Improving AD Clinical Trial Enrollment Using Machine Learning
Date:
This talk presents a machine learning approach to improve enrollment efficiency in Alzheimer’s disease (AD) clinical trials targeting individuals with high amyloid levels. Using the NACC Uniform Data Set and XGBoost models, we show that cognitive exam scores significantly enhance the ability to predict AD conversion beyond demographics and genetic risk factors alone. Power simulations demonstrate that screening with cognitive data can increase trial power by up to 45.8% compared to demographic-only models. We then extend this work to a larger population without PET scans, predicting both AD risk and likely amyloid status to identify high-priority candidates for imaging. Overall, our method offers a data-driven framework to optimize trial recruitment and make PET-based trials more scalable and inclusive.