Existing data on breast cancer in African-American women

What we know and what we need to know

Authors


  • This article is a US Government work and, as such, is in the public domain in the United States of America.

Abstract

BACKGROUND

Much of what is known about breast cancer in African-American (AA) women is based on existing cancer surveillance data. Thus, it is important to consider the accuracy of these resources in describing the impact of breast cancer in AA populations.

METHODS

National cancer surveillance data bases are described, their most recent findings are presented, their limitations are outlined, and recommendations are made for improving their utility.

RESULTS

Breast cancer characteristics have been studied well in urban (but not in rural) and Southern AA populations. The recent Surveillance, Epidemiology, and End Results (SEER) Program expansion and the continued improvement of state cancer registry operations will provide opportunities to study larger and more diverse AA subpopulations. Recommendations for improving the utility of surveillance data bases include adding new items to better describe correlates of advanced stage at diagnosis and reduced survival of AA women with breast cancer by linking surveillance data bases with other large data bases to provide area-level socioeconomic status, health insurance status, and retrieving new information about patient comorbidities and biomarkers from medical records; improving the completeness and accuracy of treatment and survival information already collected for all patients; working to improve the dissemination of appropriate cancer data to nonresearch consumer communities, including clinicians, patients, advocates, politicians, and health officials; and the development of new training programs for cancer registrars and researchers.

CONCLUSIONS

The continued improvement of cancer surveillance systems should be considered important activities in this research agenda, because these data will play a far-reaching role in the prevention and control of breast cancer in AA women. Cancer 2003;97(1 Suppl):211–21. Published 2003 by the American Cancer Society.

DOI 10.1002/cncr.11026

Much of what we know about the burden of breast cancer in African-American (AA) women is based on existing cancer surveillance data. From these routinely collected data, we have learned that AA women have lower overall incidence rates of breast cancer compared with white women but are diagnosed with later stage disease, have shorter survival, and have the highest rate of breast cancer mortality of all racial-ethnic groups in the United States.1–3 Surveillance data also have been useful in more detailed studies of breast cancer incidence patterns and trends, risk factors, screening, treatment, mortality, and survival among AA women.4–28

Although cancer surveillance is a cornerstone of our efforts to prevent and control cancer in all populations, no existing program was designed specifically to monitor breast cancer in the AA population. Thus, it is important to assess critically the utility of existing surveillance data in the study of breast cancer in AA women, a charge assigned to the authors as part of the Summit Meeting on Breast Cancer Among African-American Women in September, 2000. This summary article describes cancer surveillance programs that collect information about breast cancer in AA women in the United States, provides recent scientific observations from them, discusses their limitations, and offers recommendations for how they may be improved and used more effectively in the future.

Cancer Surveillance Data Sources

There is no single national cancer surveillance system in the United States;29 rather, multiple programs routinely collect and report data regarding the incidence, survival, and mortality patterns of breast cancer in AA women. These include the population-based Surveillance, Epidemiology, and End Results (SEER) Program, state cancer registries funded by the National Program of Cancer Registries, and the National Center for Health Statistics (NCHS) mortality data collection efforts, as well as the hospital-based registries of the National Cancer Data Base (NCDB). Each of these programs collects data for different purposes and, thus, varies with respect to the populations covered, the scope of information collected, the methods by which data are collected, and the timeliness with which data are available for broad distribution.

The SEER program

The SEER Program is a National Cancer Institute-funded initiative to collect high-quality cancer incidence and survival data for a sample of the United States population. The program has collected information about all newly diagnosed cancers occurring in nine geographic regions, including the metropolitan areas of San Francisco/Oakland, Detroit, Atlanta, and Seattle/Puget Sound; the states of Connecticut, Hawaii, Iowa, New Mexico, and Utah since the early 1970s; and the county of Los Angeles and the San Jose/Monterey metropolitan area since 1992. The expansion of the SEER Program in 2001 to include Kentucky, New Jersey, Louisiana, and the rest of California will increase coverage to over 65 million persons, about 26% of the United States population.30 Data items collected by SEER include demographic characteristics of the patient, anatomic and histologic characteristics of the cancer, stage at diagnosis, diagnostic techniques used, the kinds of treatment received within 4 months of diagnosis, and patient outcomes, including survival and subsequent cancers. The program is renowned for its stringent standards for data completeness and quality, with an estimated case ascertainment of 98%.

