A novel integrative risk scoring system for predicting atrial fibrillation in breast cancer patients: a multicenter cohort study
D Y Kim, S H Kim, H S Lim, J C YounAbstract
Background
As breast cancer (BC) survival improves, long-term cardiovascular complications such as atrial fibrillation (AF) are increasingly relevant. Conventional AF prediction models rarely incorporate cancer treatment exposures that may modify risk. We developed and internally optimized an integrative AF risk score combining traditional cardiovascular risk factors with BC-specific treatment variables.
Methods
We analyzed a multicenter cohort of 1,233 women with BC. Incident AF was assessed over follow-up. To reduce overfitting and optimize predictor selection, we applied penalized LASSO Cox regression with K-fold cross-validation. Predictors retained in the final multivariable model were converted into an integer-based point score using effect sizes from the Cox model.
Results
Over a median follow-up of 103.4 months, 2.1% of patients developed AF. Variables retained in the final multivariable model were: prior heart failure (HR 18.26, p=0.004; 6 points), age (per 10-year increase above 50 years; HR 2.42, p<0.001; 4 points), hypertension (HR 2.45, p=0.06; 3 points), radical mastectomy (HR 2.27, p=0.120; 1 point), and doxorubicin dose (per 100 mg/m²; HR 1.46, p=0.089; 1 point). Incorporating cancer-related variables (surgical extent and anthracycline exposure) improved discrimination, increasing the C-index to 0.82 (p<0.001) with an absolute gain of 0.038 versus a conventional clinical model (p=0.007). A total score ≥7 identified a high-risk group, with an estimated 28% higher AF risk per additional point.
Conclusion
An integrative risk score incorporating cancer specific risk factors with conventional cardiovascular risk factors improves AF prediction in BC survivors. This tool may support personalized cardio-oncology surveillance strategies and targeted rhythm monitoring in long-term follow-up.