Fax: (212) 305-9413
Within-stage racial differences in tumor size and number of positive lymph nodes in women with breast cancer
Version of Record online: 13 AUG 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 6, pages 1201–1208, 15 September 2007
How to Cite
McBride, R., Hershman, D., Tsai, W.-Y., Jacobson, J. S., Grann, V. and Neugut, A. I. (2007), Within-stage racial differences in tumor size and number of positive lymph nodes in women with breast cancer. Cancer, 110: 1201–1208. doi: 10.1002/cncr.22884
- Issue online: 31 AUG 2007
- Version of Record online: 13 AUG 2007
- Manuscript Accepted: 22 MAY 2007
- Manuscript Revised: 18 APR 2007
- Manuscript Received: 5 APR 2007
- R25 Award from the National Cancer Institute (NCI). Grant Number: CA94061
- National Center for Research Resources (NCRR) of the National Institutes of Health. Grant Number: ULI RR024156
- K07 Award from the NCI. Grant Number: CA95597
- K05 Award from the NCI. Grant Number: CA89155
- American Cancer Society. Grant Number: RSGT-01-024-04-CPHPS
- Department of Defense. Grant Number: BC043120
- racial disparities;
- breast cancer;
- Surveillance, Epidemiology, and End Results (SEER);
- lymph nodes;
- tumor size
Black women have higher breast cancer mortality rates, are more likely to be diagnosed at an advanced stage of disease, and have worse stage-for-stage survival than white women. It was hypothesized that differences in the tumor size and number of positive lymph nodes within each disease stage contribute to the survival disparity.
In the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database, black and white women diagnosed with a first primary tumor (TNM stage I-IIIA breast cancer) between 1988 and 2003 were identified. The demographic and clinical characteristics were compared by race. Logistic regression models of the association between race and tumor size and lymph node status were developed. Cox proportional hazards models of the association between mortality and race, tumor size, lymph node status, and other covariates were also examined.
Among 256,174 SEER cases (21,861 black and 234,313 white women), more black than white women with lymph node-negative breast cancer had tumors measuring ≥2.0 cm. Adjusted for tumor size, more black than white women had ≥1 positive lymph nodes (odds ratio [OR], 1.24; 95% confidence interval [95% CI], 1.20–1.28). The age-adjusted and TNM stage-adjusted mortality rate ratio for blacks versus whites was 1.56 (95% CI, 1.51–1.61). Adjustment for within-stage differences in tumor size and lymph node involvement were found to have a negligible effect. With adjustment for additional covariates, the rate ratio was 1.39 (95% CI, 1.35–1.44). In addition, the rate ratio reflecting racial disparity increased as the stage of disease increased.
Adjusting for within-stage differences in tumor size and lymph node status did not appear to reduce the racial disparity. The finding that disparities increased with higher stage of disease suggests that interventions aimed at reducing these differences should target women with more advanced disease. Cancer 2007. © 2007 American Cancer Society.
Major advances in breast cancer screening, diagnosis, and treatment have led to an improvement in survival over the past 20 years. However, despite a lower incidence rate of breast cancer, black women have a notably higher mortality rate from breast cancer than white women. A large portion of this disparity is attributed to more advanced stage at diagnosis among black women. According to the American Cancer Society's annual report of cancer statistics, 63% of white women but only 52% of black women were diagnosed with localized disease, and 34% of white women but 45% of black women were diagnosed with regional or distant disease.1 Some have ascribed the differences in stage at diagnosis to racial differences in the utilization of mammographic screening.2
However, even on a stage-for-stage basis, black women have poorer survival than white women.3, 4 Studies attempting to explain this disparity suggest that differences in biologic characteristics of the tumor and/or factors affecting quality of treatment are responsible.3–13 For example, black women are more likely than white women to be diagnosed with hormone receptor-negative tumors with high nuclear grade.3, 5, 6 Tumors from black women have been shown to have a higher S-phase fraction, are more likely to have cell cycle alterations (p53, p21, p16, and Ki-67), and are more likely to have a basal cell, triple-negative phenotype.3, 7 Studies by our group and others have shown that black women also appear less likely to receive appropriate treatment.8–13 Black women are more likely than white to have delayed initiation of adjuvant chemotherapy or radiation therapy, and more likely to discontinue chemotherapy prematurely, all of which can be associated with worse survival.
