Objectives: The classical Braak neuropathological staging model in Parkinson's disease (PD) suggests that brain lesions progress from the medulla oblongata to the cortex. An alternative model in which neurodegeneration first occurs in the cortex has also been proposed. These two models may correspond to different patient phenotypes. To test these two models and investigate if they were influenced by the presence of rapid eye movement sleep behavior disorder (RBD) we used multimodal MRI and Partial Least Squares Path-Modeling (PLS-PM) assuming that patients with RBD followed distinct neurodegeneration pattern.

Methods: Fifty-four patients with PD (thirty-four with RBD) and twenty-five healthy volunteers were scanned using T1-weighted, diffusion tensor and neuromelanin-sensitive imaging. Volume, signal, mean, axial and radial diffusivities were calculated in brainstem, basal forebrain and cortical regions. PLS-PM, estimating a network of causal relationships between blocks of variables was used to build and test an analytical model based on Braak staging. The overall quality of the model was assessed using Goodness of fit coefficient (Gof).

Results: PLS-PM was run on PD patients with RBD and without RBD separately. In PD with RBD, a brainstem-to-cortex model had significant Gof (0.71, p=0.01) whereas a cortex-to-brainstem model was not. In contrast, in PD patients without RBD, the brainstem-to-cortex model was not significant (Gof=0.64, p=0.27), and the cortex-to-brainstem model was highly significant (Gof=0.72, p=0.008).

Conclusions: Using the PLS-PM imaging-based model, the neurodegeneration pattern of PD patients with RBD was consistent with the Braak brainstem-to-cortex model, whereas that of patients without RBD followed the cortex-to-brainstem model.