of Four Biomarkers to Evaluate the Risk of Breast Cancer Subtypes in the Women's Contraceptive and Reproductive Experiences Study

Epidemiologic studies have suggested that some hormone-related breast cancer risk factors differentially influence breast cancer risk by estrogen receptor (ER) and progesterone receptor (PR) expression status in tumor tissue (1–3). However, it remains unclear whether these associations differ further by the expression status of human epidermal growth factor receptor 2 (HER2) and p53 protein (p53).

HER2, a transmembrane tyrosine kinase receptor protein, is normally involved in the signal transduction pathways that lead to cell growth and differentiation (4, 5). HER2 is overexpressed in approximately 15% to 25% of breast carcinoma specimens (6–8). Several epidemiologic studies have classified breast cancers as triple-negative (ER⁻/PR⁻/HER2⁻), luminal A (ER⁺ or PR⁺ plus HER2⁻), luminal B (ER⁺ or PR⁺ plus HER2⁺), and ER⁻/PR⁻/HER2⁺ subtype (7, 9–11). Triple-negative breast cancers, which are typically more aggressive and have poorer prognosis than other subtypes, are often characterized by a basal-like molecular profile (12, 13). Basal-like breast cancers, as defined by gene expression microarray analysis, exhibit overexpression of several genes in the p21 (CDKN1A) pathway that play a critical role in cell proliferation and DNA replication (MCM3, MCM4, MCM7, and MAD2L1), whereas luminal A tumors have been associated with the ER signaling pathway (14, 15). Therefore, one might expect that associations with hormone-related breast cancer risk factors would differ between triple-negative breast cancer, a proxy for basal-like breast cancer, and other breast cancer subtypes. Thus far, epidemiologic data on this topic are inconsistent (9, 11, 16).

p53, a tumor suppressor gene, inhibits the proliferation of abnormal cells (17). p53 gene mutations that occur in approximately 15% to 35% of breast carcinoma specimens (18–20) may cause loss of tumor suppressor function and gain of oncogenic activity (21). Overexpression of p53 in tumor tissue is associated with the presence of the mutations, especially missense mutations, in the p53 gene (21). It has been unclear whether the overexpression status of p53 would modify the associations between hormone-related breast cancer risk factors and breast cancer risk. Three previous epidemiologic studies have examined the effects of oral contraceptive (OC) use and reproductive factors by p53 status, but none has concurrently considered the expression status of ER, PR, and HER2 (22–24).

In this study, we examined whether OC use and reproductive factors differentially influenced risk for triple-negative, luminal A, luminal B, and ER⁻/PR⁻/HER2⁺ breast cancer. Further, we explored whether any differences in the associations of OC use or reproductive factors with triple-negative and luminal A breast cancer, two common subtypes, were confined to younger (35–44 years) or older (45–64 years) women. We also evaluated whether p53 expression status modified any observed differences between these two subtypes.

The Women's Contraceptive and Reproductive Experiences (CARE) Study is a population-based case-control study of invasive breast cancer conducted among U.S.-born White and African-American women ages 35 to 64 y who resided in one of five areas of the United States [Atlanta, Detroit, Los Angeles (LA) County, Philadelphia, or Seattle; ref. 25]. This current study is restricted to women who participated in the LA County or the Detroit Metropolitan Area, which includes Wayne, Oakland, and MacComb counties.

Case participants in the Women's CARE Study had no prior diagnosis of invasive or in situ breast cancer and were diagnosed with their first primary invasive breast cancer between June 1994 and August 1998. Control participants were women with no history of invasive or in situ breast cancer who were identified by random digit dialing. Control participants were frequency matched to the expected distribution of cases in strata defined by 5-y age groups, ethnicity (White or African-American), and residence located in the same geographic (study) region. The Women's CARE Study recruited 1,921 cases (1,072 White and 849 African-American) and 2,034 control participants (1,161 White and 873 African-American) from LA and Detroit. The interview response rates were 73.3% for cases in LA, 73.7% for controls in LA, 74.7% for cases in Detroit, and 74.1% for controls in Detroit.

The Women's CARE Study collected demographic characteristics, detailed information about OC use, complete histories of menstrual and reproductive factors, family history of breast cancer, and information pertaining to other factors from each participant during an in-person interview. Information was recorded up to a predetermined reference date for each participant, the date of diagnosis for case patients, and the date of initial telephone screening of the household for control subjects. The detailed information on OC use was collected by using a mixture of recall and recognition techniques (structured questionnaire, response cards, color pictures of all OC preparations marketed in the United States, and a life events calendar). The detailed data about OC use and breast cancer risk (26), and combined effect of OC use and hormone replacement therapy on breast cancer risk (27), appear elsewhere. All participants provided written informed consent. The study protocol was approved by the federally approved Institutional Review Boards at participating institutions.

