Racial Disparities in Breast Cancer Mortality in a Multi-Ethnic Cohort in the Southeast

Abstract

Background

Although much has been done to examine those factors associated with higher mortality among African American women, there is a paucity of literature which examines disparities among rural African Americans in South Carolina. The purpose of this investigation was to examine the association of race and mortality among BrCA patients in a large cohort residing in South Carolina for which treatment regimens are standardized for all patients.

Methods

Subjects included 1209 women diagnosed with BrCA between 2000–2002 at a large, local hospital containing a comprehensive breast center. Kaplan Meier survival curves were calculated to determine survival rates among AA and EA women, stratified by disease stage or other prognostic characteristics. Adjusting for various characteristics, Cox multivariable survival models were used to estimate the hazard ratio (HR)

Results

The 5-year overall all-cause mortality survival proportion was ~78% for AA women and ~89% for EA women, p<0.01. In analyses of sub-populations of women with identical disease characteristics, AA women had significantly higher mortality than EA women for the same type of breast cancer disease. In multivariable models, AA women had significantly higher mortality than EA women for both BrCA specific death (HR = 2.41; 1.21–4.79) and all-cause mortality (HR = 1.42; 1.06–1.89).

Conclusion

AA women residing in rural South Carolina had lower survival for breast cancer even after adjustment for disease-related prognostic characteristics.

Impact

These findings support health interventions among AA BrCA patients aimed at tertiary prevention strategies or further down-staging of disease at diagnosis.

Introduction

Breast cancer (BrCA) is the most commonly diagnosed cancer among women and ranks second as a cause of death from cancer. The American Cancer Society estimates that 254,650 women were diagnosed with BrCA and 41,170 women died of BrCA in 2009.¹ Unfortunately, the impact of disease is not equally distributed. It is a well-documented fact that African-American (AA) women have significantly poorer BrCA survival compared with their European-American (EA) counterparts.² Surveillance statistics show that AA women have an overall (10%) lower incidence than do EA women, yet AA women are more likely to die of invasive BrCA than are EA women (33.5 and 24.4 per 100,000 women, respectively).³ Additionally, trends in BrCA incidence and mortality over time evince patterns that vary markedly by ethnicity. From 1975 to 2006, Surveillance, Epidemiology, & End Results (SEER) data indicate that EA women had a 16% increase in BrCA incidence and a 28% decrease in mortality.⁴ While AA women experienced a 29% increase in incidence of BrCA, they had a 6% increase in mortality during that time.⁴

South Carolina has some of the largest health disparities in the nation, and the most dramatic of these are associated with elevated cancer mortality rates among AA.⁵ Statewide, BrCA incidence rates from 1996–2001 have remained stable with a somewhat higher age-adjusted incidence among EA women compared with AA women (129.8/100,000 and 111.6/100,000 respectively).⁶ However, in South Carolina, AA women have a 47% higher risk of death from BrCA compared with EA women, (30.6 deaths/100,000/year compared with 20.8 deaths/100,000/year).⁶

Plausible reasons for these existing disparities range from socioeconomic factors (e.g., including access to care) to biological processes.^7–16 Research conducted in South Carolina has shown that AA women typically present with aggressive tumor types more typical of those found in younger EA women.⁸ By contrast, other investigations have shown no mortality difference among AA and EA women participating in South Carolina’s National Breast and Cervical Cancer Early Detection Program, which provides free mammograms for women meeting income eligibility requirements.¹⁷ This suggests that some of the observed mortality disparity may be explained by access to mammography screening care. In all likelihood, the root of these BrCA disparities is multi-faceted requiring intervention on a variety of levels.

Although much has been done to identify factors associated with racial disparities, there is a paucity of literature examining BrCA disparities among rural AAs in South Carolina. Additionally, little has been done to examine the influence of disease characteristics on these disparities. Consequently, this investigation was undertaken to examine the association of race and mortality among BrCA patients in a large cohort residing in South Carolina.

Methods

All data utilized for this analysis were collected as a part of a local hospital tumor registry system. This hospital is the largest health care provider in the Midlands of South Carolina, an area that includes about 500,000 people in the two counties surrounding Columbia (the state’s capitol) and over 900,000 in the wider Standardized Metropolitan Statistical Area that roughly corresponds to its catchment area.¹⁸ The hospital maintains a comprehensive breast center with the aim of standardizing detection, diagnosis, and treatment of all breast cancer cases. Thus we were able, in part, to account for treatment differences that might reasonably be expected to influence outcome.

