Outcomes in Black Patients With Early Breast Cancer Treated With Breast Conservation Therapy

Int. J. Radiation Oncology Biol. Phys., Vol. 79, No. 2, pp. 392–399, 2011 Copyright Ó 2011 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/$–see front matter

doi:10.1016/j.ijrobp.2009.11.017

CLINICAL INVESTIGATION

Breast

OUTCOMES IN BLACK PATIENTS WITH EARLY BREAST CANCER TREATED WITH BREAST CONSERVATION THERAPY MICHAEL A. NICHOLS, M.D, PH.D.,* LOREN K. MELL, M.D.,* MICHAEL D. HASSELLE, M.D.,* THEODORE G. KARRISON, PH.D.,y DHARA MACDERMED, M.D.,* AMBER MERIWETHER, B.A.,* MARY ELLYN WITT, R.N.,* RALPH R. WEICHSELBAUM, M.D.,* AND STEVEN J. CHMURA, M.D., PH.D.* Departments of *Radiation and Cellular Oncology and yHealth Studies, University of Chicago, Chicago, IL Background: The race-specific impact of prognostic variables for early breast cancer is unknown for black patients undergoing breast conservation. Methods and Materials: This was a retrospective study of 1,231 consecutive patients $40 years of age with Stage I–II invasive breast cancer treated with lumpectomy and radiation therapy at the University of Chicago Hospitals and affiliates between 1986 and 2004. Patients were classified as either black or nonblack. Cox proportional hazards regression was used to model the effects of known prognostic factors and interactions with race. Results: Median follow-up for surviving patients was 82 months. Thirty-four percent of patients were black, and 66% were nonblack (Caucasian, Hispanic, and Asian). Black patients had a poorer 10-year overall survival (64.6% vs. 80.8%; adjusted hazard ratio [HR], 1.59; 95% confidence interval [CI], 1.23–2.06) and 10-year disease-free survival (58.1% vs. 75.4%; HR 1.49; 95% CI, 1.18–1.89) compared with nonblack patients. Tumor sizes were similar between nonblack and black patients with mammographically detected tumors (1.29 cm vs. 1.20 cm, p = 0.20, respectively). Tumor size was significantly associated with overall survival (HR 1.48; 95% CI, 1.12–1.96) in black patients with mammographically detected tumors but not in nonblack patients (HR 1.09; 95% CI, 0.78–1.53), suggesting that survival in black patients depends more strongly on tumor size in this subgroup. Tests for race-size method of detection interactions were statistically significant for overall survival (p = 0.049), locoregional control (p = 0.036), and distant control (p = 0.032) and borderline significant for disease-free survival (p = 0.067). Conclusion: Despite detection at comparable sizes, the prognostic effect of tumor size in patients with mammographically detected tumors is greater for black than in nonblack patients. Ó 2011 Elsevier Inc. Breast neoplasm, Continental population groups, Mammography, Outcome assessment, Health care.

INTRODUCTION

improvements in adjuvant therapy (15). However, the magnitude of the decrease in breast cancer mortality differs by race, with Caucasian women experiencing a 2.4% yearly decline since 1990, compared with only 1.1% in African-American women since 1991 (1). The reason for this discrepancy is unclear. Although historically, black women have been less likely than Caucasian women to use screening mammograms (15), in recent years such use by black women has reached the same level as that by Caucasian women (1), indicating that mammography underuse is unlikely to explain racial disparities in current trends. Black patients with breast cancer are more likely to be classified as having lower SES, but defining SES is notoriously difficult, and no consensus on the most suitable definition has been reached (5–7). Low SES has been associated with lower use of screening mammography (17, 10) and poor outcomes (28, 29). However, a recent meta-analysis examining

Breast cancer mortality has been declining for all women since 1990, but the magnitude of this decrease has been greater for white than for black patients (1). Racial disparity in breast cancer outcomes is a well-recognized problem (2–5). A variety of factors, including poorer socioeconomic status (SES) (5–7), higher incidence of competing comorbid diseases (8, 9), biologic differences in tumor characteristics (10–14), lower use of screening mammograms (15–18), and inequality of treatment (3, 19–21), have been proffered to explain the poorer outcomes observed in black patients. However, even when controlling for these and other known prognostic factors, black patients continue to have worse breast cancer–specific mortality (3, 7, 8). The decline in breast cancer mortality has been attributed in part to both screening mammography (15, 22–27) and

Conflict of interest: none. Acknowledgment—The authors thank Samuel Hellman, M.D., for suggestions and review of the manuscript before publication. Received Sept 10, 2009. Accepted for publication Nov 2, 2009.

