糖心vlog

Key Points

Question听 What are the relative contributions of social determinants of health and tumor biology to racial disparities in cancer-related death among Black and White women with estrogen receptor鈥損ositive, axillary node-negative breast cancer.

Findings听 Racial differences in indicators of aggressive tumor biology that included a genomic biomarker mediated the same proportion of the survival disparity as individual and neighborhood disadvantage.

Meaning听 Disproportionately aggressive tumor biology among Black women may be an important driver of racial disparities in survival from estrogen receptor鈥損ositive, early-stage breast cancer.

Importance听 Black women with hormone receptor鈥損ositive breast cancer experience the greatest racial disparity in survival of all breast cancer subtypes. The relative contributions of social determinants of health and tumor biology to this disparity are uncertain.

Objective听 To determine the proportion of the Black-White disparity in breast cancer survival from estrogen receptor (ER)-positive, axillary node-negative breast cancer that is associated with adverse social determinants and high-risk tumor biology.

Design, Setting, and Participants听 A retrospective mediation analysis of factors associated with the racial disparity in breast cancer death for cases diagnosed between 2004 and 2015 with follow-up through 2016 was carried out using the Surveillance, Epidemiology, and End Results (SEER) Oncotype registry. The study included women in the SEER-18 registry who were aged 18 years or older at diagnosis of a first primary invasive breast cancer tumor that was axillary node-negative and ER-positive, who were Black (Black), non-Hispanic White (White), and for whom the 21-gene breast recurrence score was available. Data analysis took place between March 4, 2021, and November 15, 2022.

Exposures听 Census tract socioeconomic disadvantage, insurance status, tumor characteristics including the recurrence score, and treatment variables.

Main Outcomes and Measures听 Death due to breast cancer.

Results听 The analysis with 60鈥�137 women (mean [IQR] age 58.1 [50-66] years) included 5648 (9.4%) Black women and 54鈥�489 (90.6%) White women. With a median (IQR) follow-up time of 56 (32-86) months, the age-adjusted hazard ratio (HR) for breast cancer death among Black compared with White women was 1.82 (95% CI, 1.51-2.20). Neighborhood disadvantage and insurance status together mediated 19% of the disparity (mediated HR, 1.62; 95% CI, 1.31-2.00; P鈥�<鈥�.001) and tumor biological characteristics mediated 20% (mediated HR, 1.56; 95% CI, 1.28-1.90; P鈥�<鈥�.001). A fully adjusted model that included all covariates accounted for 44% of the racial disparity (mediated HR, 1.38; 95% CI, 1.11-1.71; P鈥�<鈥�.001). Neighborhood disadvantage mediated 8% of the racial difference in the probability of a high-risk recurrence score (P鈥�=鈥�.02).

Conclusions and Relevance听 In this study, racial differences in social determinants of health and indicators of aggressive tumor biology including a genomic biomarker were equally associated with the survival disparity in early-stage, ER-positive breast cancer among US women. Future research should examine more comprehensive measures of socioecological disadvantage, molecular mechanisms underlying aggressive tumor biology among Black women, and the role of ancestry-related genetic variants.

Four decades after the racial disparity in breast cancer survival was first documented in the US,¹ non-Hispanic Black (Black) women continue to die at a higher rate following a breast cancer diagnosis compared with non-Hispanic White (White) women.^2-4 The greater incidence of advanced-stage disease among Black women contributes to this ongoing public health concern,^3,4 but the survival disparity exists even among women with early-stage disease.⁵ The largest relative survival gap is between Black and White women with estrogen receptor (ER)-positive, human epidermal growth factor receptor 2 (ERBB2, formerly HER2)-negative tumors (hereafter referred to as ER-positive),^6-10 which account for 60% and 75% of Black and White women with breast cancer, respectively.¹¹ A recent study found that Black women in the US with axillary node-negative, ER-positive tumors have an 80% higher likelihood of breast cancer death compared with their White counterparts.¹²

Racial disparities in breast cancer outcomes are rooted in a wide range of adverse social determinants of health^4,13-15 and a higher rate of biologically aggressive tumors among Black women.^12,16,17 Contextual factors that encompass societal conditions and policy decisions, neighborhood disadvantage, and individual socioeconomic deprivation all contribute to the racial disparity in breast cancer mortality.^8,18-22 Emerging evidence indicates that social determinants can also influence tumor biology.^14,15,23,24

