SEARCH

SEARCH BY CITATION

Keywords:

  • breast carcinoma;
  • survival;
  • race/ethnicity;
  • SEER;
  • socioeconomic status

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND

Although overall survival for invasive breast carcinoma remains high, black women experience poorer survival than whites. Less is known about the survival of Hispanics and Asians, who may share clinical and socioeconomic risk factors similar to blacks. To better understand racial/ethnic survival patterns, we investigated the effect of socioeconomic status (SES) and disease stage on racial/ethnic differences in breast carcinoma survival in a large population-based cohort.

METHODS

Using data from the Surveillance, Epidemiology, and End Results program (SEER), we identified 10,414 white, 940 black, 1100 Hispanic, and 1180 Asian females diagnosed with breast carcinoma in the Greater San Francisco Bay Area between 1988 and 1992. We used the Kaplan–Meier method to generate survival rates and Cox proportional hazards regression to estimate the risk of death by race/ethnicity, after adjustment for clinical, demographic, and census-derived SES variables.

RESULTS

The 10-year unadjusted survival rates were 81% for whites, 69% for blacks, 75% for Hispanics, and 79% for Asians. Adjusting for stage decreased the relative risk of mortality for blacks from 1.81 to 1.29; the stage-adjusted relative risk for Hispanics (1.11) and Asians (1.02) did not differ significantly from whites. Additional adjustment for age, tumor characteristics, and treatment factors did little to alter the relative risk in blacks; adding blue-collar status to the model further decreased the relative risks for blacks to 1.22. Residing in a blue-collar neighborhood was independently associated with a 1.16 increase in risk of death.

CONCLUSIONS

After adjustment for multiple factors, blacks continue to have slight but significantly poorer survival after breast carcinoma compared with whites, whereas the survival of Hispanics and Asians did not differ from whites. Cancer 2003;97:1303–11. © 2003 American Cancer Society.

DOI 10.1002/cncr.11160

Among all women in the United States, breast carcinoma is the most frequently occurring cancer and the second leading cause of cancer death.1 Although white women have the highest incidence rates of breast carcinoma, blacks have the highest mortality rates.2 Black women typically present with late-stage disease, the strongest prognostic factor for decreased survival.3 Poor survival is also associated with socioeconomic status (SES), access to care, age at diagnosis, and more aggressive tumor characteristics.4–6 Some studies report that after controlling for such factors, survival differences between blacks and whites are no longer statistically significant.7–11 Therefore, race/ethnicity may be a predictor of survival as a surrogate for other characteristics rather than as a biologic risk factor. Other studies, however, continue to show decreased survival for blacks even after adjusting for these other factors.5, 6, 12

Although the black/white survival disparity in breast carcinoma is well documented, few studies of racial/ethnic disparities have addressed the survival experience of other racial/ethnic groups. Like black women, Hispanic women also tend to be diagnosed with breast carcinoma at a later stage than white women, but it is unclear if Hispanic females have poorer survival.9, 12–15 International studies of breast carcinoma have shown better survival for Japanese than Caucasians,16 but most studies report similar survival rates for Asians and whites within the United States.12, 17–19 In addition, although race/ethnicity, disease stage, and SES are correlated, SES measures have not been included in most studies of Asian and Hispanic women. Previous studies of survival by race/ethnicity were limited by the lack of data on SES and tumor characteristics, inclusion of in situ disease with local stage disease, and inadequate statistical power due to small sample sizes. To overcome these limitations, we evaluated survival following breast carcinoma in a large population-based cohort of women from four racial/ethnic groups, particularly adding to the literature for Hispanic and Asian American women. Specifically, we asked the following questions: First, Does the black/white survival disparity persist after adjusting for disease stage, tumor characteristics and SES? Second, is there a survival differential for Hispanics and Asians, after adjustment for disease stage, tumor characteristics, and SES?