Reprint requests to: Steven J. Chmura, M.D., Ph.D., University of Chicago Hospitals, Department of Radiation & Cellular Oncology, 5758 S. Maryland Avenue / MC 9006, Chicago, IL 60637; E-mail: [email protected] Supported by funding from the American Cancer Society. 392

Race and tumor size in breast cancer d M. A. NICHOLS et al.

race and SES concluded that, even when controlling for age, stage, and SES, race is an independent indicator of breast cancer mortality (7). A higher prevalence of comorbid disease in black patients has also been posited to explain disparities in outcomes. A recent study of patients with Stage I–IV breast cancer reported that black patients were more likely than white patients to have adverse comorbid diseases, which explained differences in all-cause and competing-cause mortality (8). However, race was still independently associated with both recurrence and breast cancer mortality when controlling for the effects of comorbid disease. An equally complicated question has been the role of biology with regard to racial differences in the natural history of breast cancer (30–32). Tumors in African-American women are more likely to be estrogen receptor (ER)/progesterone receptor (PR) negative and to have high grade, a high mitotic index, alterations of p53, and the basal-like phenotype (5, 10–13, 32–35). Whether these differences are due to an inherently more aggressive tumor phenotype or simply to later presentation of disease, or both, is unclear. Recently, an analysis of three of the largest randomized controlled trials for screening mammography in North America revealed that interval cancers are associated with poor outcomes (36). Whether interval cancers occur more commonly in black patients is not known, but adverse pathologic features, such as a higher growth fraction, alteration of p53, and lower apoptotic fraction, are correlated with later presentation of disease (37) and are more common in interval clinically detected cancers appearing in mammographically screened populations (33, 38, 39). Thus, investigating potential differences in biologic tumor characteristics by race and method of detection is an increasingly important area of research. Comprehensive analyses of factors contributing to racial disparities in outcome in early breast cancer are limited. We sought to examine a racially diverse cohort of patients undergoing breast conservation with lumpectomy and radiotherapy, with or without adjuvant systemic therapy, whereby differences due to advanced presentation of disease, access to treatment, and variability of therapy were minimized. Particular attention was given to interactions between race and known prognostic factors. In so doing, preliminary analyses indicated a differential benefit from screening mammography among women who had already been screened, and further analysis focused on the nature of this discrepancy. METHODS AND MATERIALS Patient eligibility The study sample consisted of 1,231 consecutive patients treated at the University of Chicago and affiliated hospitals between January 1986 and May 2004. All patients had American Joint Committee on Cancer clinical or pathologic Stage I–II breast cancer and were treated with breast conservation surgery followed by whole breast irradiation, with or without systemic therapy. This database concerned recurrence and cosmesis in women undergoing breast conservation; data for women who underwent mastectomy were

393

not available. We excluded women with ductal carcinoma in situ without evidence of microinvasion, noncarcinoma histology, or bilateral breast cancer and patients with a history of other cancer within the previous 10 years, other than nonmelanomatous skin cancer.

Surgical therapy Surgery consisted of partial mastectomy. Axillary staging was either an axillary lymph node (LN) dissection or a sentinel LN procedure followed by axillary dissection as appropriate. In most cases when margins were found to be positive, the patient underwent reexcision. Margins were defined as positive if the tumor extended to the inked margin.

Radiation therapy Patients were treated with whole breast radiation therapy (RT) with a boost to the tumor bed. All fields were treated daily, 5 days a week. Patients were treated in the supine position with tangential fields with additional supraclavicular field at the discretion of the treating physician.

Systemic therapy Systemic chemotherapy consisted of various regimens including cyclophosphamide, methotrexate, fluorouracil, anthracyclines, and taxanes, at the treating physician’s discretion. Typically, patients received adriamycin and cyclophosphamide, with or without a taxane (AC  T), CAF (cyclophosphamide, adriamycin, and fluorouracil), or CMF (cyclophosphamide, methotrexate, and fluorouracil). Patients with either estrogen receptor or progesterone receptor positive by immunohistochemistry were scored as hormone receptor positive. Adjuvant therapy with tamoxifen was recorded, but the use of aromatase inhibitors was not standard for most of the study period, and data on their use were not available.

Follow-up evaluation After the completion of therapy, patients were seen in clinic at 3- to 6-month intervals for the first 2 years and at 6-month intervals for the first 5 years. Thereafter, patients were seen yearly. Unilateral mammograms were obtained every 6 months on the affected breast, and bilateral mammograms were obtained yearly. Chest radiographs and liver function tests were obtained yearly in the absence of symptoms or findings worrisome for metastatic spread, and more extensive testing was undertaken as appropriate to the clinical situation.

Definitions Race was self-described by the patient at the time of initial consultation. Tumors were recorded as ‘‘mammographically detected’’ if they were detected by screening mammogram and were not associated with clinical signs. Patients were stratified into ‘‘low’’ or ‘‘high’’ SES by estimating the median household income for each patient, using census tract data. Patients below the median household income for the group were recorded as being of ‘‘low’’ SES. Comorbid disease was self-reported by the patient and was scored in a binary fashion based on whether the patient had hypertension, diabetes mellitus, coronary artery disease, or chronic obstructive pulmonary disease.

