Abstract
Background Black race is associated with worse outcome in patients with breast cancer. We evaluated distant relapse-free survival (DRFS) between Black and White women with localized breast cancer who participated in NCI-sponsored clinical trials.
Methods We analyzed pooled data from eight National Surgical Adjuvant Breast and Bowel Project (NSABP) trials including 9,702 women with localized breast cancer treated with adjuvant chemotherapy (AC, n=7,485) or neoadjuvant chemotherapy (NAC, n=2,217), who self-reported as Black (n=1,070) or White (n=8,632). The association between race and DRFS was analyzed using log-rank tests and multivariate Cox regression.
Results After adjustment for covariates including age, tumor size, nodal status, body mass index and taxane use, and treatment (AC vs. NAC), Black race was associated with an inferior DRFS in ER-positive (HR 1.24 [95% CI 1.05-1.46], p=0.01), but not in ER-negative disease (HR 0.97 [95% CI 0.83-1.14], p=0.73), and significant interaction between race and ER status was observed (p=0.03). There was no racial disparity in DRFS among patients with pathologic complete response (pCR) (Log-rank p =0.8). For patients without pCR, black race was associated with worse DRFS in ER-positive (HR 1.67 [95% CI 1.14-2.45], p=0.01), but not in ER-negative disease (HR 0.91 [95% CI 0.65-1.28], p=0.59).
Conclusion Black race was associated with significantly inferior DRFS in ER-positive localized breast cancer treated with AC or NAC, but not in ER-negative disease. In the NAC group, racial disparity was also observed in patients with residual ER-positive breast cancer at surgery, but not in those who had a pathologic complete response.
Lay Summary Black women with breast cancer have worse outcome compared to White women. We investigated if this held true in the context of clinical trials which provide controlled treatment setting. Black women with cancer containing estrogen receptors (ER) had worse outcome than White women. If breast cancers did not contain ER there was no racial disparity in outcome. Moreover, we observed racial disparity in women who received chemotherapy before their cancer was removed, but only if they had cancer with ER and residual disease upon completion of treatment. If cancer disappeared upon pre-surgical chemotherapy, there was no racial disparity.
Background
Breast cancer is the second most commonly diagnosed cancer and the second leading cause of cancer death among U.S. women 1,2. Black race or African American (AA) ethnicity has been associated with worse outcome in both women and men with localized and advanced breast cancer3-5. This disparity has been attributed to patient- and tumor-specific factors such as more advanced disease at presentation 6, higher rates of obesity 7, and higher rates of unacceptable side effects to chemotherapy 8,9, as well as differences in tumor microenvironment (TME) and systemic inflammatory signatures 10. Other factors contributing to racial disparities include differences in socioeconomic status and limitations in access to care 11. Studying race and ethnicity-dependent outcome has been challenging due to heterogeneity within racial and ethnic groups and difficulty in controlling for social determinants of health.
Clinical trials provide an opportunity to control for some of the social determinants of health, such as access to care, medical comorbidities, as well as variations in treatment and adherence to therapy. However, Blacks and Hispanics are consistently underrepresented in phase III oncologic clinical trials, including trials sponsored by the U.S. National Cancer Institute (NCI) 12,13.
Efforts have been made to overcome this limitation by pooling data from NCI-sponsored adjuvant trials in localized breast cancer. For example, a pooled analysis of four trials (B-13, 14, 19, and 20) coordinated by the National Surgical Adjuvant Breast and Bowel Project (NSABP) including 5,444 patients with localized breast cancer that completed accrual prior to 1988 found that there was no racial disparity in cancer-specific outcomes 14. On the other hand, in an analysis of 35 phase III trials coordinated by the Southwestern Oncology Group (SWOG), including 19,547 patients with a variety of solid tumors and hematologic malignancies, there was a racial disparity, even after controlling for clinical prognostic factors, income, and education, among those with breast, ovarian, or prostate cancer 15. Similarly, Black race was associated with a worse outcome in those with hormone receptor (HR)-positive, HER2-negative disease after controlling for covariates including obesity in a trial including 4,817 women coordinated by the Eastern Cooperative Oncology Group (ECOG) 7. Moreover, a recent analysis of another NCI-sponsored trial (TAILORx) that included 10,273 women demonstrated a higher distant recurrence rate in Black compared to White participants with HR-positive HER2-negative disease despite similar 21-gene assay recurrence scores16.