The long tenure and high quality of the SEER data- base make it a valuable source of information about breast cancer incidence trends in AA populations. In addition, it is the only source of population-based data regarding breast cancer stage at diagnosis, survival, and stage specific survival. SEER data for a given year are available publicly about 28 months after the end of that year. Breast cancer information specific to AA women are produced yearly and are summarized in the SEER cancer statistics review.1 Public-use data sets that are useful for more detailed analyses also are produced and released annually.

Other population-based cancer registries

Many state cancer registries that are not participants in the SEER Program also collect and distribute information regarding breast cancer in AA women. Data collected by these registries generally are more limited and do not include detailed clinical information or patient survival, but they do represent an important resource for breast cancer incidence rates and trends in a more diverse sampling of regional AA populations. State cancer registries are administered jointly by state health departments and by the Centers for Disease Control and Prevention's National Program of Cancer Registries (NPCR). Altogether, state registries funded by NPCR currently collect complete cancer incidence data for about 50% of the total United States population and eventually will cover > 90%. Data from state cancer registries that cover at least 90% of their state populations have been aggregated and reported by the North American Association of Central Cancer Registries (NAACCR) as part of the Cancer in North America publications.31

NCHS

The NCHS compiles and distributes breast cancer mortality data by race-ethnicity for the entire United States population. Whereas cancer registries generally collect information from medical records, breast cancer deaths are determined from death certificates. Death certificates appear to be an acceptably sensitive means of identifying deaths due to breast cancer, because studies comparing death certificate information with medical and autopsy records suggest that they undercount breast cancer deaths by < 4%.32, 33 Mortality data for a given year are available publicly about 2 years after the end of that year. United States breast cancer mortality rates and trends for blacks and whites are summarized and published by the SEER Program to facilitate comparison with incidence data.

NCDB

The NCDB is supported by the American College of Surgeons and the American Cancer Society for the specific purpose of monitoring and improving the quality of cancer care.34 The NCDB is comprised of cancer data from over 1800 hospital-based cancer registries, covering over half of the United States cancer patient population. Participating hospital-based cancer registries represent a broad spectrum of patient care facilities throughout the United States, ranging from large academic medical centers to rural community hospitals.35 Although AA women comprised 8.2% of all breast cancer patients and 14.3% of all patients with Stage IV disease reported to the NCDB in 1995,36 specific analyses of AA breast cancer patients have not been published, perhaps because of the unknown representativeness of AA women with breast cancer reported to the NCDB. However, because NCDB data provide a unique opportunity for the comparison of patient outcomes across hospitals, hospital-level analyses of outcomes among AA women with breast cancer would be informative. NCDB data are available several years after cancer diagnosis.37

Recent Observations from Cancer Surveillance Databases

Disease characteristics

Simple tabulations describe the characteristics of a disease in a population without consideration of the size and other features of the population itself. Table 1 shows the age distributions of AA, Asian/Pacific Islander, and white (non-Hispanic) women diagnosed with invasive breast cancer between 1992 and 1997 and reported to 1 of the 11 SEER Program registries.38 These data suggest that AA women and Asian/Pacific Islander women with breast cancer have similar age distributions. Among AA women with breast cancer, 33.0% were diagnosed at age < 50 years, whereas 67.0% were diagnosed at age ≥ 50 years. Among Asian/Pacific Islander women with breast cancer, 33.0% and 67.1% were diagnosed at age < 50 years and at age ≥ 50 years, respectively. In contrast, 21.9% and 78.1% of white women were diagnosed in the same respective age groupings. Thus, higher proportions of AA women and Asian/Pacific Islander women with breast cancer were age < 50 years compared with white patients.

Table 1. Distributions of Patients with Invasive Breast Cancer by Race/Ethnicity and Age Group: SEER Program, 1992–1997
Age at diagnosis (yrs)African-AmericanWhiteAsian/Pacific Islander
No.%No.%No.%
  1. SEER: Surveillance, Epidemiology, and End Results.