Statistical analysis and definition of endpoints Comparisons of demographic and tumor characteristics between groups were performed using the chi-square test or Fisher’s exact test (if the number of patients in a given cell was <5) for proportions, and the unpaired t-test or analysis of variance (for >2 groups) for

394

I. J. Radiation Oncology d Biology d Physics

Volume 79, Number 2, 2011

Table 1. Patient and tumor characteristics Characteristic

All

Nonblack

Black

p

n Age, y Weight, kg Low SES Comorbid disease Pathologic tumor size, cm Mammographically detected Stage I II T stage T1 T2 T3 Lymph node positive Hormone receptor status Negative Positive Unknown Grade 1 2 3 Unknown Chemotherapy Final margin negative

1,231 60.1 (11.3) 75.1 (17.5) 49.6% 37.0% 1.55 (0.91) 49.3%

816 59.7 (11.4) 72.0 (15.3) 30.5% 28.1% 1.49 (0.85) 52.0%

415 61.1 (11.3) 81.0 (19.9) 87.0% 54.7% 1.67 (1.01) 44.1%

0.038 <0.001 <0.001 <0.001 0.002 0.009

68.6% 31.4%

71.9% 28.1%

61.9% 38.1%

<0.001

79.4% 20.1% 0.6% 18.0%

81.4% 18.1% 0.5% 15.6%

75.4% 23.9% 0.7% 22.7%

0.041

17.6% 58.8% 23.6%

16.2% 67.2% 16.7%

20.2% 42.4% 37.4%

<0.001

11.9% 37.6% 28.1% 22.4% 32.9% 94.5%

12.6% 39.3% 27.0% 21.1% 32.1% 94.2%

10.4% 34.2% 30.4% 25.1% 34.5% 94.9%

0.10

0.002

0.41 0.61

Abbreviation: SES = socioeconomic status. For continuous variables, values given are the mean, with standard deviation in parentheses.

continuous variables. The Mann-Whitney test was used to compare follow-up times among survivors. Differences in survival curves by group were compared with the log-rank test or stratified log-rank test. Cox proportional hazard regression with backward elimination (threshold: p < 0.10) and robust variance estimation was used to model factors prognostic for overall survival (OS), disease-free survival (DFS), locoregional control (LRC), and distant control (DC). Race was retained in all models as a covariate regardless of statistical significance. The proportional hazards assumption was tested by analyzing scaled Schoenfeld residuals (40). Overall survival, DFS, and LRC were estimated using the Kaplan-Meier method. Disease-free survival time was defined as time from first pathologic confirmation of the diagnosis to first locoregional failure (LRF), distant failure (DF), or death resulting from any cause. Locoregional control time was defined as time to first LRF, censoring deaths without LRF; DC was defined analogously. Patients who were lost to follow-up were censored at the time of the last known medical encounter. Gray’s test (41) was used to compare differences in the cumulative incidence of LRF and DF as first events. Covariates in the regression models included age, weight, and pathologic tumor size as continuous variables and LN-positive status, hormone receptor–positive status, tumor grade, whether the patient received chemotherapy, margin status, black race, presence of comorbid disease, and method of detection (clinical vs. mammographic) as dichotomous variables. Her2 data were not available. Tests for interaction were conducted by including pairwise interaction terms (coded as the product of the two variables) between race and each covariate that was significantly associated with the outcome (p < 0.05) on univariate analysis. Missing values were imputed for both receptor positive status (24.2% of the sample) and

tumor grade (6.1%), using the Markov chain Monte Carlo method of multiple imputation (42). A sensitivity analysis to test for confounding of coefficient estimates by hormone therapy use was conducted on a subsample of patients whose tumors were diagnosed between 1986 and 1999. In this analysis, tamoxifen use was included as a dichotomous covariate in the regression analysis.

RESULTS Demographic characteristics The mean age of patients was 60.1 years. Thirty-four percent were black, and 66% were nonblack (60.5% Caucasian, 2.0% Asian, and 3.4% Hispanic). Median follow-up for surviving patients was 82 months (black, 80; nonblack, 84; p = 0.19). Black patients were significantly older (61.1 vs. 59.7 years, p = 0.04) and had significantly higher weight, lower SES, and higher likelihood of having comorbid disease than did nonblack patients (Table 1). Surgery, radiation, and chemotherapy All patients underwent lumpectomy followed by RT. Final margins were negative in 94.5% of patients (Table 1). For patients who underwent axillary dissection, the median number of LNs removed was 13 (black, 15; nonblack, 12; p < 0.001). The median whole breast irradiation dose was 4,600 cGy (range, 4,000–5,040). An electron boost to the tumor bed was added to a median dose of 1,400 cGy (range, 0 cGy–2,000 cGy). The median total dose to the tumor bed was 6,000 cGy (range, 4,320 cGy–6,680 cGy). There were

Race and tumor size in breast cancer d M. A. NICHOLS et al.

