Racial disparities in breast carcinoma survival rates

Separating factors that affect diagnosis from factors that affect treatment

Abstract

BACKGROUND

Black females have lower breast carcinoma survival rates compared with white females. One possible reason is that black females have more advanced-stage breast disease. Another factor may be racial differences in the utilization of cancer treatments.

METHODS

The authors determined racial differences in 6-year stage specific survival rates, adjusting for age and treatments (using estrogen receptor [ER] status), to determine whether there were racial differences in treatment. Racial differences in the stage distributions of breast disease were used to examine the impact of racial factors on breast carcinoma diagnosis.

RESULTS

For all breast carcinoma cases, the stage specific 6-year survival rates, in general, were significantly lower for black females for all stages combined and for Stages I–III in every age group. However, examination by different treatments, as measured by ER status, revealed some different results. Only black women younger than age 50 years with ER-positive tumors and women younger than age 65 years with ER-negative tumors had significantly lower stage-specific survival rates. In addition, the stage distribution analyses showed that black females of every age group had less Stage I breast disease.

CONCLUSIONS

For younger black women (younger than age 50 years), there was evidence of racial differences in treatment for both women with ER-positive tumors and women with ER-negative tumors, as indicated by their lower stage-specific survival rates. In contrast, for black females age 65 years or older with ER-positive or ER-negative tumors, the lack of a significant difference in the stage-specific survival rate suggests that Medicare may help to alleviate racial disparities in cancer treatment. Furthermore, racial differences in the stage distributions indicated the need for earlier diagnosis for black females of every age. cancer 2003;97:2853–60. Published 2003 American Cancer Society.

DOI 10.1002/cncr.11411

It is well established that breast carcinoma survival rates for black females are lower compared with the rates for white females.1 Multivariate analyses of breast carcinoma survival rates have shown that black females have less early stage breast disease and more advanced-stage breast disease compared with white females,2, 3 even in equal-access health care systems.4, 5 Clearly, the presence of less early stage disease and more late stage disease in black females is a major contributor to their lower survival rates.

Another contributing factor in lower survival rates for black females may be racial differences in treatment.6 Randomized clinical trial studies in women with breast carcinoma have shown that monitoring patients (to insure equal utilization of medical services) and performing standardized treatments yield equal survival rates for black females and white females for patients with the same disease stage at diagnosis.7–9 Although equal treatment (including equal utilization) for patients with the same disease stage yields equal survival rates, unequal stage specific survival rates may indicate racial differences in treatment.

Disparities resulting from racial differences in treatment have been documented.6 For example, Diehr et al. showed that black females were less likely to receive radiation therapy in combination with radical/modified mastectomy and were less likely to receive rehabilitation support services after undergoing mastectomy.10 Bach et al. demonstrated that blacks were less likely to undergo surgery for resectable nonsmall cell lung carcinoma compared with whites after adjusting for age, gender, income, comorbidities, geographic region, and type of Medicare insurance.11 Those studies indicate that there can be racial disparities in cancer treatments.

Numerous studies have measured race as a predictor of survival. Examinations of multivariate analyses of survival rates indicate that some studies have identified race as an independent predictor of survival after adjustment for patient age, disease stage, estrogen receptor (ER) status, tumor grade, socioeconomic status, histology, and other risk factors.3, 12–17 However, other investigators did not find that race was a significant predictor of survival according to these multivariate analyses.7, 18–22 Those studies illustrate the complexity of examining the impact of individual factors, such as race, in analyzing breast carcinoma survival rates.

We offer an alternative approach. To determine whether there are racial differences in treatment, we examined differences in black and white stage-specific survival rates. Survival rates are affected by factors that are determined after the diagnosis of breast carcinoma, principally by cancer treatments. In patients with breast carcinoma, treatments vary according to ER status; therefore, examining survival rates according to ER status provides an indication of whether there are racial differences for different treatments. Specifically, we determined differences in the stage-specific survival rates between white females and black females among women with American Joint Committee on Cancer (AJCC) Stage I–IV breast carcinoma according to patient age (i.e., all ages, younger than age 50 years, ages 50–64 years, and age 65 years and older) and treatments (according to ER status: all breast tumors, ER-positive tumors, and ER-negative tumors). To determine whether there are racial differences in factors that affect the diagnosis of breast carcinoma, such as screening and early detection, we assessed racial differences in the stage distribution of breast carcinoma cases.

