Periodic Reporting for period 2 - iCARE4CVD (INDIVIDUALISED CARE FROM EARLY RISK OF CARDIOVASCULAR DISEASE TO ESTABLISHED HEART FAILURE)
Période du rapport: 2024-10-01 au 2025-09-30
Cardiovascular diseases remain the leading cause of morbidity, mortality and healthcare costs in Europe. Rising patient complexity, multimorbidity and variation in care require more personalised and efficient models. iCARE4CVD aims to improve cardiovascular disease management across the full continuum, from early risk to advanced disease, by strengthening precision medicine, data driven decision making and implementation pathways across Europe.
The project addresses three linked challenges: improving early diagnosis, advancing prediction models that identify individuals who benefit most from interventions and enabling personalised treatment strategies that maximise benefit and minimise harm. These aims are supported by advanced artificial intelligence integrated with a large federated and privacy preserving data infrastructure.
Objectives
1. Build large cohorts covering the full cardiovascular disease spectrum.
2. Create a federated and secure database for patient level data.
3. Identify and characterise known and novel biomarkers.
4. Engage patients and stakeholders, define PROMs and PREMs and understand preferences and adherence.
5. Improve diagnostic, prediction and therapeutic models using advanced artificial intelligence.
6. Prospectively validate new patient centred prediction models.
7. Establish cost effective and scalable patient oriented care pathways.
8. Support implementation through legal, regulatory and policy frameworks.
The project addresses three linked challenges: improving early diagnosis, advancing prediction models that identify individuals who benefit most from interventions and enabling personalised treatment strategies that maximise benefit and minimise harm. These aims are supported by advanced artificial intelligence integrated with a large federated and privacy preserving data infrastructure.
Objectives
1. Build large cohorts covering the full cardiovascular disease spectrum.
2. Create a federated and secure database for patient level data.
3. Identify and characterise known and novel biomarkers.
4. Engage patients and stakeholders, define PROMs and PREMs and understand preferences and adherence.
5. Improve diagnostic, prediction and therapeutic models using advanced artificial intelligence.
6. Prospectively validate new patient centred prediction models.
7. Establish cost effective and scalable patient oriented care pathways.
8. Support implementation through legal, regulatory and policy frameworks.
• During the second reporting period, iCARE4CVD progressed from foundational work to integrated implementation. Efforts focused on strengthening the federated infrastructure, advancing artificial intelligence based modelling and expanding engagement with patients and clinicians.
• The federated architecture is now operational for metadata exploration and secure multi cohort analysis. Fifteen metadata dictionaries covering more than one hundred fifty thousand individuals were mapped and validated, with additional cohorts undergoing harmonisation. The Cohort Explorer gained multi cohort comparison tools, ontology based knowledge graphs and improved query functions. Integration with the data clean room and the blockchain governance prototype also advanced.
• Biomarker research progressed with completion of an artificial intelligence supported pipeline for scalable evidence extraction. The first systematic review on interleukin six was completed, and digital biomarker reviews for heart failure and coronary artery disease were initiated with harmonised search methods.
• Patient participation increased through three national panels, early outputs and a presentation at a scientific congress. The PROMs and PREMs scoping review was submitted, and the INSPIRE eHealth preference survey recruited more than seven hundred individuals, informing motivational modelling and patient centred tool design.
• In Work Package 5, progress was made in risk prediction, precision medicine modelling and explainable artificial intelligence. Harmonised datasets were processed through an extended readiness pipeline. New survival and causal models were developed to estimate individual treatment effects and identify treatment modifiers, and models predicting hospitalisation and mortality were benchmarked across cohorts. Dashboards and explainable artificial intelligence tools are prepared for prospective validation.
• Trial preparation in Work Package 6 advanced with completion of the electronic case report form, the data flow architecture and the decision engine design. Although model delivery delays shifted timelines, all core documents and laboratory infrastructure are in place.
• Work Packages 7 and 8 progressed health economic frameworks, pathway mapping and regulatory readiness. Patient journeys were adopted for pathway design, resource use metrics were defined for cost effectiveness analysis and the regulatory roadmap aligned with federated analytics and emerging standards for artificial intelligence in medicine.
