October 23, 2025JAMA Oncology4 citationsOpen Access
View Full PaperRisk Prediction Model for Development of Heart Failure or Cardiomyopathy After Breast Cancer Treatment
Key Result
A multivariable risk prediction model accurately identified the 10-year risk of heart failure or cardiomyopathy in women with early-stage breast cancer, achieving a time-dependent AUC of 0.79.
Study Design
Type
Cohort (n=26,044)
Structured PICO
Can a multivariable risk prediction model accurately predict the 10-year risk of developing heart failure or cardiomyopathy in women receiving systemic treatment for early-stage breast cancer?
P
Population
26,044 women aged 18 to 79 years with newly diagnosed invasive local or regional breast cancer from 2008 to 2020, median age 61, cared for in Kaiser Permanente Southern California.
I
Intervention
Risk prediction model using multivariable elastic-net Cox proportional hazards models based on clinical variables and breast cancer treatments.
O
Outcome
Incident heart failure or cardiomyopathy (HF/CM) events at 10 years.hard clinical
A newly developed risk prediction model can accurately identify the 10-year risk of heart failure or cardiomyopathy in women treated for early-stage breast cancer, enabling risk-guided cardiac management.
Main Result
Effect estimate: AUC 0.79
Abstract
Importance: Women receiving breast cancer (BC) treatment are at risk for heart failure or cardiomyopathy (HF/CM). No standard method exists to identify women at risk who might benefit from preventive cardiac strategies and surveillance during and after treatment. Objective: To develop and validate a model predicting the 10-year risk of developing HF/CM in women receiving systemic treatment for invasive early-stage BC to inform cardiac risk management. Design, Setting, and Participants: This longitudinal cohort study of patients cared for in Kaiser Permanente Southern California, a vertically integrated health care system. The goal of the study was to predict the 10-year HF/CM risk using multivariable elastic-net Cox proportional hazards models. Women aged 18 to 79 years with newly diagnosed invasive local or regional BC from 2008 to 2020, with a median follow-up of 5.2 years, were included. The cohort was randomly split into derivation (60%) and validation (40%) cohorts. Data were analyzed from April 2024 to May 2025. Exposures: HF/CM risk predictors included age at BC diagnosis, race and ethnicity, area-level socioeconomic status, local and systemic BC treatments, cancer stage, obesity, and history of hypertension, diabetes, hyperlipidemia, smoking, and other CV conditions. Main Outcome and Measures: Incident HF/CM events. Women were categorized into low-, moderate-, and high-risk groups based on the tertiles of the risk score in the derivation cohort. Results: Of 26 044 women included in the total cohort, the median (IQR) age was 61 (52-68) years. The risk model had good calibration and high accuracy in predicting HF/CM risk in the 3 subgroups, with HF/CM risk in the validation cohort matching the estimates from the derivation for identifying women at low risk (1.7%; 95% CI, 1.1%-2.4%) and high risk (19.4%; 95% CI. 17.3%-21.5%) of HF/CM at 10 years. The model's discrimination ability was good, based on the time-dependent area under the curve of 0.79 at 10 years in the validation cohort. Conclusions and Relevance: This risk prediction model among women with early-stage BC was able to prospectively identify those at risk of HF/CM over a 10-year period based on the selected BC treatment and clinical variables available at BC diagnosis. The model can be used to inform risk-guided cardiac management for these women.