ORF27-02 - Breast Cancer Survival Association of Tumor-Proximal Adipocyte Size Validated in the Australian Breast Cancer Tissue Bank after 12-Cancer Screening

Background: Obesity is linked to cancer outcomes across 12 cancer types, and adipocytes at invasive fronts have been demonstrated to support cancer growth through providing free fatty acids. Whether tumor-proximal adipocyte morphology provides a broadly generalizable survival marker across cancers remains unclear.

Methods: To quantify adipocyte size as a function of distance to tumor boundaries at scale, we developed AdiFind, a custom-made computer vision pipeline. For every adipocyte, we measured the nearest distance to the tumor boundary and summarized size-distance profiles using clinically interpretable metrics (primary: average adipocyte area within 0–15,000 µm distance from tumor; secondary: proximal-to-distal ratio and size-distance slope). We analyzed all associated digital slides from 12 cancer types from TCGA, the Australian Breast Cancer Tissue Bank (ABCTB) and the Susan G. Komen Tissue Bank. In total, we processed 21,968 slides and measured >75 million adipocytes. Associations with survival were tested using Cox proportional hazards models with cohort-available clinical covariates; in breast cohorts we additionally evaluated effect modification by menopausal status and BMI.

Results: Across 12 cancer types, a significant association between tumor-proximal adipocyte size and survival was observed in breast cancer (N=3105, P=0.00115, HR=1.37), postmenopausal breast cancer (N=2230, P=0.00001, HR =1.56 ) and also in postmenopausal obese patients (N=456,P=0.0151, HR=2.18). Findings replicated in ABCTB (BMI-annotated; n=2,500). In BMI-stratified analyses, prognostic performance was highest in obese, postmenopausal and improved risk discrimination beyond BMI alone.

Conclusion: Tumor-proximal adipocyte size is a breast cancer-specific prognostic marker and is most informative in postmenopausal and postmenopausal obesity. These results support metabolic coupling at the tumor margin and motivate integrating spatial adipocyte phenotyping into endocrine-metabolic risk stratification in breast cancer.