Analysis of Sociodemographic, Clinical, and Genomic Factors Associated With Breast Cancer Mortality in the Linked Surveillance, Epidemiology, and End Results and Medicare Database

IMPORTANCE: Understanding interactions among health service, sociodemographic, clinical, and genomic factors in breast cancer disparities research has been limited by a disconnect between health services and basic biological approaches. OBJECTIVE: To describe the first linkage of Surveillance, Epidemiology, and End Results (SEER)–Medicare data to physical tumor samples and to investigate the interaction among screening detection, socioeconomic status, tumor stage, tumor biology, and breast cancer outcomes within a single context. DESIGN, SETTING, AND PARTICIPANTS: This population-based cohort study used tumor specimen blocks from a subset of women aged 66 to 75 years with newly diagnosed nonmetastatic, estrogen receptor–positive invasive breast cancer from January 1, 1993, to December 31, 2007. Specimens were obtained from the Iowa and Hawaii SEER Residual Tissue Repositories (RTRs) and linked with Medicare claims data and survival assessed through December 31, 2015. Data were analyzed from August 1, 2018, to July 25, 2021. EXPOSURES: Screening- vs symptom-based detection of tumors was assessed using validated claims-based algorithms. Demographic factors and zip code–based educational attainment and poverty socioeconomic characteristics were obtained via SEER. MAIN OUTCOMES AND MEASURES: Molecular subtyping and exploratory genomic analyses were completed using the NanoString Breast Cancer 360 gene expression panel containing the 50-gene signature classifier. Factors associated with overall and breast cancer–specific (BCS) survival were analyzed using Cox proportional hazards regression models combining sociodemographic, clinical, and genomic data. RESULTS: SEER-Medicare data were available for 3522 women (mean [SD] age, 70.9 [2.6] years; 3049 [86.6%] White), of whom 1555 (44.2%) were diagnosed by screening mammogram. In the SEER-Medicare cohort, factors associated with increased BCS mortality included symptomatic detection (hazard ratio [HR], 1.49 [95% CI, 1.16-1.91]), advanced disease stage (HR for stage III, 2.33 [95% CI, 1.41-3.85]), and high-grade disease (HR, 1.85 [95% CI, 1.46-2.34]). The molecular cohort of 130 cases with luminal A/B cancer further revealed increased all-cause mortality associated with genomic upregulation of transforming growth factor β activation and p53 dysregulation (eg, p53 dysregulation: HR, 2.15 [95% CI, 1.20-3.86]) and decreased mortality associated with androgen receptor, macrophage, cytotoxicity, and Treg signaling (eg, androgen receptor signaling: HR, 0.23 [95% CI, 0.12-0.45]). Symptomatic detection (HR, 2.49 [95% CI, 1.19-5.20]) and zip codes with low levels of educational attainment (HR, 5.17 [95% CI, 2.12-12.60]) remained associated with mortality after adjusting for all clinical and demographic factors. CONCLUSIONS AND RELEVANCE: Linkage of SEER-Medicare data to physical tumor specimens may elucidate associations among biology, health care access, and disparities in breast cancer outcomes. The findings of this study suggest that screening detection and socioeconomic status are associated with survival in patients with locally advanced, estrogen receptor–positive tumors, even after incorporating clinical and genomic factors.