The Oncotype DX 21-gene Breast Recurrence Score (RS) is the most commonly ordered genomic biomarker for breast cancer in the US,²⁵ and guidelines for systemic adjuvant therapy are based on this gene expression profile for patients with ER-positive, axillary node-negative tumors.²⁶ In a previously reported analysis of the the Surveillance, Epidemiology, and End Results (SEER) Oncotype Dx Database,¹² we demonstrated that ER-positive, axillary node-negative breast cancer tumors diagnosed in Black women were 30% more likely to have a high-risk RS compared with tumors from White women. The relative contributions of tumor biology and social disadvantage to the survival disparity in patients with early-stage, ER-positive disease have not been defined. We report the first mediation analysis we are aware of that incorporated tumor genomics to determine the contribution of high-risk tumor biology to the racial disparity in breast cancer death among US women with ER-positive, axillary node-negative breast cancer. Race is conceptualized as a social construct for this analysis.^27,28

This study followed the Strengthening the Reporting of Observational Studies in Epidemiology () reporting guidelines. The SEER Oncotype Dx Database provided data for this retrospective cohort study. Variables generated from the 21-gene Breast RS provided by Genomic Health are linked to invasive breast cancer cases in the SEER-18 registry diagnosed between 2004 and 2015,^29,30 with follow-up for survival through 2016. Women included in this analysis had no prior cancer diagnosis and were aged 18 years or older when diagnosed with a first primary breast cancer that was categorized in the SEER record as stage 1-2 according to the American Joint Committee on Cancer, 6th edition, axillary node-negative, ER-positive; and were categorized as non-Hispanic Black (Black) or non-Hispanic White (White) according to SEER recoding of race and Hispanic ethnicity. Vital status and cause of death were determined from cancer registry follow-up information. The RS variables included the continuous measure (0-100 scale), and month and year of testing. The RS was categorized as low risk (0-10), intermediate risk (11-25), or high risk (26-100).^31-33 Information on age at diagnosis (18-39, 40-49, 50-59, 60-69, and 鈮�70 years), year of diagnosis (2004-2007, 2008-2011, and 2012-2015) tumor size (鈮�2 cm, 2.1-5.0 cm, >5 cm), tumor grade (1, 2, 3, and unknown), progesterone receptor (PR) status (positive/borderline, negative, unknown), receipt of surgery (yes, no), and receipt of radiation therapy and chemotherapy (yes, no/unknown) as part of the first course of therapy were collected from SEER records. Type of health insurance was used as a proxy for individual socioeconomic status. Insurance information was collected in the SEER registry beginning in 2007, and was categorized per SEER recoding as: uninsured, Medicaid, insured, insured not specified, and unknown. The categories of insured and insured not specified were combined for all analyses since both groups include Medicare beneficiaries who are not Medicaid-eligible. Medicare enrollees who are Medicaid-eligible are recoded as Medicaid in SEER. Three census tract variables were collected from SEER records as indicators of structural disadvantage: median household income (continuous), percent of households below 150% of the federal poverty line (continuous), and an education index (continuous weighted score based on percentages of residents who attained less than high school, high school only, and more than high school). Patients with no information on RS or follow-up for survival were excluded from the analysis. Information on ERBB2 status was only available for cases diagnosed in years 2010 and later. All cases diagnosed between 2010 and 2015 included in the study were ERBB2 negative (69% of the analytic cohort); ERBB2-positive/borderline tumors were excluded.

The outcome of interest was death due to breast cancer identified using International Classification of Diseases, Tenth Revision (ICD-10) codes, autopsy, or death certification. Cause of death was determined based on SEER records. Women were followed from the month of breast cancer diagnosis until death or end of the study period.

Discrete-time survival analysis models estimated associations of demographic, tumor, and treatment characteristics with the hazard ratio (HR) of breast cancer death for Black compared with White women (the disparity HR). Discrete-time models are more appropriate for mediation analyses compared with Cox models because the proportional hazards assumption can never be satisfied for nested models both with and without the mediator.³⁴ We divided time at risk into 90-day intervals and modeled time to breast cancer death or censoring using complementary log-log regression, which produces estimates that can be interpreted as HR.³⁵ Individual discrete-time survival models for each predictor variable were estimated while controlling for age at diagnosis.