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Study Population

Breast carcinoma cases were identified through the population-based Greater San Francisco Bay Area Cancer Registry, a participant in both the National Cancer Institute's Surveillance, Epidemiology, and End Results program (SEER) and the California Cancer Registry. The registry covers more than 6 million residents living in nine counties in the San Francisco-Oakland and San Jose, California, region. We restricted eligibility to women diagnosed in between 1988 and 1992 to enable linkage with the 1990 Census and to allow a sufficient follow-up period. During the period 1988–1992, 18,425 females were newly diagnosed with primary, histologically confirmed, invasive breast carcinoma. We restricted study eligibility to females with adenocarcinoma and for whom breast carcinoma was the only known cancer (n = 15,667), thereby eliminating survival differences due to the effects of other cancers. After excluding females with unknown values for stage at diagnosis, race/ethnicity, census block group, or survival time, and those diagnosed at autopsy or identified by death certificate, the final sample included 13,634 females (87% of eligible patients). Most of the excluded patients were missing American Joint Committee on Cancer (AJCC) staging information. There was no difference in race/ethnicity, age, vital status, or the SES measures between the included and excluded groups.

Patients and Tumor Characteristics

The registry provided routinely collected demographic data including age at diagnosis, race/ethnicity, and residence at diagnosis. Race/ethnicity, based on the medical record, was categorized into four mutually exclusive groups: non-Hispanic whites (whites), non-Hispanic blacks (blacks), Hispanics (Hispanics), and non-Hispanic Asians/Pacific Islanders (APIs). We did not divide the API category into subgroups to maintain an adequate sample size for the multivariate comparisons. As the registry does not collect individual measures of SES, we used the patient's residence at diagnosis to derive surrogate measures based on 1990 U.S. Census data for block groups. A subdivision of census tracts, block groups are a collection of census blocks, the smallest geographic unit of measurement in the census. Block groups contain an average of 1000 residents and tend to be more homogeneous than census tracts. Using the race-specific measures for each block group, we created dichotomous variables for low/not low education (with a cutpoint of 25% of residents older than age 25 not receiving a high school diploma as low education), low/not low income (cutpoint of 20% of residents living below the poverty line as low income), and “blue collar” if 66% or more of the residents were employed in blue-collar jobs. Census-based measures are commonly used in public health research. These three measures with the specific cutpoints were selected because they have been validated in previous studies of breast carcinoma in the San Francisco Bay Area.20, 21 To further investigate the impact of poverty, the census measure of median income was used both as a continuous variable and as a dichotomous variable cut at the median value for each race/ethnicity. As results using median income did not differ from those using the poverty level measure, we only report on the poverty level variable.

The SEER program also routinely collects data on the extent of disease at diagnosis, tumor characteristics, and the first course of treatment. Stage at diagnosis was defined according to the AJCC system.22 As preliminary results showed no survival difference between Stages IIIA and IIIB, we combined the two to improve statistical power. To minimize the effect of disease heterogeneity within stage, tumor size and number of positive lymph nodes were also incorporated into the models. We also included the following clinical covariates: histology; grade; estrogen receptor (ER) status; progesterone receptor (PR) status; radiation; and surgery, categorized as breast-conserving (BCS; segmental mastectomy, lumpectomy, quadrantectomy, tylectomy, wedge resection, excisional biopsy, partial mastectomy), mastectomy, or none. As data on chemotherapy and hormonal therapy were not consistently found in the hospital record, which is the primary source of treatment information for the registry, they were not included in the analysis.

The registry determines vital status through passive (linkages with state and national death indices, Health Care Financing Administration files, Department of Motor Vehicles records, and voter registration files) and active (contacting physicians' offices, hospitals, and patients) follow-up procedures. All cases were followed through July 2001. Survival time was measured in months from the time of diagnosis to death or censoring. The outcome variable was death from breast carcinoma. Deaths from other or unknown causes were censored at the time of death. To explore the potential misclassification of breast carcinoma as the cause of death on death certificates, we repeated all analyses with all causes of death as the outcome. When all-cause death was substituted as the outcome, the results for race/ethnicity and the SES variables were minimally changed. As anticipated, the risk estimates associated with stage and grade decreased whereas the impact of age increased. Therefore, we only report on the associations with death from breast carcinoma.