> 301481.35520.5660.7
30–49365131.724,74321.4288532.9
50–69484642.148,13241.7398845.5
> 70287424.942,06236.4182020.8

Table 2 presents the distribution according to American Joint Committee on Cancer (AJCC) stage at the time of diagnosis with breast cancer among AA women, Asian/Pacific Islander women, and white women who were diagnosed between 1992 and 1997.38 These data confirm that, in the 11 SEER regions, a lower proportion of AA women, compared with Asian/Pacific Islander women or white women, were diagnosed with Stage I disease (29% vs. 43% and 43%, respectively), and a higher proportion were diagnosed with later Stage III and IV disease (17% vs. 10% and 10%, respectively). Stage distributions by racial-ethnic group have not been published by the NCDB.

Table 2. Distribution and 5-Year Relative Survival of Patients with Invasive Breast Cancer by Race/Ethnicity and Stage at Diagnosis: SEER Program, 1992–1997
Stage at diagnosisaAfrican AmericanWhiteAsian/ Pacific Islander
%Survival%Survival%Survival
  • SEER: Surveillance, Epidemiology, and End Results.

  • a

    Staging was performed according to the criteria of American Joint Committee on Cancer, 3rd edition.

Stage I2996431004399
Stage II367632853684
Stage III1037658653
Stage IV713417418

Incidence rates and trends

Age-adjusted incidence rates account for the size and the age structure of the population at risk and allow comparison of incidence between AA women and women in other racial-ethnic groups. Because age-adjusted rates are constructed statistically to allow these comparisons, they are not an accurate reflection of the risk carried by an individual woman. Rather, incidence rate ratios estimate relative risks between groups. Figure 1 shows incidence rates of invasive breast cancer specific to 5-year age groups for AA women and white women who were diagnosed between 1992 and 1997 and were reported to the SEER Program.38 The incidence rates for AA women are slightly higher compared with the rates for white women age < 50 years; after that age, the rates for white women climb substantially higher compared with the rates for AA women. The black-white crossover has provoked considerable speculation regarding possible etiologic explanations.6, 39 These recent data suggest an AA-to-white ratio of 1.05 for breast cancer diagnosed in women age < 50 years and a 0.85 ration for breast cancer diagnosed in women age > 50 years.

Figure 1.

Age specific incidence rates of invasive breast cancer in African-American women (AA) and white women. [Surveillance, Epidemiology, and End Results (SEER) Program, 1992–1997].

Figure 2 shows trends in invasive breast cancer incidence overall and for women ages 0–49 years and age > 50 years in the SEER Program between 1973 and 1997.38 Over this 25-year period, incidence rates increased 29% in white women, 39% in AA women, and 55% among women in other racial-ethnic groups. Rates of increase appeared to be more rapid in women age ≥ 50 years: The rates increased 38% in white women, 49% in AA women, and 79% among women in other groups over this period. Trends in women age < 50 years were less dramatic, with small increases in all racial-ethnic groups over the 25 years, including white women (2%), AA women (13%), and women in other groups (11%).

Figure 2.

Trends in the incidence of invasive breast cancer in African-American women (AA) and white women, overall and by age. [Surveillance, Epidemiology, and End Results (SEER) Program, 1973–1997].

There is substantial geographic variation in breast cancer incidence rates among AA women. Table 3 presents 1995–1997 age-adjusted incidence rates by region from SEER and NAACCR. Under the assumption that all participating registries are of comparable completeness, these data suggest that the incidence of invasive breast cancer in AA women can vary by as much as 117% by SEER region or state. Among women age < 50 years, age-adjusted incidence ranged from 18.1 women per 100,000 (Rhode Island) to 39.0 women per 100,000 (Iowa). Among women age ≥ 50 years, similar rates ranged from 235.4 women per 100,000 (West Virginia) to 382.7 women per 100,000 (Atlanta). Among women age ≥ 50 years, rates were generally higher in white women compared with AA women, with the exceptions of women living in Atlanta, Iowa, and Kentucky, for whom rates were higher in AA women compared with white women. Whereas the small size of the AA population in Iowa may have produced an unstable rate, breast cancer incidence rates among AA women in Atlanta and Kentucky were curiously high and deserve further scrutiny.