All data collected by the hospital tumor registry are ultimately reported to the South Carolina Central Cancer Registry, which consistently maintains a “gold-certified” rating through the National Association of American Cancer Registries (NAACR), indicating data of exceptionally high quality, validity, and completeness. Due to the fact that all data for this analysis had been previously collected for reporting purposes and were deidentified prior to analyses, this investigation was granted exemption from Institutional Review Board review.

Study Population

The study population consisted of women diagnosed with a histopathologically confirmed, first primary breast neoplasm at a large, local SC hospital between 2000 and 2002. In cases of multiple observations per patient (multiple tumors diagnosed at the same time), only the tumor with the highest stage was retained for analysis. Data used for this investigation were derived only from those women with a race designation of either “Black” or “White”. Other races were excluded because of low frequency count; there were only 8 women that were identified as “Hispanic” and 43 identified as “Other”. A total of 1,209 BrCA cases were extracted from the registry.

Covariates

Using the American Joint Committee on Cancer (AJCC) 5^th Edition criteria, stage of cancer for each patient was collapsed to create 5 mutually exclusive categories of Stage 0, Stage I, Stage II (included A and B), Stage III (included A and B), and Stage IV. Elston score on a scale of 3 to 9 was determined by the pathologist based upon tumor architecture, mitotic activity, and nuclear pleomorphism. This scoring was then collapsed for analysis into 3 groups: low (Elston score 3–5), moderate (Elston score 6–7), and high (Elston score 8–9). Estrogen receptor (ER) status was categorized as positive or negative. Human epidermal growth factor receptor 2 (Her2) status was classified as positive, negative, or borderline according to the ImmunoHistoChemistry (IHC) test. Insurance coverage was coded as public, private, not insured, and unknown. Public insurance coverage includes Medicaid and Medicare.

Statistical Analysis

All analyses were performed using SAS version 9.2 (Cary, NC). Descriptive statistics were calculated and compared by race using either a Chi-square test or t test as appropriate. Kaplan Meier survival curves were calculated and the log-rank test statistic was used to assess for statistical differences between race groups. Cox proportional hazards modeling was used to test the association between race and mortality (both BrCA-specific and overall) after adjusting for healthcare insurance and various tumor characteristics. Age was included in all Cox proportional hazards models.

Due to issues of collinearity, stage and Elston score were not included in the same Cox proportional hazards model; hence, two final models were created. The first excluded Elston grade and adjusted for age, insurance, cancer stage, ER, and Her2. The second model excluded cancer stage and adjusted for age, insurance, ER, Elston grade, and Her2. The proportional hazards assumption was examined visually by inspecting graphs of survival function against log survival time and in the Cox model by creating interactions of the predictors and a function of survival time. An alpha level of 0.05 was used to determine significance for all tests.

Results

A total of 1,209 BrCA patients diagnosed between 2000–2002 were included in the analysis. The race distribution was 31% AA and 79% EA. The mean age for EA patients was 59 years (SD, 13; range, 25–92 years) and for AA patients, the mean was 55 (SD, 13; range, 20–97). The majority of EA and AA women had private insurance (59.47% and 52.27%, respectively). Other demographic characteristics are displayed in Table 1.

Table 1.

Characteristics of Breast Cancer Patients by Ethnicity, 2000–2002^*

Characteristic	Race		P-value
	Black†, n (%)	White†, n (%)
No of patients	375 (31.0%)	834 (69.0%)
Age
<40	36 (9.6%)	51 (6.1%)	<0.01
40–49	99 (26.4%)	157 (18.8%)
50–59	112 (29.9%)	237 (28.4%)
60–69	64 (17.1%)	194 (23.4%)
≥ 70	64 (17.1%)	195 (23.4%)
Insurance
Public‡	154 (41.1%)	288 (34.5%)	0.09
Private	196 (52.3%)	496 (59.5%)
Not Insured	16 (4.3%)	26 (3.1%)
Unknown	9 (2.4%)	24 (2.9%)
Histology Description
Ductal Carcinoma	284 (75.7%)	586 (70.3%)	0.03
Lobular Carcinoma	18 (4.8%)	73 (8.8%)
Mixed Ductal & Lobular	10 (2.7%)	39 (4.7%)
All Others	63 (16.8%)	136 (16.3%)
Behavior of Cancer
In Situ	65 (17.3%)	155 (18.6%)	0.60
Invasive	310 (82.7%)	679 (81.4%)
Hormone Treatment
Administered	154 (41.1%)	472 (56.6%)	<0.01
None	221 (58.9%)	362 (43.4%)
Cancer Stage
Stage 0	65 (17.6%)	156 (19.1%)	<0.01
Stage I	97 (26.3%)	328 (40.1%)
Stage II	147 (39.8%)	267 (32.6%)
Stage III	33 (9.8%)	49 (5.9%)
Stage IV	24 (6.5%)	18 (2.2%)
Elston Grade
Low	34 (14.5%)	158 (31.6%)	<0.01
Moderate	69 (29.4%)	182 (36.4%)
High	132 (56.2%)	160 (32.0%)
Estrogen Receptor Status
Positive	165 (58.3%)	520 (80.5%)	<0.01
Negative	118 (41.7%)	126 (19.5%)
Her2§
Positive	35 (14.2%)	75 (13.4%)	0.66
Borderline	25 (10.1%)	69 (12.3%)
Negative	187 (75.7%)	415 (74.2%)