Neoadjuvant chemotherapy (NAC) has emerged as an effective means to downstage locally advanced disease and to allow for breast conservation and less axillary surgery 17. Although single-institution, retrospective studies have demonstrated a worse distant recurrence free survival (DRFS) in Blacks compared to Whites treated with NAC 18,19, this has been previously investigated in only one prospective clinical trial, which showed a trend towards worse DRFS in Blacks compared to Whites, though not statistically significant20.
Given the discrepant results of AC trial analyses and lack of NAC studies, we aimed to evaluate racial disparity in DRFS using pooled analysis of 8 NSABP trials including 9,702 women with localized breast cancer receiving NAC or AC regimens including doxorubicin and cyclophosphamide with or without a taxane; this analysis did not include 4 trials reported in a prior NSABP analysis14. NSABP is an NCI-sponsored clinical trials cooperative group that has research sites at nearly 1,000 cancer treatment centers in the United States, Canada, Puerto Rico, Australia, and Ireland.
Methods
Study/patient selection
A pooled analysis of eight NSABP trials was performed (Appendix 1), including five trials of only adjuvant chemotherapy (B-15, B-22, B-23, B-28, B-30), one trial including both adjuvant and neoadjuvant chemotherapy arms (B-18 arm 1 and arm 2), and two trials including only neoadjuvant chemotherapy (B-27 arm 1 and arm 2, B-40 arm 1A); patients enrolled in the B-27 3rd arm including both AC and NAC were not included in this analysis. The accrual dates for the trials are included in Supplemental Table 1S. These eight trials were selected by reviewing treatment with doxorubicin, cyclophosphamide with or without a taxane for localized invasive cancer. Of the trials included in our analyses, onlyB-40 had information regarding HER2 expression available. Therefore, we could not perform subgroup analysis according to HER2 expression status. Additionally, B-36 was excluded as patients were treated with trastuzumab at the discretion of treating physician. Of the remaining trials, there were 14 involving doxorubicin and cyclophosphamide. To minimize the effects of treatment variations, we only included the trials in which patients received doxorubicin 60mg/m2 IV and cyclophosphamide 600 mg/m2 IV with or without a taxane every 3 weeks (n=9). B-16 was excluded due to its limited eligibility criteria. Lastly, B-29 (arm 3) was excluded because it included <8% Black participants. AC treatment group was defined as patients who received surgery first, followed by chemotherapy. NAC treatment group was defined as those who received all chemotherapy prior to surgery (B-27 arm 3 was excluded due to receipt of taxane after surgery). Of the 10,522 patients in these trials, only patients of Black (n=1,070, 11.0%) or White (n=8,632, 89.0%) race were included in the analyses. Race was self-reported, and ethnicity was not considered (i.e., White race encompasses both Hispanic and non-Hispanic Whites).
Patient characteristics, entire cohort.
ǂPathologic tumor size and node status for adjuvant group; clinical (pre-treatment) tumor size and node status for neoadjuvant group
**Obesity defined as BMI ≥ 30kg/m2
Statistical analysis
Patient and tumor characteristics such as age (years), ER status (positive vs. negative), tumor size (≥2 cm vs. < 2 cm), nodal status (positive vs. negative), obesity status (obese vs. non-obese) and taxane use (yes vs. no), were compared using Wilcoxon rank sum tests for continuous variables and Pearson’s chi-squared tests for categorical variables. Obesity was defined as BMI ≥30kg/m2. Pathological tumor size and node status were used for AC group while clinical tumor size and node status were used for NAC group. Pathologic complete response (pCR) was defined as no residual disease in the breast and was also compared between Black and White patients in the NAC cohort.
