HER2 status and disparities in luminal breast cancers

Abstract National Comprehensive Care Network guidelines for adjuvant treatment of invasive breast cancer are based on HER2 and hormone receptor (HR) status, where HR+ disease encompasses all estrogen receptor (ER)+ and/or progesterone receptor (PR)+ tumors. We sought to explore clinical and demographic differences among patients with HR+ breast cancer subtypes, and the role of HER2 status, age, race/ethnicity, and socioeconomic status (SES) in disease risk. We evaluated breast cancer subtype distribution, defined by HR and HER2 status, using patient clinical, demographic, and socioeconomic characteristics. Differences in HR categories by demographic and tumor characteristics were examined using chi‐squared tests. Multinomial logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (CI) to quantify associations between breast cancer HR status and demographic factors. We found that differences in HR+ (ER−/PR+ vs. ER+/PR− or ER+/PR+) tumor biology are likely clinically significant and may play a role in breast cancer, regardless of HER2 status. While clinical and patient characteristics differed within each luminal subtype, we found disparities in SES only among Luminal A (HR+/HER2−) tumors. Among HR+/HER2− cases, we observed that ER−/PR+ patients tend to live in areas of higher poverty (OR = 1.20, 95% CI = 1.03–1.40) and are 70% more likely to be aged 50 years or older. However, this pattern was not found in women with Luminal B (HR+/HER2+) disease (Poverty OR = 0.98, 95% CI = 0.76–1.27; Age OR = 1.01, 95% CI = 0.81–1.26). Racial/ethnic disparities among non‐Hispanic black and Hispanic women persisted across HR+/HER2− cases compared to non‐Hispanic white women. Our findings suggest that while race/ethnicity and SES are correlated, each plays an independent role in contributing to disease among Luminal A tumors. Further study is needed to investigate how tumor biology, race/ethnicity, and socioeconomic disparities among HR+/HER2− cases may contribute to poorer patient prognosis.


Introduction
Breast cancer is the most common cancer diagnosed in women in the United States, accounting for 29% of newly diagnosed female cancers. An estimated 246,660 new cases and 40,450 deaths are anticipated from the disease in 2016 [1]. Breast cancer most commonly arises from the mammary ductal epithelium, and its systemic treatment is guided by its molecular characteristics. Specifically, National Comprehensive Care Network (NCCN) guidelines for the treatment of invasive breast cancer outline systemic adjuvant therapies based on hormone receptor (HR) and human epidermal growth factor receptor 2 (HER2) statuses, where positive HR status is defined as expression of the estrogen (ER) and/or progesterone (PR) receptors [2]. Clinically, tumor cells are evaluated for these biological markers to approximate cancer molecular subtype based on expression profiling (Luminal A/B, HER2−enriched, basal-like).
Luminal breast cancers account for about 60% of all cases. They are HR-positive (HR+) and can be further classified based on HER2 receptor status [3]. Luminal A breast cancers are HER2−negative (HR+/HER2−) and include ER+/PR+, ER+/PR−, and ER−/PR+ status. Their adjuvant treatment includes endocrine therapy with or without multimodality chemotherapy, based on tumor size, lymph node status, and, more recently, the 21-gene recurrence score [2]. In contrast, Luminal B tumors tend to be more aggressive, demonstrate HER2−enrichment (HR+/HER2+), and encompass ER+/PR+, ER+/PR−, and ER−/PR+ cases. Recommended treatment for Luminal B tumors includes anthracycline-based trastuzumabcontaining multimodality chemotherapy, followed by a 1-year course of trastuzumab and 5 years of endocrine therapy [2]. Together, luminal breast cancer subtypes are associated with the best short-term prognoses for patients, attributable to favorable responses to hormonal therapy [4,5].
Across breast cancer subtypes, there exist demographic and socioeconomic status (SES) differences [6]. Previous reports have found that differences in SES may reflect underlying differences in exposures to known breast cancer risk factors, as women with higher SES tended to have lower parity and more frequent use of exogenous hormones [7][8][9]. Additionally, SES has been found to interact with race/ethnicity among female breast cancers [10]. SES has previously been associated with negative HR status (HER2−enriched and basal-like subtypes) in breast cancer [7,11,12]. SES disparities have also persisted in HR+ subtypes (HR+/HER2− and HR+/HER2+) [13]. We therefore sought to explore further clinical and demographic differences in order to identify disparities among female patients with luminal (HR+) breast cancer subtypes using Surveillance, Epidemiology, and End Results (SEER) data by assessing estrogen and progesterone receptor status, and the role of SES in risk of disease.

