Introduction: Non-Hispanic Black (African American, AA) women have the worst breast cancer outcomes of any racial/ethnic group. In comparison to non-Hispanic white (white) women, disparities in breast cancer survival for AA women have persisted over the past several decades, although there has been some recent improvement both at the national level and for AA women in Florida. However, we recently reported that AA women in Florida still have twice the rate of breast cancer death compared to their white counterparts. Therefore, more targeted approaches are needed to eliminate survival disparities for AA women diagnosed with breast cancer. The purpose of this study was to conduct a multiple mediation analysis to identify the most important factors causing the disparity in breast cancer survival for AA women. Identifying the most important mediators will provide opportunities for targeted policies, programs, and interventions to ultimately eliminate the prognostic effect of race/ethnicity for Florida women diagnosed with breast cancer. Methods: The study population is comprised of women in Florida diagnosed with breast cancer between 2004 and 2015 who were reported to the Florida Cancer Data System. The following putative mediators were analyzed in this study: residence in high poverty neighborhoods, having Medicaid or no insurance, advanced disease stage at diagnosis, high tumor grade, negative hormone receptor status, and receipt of surgery, based on our causal diagram. The outcome of interest was 5-year breast cancer-specific death. Sociodemographic and clinical characteristics were compared between African American and white women. The Cox model was used to obtain hazard ratios (HR) with 95% confidence intervals (CI) for the direct effect of AA race/ethnicity and indirect effects via mediators. Age was treated as a confounder in the model and mediators were treated as independent. The relative effect for African American ethnicity and each mediator, which is the percentage of the total effect of AA race/ethnicity explained (directly and indirectly) on breast cancer survival, are also provided. Results: African American women were more likely to be younger at diagnosis, living in high poverty neighborhoods, to be insured via Medicaid or uninsured, to have more advanced disease (stage, grade) with hormone receptor negative tumors, and less likely to receive surgery as treatment. From our mediation model, after accounting for all potential mediators, the direct effect associated with African American ethnicity was a 17% (HR = 1.17: 95% CI, 1.11-1.23) increased rate of breast cancer death, which represented 3.9% of the total racial/ethnic effect. The indirect and relative effects of mediators are provided in Table 1.

Conclusions: This analysis identified receipt of surgery, advanced disease stage, hormone receptor negative disease, insurance status, and high tumor grade as the most important mediators of the racial/ethnic disparity in breast cancer survival for African American women in Florida. These results have ramifications for increasing access to screening services, providing options for health insurance, and ensuring prompt receipt of curative surgery to eliminate the prognostic effect of race/ethnicity for African American women diagnosed with breast cancer in the Florida. Study limitations are that these results are hypothesis generating based on the current status of methodology to examine multiple mediators in survival models.

Table 1.

Direct and Indirect Effects for the Association of African American Race/Ethnicity with 5-year Breast Cancer-Specific Survival

Direct & indirect effectsCox Proportional Hazard Model
HR95%. CIrel eff
AA race/ethnicity (Direct Effect) 1.17 1.11, 1.23 3.9% 
Surgery 3.18 2.93, 3.47 28.3% 
Advanced stage 3.09 2.84, 3.36 27.5% 
Hormone receptor negative 1.58 1.51, 1.65 11.1% 
Medicaid-Uninsured 1.48 1.40, 1.57 9.6% 
High grade 1.47 1.42, 1.53 9.4% 
High poverty 1.34 1.23, 1.47 7.2% 
Hormone treatment 1.11 1.09, 1.14 2.5% 
Immunotherapy 1.08 1.02, 1.15 1.9% 
Radiation 1.04 1.03, 1.05 1.0% 
Chemotherapy 0.91 0.89, 0.92  
Total Effect 60.10 50.8, 71.1 100.0% 
Abbreviations: AA, African American; HR, hazard ratio; CI, confidence interval.  
Direct & indirect effectsCox Proportional Hazard Model
HR95%. CIrel eff
AA race/ethnicity (Direct Effect) 1.17 1.11, 1.23 3.9% 
Surgery 3.18 2.93, 3.47 28.3% 
Advanced stage 3.09 2.84, 3.36 27.5% 
Hormone receptor negative 1.58 1.51, 1.65 11.1% 
Medicaid-Uninsured 1.48 1.40, 1.57 9.6% 
High grade 1.47 1.42, 1.53 9.4% 
High poverty 1.34 1.23, 1.47 7.2% 
Hormone treatment 1.11 1.09, 1.14 2.5% 
Immunotherapy 1.08 1.02, 1.15 1.9% 
Radiation 1.04 1.03, 1.05 1.0% 
Chemotherapy 0.91 0.89, 0.92  
Total Effect 60.10 50.8, 71.1 100.0% 
Abbreviations: AA, African American; HR, hazard ratio; CI, confidence interval.  

Citation Format: Robert B. Hines, Xiang Zhu, Eunkyung Lee, Carolyn Rapp, Albert Volk, Asal M. Johnson. A mediation analysis of racial disparity in breast cancer survival [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P3-14-14.