The outcome evaluated was distant relapse-free survival (DRFS), defined as time to first distant relapse or a second primary cancer. Death prior to a distant recurrence or second primary cancer was censored. The median follow-up was 115.2 months, and we censored the follow-up beyond 210 months. Kaplan-Meier survival curves were estimated for each racial group (adjusted Kaplan-Meier curves and log-rank tests21 were also estimated, see Appendix 2) and log-rank tests were used to compare survival between Black and White. Multivariate Cox proportional hazard models stratified were used to compare survival between Black and White patients, while adjusting for age, ER status, tumor size, nodal status, obesity status and taxane use and treatment. Because prior studies indicated racial disparities were evident primarily in ER-positive breast cancer7,16, separate Cox models were used for ER-positive and ER-negative disease. A formal comparison of racial disparity between ER-positive and ER-negative disease was made by testing the interaction between race and ER status. Within the NAC cohort, separate analyses were also performed by pCR, because pCR pertains good long-term prognosis. Since prior studies indicated racial disparity in ER positive breast cancer7,16, separate analyses for patients with ER+ and ER-disease were performed in patients with residual cancer after NAC.
The proportionality of the Cox models was examined based upon Schoenfeld residuals22,23, and no violation was identified.
Statistical significance was specified a priori as p<0.05 and the p-values reported are two-sided. All analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC 2014).
This study was reviewed by the Albert Einstein College of Medicine Institutional Review Board and was deemed exempt status (IRB# 2018-8959).
Results
Patient characteristics
There were 9,702 patients included in our analysis, all women, of whom 1,070 (11.0%) were Black and 8632 (89.0%) were White. Of note, patients characterized as “other”, and Hispanics were not included in the analysis. It was not clear from the data provided what race was included in the “other”. Likewise, patients in the Hispanic group were not categorized as Black or White. Of these, 7,485 (77.1%) were in the AC treatment group and 2,217 (22.9 %) were in the NAC treatment group (Table 1). Compared to AC treatment group, patients in the NAC treatment group had more node negative disease (68.9% vs 15.7% p<0.001) and more often had tumors ≥2cm (93.3% vs 55.3%, p<0.001). The NAC treatment group also had a higher proportion of Black patients (12.5% vs 10.6%, p=0.01). There was no difference between treatment groups in age at diagnosis, prevalence of obesity, and receipt of taxane (Table 1). The differences in tumor size and node status were attributable to the differences in enrollment criteria for the trials (Appendix 1). Furthermore, in the AC treatment group, we compared the pathologic tumor size and node status, and in the NAC treatment group, the clinical, or pre-treatment tumor size and node status.
Clinical outcomes in the entire cohort
In the entire cohort, Black patients had a worse DRFS compared to White patients (Log-Rank p<0.001) (Figure 1A), and Black race remained significantly associated with a worse DRFS after adjusting for age, ER status, tumor size, node status, obesity, receipt of taxane, and treatment (AC vs. NAC) (HR 1.13 95% CI 1.01-1.25, p=0.03) (Figure 1B). However, multivariate analysis stratifying patients by ER status showed that Black race was significantly associated with a worse DRFS in ER-positive disease (HR 1.24 [95% CI 1.05-1.46], p=0.01) (Figure 1C), but not in ER-negative disease (HR 0.97 [95% CI 0.83-1.14], p=0.73) (Figure 1D). The Cox model including the interaction term revealed significant interaction between race and ER status (p=0.03) (Table 2S).
