Building on prior work by FH, the first main objective of this project is to stratify the trajectories of patients with a confirmed PD diagnosis and to predict the individual rate of disease progression from baseline, using data from PPMI, LuxPARK and ICEBERG. One of the challenges to be addressed in this context is that patients in many cohort studies already demonstrate a high variability of disease progression at study baseline, which can lead to severe biases of models, if they are naively trained. One possibility to address such bias is to rely on a linear mixed modelling framework to interpolate between patient trajectories and align them to a common disease time scale. However, non-linear techniques, including neural networks with mixed effects, have also been devised recently, and these techniques may potentially be better able to capture the long-term behaviour of disease progression.
The specific aims of this project are thus:
The second main objective is to study the potential benefit of stratifying patients by disease progression for clinical trial design, as e.g. demonstrated in prior work by FH for other indication areas. We will investigate the sample size reduction that could be achieved using such stratification, and whether future clinical trials for a drug would preferably recruit patients that have been predicted as fast progressors at the screening phase. Different settings will be considered regarding patient inclusion/exclusion criteria, endpoints, control arm/active comparator as well as anticipated effect sizes of a future drug. DC2 focuses on disease progression assessed via traditional clinical assessments, whereas DC4 will explore non-traditional digital speech assessments for monitoring of disease progression. Both DCs will closely work together.
This project is part of the "Precision Neurology" work package.
Artificial Intelligence in Parkinson’s Disease