Assessment of Racial Disparity in Survival Outcomes for Early Hormone Receptor–Positive Breast Cancer After Adjusting for Insurance Status and Neighborhood Deprivation: A Post Hoc Analysis of a Randomized Clinical Trial

Key Points

Question  What is the association between individual (ie, insurance) and structural (ie, neighborhood deprivation index) social determinants of health and racial disparities in clinical outcomes among women with breast cancer in the Trial Assigning Individualized Options for Treatment?

Findings  In this post hoc analysis of a randomized clinical trial including 9719 women with breast cancer, Black race compared with White race was associated with inferior relapse-free interval and overall survival after adjustment for insurance, neighborhood deprivation, and early discontinuation of endocrine therapy.

Meaning  The findings of this study suggest that individual insurance and neighborhood deprivation independently correlated with overall survival, with Black women with early hormone receptor–positive breast cancer continuing to experience shorter relapse-free intervals and overall survival compared with White women.

Importance  Racial disparities in survival outcomes among Black women with hormone receptor–positive breast cancer have been reported. However, the association between individual-level and neighborhood-level social determinants of health on such disparities has not been well studied.

Objective  To evaluate the association between race and clinical outcomes (ie, relapse-free interval and overall survival) adjusting for individual insurance coverage and neighborhood deprivation index (NDI), measured using zip code of residence, in women with breast cancer.

Design, Setting, and Participants  This was a post hoc analysis of 9719 women with breast cancer in the Trial Assigning Individualized Options for Treatment, a randomized clinical trial conducted from April 7, 2006, to October 6, 2010. All participants received a diagnosis of hormone receptor–positive, ERBB2-negative, axillary node–negative breast cancer. The present data analysis was conducted from April 1 to October 22, 2021.

Main Outcomes and Measures  A multivariate model was developed to evaluate the association between race and relapse-free interval and overall survival adjusting for insurance and NDI level at study entry, early discontinuation of endocrine therapy 4 years after initiation, and clinicopathologic characteristics of cancer. Median follow-up for clinical outcomes was 96 months.

Results  A total of 9719 women (4.2% [n = 405] Asian; 7.1% [n = 693] Black; 84.3% [n = 8189] White; 4.4% [n = 403] others/not specified) were included; 9.1% of included women [n = 889] were Hispanic or Latino. Median (SD) age was 56 (9.2) years. In multivariate models, Black race compared with White race was associated with statistically significant shorter relapse-free interval (hazard ratio [HR], 1.39; 95% CI, 1.05-1.84; P = .02) and overall survival (HR, 1.49; 95% CI, 1.10-2.99; P = .009), adjusting for insurance and NDI level at study entry and other factors. Although uninsured status was not associated with clinical outcomes, patients with Medicare (HR, 1.30; 95% CI, 1.01-1.68; P = .04) and Medicaid (HR, 1.44; 95% CI, 1.01-2.05; P = .05) had shorter overall survival compared with those with private insurance. Participants living in neighborhoods in the highest NDI quartile experienced shorter overall survival compared with those in the lowest quartile (HR, 1.34; 95% CI, 1.01-1.77; P = .04), regardless of self-identified race.

Conclusions and Relevance  The findings of this post hoc analysis of a randomized clinical trial suggest that Black women with breast cancer have significantly shorter relapse-free interval and overall survival compared with White women. Early discontinuation of endocrine therapy, clinicopathologic characteristics, insurance coverage, and NDI do not fully explain the observed disparity.

