Widening disparity in survival between white and African-American patients with breast carcinoma treated in the U. S. Department of Defense Healthcare system†
The opinions expressed in this article represent the private views of the authors and should not be construed as reflecting the official views of the Departments of the Army, Air Force, or Defense.
In the U. S., age-adjusted breast carcinoma mortality rates among white and African-American women have been diverging during the last 20 years. Some investigators speculate that the widening disparity is due to inequalities in access to healthcare, with African Americans having less access to necessary healthcare and improved therapies. Others argue that differences in tumor biology or some extrinsic influences on cancer etiology and behavior may account for the widening disparity. To examine this issue further, the authors compared trends in survival among white and African-American women diagnosed with breast carcinoma in the U. S. Department of Defense (DoD), an equal access healthcare system.
The medical records of all women diagnosed with primary breast carcinoma between 1980–1999 were retrieved from the U. S. DoD Automated Central Tumor Registry (ACTUR). Variables selected for further analysis were date of diagnosis, date of birth, vital status and date of death if applicable, race (black, white, and others), and stage of tumor. Because the database does not contain causes of death, overall survival was investigated. The effect of year of diagnosis and race on overall survival was analyzed using the Cox proportional hazards model stratified by age at diagnosis (1-year age groups). Calculations were performed separately by disease stage for all stages combined and stratified by stage. Statistical analyses were performed using the statistical software package SAS.
After deleting observations with missing or implausible information regarding patient age, gender, and follow-up time, the final dataset was comprised of 23,612 women with primary breast carcinoma. The survival of African-American women compared with white women demonstrated an increasing ratio with calendar period. Although the hazard ratio was 1.269 for women diagnosed with breast carcinoma during the calendar period 1980–1984, it increased to 1.849 for those diagnosed between 1995–1999, which is a ratio of 1.46. For this period, the interaction between race and period was found to be significant (P = 0.04).
The results of the current study demonstrated that breast carcinoma survival rates among white and African-American patients, adjusted for age and stage, are diverging in the U. S. DoD healthcare system. Thus, inequalities in access to healthcare most likely are not solely responsible for the widening racial disparities in outcome reported among women diagnosed with breast carcinoma. Cancer 2003;98:894–9. © 2003 American Cancer Society.
In the U. S., a widening disparity in breast carcinoma mortality rates between white and African-American women has been reported over the past 20 years.1, 2 There are at least two possible reasons for this observation. Some investigators argue that differences in tumor biology or extrinsic influences on cancer etiology and behavior account for the increasing disparity.3, 4 However, others attribute the widening racial disparity to the fact that African-American women often receive less than adequate healthcare.1, 5, 6 Indeed, health insurance coverage appears to be an important factor in determining outcome after a diagnosis of breast carcinoma, and a significant proportion of African Americans lack health insurance coverage or have inadequate coverage, which may limit their access to necessary medical care and improved therapies.7
To our knowledge, no studies to date have examined survival trends among white and African-American women with breast carcinoma in an equal access healthcare system. We therefore undertook the current study to compare age-adjusted breast carcinoma survival trends among white and African-American women in the U. S. Department of Defense (DoD) Healthcare system. All beneficiaries of the DoD system receive free and equal access to healthcare. Thus, disparities in outcome after a diagnosis of breast carcinoma should not be ascribed to inequalities in access to healthcare. We compared survival trends for white and African-American women diagnosed with breast carcinoma during the calendar periods 1980–1984, 1985–1989, 1990–1994, and 1995–1999.