Statistical Analyses

We used the Mantel–Haenszel chi-square test to investigate the bivariate relationships between race/ethnicity and patient/tumor characteristics. Survival curves were generated using Kaplan–Meier estimates and differences were compared using the log rank test. We used Cox proportional hazards regression to 1) assess the impact of race/ethnicity on the risk of death due to breast carcinoma while simultaneously adjusting for SES, stage, clinical characteristics, and age; and 2) evaluate the independent effects of race/ethnicity, SES, stage, and grade on survival while controlling for ER status, PR status, type of surgery, radiation, and age. The resulting hazards ratios (HR), which measure the risk of mortality after adjustment for other factors, are presented with their 95% confidence intervals (CI). Interaction terms were included in the models to explore potential interaction between race/ethnicity and each of the census-based SES measures. Statistical computing was performed by using STATA statistical software, version 7.0 (Stata, College Station, TX) and SAS version 6.12 (SAS Institute, Cary, NC).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Subject Description

Table 1 displays the characteristics of the 13,634 patients. Although the majority of females were white, there were substantial numbers of patients in each of the other three racial/ethnic groups: 940 blacks, 1100 Hispanics, and 1180 APIs. All patient and disease characteristics differed significantly by race/ethnicity, but the racial/ethnic group that differed was not the same for each characteristic. Whites were significantly older at diagnosis than other racial/ethnic groups. Although the majority of the patients were diagnosed with early-stage disease (76% were AJCC Stages I and IIA), stage at diagnosis varied considerably by race/ethnicity. For example, 78% of whites, 64% of blacks, 70% of Hispanics, and 71% of APIs presented with Stage I or IIA disease.

Table 1. Relative Frequencies of Characteristics of Females Diagnosed with Invasive Breast Carcinoma in the Greater San Francisco Bay Area, SEER Data, 1988–1992, by Race/Ethnicitya
CharacteristicPercent White (N = 10,414)Percent Black (N = 940)Percent Hispanic (N = 1100)Percent API (N = 1180)Total (N = 13,634)
  • SEER: Surveillance, Epidemiology, and End Results program; API: Asian/Pacific Islander

  • a

    Percentages may not sum to 100 due to rounding. All characteristics were significantly different by race/ethnicity (P < 0.02).

  • b

    American Joint Committee on Cancer staging converted from SEER extent of disease.

  • c

    Living in 1990 census block group with 66% or greater proportion of adults (16 + years old) working in blue-collar jobs.

  • d

    Living in 1990 census block group with 25% or greater proportion of adults (25 + years old), of the same racial/ethnic group, with no high school diploma.

  • e

    Living in 1990 census block group with 20% or greater proportion of residents, of the same racial/ethnic group, living below poverty line.

  • f

    Vital status followed through August 2001.

Mean age (SD)61.8 (14.4)56.7 (14.5)56.8 (14.7)55.5 (13.9) 
Age at Diagnosis (yrs)     
 0–341.95.06.65.42.8
 35–4921.930.027.733.123.9
 50–6430.331.833.032.930.8
 65–7424.919.819.419.223.6
 75 +21.113.413.39.418.9
AJCC stage at diagnosisb     
 I47.733.736.839.645.2
 IIA30.029.932.632.030.3
 IIB12.817.916.416.113.7
 III5.411.39.27.66.3
 IV4.17.25.14.84.4
Extent of lymph node involvement     
 Negative lymph nodes55.246.148.951.753.8
 1–3 positive lymph nodes18.219.520.720.318.7
 4+ positive lymph nodes12.417.015.914.013.2
 Unknown14.217.514.514.014.4
Tumor size (cm)     
 ≤ 261.746.552.052.959.0
 2.1–5.028.738.134.935.630.0
 > 5.04.78.66.87.05.0
 Direct extension1.32.51.91.91.0
 Unknown size3.74.44.42.64.0
Histology     
 Infiltrating duct74.772.872.677.574.6
 Lobular7.74.77.22.87.0
 Infiltrating duct and lobular4.33.04.23.14.1
 Inflammatory carcinoma1.01.82.11.31.2
 Other12.317.814.015.313.1
Grade     
 Well differentiated6.93.82.95.76.2
 Moderately differentiated26.223.422.326.125.7
 Poorly differentiated24.133.229.629.225.6
 Undifferentiated/anaplastic1.52.63.11.01.7
 Unknown41.337.042.138.140.8
Estrogen receptor status     
 Positive36.827.533.637.517.1
 Negative/borderline16.121.519.619.836.0
 Unknown47.151.146.742.647.0
Progesterone receptor status
 Positive30.723.427.833.721.9
 Negative/borderline21.325.024.822.130.2
 Unknown48.051.647.444.247.9
Initial surgery type     
 No surgery done2.45.53.03.03.0
 Breast-conserving34.934.028.525.134.0
 Mastectomy62.760.468.672.064.0
Radiation     
 None63.361.064.867.164.0
 Yes36.739.035.232.936.0
Blue-collar status, community levelc     
 No76.439.752.359.870.5
 Yes23.660.347.740.229.5
Low education, community leveld     
 No89.650.433.764.280.2
 Yes10.449.666.335.819.8
In poverty, community levele     
 No97.357.379.785.392.1
 Yes2.742.720.314.87.9
Vital statusf     
 Dead40.750.341.032.040.6
 Alive59.349.759.068.059.4
Death due to breast carcinoma     
 No83.172.577.281.081.7
 Yes16.927.622.819.018.3