Paraffin-embedded tumor blocks were obtained for 1,333 cases (LA, 919; Detroit, 414), ∼80% of those requested. Tumor blocks were carefully reviewed and evaluated in Dr. Press' laboratory at the University of Southern California (referred to as the centralized laboratory). This laboratory was among the first to validate use of monoclonal ER antibodies (28–32) and monoclonal PR antibodies (33–35) for localization of ER and PR in tissue sections by immunohistochemistry and to subsequently adapt these methods to analysis of archival tissues (36, 37).

We excluded 127 case samples because tumor blocks contained only carcinoma in situ (n = 56) or no tumor tissue (n = 46), were H&E-stained tissue sections (n = 8), had insufficient tissue for assay (n = 3), or had other problems (n = 14). We successfully determined ER, PR, HER2, and p53 expression status for 1,206 case subjects (LA, 839; Detroit, 367) in the centralized laboratory. In one previous study of the association between percent mammographic density and subtypes of breast cancer, we reported on 352 of the LA cases for whom we also had mammograms (7). In a prior report comparing ER/PR status from the centralized pathology laboratory with the classification obtained from a population-based cancer surveillance registry, we reported on concordance for ER and PR status between the two sources for 919 cases (38).

The status of ER and PR was determined using previously published immunohistochemistry methods (31, 36). Immunostaining results for ER and PR expression were interpreted in a blinded fashion and scored semiquantitatively based on the visually estimated percentage of positively stained tumor cell nuclei. The intensity of nuclear staining was scored for individual tumor cell nuclei as negative (−)/no staining, plus one (+1)/weak intensity, plus two (+2)/intermediate intensity, or plus three (+3)/strong intensity. A minimum of 100 tumor cells was scored with the percentage of tumor cell nuclei in each category recorded. In this study, an overall score of ≥1% immunostained tumor cell nuclei was considered positive for ER or PR status.

HER2 expression status was determined by immunohistochemistry using the 10H8 monoclonal antibody (8, 39) to assess HER2 membrane protein immunostaining. No immunostaining (0) or weak (1+) membrane immunostaining was considered low HER2 expression (HER2⁻). Moderate (2+) or strong membrane immunostaining (3+) was considered HER2 overexpression (HER2⁺).

The expression status of p53 was determined by immunohistochemistry using the monoclonal mouse antibodies DO7 (Oncogene Science, Inc.) and BP 53-12-1 (Biogenex) to measure p53 nuclear protein immunostaining. Based on previous study findings comparing p53 mutations in exons 2 to 11 with p53 expression levels (40, 41), any nuclear staining for p53 was deemed positive (42).

We used t tests to evaluate differences in continuous variables and Pearson χ² tests to evaluate differences in the frequency distributions of categorical variables comparing case patients with control subjects.

We assessed the association of triple-negative, ER⁻/PR⁻/HER2⁺, luminal A, and luminal B breast cancer with the following factors: OC use (never, ever), duration of OC use, age at first OC use, duration between menarche and first OC use, time since last OC use to reference date, number of full-term (>26-wk gestation) pregnancies, age at first full-term pregnancy (defined for each woman as the age at which that pregnancy ended), and duration of breast-feeding. We estimated odds ratios (OR) and corresponding 95% confidence intervals (95% CI) using multivariable polychotomous unconditional logistic regression for case-control comparisons.

Tests for trend were conducted by fitting ordinal values corresponding to exposure categories and testing whether the slope coefficient differed from zero. We also conducted Wald χ² tests for homogeneity of the associations with OC use or reproductive factors across breast cancer subtypes by fitting a multivariable polychotomous unconditional logistic regression model for dichotomous or ordinal variables.

