B0233
Title: Double anchoring events based sigmoidal mixed model: an application in Alzheimer's disease progression
Authors: Panpan Zhang - Vanderbilt University Medical Center (United States) [presenting]
Abstract: Understanding the temporal evolution of Alzheimer's disease (AD) biomarkers over the entire continuum of AD is important yet challenging due to the slow progression of AD and the limited resources to collect longitudinal biomarkers from the ageing population with a fully observed clinical spectrum of AD. Sigmoidal mixed models (SMM) have been proposed to characterize non-linear trajectories over a "time to an anchoring event" time scale. However, the use of an anchoring event (e.g., the initial diagnosis of AD) naturally excludes subjects without the anchoring event observed and thus results in selection bias. A double anchoring event-based sigmoidal mixed model (DSMM) is proposed to include a secondary anchoring event (e.g., the initial diagnosis of mild cognitive impairment) such that subjects with either primary or secondary anchoring event observed can be included in the construction of AD progression model. The proposed DSMM is applied to the Alzheimer's disease neuroimaging initiative (ADNI) data to characterize the trajectories of subject memory performance toward AD onset and has shown to perform better than standard SMM and a two-stage SMM approach in capturing memory performance trajectories. This method provides a methodological foundation for trajectory modelling in many neurodegenerative diseases with slow disease progression.