Association between race and DRFS, entire cohort. A) Kaplan-Meier curve describing distant-relapse free survival in entire cohort. Log-rank test was used to compare survival in White vs Black race patients. B) Cox-regression model adjusting for co-variates and stratified by protocol. Comparison groups: White race, ER-negative, node negative, tumor <2cm, non-obese (BMI <30kg/m2), no taxane, adjuvant chemotherapy. C) Cox-regression model in ER-positive disease only. D) Cox-regression model in ER-negative disease only. ‡Pathologic tumor size and node status for the adjuvant group; clinical (pre-treatment) tumor size and node status for the neoadjuvant group
Given the differences in tumor characteristics between AC (Table 3S) and NAC (Table 4S) treatment groups, we repeated the analyses stratified by use of AC or NAC. Although, Black patients had worse DRFS compared to White patients the association was not statistically significant after adjustment for covariates including obesity (Figures 1S and 2S).
Clinical outcomes in the NAC treatment group
As pCR portends to better long-term outcome, the NAC treatment group was further stratified by pCR status. There were 394 patients (17.8%) who achieved pCR, of whom 61(15.5%) were Black and 333 (84.5%) were White. There were 1,731 (81.5%) patients who did not achieve pCR, of whom 198 (11.4%) were Black and 1,533 (88.6%) were White. pCR status was unknown in 92 (4.3%) of the patients. Among patients who achieved pCR, there was no racial disparity in DRFS (Log-Rank p=0.81) (Figure 2A). Among those who did not achieve pCR, however, Black patients had a significantly worse DRFS (Log-Rank p=0.002) (Figure 2B). However, after adjusting for covariates, Black race was not significantly associated with a worse DRFS (HR 1.23 95% CI [0.98-1.53], p=0.08) (Figure 2C).
Association between race and DRFS in neoadjuvant treatment cohort, stratified by pCR. A) Kaplan-Meier curve describing distant-relapse free survival in patients who achieved pathologic complete response following neoadjuvant chemotherapy. Log-rank and Wilcoxon tests were used to compare survival in White vs Black race patients. B) Kaplan-Meier curve describing distant-relapse free survival in patients who did not achieve pathologic complete response following neoadjuvant chemotherapy. Log-rank and Wilcoxon tests were used to compare survival in White vs Black race patients. C) Cox-regression model adjusting for co-variates in no pCR only. D) Cox-regression model in ER-positive disease with no pCR. E) Cox-regression model in ER-negative disease with no pCR. Comparison groups: White race, ER-negative, node negative, tumor <2cm, non-obese (BMI <30kg/m2), no taxane.
Because prior studies show racial disparity in patients with ER+ disease the NAC treatment group without pCR was stratified by ER status (Figure 2D, E). In ER-positive disease, Black race was significantly associated with a worse DRFS even after adjusting for all covariates (HR 1.67 [95% CI 1.14-2.45], p=0.009) while no significant association with DRFS was found in ER-negative group (HR 0.91 [95% CI 0.65-1.28], p=0.59). Although the estimates of racial disparity were very different between the ER groups, the interaction between race and ER status was not significant (p=0.13) (Table 5S).
In summary, there was no racial disparity in DRFS among patients who achieved pCR. For patients who did not achieve pCR, black race was associated with worse DRFS in ER-positive disease, but not in ER-negative disease.
Discussion
In this pooled analysis of data from eight prospective NCI-sponsored clinical trials, we found that after adjusting for covariates including age, tumor size, node status, obesity, receipt of taxane and treatment (AC vs. NAC), Black race was associated with a worse DRFS in ER-positive disease but not in ER-negative disease. Formal testing in the entire cohort demonstrated a significant interaction between race and ER status. Within the NAC cohort, there was no racial disparity in DRFS among patients who achieved pCR. For patients who did not achieve pCR, black race was only associated with worse DRFS in ER-positive disease, but not in ER-negative disease. However, formal testing did not reveal a significant interaction between race and ER status, likely due to a limited sample size.
Utilizing data from patients who qualify for participation in national NCI-sponsored clinical trials minimizes some of the potential factors contributing to racial disparities noted in population-based studies, including access to care and significant comorbidities. Furthermore, utilizing clinical trial data ensured administration of and adherence to a pre-determined treatment regimen. However, the social determinants of health were not considered in these trials. The patient accrual took place more than twenty years ago, when there was less known about how these parameters may affect the long-term outcome of some patients.