We included the following factors, selected a priori, as potential confounders in all multivariable models: study site (LA or Detroit), race (white or African-American), education as a proxy for social economic status (high school or lower level of education, technical school or some college, or college graduate), age (35–39, 40–44, 45–49, 50–54, 55–59, or 60–64 y), family history of breast cancer [first degree (mother, sister, or daughter); no first-degree family history including 4% of participants with uncertain answers], age at menarche (≤11, 12, 13, >13 y), menopausal status (premenopausal, postmenopausal, unknown), and body mass index (BMI) 5 y before the reference date (continuous variable, kg/m²). We also included parity (nulligravid, pregnant but no full-term pregnancy, or parity 1, 2, 3, ≥4) in models as a potential confounder when it was not the exposure of interest. When parity was the exposure of interest, we chose women who had never been pregnant as our reference group and treated women who had been pregnant but had never carried to term as a separate group that was excluded when testing for trend across categories of parity. When restricting analyses to parous women, a single model was fit to assess the joint effects of age at first full-term pregnancy and breast-feeding duration.

Using two major subtypes, triple-negative and luminal A breast cancer, we explored whether any differences in the associations of OC use or reproductive factors were confined to younger (35–44 y) or older (45–64 y) women. We also evaluated whether p53 expression status modified any observed differences between these two subtypes.

We excluded 19 control subjects and 9 case patients from the analyses who were missing information on OC use (6 controls, 4 cases), BMI (9 controls, 4 cases), or parity (4 controls, 1 case). This resulted in 2,015 controls and 1,197 cases available for the current analysis. Of the 1,197 cases, 28.0% were triple-negative, 8.1% were ER⁻/PR⁻HER2⁺, 53.9% were luminal A, and 10.0% were luminal B breast cancer. Among 335 cases with triple-negative breast cancer, 45.1% were p53⁺, whereas among 645 cases with luminal A breast cancer, 29.2% were p53⁺ breast cancer.

In reporting the results of trend tests or homogeneity tests, we considered a two-sided P value of <0.05 as statistically significant. All analyses were performed using the SAS statistical package (version 9.2; SAS Institute).

As previously reported among all participants of the Women's CARE Study (26, 43, 44), there were no differences in the OC use characteristics between case and control patients; in contrast, case participants were more likely to have a first-degree breast cancer family history (P_χ² < 0.0001), to have fewer full-term pregnancies (P_{t test} = 0.008), and to have never breast-fed (P_χ² = 0.02) compared with control participants (Supplementary Table S1).

Examination of multiple aspects of OC use (ever use, duration of use, age at first use, interval between age at menarche and age at first OC use, and time since last OC use to reference date) showed that OC use was not associated with any subtype defined by ER, PR, and HER2 (Table 1).

Number of full-term pregnancies was not associated with risk of triple-negative breast cancer (P_trend = 0.84) but was inversely associated with risk of the other three subtypes (all P_trend < 0.05; Table 2). The difference in trends across the four subtypes of breast cancer was marginally statistically significant (homogeneity test: P = 0.06).

Among parous women, older age at first full-term pregnancy was not associated with any subtype defined by ER, PR, and HER2 (Table 2). A statistically significant protective effect of longer duration of breast-feeding was observed for the two major subtypes: triple negative (P_trend = 0.03) and luminal A (P_trend = 0.004).

Analyses of triple-negative and luminal A subtype within age strata (younger women: 35–44 years and older women: 45–64 years) showed that older women who initiated OC use before age 18 years had a 2.9-fold increased risk for triple-negative tumors compared with women in the same age group who had never used OCs (OR, 2.87; 95% CI, 1.44–5.74); no such association for luminal A cancers was observed among older women (OR, 1.36; 95% CI, 0.75–2.48) or for either subtype among younger women (Table 3).

Among older women, those who had their first full-term pregnancies at or after age 25 years seemed to have a higher risk of luminal A breast cancer than those who had their first full-term pregnancies before age 20 years (P_trend = 0.04); however, the risk estimates declined in the oldest age group and the CI for the risk estimate for women with a first full-term pregnancy at or after age 30 years included 1.0 (OR, 1.22; 95% CI, 0.71–2.09; Table 4).

Similar to the analyses across all age groups, risk of luminal A cancers decreased with an increasing number of full-term pregnancies for both younger and older women (P_trend = 0.09 and P_trend = 0.006, respectively; Table 4). Increasing duration of breast-feeding was inversely associated with risk for both subtypes in younger and older women.

We evaluated whether any associations between OC use or reproductive factors and breast cancer risk differed by p53 status alone and did not observe any statistically significant differences in these associations (results not shown). Subclassification of the triple-negative and luminal A subtypes by p53 status did not further differentiate the associations of these subtypes with OC use (Table 5) or reproductive history (Table 6). Although we found that number of full-term pregnancies was statistically significantly associated with luminal A tumors without p53 overexpression (P_trend = 0.002) and that duration of breast-feeding was statistically significantly associated with both triple-negative (P_trend = 0.03) and luminal A tumors (P_trend = 0.008) without p53 overexpression, these associations did not differ statistically from those for the corresponding subtypes where p53 was overexpressed (all homogeneity tests P > 0.20; Table 6).