When discussing access to healthcare, one should consider two broad issues. One pertains to early detection and prevention services and the other to specific breast carcinoma treatments such as adjuvant therapy. Both will influence breast carcinoma mortality rates, in which the denominator is the entire population at risk and the numerator is the number of deaths caused by breast carcinoma. However, trends in breast carcinoma mortality do not always correlate with trends in overall survival.8 Overall survival refers to the period of time from the diagnosis of cancer to death. Increases in overall survival may simply represent the effect of finding the disease earlier (stage migration), rather than postponement in the time of death from more effective treatments. To account for the effect of stage migration, we also calculated stage-adjusted breast carcinoma survival trends within the DoD system. Stage-adjusted breast carcinoma survival rates are largely influenced by specific breast carcinoma treatments such as adjuvant therapy and not by early detection and prevention services.
MATERIALS AND METHODS
The medical records of all women diagnosed with breast carcinoma during the period 1980–1999 were retrieved from the U. S. DoD Automated Central Tumor Registry (ACTUR). All patients in the study were beneficiaries of the DoD health care system and were entitled to equal access to medical care, either free of charge or at a nominal fee, at military medical facilities.
All women with a diagnosis of primary breast carcinoma were selected from the database. Variables selected for further analysis were date of diagnosis, date of birth, vital status and date of death if applicable, race (black, white, and others), and stage and grade of tumor. The few cases of male breast carcinoma in the database were not considered.
Because the database does not contain causes of death, overall survival was investigated.
The effect of year of diagnosis and race on overall survival was analyzed using the Cox proportional hazards model stratified by age at diagnosis (1-year age groups). Calculations were performed separately by stage, for all stages combined, and stratified by stage.9 Statistical analyses were performed using the statistical software package SAS (SAS Institute, Inc., Cary, NC).
The original dataset contained 24,393 observations. After deleting those observations with missing or implausible information concerning age, gender, and follow-up time, 23,612 women remained for further analysis. Table 1 provides a description of some variables in the study population. The median age at the time of breast carcinoma diagnosis was 56 years.
Table 1. Distribution of Age and Race
|Race|| || |
Table 2 shows the distribution of disease stage by race for blacks and whites (others not included) overall and by calendar period. The stage distribution shows that white women were diagnosed at an earlier stage of disease compared with black women. However, for a large percentage of women the stage of disease was unknown, and this was particularly true for women diagnosed with breast carcinoma during the years 1980–1989. In the calendar periods 1990–1994 and 1995–1999, this percentage was approximately 12%, and there was a stronger tendency toward the earlier detection of breast carcinoma in white women.
Table 2. Distribution of Disease Stage at the Time of Diagnosis by Race
|White|| || || || || || || |
|Black|| || || || || || || |
|Total|| || || || || || || |
Table 3 shows the results of the Cox regression analysis which is based on 4795 deaths. The reference calendar period was 1980–1984. We observed an overall improvement in survival over calendar period studied (model “all stages combined, all women”) with hazard rate ratios declining from 0.95 (P = 0.2) for the period 1985–1989 to 0.89 (P = 0.08) for the period 1995–1999. This decline was found to be stronger in white women (model “all stages combined, whites”), with hazard rate ratios declining from 0.95 (P = 0.21) for the period 1985–1989 to 0.84 (P = 0.02) for the period 1995–1999 compared with black women (model “all stages combined, blacks”), in whom little overall improvement in survival was observed.
