The wealth of biomarker data now available offers an unprecedented opportunity to better understand the complex pathways of Alzheimer’s Disease (AD), essential to improve prevention strategies and individual healthcare. However, current statistical tools are inefficient at modelling jointly the multiple AD biomarkers and capture their relationships rigorously.
We propose an original modelling framework of AD to describe the complex dynamics of the disease dimensions and apprehend their causal temporal relationships, via several correlated underlying pathological processes. Each of the various processes is measured by several repeated markers. To reduce potential selection biases, we account for informative missing data due to dementia and death.
The dynamic model will be applied on the French prospective cohort Paquid to disentangle the temporal relationships between depression, cognition and functional autonomy, and identify relevant target domains which may slow down the evolution of AD, at each stage of the disease.