African-American/White differences in breast carcinoma

p53 Alterations and other tumor characteristics


  • See related editorial on pages 1261–3, this issue.

  • Certain data used in the current study were obtained from the Connecticut Tumor Registry located in the Connecticut Department of Public Health. The author(s) assume full responsibility for analyses and interpretation of these data.



Despite mounting evidence that breast tumors in African-American (AA) women are more aggressive compared with breast tumors in white (W) women, little is known regarding racial/ethnic differences in genetic alterations that may be of prognostic importance.


In this population-based cohort of 322 AA women (45%) and W women (55%) who were diagnosed with breast carcinoma between 1987–1989, the authors evaluated available archived tumor tissue (n = 247 samples) for racial differences in selected genetic alterations and other prognostic indicators. Tumor characteristics were assessed by immunohistochemistry and/or expert review.


Alterations in p53 were significantly more common in AA women compared with W women (odds ratio, 4.00; 95% confidence interval, 1.77–9.01) and remained statistically significant in models that were adjusted for disease stage at diagnosis, according to American Joint Committee on Cancer (AJCC) criteria, and for other prognostic indicators. No racial difference with regard to HER-2/neu status was observed, but alterations in c-met were more common in AA women once the model was adjusted for negative confounders (not significant). Among other tumor characteristics, significant findings included later AJCC stage and higher histologic and nuclear grade tumors in AA women. In addition, the burden of aggressive tumor characteristics was greater in AA women because they were more likely to be at high risk on multiple factors (e.g., both high histologic grade and high nuclear grade [P = 0.03] and negative status for both estrogen receptors and progesterone receptors [P = 0.01]).


Data from this population-based cohort confirmed that breast tumors in AA women most likely are more aggressive compared with breast tumors in W women and offer new evidence for possible racial/ethnic differences with regard to p53 alterations. Cancer 2004. © 2004 American Cancer Society.

Several studies now suggest that breast tumors diagnosed in African-American women may be more aggressive compared with the breast tumors that occur in white women.1–5 In addition to the later disease stage at the time of diagnosis2, 3, 5 and the related variables of larger tumor size and lymph node involvement,2–4 compared with tumors in white women, tumors in African-American women reportedly are more likely to have high-grade nuclear atypia,1 high mitotic activity,1 higher S-phase fraction,3 and necrosis1, 4 and reportedly are poorly differentiated (higher grade),1, 6 estrogen receptor (ER) negative,1, 3, 5, 7, 10 and progesterone receptor (PR) negative.3–6, 8, 11 Much less is known about potential racial/ethnic differences in genetic alterations that may be associated with recurrence and overall survival in patients with breast carcinoma.

Although it is known that genetic mutations and other molecular alterations play a role in tumorigenesis, it also is believed that some affect prognosis. Among those that have been studied fairly well, it has been shown that alterations in p53, a tumor suppressor gene,12, 13 and in HER-2/neu (HER-2), an oncogene,14, 15 have a negative effect on prognosis in patients with breast carcinoma. In addition, there is some evidence that alterations in the c-met protooncogene, which encodes the met protein, the receptor for scatter factor/hepatocyte growth factor (SF/HGF), also result in decreased survival of patients with breast carcinoma.16–19

From a population-based cohort of African-American and white patients with breast carcinoma for whom we collected extensive in-person interview and medical record data near the time of diagnosis, we conducted a centralized pathology review of archived tumor tissue to examine racial/ethnic differences in a number of tumor characteristics of putative prognostic significance. Although these data add to the growing literature concerning race-linked tumor aggressiveness, a primary focus of this investigation was to determine whether there are differences in breast tumors among African-American women and white women with respect to molecular alterations in p53, HER-2, and c-met.