395

Table 2. Univariate analysis of outcomes by race 10-year overall survival

10-year disease-free survival

Characteristic

All

Nonblack

Black

p*

All

Nonblack

Black

p*

All Age, y <60 $60 SES Low High Comorbid disease No Yes Mammographically detected No Yes Stage I II T stage T1 T2–3 N stage N0 N1 Hormone receptor Negative Positive Unknown Grade 1 2 3 Unknown Chemotherapy No Yes Final margin Negative Positive

75.3%

80.8%

64.6%

<0.001

69.7%

75.4%

58.1%

<0.001

84.0% 67.2%

86.0% 75.7%

79.6% 52.3%

<0.001

63.4% 76.5%

72.0% 78.9%

47.9% 70.8%

<0.001

71.4% 79.1%

82.2% 80.3%

64.5% 64.2%

<0.001

65.1% 74.0%

75.5% 75.4%

58.3% 54.6%

<0.001

83.7% 64.3%

85.9% 71.4%

76.3% 56.7%

<0.001

78.2% 58.5%

80.7% 66.1%

69.6% 50.2%

<0.001

70.4% 80.8%

76.0% 85.6%

61.3% 69.0%

<0.001

63.5% 76.5%

70.0% 80.7%

52.6% 66.1%

<0.001

78.5% 68.4%

83.2% 74.5%

67.5% 59.7%

<0.001

74.4% 59.4%

79.0% 66.2%

63.5% 49.7%

<0.001

77.9% 65.3%

82.8% 72.2%

67.5% 54.8%

<0.001

73.4% 55.3%

78.3% 62.9%

62.7% 43.3%

<0.001

77.5% 65.2%

82.3% 72.1%

67.0% 56.4%

<0.001

72.6% 56.3%

77.7% 62.9%

60.9% 43.3%

<0.001

74.0% 77.9% 69.6%

83.9% 80.3% 79.1%

58.6% 69.9% 61.3%

<0.001

67.7% 72.4% 64.3%

76.1% 75.5% 73.9%

54.0% 61.7% 55.8%

<0.001

74.9% 77.4% 75.8% 69.9%

80.4% 82.2% 80.4% 77.4%

63.1% 66.4% 68.2% 57.6%

<0.001

69.5% 70.5% 70.5% 66.3%

75.4% 75.9% 75.8% 73.1%

56.1% 57.8% 61.4% 54.9%

<0.001

74.4% 77.5%

80.0% 82.9%

62.9% 67.7%

<0.001

70.2% 68.7%

75.6% 75.1%

59.0% 56.4%

<0.001

75.7% 70.8%

81.6% 71.3%

64.4% 69.5%

<0.001

70.4% 59.4%

76.2% 64.2%

58.7% 49.0%

<0.001

Abbreviation: SES = socioeconomic status.

no significant differences between black and nonblack patients with regard to mean dose (6,048 vs. 6,032 cGy, p = 0.23, respectively) or the use of a boost (99.1% vs. 99.4%, p = 0.20). Adjuvant chemotherapy was given to 33.2% of patients: 22.8% of LN-negative patients and 80.4% of LN-positive patients. Among LN-positive patients, 80.5% of black and 80.4% of nonblack patients received chemotherapy (p = 0.99). Among LN-negative patients, 21.1% of black patients and 23.6% of nonblack patients received chemotherapy (p = 0.37). Of the patients who received chemotherapy, 73.6% received anthracycline-based chemotherapy (black patients, 76.5%; nonblack patients, 72.2%; p = 0.21). For systemic hormonal therapy, 52.3% of women received tamoxifen. For hormone receptor–positive women, 63.2% received tamoxifen, including 67.4% of nonblack and 53.3% of black patients (p < 0.0001). Of the receptor-negative patients, 16.6% received tamoxifen, including 13.9% of nonblack and 19.4% of black patients (p = 0.21).

Tumor characteristics There were significant differences between black and nonblack patients with regard to most tumor characteristics (Table 1). Black patients were more likely to present with LN-positive disease (22.7% vs. 15.6%, p = 0.002), receptornegative disease (20.2% vs. 16.2%, p < 0.001), and clinically detected tumors (52.0% vs. 44.1%, p = 0.009). Mean tumor size was larger for black patients (1.67 cm vs. 1.49 cm, p = 0.002). However, when stratified by method of detection, black and nonblack patients with mammographically detected tumors had a similar mean tumor size (1.29 cm [black] vs. 1.20 cm [nonblack], p = 0.20), history of prior mammography (85.2% [black] vs. 86.8% [non-black], p = 0.59), and interval since the previous screening mammogram (1.26 years [black] vs. 1.56 years [nonblack[, p = 0.07). In patients with mammographically detected tumors, the likelihood of being LN negative was similar for black and nonblack patients (86.9% vs. 89.1%, p = 0.28, respectively),