Table 3. Relative Survival for Calendar Period by Race and Stage
| || ||1990–1994||1.300||0.474–3.566||0.6106|
| || ||1995–1999||1.105||0.348–3.510||0.8660|
| || ||1990–1994||0.997||0.787–1.262||0.9784|
| || ||1995–1999||1.374||0.947–1.995||0.0945|
| || ||1990–1994||1.104||0.926–1.315||0.2717|
| || ||1995–1999||1.104||0.833–1.461||0.4914|
| || ||1990–1994||1.533||1.249–1.880||< 0.0001|
| || ||1995–1999||1.482||1.119–1.964||0.0060|
| || ||1990–1994||1.296||1.126–1.493||0.0003|
| || ||1995–1999||1.376||1.090–1.736||0.0073|
| ||All stages “combined”||1985–1989||0.953||0.885–1.026||0.2021|
| || ||1990–1994||0.897||0.828–0.972||0.0078|
| || ||1995–1999||0.891||0.782–1.014||0.0811|
| ||Stratified by “stage”||1985–1989||1.082||1.001–1.170||0.0475|
| || ||1990–1994||1.214||1.113–1.324||< 0.0001|
| || ||1995–1999||1.299||1.133–1.489||0.0002|
| || ||1990–1994||1.337||0.482–3.709||0.5774|
| || ||1995–1999||0.905||0.270–3.029||0.8715|
| || ||1990–1994||1.082||0.839–1.396||0.5420|
| || ||1995–1999||1.537||1.029–2.297||0.0357|
| || ||1990–1994||1.069||0.878–1.301||0.5053|
| || ||1995–1999||0.990||0.711–1.380||0.9546|
| || ||1990–1994||1.405||1.116–1.768||0.0038|
| || ||1995–1999||1.242||0.891–1.730||0.2003|
| || ||1990–1994||1.286||1.097–1.508||0.0019|
| || ||1995–1999||1.323||1.009–1.733||0.0425|
| ||All stages “combined”||1985–1989||0.950||0.876–1.030||0.2134|
| || ||1990–1994||0.883||0.808–0.965||0.0062|
| || ||1995–1999||0.842||0.724–0.980||0.0260|
| ||Stratified by “stage”||1985–1989||1.067||0.979–1.163||0.1375|
| || ||1990–1994||1.189||1.079–1.310||0.0005|
| || ||1995–1999||1.198||1.022–1.405||0.0257|
| || ||1990–1994||1.117||0.093–13.364||0.9304|
| || ||1995–1999||.||.||.|
| || ||1990–1994||2.024||0.519–7.890||0.3099|
| || ||1995–1999||1.701||0.294–9.855||0.5532|
| || ||1990–1994||1.577||0.855–2.908||0.1448|
| || ||1995–1999||1.359||0.594–3.106||0.4674|
| || ||1990–1994||2.395||1.078–5.321||0.0320|
| || ||1995–1999||3.222||1.277–8.125||0.0132|
| || ||1990–1994||0.980||0.598–1.606||0.9356|
| || ||1995–1999||2.054||1.044–4.042||0.0371|
| ||All stages “combined”||1985–1989||0.958||0.737–1.246||0.7508|
| || ||1990–1994||0.920||0.704–1.204||0.5449|
| || ||1995–1999||0.972||0.679–1.392||0.8776|
| ||Stratified by “stage”||1985–1989||1.126||0.830–1.529||0.4450|
| || ||1990–1994||1.391||1.010–1.916||0.0430|
| || ||1995–1999||1.749||1.140–2.684||0.0105|
However, for disease stages analyzed separately, there was a slight worsening (for the most part not significant) in survival for the majority of disease stages over the calendar period in white women, and a worse survival in black women that was significant for AJCC Stages III and IV disease. In an overall analysis stratified for stage, a significantly worse survival was found with hazard rate ratios of 1.20 (95% confidence interval [95% CI], 1.02–1.41) and 1.749 (95%CI, 1.14–2.68) for 1995–1999 versus 1980–1984, for white and black women, respectively. The overall improvement in survival was the result of a shift in stage over the calendar period (earlier diagnosis) as noted in Table 2. This shift was more pronounced in white women compared with black women.
The survival of black women compared with white women showed an increasing ratio with calendar period as presented in Table 4. Although the hazard ratio was 1.269 in 1980–1984, it increased to 1.849 in 1995–1999, a ratio of 1.46. For this period, the interaction between race and period was found to be significant (P = 0.04).
Table 4. Survival of Blacks versus Whites over Calendar Periods, Stratified by Stage
|Blacks vs. whites||1980–1984||1.269||1.03–1.57||0.03|
| ||1995–1999||1.849||1.38–2.48||< 0.001|
Age at the time of diagnosis appeared to have little impact on the result. If we considered women age < 60 years separately (a group that has a much lower mortality from other diseases), very similar results were found. For example, the hazard rate ratio for the period 1990—1995 compared with that for 1980–1984 was 1.27 (P < 0.001) for women age < 60 years and was 1.21 for all women (P < 0.001).