Histologically, 75% of the females had infiltrating duct adenocarcinoma. Grade, ER status, and PR status were missing for a substantial number of study subjects, as these variables were not routinely available until 1990. Of women with known ER status, blacks were more likely to be ER negative (44%) than any of the other racial/ethnic groups (whites, 30%; Hispanics, 37%; APIs, 35%; P = 0.001). Almost two-thirds of the patients had mastectomies. Hispanics and APIs were less likely to have BCS than either whites or blacks. Radiation therapy was administered to 36% of the women, but only 79% of patients with BCS received radiation treatment (data not shown). Among women with BCS, receipt of radiation therapy did not differ by race/ethnicity, with 78–81% of women receiving radiation therapy (P = 0.87).

Census-based SES analyses showed that, of all patients, 30% lived in blue-collar neighborhoods, 20% in low education neighborhoods, and 8% in poor neighborhoods. All three SES measures were significantly associated with race/ethnicity (Table 1). Black patients were significantly more likely to live in blue-collar, low education, or low income neighborhoods than whites, whereas Hispanic patients were significantly more likely to live in low education neighborhoods. In addition, we found significant associations between SES and stage of disease. Subjects in low income neighborhoods were significantly more likely to be diagnosed with Stage IIB or higher disease than were those who did not live in low income neighborhoods (33% vs. 24%, P = 0.001). Similarly, a higher percentage of women with late-stage disease lived in the low education (30% vs. 23%, P = 0.001) and blue-collar (29% vs.23%, P = 0.001) neighborhoods.

During the follow-up period of January 1988 through July 2001, 5539 (41%) subjects died (Table 1) and breast carcinoma was listed as the underlying cause of death for 2491(18%). Figure 1 shows the unadjusted survival curve by the four racial/ethnic groups, confirming the generally high long-term survival following this disease. Whites had the highest breast carcinoma survival, whereas blacks and Hispanics experienced significantly lower rates of survival (P < 0.0001).The survival rate for APIs did not differ from that of whites. Unadjusted 5-year survival rates were 88%, 77%, 84%, and 86% for whites, blacks, Hispanics, and APIs, respectively and the 10-year rates were 81%, 69%, 75%, and 79%. Adjusting for stage of disease decreased the survival difference across race/ethnicity, but a slight survival deficit remained significant for blacks (Fig. 2). The 5-year stage-adjusted survival rates were 96% for whites, Hispanics, and APIs and 94% for blacks. The 10-year rates were 93% for both whites and APIs, 92% for Hispanics, and 91% for blacks.

thumbnail image

Figure 1. Breast carcinoma survival estimates, by race/ethnicity, Greater San Francisco Bay Area, Surveillance, Epidemiology, and End Results program data, 1988–1992. API: Asian/Pacific Islander.

Download figure to PowerPoint

thumbnail image

Figure 2. Breast carcinoma survival estimates, by race/ethnicity, adjusted for stage, Greater San Francisco Bay Area, Surveillance, Epidemiology, and End Results program data, 1988–1992. API: Asian/Pacific Islander.

Download figure to PowerPoint

The impact of the demographic and clinical characteristics on the HRs for race/ethnicity is reported in Table 2. These HRs reflect the excess risk of death for black, Hispanic, and API race/ethnicity, relative to a white reference group. In unadjusted analyses, black or Hispanic race/ethnicity was associated with an increased risk of mortality from breast carcinoma. To understand the impact of confounders on race/ethnicity, we added stage to the Cox model. After adjustment for stage, the excess risks associated with race/ethnicity decreased from 81% to 29% (HR of 1.81 to 1.29) for blacks, but the stage-adjusted excess risks for Hispanics (11%) or APIs (2%) were not significantly different from whites. Subsequent models stepped in the clinical variables and age, followed by a final model of race/ethnicity, stage, clinical characteristics, age, and SES measures. Additional adjustment for age and for tumor and treatment factors did little to alter the excess risk among blacks. The addition of blue-collar status to the models further decreased the excess risk to 22%, whereas the other SES measures did not alter the HR. In contrast, although the stage-adjusted risk for Hispanics was not significantly different from that for whites, the addition of age, clinical characteristics, and SES variables continued to decrease the risk to 1.01. For APIs, the addition of other risk factors brought the risk to 0.96.