MATERIALS AND METHODS

Data Collection

The data were obtained from population-based data collected by the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute for the period from January 1992 to December 1999.23 The invasive breast carcinomas (i.e., in situ carcinomas were excluded) used in this study were diagnosed among residents of 11 geographic areas: Connecticut, Hawaii, Iowa, New Mexico, Utah, Atlanta, Detroit, Seattle-Puget Sound, San Francisco-Oakland, San Jose-Monterey, and Los Angeles.23

The breast carcinoma-specific survival rates were based on invasive breast disease in individuals who had breast carcinoma designated as their underlying cause of death on their death certificates. To examine the long-term effects of treatment, we examined survival rates at the longest possible follow-up. For our data, 6-year breast carcinoma-specific survival rates and their standard errors were calculated for black females and white females using the SEER cancer incidence public-use CD-ROM (1973–1999) with its SEER*Stat computer program (version 4.2; National Cancer Institute, Bethesda, MD).23

The variables analyzed were ER status, disease stage at diagnosis, and patient age. The ER status was coded by the SEER Program data collectors based on laboratory results in the medical records at the time of data collection since 1992 for the 11 cancer registries. The ER status data are reported for ER-positive tumors and ER-negative breast tumors. Tumors with unknown ER status (this group includes not done, borderline/undetermined, ordered but results not in chart, or unknown for ER status) were excluded.24 The ER status was used because there are different beneficial cancer treatments for women with ER-positive tumors and women with ER-negative tumors.25

Data concerning disease stage at diagnosis are collected according to the SEER Extent of Disease codes and coding instruction manual.24 The stage at diagnosis used in this report was based on the AJCC's classification of tumors (Stages I–IV).26

In addition to women of all ages combined, women younger than age 50 years, women ages 50–64 years, and women age 65 years and older were examined. The older age group was chosen to represent possible Medicare-eligible patients.

Statistical Methods

The difference in the 6-year survival rate between white females and black females was determined for each type of stage-specific survival (all stages combined and Stages I–IV individually). A test was performed to determine whether the stage specific survival rates for white females and black females were equal:

• 

in which Z is the standard normal deviate, S_i^W is the Stage i survival rate for white females, SE(S_i^W) is the standard error for the Stage i survival rate for white females, S_i^B is the Stage i survival rate for black females, and SE(S_i^B) is the standard error for the Stage i survival rate for black females. A Bonferroni correction for the 20 multiple comparisons (4 age groups × 5 types of survival; all stages, Stage I, Stage II, Stage III, and Stage IV) was used, so that a result was significant if the P value was < 0.0025 (0.05/20).

A test to determine whether the fraction of patients with Stage i disease was the same for black females and white females also was calculated as follows:

• 

in which Z is the standard normal deviate, F_i^W is the fraction of Stage i tumors based on Stage I–IV disease in white females, SE(F_i^W) is the standard error for the fraction of Stage i tumors in white females, F_i^B is the fraction of Stage i tumors based on Stage I–IV disease in black females, and SE(F_i^B) is the standard error for fraction of Stage i tumors in black females. A Bonferroni correction for 16 multiple comparisons (4 age groups × the 4 stages) was used, so that a result was significant if the P value was < 0.0031 (0.05/16).

RESULTS

The differences in stage distributions between white and black females are reported in Table 1 by age (all ages, younger than age 50 years, ages 50–64 years, and age 65 years and older) and ER status (all breast tumors, ER-positive tumors, and ER-negative tumors). Black females had less Stage I disease and more Stage II, III, and IV disease compared with white females. This was true for all breast tumors and for ER-positive tumors in every age group. For ER-negative tumors, black females had less Stage I disease in every age group; however, in general, they did not have more late-stage disease, except for black women ages 50–64 years, who had more Stage III disease compared with white females. Generally, black females had less early-stage disease in every age group for each ER status and had more late-stage disease in every age group for all breast tumors and ER-positive tumors only.