Mediation analyses examined potential explanations for the racial disparity in breast cancer death. To model census tract socioeconomic indicators as a mediator, we first conducted principal components analysis of the 3 census tract variables (poverty, income, and education). We extracted the predicted score for the first component, which accounted for 74% of the variance across these variables, and used it as a census tract socioeconomic index (SEI) mediator variable in the mediation analyses. We compared the disparity HR before (baseline HR defined as HR_B) and after (mediated HR defined as HR_M) controlling for each potential mediator. Both sets of models were adjusted for age. The mediation proportion was calculated as the proportionate reduction in the disparity HR as ([HR_B鈭�1闭鈭抂贬搁_M鈭�1]) / (HR_B鈭�1). The method of rescaled coefficients was used as a second approach because it accounts for the nonconstant variance between the unmediated and mediated models.³⁶ This method rescales the model variance in the reduced model (without the mediator) to equal the variance in the full model (with the mediator) and calculates a mediation proportion and P value for the corresponding difference in the disparity coefficient before and after controlling for the mediator.³⁷ We examined the proportionate reduction in the racial disparity HR when adjusting for the following domains: census tract SEI; health insurance type; 21-gene RS; tumor histopathologic characteristics (PR status and tumor grade modeled together); tumor size, and treatment (all combinations of surgery type, and initiation of radiation and chemotherapy). Cases diagnosed prior to 2007 (6% of the analytic cohort) were excluded from all analyses that included insurance type as a mediator variable in the model because that data element was not collected in SEER until 2007. Cases missing data for other predictor variables were excluded from analyses of the variable that was missing data, but were included in analyses of nonmissing variables. We examined the role of census tract SEI in mediating racial differences in the probability of a high-risk RS with logistic regression while adjusting for age. Model-based mediation proportion was then calculated as described.

All relevant tests were 2-sided, and associations were considered statistically significant with 伪鈮�.05. All analyses were performed in SAS (version 9.4; SAS Institute,) and Stata statistical software (version 17,). The study was approved by the institutional review board (IRB) at the University of Illinois at Chicago (IRB Protocol #2019-0170). The analysis was conducted between March 4, 2021, and November 15, 2022.

A total of 60鈥�137 women (mean age [IQR], 58.1 [50-66] years) were available for the analysis (eFigure in Supplement 1), including 54鈥�489 (90.6%) White women and 5648 (9.4%) Black women. White women were more likely than Black women to be in the highest income and education groups and in the lowest poverty group, and to have health insurance (Table 1). Black women were more likely to have tumors with aggressive histopathologic characteristics and a high risk RS, to be diagnosed at stage 2, and to receive chemotherapy (Table1).

Median (IQR) follow-up time was 56 (32-86) months. Black women were slightly less likely to have at least 60 months of follow-up compared with White women (41% vs 47%) and were 80% more likely to die from breast cancer compared with White women (age-adjusted HR, 1.82; 95% CI, 1.51-2.20). The risk of death was greater for women with Medicaid compared with insured women, and for those residing in census tracts with lower income and education, and higher rates of poverty (Table 2). Women with a high-risk RS were nearly 6 times as likely to die as a result of breast cancer compared with those with a low RS (HR, 5.95; 95% CI, 4.66-7.61). Histopathologic characteristics associated with aggressive biology and larger tumor size were each associated with greater risk of breast cancer death. In a fully adjusted model that adjusted for all domains (Table 3), Black women were nearly 40% more likely than White women to die as a result of breast cancer (HR, 1.38; 95% CI, 1.11-1.71). In analyses stratified by SEI (Table 4), Black women in each census tract SEI category were more likely than White women to die from breast cancer (age-adjusted HR, 1.50; 95% CI, 1.10-2.05 in the most disadvantaged tract; HR, 1.88; 95% CI, 1.33-2.65 in the middling tract; and HR, 1.64; 95% CI, 0.98-2.73 in the most advantaged tract).

Table 3 presents the point estimates for the disparity HR from the unmediated and mediated models (before and after controlling for each mediator in an age-adjusted model). The Figure shows the proportionate reduction in the disparity in Black compared with White women with cancer after adjustment for each domain based on the excess HR approach and the rescaled coefficients approach. With rescaled coefficients (the more conservative comparison), census tract SEI and health insurance together accounted for 19% of the disparity. The RS and tumor histopathologic characteristics combined accounted for 20% of the greater risk of death for Black women. Racial differences in tumor size and in treatment individually accounted for 8% and 6% of the disparity. When modeled together, all domains combined accounted for 44% of the disparity in breast cancer survival between Black and White women with ER-positive, axillary node-negative tumors.

