Electronic health record–based gut disruption and perfluoroalkyl and polyfluoroalkyl substances drinking-water proxies in postmenopausal hormone receptor–positive breast cancer: A nested case-control study.

Abstract

e24180 Background: Environmental exposures and gut microbiome disruption may influence hormone receptor-positive (HR+) breast cancer (BC), but large-scale studies are limited by lack of exposure biomarkers and stool profiling. To address this gap, we leveraged extensive electronic health record data in the NIH All of Us research program dataset to find proxies for these potentially related biomarkers. Methods: We conducted a matched nested case-control study in the NIH All of Us Controlled Tier using conditional logistic regression. We matched cases and controls on age, race, sex, and observation time. We defined a proxy of postmenopausal status as age ≥ 55 years at the index date. We defined incident HR+ BC as breast cancer with evidence of endocrine therapy. We derived a Perfluoroalkyl and Polyfluoroalkyl Substances (PFAS) drinking-water proxy, cumulative environmental load (CEL), by linking ZIP3 to U.S. EPA UCMR5 PFAS occurrence data published for 2023-2025. We created CEL by detect-frequency-weighting PFAS concentrations, winsorizing, applying log1p transformation, and SD-standardizing. We also evaluated a rank-based CEL in sensitivity analysis. We defined a gut disruption proxy (GDP) as ≥ 30 cumulative antibiotic days in the prior 5 years and/or prior GI dysfunction diagnoses (IBS, IBD, or C. difficile). We adjusted for area-level deprivation and insurance-related covariates as available, and tested a CEL and GDP interaction. Results: We excluded 24 non-informative matched sets and analyzed 13,696 observations across 4,582 strata (4,582 cases; 9,114 controls). GDP was associated with higher odds of HR+ BC (adjusted OR 4.37; 95% CI 3.82-4.99; p < 0.001). CEL was not associated with HR+ BC under the primary specification (OR 1.00 per SD; 95% CI 0.96-1.04; p = 0.92), and GDP did not modify the CEL association (interaction OR 0.97; 95% CI 0.82-1.15; p = 0.75). Rank-based CEL showed a modest association (OR 1.12; 95% CI 1.07-1.17; p < 0.001) without interaction (p = 0.25), suggesting potential sensitivity of PFAS–HR+ BC associations to exposure scaling and distributional assumptions. GDP components were each associated with HR+ BC, including antibiotic burden (OR 1.73 per SD; 95% CI 1.64-1.82; p < 0.001) and GI dysfunction history (OR 4.84; 95% CI 3.82-6.13; p < 0.001). Conclusions: An EHR-derived gut disruption proxy was strongly associated with incident postmenopausal HR+ BC, though this could reflect healthcare utilization bias within the dataset in addition to biological effects. Broadness of the ZIP3-linked UCMR5 PFAS drinking-water proxy likely contributed to no association under the primary log1p SD-standardized specification and no evidence of synergy with gut disruption, although rank-based scaling yielded a modest association.