Table 2. Effect of Demographic and Clinical Characteristics on Racial/Ethnic Survival Estimates of Females Diagnosed with Invasive Breast Carcinoma in the Greater San Francisco Bay Area, SEER Data, 1988–1992 (n = 13,634)
Variables in modelHazards ratio (95% confidence interval)
WhiteBlackHispanicAPI
  • SEER: Surveillance, Epidemiology, and End Results program; API: Asian/Pacific Islander.

  • a

    Clinical variables include histology, grade, surgery, radiation therapy, estrogen receptor status, and progesterone receptor status.

  • b

    Socioeconomic status variables include neighborhood measures of blue collar, low education, and below poverty status.

Race/ethnicity only1.01.81 (1.59–2.07)1.39 (1.22–1.59)1.14 (0.99–1.31)
Race/ethnicity and stage1.01.29 (1.13–1.47)1.11 (0.97–1.27)1.02 (0.89–1.17)
Race/ethnicity, stage, clinical,a age1.01.28 (1.12–1.46)1.05 (0.92–1.20)0.98 (0.85–1.13)
Race/ethnicity, stage, clinical,a age, and SESb1.01.22 (1.05–141)1.01 (0.87–1.17)0.96 (0.83–1.11)

Table 3 summarizes the results of the proportional hazards model, which looked individually at each variable (race/ethnicity, stage, grade, and SES variables) while simultaneously controlling for these variables and other clinical and demographic factors. After adjustment, the risk of death from breast carcinoma remained significantly higher for blacks than whites, whereas there was no mortality difference for Hispanics and APIs compared with whites (the same HRs are shown in Table 2). Clinical variables, particularly stage, exerted the strongest impact on mortality. Subjects diagnosed in Stages II and III had two to four times the risk of death, whereas those in Stage IV had nearly 14 times the risk of death from breast carcinoma as those in Stage I. Compared with subjects with well differentiated tumors (Grade 1), increasing grade was significantly associated with increased risk of death from breast carcinoma. The only SES measure independently associated with death was residing in a blue-collar neighborhood, with a modest increase in mortality. No interaction was found between race/ethnicity and SES measures.

Table 3. Multivariate Proportional Hazards Results of Females Diagnosed with Invasive Breast Carcinoma in the Greater San Francisco Bay Area, SEER Data, 1988–1992 (n = 13,634)a
FactorHazard ratio95% Confidence interval
  • a

    Adjusted for other factors, including age at diagnosis, histology, tumor size, number of lymph nodes, estrogen and progesterone receptor status, surgery, and radiation therapy.

  • b

    American Joint Committee on Cancer staging converted from SEER extent of disease.

  • c

    Living in 1990 census block group with 66% or greater proportion of adults (16 + years old) working in blue-collar jobs.

  • d

    Living in 1990 census block group with 25% or greater proportion of adults (25 + years old), of the same racial/ethnic group, with no high school diploma.

  • e

    Living in 1990 census block group with 20% or greater proportion of residents, of the same racial/ethnic group, living below poverty line.

Race/ethnicity  
 White (referent)1.00
 Black1.221.05–1.41
 Hispanic1.010.87–1.17
 Asian/Pacific Islander0.960.83–1.11
AJCC stageb  
 I (referent)1.00
 IIA2.422.03–2.89
 IIB3.572.75–4.62
 III4.773.71–6.14
 IV13.8311.02–17.36
Grade  
 Well differentiated (referent)1.00
 Moderately differentiated1.751.27–2.42
 Poorly differentiated2.922.12–4.02
 Undifferentiated2.661.79–3.94
 Unknown1.991.45–2.74
Blue-collar, community levelc  
 No1.00
 Yes1.161.05–1.27
Low education, community leveld  
 No1.00
 Yes0.990.88–1.12
Below poverty, community levele  
 No1.00
 Yes1.000.86–1.16

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

We evaluated the effect of SES and numerous clinical characteristics on racial/ethnic differences in survival following breast carcinoma in this large population-based study. Although crude survival rates were significantly lower for blacks and Hispanics than whites, the survival deficit was reduced, but still present, for blacks and eliminated for Hispanics after adjustment for other risk factors. APIs had a survival experience similar to whites.