Trial Registration  ClinicalTrials.gov Identifier: NCT00310180

Breast cancer is the second leading cause of cancer death, with 43 600 women in the US expected to die in 2021 of breast cancer.¹ Black women have higher death rates from breast cancer compared with any other racial or ethnic groups in the US. The mortality gap was reported to be 39% higher in Black women compared with White women in 2015, although the incidence rate appears to have converged among Black and White women.² Poor survival outcomes among Black women remain unexplained and are likely due to a combination of factors including genetics, social determinants of health, differences in screening, stage at diagnosis, tumor biological characteristics, treatment difference, and other nonbiological factors.^3-7

A prior study reported that Black women with hormone receptor–positive, human epidermal growth factor receptor 2 (ERBB2)–negative, axillary node–negative breast cancer participating in the Trial Assigning Individualized Options for Treatment (TAILORx)⁸ had worse clinical outcomes, including shorter relapse-free interval and overall survival, despite similar 21-gene assay recurrent score results and comparable systemic therapy.⁹ However, the study did not adjust for endocrine therapy adherence, early discontinuation of therapy, or structural inequity (eg, neighborhood deprivation).⁹

Nonadherence and early discontinuation are substantial barriers to the delivery of endocrine therapy and may result in recurrence and mortality.¹⁰ Although in some studies, Black women compared with White women have been shown to have a higher likelihood of early discontinuation of endocrine therapy,^11-15 after adjusting for social determinants of health, such as neighborhood deprivation and insurance effect, Black women participating in the TAILORx trial had significantly lower rates of early discontinuation of endocrine therapy (hazard ratio [HR], 0.73; 95% CI, 0.57-0.93).¹⁶ This finding suggested that the observed racial disparities in early discontinuation of endocrine therapy seen in other studies^11-15 may have been explained by social determinants of health.

In the present study, we aimed to perform a post hoc analysis evaluating the association between race and clinical outcomes (ie, relapse-free interval and overall survival) adjusting for clinicopathologic characteristics, insurance as a proxy for individual social determinants of health, neighborhood deprivation index as a proxy for structural inequity, and early discontinuation of endocrine therapy among women enrolled in the TAILORx trial.

TAILORx was a prospective, National Cancer Institute–funded randomized clinical trial that was coordinated by the Eastern Cooperative Oncology Group (ECOG) and subsequently the ECOG-American College of Radiology Imaging Network Cancer Research Group.⁸ The trial was approved by the National Cancer Institute Central Institutional Review Board and appropriate local institutional review boards. Participants were registered between April 7, 2006, and October 6, 2010; post hoc analysis was conducted from April 1 to October 22, 2021.

The trial population included women with hormone receptor–positive, ERBB2-negative, axillary node–negative breast cancer. Women were required to provide written informed consent, including a willingness to have treatment assigned or randomized based on the recurrence score results (determined with Oncotype DX, Genomic Health Inc). Women with a recurrence score of 10 or lower were assigned to receive endocrine therapy only, and women with a score of 26 or higher were assigned to receive chemotherapy plus endocrine therapy. Women with a midrange score of 11 to 25 underwent randomization and were assigned to receive either endocrine therapy alone or chemotherapy plus endocrine therapy. Self-reported race was categorized as Asian, Black, American Indian/Alaska Native, Native Hawaiian or Other Pacific Islander, White, multirace, and unknown or not reported. The trial protocol and statistical analysis plan are available in Supplement 1. Additional details on the TAILORx trial have been reported,^8,17,18 and the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline has been followed.

Insurance status at study entry was determined using standard categories collected for all National Clinical Trials Network studies at the time of registration and served as an individual-level social determinant of health. As described in a prior study,¹⁶ the neighborhood deprivation index (NDI) represented a measure of structural inequity. The NDI level was calculated using the Agency for Healthcare Research and Quality socioeconomic status index¹⁹ and by linking patients’ 5-digit zip codes at the time of registration to county-level data using the 2016-2017 Health Resource and Services Administration Area Health Resources File, which includes data on population characteristics and economics.²⁰ The index is a weighted combination of the percentage of households with a mean number of 1 person or more per room, the median value of owner-occupied housing units, the percentage living below the poverty level, the median household income, the percentage aged 25 years or older with a bachelor’s degree or higher, the percentage aged 25 years or older with less than a 12th-grade education, and the percentage aged 16 years or older in the labor force who are unemployed.¹⁹ When a zip code represented multiple counties, aggregate means and totals from those multiple counties were used to represent the county-level estimates for that zip code. The index was not calculated for cases with unknown zip codes as well as those from Puerto Rico or international sites. The NDI scores range between 0 and 100, with higher scores representing greater neighborhood deprivation.

