ASPO Abstracts
Identifying Phenotypes Associated with Advanced Disease Presentation in Breast Cancer
Category: Cancer Health Disparities
Conference Year: 2020
Abstract Body:
Purpose
Describe the combinations of individual and community factors (“phenotypes”) that are associated with
advanced disease presentation in breast cancer.
Methods
We derived phenotypes associated with advanced disease (regional or distant stage disease) in breast cancer
patients using K-medoid clustering, an unsupervised machine learning technique. The phenotypes were
derived from 6,132 adult, female patients who were diagnosed with advanced breast cancer in 2016 and lived
in Ohio. The K-medoid algorithm was applied to 7 compositional variables (age, race, ethnicity, marital status,
insurance status, receptor positivity) and 12 contextual variables (% households without access to a car, %
without high school degree, median rent, % homeowners, % crowded households, % below poverty, % white
collar workers, % Black, % under 18, food desert status, health professional shortage area, rurality).
Results
We identified 8 phenotypes associated with advanced disease presentation (named A, B, C, D, E, F, G, H).
The most common A phenotype (n = 1,168, 19.0%) was comprised of individuals who were more likely to be
White and younger and live in communities that are more educated, have lower poverty rates, and have more
white-collar workers. Two noteworthy phenotypes were B and C. The B phenotype (n = 569, 9.3%) was
comprised of individuals who are more likely to be younger and live in rural communities. The C phenotype (n
= 528, 8.6%) was comprised of individuals who are more likely to be White and live in food desert communities
with fewer white-collar workers, homeowners, and households with access to a car.
Conclusions
Risk factors associated with advanced disease presentation are multifactorial, occurring at the individual and
community levels. Therefore, we identified phenotypes associated with advanced breast cancer, an approach
that differs from traditional cancer disparities studies that often utilize parametric, regression-based
approaches that identify independent risk factors associated with advanced disease. Characterizing multilevel
phenotypes associated with advanced disease presentation could inform disparities elimination efforts.
Keywords: Breast neoplasms, stage at presentation