Black women were more likely than White women to have a high-risk RS (Table 1). Socioeconomic disadvantage was associated with a greater RS. There was a 1 percentage point increase in the prevalence of high-risk RS comparing values at 2 standard deviations below vs 2 standard deviations above the mean of the SEI (P鈥�<鈥�.001). However, census tract socioeconomic disadvantage accounted for only a small proportion of the difference in prevalence of a high risk RS between Black and White women (8% reduction with rescaled coefficients, P鈥�=鈥�.02).

The findings of this population-based study with nationally representative data demonstrated that racial differences in neighborhood- and individual-level socioeconomic disadvantage and indicators of aggressive tumor biology accounted for equal proportions of the racial disparity in breast cancer death among US women with axillary node-negative, ER-positive breast cancer. This is the first mediation analysis of the racial disparity in breast cancer survival we are aware of to include a tumor genomic variable as a mediator, which provided a more robust measure of tumor biology than mediation analyses that used only tumor histopathologic characteristics.^8,20 We found that including the RS along with standard histopathologic measures increased the proportion of the disparity accounted for by tumor biology.

The main findings of this analysis in terms of the role of social determinants of health are consistent with previous studies that showed both individual- and neighborhood-level measures of socioeconomic deprivation drive racial disparities in breast cancer survival.^8,18-22 Earlier studies that examined the role of tumor biology focused on racial differences in hormone receptor status and tumor grade.^8,20 For example, Warner and colleagues⁸ analyzed patients treated at institutions belonging to the National Comprehensive Cancer Network and found that racial differences in the status of ER and tumor grade together mediated 24% of the overall racial disparity in breast cancer death for all subtypes combined. None of the prior studies analyzed the proportion of the disparity mediated by tumor biologic variables specifically for patients with ER-positive tumors. As a result, the contribution of tumor biology to the survival disparity in those studies was primarily due to racial differences in the prevalence of aggressive ER-negative tumors rather than biological differences within a particular subtype. By focusing our analysis on women with ER-positive tumors and by including tumor genomics as a mediating variable, we were able to more precisely quantify the relative excess of breast cancer death mediated by aggressive tumor biology in Black women with the most common breast cancer subtype.¹¹

In a post hoc analysis of the randomized TAILORx trial³⁸ of patients with ER-positive, axillary node-negative tumors, Sadigh and colleagues²² reported an overall survival disparity between Black and White women with an HR of 1.49 (95% CI, 1.10-2.99) after adjusting for both structural (ie, neighborhood disadvantage) and individual social determinants of health (insurance type), and tumor characteristics (RS, tumor grade, and size). The measures of social disadvantage analyzed and the associations with breast cancer death reported by Sadigh et al are consistent with this report, but those authors did not conduct a formal mediation analysis to determine the relative contribution of each domain to the overall survival disparity. Moreover, the study reported here include an 8-fold larger sample of Black patients (5648 vs 693) and used a data set that was more representative of the US population. The TAILORx³⁸ and cooperative group trials^7,39 support our finding of a role for tumor biology in the racial disparity in ER-positive breast cancer survival. Those prospective clinical trials, with uniform stage and treatment for all study participants, reported that the survival disparity could not be explained by racial differences in treatment. This finding suggests that disproportionately aggressive tumor biology contributes to worse outcomes for Black women.

The finding of a racial disparity even among women residing in the most advantaged neighborhoods suggests that adverse social conditions that accrue to Black women regardless of socioeconomic status, such as racial animus, experiences of discrimination and marginalization, may act as upstream factors driving inequities in cancer care. Moreover, recent conceptual models of health disparities recognize that lifelong exposure to discrimination and hostility can influence tumor biology in Black women at the cellular and molecular level.^13-15 Although neighborhood-level deprivation accounted for only a small component of the racial difference in the RS in this study, it is possible that socio-ecological factors would mediate a larger proportion if more comprehensive measures of social determinants over the life course were available for the analysis. Furthermore, it is possible that tumor biological mechanisms not reflected in the RS^40,41 could contribute to the survival disparity. Unfortunatley, we were unable to examine whether ancestry-related genetic variants underlie racial differences in tumor biology.¹⁵

