Authors
Xiaolin Gao, Kexin Chen, Haowen Liang, Andong Huang, Qiexinhao Li, Jianpeng Liang, Qiutong Pan, Ruilang Zhong, Jiayao Zhang, Ruirui Liang, Weipeng Zheng, Kenie Wang, Jun Peng, Tao Yang
Published in
Frontiers in cardiovascular medicine. Volume 13. Pages 1826258. Epub Aug 03, 2026.
Abstract
Cardiovascular disease (CVD) is a major long-term complication in cancer survivors treated with chemotherapy, yet risk varies markedly across individuals. Using the UK Biobank, we studied chemotherapy-treated cancer patients without baseline CVD and identified data-driven cardio-oncology risk phenotypes with unsupervised clustering (UMAP embedding and HDBSCAN). We evaluated associations between phenotypes and incident CVD using Cox proportional hazards models and assessed 5-year risk discrimination with and without phenotype labels. Among 18,693 patients, 9 stable phenotypes were identified, characterized by distinct patterns of cancer treatment exposure, baseline cardiovascular medication use, cardiometabolic burden, renal function, and lifestyle factors. Compared with the reference phenotype, all phenotypes had higher risks of composite CVD, and the highest-risk phenotype demonstrated a hazard ratio of 4.06 (95% CI, 3.44-4.78). Phenotype-defining features were interpreted using SHAP values. In subtype analyses, heterogeneity was most apparent for ischemic heart disease, angina pectoris, and acute myocardial infarction. Adding phenotype labels modestly improved 10-year discrimination, with AUC values increasing from 0.648 to 0.692. These findings suggest that phenotype-based profiling may provide an exploratory framework for risk stratification in cardio-oncology and may help characterize risk heterogeneity across cancer survivors treated with chemotherapy.
PMID:
42609512
Bibliographic data and abstract were imported from PubMed on 18 Aug 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 5
- Comments 0