CARAMEL: Retrospective Study for Personalized Risk Assessment of Cardiovascular Disease in Menopausal and Perimenopausal Women Using Real World Data

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 rece