Survival in Blacks

The disparity between blacks and whites in survival after breast carcinoma has been controversial. Much of the debate has focused on factors for which race likely serves as a surrogate, such as stage, tumor characteristics (ER status, grade), SES, access to care (delayed diagnosis and treatment, treatment differences), body size, and comorbidity.5, 8, 15 Although the unadjusted risk of death in this study was 1.8 times higher for blacks than for whites, most of the excess risk was explained by disease stage. After controlling for stage, the excess risk was substantially reduced. With further adjustment for other clinical factors and SES, we found a small (22%) albeit significant excess risk of death for blacks, consistent with other studies.6, 23–25 This residual risk may result from uncontrolled confounding, as detailed treatment measures, body size, and comorbidity status were unavailable.

Several survival studies, however, found no differences in outcome between black and white patients after controlling for clinical factors and SES.7–11, 26, 27 This inconsistency may reflect study design issues such as differences in outcome definition (any cause of death vs. death from breast carcinoma), risk factor inclusion, or statistical power. For example, a Michigan study looking at all-cause mortality found no black/white survival difference after adjustment for stage in a managed-care population of breast carcinoma patients, suggesting that health care access eliminated disparity.27 In our data, insurance status was not available. A recent Detroit study evaluated the odds of breast carcinoma death at 2 years postdiagnosis, rather than using a survival analysis. These authors found that enrollment in Medicaid fee-for-service programs, not black race, was associated with death from breast carcinoma.8 Some studies reporting negative results had risk estimates for black race/ethnicity similar to those reported in the current study. However, as our study sample is 10 times larger than most, the difference may be due to statistical power.7, 9, 10, 26, 27

Survival in Hispanics

In this cohort, the prognosis for survival following breast carcinoma diagnosis was similar for Hispanic and white females, after controlling for stage. The literature on breast carcinoma survival among Hispanics is sparse and inconclusive. Similar to our results, no survival differences were reported in two successive cohorts at the M. D. Anderson Cancer Center in Texas or in a recent Florida population.6, 9, 24 However, a large study from the Orange, San Diego, and Imperial County regions of the California Cancer Registry reported that Hispanic patients were significantly more likely to die of breast carcinoma compared with white women.12 Studies comparing Hispanic and white breast carcinoma mortality rates in New Mexico have produced conflicting results. Breast carcinoma mortality rates in the 1958–1987 period were lower for Hispanics in one study,28 whereas a subsequent study reported higher mortality rates for Hispanics during the 1973–1992 period.13 Both New Mexico studies investigated survival trends over time. Whereas whites experienced improvements in stage at diagnosis and survival (most notably in the 1983–1992 period), Hispanic mortality rates did not change.

Survival in APIs

The current study found similar survival rates for white and API females, after controlling for age and stage. A 1982 American College of Surgeons survey reported higher survival rates for Chinese-American and Japanese-American women than for American white women, even after adjustment for stage, age, and histology.29 A SEER-wide study of breast carcinoma diagnosed in females in the 1988–1994 period found higher stage-adjusted survival rates for Japanese females than for Chinese, Filipinas, or whites.18 A previous study of breast carcinoma survival in the San Francisco Bay Area showed that for women diagnosed with local disease, Filipinas had significantly lower relative survival than whites.17 By combining all APIs into one category, we may have obscured the higher survival of the Japanese with the lower survival of the Filipinas. In this cohort, only 16% of the APIs were classified as Japanese, 37% were Chinese, and 33% were Filipinas. Other reports, however, found similar survival for APIs and whites, whether looking at Asians as a combined group or in the various subgroups.12, 19