Table 3. Age-Adjusted Rates of Invasive Breast Cancer in African-American Women and White Women by Age Group (Birth 0–49 years or ≥ 50 years) from the SEER Program and Selected State Cancer Registries, 1995–1997a
SourceAge < 50 yrsAge ≥ 50 yrs
AAWhiteAAWhite
  • SEER: Surveillance, Epidemiology, and End Results; AA: African American; SF: San Francisco.

  • a

    Age-adjusted to the 1970 U.S. standard; rates are shown only when they were based on counts exceeding 10 patients.

  • b

    From the National Cancer Institute.38

  • c

    From the North American Association of Cancer Registries.31

SEER Programb31.531.4317.8374.6
 Atlanta (metropolitan)38.434.1382.7377.3
 Connecticut29.036.3281.9386.3
 Detroit (metropolitan)31.830.8315.3360.6
 Greater SF Bay Area28.333.8298.7430.7
 Hawaii30.3396.8
 Iowa39.028.7374.8342.4
 Los Angeles County29.529.7309.9354.3
 New Mexico30.4256.7328.4
 Seattle (metropolitan)27.132.0326.3413.8
 Utah24.2316.9
State cancer registriesc    
 Arizona19.726.4242.8344.9
 California26.027.2303.9378.9
 Colorado23.426.4281.1356.7
 Delaware32.127.9330.6399.7
 Florida27.127.8253.4345.5
 Idaho23.2349.4
 Illinois29.027.0307.6350.2
 Kentucky29.125.3322.4318.9
 Louisiana27.325.1260.3325.8
 Minnesota30.926.5265.2365.7
 Montana25.9322.3
 Nebraska23.823.6334.8352.4
 New Hampshire27.5374.5
 New Jersey26.630.5291.2375.9
 North Carolina28.627.0268.4319.7
 Rhode Island18.129.4308.4369.5
 West Virginia24.023.6235.4301.5
 Wisconsin26.425.8281.1354.4

Mortality

Across all ages, breast cancer mortality rates in AA women continued to be higher compared with the mortality rates for white women or among women in other racial-ethnic groups. Overall age-adjusted breast cancer death rates for AA women were 1.3 times the rates for white women and 2.7 times the rates among women in other racial-ethnic groups according to 1993–1997 NCHS data for the entire United States. Figure 3 demonstrates trends in United States breast cancer mortality overall and for women age < 50 years and age ≥ 50 years between 1969 and 1997. Over this 28-year period, overall age-adjusted breast cancer death rates decreased 15% in white women but increased 22% in AA women and 14% among women in other racial-ethnic groups. Mortality trends over this period were disparate by age group. In women age ≥ 50 years, the death rate decreased 10% in white women but increased 35% in AA women and 22% among women in other racial-ethnic groups. The breast cancer mortality rate in women age < 50 years decreased in all racial-ethnic groups, but the total percent decrease was greater among white women (33%) compared with AA women (11%) and women in other groups (6%). These data update and confirm the widening AA/white disparity in breast cancer mortality that has been reported previously by other authors16, 17, 21, 23 and underscore the differences in trend by age group.

Figure 3.

Trends in mortality due to breast cancer in African-American women (AA) and white women in the United States, overall and by age, 1969–1997.

Survival patterns

Table 2 presents 5-year relative survival rates for patients with invasive breast cancer by race-ethnicity and stage at diagnosis. Relative survival represents the percentage of women who survive breast cancer for 5 years after diagnosis adjusted for baseline mortality from other causes. These data from women who were diagnosed in the 11 SEER regions between 1992 and 199738 confirm the poorer survival of AA women with breast cancer, even within groups controlled for stage. Survival rates were most disparate for women diagnosed with AJCC Stage III disease: The 5-year relative survival rate for AA women was 37%, substantially lower compared with the rate for whites (58%) and Asians/Pacific Islanders (53%).

The NCDB provides some stage specific and age specific, relative survival information specific to AA women who were diagnosed between 1985 and 1988.12 For women in the NCDB with AJCC Stage I disease, the 5-year relative survival rates were 90%, 86%, and 78%, respectively, for patients age < 35 years, ages 35–49 years, and age ≥ 50 years. For women in the NCDB with Stage II disease, similar age specific, the 5-year relative survival rates were 68%, 70%, and 66%, respectively. Comparable rates were not available from the SEER Program, because AJCC stage was not collected for patients who were diagnosed before 1988.

