Introduction

Hispanic women may be prone to developing breast cancer molecular subtypes [defined by estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2)] associated with poor prognosis, such as ER−/PR−, and triple negative (ER−/PR−/HER2−) breast cancers [1–3]. Studies among Hispanics point to a 20–40 % increased risk of developing triple negative and ER−/PR− breast cancers, compared with non-Hispanic white women [2, 3]. Women diagnosed with ER−/PR− or triple negative tumors have significantly greater risk of death than more favorable prognosis subtypes (i.e., ER+/PR+) [4, 5]. Thus, defining the population-based distribution of breast cancer subtypes among Hispanic women provides information of clinical, prognostic, and therapeutic value.

However, only two studies have examined the patterns of breast cancer subtypes among Hispanic women [6, 7]. Hines and colleagues assessed breast tumors of 69 Hispanic women [6], and found that the four most prevalent subtypes were ER+/PR+/HER2− (41 %), triple negative (17 %), ER−/PR−/HER2+ (15 %), and ER+/PR+/HER2+ (13 %). Ortiz et al. [7] showed a similar subtype distribution among 663 women with breast cancer in Puerto Rico, and reported a higher risk of death for triple negative and ER+/PR+/HER2+ tumors, compared with the ER+/PR+/HER2− subtype. To expand on this work, data from the California Cancer Registry (CCR) on 16,380 Hispanic women diagnosed with invasive breast cancer were used to characterize the molecular subtypes defined by joint hormone receptor (HR) and HER2 status. Our objective was to conduct a large, population-based assessment of the distribution and survival by breast cancer subtype and examine associations with demographic and clinical attributes.

Methods

Study population

We obtained data from the CCR on all Hispanic female California residents aged 25 years and older, diagnosed with a first primary invasive breast cancer between 2005 and 2010. Patient sociodemographic information included age at diagnosis, race/ethnicity, birthplace, insurance status, marital status, and residential address at diagnosis. Race/ethnicity and birthplace data are abstracted from medical records or death certificates [8]. The North American Association of Central Cancer Registries Hispanic Identification Algorithm (NHIA) was used to improve the classification of Hispanic ethnicity [9]. Patient clinical information included the American Joint Committee on Cancer (AJCC) stage, tumor grade, tumor size, lymph node involvement, tumor metastasis, and first course of treatment (surgery, chemotherapy, and radiation).

Nativity

Patient nativity was classified as previously described [10], based on (1) cancer registry-based data from medical records and/or death certificates and (2) imputation using the first five digits of the patient’s social security number (SSN), for those with unknown birthplace (34.5 %). SSN digits are linked to the state and year of issuance, from which nativity was imputed as follows: women who received their SSN before age 21 years were considered as United States (US)-born, whereas those who received their SSN on or after age 21 years as foreign-born. The age threshold was determined and validated based on a prior cohort of Hispanic cancer patients [11].

Neighborhood socioeconomic status (SES) and Hispanic enclave

Each patient’s residential address at diagnosis was geocoded to a census block group. Participants with incomplete residential address information (7.3 %) were assigned an SES value based on their county of residence. Neighborhood SES was determined based on an index that incorporates 2000 Census (for cases diagnosed in 2005) and 2006–2010 American Community Survey data (for cases diagnosed after 2005) on education, occupation, unemployment, household income, poverty, rent, and house values [22]. Hispanic enclave was based on 2000 Census variables (% linguistically isolated,  % linguistically isolated who speak Spanish, % speaking limited English, % speaking limited English who spoke Spanish, % recent immigrants, % Hispanic, and % foreign-born), developed via principal components analysis. Participants were assigned to a neighborhood SES quintile and Hispanic enclave quintile based on the distribution of each variable across California block groups.

Breast cancer subtype definition

A detailed description of the methods used for classification of breast cancer subtypes has been published elsewhere [12]. Based on joint tumor expression of ER, PR, and HER2, as described in the pathology record, breast cancers were classified into four distinct subtype categories: HR+/HER2− was defined as ER+ or PR+ and HER2+; HR+/HER2+ as ER+ or PR+ and HER2+; HR−/HER2+ as ER− and PR− and HER2+; and triple negative as ER−, PR−, and HER2− [4, 13–17]. Participants missing the tumor marker information needed to assign to a subtype were excluded [n = 3,253 (16.6 %)].

