Effects of breast cancer risk factors on the tumor microenvironment: a morphological deep-learning analysis of three prospective cohorts

作者：Clara Bodelón, Mohamed Amgad, James M. Hodge, Maha AT Elsebaie, Mariah Landry, Cheng Peng, Rulla M. Tamimi, Peter Kraft, Lauren E. McCullough, Mark E. Sherman, Mia M. Gaudet, Alpa V. Patel, Lee Cooper, Lauren R. Teras · 发表于：npj Breast Cancer · 年份：2026 · DOI：10.1038/s41523-026-00994-1 · 研究领域：AI in cancer detection、Radiomics and Machine Learning in Medical Imaging、Biomarkers in Disease Mechanisms

Emerging evidence suggests a critical role of the tumor microenvironment (TME) in breast cancer (BC) development and outcomes, yet factors that modify the TME are poorly understood. We investigated the relationship between BC etiological factors and the tumor and TME using 110 histological features of the epithelium, stroma, and immune infiltration, computationally quantified in 3724 H&E slides from three prospective cohort studies. Age, race, hormonal, and lifestyle factors were associated with features of the breast TME. Menopausal hormone therapy was associated with epithelial and stromal features found in less aggressive tumors, while higher body mass index (BMI) was associated with two histological features not captured by grade, and both were associated with poor prognosis. These two features mediated the BMI and BC-specific mortality association by 18.1%. Our findings provide novel insights into the role of etiological factors on the TME including modifiable factors that have implications for prevention and outcomes.