What accounts for the racial disparity in survival from estrogen receptor-positive, axillary node-negative breast cancer in the United States? An analysis of the SEER-Oncotype database.

Abstract

6534 Background: In a previous analysis of the SEER-Oncotype database (Hoskins, et al. JAMA Oncol, 2021), we reported that among women with estrogen receptor (ER)-positive, axillary node-negative breast cancer (BC), Black women were more likely than non-Hispanic White (White) women to have a high risk Oncotype Recurrence Score (RS), and the adjusted hazard for BC-specific death was 1.5-2.5 times higher for Black than for White women within each RS risk category. In this study we examined the role of health insurance, census tract socioeconomic status and tumor biology in the racial disparity in ER-positive BC mortality among US women. Methods: We obtained BC-specific survival data from the SEER-Oncotype database for women diagnosed with first primary, stages I-II, ER-positive, node-negative BC between 1/1/2004 -12/31/2015 who had an Oncotype Recurrence Score (RS) through Genomic Health Laboratory. The racial (Black:White) BC mortality disparity was estimated as a disparity hazard ratio (HR) from a series of Cox proportional hazards models of time to BC death. The baseline model adjusted for SEER registry and age at diagnosis (included in all models). We estimated the disparity HR after controlling for variable domains one at a time, as well as cumulatively in this order: neighborhood SES index (a composite measure including census tract education, income and poverty) and insurance (none, public, private); tumor biology (RS, PR status, tumor grade); tumor size; and treatment (surgery type, initiation of radiation and chemotherapy). Results: The analysis included 57,428 White and 6,003 Black women with node-negative, ER-positive BC (median follow-up = 54 months). The total disparity HR for BC death was 1.67 (95% CI: 1.37, 2.02). Tract SES and insurance together accounted for 21% of the disparity HR (adjusted HR 1.52, 95% CI: 1.22, 1.88); most of this was due to differences in insurance status, with tract SES accounting for 7% of the disparity when considered in isolation. Tumor biology (RS, PR status and grade) accounted for 30% of the disparity (adjusted HR = 1.41, 95% CI:1.14, 1.75); together, the domains of social determinants and tumor biology accounted for 50% of the disparity HR (adjusted HR = 1.30, 95% CI: 1.04, 1.62). Tumor size and treatment initiation each explained roughly 10% of the disparity HR when considered in isolation but did not account for any of the disparity once other factors were accounted for. Results obtained from additional methods for mediation analyses, including a method of rescaled coefficients and structural equation modeling (SEM) applied to discrete-time survival analysis will also be presented. Conclusions: In this study of node-negative, ER-positive BC, much of the BC survival disparity among US women could be explained by racial differences in measured tumor biology and social determinants of health.