As described in a prior study,¹⁶ early discontinuation was measured as a time-dependent binary variable defined as stopping medication within 4 years of start for reasons other than distant recurrence or death. Patients who were lost to follow-up or withdrew from the study within 4 years of starting therapy but reported receipt of endocrine therapy on their last follow-up were analyzed as still receiving endocrine therapy with duration censored.

Study end points included relapse-free interval and overall survival as defined by the Standardized Definitions for Efficacy Endpoints criteria,²¹ with a median follow-up of 96 months after enrollment. Overall survival rates were estimated using the Kaplan-Meier method.⁹

For each of the 2 study end points (relapse-free interval and overall survival), a multivariate model was developed to assess correlation with race adjusted for other factors. There was not a prospective statistical analysis plan written for this analysis, but the modeling was built on a prior publication.^9,16 The 2 factors added to the models in the present study were NDI and early termination of endocrine therapy. We also attempted to adjust for possible outcomes associated with early termination of endocrine therapy by incorporating a time-dependent covariate. Except for duration of endocrine therapy, all covariates were evaluated only at baseline. Factors included were self-reported race (Asian, Black, White, other, and unknown or not reported), age (≤40, 41-50, 51-60, 61-70, and ≥70 years), continuous recurrence score, tumor size (>2  vs ≤2 cm), histologic grade (high vs medium vs low vs unknown), insurance status (private, Medicare, Medicaid, uninsured, and international), NDI level (4 quartiles), early discontinuation of endocrine therapy within 4 years of start (yes/no), and lack of receipt of endocrine therapy (yes/no). Early discontinuation of medication was treated as a time-dependent covariate with patients analyzed as entering the early discontinuation state 150 days after the last treatment was given. The data for formal classification of the tumor by molecular subtypes, such as luminal A vs B, are not available in TAILORx. However, we included histologic grade and recurrence score in the models, which generally show an association with luminal A vs B in this population.^22,23 Because patients needed to be within 84 days of their primary surgery at the time of enrollment, age at enrollment (included in the model) is either similar to or within 1 year from age at diagnosis.

Tests for individual variables, HRs, and 95% CIs are reported for each factor. The only interaction formally examined was for race by the NDI. The power for interactions of race with other factors is limited, owing to small numbers of non-White patients in subsets. Missing data were handled for most factors by including unknown (or other/unknown) as a separate category. There were 3 cases missing tumor size that were excluded from the models. Proportional hazards were examined as previously described in Albain et al⁹ (estimating effects separately in the follow-up intervals 0-5 and >5 years). Two-sided P values <.05 were considered statistically significant. Analyses were done in R, version 4.0.1 (R Foundation for Statistical Computing).

A total of 10 253 eligible women were registered, and 9719 with follow-up information were included in the analysis (eFigure in Supplement 2). Baseline demographic characteristics of included patients are reported in the eTable in Supplement 2 and reported in more detail in a prior study.⁹ A total of 4.2% (n = 405) of the women were Asian, 7.1% (n = 693) were Black, 84.3% (n = 8189) were White, and 4.4% (n = 432) were of other or unknown or not reported race. Furthermore, 9.1% (n = 889) of the women were Hispanic or Latino. Median (SD) age was 56 (9.2) years. Endocrine therapy was terminated in 11.9% (n = 1152) of the women within 4 years after the start of medication, and 1.8% (n = 179) did not initiate endocrine therapy.

Table 1 and Table 2 report the multivariate models of the estimated HRs for relapse-free interval and overall survival based on race adjusting for social determinants of health, early discontinuation of endocrine therapy, and demographic and clinicopathologic characteristics. After adjustment for relevant covariates, compared with White women, women with self-reported Black race were more likely to experience significantly shorter relapse-free intervals (HR, 1.39; 95% CI, 1.05-1.84; P = .02) and shorter overall survival (HR, 1.49; 95% CI, 1.10-2.99; P = .009).