Table 1. Number of Breast Carcinomas and the Differences in Proportion of Carcinomas between White Females and Black Females by Stage, Age and Estrogen Receptor Status

Stage	All ages				Ages < 50 yrs				Ages 50–64 yrs				Ages 65 yrs and older
	White	Black	Stage diff (%)	P value	White	Black	Stage diff (%)	P value	White	Black	Stage diff (%)	P value	White	Black	Stage diff (%)	P value
NS: not significant (two-sided P value > 0.05); ns: not significant due to multiple comparisons (P value between 0.05 and 0.0031 [0.05/16]); Stage diff: the difference in proportion of carcinomas in stage between white females and black females. a P value < 0.0031.
All breast carcinomas
I–IV	106,536	10,667	—	—	25,333	3703	—	—	33,785	3506	—	—	47,418	3458	—	—
I	52,784	3785	14a	< 0.0001	9959	1055	11a	< 0.0001	17,129	1304	14a	< 0.0001	25,696	1426	13a	< 0.0001
II	41,557	4891	−7a	< 0.0001	12,241	1970	−5a	< 0.0001	12,980	1562	−6a	< 0.0001	16,336	1359	−5a	< 0.0001
III	7385	1178	−4a	< 0.0001	2228	445	−3a	< 0.0001	2200	359	−4a	< 0.0001	2957	374	−5a	< 0.0001
IV	4810	813	−3a	< 0.0001	905	233	−3a	< 0.0001	1476	281	−4a	< 0.0001	2429	299	−4a	< 0.0001
Unknown	7959	1076	—	—	1768	336	—	—	2220	318	—	—	3971	422	—	—
Estrogen receptor-positive tumors
I–IV	66,351	4686	—	—	13,644	1371	—	—	21,116	1518	—	—	31,591	1797	—	—
I	35,072	1894	12a	< 0.0001	5843	450	10a	< 0.0001	11,330	622	13a	< 0.0001	17,899	822	11a	< 0.0001
II	25,220	2124	−7a	< 0.0001	6385	718	−6a	< 0.0001	7939	685	−8a	< 0.0001	10,896	721	−6a	< 0.0001
III	3985	412	−3a	< 0.0001	1068	133	−2	ns	1196	125	−3a	0.0004	1721	154	−3a	< 0.0001
IV	2074	256	−2a	< 0.0001	348	70	−3a	< 0.0001	651	86	−3a	< 0.0001	1075	100	−2a	< 0.0001
Unknown	2726	275	—	—	572	71	—	—	819	98	—	—	1335	106	—	—
Estrogen receptor-negative tumors
I–IV	19,363	3026	—	—	6925	1343	—	—	6399	1028	—	—	6039	655	—	—
I	7496	902	9a	< 0.0001	2244	338	7a	< 0.0001	2598	342	7a	< 0.0001	2654	222	10a	< 0.0001
II	9087	1571	−5a	< 0.0001	3715	778	−4	ns	2911	496	−3	NS	2461	297	−5	ns
III	1908	385	−3a	< 0.0001	732	166	−2	NS	605	135	−4a	0.001	571	84	−3	ns
IV	872	168	−1	ns	234	61	−1	NS	285	55	−1	NS	353	52	−2	NS
Unknown	1008	181	—	—	335	75	—	—	335	58	—	—	338	48	—	—

The 6-year survival rates for white females and black females by stage and their racial differences are presented in Table 2 according to age and ER status. For all breast tumors, the differences in survival rates between white females and black females indicated that black females had significantly poor survival rates compared with white females for all stages combined and for Stages I–III in every age group. Examination by treatments, as indicated by the ER status, showed a different picture: For ER-positive tumors, black females had significantly lower survival rates for all stages combined in every age group, although there were no significant differences in any stage-specific survival rates for black females age 50 years and older. For ER-positive tumors, only black females younger than age 50 years had significantly lower stage-specific survival rates. For ER-negative tumors, black females also had significantly lower survival rates for all stages combined in every age group, although there were no significant differences in any stage specific survival rates for black females age 65 years and older. Accordingly, black females with ER-negative tumors had lower survival rates with Stage II disease in the group younger than age 50 years and had lower survival rates with Stage III disease in the group ages 50-64 years.