^*Missing values were excluded.

^†The terms Black and White are used for disease registration purposes. For this manuscript, they are essentially synonymous with African American and European American

^‡Includes Medicare and Medicaid

^§Human epidermal growth factor receptor 2

Table 2 presents the 3- and 5-year survival proportion for breast cancer-specific and all-cause mortality among AA and EA women. Similar to statistics for state-wide data, significant racial differences were observed in the overall 5-year survival proportion (~78% for AA women and ~89% for EA women, p< 0.01). No significant racial differences were evident among the different stages of disease. However, there were significant overall racial differences in survival in both low- and high-Elston grade populations, with the largest difference evident at 5 years. Significant racial differences in survival also were observed by ER and Her2 status. Among those who had private insurance, AA women had a significantly lower 3- and 5- year survival proportion of all-cause mortality compared to EA women (83% vs. 95%, respectively, p-value<0.01). Interestingly, no significant differences were noted for public insurance or uninsured patients.

Table 2.

Three- and Five- Year Survival for Breast Cancer and All-Cause Mortality, By Race

Variable	Race	Breast Cancer Mortality				All-Cause Mortality
		N	3-Year Survival (95% CI)	5-Year Survival (95% CI)	P-value‡	N	3-Year Survival (95% CI)	5-Year Survival (95% CI)	P-value§
Overall	AA	306	92% (89%–95%)	91% (87%–94%)	<0.01	375	86% (82%–89%)	78% (73%–82%)	<0.01
	EA	715	98% (97%–99%)	98% (97%–99%)		834	93% (91%–95%)	89% (87%–91%)
Cancer Stage
Stage 0	AA	61	100%	98% (88%–100%)	0.51	65	100%	95% (86%–98%)	0.07
	EA	151	99% (95%–100%)	99% (95%–100%)		156	99% (96%–100%)	99% (96%–100%)
Stage I	AA	78	100%	100%	0.25	97	94% (87%–97%)	84% (75%–90%)	0.11
	EA	294	99% (97%–100%)	99% (97%–100%)		328	96% (92%–97%)	92% (88%–95%)
Stage II	AA	123	95% (89%–98%)	95% (89%–98%)	0.05	147	88% (82%–93%)	83% (76%–88%)	0.41
	EA	215	99% (95%–100%)	99% (95%–100%)		267	92% (88%–95%)	85% (80%–89%)
Stage III	AA	25	68% (46%–83%)	64% (42%–80%)	<0.01	36	66% (49%–80%)	55% (37%–69%)	0.13
	EA	35	97% (15%–77%)	94% (78%–98%)		49	82% (68%–90%)	75% (60%–87%)
Stage IV	AA	8	33% (12%–56%)	27% (8%–50%)	0.19	18	33% (16%–52%)	13% (3%–29%)	0.09
	EA	15	50% (15%–77%)	50% (15%–77%)		24	44% (22%–65%)	33% (14%–55%)
Elston Grade
Low	AA	26	100%	100%	------	34	94% (78%–98%)	78% (60%–89%)	<0.01
	EA	142	100%	100%		158	97% (93%–99%)	96% (92%–98%)
Moderate	AA	56	100%	98% (88%–100%)	0.51	69	93% (83%–97%)	87% (76%–93%)	0.52
	EA	155	99% (94%–100%)	99% (94%–100%)		182	92% (87%–95%)	88% (82%–92%)
High	AA	101	85% (76%–91%)	84% (75%–90%)	<0.01	132	77% (69%–83%)	69% (60%–76%)	0.04
	EA	121	97% (92%–99%)	97% (91%–99%)		160	87% (81%–92%)	80% (73%–86%)
Estrogen Receptor Status
Positive	AA	134	96% (90%–98%)	96% (90%–98%)	<0.01	165	90% (85%–94%)	81% (74%–86%)	0.04
	EA	439	99% (97%–100%)	99% (97%–100%)		520	90% (85%–94%)	89% (86%–91%)
Negative	AA	92	82% (72%–89%)	79% (68%–86%)	0.02	118	72% (63%–80%)	65% (56%–73%)	0.03
	EA	103	93% (86%–97%)	91% (83%–95%)		126	86% (78%–91%)	78% (70%–85%)
Her2*
Positive	AA	31	81% (62%–91%)	74% (55%–86%)	<0.01	35	80% (62%–90%)	67% (49%–80%)	0.03
	EA	66	95% (86%–98%)	95% (86%–98%)		75	88% (78%–93%)	85% (74%–91%)
Negative	AA	150	93% (87%–96%)	93% (87%–96%)	<0.01	187	84% (78%–89%)	79% (72%–84%)	0.02
	EA	346	99% (97%–99%)	99% (97%–99%)		415	93% (90%–95%)	88% (78%–94%)
Borderline	AA	18	89% (62%–97%)	89% (62%–97%)	0.34	25	84% (63%–94%)	63% (41%–79%)	<0.01
	EA	61	97% (87%–99%)	95% (85%–98%)		69	93% (83%–97%)	88% (78%–94%)
Insurance
Public†	AA	111	91% (83%–95%)	91% (83%–91%)	0.04	154	82% (74%–87%)	71% (63%–78%)	0.25
	EA	211	96% (92%–98%)	96% (92%–98%)		288	86% (82%–90%)	78% (73%–83%)
Private	AA	174	92% (87%–96%)	91% 86%–95%)	<0.01	196	89% (84%–93%)	83% (77%–87%)	<0.01
	EA	461	99% (98%–100%)	99% (98%–100%)		496	97% (95%–98%)	95% (93%–97%)
Not Insured	AA	14	86% (54%–96%)	86% (54%–96%)	0.34	16	81% (52%–94%)	81% (52%–94%)	0.51
	EA	22	95% (70%–99%)	95% (70%–99%)		26	88% (67%–96%)	84% (62%–94%)