I. J. Radiation Oncology d Biology d Physics

396

Volume 79, Number 2, 2011

Table 3. Multivariate analysis of outcomes by race Nonblack Overall survival

Disease-free survival

Locoregional control

Distant control

Covariate

HR (95% CI)

HR (95% CI)

HR (95% CI)

HR (95% CI)

Mammographic detection Tumor size Lymph node positive Positive margin Age Comorbid disease

0.61 (0.43–0.85) 1.20 (1.02–1.42) 1.65 (1.09–2.49) NS 1.05 (1.03–1.07) 1.59 (1.12–2.26)

0.70 (0.51–0.95) 1.23 (1.05–1.43) 1.52 (1.05–2.22) NS 1.03 (1.01–1.05) 1.46 (1.06–2.01)

1.01 (0.51–2.02) 1.49 (1.08–2.06) NS NS 0.95 (0.91–0.99) NS

0.55 (0.29–1.03) 1.63 (1.29–2.07) NS 2.08 (0.94-4.59) 0.96 (0.93–0.99) NS

Black Covariate

Overall survival

Disease-free survival

Locoregional control

Distant control

Mammographic detection Tumor size Lymph node positive Positive margin Age Comorbid disease

0.85 (0.57–1.27) 1.18 (1.03–1.36) 1.59 (1.04–2.45) NS 1.04 (1.03–1.06) 1.64 (1.08–2.48)

0.80 (0.55–1.18) 1.20 (1.04–1.40) 1.62 (1.09–2.41) NS 1.04 (1.02–1.05) 1.71 (1.16–2.51)

0.75 (0.31–1.81) NS NS 6.03 (2.09–17.4) NS NS

0.65 (0.26–1.60) 1.33 (0.98–1.79) 1.97 (0.93–4.19) NS NS NS

Abbreviations: HR = hazard ratio; CI = confidence interval; NS = not significant (p > 0.10). Mammographic detection included in all models regardless of statistical significance.

but black patients were more likely to be ER negative / PRnegative (23.5% vs. 15.1%, p = 0.01). Outcomes Black patients had poorer 10-year OS (64.6% vs. 80.8%, p < 0.001) and DFS (58.1% vs. 75.4%, p < 0.001) (Table 2). Differences between black and nonblack patients remained statistically significant after stratification by age, SES, presence of comorbid disease, method of detection, and various tumor characteristics (Table 2). The 10-year cumulative incidence of LRF as a first event was 6.0% (95% CI, 3.2–8.8%) for black and 3.7% (95% CI, 2.1–5.3%) for non-black patients (p = .13). The 10-year cumulative incidence of DF as a first event was 7.9% (95% CI, 4.8–11.0%) for black and 6.7% (95% CI, 4.6–8.8%) for nonblack patients (p = 0.44). Regression analysis On multivariate Cox regression analysis, black race was associated with significantly poorer OS (adjusted hazard ratio [HR] 1.59; 95% CI, 1.23–2.06, p < 0.001) and DFS (HR

1.49; 95% CI, 1.18–1.89, p = 0.001). Variables included in these models included age, weight, race, SES, method of tumor detection, comorbid disease, pathologic tumor size, ER/ PR status, and lymph node status. Increasing tumor size was associated with poorer OS (HR 1.18; 95% CI, 1.06–1.32, p = 0.002), DFS (HR 1.21; 95% CI, 1.09–1.35, p = 0.001), and DC (HR 1.45; 95% CI, 1.19–1.76, p < 0.001). Mammographic detection was associated with significantly improved OS (HR 0.70; 95% CI, 0.54–0.91, p = 0.009), DFS (HR 0.74; 95% CI, 0.58–0.94, p = 0.013), and DC (HR 0.46; 95% CI, 0.26–0.81, p = 0.008). No significant pairwise interactions between race and other covariates were identified (data not shown). Sensitivity analysis revealed no changes in the HR estimates with or without exclusion of tamoxifen use as a covariate on multivariate analysis in relation to tumor size, race, and method of detection. To better analyze racial differences in factors correlated with outcomes, we stratified the analysis by race (Table 3). Hazard ratios for mammographic detection were significant for OS and DFS in nonblack patients but were not significant for any endpoint in black patients. Tests for interactions

Table 4. Effect of tumor size (adjusted hazard ratio per 1 cm increase), by race and detection method Nonblack

Black

Mammographic

Clinical

Mammographic

Clinical

Endpoint

HR (95% CI)

HR (95% CI)

HR (95% CI)

HR (95% CI)

Overall survival Disease-free survival Locoregional control Distant control

1.09 (0.78–1.53) 1.06 (0.80–1.41) 1.21 (0.61–2.40) 1.03 (0.49–2.17)

1.29 (1.05–1.58) 1.33 (1.09–1.62) 1.86 (1.18–2.93) 1.90 (1.44–2.51)

1.48 (1.12–1.96) 1.47 (1.07–2.03) 1.98 (1.25–3.14) 1.83 (1.22–2.74)

1.14 (0.99–1.32) 1.17 (1.00–1.36) 0.74 (0.33–1.65) 1.25 (0.97–1.60)

Abbreviations: HR = hazard ratio; CI = confidence interval.