In multivariate models, neither patients’ insurance at registration nor NDI level correlated with the relapse-free interval (Table 1). However, patients with Medicare (HR, 1.30; 95% CI, 1.01-1.68; P = .04) and Medicaid (HR, 1.44; 95% CI, 1.01-2.05; P = .05) had inferior overall survival compared with those with private insurance (Table 2). Similarly, participants living in neighborhoods in the highest NDI quartile experienced shorter overall survival compared with those living in neighborhoods in the lowest quartile (HR, 1.34; 95% CI, 1.01-1.77; P = .04). The overall survival HRs for Black vs White patients were 1.05 (95% CI, 0.60-1.84) in the most-deprived subset, 1.81 (95% CI, 1.01-3.25) in the second NDI quartile, 1.70 (95% CI, 0.93-3.12) in the third NDI quartile, and 1.67 (95% CI, 0.88-3.17) in the least-deprived subset. The relapse-free interval HRs for Black vs White patients were 0.98 (95% CI, 0.62-1.80) in the most-deprived subset, 1.33 (95% CI, 0.73-2.43) in the second NDI quartile, 2.03 (95% CI,1.22-3.37) in the third NDI quartile, and 1.17 (95% CI, 0.63-2.15) in the least-deprived subset. Although the power for racial comparisons within the NDI subsets is limited, the interaction test for differences in Black vs White women across all 4 NDI quartiles showed persistent racial disparity in overall survival (Figure); disparity was also persistent in the relapse free-interval across all deprivation index quartiles.

Larger tumor size, higher histological grade, higher recurrence score and older age were significantly associated with inferior relapse-free interval (tumor size >2 cm: HR, 2.08; 95% CI, 1.75-2.48; P < .001; high histologic grade: HR, 2.11; 95% CI, 1.57-2.83; P < .001; age ≥71 years: HR, 0.46; 95% CI, 0.28-0.75; P = .002) (Table 1) and overall survival (tumor size >2 cm: HR, 1.65; 95% CI, 1.37-1.98; P < .001; high histologic grade: HR, 1.37; 95% CI, 1.02-1.82; P = .03; age ≥71 years: HR, 3.31; 95% CI, 1.81-6.05; P < .001) (Table 2). The models also included spline terms to adjust for recurrence score effects. Early discontinuation of endocrine therapy was not correlated with relapse-free interval (HR, 1.05; 95% CI, 0.73-1.51; P = .80), but was associated with shorter overall survival (HR, 2.80; 95% CI, 2.22-3.54; P < .001). However, not initiating endocrine therapy was associated with both shorter relapse-free interval (HR, 2.12; 95% CI, 1.21-3.73; P = .009) and shorter overall survival (HR, 2.38; 95% CI, 1.39-4.08; P < .001).

Prior studies have reported racial disparities in survival outcomes among Black women with hormone receptor–positive breast cancer.^9,24-26 However, the degree to which this association between race and survival can be explained by social determinants of health, such as insurance coverage and neighborhood deprivation, has not been considered in most of these studies. Our retrospective analysis of women with breast cancer enrolled in the TAILORx trial noted that individual (ie, insurance) and structural (ie, neighborhood deprivation) social determinants of health independently correlated with overall survival, but not with relapse-free interval. However, persistent associations with self-reported race were identified with Black women experiencing shorter relapse-free intervals and overall survival compared with White women, even after adjusting for social determinants of health, such as insurance and neighborhood deprivation, as well as early discontinuation or noninitiation of endocrine therapy and other clinicopathologic characteristics.