Strengths of this study include the size and population-based sampling of the nationally representative data set, analysis of tumor genomics as a measure of tumor biology, inclusion of both structural and individual-level measures of disadvantage, and breast cancer鈥搒pecific death as the outcome. However, this study has limitations. The SEER registry does not contain data on body mass index (BMI), which can confound survival analyses comparing Black and White women with breast cancer.^42,43 The SEER registry does not collect data on delays in treatment or dose reductions for chemotherapy and radiotherapy, and racial differences in those treatment parameters may mediate some of the survival disparity.⁴⁴ Furthermore, SEER combines the 鈥渘o鈥� and 鈥渦nknown鈥� responses for those variables, which can lead to under-reporting of chemotherapy use.⁴⁵ The SEER registry does not include variables on use of adjuvant endocrine therapies. Although nonadherence or early discontinuation may account for some of the disparity, randomized clinical trials reported that racial differences in the use of adjuvant endocrine therapy do not fully explain the higher rate of death in Black women with ER-positive breast cancer.^7,22 The SEER registry does not include data on ERBB2 status for cases diagnosed prior to 2010. Based on the number of ERBB2-positivite/ER-positive cases diagnosed from 2010 to 2015 with oncotype scores available in SEER (excluded from this analysis), we estimate that the rate of contamination of the analytic cohort by ERBB2-positivite tumors is less than 2%. Median follow-up time of 56 months is relatively short for early stage, ER-positive breast cancer. However, the racial disparity in survival from this subtype of breast cancer emerges within the first few years after diagnosis,⁴⁶ so the estimates reported are likely to remain fairly stable with longer follow-up. Data on type of health insurance is not available for cases diagnosed prior to 2007 (6% of the analytic cohort), necessitating removal of those cases from mediation analyses that included this variable. Finally, RS data are only available for cases whose treating oncologist ordered the test. We cannot exclude the possibility that this introduced selection bias in the study cohort, which could bias the results.

This study of Black and White women in the US demonstrated that structural and individual-level disadvantage and tumor biology mediated the same proportion of the racial disparity in survival following a diagnosis of ER-positive, axillary node-negative breast cancer. Eliminating this survival gap will require a better understanding of the multilevel effects of social disadvantage on cancer care, a clearer picture of the biological mechanisms underlying the aggressive tumor phenotype that is more prevalent in Black women, and insight into the complex relationship between adverse social conditions, ancestry-related genetic variants, and tumor biology.

Accepted for Publication: November 29, 2022.

Published Online: February 16, 2023. doi:10.1001/jamaoncol.2022.7705

Corresponding Author: Kent F. Hoskins, MD, Division of Hematology and Oncology, University of Illinois at Chicago, 840 S Wood St (MC 713), Chicago, IL 60612 (khoski@uic.edu).

Author Contributions: Dr Rauscher had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Hoskins, Calip, Danciu.

Acquisition, analysis, or interpretation of data: Hoskins, Calip, Huang, Ibraheem, Rauscher.

Drafting of the manuscript: Hoskins, Huang.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Calip, Huang, Rauscher.

Administrative, technical, or material support: Calip, Danciu.

Supervision: Hoskins.

Conflict of Interest Disclosures: Dr Hoskins reported grants from Pfizer, support to institution for clinical trial from Merck, Novartis, Abbvie, and Genetech, and nonfinancial support from Agendia data access, analysis, and technical writing assistance outside the submitted work. Dr Calip reported personal fees from Flatiron Health employment, ownership of Roche stock, and grants from Pfizer outside the submitted work. Dr Ibraheem reported support to insitution for clinical trials from Lilly and Gilead, and reported grants from Bristol Myers Squibb outside the submitted work. Dr Danciu reported support to institution for clinical trials from Pfizer, Novartis, Sanofi, and Seagen, and is an advisory board member for Biotheranostics. No other disclosures were reported.

Funding/Support: Gregory S. Calip was supported by the National Cancer Institute (U54CA202995, U54CA202997, and U54CA203000). Garth H. Rauscher was supported by the National Cancer Institute (P01CA154292).

Role of the Funder/Sponsor: The funding agencies 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.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data Sharing Statement: See Supplement 2.