Table 2. Six-Year Survival Rates by Race, Age, Stage, and Estrogen Receptor Status

Stage	All ages				Ages < 50 yrs				Ages 50–64 yrs				Ages 65 yrs and older
	White	Black	Surv diff	P value	White	Black	Surv diff	P value	White	Black	Surv diff	P value	White	Black	Surv diff	P value
NS: not significant (two-sided P value > 0.05); ns: not significant due to multiple comparisons (P value between 0.05 and 0.0025 [0.05/20]); Surv diff: the difference in stage specific 6-year survival rates between white and black females. a P value < 0.0025.
All breast carcinomas
I–IV	0.85	0.72	0.13a	< 0.0001	0.84	0.69	0.15a	< 0.0001	0.86	0.74	0.13a	< 0.0001	0.85	0.74	0.11a	< 0.0001
I	0.96	0.93	0.04a	< 0.0001	0.96	0.91	0.05a	0.0002	0.97	0.93	0.04a	< 0.0001	0.96	0.94	0.02	ns
II	0.83	0.74	0.09a	< 0.0001	0.84	0.72	0.12a	< 0.0001	0.84	0.77	0.07a	< 0.0001	0.82	0.75	0.07a	< 0.0001
III	0.56	0.38	0.19a	< 0.0001	0.60	0.37	0.23a	< 0.0001	0.56	0.35	0.20a	< 0.0001	0.54	0.41	0.13a	0.0011
IV	0.18	0.12	0.06a	0.0024	0.24	0.13	0.11	ns	0.19	0.11	0.09a	0.0019	0.15	0.14	0.01	NS
Unknown	0.78	0.68	—	—	0.83	0.66	—	—	0.85	0.74	—	—	0.72	0.65	—	—
Estrogen receptor-positive tumors
I–IV	0.89	0.80	0.10a	< 0.0001	0.89	0.77	0.12a	< 0.0001	0.90	0.81	0.10a	< 0.0001	0.89	0.81	0.08a	< 0.0001
I	0.97	0.95	0.02a	0.0014	0.97	0.95	0.02	NS	0.98	0.95	0.03	ns	0.97	0.95	0.02	NS
II	0.87	0.80	0.08a	< 0.0001	0.88	0.77	0.11a	< 0.0001	0.88	0.82	0.06	ns	0.86	0.80	0.06	ns
III	0.66	0.48	0.18a	< 0.0001	0.71	0.49	0.22a	0.0003	0.65	0.49	0.15	ns	0.62	0.46	0.16	ns
IV	0.23	0.16	0.07	NS	0.31	0.16	0.15	NS	0.24	0.11	0.12	ns	0.19	0.21	−0.01	NS
Unknown	0.89	0.76	—	—	0.90	0.74	—	—	0.93	0.73	—	—	0.86	0.79	—	—
Estrogen receptor-negative tumors
I–IV	0.76	0.66	0.10a	< 0.0001	0.78	0.66	0.11a	< 0.0001	0.77	0.68	0.10a	< 0.0001	0.73	0.62	0.11a	< 0.0001
I	0.92	0.87	0.05a	0.0017	0.93	0.87	0.06	ns	0.93	0.88	0.05	ns	0.92	0.88	0.04	NS
II	0.75	0.68	0.06a	0.0001	0.78	0.68	0.10a	< 0.0001	0.76	0.72	0.05	NS	0.68	0.64	0.04	NS
III	0.44	0.29	0.15a	< 0.0001	0.49	0.36	0.13	ns	0.44	0.24	0.19a	0.0020	0.36	0.22	0.15	ns
IV	0.12	0.09	0.03	NS	0.17	0.15	0.02	NS	0.12	0.08	0.04	NS	0.09	0.05	0.05	NS
Unknown	0.78	0.65	—	—	0.80	0.63	—	—	0.83	0.72	—	—	0.72	0.58	—	—

DISCUSSION

We will discuss the basis for using stage specific survival rates to measure treatment disparities, the inferences from the racial differences in the stage-specific rates, the racial differences in the stage distribution of tumors, and limitations of the methodology employed. We used black and white differences in stage-specific survival rates as indicators of racial differences in treatment. We believe that this approach is meaningful for a number of reasons. First, survival rates measure the impact of those events that occur after diagnosis, principally the impact of treatment. In contrast, stage distributions measure the impact of events affecting diagnosis. In breast carcinoma, there are two major phenomena: beneficial early detection due to mammography and beneficial treatments due to adjuvant therapies.27 To examine the role of treatment, there is a need to separate the impact of early detection from treatment. Early detection involves stage shifts of late-stage disease to earlier stages. The stage distributions capture these diagnostic phenomena. Thus, examination of the stage specific rates allows consideration of treatment without the confounding factor that black females may have more late-stage disease.