This was a follow-up study of 145 African-American women (45%) and 177 white women (55%) who were diagnosed with breast carcinoma between January 1987 and May 1989. Using a rapid case-ascertainment system, as described previously,20, 21 we identified all women with newly diagnosed breast carcinoma from the 22 Connecticut hospitals in which 98% of cases that occurred in African Americans in Connecticut were diagnosed. All eligible African-American women and a random sample of white women, who were frequency matched based on hospital and date of diagnosis (1–3-week period), were enrolled. In-home interviews were conducted for all women (within 3–6 months of diagnosis for 90% of the sample and within 1 year for the remaining 10% of the sample) by trained interviewers using a standardized instrument. This instrument was a modified version of the questionnaire used in the National Cancer Institute Black/White Cancer Survival Study22 and collected extensive information regarding sociodemographic, health history, medical care, and psychosocial factors. Among the eligible participants (423 of 453 identified women), the participation rate was 76% and did not vary significantly by race/ethnicity. A comparison of enrolled patients with patients who were not enrolled but were listed in the Connecticut Tumor Registry, a Surveillance, Epidemiology, and End Results (SEER) Program site, showed no significant differences with regard to SEER stage at diagnosis23 (P = 0.35); further, this association did not vary significantly across racial groups. Approvals from the Human Investigation Committees of all participating institutions20, 21 were obtained for all phases of this study. Accordingly, informed consent was obtained at the time of the in-person interviews but was waived (with approvals from the funding agency, Yale School of Medicine Human Investigation Committee, and all participating hospital Institutional Review Boards) at the time of the follow-up study, during which the archived tumor tissue was accessed.

Archived tumor specimens (n = 247 specimens) were retrieved from the hospital of diagnosis. Because of the time lapse (approximately 10 years), 4 hospitals were unable to provide all or some tissues, resulting in a loss of blocks for 21 patients. Additional losses were attributed to lost blocks (n = 7 specimens) or insufficient tumor tissue (n = 47 specimens). Because smaller amounts of tissue were removed for in situ or localized tumors, missing laboratory results were correlated with disease stage at diagnosis. Notwithstanding the association between race and disease stage observed in these data,20 the distribution of race by availability of blocks was similar to the full sample (47.4% African-American women vs. 52.6% white women). Missing data concerning tumor characteristics did not differ significantly by race, and the association between missing data and disease stage at the time of diagnosis held for both groups. Finally, adjustment for hospital of diagnosis did not significantly alter the results.

The results reported here for histologic grade, nuclear grade, ER status, PR status, and genetic mutations (p53, HER-2, and c-met) were obtained from analysis of archival tumor tissue. Testing was standardized, in that all assays were conducted within a single laboratory and were reviewed and interpreted by one pathologist (M.L.), thus minimizing the variation due to differences in test kits, laboratory methods, and interobserver variability. Working from original pathology reports and medical records, two study physicians (R.D. and M.M.C.) assigned the AJCC stage at diagnosis24 using the TNM classification system (T, tumor size; N, absence or presence and extent of regional lymph node metastasis; and M, absence or presence of distant metastasis). Lymph node involvement was uncertain for 11 women, precluding AJCC stage assignment. Results reported for necrosis, lymphatic invasion, skin involvement, and nipple involvement (from original pathology reports) also were reviewed systematically. Interrater differences were resolved by case conference.

For women with available tumor specimens, hematoxylin and eosin-stained slides were evaluated by the study pathologist, who was blinded to all patient identifiers (including race). In addition, batches of slides/blocks were balanced with respect to race/ethnicity. After determining the availability of tumor tissue and confirming tumor type, histologic grade and nuclear grade were assessed using the criteria of Bloom and Richardson.25 Immunohistochemistry (IHC) was used to assay ER and PR status and the presence of genetic alterations on p53, HER-2, and c-met. Using standard protocols, immunoperoxidase staining was performed on thin sections (4–5 μm) that were cut from formalin fixed, paraffin embedded blocks; this staining was performed within 5 days of sectioning to minimize oxidation or other surface events that may have an impact on the binding of antibodies. Paraffin sections were pretreated if necessary to optimize antigen availability. Positive and negative controls were run with each stain set.

Specific Procedures

ER and PR

Immunostains for ER and PR levels were performed using their respective antibodies and staining kits from Abbott Laboratories (North Chicago, IL). Paraffin sections that were stained for ER were pretreated with pronase.


Staining for p53 was performed using antibody D07 (Dako Corporation, Inc., Carpinteria, CA) Slides were pretreated for 5 minutes at 95 °C with 10 mM sodium citrate buffer, pH 6.0.


The anti-HER-2/neu monoclonal antibody Ab3 (clone 3B5; Oncogene Science, Inc., Manhasset, NY) was used to detect alterations in HER-2. This antibody was raised against a synthetic peptide encompassing the Tyr1248 C-terminal autophosphorylation site of HER-2/neu.