^*Human epidermal growth factor receptor 2

^†Includes Medicare and Medicaid

^‡Comparison between AA and EA among breast cancer mortality

^§Comparison between AA and EA among all-cause mortality

As with overall survival, AA women had significantly lower breast cancer survival rates for most tumor characteristics and insurance categories. Unlike overall survival models, AA women who had public insurance were significantly more likely to die of breast cancer than EA women (91% vs. 96%, p=0.04)

Table 3 presents the results of several BrCA-specific and overall survival Cox proportional hazards models. AA women had more than a four-fold excess risk of death from BrCA (HR=4.49; 95% CI, 2.42–8.35) and about two-fold excess risk of death from all causes (HR=2.04; 95% CI, 1.55–2.67). After controlling for age, insurance, stage, Elston grade, ER, and Her2, AA women still had a higher risk of death from both BrCA and all-cause mortality. Two other models were run to examine the separate effects of stage and Elston score. The first model, adjusted for age, insurance, stage, ER, and Her2, showed that AA women had more than twofold excess risk of death for BrCA (HR=2.41; 95% CI, 1.21–4.79) and 1.42 (95% CI, 1.06–1.89) times for all-cause mortality. The second model was adjusted for age, insurance, Elston score, ER, and Her2; the hazard ratios for BrCA and all-cause mortality were 3.45 (95% CI, 1.79–6.65) and 1.54 (95% CI, 1.16–2.05), respectively.

Table 3.

Adjusted Hazard Ratios Among AA Women Compared With EA Women for Breast Cancer-Specific and Overall Survival Deaths with Selected Characteristics Added into the Model

Variables *	Hazard Ratio for Race (95% CI†)
	Breast Cancer Deaths	All Deaths
Race	4.49 (2.42–8.35)	2.04 (1.55–2.67)
Race, insurance	4.25 (2.26–7.99)	1.81 (1.37–2.39)
Race, stage	2.69 (1.41–5.13)	1.70 (1.29–2.24)
Race, Elston grade	3.94 (2.09–7.39)	1.80 (1.37–2.37)
Race, estrogen receptor status	3.27 (1.70–6.18)	1.79 (1.36–2.36)
Race, human epidermal growth factor receptor 2	4.62 (2.48–8.61)	2.03 (1.55–2.67)
Final model-1‡	2.41 (1.21–4.79)	1.42 (1.06–1.89)
Final model-2§	3.45 (1.79–6.65)	1.54 (1.16–2.05)