TNM staging for breast cancer combines information regarding tumor size, the number of positive axillary lymph nodes, and distant metastasis; each of these factors is reportedly associated with patient prognosis.14 However, each individual stage encompasses a wide range of tumor sizes and lymph node involvement. For example, before the 2003 revision of the American Joint Committee on Cancer (AJCC) staging manual, stages IIA and IIB included patients with any number of positive lymph nodes. The large difference in survival between patients with stages IIA or IIB disease was in fact a primary justifications for the 2003 revision of the staging manual.15 In an observational study at the University of Chicago spanning >50 years, researchers found that women with stage IIA (T1N1) cancer and only 1 positive lymph node had a disease-free survival of 78%, whereas the survival of similarly staged women with 4 to 9 positive lymph nodes dropped to 39%.16 In an earlier Surveillance, Epidemiology, and End Results (SEER)-based study that examined the correlation between tumor size and number of positive lymph nodes in 24,740 women, investigators reported that the 5-year relative survival of women with T1 (<2.0 cm) tumors and 1 to 3 positive lymph nodes was in the range of 87% to 96%, whereas women with similarly sized tumors and ≥4 lymph nodes had a 5-year survival rate of 64%.14
We hypothesized that within-stage differences in tumor size and lymph nodes between blacks and whites may account for some of the stage-for-stage disparities in survival that are observed. We used data from the National Cancer Institute's SEER database to evaluate if the number of positive lymph nodes and tumor size would explain racial disparities in survival.
MATERIALS AND METHODS
We identified black and white women diagnosed with histologically confirmed, invasive, AJCC17 stage I-IIIA breast cancer between January 1, 1988, and December 31, 2003 (n = 290,345) in the SEER program using SEER*Stat software (version 6.2.3) (available at: www.seer.cancer.gov/seerstat). The SEER program has recorded incident cancers since 1973, but has only included AJCC staging information since 1988, as well as cases from 9 registries from Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Puget Sound, and Utah. In 1992 the program expanded to include an additional 4 registry sites: San Jose-Monterey, Los Angeles, Alaska Natives, and Rural Georgia. In 2000, the SEER program was expanded again to include the entire state of California as well as Kentucky, Louisiana, and New Jersey. The SEER Public Use Data include SEER incidence and population data grouped by age, sex, race, year of diagnosis, and geographic area, and provide information regarding stage of disease; tumor grade, size, and histology; lymph node status; and overall survival for each registered patient.
We excluded patients diagnosed with sarcomas or lymphomas of the breast (n = 2). We also sequentially excluded those with no tumor present (n = 190), microscopic foci (n = 5560), no recorded tumor size (n = 1608), Paget disease of the nipple (n = 1), inflammatory breast cancer (n = 227), no reported lymph node dissection (n = 20,947), or a nonspecified number of positive lymph nodes (n = 5258).
Tumor size, number of positive lymph nodes, age, and year of diagnosis were taken directly from the SEER database. The histologic grade of the tumor was categorized as high if it was grade 3 or 4, low if it was 1 or 2, and unknown if it was not recorded. Estrogen receptor (ER) and progesterone receptor (PR) status was considered positive if either ER or PR was recorded as positive, negative if both ER and PR were negative, and unknown if neither measure was recorded. Patients were categorized by marital status at diagnosis as married or unmarried (single, divorced, widowed, or separated). Residence in a major metropolitan area was defined based on the county-level 2003 urban-rural continuum code provided by SEER. Patients were categorized as living in a metropolitan area if their county of residence was located in an urban area with a population of ≥250,000. Patients were categorized by poverty index based on the percentage of families below the poverty level in their county of residence in the year 2000. SEER uses census data as surrogates for individual data concerning residential and socioeconomic factors. The validity of using census-based data for this purpose has been discussed elsewhere.18–21
We used the chi-square statistical test to compare the distributions of demographic (age, marital status, metropolitan residence, socioeconomic status [SES]) and clinical characteristics (stage, tumor size, number of positive lymph nodes, number of lymph nodes assessed, year of diagnosis, tumor grade, and ER/PR status) of the breast cancer patients by race. We conducted univariate analyses of tumor size and number of positive lymph nodes to identify differences in their distributions by race. We used the Kolmogorov- Smirnoff procedure to test for the normality of the distributions of the continuous variables and, for nonnormally distributed data, the nonparametric Wilcoxon test to assess statistically significant differences in the distribution of tumor sizes between races. We categorized tumor size in 0.5-cm increments from 0 to ≥5.0 cm, and also used a chi-square test to evaluate the significance of the difference in distribution of white and black patients in these categories.