Another reason that this approach is meaningful is that there are a number of stage-specific, beneficial treatments for women with breast carcinoma. The NIH Consensus Panel recommended breast-conserving surgery with radiation therapy for patients with Stage I and II breast carcinoma and for some patients with Stage III breast carcinoma.28 In general, women who have ER-positive tumors are then treated with tamoxifen as the adjuvant systemic therapy, whereas women with ER-negative tumors are treated with chemotherapy, although there are treatment differences that depend on patient age, risk, tumor size, and lymph node status in addition to ER status.25 Thus, because there are beneficial treatments for women with Stage I, II, and III breast disease, racial differences in the stage specific survival rates according to ER status may reflect racial differences in the utilization of these beneficial therapies for women with breast carcinoma.

A third reason for this approach is that multivariate analyses may have statistical interactions. We have shown that differences in stage-specific survival rates are not the same across age groups or ER groups, causing statistical interactions in the multivariate analyses. In this situation, examination of the stage-specific survival rates by age and ER status, as we have done, is recommended.29 Finally, the approach is general, so that other disease sites that have stage-specific rates can be examined in a similar manner. More important, the approach can be used to examine health disparities in other racial ethnic groups, such as Hispanics, American Indians, and Asian Americans and Pacific Islanders, for whom data regarding possible treatment disparities are sorely lacking. Thus, the examination of racial differences in stage-specific survival rates offers a powerful approach to examine cancer disparities in racial/ethnic groups.

Two sets of stage specific survival rate results appear to have particularly important significance. First, black women younger than age 50 years have significantly lower stage-specific survival rates for ER-positive and ER-negative tumors. Because unequal survival rates may be due to unequal treatments, these results appear to provide evidence of unequal treatments for these younger black women. Lower survival rates among younger black females may have been due, in part, to the fact that young black females have more ER-negative tumors (which indicate a poorer prognosis) compared with white females.30 However, we controlled for this effect by examining survival rates for ER-positive tumors and ER-negative tumors separately. In each case, black females had lower rates, indicating that lower survival rates occur independent of the differences in the proportion of ER-positive and ER-negative tumors in the racial groups. Examination of risk factors in young black women and young white women indicated little evidence of differences in risk factors between the racial groups.31 Thus, it is unlikely that differences in survival rates are due to racial differences in risk factors in younger age groups. It is believed that genetic factors are another potential source of differences in survival rates. Black females may have more ER-negative tumors, and BRCA1 tends to be associated with ER-negative tumors in younger women. However, we have seen the survival differences in both women with ER-positive tumors and women with ER-negative tumors. Thus, it is unlikely that genetic factors are the source of the survival differences we report. We are left with the conclusion that racial differences in the treatment of women with breast carcinoma may be the source of differences in stage-specific survival rates observed in younger women.

Two broad categories of factors that potentially influence the receipt of optimal cancer care are structural factors, such as insurance coverage, medical facilities, and staff to provide standard treatment; and information issues involving physician/clinicians and patients.32 Lack of information for physicians and clinicians regarding appropriate treatments and for patients to choose the appropriate treatments can be an important factor in treatment disparities.33 Furthermore, utilization of cancer care, even when appropriate treatments are known, is influenced greatly by behaviors and decision making by health care providers and patients.32 In women age 65 years and older, survival rates were higher in white females compared with black females, although stage-specific survival rates were not significantly different. One possible interpretation of our results suggests that health insurance, as provided by Medicare, may help to alleviate these barriers. The structural barriers may be overcome by offering health insurance and by providing information to physicians and patients concerning which treatment procedures are covered by Medicare. Other possible explanations for the lack of a significant difference include the loss of statistical power in the subset analyses by age and ER status. However, the fact that the lack of significance is found for both women with ER-positive tumors and women with ER-negative tumors may argue against this explanation.