Data sources and case selection
Data were obtained from the National Cancer Institute's SEER program. The SEER program collects cancer incidence and mortality data from 20 population-based cancer registries covering approximately 28% of the US population [14]. SEER*Stat is a free program provided by SEER to access and analyze information in the publicly available database. A case listing session in SEER*Stat was run on the SEER 18 incidence dataset to obtain demographic, tumor characteristics, and socioeconomic information on breast cancers [15]. The SEER 18 incidence dataset includes information from the following registries: Alaska Native Tumor Registry, Arizona Indians, Cherokee Nation, Connecticut, Detroit, Atlanta, Greater Georgia, Rural Georgia, San Francisco-Oakland, San Jose-Monterey, Greater California, Hawaii, Iowa, Kentucky, Los Angeles, Louisiana, New Jersey, New Mexico, Seattle-Puget Sound, and Utah. The dataset used is publicly available and was exempt from human subjects review. Therefore, an IRB protocol was not required for this study.
The variables of interest included age at diagnosis (<50, 50-64, 65-74, 75+ years), race/ethnicity (non-Hispanic (NH) white, NH black, NH Asian, or Pacific Islander, NH American Indian, Hispanic), American Joint Commission of Cancer (AJCC) clinical stage, and tumor size (≤0.5 cm, >0.5 cm). An approximation of SES was evaluated using contextual measures of area-based poverty. The percentage of persons and families whose incomes were below 200% of the poverty level were calculated in SEER using county attribute data from the US Census Bureau's 2008-2012 American Community Survey. We categorized our cohort of breast cancer cases into quartiles based on the percentage of persons with incomes below 200% of the poverty level distribution within the dataset. Quartile 1 (Q1) (<24%) represented the fourth of the cohort that reside in areas where less than 24% of residents are from low-income category. We defined quartile 2 (Q2) (24% ≤ 31% low income), quartile 3 (Q3) (31% ≤ 39%), and quartile 4 (Q4) (>39%), such that Q4 contained the fourth of our cohort that reside in areas with the highest proportion of low-income residents. Follow-up for each case was current within 22 months of the annual submission date (1 November 2014).

Statistical analysis
Pair-wise comparisons of hormone receptor (HR) categories by demographic and tumor characteristics were examined using chi-squared tests. Multinomial logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (CI) to quantify associations between breast cancer hormone receptor status and various demographic factors. Analyses were stratified by HER2 status and the reference outcome category was ER+/PR+ disease. Demographics variables assessed, included: age at diagnosis (<50, 50-64, 65-74, 75+), race/ethnicity (NH white, NH black, NH Asian or Pacific Islander, NH American Indian, Hispanic), and area-based poverty level (quartiles). These factors were adjusted for age, race/ethnicity, and poverty level based on patients having complete information for each of these covariables. All data were analyzed using SAS version 9.4 statistical software (SAS Institute; Cary, NC). All statistical tests were two-sided. A P < 0.05 was considered to be statistically significant.
An additional 19,020 breast cancer patients had borderline or unknown status for ER, PR, or HER2, and were not included in this analysis. Among these women, 67.7% were NH white (12,876 cases), 11.7% were NH black (2227 cases), 10.9% were Hispanic (2068 cases), 7.4% were Asian or Pacific Islander (1404 cases), 0.5% were American Indian/ Alaska Native (91 cases), and 1.9% (354 cases) had unknown race/ethnicity (data not shown). Eighty three percent of unknown cases were diagnosed with invasive breast cancer at age 50 years and older (15,793 cases). 61.5% of cases were diagnosed with early stage (AJCC stage 0-II) disease (11,701 cases), 17.8% were diagnosed with later stage (AJCC stage III-IV) cancer, and 20.1% (3929 cases) did not have information on AJCC stage. 59.1% of cases (11,242 cases) resided in areas where at least 31% of residents were living in poverty (Q3 or Q4). Together, cases with unknown receptor status tended to be NH white, older, and diagnosed with early-stage disease (data not shown).
Overall, we observed that, regardless of HER2 status, ER−/PR+ cases were more likely to be diagnosed in young patients (age <50 years) and to present with later stage (stage III-IV) disease, but were less likely to be NH white, as compared to ER+/PR− or ER+/PR+ patients.

Socioeconomic disparities in Luminal A breast cancers
We also noted significant differences in area-based poverty among patients with Luminal A (HR+/HER2−) tumors. To further explore the relationship of SES and HR+ status (ER+/PR+, ER+/PR−, or ER−/PR+) within each breast cancer subtype, we used multinomial logistic regression models. Because race and age are associated with SES, we adjusted for these variables to determine whether socioeconomic disparities persisted in this population. Table 3 summarizes results from models adjusted for area-based poverty, age, and race/ethnicity. Using ER+/ PR+ tumors as the referent outcome in each subtype and poverty quartile Q1 as the referent covariable, we found that women with Luminal A breast cancer who live in counties with higher poverty were more likely to be diagnosed with ER−/PR+ disease (Q4: OR = 1.20, 95% CI = 1.03-1.40) ( Table 3). Women diagnosed with ER−/ PR+ Luminal A disease were 1.7-fold more likely to be under 50 years of age compared to women with ER+/ PR+ tumors in our area-based poverty-adjusted models. In addition, NH blacks were at an increased risk of being  (Table 3).

