Impact of food access and poverty on somatic genomic profiles and clinical outcomes in metastatic breast cancer.

Abstract

1015 Background: Patients living in food deserts or experiencing food insecurity have higher cancer mortality rates. While neighborhood poverty correlates with more aggressive tumor biology, including higher OncotypeDX scores and rates of TP53 mutations, the associations among food access, somatic genomic profiles, and clinical outcomes in metastatic breast cancer (MBC) remain understudied. Methods: This multi-institutional retrospective cohort study included patients with MBC and circulating tumor DNA (ctDNA) testing through the Guardant360 assay from Washington University in St. Louis (N = 651), Massachusetts General Hospital (N = 349), Weill Cornell Medicine (N = 187), and Northwestern University (N = 86). Patient census tracts were linked with the USDA Food Access Research Atlas, which provides data on census-level food access throughout the United States. Deidentified data were shared across sites. We categorized patients by: Food access: Low access (LA) if ≥33% of the census tract lived > 1 mile (urban) or > 10 miles (rural) from a food store Income and food access: Low-income and low access (LILA) if the census tract was both LA and poverty rate was ≥20% We assessed descriptive differences and determined associations between LA/LILA and ctDNA profiles using multivariate analysis adjusted for race, stage, and subtype. Results: Of the 851 patients with census tract data, 46.8% lived in LA and 8.0% lived in LILA areas. Although White patients were more likely to live farther from food stores than Black patients (48.9% vs 34.4%, p = 0.003), Black patients were significantly more likely to live in LILA areas (19.2% vs 6.1%, p = < 0.001). On multivariate analysis, patients in the overall cohort living in LA areas were more likely to have RTK/RAS pathway mutations (Odds ratio [OR] 2.05, 95% confidence interval [CI] 1.36-3.09, p = 0.001), confirmed in the hormone receptor positive, HER2 negative (HR+/HER2-) cohort (OR 2.18, 95% CI 1.34-3.55, p = 0.002). Patients in the overall cohort living in LILA areas were more likely to have CCNE1 copy number variants [cnv] (OR 2.81, 95% CI 1.18-6.73, p = 0.02). Patients in LA areas had significantly shorter overall survival (OS) after first ctDNA test (24 months [mos] vs 31 mos, p = 0.01). In HR+/HER2- MBC, Black patients in LA areas had the shortest OS (11 mos) compared to those in high access areas (38 mos) and White patients in either setting (27-33 mos, p = 0.02). Conclusions: Patients with MBC living in food deserts appear to have distinct ctDNA profiles, including higher rates of RTK/RAS pathway mutations, which drive tumor growth, and CCNE1 cnv, associated with poor prognosis. We also found a significant survival disadvantage associated with living in LA areas, particularly among Black patients. These findings demonstrate the necessity of considering the intersection of tumor biology and social determinants of health in understanding and treating MBC.