A recent study reported that, among women ages 20 to 45 years, use of OCs for at least 1 year was associated with a 2.5-fold increased risk for triple-negative breast cancer (45); this association was not observed in another population-based case-control study that included women ages 20 to 74 years (16). We previously reported that OC use was not associated with breast cancer risk among participants in the Women's CARE Study (26). In the current analysis, OC use was not associated with any subtype except for a 2.9-fold increase in risk for triple-negative breast cancer among older women (ages 45–64 years) who started OC use before age 18 years. All of these older triple-negative breast cancer patients began OC use before 1971 when most OCs contained high doses of synthetic estrogen. In contrast, exposure to OCs at an early age in older women did not significantly increase the risk of luminal A breast cancer in our data.

The association with OC use at a young age among older women that is restricted to triple-negative breast cancer can be viewed as inconsistent with the hypothesis that hormone-related risk factors act through estrogen mediated by its receptor (46). However, there is evidence that ER⁺ progenitor cells produce paracrine signals when exposed to estrogen, which cause the proliferation of nearby ER⁻ cells (47). Furthermore, the Breast and Prostate Cancer Cohort Consortium showed that two common haplotypes of the 17β-hydroxysteroid dehydrogenase 1 gene are associated with risk of ER⁻ but not ER⁺ breast cancer (48). This gene encodes 17HSD1, which affects the conversion of estrone to estradiol. The increase in risk for triple-negative invasive breast cancer suggests that early high-dose OC initiation may increase tumor aggressiveness.

We previously reported that multiparity and early age at first birth were associated with reduced relative risk of ER⁺/PR⁺ but not of ER⁻/PR⁻ tumors, whereas duration of breast-feeding was associated with lower relative risk of both receptor-positive and receptor-negative breast cancer among all cases [who had ER/PR status from Surveillance, Epidemiology, and End Results (SEER)] and control participants of the Women's CARE Study (49). Consistent with our previous findings, in the current analysis, we found that parity was inversely associated with risk for luminal subtypes. Age at first full-term pregnancy was positively associated with luminal A tumors among older women (ages 45–64 years). Neither parity nor age at first full-term pregnancy was associated with triple-negative tumors. These findings, and particularly the inverse association of parity with all subtypes except triple-negative breast cancer, suggest that some hormone-related risk factor profiles differ between triple-negative and other breast cancer subtypes. In addition, the lack of heterogeneity in the effect of breast-feeding in our data suggests that there are different pathways involving the association between breast-feeding and the risk of breast cancer, such as paracrine signals produced by ER⁺ progenitors described previously (47).

Consistent with our findings, a pooled analysis of two population-based case-control studies conducted in Western Washington State found that nulliparity was marginally associated with an increased risk of ER⁺ breast cancer but was not associated with triple-negative breast cancer, whereas breast-feeding protected against both ER⁺ and triple-negative subtypes (9). Our results are also consistent with those from a population-based case-control study conducted in Poland, which found that parity was inversely associated with the luminal A subtype but not with the basal-like subtype (triple-negative plus the expression of HER1 and/or cytokeratin 5; ref. 11). Our results are not consistent with those from a population-based case-control study conducted in North Carolina in which parity and early age at first full-term pregnancy were associated with an increased risk of basal-like (ER⁻, PR⁻, HER2⁻, HER1⁺, and/or cytokeratin 5/6⁺) breast cancer and long duration of breast-feeding was associated with reduced risk of basal-like but not luminal A breast cancer (16). These inconsistencies could be due to some important differences between the North Carolina study and ours, such as the participants' age distribution (ranges: 20–74 versus 35–64 years), the breast cancer case patients' stage (in situ and invasive versus only invasive), control subjects' participant rates (57% versus 74%), analysis approach (each subtype separately compared with control group using unconditional logistic regression models versus all subtypes simultaneously compared with control group using unconditional polychotomous logistic regression models), and different reference groups when testing the effects of age at first full-term pregnancy and breast-feeding (reference group including nulliparous women versus analysis restricted to parous women).