Staining for human met was performed using rabbit polyclonal antibody C-12 (sc-10; Santa Cruz Biotechnology, Santa Cruz, CA). The antibody is specific for c-met, with an epitope at the carboxy terminus of c-met p140 of human origin.

Evaluation of immunostaining for ER, PR, HER-2, c-met, and p53 was based on the intensity of staining and number of infiltrating (invasive) cells affected: 0, scant to few stained cells; 1 +, < 30% of the cells demonstrating weak-to-moderate staining; 2 +, 30–70% of cells stained weakly or moderately; 3 +, > 70% of cells stained with any intensity. Immunoperoxidase staining for ER, PR, and p53 was evaluated in the nucleus (yellow-brown color). Only membranous staining (brown color) was evaluated for HER-2, whereas both cytoplasmic and membranous staining were evaluated for c-met. Scores ≥ 1 + were considered positive for p53; however, for HER-2 and c-met, both of which are expressed normally, scores of 0 and 1 + were considered normal, whereas scores of 2 + or 3 + were considered positive. Because cut-off points are somewhat controversial with different criteria applied across studies and in clinical settings, statistical analyses using more stringent cut-off points (i.e., considering only > 70% staining positive for c-met and HER-2) were carried out but did not substantially alter the reported findings.

Racial differences in tumor characteristics were evaluated with logistic regression using unconditional maximum likelihood analysis.26 Odds ratios (ORs) and 95% confidence intervals (95% CIs) are reported. Some analyses were restricted to the 247 women for whom tumor tissue was available, with dummy variables included in multivariate models for missing data on histologic and nuclear grade. Categorizations (where applicable) of variables that were included in multivariate analyses were determined from preliminary analyses and are detailed in the tables.

Race/ethnicity was based on self-identification at the time of the interview. Other variables that were included in multivariate analyses included age, three measures of socioeconomic status (SES), education, family income, and occupational ranking; selected reproductive factors, including menopausal status, parity, age at first birth, use of hormone replacement therapy and oral contraceptives, history of oophrectomy and benign breast disease (both taken from medical records); and, finally, biomedical and lifestyle factors, including self-reported alcohol consumption, regular smoking, and severe obesity, as defined according to a body mass index (BMI) ≥ 32.3 kg/m2.27, 28 The BMI was calculated from the height and weight recorded in the medical record at the time of diagnosis.


Compared with white women, African-American women were more likely to have been socioeconomically disadvantaged based on education, family income, and occupational rank (Table 1). In contrast, African-American women had superior health insurance coverage, largely due to more extensive coverage and lower deductibles associated with public versus some private insurance. Other significant differences included that African-American women were more likely to have been single, severely obese, and never to have consumed alcohol.

Table 1. Selected Characteristics of Female Patients with Breast Carcinoma by Race/Ethnicity: Connecticut, 1987–1989
Patient characteristicNo. of patients (%)OR95% CI
African Americans (n = 145)aWhites (n = 177)a
  • OR: odds ratio; 95% CI: 95% confidence interval.

  • a

    Numbers for each characteristic may not sum to the total because of missing data.

  • b

    Adjusted for age (continuous variable) and marital status (married or living as married/not married).

  • c

    Dichotomized at the median (for family income; the median was approximated from midpoints of categories).

  • d

    Duncan Socioeconomic Index,46, 47 adapted for spouse pairs (≤ median vs. > median).

  • e

    Adjusted for age (continuous variable).

  • f

    Composite measure of the extent of coverage for hospitalization, outpatient visits, prescription medicines, laboratory tests, check-ups, screening mammograms: any/some and amount of deductible (dichotomized at median).