Logistic regression models were used to identify predictors of dichotomized tumor size (<2 cm, ≥2 cm) and of any versus no positive lymph nodes within each stage. Cox proportional hazards models were developed to estimate the differences in overall mortality by race. First we adjusted for age at diagnosis. We then sequentially adjusted for 1) stage; 2) tumor size and lymph nodes; 3) stage, size, and lymph nodes; 4) stage plus clinical (year of diagnosis, tumor grade, ER/PR status) and demographic characteristics (marital status, metropolitan residence, zip code level SES); and 5) stage, size, lymph nodes, and clinical and demographic characteristics.
We included interaction terms for AJCC stage and race in the Cox models. We then compared the overall fit (−2 log likelihood) of the model that included variables for AJCC stage with that of the model that also included continuous variables for tumor size and number of positive lymph nodes. We compared the overall fit (-2 Log likelihood ratio test) of the model that included variables for AJCC stage with that of the model that also included continuous variables for tumor size and the number of positive lymph nodes. By comparing the parameter estimates for race between the 2 models, we were able to assess the presence of residual confounding due to racial differences in tumor size and lymph nodes within each AJCC stage. All statistical analyses were performed using SAS software for Windows (version 9.0.3; SAS Institute Inc, Cary, NC).
We identified 256,174 women (21,861 of whom were black [8.5%] and 234,313 of whom were white [91.5%]) and were diagnosed with stage I to stage IIIA breast cancer who met our inclusion criteria. The median follow-up was 45 months (interquartile range, 22–92 months). The clinical and demographic characteristics of the 2 groups are presented in Table 1. The median age at diagnosis was 55 years among black women and 60 years among white women. The mean tumor sizes were 2.5 cm and 1.6 cm in black versus white women, respectively (P < .0001). The mean number of positive lymph nodes among women with lymph node involvement was 4.3 (standard deviation [SD] = 4.9) among black women and 4.0 (SD = 4.9) among white women, respectively (P < .0001). Across the range of tumor stages, black women had fewer stage T1 tumors (52.2% vs 65.9%), and significantly more stage T2 (40.4% vs 30.1%) and T3 (7.4% vs 4.0%) tumors compared with whites (P < .0001). A greater proportion of black women had at least 1 positive lymph node when compared with whites (42.1% vs 32.4%; P < .0001).
|Tumor size, cm|
|Positive lymph nodes|
|Tumor size, cm||2.1/1.7||1.7||2.5/2.0||1.9||2.0/1.6||1.7|
|Lymph nodes examined||12.9/12.0||8.4||13.4/13.0||8.3||12.8/12.0||8.4|
|Positive lymph nodes|
|Lymph node +/−||1.4/0.0||1.8/0.0||1.3/0.0|
|Lymph node + only||4.1/2.0||4.3/2.0||4.0/2.0|
Figure 1 illustrates the size (largest dimension in cm) distribution of tumors by race among women with lymph node-negative T1-T3 breast cancer. Tumors measuring <2.0 cm in greatest dimension accounted for 57% of cancers in white women but only 44% of cancers in black women, whereas there were comparatively fewer white women with tumors measuring ≥2.0 cm in size. Two-tailed P-values for both the Wilcoxon rank sum test and the chi-square test were less than .0001.
In a logistic regression model controlling for tumor size, we found that black women were 24% more likely than white women to have at least 1 positive lymph node (odds ratio [OR], 1.24; 95% confidence interval [95% CI], 1.20–1.28). We tested the possibility that this effect was due to white women being more likely to undergo an axillary lymph node dissection or having a greater number of lymph nodes examined than blacks. The proportions of black and white women meeting our other inclusion criteria who were excluded from our analysis for either having not undergone a lymph node dissection or being listed as having positive lymph nodes of unspecified number did not differ statistically (9.4% vs 9.2%; P = .31). We found that black women have a greater mean number of lymph nodes examined when compared with whites (P < .0001). However, black women were found to have a higher ratio of positive lymph nodes to lymph nodes examined when compared with whites (0.13 vs 0.10; P < .0001). Figure 2 illustrates the association between the mean number of positive lymph nodes and tumor size in black and white women with lymph node-positive disease. It shows that for the most commonly found tumor sizes (1.0–2.0 cm), black women had a greater mean number of positive lymph nodes than whites. Among women diagnosed with tumors measuring between 2.5 and 4.0 cm, black and white women had a similar number of positive lymph nodes; however, as tumor size increased above 4.0 cm, white women were found to have significantly more positive lymph nodes.