The findings of this study are in line with the existing literature that Black women receiving chemotherapy for localized breast cancer have an inferior DRFS 14,24-28. Consistent with prior reports on racial disparity in ER+ disease7,16, Black race was significantly associated with higher distant recurrence rates in ER-positive disease when adjusted for other prognostic covariates in the overall cohort. In addition, we showed here that Black race was associated with worse DRFS in patients with ER+ residual tumors following neoadjuvant therapy. The decision to stratify according to ER status and pCR was also based on parameters affecting long term outcomes in patients. ER positive, compared to ER negative disease tends to have a weaker response to chemotherapy, and pCR is associated with a better long-term outcome. Further research is needed to better understand whether the observed disparity is due to adherence to endocrine therapy, and/or biological, environmental and/or social determinants of health.
Consistent with prior reports, we did not observe substantial racial disparities in pCR rates after NAC29-32. In fact, Black women in our analysis had somewhat higher pCR rates than Whites (21.9% vs. 17.2%, p=0.0051), which is likely due to a higher prevalence of basal-like or triple negative subtype in Blacks 33,34 that is more chemosensitive 35-39. Due to limitations in the dataset, we were not able to evaluate molecular subtypes in this study directly, although the higher prevalence of ER-negative disease among Blacks would suggest that Black women in our group have a higher prevalence of basal-like subtype in line with the existing literature40.
Strengths of this study include the large number of patients included in the analysis, the long follow-up duration, inclusion of subjects who had access to care and absence of comorbidities that would preclude optimal adjuvant systemic chemotherapy and endocrine therapy, and the use of standard state-of-the-art chemotherapy regimens in controlled clinical settings. Limitations include use of self-reported race, absence of information regarding social determinants of health, and absence of information on treatment administration and adherence. Race is a social construct, and self-identified race has an imperfect correlation with geographical ancestry as determined by genetic studies 41-43. Furthermore, both Black and White races are heterogeneous populations, and there is emerging evidence that presentation and outcomes within Black race vary widely depending on additional ethnic variables such as country of origin 44-46. This intra-racial heterogeneity merits further exploration in future analyses and prospective collection of genetic ancestry may be a step towards personalized prognosis and treatment options. Regarding social determinants of health, patient eligibility for enrollment to clinical trials controls for some but not all of these factors. One study found that after adjusting for poverty and Medicaid enrollment, race was not in and of itself a risk factor for worse survival 47. However, a pooled analysis of two ECOG clinical trials found that insurance and neighborhood SES did not affect outcomes 48.
In conclusion, in this pooled analysis of eight NSABP trials including 9,702 women with localized breast cancer treated with adjuvant or neoadjuvant systemic chemotherapy including cyclophosphamide and doxorubicin without a taxane, Black women with ER-positive disease had a worse DRFS compared to White women, especially in those with residual cancer following NAC. Further research is needed to fully understand the mechanisms behind this disparity.
Data Availability
All data produced in the present study are available upon reasonable request to the authors
Notes
Role of the funder
Supported in part by grants from the Department of Health and Human Services and the National Institutes of Health (5P30CA013330, 1UG1CA189859, T32CA200561), Peter T Rowley (DOH01-ROWLEY-2019-00037), U10CA180868, UG1CA189867, and U10CA180822.
Author disclosures
None
Author contributions
GK, JMP, GSK, DE, XX, MHO—conceptualization; GK, JQ, SA, TJ,GY —data curation; JQ, JL, SA, GY, XX—Formal analysis & methodology; GK, JMP— writing – original draft; All authors—Writing – review & editing
Prior presentations
This work was presented at American Association for Cancer Research Annual Meeting on April 10, 2021 (virtual meeting).
Acknowledgements
We would like to acknowledge NRG Oncology/NSABP for providing us the data that was used for this study.