Limitations of Cancer Surveillance Data

Representativeness of AA populations covered by surveillance programs

Population-based cancer surveillance programs administered by SEER and the NPCR were designed to monitor cancer in geographically defined populations, not racially or ethnically defined populations. Thus, the AA populations covered by these programs may not be entirely representative of the AA population at large. Existing SEER Program regions were selected both for their ability to operate a population-based registry and for the ethnic composition of their populations; however, together, they do not include all subpopulations of AA women in the United States. Although census data indicate that 13% of the 1990 United States female population was identified as AA, only 8% of the female population covered by the 11 SEER regions were identified as AA. Most of the AA women currently covered by SEER reside in five largely urban regions, including Detroit (30%), Los Angeles (27%), Atlanta (19%), San Francisco/Oakland (13%), and Connecticut (7%), with small proportions resident in Seattle (3%) and the San Jose/Monterey, Iowa, New Mexico, Hawaii, and Utah regions (1% each). No SEER county with a substantial AA population (at least 5%) was classified as rural by the 1990 Census.40 Thus, SEER breast cancer data for AA women reflect the incidence and survival patterns for a highly urban AA female population. In addition, this population likely has higher socioeconomic status compared with the AA female population at large, because 20 SEER counties with substantial numbers of AA residents had significantly higher levels of education and income compared with non-SEER counties that had similar numbers of AA residents.40 Because higher regional socioeconomic status has been associated previously with elevated breast cancer rates in AA populations,5, 19 breast cancer patterns gleaned from SEER data may overestimate incidence for the AA female population at large. This possibility has been raised previously with the observation that breast cancer rates in predominantly black, rural counties in North Carolina and Georgia are substantially lower compared with the overall SEER rate.41 Geographic representation of the AA population also is skewed in the SEER Program, because only one SEER region (metropolitan Atlanta) includes AA residents in the southern United States, the region with the highest density of AA population. Thus, the SEER database probably has provided an accurate reflection of the impact of breast cancer in urban, Western, and Midwestern communities over the last 3 decades, but it has not provided information about breast cancer in rural and southern AA communities. The inclusion of rural and southern AA populations was an important objective of the program's expansion, which increased SEER coverage to 24% of the total AA population in the United States.

The continued aggregation of state cancer registry data by the NPCR and the NAACCR will be particularly helpful in the study of breast cancer incidence among AA women, because it will facilitate comparisons of breast cancer characteristics in well-studied urban populations with those in previously understudied rural and Southern United States populations. In addition, the boosted numbers will allow for more accurate and stable estimates of breast cancer incidence among AA women across the United States as well as providing more statistical power for subgroup analyses by age.

Racial-ethnic classifications

Accurate and reliable cancer statistics for the AA population require accurate and reliable classification of each patients' race. Cancer registrars abstract patient race-ethnicity from hospital medical records. Although information in medical records may derive from the patient's own self-report, it may be based on subjective assessment by medical personnel; thus, it has been reported that medical record classifications vary in reliability.42 In addition, for patients without information in the medical record, cancer registries may assign race-ethnicity from a variety of available information, including United States Census surname lists, interview data, or death certificates.

If self-reported race-ethnicity can be considered a gold standard, then the accuracy of cancer registry racial-ethnic classifications probably differs by group of interest, with higher rates of misclassification for Hispanics and Asian subgroups.43, 44 AA women appear to be classified accurately, at least in the Greater (San Francisco) Bay Area Cancer Registry, where preliminary analyses indicate that registry classification of AA race corresponds well with self-report. Of 452 women with breast cancer who were identified by the registry as AA, 96% also were self-identified as AA when they were interviewed for a multiethnic case–control study (unpublished data). Of 500 women in the same study who identified themselves as AA, 87% were identified as AA by the registry. The remaining 13% of women had been misclassified as white or Hispanic. At least in the Greater San Francisco Bay Area, the sensitivity and predictive value positive of cancer registry-assigned AA race are very high, suggesting that cancer registries can accurately capture the numbers of breast cancer patients who identify with the AA community. However, some comparison groups, including Hispanics and specific Asian subgroups, remain more difficult to capture accurately, which may influence comparative analyses.