^*All models include age

^†Confidence interval

^‡Adjusted for age, insurance, estrogen receptor status, human epidermal growth factor receptor 2, and stage

^§Adjusted for age, insurance, estrogen receptor status, human epidermal growth factor receptor 2, and Elston grade

Discussion

In this multi-ethnic cohort in South Carolina, of whom nearly one-third were African Americans, significant racial disparities were found in BrCA mortality after accounting for tumor characteristics and other prognostic characteristics. After adjusting for important covariates, the excess risk of death from BrCA in AA women, was 2.41 times the risk for EA women.

Many factors may contribute to differential breast cancer prognosis for AA and EA women, and include biological aspects, ^{8–9, 14–16, 19} socioeconomic status, ^{7, 10–13, 20} and health care access. ²¹ Patient’s refusal of therapy has also been shown to be an important contribution to breast cancer prognosis, especially for AA women.²² With this investigation, we were able to control for some aspects related to socioeconomic status and health care access by utilizing a population of patients who received all their breast cancer care at an institution with standardized protocols to ensure standardization of the physician recommendation and receipt of different treatment regimens. Previous quality care investigations demonstrated that all breast cancer patients within this program received the recommended treatment protocol. ²³ Consequently, our findings suggest that some of the mortality disparities observed may be related to differences in the biological basis of disease development and progression. If this hypothesis should prove to be true upon additional investigation, future work should focus on elucidating better and more specialized treatment regimens for AA women diagnosed with breast cancer.

These findings also argue for investigation into the underlying causes of carcinogenesis that could result in health interventions among AA BrCA patients aimed at tertiary prevention strategies or disease recurrence. For example, modification of diet and physical activity have been shown to be potentially important secondary tertiary prevention efforts. ^24–25 Our previous work shows that diet and physical activity interventions can be effective in reducing body mass (and thereby, adiposity) among BrCA survivors in the short term.²⁶ Work from our group and others shows that adiposity, as estimated by body mass index [BMI=weight(kg)/height(m)²] can influence survival.^{24–25, 27–29} There also is evidence that increased vegetable consumption post-diagnosis improves survival.^{24, 30–31} Other strategies might include consistent post treatment surveillance and approaches to strengthen self-care activities. Indeed, some psychosocial interventions that have been designed to only positively impact quality of life have been shown to affect survival.³² Very few current investigations focus on meaningfully large AA populations; so, increasing representation of this high-risk group in future research is a critical need.

As with any epidemiological investigation, this study is not without its limitations. While the hospital attempts to standardize treatment recommendations and courses for all BrCA patients diagnosed in their system,²³ individual patients may choose not to undergo a recommended treatment due to financial reasons, cultural perspective, or personal beliefs. If such choices are driven by culture, we would expect our findings to be biased. Additionally, we were using data collected for vital status reporting purposes, not research, and were therefore somewhat limited in scope. Consequently, we did not have information on factors that might impact recurrence and subsequently mortality such as body mass, treatment modality, and comorbidities. Also, the only socioeconomic variable we have from the registry is health insurance; therefore, complete adjustment for socioeconomic status in our finding is somewhat limited.

On the other hand, there are many strengths to this investigation which should be noted. Though cohorts have been formed to investigate racial disparities in cancer risks in underserved populations throughout the south,³³ this is one of the first BrCA mortality studies to be conducted in a multi-ethnic cohort representing the Southeastern United States. South Carolina is characterized as a largely rural state with a high representation of AA residents.¹⁸ It also has many areas which are medically underserved (including the area from which this cohort is derived). Thus, this article focuses on a population which has been chronically underrepresented both in terms of careful monitoring of services and participation in research. In addition, due to the large AA women representation and lengthy follow-up period, we had ample sample sizes and power to examine the various breast cancer sub-populations. These stratified analyses were instrumental in being able to elucidate the impact of tumor characteristics and race on mortality.

In conclusion, we found evidence for significantly poorer survival among AA BrCA patients diagnosed with identical stage breast cancer as EA patients. Furthermore, after adjusting for multiple prognostic indicators, AA women were still more likely to die from both BrCA and all-cause mortality compared to EA women. Evidence derived from this cohort suggests a possible biological processing basis for some of the BrCA mortality disparities seen in South Carolina.