• The federated architecture is now operational for metadata exploration and secure multi cohort analysis. Fifteen metadata dictionaries covering more than one hundred fifty thousand individuals were mapped and validated, with additional cohorts undergoing harmonisation. The Cohort Explorer gained multi cohort comparison tools, ontology based knowledge graphs and improved query functions. Integration with the data clean room and the blockchain governance prototype also advanced.
• Biomarker research progressed with completion of an artificial intelligence supported pipeline for scalable evidence extraction. The first systematic review on interleukin six was completed, and digital biomarker reviews for heart failure and coronary artery disease were initiated with harmonised search methods.
• Patient participation increased through three national panels, early outputs and a presentation at a scientific congress. The PROMs and PREMs scoping review was submitted, and the INSPIRE eHealth preference survey recruited more than seven hundred individuals, informing motivational modelling and patient centred tool design.
• In Work Package 5, progress was made in risk prediction, precision medicine modelling and explainable artificial intelligence. Harmonised datasets were processed through an extended readiness pipeline. New survival and causal models were developed to estimate individual treatment effects and identify treatment modifiers, and models predicting hospitalisation and mortality were benchmarked across cohorts. Dashboards and explainable artificial intelligence tools are prepared for prospective validation.
• Trial preparation in Work Package 6 advanced with completion of the electronic case report form, the data flow architecture and the decision engine design. Although model delivery delays shifted timelines, all core documents and laboratory infrastructure are in place.
• Work Packages 7 and 8 progressed health economic frameworks, pathway mapping and regulatory readiness. Patient journeys were adopted for pathway design, resource use metrics were defined for cost effectiveness analysis and the regulatory roadmap aligned with federated analytics and emerging standards for artificial intelligence in medicine.
iCARE4CVD delivered advances beyond current practice in cardiovascular research, artificial intelligence enabled healthcare and federated infrastructures.
A major innovation is the operationalisation of a pan European federated ecosystem that uses metadata ontologies, artificial intelligence supported mapping and secure multi party computation. This enables multi cohort patient level analytics while protecting privacy, which is rarely achieved at this scale. The knowledge graph approach and artificial intelligence assisted code mapping support rapid and high quality harmonisation across heterogeneous datasets.
In biomarker science, the project introduced a scalable evidence synthesis pipeline that generates structured and reproducible biomarker evidence. The interleukin six analysis demonstrates its quality and efficiency.
In machine learning, precision treatment modelling advanced through modern causal inference methods applied to heart failure. The integration of targeted maximum likelihood estimation, double machine learning, mediation analysis and survival modelling across harmonised real world cohorts represents a significant methodological step. These models provide patient specific treatment effect estimates and support the development of a clinically deployable decision engine.
Patient involvement through multi country panels, preference studies and motivational research introduced behavioural insights and patient reported outcomes early in model development.
The integration of artificial intelligence models into a federated platform that can support a regulatory ready clinical decision support system sets a new benchmark for translational cardiovascular medicine.
A major innovation is the operationalisation of a pan European federated ecosystem that uses metadata ontologies, artificial intelligence supported mapping and secure multi party computation. This enables multi cohort patient level analytics while protecting privacy, which is rarely achieved at this scale. The knowledge graph approach and artificial intelligence assisted code mapping support rapid and high quality harmonisation across heterogeneous datasets.
In biomarker science, the project introduced a scalable evidence synthesis pipeline that generates structured and reproducible biomarker evidence. The interleukin six analysis demonstrates its quality and efficiency.
In machine learning, precision treatment modelling advanced through modern causal inference methods applied to heart failure. The integration of targeted maximum likelihood estimation, double machine learning, mediation analysis and survival modelling across harmonised real world cohorts represents a significant methodological step. These models provide patient specific treatment effect estimates and support the development of a clinically deployable decision engine.
Patient involvement through multi country panels, preference studies and motivational research introduced behavioural insights and patient reported outcomes early in model development.
The integration of artificial intelligence models into a federated platform that can support a regulatory ready clinical decision support system sets a new benchmark for translational cardiovascular medicine.