AA identification appears to be captured accurately by mortality data, for which racial-ethnic classification is derived from the death certificate. For over 3000 persons who participated in a national interview study and reported their race-ethnicity as black, 98.2% also were classified as black on the death certificate.45 Again, agreement between self-report and the death certificate was lower for Hispanics and Asians/Pacific Islanders,45 possibly biasing comparative mortality analyses.

Accurate denominators for cancer rates

Equally important to the accuracy of cancer rates is the accurate enumeration of the population at risk. Population counts, which are used as denominators in the calculation of cancer rates, are obtained from the United States Census. Undercounting of populations, which has been suspected for some urban AA subgroups, would result in artificially high cancer rates. Flawed population projections with regard to race or age also would bias cancer rates. A more subtle form of methodologic bias relates to the different means of classification used by cancer rate components. Although census racial-ethnic classifications are based on self-report, cancer case counts, which form the numerator of cancer rates, are not necessarily based on self-report, as discussed above. Thus, cancer rates may be biased when case counts do not match denominators, and the magnitude and direction of the bias would be unknown.

The 1990 United States Census defined race and ethnicity separately, such that a person of any race (white, black, American Indian/Alaskan Native, and Asian or Pacific Islander) also may be of Hispanic ethnicity. Some among the AA population, particularly those of Dominican, Puerto Rican, or other Caribbean ancestry, consider themselves both black and Hispanic. Attention must be paid to the consistent classification of these individuals for the purposes of cancer surveillance, because cancer rates also may be biased when either the numerator or denominator of AA breast cancer rates differentially include or exclude Hispanic blacks. The issue of classifying multiracial individuals has been magnified lately, because the 2000 United States Census allowed individuals to report more than one racial-ethnic background, representing a radical shift in the way that governmental agencies have considered race and/or ethnicity. Although estimates suggest that few individuals actually report a multiracial background, it is uncertain how this means of collecting data will influence both the size and the composition of AA population estimates.46 Although factors that influence the reporting of race have been discussed by other authors,47 it seems certain that race specific cancer incidence trend analyses will be hampered substantially by uncertainties regarding the 2000 Census population estimates.

Yet another procedural change that will influence the interpretation of cancer rates and trends is the requirement for all federal health data to be age adjusted with the 2000 standard population. The switch to the 2000 standard population will make cancer incidence and NCHS mortality data comparable for the first time, because these data previously were age adjusted with different standard populations—incidence data with the 1970 age standard and mortality data with the 1940 age standard. However, rates adjusted to the 2000 age standard, which reflects an older population compared with the 1970 standard, will not be comparable to and will appear substantially higher than previously published rates. To compound the problem, the amount of increase will differ by race because of differences in age specific incidence. Comparing overall cancer mortality rates for 1992–1996 adjusted to the three different age standards (1940, 1970, and 2000), rates for the AA population were 36%, 33%, and 31% higher compared with the white population.

Completeness and quality of data currently collected by cancer registries

All cancer registry data is subject to routine edits and quality-control procedures, but its quality can be influenced by the availability of treatment information in medical records, particularly when treatments have been given by multiple providers. The completeness and accuracy of treatment data is now being scrutinized for all cancer registry patients. In a recent study of older women with breast cancer that utilized linked SEER and Medicare claims data, over 18% of women who were identified by Medicare claims as recipients of radiation therapy were not so identified by SEER, whereas 7% of those identified as recipients of radiation therapy by SEER were not identified by Medicare, with disagreement particularly high for women who underwent breast-conserving surgeries.48 Thus, the increasing use of outpatient clinics or providers other than the diagnosing facility for chemotherapy and radiation therapy may have an impact on the completeness and accuracy of treatment information for patients with breast cancer. In contrast, SEER data regarding breast cancer surgery seems to be very complete and of excellent quality. In a recent study of older patients with breast cancer that utilized linked SEER and Medicare claims data, 94% of patients who underwent surgery for breast cancer according to SEER were identified by Medicare.49

Survival information is another data item for which there have been concerns about completeness and accuracy, although only a small proportion of cancer registries follow patients for vital status. In the SEER Program, patient identifiers (e.g., name, birth date, and social security number) are linked with an array of vital statistics and administrative data bases to confirm vital status. Quality standards dictate that, at any given time, 95% of all patients have at least one report of vital status in the last 18 months. However, the completeness and accuracy of updated vital status and cause of death information can be influenced negatively by several factors, including change of patient name and relocation of patients to other states. The completeness and accuracy of both vital status and cause of death information can be improved by routinely linking cancer registry data bases with the National Death Index, which detects deaths of patients with cancer who move to another state.