  • g

    Weight (kg)/height (m2): severe obesity ≥ 32.3 kg/m2.27, 28

 ≥ 50 yrs 78 (53.8)123 (69.5)0.510.32–0.81
 < 50 yrs 67 (46.2) 54 (30.5)1.00
 0–11 yrs 47 (32.6) 31 (17.5)2.93b1.62–5.23
 ≥ 12 yrs 97 (67.4)146 (82.5)1.00
Annual family incomec    
 ≤ $24,999 84 (64.6) 68 (43.9)2.69b1.49–4.87
 ≥ $25,000 46 (35.4) 87 (56.1)1.00
Occupational rankcd    
 Low score101 (72.1) 55 (33.3)6.18b3.49–10.93
 High score 39 (27.9)110 (66.7)1.00
 No 88 (60.7) 68 (38.4)3.79e2.27–6.32
 Yes 57 (39.3)109 (61.6)1.00
Insurance indexf    
 < Median 67 (47.2)109 (61.9)0.58b0.36–0.94
 ≥ Median 75 (52.8) 67 (38.1)1.00
Menopausal status    
 Pre/perimenopausal 52 (36.9) 50 (28.2)1.84e0.86–3.93
 Postmenopausal 89 (63.1)127 (71.8)1.00 
Body mass index (kg/m2)g    
 ≥ 32.3 37 (25.7) 12 (6.8)5.02e2.45–10.27
 < 32.3107 (74.3)164 (93.2)1.00
Alcohol consumption    
 Ever 80 (56.3)146 (82.9)0.21e0.12–0.37
 Never 62 (43.7) 30 (17.1)1.00
Regular smoker    
 Ever 72 (50.7)108 (61.4)0.65e0.41–1.01
 Never 70 (49.3) 68 (38.6)1.00

African-American women were significantly more likely than white women to be diagnosed with later AJCC stage breast carcinoma (age-adjusted OR, 2.01; 95% CI, 1.25–3.25) and had both larger tumors and positive lymph nodes (Table 2), as reported previously.20, 21 Histologic Grade 3 tumors also were more common for African-American women (41.4% vs. 21.3% in white women); and, with each increase in histologic grade, the age-adjusted proportional OR was 2.20 (95% CI, 1.08–4.49). A similar pattern was observed for nuclear grade. These associations remained statistically significant when they were adjusted further for AJCC stage at diagnosis and for three measures of SES. Approximately 54% of African-American women were diagnosed with ER-negative –tumors, compared with 39% of white women (P < 0.05); with multivariate adjustment, the OR was relatively unchanged, although the difference was reduced to borderline significance (OR, 1.81; 95% CI, 0.92–3.57; P = 0.09). Consistent with the trend toward more aggressive tumors, but not reaching the level of statistical significance, PR-negative tumors, lymphatic invasion, and necrosis also were more common in African-American women. Although the data are not shown, tumors that were diagnosed in African-American women, compared with tumors that were diagnosed in white women, were more likely to have both high histologic grade and high nuclear grade (P = 0.03) and ER-negative/PR-negative status (P = 0.01).

Table 2. Tumor Characteristics of Female Patients with Breast Carcinoma by Race/Ethnicity: Connecticut, 1987–1989
Tumor characteristicaNo. of patients (%)OR (95% CI)
African Americans (n = 145)bWhites (n = 177)bAge-adjustedcMultivariate
  • OR: odds ratio; 95% CI: 95% confidence interval; AJCC: American Joint Committee on Cancer.

  • a

    The last ordered category is the referent group.

  • b

    Numbers for each characteristic do not sum to the total because of missing data.

  • c

    Age-adjusted odds ratio from the logistic regression model; numbers for models as shown.

  • d

    American Joint Committee on Cancer TNM classification system24: ≥ Stage IIA, T0N1M0 (no evidence of primary tumor [T0], metastasis to moveable ipsilateral axillary lymph node/s [N1], no distant metastasis [M0]); or T1N1M0 (tumor ≤ 2 cm in greatest dimension [T1], N1 and M0 as above); or T2N0M0 (tumor > 2 cm but ≤ 5 cm in greatest dimension [T2], no regional lymph node metastasis [N0], and M0 as above); Stage I, T1N0M0.

  • e

    The proportional odds ratio was obtained from the cumulative logit model.

  • f

    The multivariate adjusted odds ratio from the logistic regression model was adjusted for age (continuous) and American Joint Committee on Cancer stage (≥ Stage IIA vs. in situ/Stage I) and for 3 measures of socioeconomic status (see Table 1). The numbers of patients included in the logistic regression models were as follows: histologic grade (n = 103 patients), nuclear grade (n = 114 patients), estrogen receptor status (n = 198 patients), progesterone receptor status (n = 196 patients), necrosis and lymphatic invasion (n = 261 patients), and skin and nipple involvement (n = 260 patients).

