Objective: To determine the value of MRI metrics in predicting global cognitive ability in Parkinson’s disease (PD).

Background: Progressive cognitive impairment is a dominant feature of the clinical profile of PD, with most patients eventually developing dementia. We and others have shown robust group-level associations between cognitive impairments and multiple MRI measurements, including grey matter (GM) volume, white matter hyperintensities (WMH), and DTI metrics (fractional anisotropy, FA; mean diffusivity, MD). However, the utility of MRI metrics in predicting cognitive decline, at the individual level, is not yet known.

Methods: 177 PD participants representative of the cognitive spectrum and 51 healthy controls were assessed approximately every two years, up to eight years. Each assessment included a neuropsychological battery (consistent with Movement Disorder Society Task Force level II criteria) and MRI scan. Global cognition was expressed as an aggregate z score derived from multiple tests within each of five cognitive domains. MR data included a T1-weighted 3D SPGR image, a T2-weighted FLAIR image, and diffusion tensor imaging (DTI) at each time point. Global MRI metrics included total GM volume, total WMH volume, and average skeletal FA and MD. 589 scans were included in the analysis. Beyond patient characteristics (age, sex, and group status of PD/control), global MRI metrics from all individuals and all time points were entered into Bayesian hierarchical models to predict global cognitive ability. We used WAIC, a measure of out-of-sample individual predictive accuracy, to determine which MRI metrics were most important.

Results: The three most important variables for predicting global cognitive ability for an individual were age, a diagnosis of PD, and GM volume. WMH volume also increased predictive accuracy, but not once GM had been added to the model. Average skeletal FA and MD were not useful predictors of global cognitive ability.

Conclusions: Total GM volume was the only global metric to increase predictive accuracy of cognitive ability at the individual level. Thus, GM volume has a closer association with cognition than global WMH, skeletal FA, or MD, and may have utility as a proxy measure for cognitive ability. The predictive value of regional and network GM, WMH, FA, and MD should be investigated in future analyses.

« Back to 2018 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/longitudinal-mri-to-predict-cognitive-impairment-in-parkinsons-disease/