Statistical analysis

Associations between breast cancer subtypes and patient attributes

Adjusted polytomous regression models were used to estimate odds ratios (ORs) and 95 % confidence intervals (CIs) by sociodemographic and clinical attributes for triple negative, HR+/HER2+ and HR−/HER2+ versus HR+/HER2- subtypes. Tests for trend were considered statistically significant at p ≤ 0.05.

Kaplan–Meier survival curves depicting time from diagnosis to death were calculated for triple negative, HR+/HER2+ and HR−/HER2+ versus HR+/HER2− subtypes. Surviving participants were censored at the time of last known follow-up. Wilcoxon rank order test was used to test homogeneity of survival by subtype (SAS Institute v 9.3, Cary, NC).

Adjusted Cox proportional hazard modeling was performed to estimate the risk of death from breast cancer for all subtypes adjusted for patient sociodemographic and clinical characteristics. Survival time, in months, was defined as time from diagnosis to whichever of the following occurred first: death from breast cancer, last known contact, death due to other causes, or end of study follow-up (December 31, 2010). We tested the proportionality assumption using scaled Schoenfeld residuals. AJCC stage levels were included in the models as a stratifying variable, allowing the underlying hazard function to vary by stage.

Results

HR+/HER2− was the most common subtype (62.6 %), followed by triple negative (15.6 %), HR+/HER2+ (13.8 %) and HR−/HER2+ (8.1 %) (Table 1) . Over 50 % of breast cancers were diagnosed among foreign-born Hispanics, overall and within each subtype, compared with US-born Hispanics. Women in the lowest SES and highest Hispanic enclave neighborhoods comprised the greatest proportions of participants among all breast cancer subtypes.

Table 1 Demographic and tumor characteristics of the breast cancer subtypes in Hispanic women with invasive breast cancer, California, 2005–2010
Full size table

Association between breast cancer subtypes and patient attributes

Foreign-born Hispanic women were significantly more likely than US-born Hispanic women to be diagnosed with HR−/HER2+ versus HR+/HER2− breast cancer (OR = 1.17, 95 % CI 1.02–1.35); other subtypes did not differ by nativity (Table 2). Compared to women living in the highest SES neighborhoods, those in lower SES neighborhoods had a 1.32–1.42 fold greater risk of triple negative (p < 0.05) and 1.17–1.43 fold greater risk of HR−/HER2+ relative to HR+/HER2− breast cancer (p < 0.05). Hispanic women aged 45–49 years had a lower risk of the triple negative (OR = 0.72, 95 % CI 0.60–0.87) and HR-/HER2+ (OR = 0.59, 95 % CI 0.47–0.75) subtypes, compared with women aged 50–54 years.

Table 2 Adjusted odds ratios of sociodemographic and tumor characteristics associated with breast tumor subtypes (vs. HR+/HER2−) in Hispanic Women, California, 2005–2010
Full size table

Breast cancer-specific survival and mortality by breast cancer subtype

Breast cancer-specific survival significantly differed between tumor subtypes (p < 0.0001; Fig. 1). Over the approximately 5.5 years of follow-up, Hispanic women with triple negative and HR−/HER2+ breast cancer had the lowest probability of survival. After multivariate adjustment (Table 3), Hispanic women diagnosed with triple negative breast cancer were four times more likely to die from the disease compared to those diagnosed with the HR+/HER2− subtype (HR = 4.05, 95 % CI 3.35–4.90). Overall, the risk of death from breast cancer followed a step-wise pattern by neighborhood SES, such that the risk of death from breast cancer increased as a woman’s neighborhood SES level decreased; compared to quintile 5 (highest SES), HRs were 1.38 (95 %CI 0.98–1.94) for quintile 4 and 1.76 (95 % CI 1.25–2.49) for quintile 1.

Fig. 1
figure 1

Kaplan–Meier curve of Hispanic women by breast cancer subtype, California 2005–2010

Full size image
Table 3 Breast cancer-specific mortality, by tumor subtype, in Hispanic Women in California, 2005–2010
Full size table

For US-born Hispanic women, the risk of breast cancer-specific mortality was significantly greater for all subtypes compared with HR+/HER2− tumors (Table 4), whereas among foreign-born Hispanic women, only those diagnosed with triple negative and HR−/HER2+ subtypes had significantly increased risk of death from breast cancer.

