A Simple Clinical Risk Score to Improve Prediction of Anthracycline-Induced Cardiotoxicity

Abstract

Background Anthracycline-induced cardiotoxicity is a major limitation of cancer therapy, highlighting the need for accurate and practical risk stratification. Although current guidelines recommend the Heart Failure Association–International Cardio-Oncology Society (HFA-ICOS) score for baseline risk assessment, its complexity may limit routine clinical use. Objectives This study sought to develop and validate a simplified risk score based on readily available clinical and echocardiographic variables for predicting anthracycline-related cardiotoxicity, and to compare its performance with the HFA-ICOS model. Methods We retrospectively analyzed 2,612 adult cancer patients with baseline left ventricular ejection fraction (LVEF) of ≥50% treated with anthracyclines. Echocardiography was performed at baseline, during therapy, and up to 2 years after treatment. Cardiotoxicity was defined as a decrease in LVEF of >10 percentage points to a value <50% or the development of heart failure. The predictive performance of the HFA-ICOS score was assessed using the area under the receiver operating characteristic curve (AUC). A simplified model was derived using logistic regression with cross-validation and was further validated with machine learning techniques. External validation was conducted in an independent cohort of 819 patients. Results Over the first 2 years, the likelihood of cardiotoxicity was higher in the moderate (HR: 1.44, 95% CI: 1.02-2.06) and high (HR: 1.84, 95%:CI: 1.36-2.50) HFA-ICOS risk categories compared to the low-risk group. The cumulative incidence of cardiotoxicity was 8.3% (95% CI: 7.1%-9.1%) at 12 months, with only 5 additional events observed in patients still at risk through 2 years. The HFA-ICOS score showed good discrimination at 1 year (AUC: 0.76, 95% CI: 0.67-0.86). A simplified model including postchemotherapy LVEF and cumulative anthracycline dose (RE-ACT score) demonstrated higher accuracy at 1 year (AUC: 0.83, 95% CI: 0.74-0.93) though not statistically different. Combining both scores in the derivation cohort further improved predictive performance (AUC: 0.87, 95% CI: 0.80-0.95; P < 0.001 vs HFA-ICOS score). In the validation cohort, the RE-ACT score achieved an AUC of 0.88 (95% CI: 0.80-0.95) at 1 year. Conclusions HFA-ICOS score effectively stratifies baseline risk, while the RE-ACT score provides additional post-treatment prognostic value. Their integration supports a practical 2-step strategy for personalized surveillance in cardio-oncology.