AJCC staged    
 ≥ IIA100 (69.9)89 (53.0)2.01
 In situ/I43 (30.1)79 (47.0)(1.24–3.24)
AJCC tumor size    
 > 2 cm81 (55.9)73 (41.2)1.85
 In situ/≤ 2 cm64 (44.1)104 (58.8)1.17–2.93
AJCC lymph node status    
 Positive69 (50.0)58 (35.2)1.72
 Negative69 (50.0)107 (64.8)1.07–2.75
Histologic grade    
 322 (41.4)16 (21.3)2.20e2.68ef
 225 (47.2)43 (57.3)1.08–4.49)(1.04–6.90)
 16 (11.3)16 (21.3)  
Nuclear grade    
 326 (41.3)19 (24.4)2.00e2.70ef
 229 (46.0)42 (53.9)(1.04–3.85)(1.16–6.25)
 18 (12.7)17 (21.8)  
Estrogen receptor    
 Negative61 (53.5)50 (38.8)1.661.81f
 Positive53 (46.5)79 (61.2)(1.00–2.80)(0.92–3.57)
Progesterone receptor    
 Negative78 (69.0)80 (62.5)1.361.51f
 Positive35 (31.0)48 (37.5)(0.79–2.35)(0.76–2.98)
 Present44 (30.6)39 (22.2)1.411.23f
 Not present100 (69.4)137 (77.8)(0.84–2.36)(0.64–2.37)
Lymphatic invasion    
 Present31 (21.5)24 (13.7)1.761.39f
 Not present113 (78.5)151 (86.3)(0.97–3.21)(0.64–3.02)
Skin involvement    
 Present9 (6.3)8 (4.6)1.850.91f
 Not present134 (93.7)167 (95.4)(0.67–5.15)(0.21–3.87)
Nipple involvement    
 Present13 (9.1)11 (6.3)1.630.74f
 Not present130 (90.9)164 (93.7)(0.69–3.87)(0.25–2.15)

In Table 3, racial/ethnic differences in genetic alterations are presented, adjusting for age and then for potential confounders, as determined from the literature29–32 and from our own data. Breast tumors in African American women were significantly more likely than tumors in white women to carry alterations in p53 (age-adjusted OR, 4.00; 95% CI, 1.77–9.01). Although it produced a large confidence interval because of the relatively small numbers, a full model adjusted simultaneously for age, SES, tumor characteristics (AJCC stage, histologic grade, ER, PR, HER-2 and c-met), severe obesity, and smoking history (alcohol dropped out) produced results that were consistent with more parsimonious models (OR, 4.29; 95% CI, 1.32–13.94; P = 0.02). The proportions of tumors that were positive for HER-2 overexpression did not differ significantly across racial group, similar to the results for c-met. Adjusting for known correlates for each (benign breast disease was pruned from the HER-2 model), multivariate adjustments that were undertaken to explore possible negative confounding suggested the possibility of an excess of c-met-positive tumors in African-American women (OR, 2.08; 95% CI, 0.75–5.78; P = 0.16). Although statistical power was limited in testing interactions, the racial/ethnic differences observed in these and other tumor characteristics were not more common in younger women, as reported by others.33 The one exception was that the racial difference in p53 overexpression was found to be greater among women age < 50 years, but this interaction term was not statistically significant (P = 0.17).

Table 3. Genetic Alterations in Female Patients with Breast Carcinoma by Race/Ethnicity: Connecticut, 1987–1989
Genetic alterationNo. of patients (%)ORb(95% CI)Multivariate OR (95% CI)
African Americans (n = 145)aWhites (n = 177)a
  • OR: odds ratio; 95% CI: 95% confidence interval.

  • a

    Numbers for each characteristic may not sum to total because of missing data.

  • b

    The odds ratio was adjusted for age (continuous).

  • c

    The multivariate odds ratio (n = 192 patients) was adjusted for age (continuous), American Joint Committee on Cancer stage (≥ Stage IIA vs. in situ/Stage I), histologic grade (3, 2, 1), estrogen receptor and progesterone receptor status (negative vs. positive), HER-2 and c-met status (positive vs. negative), severe obesity (body mass index ≥ 32.3 kg/m2 vs. < 32.3 kg/m2), 3 measures of socioeconomic status, and regular smoker (see Table 1).