Discussion
In this study, we explored associations among luminal breast cancers and SES as assessed by age, race/ethnicity, and a measurement of county-level poverty and found that within Luminal A (HR+/HER2−) and Luminal B (HR+/HER2+) cancers, clinical and demographic characteristics varied. Clinical differences among luminal breast cancers can be attributed to the opposing effects of estrogen and progesterone on tumor progression. Estrogen supports tumor growth but suppresses progression, whereas, progesterone supports tumor progression and is associated with more aggressive disease [16,17]. In the absence of estrogen signaling (ER− tumors), high progesterone levels in women have been shown to support tumor progression without opposition from estrogen [18,19]. Consistent with these findings, we observed in this study that regardless of HER2 status, women with ER−/PR+ tumors were more likely to present with later stage (stage III-IV) disease compared to ER+/PR+ or ER+/PR− cases. These results suggest that differences in HR+ (ER−/PR+ vs. ER+/PR− or ER+/PR+) tumor biology are likely to be clinically significant and play a role in breast cancer disease, regardless of HER2 status.
Demographic characteristics of patients, including age, also varied within each luminal breast cancer subtype. Progesterone levels are higher in premenopausal women, typically those diagnosed with breast cancer under the age of 50 years, compared to postmenopausal women  [20,21]. Our study showed that women under the age of 50 were at an increased risk of developing ER−/PR+ Luminal A disease, while women over the age of 60 were at a decreased risk compared to ER+/ PR+ disease. This observation is consistent with reports that high progesterone levels (occurring only in the luteal phase and in pregnancy) will induce breast cancer cell invasiveness and metastasis in the absence of estrogen or the ER. In contrast, age was not associated with increased risk of ER−/PR+ disease among Luminal B cases. This observation may be explained by findings using experimental models, that overexpression of HER2 supports aggressive tumor growth in luminal breast cancers [22].
Racial and ethnic differences were also noted among Luminal A cancers, as NH black women were most likely to develop ER−/PR+ disease. Indeed, previous studies have demonstrated that among Luminal A tumors, raceassociated biological factors contribute to poorer outcomes in black women compared to NH white women [23,24]. Among women with Luminal B tumors, NH Asian or Pacific Islander or American Indian/Alaska Native individuals were at an increased risk of developing ER−/PR+ tumors. However, caution should be taken in interpreting results in these racial/ethnic categories due to small sample size. Earlier findings have observed that Hispanic white women are more likely to present with more aggressive tumors and be diagnosed with ER+/PR− disease [25]. In our study, we confirmed that Hispanic individuals were at an increased risk of developing ER+/PR− or ER−/PR+ disease compared to ER+/PR+ status in both Luminal A and B breast cancer subtypes after adjusting for age, race, and area-based poverty.  While clinical and patient characteristics differed within each luminal breast cancer subtype, we found disparities in SES persist only among Luminal A tumors. Luminal A breast cancers are associated with the most favorable short-term prognosis due to favorable responses to endocrine therapy [4,5]. However, assessment of long-term prognosis demonstrates similar or worse overall survival for Luminal A cases as compared to other subtypes [26]. In our study, among Luminal A cases, we observed that ER−/PR+ disease was associated with residing in areas of higher poverty even after adjusting for age and race/ethnicity. Caution should be exercised in the interpretation for results of ER−/PR+ cases due to small sample size. The relationship between SES and development of Luminal A breast cancers demonstrates that while race and SES are correlated, each plays an independent role in contributing to disease among Luminal A tumors. These disparities were not observed among Luminal B tumors, suggesting that HER2 status may be associated with risk factors that affect the SES of patients.
Our study is the first to assess disparities among HRpositive (HR+) breast cancer subtypes in the context of HER2 status using SEER patient data. Previous studies have investigated the role of HR+ status without HER2 information, or have analyzed the role by breast cancer subtype. For example, a recent study by Parise et al. that used the California Cancer Registry found that SES moderately altered racial disparities and risk of mortality in particular breast cancer subtypes [27]. Similar results have been observed among other tumor sites [10]. While our findings suggest differences within Luminal A or B breast cancer subtypes, available data for HER2 are limited to 3 years of diagnosis, which does not allow further analysis for SES disparities in mortality at this time. The use of data from the population-based SEER program is a strength of this study, as it allows for the inclusion of a considerable number of pathologically verified cases making our results more generalizable to the larger United States population. Another strength is that these data are of high quality and database entries are standardized and continuously monitored for accuracy. However, use of SEER data does have limitations. Family history, lifestylerelated factors (e.g., obesity, reproductive factors, and environmental exposures), modality of diagnosis, and chemotherapy data, are not available for study. We were also unable to examine patients whose ER, PR, or HER2 status was unknown.
Scientific, clinical, and public health implications can be inferred from this study. First, our findings are consistent with preclinical and clinical data regarding the opposing effects of the estrogen and progesterone receptors in breast cancer growth and progression. Further research