Race and tumor size in breast cancer d M. A. NICHOLS et al.

397

Fig. Estimated hazard ratio for distant failure with respect to tumor size in patients with mammographically (A) and clinically (B) detected tumors. Hazard ratios are plotted relative to the baseline hazard for each race. Dotted lines represent 95% confidence intervals. (C) Composite graph of estimated hazard ratios for distant failure with respect to tumor size in patients with mammographically detected (MD) and clinically detected (CD) tumors. Hazard ratios are plotted relative to the baseline hazard (clinically detected tumor, nonblack race, size = 0 cm).

between race and method of detection were not statistically significant (p > 0.10). We further fit regression models stratified by race and method of detection. The estimated HRs for the remaining covariates generally agreed across the four strata. The HR for tumor size was lower for each endpoint in nonblack patients with mammographically detected tumors (Table 4), suggesting that the effect of tumor size could differ based on race and method of detection. In women with clinically detected tumors, the HR for tumor size was higher for nonblack women. Analysis of models including all covariates in the multivariate analyses, with full sets of interaction terms for race, size, and detection method, revealed that the race-size-detection interaction was statistically significant for the endpoints OS (p = 0.049), LRC (p = 0.036), and DC (p = 0.032) and was borderline significant for DFS (p = 0.067), supporting the conclusion that the effect of tumor size differs by race and method of detection. Tests for racesize interactions in clinically detected tumors were significant only for DC (p = 0.045), whereas tests for race-size interac-

tions in patients with mammographically detected tumors were borderline significant for the endpoints OS (p = 0.069), DFS (p = 0.074), and LRC (p = 0.074) and nonsignificant for DC (p = 0.14). To illustrate the variation in the effect of tumor size by race and method of detection, we plotted HRs for DF as a function of size in each of the four strata. In patients with mammographically detected tumors (Figure, A), the effect of tumor size on the HR for DF changed minimally in nonblack patients compared with black patients. For black patients, the HR increased approximately 1.8-fold for every 1-centimeter increase in tumor size. In patients with clinically detected tumors (Figure, B), the HR for DF increased in both groups with increasing tumor size. Part C of the figure overlays the four HR curves, with each HR plotted relative to the hazard for a nonblack patient with a clinically detected tumor of hypothetical size 0 cm. Size had an important effect in all groups except in nonblack patients with mammographically detected tumors, suggesting a more indolent clinical behavior of breast cancer in this subgroup.

398

I. J. Radiation Oncology d Biology d Physics

Volume 79, Number 2, 2011

DISCUSSION

starting size would be predicted to undergo transformation to a distant metastatic phenotype at different rates as tumor size increases. The strengths of this study include the fact that this was a large, racially heterogeneous cohort of patients receiving relatively homogenous treatment. We performed a comprehensive analysis including comorbid disease, SES, method of detection, and other tumor characteristics among a large cohort of narrowly defined patients to define prognostic factors. Furthermore, analyses in large populations, such as from Surveillance Epidemiology and End Results data, typically exclude key prognostic factors, such as method of detection, and lack analyses of disease recurrence endpoints. With the exception of adjuvant tamoxifen use, we did not observe any variation in treatment of patients in regard to race; however, our sensitivity analysis did not find a strong likelihood of confounding by use of hormone therapy. The limitations of this study are that it was a retrospective analysis from a single institution and subject to referral bias. Although we used a common definition of SES, this is limited in its ability to measure actual income, education, and access to health care. For comorbid disease we relied on selfreported disease ascertained from the medical record. This analysis was limited to the main comorbid contributors, but it may have omitted important contributors to early mortality (8). Because this database was created specifically for the purpose of evaluating breast conservation outcomes and cosmesis, we lack data on mastectomy outcomes. It is possible that women undergoing mastectomy have more advanced disease that may have changed some of our observations. However, one may argue that exclusion of these patients lessened the differences between racial groups if black women are expected to present with more advanced disease. We also lack data regarding her2/neu status and duration of tamoxifen or aromatase inhibitor therapy. It is also possible that the inclusion of Hispanic and Asian women altered the data if one wishes to make a direct comparison between black women and Caucasian women. Because of these limitations and the potential pitfalls of retrospective subset analyses, we think that an important next step is to validate this finding in a separate cohort including patients treated with breast conservation as well as mastectomy.

