Cardiovascular Risk Factors, Menopausal Women, Perimenopausal Women, Real World Data
Conditions
Keywords
Cardiovascular Disease, Cardiovascular risk factors, Menopausal women, Perimenopausal women, real world data, personalised prevention, computational modelling, women-specific risks
Brief summary
This retrospective observational study, part of the EU-funded CARAMEL project, aims to develop and validate personalized cardiovascular disease (CVD) risk assessment models specifically designed for menopausal and perimenopausal women (ages 40-60). The study leverages Real World Data (RWD) collected from multiple international clinical partners, including electronic health records (EHR), diagnostic imaging data, and signal data. The main objective is to improve the prediction of CVD precursors such as hypertension and dyslipidemia, as well as mid- and long-term risk of CVD events, through advanced artificial intelligence (AI) models. These models will be trained on multimodal data to capture complex, individualized risk trajectories that current risk calculators fail to address, particularly in women. Special focus is placed on under-researched, women-specific risk factors and their interactions with traditional predictors. The study includes several research objectives: (1) predicting the onset of hypertension and dyslipidemia using EHR data; (2) modeling the long-term risk of fatal and non-fatal cardiovascular events and disease trajectories; (3) identifying novel imaging biomarkers from routine screening tests such as mammography, DXA, ultrasound, and cardiac MRI; (4) developing multimodal prediction models combining imaging and clinical data; (5) creating automated AI tools for imaging biomarker extraction; and (6) using signal data from cardiac devices to predict disease progression and events. The study population consists of middle-aged women with retrospective data available across different health systems. The expected outcome is a validated set of stratified, personalized CVD risk models that can support targeted prevention strategies and enable more equitable, sex-specific care. This will contribute to reducing the burden of CVD in women and addressing critical gaps in early detection, clinical decision-making, and health policy. This project has received funding from the European Union's Horizon Europe Research and Innovation Programme under Grant Agreement No 101156210.
Interventions
None listed
Sponsors
VISUAL INTERACTION & COMMUNICATION TECHNOLOGIES - VICOMTECH
Clinic for Cardiovascular Diseases Magdalena
Biokeralty Research Institute
Keralty SAS. Colombia
ETHNIKO KAI KAPODISTRIAKO PANEPISTIMIO ATHINON
Fundación Pública Andaluza para la gestión de la Investigación en Sevilla
University of Dublin, Trinity College
TREE Technology S.A.
Dublin City University
Tampere University
Ben-Gurion University of the Negev
Biogipuzkoa Health Research Institute
Vilnius University Hospital Santaros Klinikos
Hospital Universitario Virgen Macarena
Study design
Observational model
OTHER
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
FEMALE
Age
40 Years to 60 Years
Healthy volunteers
No
Inclusion criteria
Self-identified as female in the electronic health record (EHR). Age between 40 and 60 years at the time of data collection/index date. Availability of at least 5-6 years of retrospective data in the EHR, depending on the research objective. At least one healthcare encounter (visit, imaging, lab test, diagnosis, etc.) within the defined age range. For imaging substudies (e.g., RO3-RO5): availability of at least one relevant imaging test (e.g., DXA, digital mammography, cMRI, CCTA, US) during the age range. For signal-based analysis (RO6): presence of ECG monitoring data from implanted devices and at least 2 years of follow-up.
