Enhancing implementation and accuracy of ESC cardio-oncology guidelines: first insights from the ML-CardioTox trial using an automated consultation tool
D Legallois, C Dolladille, A F Plane, T Goncalves, T Pezel, A Da Silva, J AlexandreAbstract
Background/Introduction
The European Society of Cardiology (ESC) cardio-oncology guidelines were published in 2022 and represent a cornerstone document aimed at harmonising and optimising cardiovascular care in patients with cancer (1). These recommendations provide a detailed framework for baseline cardiovascular risk stratification, particularly for therapies associated with cardiotoxicity such as anthracyclines. However, despite their comprehensive nature, real-time assessment of cardiotoxicity risk during routine clinical consultations may be perceived as complex and challenging to implement in daily practice with significant heterogeneity across Europe.
Purpose
Our first aim is to facilitate the implementation of ESC cardio-oncology guidelines in clinical practice using an automated consultation tool. Our second aim is to use this tool as a platform to conduct the ML-CardioTox trial, designed to collect data from 600 patients followed in cardio-oncology outpatient clinics, treated with anthracyclines or anti-HER2 therapies, and managed in accordance with ESC recommendations. Baseline and follow-up data will be analysed using machine-learning approaches with the long-term objective of refining the HFA-ICOS risk score. Data collection is performed using a digital solution that enables automated cardiovascular risk stratification. The aim of this ancillary analysis is to evaluate the ability of the software to accurately stratify the risk of cardiotoxicity in individual patients.
Methods
The study started in October 2025 and is currently ongoing as a multicentre study. At the time of abstract submission, 23 patients had been enrolled. For each patient, baseline cardiotoxicity risk stratification was manually calculated using the HFA-ICOS score as implemented in the ESC Pocket Guidelines application or ESC guidelines and compared with the automated risk assessment generated by the software when proposing a surveillance strategy in accordance with these recommendations.
Results
Abstract Picure 1 summarises the clinical and paraclinical characteristics of the 20 patients relevant for HFA-ICOS score calculation and treated with anthracyclines. The figure also presents cardiotoxicity risk categories automatically generated by the digital solution (8 patients classified as low risk, 4 as moderate risk, and 8 as high risk). Manual calculation of baseline cardiotoxicity risk was strictly identical to the automated assessment, with a 100% concordance between the two approaches (Abstract Picture 2).
Conclusion(s)
In this preliminary cohort of patients, automated cardiotoxicity risk stratification performed by the software, based on the HFA-ICOS score, demonstrated reliable agreement with manual assessment. Beyond risk stratification, the automated consultation tool may serve as a platform for the creation of large cardio-oncology databases across Europe and contribute to reducing heterogeneity in the cardiovascular care of patients with cancer.HFA-ICOS scores of included patients Comparison of risk stratification