With the exception of the Health Insurance Plan of New York Trial, randomized trials of screening mammography have not been conducted in racially diverse populations. Furthermore, whereas randomized trials have shown that screening mammography improves breast cancer–specific survival (23–26), race-based outcomes have not been reported, to our best knowledge, so it is unknown whether the effectiveness of current mammography practices is the same in black and nonblack patients. Although both black and nonblack patients benefit from screening mammography, the magnitude of benefit may vary by race, and the potential for significant improvement exists. The current analysis did not directly compare screened with non-screened patients. This was a narrowly defined group of relatively favorable patients who were amenable to undergoing breast conservation. The treatment between racial groups was similar with the exception of tamoxifen use. Among patients with mammographically detected tumors, black women were no less likely than nonblack women to have undergone a previous screening mammogram (85.2% vs. 86.8%, p = 0.59). Interestingly, the interval since the previous mammogram was shorter (p = 0.07, results section) in black women, yet they had poorer outcomes than did nonblack patients. In our study, tumor size was an important prognostic indicator for black women whose tumors were detected mammographically, whereas in nonblack patients it was not predictive of outcome for any endpoint. There were many significant differences among patient and tumor characteristics in this study. This may be secondary to presentation at a later time in the natural history of disease among black patients or because their disease is more likely to have unfavorable characteristics with current screening recommendations. One possible explanation for the discrepancy in the prognostic effect of tumor size, particularly in mammographically detected tumors, is that in black patients there is a weaker effect of lead time and/or length bias. Either process suggests a more aggressive behavior of tumors in black patients, inasmuch as mammographically detected tumors are known to have indolent features relative to clinically detected tumors (15, 36, 39). If a baseline inequality in metastatic potential exists, then tumors of identical

REFERENCES 1. Smigal C, Jemal A, Ward E, et al. Trends in breast cancer by race and ethnicity: Update 2006. CA Cancer J Clin 2006;56: 168–183. 2. Ghafoor A, Jemal A, Ward E, et al. Trends in breast cancer by race and ethnicity. CA Cancer J Clin 2003;53:342–355. 3. Joslyn SA. Racial differences in treatment and survival from early-stage breast carcinoma. Cancer 2002;95:1759–1766. 4. Hirschman J, Whitman S, Ansell D. The black:white disparity in breast cancer mortality: The example of Chicago. Cancer Causes Control 2007;18:323–333. 5. Crowe JP Jr., Patrick RJ, Rybicki LA, et al. Race is a fundamental prognostic indicator for 2325 northeastern Ohio women with infiltrating breast cancer. Breast J 2005;11:124–128.

6. El-Tamer MB, Homel P, Wait RB. Is race a poor prognostic factor in breast cancer? J Am Coll Surg 1999;189:41–45. 7. Newman LA, Griffith KA, Jatoi I, et al. Meta-analysis of survival in African American and white American patients with breast cancer: Ethnicity compared with socioeconomic status. J Clin Oncol 2006;24:1342–1349. 8. Tammemagi CM, Nerenz D, Neslund-Dudas C, et al. Comorbidity and survival disparities among black and white patients with breast cancer. JAMA 2005;294:1765–1772. 9. Houterman S, Janssen-Heijnen ML, Verheij CD, et al. Comorbidity has negligible impact on treatment and complications but influences survival in breast cancer patients. Br J Cancer 2004; 90:2332–2337.

Race and tumor size in breast cancer d M. A. NICHOLS et al.