Exclusion criteria
Prior diagnosis of cardiovascular disease before the observation window (only applicable to specific ROs, e.g., RO2, RO4). Insufficient data quality or missing key variables needed for modeling (e.g., absence of blood pressure or lipid profile). Patients with incomplete or inconsistent records (e.g., duplicate IDs, mismatched time frames). For signal-based RO6: hospitalizations or diagnoses unrelated to cardiovascular health that may bias AI model training.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Occurrence and Predicted Risk of Cardiovascular Disease (CVD) Events (fatal and non-fatal) | up to 10 years | The study will retrospectively evaluate the occurrence of cardiovascular disease (CVD) events and develop predictive models to estimate individual risk profiles for such events. CVD events include both fatal and non-fatal occurrences such as myocardial infarction, stroke, heart failure, arrhythmias, and atherosclerotic disease. Events will be identified using structured electronic health records (EHR) and coded using ICD-10 classifications. Risk will be modeled using multimodal data sources (EHR, imaging, and signals) to predict short- and long-term outcomes, stratified by individual characteristics. The outcome integrates: Event-based measures: Time to first fatal or non-fatal CVD event. Risk-based measures: Individual predicted probabilities of experiencing a CVD event or precursor condition (e.g., hypertension, dyslipidemia) over different time frames. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| RO2. Personalized Risk Prediction of CVD Events and CVD trajectories | Up to 16 years | The occurrence of CVD events, which will be classified in fatal (if they are registered as the cause of death) or not fatal (if they are not registered as cause of death). * Fatal CVD events include the following ICD codes registered in the EHR: * I10-16 Hypertensive disease * I20-25 Ischemic heart disease * I46-52 Arrhythmias and heart failure, excluding I51.4 (Myocarditis unspecified) * I60-69 Cerebrovascular diseases * I70-73 Atherosclerosis/AAA * Not fatal CVD events include only the following ICD codes: * I21-I23 Not fatal myocardial infarction * I60-69 Not-fatal stroke |
| RO3. Novel Imaging Biomarkers and Patterns for CVD Risk Assessment | Baseline | Evaluates the predictive performance of multimodal models combining imaging features (e.g., cardiac MRI, DXA, digital mammography) and electronic health record (EHR) variables to estimate the mid- and long-term risk of cardiovascular events (CVD) in women aged 40-60. The endpoint is the first occurrence of a fatal or non-fatal CVD event after the imaging test, as documented in the EHR. The models will be compared against standard risk assessment tools (e.g., SCORE2). |
| RO1. Personalized risk prediction of CVD precursors | up to 8 years | First observation of HT or DY registered in the EHR, registered as a diagnostic code, or as a laboratory result or test. These include: * Diagnosis of HT registered in the EHR with either of the following ICD10 codes: * I10 Essential (primary) hypertension * I11.0 Hypertensive heart disease with heart failure * I11.9 Hypertensive heart disease without heart failure * I12.0 Hypertensive chronic kidney disease with stage 5 chronic kidney disease or end stage renal disease * I13.0 Hypertensive heart and chronic kidney disease with heart failure and stage 1 through stage 4 chronic kidney disease, or unspecified chronic kidney disease * I13.1 Hypertensive heart and chronic kidney disease without heart failure * I13.2 Hypertensive heart and chronic kidney disease with heart failure and with stage 5 chronic kidney disease, or end stage renal disease * Diagnosis of DY registered in the EHR with either of the following ICD10 codes: * E78.1 Pure hyperglyceridemia * E78.2 Mixed hyperlipidemia |
| RO5. Automatic imaging marker and pattern extraction | Baseline | The performance and clinical relevance of AI-based tools for the automatic extraction of cardiovascular imaging biomarkers in women aged 40-60. These tools will be used to segment anatomical regions and calculate quantitative measures from multimodal imaging (e.g., ultrasound, DXA, cardiac CT, cMRI, mammography). |
| RO6. Signal-based CVD prediction models | Up to 16 years | Occurrance of CVD events, which include: * The occurrence of Arrhythmia episodes including atrial fibrillation, ventricular tachycardia, and bradyarrhythmias. * The occurrence of Heart failure and structural heart disease, particularly severe left ventricular dysfunction and cardiomyopathy. * The occurrence of Ischemic events such as myocardial infarction (MI), coronary artery disease (CAD), and cerebrovascular accidents (CVA). * Device-related events, including the transition from loop recorders to pacemakers or ICDs due to worsening conditions. Unit of Measure: Recorded episodes (frequency/time) or binary outcome (present/absent). |
| RO4. Multimodal EHR and ImageBased CVD Prediction Models | Up to 16 years | The occurrence of CVD events, which will be classified in fatal (if they are registered as the cause of death) or not fatal (if they are not registered as cause of death). * Fatal CVD events include the following ICD codes registered in the EHR: * I10-16 Hypertensive disease * I20-25 Ischemic heart disease * I46-52 Arrhythmias and heart failure, excluding I51.4 (Myocarditis unspecified) * I60-69 Cerebrovascular diseases * I70-73 Atherosclerosis/AAA * Not fatal CVD events include only the following ICD codes: * I21-I23 Not fatal myocardial infarction * I60-69 Not-fatal stroke |
Outcome results
None listed