To our knowledge, this study is the first to examine whether p53 status modifies the association of OC use and reproductive factors with risk of invasive triple-negative or luminal A breast cancer. p53 gene mutations may cause loss of tumor suppressor function and gain of oncogenic activity (21). Approximately 69% of p53 immunohistochemistry–positive cancers have p53 gene mutations (41). Whereas 92% of missense p53 gene mutations are associated with p53 immunohistochemistry positivity, only 45% of p53 nonsense or other gene mutations are associated with immunohistochemistry positivity (41). Because the vast majority of p53 gene mutations are missense mutations, p53 overexpression and p53 gene mutation are highly correlated. We speculated that this would result in a stronger association between hormone-related risk factors and the risk of breast cancer with p53 overexpression because hormones drive cell proliferation and, therefore, increase the probability for the accumulation of random genetic errors (50). However, our data did not show any differences in associations of OC use or reproductive factors when triple-negative tumors and luminal A tumors were further classified by p53 overexpression status. We did observe that the number of full-term pregnancies was statistically significantly associated with luminal A tumors without p53 overexpression and that duration of breast-feeding was statistically significantly associated with both triple-negative and luminal A tumors without p53 overexpression. These findings may have resulted from the larger sample sizes for tumor subtypes lacking p53 overexpression than subtypes that overexpress p53.

Where p53 has previously been examined alone, reproductive factor associations did not differ by p53 status (22–24), which is consistent with our results. One previous study showed that longer duration of OC use was more strongly associated with an increase in the risk of p53⁺ than p53⁻ breast cancer among women at ages 20 to 54 years (24). Another study found that this difference in the association with OC use between p53⁻ and p53⁺ tumors existed among younger (age <45 years) but not older women (22). In a third study, OC use was not associated with either p53⁺ or p53⁻ tumors (23).

This study has several limitations. We obtained paraffin-embedded tissue for 80% of the samples requested; however, we did not request tissue for all eligible case patients because of funding constraints. We compared demographic characteristics, family history of breast cancer, aspects of OC use, reproductive factors, tumor size, and tumor stage between our eligible case patients with and without known ER, PR, HER2, and p53 status from the centralized laboratory (results not shown). Case patients with information on these four biomarkers were more likely to have been diagnosed in LA than in Detroit, to have been better educated, and to have had larger tumors than those without this information. Parous case patients with known ER, PR, HER2, and p53 status had, on average, breast-fed 1.4 months longer, but no statistically significant differences were detected for other factors examined. The difference in duration of breast-feeding could lead us to underestimate any protective effect of breast-feeding on overall breast cancer risk in the patients with biomarker results, but it is unlikely that this bias would differ across tumor subtypes. Further, although previous research shows that p53 expression and p53 mutation status determined by fluorescent in situ hybridization analysis are strongly correlated, assessment of p53 expression by immunohistochemistry may misclassify some tumors.

In conclusion, our results support the contention that some hormone-related risk factor profiles differ between triple-negative and other breast cancer subtypes defined by ER, PR, and HER2 status, but no further consistent differences are noted when p53 is also considered.

No potential conflicts of interest were disclosed.

We thank Dr. Karen Petrosyan, Armine Arakelyan, Hasmik Toumaian, and Judith Udove for technical assistance in the performance of the immunohistochemical assays for this study and the collaborators who contributed to the development and conduct of the Women's CARE Study but who did not directly contribute to the current study: Janet Daling, Dennis Deapen, Jonathan Liff, Sandra Norman, and Phyllis Wingo.

Grant Support: National Institute for Child Health and Human Development grant NO1-HD-3-3175 and National Cancer Institute grant CA48780. Data collection for the Women's CARE Study was supported by the National Institute of Child Health and Human Development and National Cancer Institute, NIH, through contracts with Emory University (N01-HD-3-3168), Fred Hutchinson Cancer Research Center (N01-HD-2-3166), Karmanos Cancer Institute at Wayne State University (N01-HD-3-3174), University of Pennsylvania (NO1-HD-3-3276), and University of Southern California (N01-HD-3-3175) and Interagency Agreement with Centers for Disease Control and Prevention (Y01-HD-7022). Collection of cancer incidence data in LA County by University of Southern California was supported by California Department of Health Services as part of statewide cancer reporting program mandated by California Health and Safety Code, Section 103885. Support for use of SEER cancer registries through contracts N01-PC-67006 (Atlanta), N01-CN-65064 (Detroit), N01-PC-67010 (LA), and N01-CN-0532 (Seattle).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.