In addition to examining survival rates, we wanted to determine whether there were racial differences in the diagnosis of breast carcinoma. In general, the stage distributions of tumors are affected by factors that affect the diagnosis of breat carcinoma. These factors may include environmental risk factors; behavioral/cultural factors related to obtaining early detection and timely cancer diagnosis; access to medical services for early detection, testing, and cancer diagnosis; socioeconomic status; and other relevant factors that affect the timely detection and diagnosis of malignant disease. The impact of all of these factors is captured in the stage distributions of tumors at the time of diagnosis.

We interpret our findings that black females have less early-stage disease in every age group for each ER status to mean that there is a need for more early diagnosis of breast carcinoma for black females. In the 1980s, surveys indicated that black women underwent fewer mammograms.34 However, more recently, the 1998 National Health Interview Survey indicated that black females' use of mammography was similar to white females for women younger and older than age 65 years.35 Consequently, the recent stage distribution differences may be due to the delays in diagnosis related to lack of follow-up of abnormal mammograms until the disease is in a later stage.36–38 The greater amount of late-stage disease in black females also may reflect the greater difficulty in getting underserved minority populations to participate in screening activities, leading to later stages of diagnosis for these populations. In addition, several studies have shown that regular mammography, rather than a one-time-only screening, is another important factor in increasing survival rates.39–41 Thus, the racial differences in stage distribution may reflect the need for black females of all ages to reduce the time from abnormal mammograms to diagnosis with more timely follow-up of abnormal screening results and the need for outreach to hard-to-reach, underserved populations as well as the need to have regular mammograms.

There are limitations to these procedures. An implicit assumption of this methodology is that the racial differences in stage-specific survival rates are not due to biologic differences in the aggressiveness of stage-specific breast disease. If there are biologic differences in the aggressiveness of stage-specific disease, with black females having more aggressive stage-specific disease, then the differences in survival rates may be due to these biologic differences. However, we have reported that the stage specific differences vary by age and ER status. Tumor aggressiveness according to race would have to vary in the same way to account for these trends, which seems unlikely. In addition, randomized clinical treatment trials have shown that, when black patients and white patients are monitored in the same manner and are given equivalent treatments, there are equal survival rates.7–9 Other possible limitations are that racial differences in delays from diagnosis to treatment can lead to poorer survival rates for black females. In addition, the racial differences in the number of persons who may be eligible for a treatment but are not given the treatment also may affect survival rates. The SEER public-use data base does not report the data necessary to evaluate these factors. Another potential limitation is that the partitioning of survival rates into factors that are affected by diagnosis and factors that are affected after diagnosis is not always clear-cut, particularly for early stage breast carcinoma. For example, the early detection of small tumors may increase early-stage survival rates. Thus, the group with more screening not only will have more early-stage tumors, they may have better early-stage survival rates. Thus, early detection can affect both the stage distribution and the survival of women with Stage I disease. However, the impact of early detection on the survival rates for women with later stage disease should be much less. Another issue is that the stage distribution must sum to 100%. An increase in one stage is followed by a decrease in another stage. Differences of stage distributions that involve an increase in early-stage disease, such as mammography, may cause the appearance of declines in late-stage disease. Examination of black and white incidence rates substantiates that black females have more late-stage disease (Stage IV: 7.6 per 100,000 for black females vs. 4.9 per 100,000 for white females).23 Finally, this analysis is a point in time: Survival rates change with more follow-up, so that conclusions drawn at one time may change with further follow-up.

In the current study, we have presented results that examine the differences in the stage-specific survival rates between white females and black females as well as differences in the stage distribution of their breast carcinomas. The racial differences in stage-specific survival rates varied by age and ER status. Black females younger than age 50 years with ER-positive or ER-negative tumors had significantly lower survival rates, suggesting the need to focus on racial differences in treatments for younger black females. For black women ages 65 years and older with ER-positive or ER-negative tumors, there were no significant differences in stage-specific survival rates. This result suggests that Medicare may be reducing racial disparities in breast carcinoma survival. In addition, the racial differences in stage distribution indicate that black females of every age need to focus more on early diagnosis.