Table 2 presents the results of our Cox proportional hazards regression analysis of the association between race and mortality in women with stages I to stage IIIA breast cancer, adjusted for clinical and tumor characteristics. All variables except metropolitan residence were found to be significantly associated with mortality at P < .0001. Only more recent year of diagnosis and being married were associated with improved mortality. The series of models demonstrates that the age-adjusted survival disparity between blacks and whites is reduced from 1.76 (95% CI, 1.71–1.81) to 1.56 (95% CI, 1.51–1.61) after controlling for TNM stage, but it is not appreciably changed after adjustment for residual differences in tumor size or lymph node involvement (1.54; 95% CI, 1.50–1.59). The actual change in the race parameter was e(0.01)(2.2% on the log scale), or 0.015 (<1% difference in the hazard ratio), well below the 10% change in the beta coefficient commonly used to assess the presence of confounding between the nested and full models. Additional adjustment for tumor grade, ER/PR status, year of diagnosis, marital status, metropolitan residence, and poverty index further reduced the black versus white hazards ratio to 1.39 (95% CI, 1.35–1.44). Although not shown in Table 2, we found that the addition of a variable for the ratio of positive lymph nodes to lymph nodes examined had no discernible effect on the survival disparity between blacks and whites in either the full or the nested models.
|Variable||Model 1: Race only||Model 2: Race + stage||Model 3: Race + tumor size + lymph nodes||Model 4: Model 2 + model 3||Model 5: Model 2 + clinical and demographic characteristics||Model 6: Model 3 + clinical and demographic characteristics||Model 7: Model 4 + clinical and demographic characteristics||Model 8: Model 5 + race × stage interaction terms|
|HR||95% CI||HR||95% CI||HR||95% CI||HR||95% CI||HR||95% CI||HR||95% CI||HR||95% CI||HR||95% CI|
|Race (Ref: white)|
|TNM Stage (Ref: stage I)|
|Tumor size, cm||1.28||1.27–1.29||1.21||1.20–1.23||1.24||1.23–1.25||1.18||1.17–1.20|
|No. of positive lymph nodes (Ref: 0)|
|Year of diagnosis||0.98||0.97–0.98||0.98||0.98–0.98||0.98||0.98–0.98||0.98||0.98–0.98|
|Tumor grade (Ref: low)|
|ER/PR status (Ref: positive)|
|Marital status (Ref: unmarried*)||0.82||0.80–0.84||0.82||0.81–0.84||0.82||0.81–0.84||0.82||0.81–0.84|
|Poverty index †||1.01||1.01–1.01||1.01||1.01–1.01||1.01||1.01–1.01||1.01||1.01–1.01|
|Stage IIA × race||1.06||0.98–1.15|
|Stage IIB × race||1.15||1.06–1.25|
|Stage IIA × race||1.30||1.17–1.45|
In addition, we identified a positive statistical interaction between race and stage that demonstrated that, as stage increases, the racial survival disparity increases significantly more than would be expected from multiplying their individual effects (P < .0001), based on a generalized Wald test to simultaneously test all 3 interaction terms. Figure 3 is a graph of the hazards ratios of blacks versus whites from stage I to stage IIB disease after adjustment for other known clinical and demographic confounders. It demonstrates not only that blacks have worse survival than whites across all stages of disease, but also that this disparity significantly increases with each increase in TNM stage.
We found that within each category of TNM stage, blacks had significantly larger tumors than whites. In addition, among all women with lymph node-positive breast cancer, blacks had a significantly higher mean number of positive lymph nodes, but that this effect was only present among tumors measuring <2.0 cm. Despite this, finer adjustment of the survival models to account for these within-stage differences appeared to have no additional effect on the mortality hazards ratio for race. Although statistically significant differences were observed within TNM stage between black and white women with breast cancer, from a clinical perspective these differences were so modest as to have no measurable effect on survival. Finally, we demonstrate that racial disparities in survival between black and white women with breast cancer increase with later stage at diagnosis.