Oral adjuvant endocrine therapy is an important part of treatment for hormone receptor–positive cancer, which includes most breast cancer cases.⁷ Endocrine therapy reduces the risk of recurrence and cancer-specific mortality and improves overall survival.^27-29 Consistent with these findings, our study shows that not initiating endocrine therapy was associated with a shorter relapse-free interval and shorter overall survival. However, once endocrine therapy was initiated, early discontinuation (within 4 years of the start of medication) was not associated with the duration of the relapse-free interval and was associated only with decreased overall survival, with both outcomes evaluated at a median of 96 months’ follow-up after enrollment. Our findings suggest that early discontinuation of endocrine therapy may be a marker for other variables that result in higher mortality risk not currently captured by insurance or neighborhood deprivation. For example, a previous study reported an association between early discontinuation of endocrine therapy and depression,³⁰ which has been shown separately to be associated with higher mortality risk.³¹ In the present study, we did not measure adherence to endocrine therapy. However, in a previous study,³² low adherence rate (<80%) to endocrine therapy was significantly associated with a higher risk of all-cause mortality at 5 years. Furthermore, a high (≥80%) adherence rate over a longer time (eg, 5 years compared with 3 years) was correlated with decreased breast cancer–specific mortality and recurrence at 5 years.³² Self-reported Black race has been associated with endocrine therapy nonadherence, defined as not taking medication every day or missing more than 2 pills in the past 14 days, but not early discontinuation of endocrine therapy, adjusting for clinical characteristics and social determinants of health.⁷ However, in a prior evaluation of TAILORx, Black women had lower rates of early discontinuation of endocrine therapy after adjusting for individual social determinants of health and structural inequity.¹⁶

Although, as reported in a prior study, uninsured women or those with Medicaid were more likely to terminate endocrine therapy early,¹⁶ being uninsured or underinsured in the present study was not associated with inferior relapse-free interval. However, similar to prior studies in cancer,^33,34 patients with Medicare and Medicaid have significantly lower overall survival compared with those with private insurance after adjusting for clinicopathologic factors, early discontinuation of therapy, and patient demographic characteristics. This lower survival might be due to factors such as reduced access to health care or decreased quality of care, with some physicians not accepting patients covered by Medicaid or Medicare or services from safety-net hospitals and health clinics being disproportionately concentrated in areas remote from where underinsured individuals reside, or that underinsured patients decline care due to cost concerns.³⁵ Decreased access to care may affect receipt of preventive services and result in a delay in getting appointments, being referred to specialists, or receiving supportive care.

Furthermore, living in a neighborhood with a higher NDI level was not associated with relapse-free interval but was consistent with prior studies in patients with cancer, showing decreased overall survival compared with living in more affluent neighborhoods.^36,37 The influence of neighborhood on overall mortality among patients with breast cancer is most likely multifactorial. For Black patients with breast cancer in particular, neighborhood has been associated with aggressive tumor subtypes and increased mortality.^38,39 Consequently, a plausible pathway of how neighborhood exerts its effects on overall survival could be a combination of access to care, activation of the stress response secondary to social and environmental challenges, and neighborhood exposures resulting in epigenetic changes.^40-42 The weathering hypothesis, which postulates that the persistent social, political, and economic marginalization faced by Black women results in higher comorbidities and premature death, may be an applicable framework to understand the racial- and neighborhood-based disparities noted in our study.^42,43 In addition, the concept of allostatic load, which describes physiologic dysregulation secondary to chronic activation of the hypothalamic-pituitary-adrenal axis and the autonomic nervous system in the setting of elevated psychosocial stressors (eg, neighborhood deprivation, social isolation, financial hardship, unemployment), provides an avenue to quantitively measure the structural inequity and inequality that Black women face, which may increase their subsequent risk for illnesses such as cancer.^44-46 Prior studies in breast cancer have reported an association between elevated allostatic load, Black race, poor tumor prognostic features (large tumor size and poor tumor differentiation), and aggressive subtypes (estrogen receptor–negative).^47,48 This information provides opportunities for future clinical trials in breast cancer to evaluate the association of longitudinal changes in allostatic load and tumor genetics. As shown with other studies,^49,50 higher histologic grade, larger tumor size, and older age were associated with higher recurrence and decreased overall survival in the present study.

