CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Longitudinal modeling of disease progression biomarkers in the latent disease timescale: Example of Alzheimer's disease Authors:  Cecile Proust-Lima - INSERM (France) [presenting]
Jeremie Lespinasse - Univ Bordeaux (France)
Carole Dufouil - INSERM - Univ Bordeaux (France)
Abstract: Alzheimer's disease and related disorders (ADRD) are characterized by progressive changes in multiple components, including protein accumulation in the brain, brain atrophies, and cognitive dysfunction. Understanding the sequence and timing of such deteriorations is paramount to refine patient stratification and facilitate earlier diagnosis. However, their modeling faces a fundamental statistical challenge: the timescale is not known. Usual timescales are inappropriate: (i) time of clinical diagnosis is not an option as most of the deteriorations appear years to months before, (ii) time since inclusion does not have any biological meaning, and (iii) chronological age induces too much inter-individual heterogeneity as people do not age similarly and ADRD onset may arise at different ages. We discuss how the mixed model theory applied to multivariate longitudinal biomarkers can be used to realign individual trajectories into a common latent disease time while taking into account the specificities of the biomarkers. We then illustrate the method to describe the sequence of progression of 12 biomarkers in the French clinic-based Memento study. Beyond the sequence of biomarker degradation, this methodology may evaluate at what stage of the disease an individual is by providing a prediction of his/her individual disease time. This has the potential for earlier diagnosis.