Table 4 Breast cancer-specific mortality by Hispanic nativity
Full size table

Discussion

Our study is among the first, certainly the most comprehensive with 16,380 participants, population-based analyses of breast cancer subtypes among US Hispanic women. These results confirm and extend emerging patterns for the molecular breast cancer subtypes. Similar to other racial/ethnic groups [13, 18–20], residence in a low SES neighborhood was significantly associated with an increased risk of diagnosis and dying from HR−/HER2+ and triple negative breast cancers. Foreign-born Hispanics were at greater risk of diagnosis with HR−/HER2+, compared with US-born Hispanics. The risk of death from triple negative breast cancer was substantial, as both US- and foreign-born Hispanic women diagnosed with this subtype had an approximately fourfold greater risk of death than those with HR+/HER2− breast cancer.

Although our case-only analyses cannot speak directly to cancer etiology or risk, these data underscore the potential impact of SES, a social determinant of health, on risk factors that may be etiologically important in increasing women’s risk of developing poor prognostic breast cancer subtypes. For instance, women of low SES status, or residing in low SES neighborhoods, may have lower access to and consume fewer healthy foods (i.e., vegetables and nutrient-rich foods), fewer opportunities to engage in physical activity, and higher levels of obesity [21, 22], as indicated by studies, are associated with increased risk of ER− [23–25] or triple negative breast cancers [26]. Moreover, low SES may be related to younger age at first birth and lack of breastfeeding [27, 28], factors associated with an elevated risk of triple negative or basal-like breast cancers [29–31].

Importantly, if socioeconomic differences play a role in the risk of developing certain tumor subtypes, then the breast cancer-specific mortality disadvantage observed among Hispanic and other racial/ethnic women may be partly due to the intrinsic aggressiveness of the tumor subtype, and the more commonly recognized factors associated with low SES or living in low SES neighborhoods [32]. Consistent with studies among other racial/ethnic groups [4, 5, 33, 34], our study confirms the adverse characteristics of certain breast cancers, such that compared with the HR+/HER2− subtype, Hispanic women diagnosed with HR+/HER2+, HR−/HER2+, and triple negative subtypes had significantly greater risk of death.

While our study adds valuable information on breast cancer subtypes among Hispanic women, it has some limitations. Tumor subtype information was missing for approximately 16.6 % of potentially eligible participants, although the magnitude and direction of biases related to missing data are unknown. We did not have individual-level data on income or education; nevertheless, neighborhood-level measures may capture information on environmental factors that occur as a result of its socioeconomic condition and deprivation [20]. The distribution of breast cancer subtypes among Hispanic women in California may not reflect distributions among other US Hispanic women; accordingly, our findings may be most generalizable to women of primarily Mexican descent, the largest Hispanic subgroup in California [35]. The imputation method used to classify nativity is subject to error and may have led to some misclassification; however, most participants (about 65 %) were classified based on cancer registry birthplace data, previously shown to have high accuracy [11, 36]. Similarly, Hispanic ethnicity may be subject to misclassification [37, 38], although classification should have improved with application of a registry-wide algorithm [9]. While Hispanics are a diverse population, we could not include information on Hispanic origin because it was missing for approximately 40 % of participants. Breast cancer-specific survival analyses are subject to the accuracy of the underlying cause of death code, which has been shown to be 84–90 % accurate [39, 40]. Finally, those tumor characteristics associated with aggressive breast cancer subtypes, including HR- and younger age at diagnosis, are also associated with interval-detected cancers, which may lead to length bias and lower survival than screen-derived cancers [41, 42].

Conclusion

In the largest population-based study of breast cancer subtypes among US Hispanic women conducted to date, we found that residing in a low SES neighborhood was significantly associated with increased risk of developing and dying from HR− breast cancers. Foreign-born women were at greater risk of HR−/HER2+ tumors, although nativity was not associated with other subtypes. Similarity in the distribution and prognosis of breast cancer subtypes among California Hispanic women and previous studies in other racial/ethnic groups underscores the need to understand how sociodemographic factors, such as SES, contribute to the distinct patterns of breast cancer subtype incidence and mortality among women of all racial/ethnic backgrounds.