  • d

    The multivariate odds ratio (n = 197 patients) was adjusted for age, American Joint Committee on Cancer stage, menopausal status (premenopausal/perimenopausal vs. postmenopausal), parity (0 vs. ≥ 1), age at first birth (≥ 25 yrs vs. < 25 yrs), hormone replacement therapy and oral contraceptives (ever vs. never), history of oophrectomy, and history of benign breast disease from medical record data (yes vs. no).

  • e

    The multivariate-adjusted odds ratio (n = 174 patients) was adjusted for age, American Joint Committee on Cancer stage, histologic grade, estrogen receptor and progesterone receptor status, HER-2 status, p53 status, severe obesity, 3 measures of socioeconomic status, menopausal status, hormone replacement therapy, and history of oophrectomy.

 Positive28 (24.5)9 (7.1)4.004.29c
 Negative85 (75.2)117 (92.9)(1.77–9.01)(1.32–13.94)
 Positive44 (38.9)52 (40.6)0.961.38d
 Negative69 (61.1)76 (59.4)(0.57–1.64)(0.71–2.69)
 Positive27 (23.7)32 (25.2)0.972.08e
 Negative87 (76.3)95 (74.8)(0.53–1.77)(0.75–5.78)


To our knowledge, this is the first population-based investigation to report that there are significant differences between African American and white women with regard to the prevalence of tumors that overexpress p53. We did not observe significant differences in overexpression of HER-2 or c-met across racial/ethnic groups. In addition to examining genetic alterations, we have demonstrated racial differences in a number of individual tumor characteristics that are associated with a poor prognosis34 as well as a greater likelihood that African-American women will have multiple tumor characteristics associated with a poor prognosis. Although some differences were not significant statistically, with further adjustment, many attained statistical significance: African American women were significantly more likely to have later stage tumors, larger tumors, positive lymph nodes, and tumors with higher histologic and nuclear grades. Other factors that were more common in African-American women compared with white women, but that did not reach statistical significance with age adjustment, included negative ER and PR status, necrosis, lymphatic invasion, skin involvement, and nipple involvement. There are inconsistencies in results reported across studies, with few studies to our knowledge including data concerning all the tumor characteristics reported herein; however, the findings of the current study are consistent with the emerging literature showing that African-American women are more likely than white women to be diagnosed with aggressive tumors.33, 35, 36

Among the few studies that have reported on p53 alterations in African Americans,3, 29, 37, 38 only 1 group reported a significant difference in the prevalence of p53 mutations,37 although others reported racial differences in the specific pattern of p53 mutations.29, 38 In a hospital-based study, Caleffi et al. reported a significant racial difference in the proportion of p53 mutations based on DNA sequencing (9 of 22 African-American patients vs. 34 of 170 white patients).37 Our investigation extends this work with larger numbers and by demonstrating a significant racial difference in p53 protein expression, as measured by IHC, that persisted with extensive adjustment for prognostic and other biomedical factors. Although our criteria for positive p53 samples did not exclude explicitly those with < 10% of stained cells, the proportion of tumors in white women that was positive for p53 alterations in this investigation was inconsistent with a high false-positive rate, in that it was lower than generally reported (14–26%).34 This difference likely reflects the population-based nature of these data as much of the literature on the prevalence of p53 mutations in patients with breast carcinoma is drawn from hospital-based/referral center studies of mostly white women. In such studies, selection biases are difficult to gauge but most likely over-represent cases that occur in women with more aggressive tumors.39 Lower than expected p53 overexpression in the total sample also may be attributable in part to antigenic degeneration that occurs over time.40 It is important to note that any misclassification of p53 status should be nondifferential with respect to race, in that all testing in this investigation was conducted with positive and negative controls, without patient identifiers, and with each test batch racially balanced. Thus, irrespective of the overall prevalence of p53 alterations, the finding of a relative difference in the presence of alterations in these genes across racial groups, as reported herein, should be valid, if not attenuated.