10. Carey LA, Perou CM, Livasy CA, et al. Race, breast cancer subtypes, and survival in the Carolina Breast Cancer Study. JAMA 2006;295:2492–2502. 11. Joslyn SA. Hormone receptors in breast cancer: Racial differences in distribution and survival. Breast Cancer Res Treat 2002;73:45–59. 12. Woodward WA, Huang EH, McNeese MD, et al. AfricanAmerican race is associated with a poorer overall survival rate for breast cancer patients treated with mastectomy and doxorubicin-based chemotherapy. Cancer 2006;107:2662–2668. 13. Jones BA, Kasl SV, Howe CL, et al. African-American/White differences in breast carcinoma: p53 alterations and other tumor characteristics. Cancer 2004;101:1293–1301. 14. Porter PL, Lund MJ, Lin MG, et al. Racial differences in the expression of cell cycle-regulatory proteins in breast carcinoma. Cancer 2004;100:2533–2542. 15. Smith-Bindman R, Miglioretti DL, Lurie N, et al. Does utilization of screening mammography explain racial and ethnic differences in breast cancer? Ann Intern Med 2006;144:541–553. 16. Blackman DK, Bennett EM, Miller DS. Trends in self-reported use of mammograms (1989-1997) and Papanicolaou tests (1991-1997): Behavioral Risk Factor Surveillance System. MMWR CDC Surveill Summ 1999;48:1–22. 17. Swan J, Breen N, Coates RJ, et al. Progress in cancer screening practices in the United States: Results from the 2000 National Health Interview Survey. Cancer 2003;97:1528–1540. 18. Tangka FK, Dalaker J, Chattopadhyay SK, et al. Meeting the mammography screening needs of underserved women: The performance of the National Breast and Cervical Cancer Early Detection Program in 2002-2003 (United States). Cancer Causes Control 2006;17:1145–1154. 19. Wojcik BE, Spinks MK, Optenberg SA. Breast carcinoma survival analysis for African American and white women in an equal-access health care system. Cancer 1998;82:1310–1318. 20. Shavers VL, Brown ML. Racial and ethnic disparities in the receipt of cancer treatment. J Natl Cancer Inst 2002;94:334–357. 21. Unger JM, Coltman CA, Afr, Barlogie B, et al. African Americans have worse survival in hormone-related cancers: A Southwest Oncology Group (SWOG) study. Proc Am Soc Clin Oncol 2003;22:529. 22. Roberts MM, Alexander FE, Anderson TJ, et al. Edinburgh trial of screening for breast cancer: Mortality at seven years. Lancet 1990;335:241–246. 23. Nystrom L, Andersson I, Bjurstam N, et al. Long-term effects of mammography screening: Updated overview of the Swedish randomised trials. Lancet 2002;359:909–919. 24. Shapiro S. Periodic screening for breast cancer: The HIP Randomized Controlled Trial. Health Insurance Plan. J Natl Cancer Inst Monogr 1997;2730. 25. Miller AB, To T, Baines CJ, et al. The Canadian National Breast Screening Study-1: breast cancer mortality after 11 to 16 years of follow-up. A randomized screening trial of mammography in women age 40 to 49 years. Ann Intern Med 2002;137:305–312. 26. Moss SM, Cuckle H, Evans A, et al. Effect of mammographic screening from age 40 years on breast cancer mortality at 10

27. 28. 29. 30.

31. 32.

33.

34.

35. 36. 37. 38. 39. 40. 41. 42.

399

years’ follow-up: A randomised controlled trial. Lancet 2006; 368:2053–2060. Frisell J, Lidbrink E, Hellstrom L, et al. Followup after 11 years: Update of mortality results in the Stockholm mammographic screening trial. Breast Cancer Res Treat 1997;45:263–270. Cross CK, Harris J, Recht A. Race, socioeconomic status, and breast carcinoma in the U.S: What have we learned from clinical studies. Cancer 2002;95:1988–1999. Baquet CR, Commiskey P. Socioeconomic factors and breast carcinoma in multicultural women. Cancer 2000;88:1256– 1264. Nanda R, Schumm LP, Cummings S, et al. Genetic testing in an ethnically diverse cohort of high-risk women: A comparative analysis of BRCA1 and BRCA2 mutations in American families of European and African ancestry. JAMA 2005;294:1925– 1933. Struewing JP, Hartge P, Wacholder S, et al. The risk of cancer associated with specific mutations of BRCA1 and BRCA2 among Ashkenazi Jews. N Engl J Med 1997;336:1401–1408. Miller BA, Hankey BF, Thomas TL. Impact of sociodemographic factors, hormone receptor status, and tumor grade on ethnic differences in tumor stage and size for breast cancer in US women. Am J Epidemiol 2002;155:534–545. Porter PL, El-Bastawissi AY, Mandelson MT, et al. Breast tumor characteristics as predictors of mammographic detection: Comparison of interval- and screen-detected cancers. J Natl Cancer Inst 1999;91:2020–2028. Morris GJ, Naidu S, Topham AK, et al. Differences in breast carcinoma characteristics in newly diagnosed African-American and Caucasian patients: A single-institution compilation compared with the National Cancer Institute’s Surveillance, Epidemiology, and End Results database. Cancer 2007;110: 876884. Amend K, Hicks D, Ambrosone CB. Breast cancer in AfricanAmerican women: Differences in tumor biology from European-American women. Cancer Res 2006;66:8327–8330. Shen Y, Yang Y, Inoue LY, et al. Role of detection method in predicting breast cancer survival: Analysis of randomized screening trials. J Natl Cancer Inst 2005;97:11951203. Koscielny S, Tubiana M, Le MG, et al. Breast cancer: Relationship between the size of the primary tumour and the probability of metastatic dissemination. Br J Cancer 1984;49:709715. Gilliland FD, Joste N, Stauber PM, et al. Biologic characteristics of interval and screen-detected breast cancers. J Natl Cancer Inst 2000;92:743–749. Ma L, Fishell E, Wright B, et al. Case-control study of factors associated with failure to detect breast cancer by mammography. J Natl Cancer Inst 1992;84:781–785. Grambsch PTT. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;81:515–526. Gray R. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988;16:1140–1154. Horton NJLS. Multiple imputation in practice: Comparison of software packages for regression models with missing variables. J Am Stat Assoc 2001;55:244–254.