This project aims to create and validate women-specific predictive models for Alzheimer’s disease risk by integrating neuroendocrine biomarkers and reproductive history, leveraging large-scale datasets from the U.S. and Latin America (N > 100,000). Using multicohort harmonization, Cox models, and machine learning approaches (e.g., XGBoost with SHAP), the study will perform internal and external validation, estimate population attributable fractions (PAR), and assess preventable fractions (PPF). The main deliverable is a bilingual (English/Spanish) web-based risk calculator providing 10-year risk estimates for middle-aged women, culturally adapted, usability-tested, and available in printable formats for low-resource settings. Within 36 months, the project will advance from data harmonization to prototype development, wide-scale validation, and implementation planning, with expected outcomes including an AUC ≥ 80% and more than 50% of female risk being explainable or preventable through neuroendocrine factors.
PI: Francesca Farina, Claudia Duran-Aniotz & Agustín Ibañez
Support: Wellcome Leap CARE