It has been well established that a proportion of the racial disparity noted in outcome is due to differences in tumor characteristics and clinical and demographic factors, which in our group resulted in a drop in the black versus white mortality hazards ratio from 1.76 to 1.39. Even after controlling for known clinical and biologic factors that affect prognosis, black women still had a 39% higher mortality rate. Although black women were significantly more likely than white women to have been diagnosed with lymph node-positive disease, this association was limited to tumors measuring 1.0 to 3.0 cm in largest dimension. Among women with tumors measuring >4.0 cm, white women had significantly more positive lymph nodes than black women. This was an unexpected finding. The finding that black women with lymph node-positive breast cancer have more positive lymph nodes overall compared with white women appears to be driven by more lymph node involvement in the most commonly diagnosed, smaller tumor sizes.
A new interesting finding was the statistically significant interaction between race and stage with respect to mortality (P ≤ .0001). Figure 3 demonstrates that as stage at diagnosis increases, the racial disparity in survival also increases. One plausible explanation for this effect is that when breast cancer is diagnosed at a more advanced stage, curative therapy is more complex, requiring more extensive surgery as well as courses of both radiation and chemotherapy. Studies have suggested that the disparity noted in breast cancer survival is related to differences in the quality of adjuvant treatment, treatment completion, and adherence.11, 12, 22 Bradley et al.23 found that after controlling for other covariates, black women were less likely than whites to undergo surgery, despite being more likely to have worse prognostic features. Our group previously reported that a substantial fraction of women with early-stage breast cancer received <75% of their chemotherapy regimen and that early termination was associated with both black race and poorer overall survival.9 The factors that prevent black women from receiving the same quality of care as white women may be exacerbated by the more complex treatment regimens used for more advanced breast cancer.
The current study had a number of limitations, the majority of which reflected the limited range of data collected by the SEER registry. Specifically, our study could not account for differences in treatment. Without data regarding treatment (surgical, adjuvant, hormonal therapy, and supportive care), we were unable to take the actual treatment received into account. SEER data also do not include comorbid conditions, which may be confounders of the correlation between race and survival. Blacks are more likely to have more comorbid conditions based on the Charlson comorbidity index.9, 24 Obesity is also associated with worse prognosis and is more prevalent in blacks.25–27 Finally, the level of detail concerning the SES of our population limits our ability to understand its role in breast cancer survival. The effect of SES is paradoxical in breast cancer, given the fact that women of lower SES have both lower incidence and lower survival rates.28 SES is of particular concern when considering the frequency with which multiple risk factors, including decreased access to screening and treatment,29, 30 comorbid conditions,31 and a greater likelihood of biologically unfavorable tumor characteristics,32–34 are often clustered together in individuals of lower SES. Given the extent to which these socioeconomic disparities coincide with race in our population, we cannot address the extent to which our estimates of the racial survival disparity are mediated by SES.
Although we were able to confirm that there are statistically significant differences within stage of disease between black and white women with regard to tumor size and lymph node involvement, these differences are not clinically significant with respect to survival over and above the standard AJCC stage categories. Our finding that the racial survival disparity increased with greater stage of disease was unexpected. If this latter finding is confirmed, efforts aimed at reducing racial disparities should focus on eliminating barriers to quality care in black women with higher-stage disease.
Supported by an R25 Award from the National Cancer Institute (NCI) (CA94061 to R.B.M.), and a T32 Award (ULI RR024156 to R.B.M.) from a grant from the National Center for Research Resources (NCRR) of the National Institutes of Health; a K07 Award from the NCI (CA95597 to D.H.); a K05 Award from the NCI (CA89155 to A.I.N.); a grant from the American Cancer Society (RSGT-01-024-04-CPHPS to A.I.N.); and a grant from the Department of Defense (BC043120 to A.I.N.).
- 1RiesLAG, MelbertD, KrapchoM, et al. (eds). SEER Cancer Statistics Review, 1975–2004. Based on November 2006 SEER data submission, posted to the SEER web site, 2007. Bethesda, MD: National Cancer Institute; 2007. Available at: http://seer.cancer.gov/csr/1975_2004/.
- 17BeahrsOH, HensonDE, HutterRVP, KennedyBJ (eds). Manual for Staging of Cancer. 4th ed. Philadelphia: Lippincott; 1992.