Strengths of our study include the prospective nature of the TAILORx trial, a large sample of women who self-reported Black race, and homogeneity of breast cancer characteristics (ie, hormone receptor–positive and ERBB2-negative). Our study has several limitations. First, the study did not include data on adherence to endocrine therapy, and we only evaluated the association between early discontinuation of endocrine therapy and clinical outcomes. It is possible that some of the race disparity in clinical outcomes is explained by the differences in adherence to endocrine therapy. Second, similar to a prior study,¹⁶ the early discontinuation rate was calculated based on the report of receipt or nonreceipt of medication in the last follow-up. Because exact durations and reasons for discontinuation of endocrine therapy were not recorded, if distant recurrence or death occurred within 3 months of the last endocrine therapy, then the reason for discontinuation was assumed to be distant recurrence or death. Third, we only analyzed the association between clinical outcomes and race, adjusting for insurance and NDI level at study entry. It is possible that a patient’s insurance or place of residence had changed and the association of these changes with clinical outcomes was not assessed. Fourth, given our data included only limited social determinants of health (ie, insurance and NDI), the intersection of social determinants of health and race may not be fully addressed in the context of other individual and structural social determinants of health (eg, percentage of single female-held household, residential segregation).

The results of this post hoc analysis of a randomized clinical trial suggest that, after controlling for clinicopathologic variables, treatment factors, tumor genomics, individual-level insurance status, and neighborhood deprivation, self-reported Black race remains associated with a shorter relapse-free interval and overall survival among patients with hormone receptor–positive ERBB2-negative breast cancer. Possible explanations are most likely a combination of a higher risk genetic profile (eg, increased germline or somatic mutations) compounded by social, behavioral, and environmental factors.⁵¹ These study findings suggest a need for more nuanced socioecological models rooted in ancestry, the stress response, and social determinants of health, both structural and intermediary, and their longitudinal changes to better understand racial disparities in breast cancer outcomes.

Accepted for Publication: October 26, 2021.

Published Online: February 17, 2022. doi:10.1001/jamaoncol.2021.7656

Corresponding Author: Gelareh Sadigh, MD, Department of Radiology and Imaging Sciences, Emory University School of Medicine, 1364 Clifton Rd, Ste BG27, Atlanta, GA 30322 (gsadigh@emory.edu).

Author Contributions: Drs Gray and Sparano had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Sadigh, Gray, Sparano, Yanez, Garcia, Timsina, Sledge, Cella, Wagner, Carlos.

Acquisition, analysis, or interpretation of data: Sadigh, Gray, Sparano, Yanez, Timsina, Obeng-Gyasi, Gareen, Sledge, Whelan, Cella, Wagner, Carlos.

Drafting of the manuscript: Sadigh, Yanez, Obeng-Gyasi, Cella, Carlos.

Critical revision of the manuscript for important intellectual content: Sadigh, Gray, Sparano, Garcia, Timsina, Obeng-Gyasi, Gareen, Sledge, Whelan, Cella, Wagner, Carlos.

Statistical analysis: Gray, Yanez, Timsina, Cella.

Obtained funding: Sparano, Wagner.

Administrative, technical, or material support: Sparano, Obeng-Gyasi, Whelan, Cella, Wagner.

Supervision: Sparano, Carlos.

Conflict of Interest Disclosures: Dr Sadigh reported receiving grants from ECOG-ACRIN pilot grant outside the submitted work. Dr Gray reported grants from US National Cancer Institute (NCI) during the conduct of the study. Dr Sparano reported grants from National Cancer Institute during the conduct of the study. Dr Yanez reported grants from Northwestern University during the conduct of the study. Dr Gareen reported grants from National Cancer Institute during the conduct of the study. Dr Whelan reported non-financial support from Genomic Health Research support outside of this submitted work. Dr Cella reported grants from NCI during the conduct of the study. Dr Wagner reported personal fees from Celgene Inc and personal fees from Athenex Inc outside the submitted work. Dr Carlos reported grants from NCI during the conduct of the study; other from JACR Salary support as editor outside the submitted work. No other disclosures were reported.

Funding/Support: This study was conducted by the ECOG-ACRIN Cancer Research Group (Peter J. O'Dwyer, MD and Mitchell D. Schnall, MD, PhD, group co-chairs) and supported by the National Cancer Institute of the National Institutes of Health under award numbers U10CA180820, U10CA180794, UG1CA189828, UG1CA233160, UG1CA233320, UG1CA233247, and U10CA180863, and Canadian Cancer Society award 704970.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government.