The collection of more data items would help cancer registry data analyses to describe better the burden, incidence, and outcomes of AA women with breast cancer, as suggested previously. To facilitate routine and ongoing collection, proposed new data items need to be available readily and recorded routinely in the medical record, because most information in cancer registries is abstracted from medical records. Although SEER registries already collect estrogen and progesterone receptor status, it would be opportune to collect additional tumor information from these records, including biomarkers (e.g., Her-2/neu status). Other potentially useful information routinely recorded in cancer patients' medical records includes the presence of comorbidity at the time of diagnosis, that is, other health conditions that may influence treatment decisions or outcome. Adjustment of cancer survival analyses for comorbid conditions may explain some of the poorer outcomes seen among older AA women. Technical assistance and additional financial resources will be needed for registrars and central registries that currently are without the necessary infrastructure to collect these items.

Additional means of augmenting and enhancing cancer registry data have been identified previously and, in some instances, implemented. Linking cancer registry data bases with other data bases is a cost-effective way of adding new data items. When patient address at diagnosis is geocoded to the census tract and block group level, area-level socioeconomic status can be added to each patient record.50 However, examination of breast cancer trends by socioeconomic status requires yearly population estimates by age, gender, and subcounty geographic area (e.g., census tract, block group). Currently, the United States Census provides age specific, gender specific, and race specific population data for subcounty geographic areas only once every 10 years. Other database linkages that may yield particularly salient data items for the study of breast cancer in AA populations includes health insurance status at the time of diagnosis (linkage with administrative data bases) and treatment beyond 4 months after diagnosis (linkage with Medicare, other insurance claims data bases). The entire SEER data base has been linked successfully with the Health Care Financing Administration/Medicare data base.51

Diffusion and dissemination of data

In addition to the SEER Program, the NAACCR, the NPCR, and the NCDB, the American Cancer Society and the National Cancer Registrars Association distribute breast cancer statistics and information to the public and to professional communities in print and on the Internet. Despite the ongoing efforts of these organizations to make cancer statistics available to nonresearch communities, including clinicians, patients, advocates, politicians, and health officials, there remains concern that cancer surveillance information is difficult to access and to understand. Work is needed to identify the data needs of each of these consumer communities and to develop more accessible modes as well as more easily understood means of disseminating syntheses of key cancer surveillance information to particular audiences with particular cancer information needs. In this way, the wealth of cancer data may best be used to motivate action and investment in programs that reduce breast cancer mortality in the AA population and in other under-served populations.

Risk Communication

Elevated risks of breast cancer in young AA women have been reported widely,6, 39 although the most recent SEER data show that age-adjusted rates of invasive disease in AA women age < 50 years were 1.05 times those in white women of similar age (Fig. 1). Cancer surveillance researchers may not be communicating adequately the magnitude of this elevation in relative risk and the difference between relative and absolute risk. In addition, it remains unclear the extent to which age-adjustment procedures explain differences in breast cancer rates in young AA and white women. Crude cancer rates are not used for routine rate comparisons, because they do not adjust for the age distribution of each underlying population, but they do estimate the actual numbers of women affected in each population per year. Crude rates of invasive breast cancer in women age < 50 years during 1992–1997 actually were lower for AA women compared with white women (38.3 women per 100,000 compared with 40.8 women per 100,000).38 Thus, for women in this age group, the slight increase in the age-adjusted rate of breast cancer in AA women may relate to a younger age distribution. In addition, regardless of race or age adjustment, rates of breast cancer for women < 50 years are one-eighth to one-tenth the rates for women age > 50 years, underscoring the low absolute risk of breast cancer in women age < 50 years compared with older women. The possibility that young AA women are being frightened unnecessarily about their breast cancer risk is of concern.