Alterations in the p53 tumor suppressor gene, measured as either mutations or p53 overexpression (IHC),39 are associated with a poorer outcome in patients with breast carcinoma,12, 13 with some evidence that the efficacy of chemotherapy is modulated by presence of these mutations.39, 41, 42 If our finding of a greater prevalence of p53 alterations in tumors diagnosed in African-American women is confirmed in other population-based studies, then it may have implications for therapeutic decisions, and it may help to further our understanding of the racial disparity in breast carcinoma survival.

We did not find significant racial/ethnic differences in either of the other genetic alterations assayed. With the exception of one small study,43 the current study results are consistent with other studies that have not observed racial differences in HER-2-positive tumors,3, 44 suggesting that HER-2 is unlikely to play an explanatory role in the observed survival differences. Although overexpression of met (c-met), the receptor for SF/HGF, may be associated with worse outcomes in patients with breast carcinoma,16–19 we know of no other reports on possible racial/ethnic differences. Because of the possibility of negative confounding (i.e., obesity and negative hormone receptor status were less common in women with c-met-positive tumors but were more common in African-American women), larger studies that include data regarding potential confounding factors will be needed to determine whether c-met is more common in African Americans. However, as noted earlier, there is ambiguity concerning the clinical significance of intermediate staining (i.e., 2 +) for both HER-2 and c-met. Although our analyses in which only the highest staining categories (3 +) were designated positive produced a negligible change in the reported association of HER-2 with race/ethnicity, using a higher threshold for positive staining brought the multivariate adjusted OR for c-met closer to 1.00 (still not statistically significant).

Consistent with other reports in the literature, African-American women in this study were more likely than white women to have ER-negative tumors,1, 3, 5, 7–10 although this difference was only marginally significant when it was adjusted for age and other prognostic factors. Unlike other reports,3–6, 8, 11, 29, 38 the racial difference in patients with PR-negative tumors was not significant in our data. However, we did observe that African Americans were significantly more likely than whites to be diagnosed with ER-negative/PR-negative tumors. Although these differences were only marginally significant after adjusting for age and AJCC stage (P = 0.095), these findings are consistent with those reported by Gapstur et al., who found that ER-negative/PR-negative tumors were more common in black patients than in non-Hispanic white patients after adjusting for age, tumor size, and histology,11 and with more recent findings based on analyses of SEER data that were adjusted only for age, SEER site, and year of diagnosis.8 Results of a recent study by Chu et al. using SEER data suggested that joint expression of ER and PR may capture distinct tumor types based on correlations with other tumor characteristics and age at diagnosis. Those authors also demonstrated distinct patterns of associations across racial/ethnic groups.45 Thus, the racial/ethnic differences in joint expression of ER and PR reported by some8, 11 and, to an extent, confirmed here may be more informative in explaining the observed racial differences in survival than either of the hormone receptors alone.

In addition to the well established disadvantage in stage at diagnosis and tumor grade, this investigation shows a clear pattern of higher risk for African-American women with respect to a number of other prognostic indicators, including alterations in p53 and, possibly, in the lesser known c-met. What distinguishes our work from much of the work in this area are the population-based data supplemented with extensive patient information derived from in-person interviews and medical record abstraction, self-identified race/ethnicity, centralized pathology review, and testing of tumor characteristics in a standardized setting. Although our results unlikely are due to selection bias or uncontrolled confounding (e.g., SES), the loss of statistical power associated with multivariate adjustment resulted in some associations that were of borderline significance, whereas we may have missed others. Nevertheless, these findings provide even more evidence that the breast carcinoma tumors diagnosed in African-American women generally are more aggressive compared with tumors that are diagnosed in white women.


The authors thank Rajni Mehta of the Rapid Case Ascertainment Shared Resource, Yale Cancer Center and the Connecticut institutions that participated in this study: Hartford Hospital, Yale-New Haven Hospital, Bridgeport Hospital, Waterbury Hospital, Hospital of St. Raphael, New Britain General Hospital, Norwalk Hospital, St. Vincent's Medical Center, The Stamford Hospital, Middlesex Hospital, Mt. Sinai Hospital, St. Mary's Hospital, Lawrence & Memorial Hospital, Manchester Hospital, Greenwich Hospital, Veterans Memorial Medical Center, Bristol Hospital, St. Francis Hospital and Medical Center, St. Joseph Medical Center, University of Connecticut Health Center/John Dempsey Hospital, Park City Hospital, and William W. Backus Hospital.