The dataset provided information regarding initial surgical treatment, and showed little difference between African-American women and white women. The percentage of patients who were treated initially with total mastectomy with or without breast reconstruction was 62.2% for African-American women and 62.0% for white women, whereas the percentage of patients treated with breast-conserving surgery was 27.4% and 24.6%, respectively. The remaining patients either did not undergo any surgery or no information concerning the type of surgical procedure was available. In addition, the dataset indicated that the proportion of women who underwent dissection of the axillary lymph nodes was virtually identical in the two groups.
The dataset also provided information regarding systemic therapy, with no indication as to whether it was administered in the adjuvant setting. A higher percentage of African-American women received chemotherapy (43.6% vs. 29.5%), whereas a higher percentage of white women received hormonal therapy (24.1% vs. 16.9%). The remaining patients either received no systemic therapy or no information was provided concerning systemic therapy. African-American women are reportedly more likely to present with lymph node-positive, estrogen receptor–negative tumors, and this might account for the increased use of chemotherapy and decreased use of hormonal therapy found in this group of patients.
The DoD database contains no information regarding screening mammography or breast carcinoma follow-up procedures.
U. S. population statistics indicate that, since approximately 1981, there has been a widening disparity in breast carcinoma mortality rates between white and African-American women.1 Some investigators attribute the widening racial disparity to inequalities in access to healthcare.1, 5, 6 However, in the current study, we found that the racial disparity in breast carcinoma survival, when adjusted for age and stage, also was increasing in the DoD system, an equal access healthcare system. These results suggest that inequalities in access to healthcare most likely are not solely responsible for the widening disparity. We can only speculate as to what other factors might be responsible. Some investigators suggest that diet and obesity may play an important role.10 Indeed, the incidence of obesity among African-American women is increasing, and obesity may raise the levels of circulating endogenous estrogen, thereby contributing to a poorer prognosis after a diagnosis of breast carcinoma.11, 12
In addition, the use of adjuvant systemic therapy for breast carcinoma has become widespread in recent years, and racial differences in the response to drugs have been reported.13 For instance, adjuvant tamoxifen therapy is particularly effective in women with estrogen receptor-positive (ER+) tumors, and white women are reported to have a greater proportion of ER+ tumors than African-American women.14 It is important to emphasize that tamoxifen is not any less effective in African-American women. Rather, fewer African-American breast carcinoma patients are eligible for tamoxifen because of the lower incidence of ER+ tumors in this population. Thus, the widening disparity noted in survival outcomes between white and African-American women might be attributable at least in part to the introduction of tamoxifen as adjuvant therapy for breast carcinoma, and its increased use in recent years. An analysis of the National Surgical Adjuvant Breast and Bowel Project (NSABP) breast carcinoma treatment trials found no differences in breast carcinoma mortality between white and African-American patients enrolled in clinical trials.15 However, the NSABP only enrolled patients with ER+ tumors in their adjuvant tamoxifen therapy trials.
We cannot exclude racial bias as another possible cause for the widening disparity in outcomes between white and African-American breast carcinoma patients in the DoD system. Racial bias in medical care is difficult to identify, but has been reported in other equal access healthcare systems. For example, racial bias in the use of invasive cardiovascular procedures was reported in the Department of Veterans Affairs Medical System.16 In that study, white patients were more likely than African-American patients to undergo cardiac catheterization. Similarly, Chen et al. found that among Medicare patients, white patients were significantly more likely than African-Americans to undergo cardiac catheterization.17 Surprisingly, this racial difference in care did not vary according to whether patients were treated by black or white physicians. Racial bias in the treatment of breast carcinoma might, for example, include differences in the types of treatments administered, delays in the initiation of therapy, or even failure to administer adjuvant therapy in some patients.