The Logistics of Cancer Surveillance Efforts

In the face of an increasing number of new diagnoses to register, cancer registries nationwide are having difficulty maintaining a skilled workforce. Many hospital registrars receive only on-the-job training, and the demand for certified tumor registrars, for whom certification requires rigorous training, far outweighs the supply. Training programs for hospital cancer registrars, particularly in rural and inner-city hospitals, need to be organized and funded. In the research setting, there are a limited number of scientists with expertise in cancer surveillance research, which requires knowledge of cancer biology, epidemiologic methods, biostatistics, disease registration, and increasingly sophisticated informatics. More researchers are needed to undertake the many descriptive epidemiologic analyses that could further point to new avenues for breast cancer research in the AA community, and stronger emphasis should be placed on recruiting those with sensitivity to cultural issues in that community. Finally, concerns about the confidentiality of medical record information may threaten efficient and high-quality cancer surveillance. Despite the fact that cancer is a reportable condition, and its registration is required by law in nearly all states, timely cancer surveillance efforts may be hampered by poorly designed restrictions or medical privacy legislation.

Final Recommendations

The continued evaluation and improvement of cancer surveillance systems should be considered important activities in the research agenda for breast cancer in AA women, because surveillance data will play a far-reaching role in the prevention and control of this disease. Careful monitoring of trends in breast cancer incidence, stage, mortality, and survival will allow us to quantify the impact of changes in exposure to risk factors, improvements in access to screening programs, and advances in treatment. The ongoing monitoring of treatments and their outcomes will allow for hospitals and other health care providers continually to evaluate and ameliorate the quality of care. The timely dissemination of breast cancer surveillance data can help researchers, clinicians, community groups, and others to institute new screening programs, initiate trials of novel treatment protocols, implement community-based interventions, and otherwise improve prevention and control efforts. With these efforts in mind, we present the following recommendations for improvement:

Improve data quality

Specific data items.

Preliminary data indicate that cancer registry classification of AA race from the medical records corresponds well with self-report. However, efforts are needed to obtain accurate racial-ethnic classifications for some comparison groups and to improve the completeness and accuracy of data regarding treatment and survival for all cancer registry patients.

Cancer rates.

Equally important to the accuracy of cancer rates is the accurate enumeration of the population at risk. In light of the multiple racial categorizations of the 2000 Census, work is needed to confirm that the census data, used as cancer rate denominators, correspond with numerators. To facilitate ongoing breast cancer surveillance in AA women, accurate population estimates are needed by age, gender, and subcounty geographic areas for all years, including intercensal years.

Add information to current cancer registry data

Adding new items to cancer registries would help us to describe better the correlates of poorer stage at the time of diagnosis and survival among AA women with breast cancer.

Medical record information.

Because most information in cancer registries is abstracted from medical records, it would be opportune to collect additional, routinely recorded information from these records, such as comorbidities and biomarkers. However, technical assistance and additional resources will be needed for registrars and central registries who do not have the necessary infrastructure to collect these items.

Record linkages.

Because new computer technologies facilitate record linkage, several new data items, including area-level socioeconomic status (linkage with census data), health insurance status at the time of diagnosis (linkage with administrative data bases), treatment beyond 4 months of diagnosis (linkage with Medicare and other insurance claims data bases), and comorbidities (linkage with hospital discharge data bases) easily may be added to surveillance.

Make existing data more useful to consumers

Data availability.

Work is needed to identify the data needs of nonresearch consumer communities, including clinicians, patients, advocates, politicians, and health officials, and to develop specific tools for disseminating cancer targeted to the specific information needs of different nonresearch audiences.

Effective communication.

Work also is needed to develop understandable and culturally sensitive means of communicating cancer surveillance information to these consumer communities.

Improve training opportunities

Registries.

Cancer registries currently face a dearth of skilled employees in the face of an increasing number of new cases in an aging population and a growing number of state cancer registries. Training programs for hospital cancer registrars, particularly in rural and inner-city hospitals, need to be organized and funded.

Research.

More training opportunities in cancer surveillance research are needed to increase the numbers of researchers with expertise in this area. Stronger emphasis should be placed on recruiting those with sensitivity to cultural issues in the AA community.

Ancillary