Abstract:
|
Disease progression in Alzheimer’s Disease (AD) is measured by a variety of clinical assessments and biomarkers, with each capturing different aspects of the disease. When conducting clinical trials to test the efficacy of treatments for AD, it is possible for multiple outcomes to demonstrate support for efficacy without reaching statistical significance. By combining information across outcomes using composite end points like ADCOMS and global statistical tests it is possible to design trials that have higher power with lower sample sizes. We discuss methods for selecting items into composite end points, including partial least squares regression, elastic net, and global statistical tests, and present results from simulations and analysis of historical controls to demonstrate the properties of composite endpoints and how they can better align with disease progression and more accurately measure efficacy than individual outcomes, while controlling type 1 error.
|