Population statistics indicate that the age-adjusted breast carcinoma mortality rate in the U.S. has declined by approximately 16% since 1990.18 Nevertheless, we did not note overall improvements in breast carcinoma survival rates in the DoD system (adjusted for age and stage) during the four calendar periods of this study. There are several possible explanations for this observation. First, as previously noted, survival trends do not always correlate with mortality trends.8 Trends in cancer mortality are closely linked to trends in cancer risk and incidence, which bear no relationship to trends in survival. The dramatic increase in breast carcinoma incidence in recent years has been attributed largely to the increased use of mammography screening, which results in the detection of some nonprogressive tumors.19, 20 However, the increase in incidence attributable to mammography screening was less than predicted, suggesting that there has been a true decline in the inherent risk of developing the disease.21 Thus, the recent decline in breast cancer mortality in the U.S. might be explained in part by a reduction in breast carcinoma risk, which would not be reflected in the survival statistics.
Second, failure to observe an improvement in survival might also be attributable in part to the survival endpoint used in the current study. The national mortality statistics report deaths due to breast carcinoma (“disease-specific” mortality). However, in our database, we were not able to distinguish between patients who died of breast carcinoma and those who died of other causes. We therefore measured overall survival after a diagnosis of breast carcinoma, rather than disease-specific survival. Therefore, small improvements in disease-specific survival might be difficult to discern in the current study, particularly if there were increases in deaths from other causes.
Finally, there is considerable variation in breast carcinoma mortality and survival rates reported in different countries, regions within countries, hospitals, and practice settings.22, 23 For example, Sainsbury et al. report better survival rates among surgeons who treat > 30 new cases of breast carcinoma each year and have a full range of treatment options available in a multidisciplinary setting.24 Surgeons who treat fewer cases were reported to have worse outcomes. A recent study also demonstrated that breast carcinoma patients treated in large referral centers by surgeons specializing in cancer care appear to do better than those treated outside such centers.25 Thus, failure to observe improvements in breast carcinoma survival during the past 20 years in the DoD system also might be attributable in part to the practice setting, in which many patients are treated outside large referral centers.
A proportion of the decline noted in breast carcinoma mortality in the U.S. has been attributed to adjuvant systemic therapy.26 The widespread use of adjuvant systemic therapy was initiated in the 1980s after the publication of results of randomized trials demonstrating that it reduced breast carcinoma mortality by approximately 25%.27 Olivotto et al. reported that the increased use of adjuvant systemic therapy increased both disease-specific and overall breast carcinoma survival rates among women in the Canadian province of British Columbia.28 These authors compared survival rates from 3 different years: 1974, 1980, and 1984. However, they did not adjust for stage of disease at the time of presentation, and at least some of the improvement in survival during the three calendar periods in their study might be attributable to stage migration. Indeed, we also found improvement in overall survival rates during the four calendar periods of the current study, but this effect disappeared after stratification for stage of disease at the time of presentation.
Despite our failure to note overall improvements in breast carcinoma stage-adjusted survival, the results of the current study do demonstrate that disparity in survival rates between white and African-American patients has increased in the DoD healthcare system. These results suggest that the widening racial disparity in breast carcinoma outcomes in the U.S. cannot be ascribed entirely to inequalities in access to healthcare. Nevertheless, inequalities in access to healthcare are almost certainly an important contributing factor. Indeed, previous studies have shown that African-American individuals in the DoD healthcare system have better survival rates from carcinomas of the prostate and breast than African Americans in the population as a whole.29, 30 Therefore, inequalities in healthcare appear to contribute to the racial disparity in outcomes among women with breast carcinoma, and efforts to improve access to healthcare care should remain a top priority. However, there also clearly is a need to identify other factors responsible for the widening racial disparity in outcome in patients with breast carcinoma.