Survival and SES

In this study, black patients were significantly more likely than white patients to be categorized as having low SES by each of the three measures. Hispanic race/ethnicity was associated with living in a low education neighborhood, but this SES measure was not associated with survival. All three SES measures were strongly tied to disease stage, as women of lower SES presented with more advanced disease. Blue-collar status was the only SES measure in this study linked to survival. It was associated with a 16% excess risk of death, regardless of race/ethnicity. Our SES measures were the same as those used in a Washington study, which reported a significantly increased risk of death following breast carcinoma for women in blue-collar neighborhoods (HR = 1.52; 95% CI 1.28–1.88).10 In that study, however, race was not a significant predictor of survival in the multivariate results. Most previous studies did not measure blue-collar residence but instead evaluated poverty, low education, and insurance type. A survival study of 1089 breast carcinoma patients treated at two Chicago public hospitals found that income, not race, was associated with survival.26 Recently, a Detroit study found that Medicaid status and poverty predicted breast carcinoma death, not race/ethnicity.8 We did not find any association between poverty and breast carcinoma death. The San Francisco Bay Area is known for its high incidence of breast carcinoma, particularly among affluent white women.30 Indeed, in this study, less than 3% of white women resided in below-poverty communities compared with 48% of black women. Despite the inclusion of more than 3000 nonwhite women, only 8% of the total study resided in below-poverty communities, a result driven by the large number of affluent whites.

Although there was no evidence of an interaction between race/ethnicity and SES in these data, SES has been found to operate differently in diverse racial/ethnic groups. Census-based poverty indicators were associated with survival for white, but not Hispanic, women in a SEER study.14 Similarly, a Cleveland study reported that lower SES was associated with decreased survival in white, but not black, breast carcinoma patients.11 Together with these studies, our results suggest that SES is an important factor to consider in cancer control activities.

Measurement Issues

The effects of measurement and misclassification on study findings are key concerns. Race/ethnicity and SES are difficult constructs to define at both the conceptual and practical levels. The two are often entangled, as both encompass treatment choice and quality, healthcare behaviors, comorbid conditions, and overall health status. Race/ethnicity is assumed to measure shared lifestyle, attitudes, experiences, and genetics. In SEER, race/ethnicity is assigned on the basis of medical record review and surname lists. Research has shown that misclassification occurs. For example, non-Hispanic women may be classified as Hispanic based on their husbands' names.31 Furthermore, categorizing subjects into four racial/ethnic groups leads to loss of precision, as aggregation of groups has been shown to obscure relevant differences in cancer incidence and survival.18 Similarly, SES measures such as education, occupation, and income may serve as surrogate markers for environmental exposures, health care choices, and access to medical services.32 To reduce misclassification in the census-based measures we used to describe the neighborhood effects on a patient's outcome, we used the race-specific block group characteristics instead of the larger tract group traits. As we measured the neighborhood or “contextual” SES effect, not the individual SES risk, there may be residual confounding by personal SES. However imprecise, both black race/ethnicity and living in a blue-collar neighborhood did explain some of the mortality risk and should be included in future studies. Further methodologic research should focus on exploring and refining SES measures in multiracial populations.

Population-based studies such as this provide an unbiased sample of patients with heterogeneity in diagnosis and treatment. In addition, this study comprised a large number of black, Hispanic, and API subjects, allowing us the statistical power to simultaneously evaluate many prognostic factors. Generalizing these results to communities outside the San Francisco Bay Area should be done carefully because SES, urbanization, racial/ethnic diversity, and access to care vary throughout the United States. In particular, the null association with poverty may be generalizable only to other affluent urban communities. Our results, however, fit within the published literature. Although this study had a number of demographic and clinical variables, two other factors important to survival (namely, insurance status and comorbidity) were unavailable. Despite these limitations, this study contributes to the understanding of the associations of stage, SES, race/ethnicity, and survival, particularly in Hispanics and APIs, two rapidly growing but understudied populations.

Conclusion and Implications

This large study demonstrated that after adjustment for clinical and SES factors, black breast carcinoma patients experienced a slight but significant decrease in survival, whereas the survival experience of Hispanics and APIs did not differ from whites. Like race/ethnicity, the impact of the census-level SES measures was attenuated when disease stage was included in the model. Considering that the relative contribution of SES is similar to that of race/ethnicity, SEER should consider collecting SES data on a routine basis. Given that stage of disease is so strongly associated with survival, early diagnosis should continue to be an important public health focus. These data highlight the need to focus screening interventions on nonwhite females and females living in lower SES communities.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES