Longitudinal brain morphometry in Parkinson's disease clinical subtypes: Distinct signatures forecast behavioral change within subtypes

Abstract

Background: Parkinson's disease (PD) is clinically heterogenous. We previously identified a multi-domain classification of distinct PD subtypes in participants without dementia, yielding Motor Only, Psychiatric & Motor, and Cognitive & Motor subtypes. These PD subtypes may exhibit different rates of brain atrophy. We aim to determine baseline and longitudinal morphometric brain signatures of PD clinical subtypes and their relationship to behavioral change.

Methods: 213 PD and 73 control participants completed T1-weighted structural MRIs and motor, cognitive, and psychiatric measures at baseline and every 1-3 years. We used FreeSurfer-v7.3 to obtain cortical thickness and brain volume measures, with manual edits as necessary. We used hierarchical linear models for longitudinal analyses-modeling time, subtype and their interaction; predictors include baseline age, motor symptom duration, sex, and education.

Results: Morphological brain MRI measures differentiate PD clinical subtypes at baseline, and the subtypes demonstrate different longitudinal rates of atrophy. The Motor Only subtype has preserved baseline morphometry and slow rate of atrophy, the Psychiatric & Motor has widespread baseline cortical thinning and slow rate of atrophy, and the Cognitive & Motor has baseline global and regional volume loss and fast rate of atrophy. Additionally, baseline morphometric measures predict longitudinal behavioral change within the Cognitive & Motor subtype.

Conclusion: Multi-domain PD clinical subtypes exhibit distinct baseline brain morphometric patterns and predictable longitudinal volumetric changes that substantiate previous observations of differential risks of progression, dementia, and death. These findings suggest underlying biological differences across subtypes that may help inform clinical expectations and guide clinical trial design.

Fig. 1.. Baseline volumetric differences across PD clinical subtypes and controls.

A. Global volumetric differences across groups. The mean and distribution of predicted volumes after adjusting for co-variates of age, sex, education, motor symptom duration, and total intracranial volume is plotted by group. The Cognitive & Motor subtype exhibits significantly larger left and right lateral ventricles and smaller brain volume without ventricles at baseline compared to the ‘motor only’ subtype (p = 0.037, 0.01 & 0.013 respectively), and a significant difference compared to control in brain volume without ventricles (p = 0.01). Other pairwise comparisons in these measures did not reach significance. B. Significant differences in cortical thickness across subtypes and controls are displayed on Desikan-Killiany atlas [63] using ggseg tools [64] in R. Each subtype compared to controls and to the other two subtypes is visualized separately; green (for Motor Only), blue (for Psychiatric & Motor) and red (for Cognitive & Motor) colors indicate thinner cortex in the specific subtype compared to the comparison group (p < 0.05). No cortical region was thinner in the control group than any of the subtypes. See supplementary material for a table of estimates and p values.

Fig. 2.. Longitudinal trajectories of brain volumes over time across PD clinical subtypes and controls

Longitudinal line plots illustrate intra-individual trajectories of volumetric change over time (visit year). Lighter lines represent individual participant raw data, while bold lines depict modeled group-level trajectory for each group (including age at baseline, motor symptom duration at baseline, sex, education, and total intracranial volume as predictors). A. Individuals with the Cognitive & Motor subtype have significantly more ventricular enlargement and brain atrophy (decreasing brain segmentation volume without ventricles) over time compared to the control group, the Motor Only subtype, and the Psychiatric & Motor subtype (p-values are <0.0001, 0.001, and 0.01 for the left lateral ventricle, <0.0001, <0.0001, and 0.01 for the right lateral ventricle, and <0.0001, 0.005, and 0.01 for the brain segmentation volume without ventricles, respectively). B. The Cognitive & Motor similarly showed regional subcortical accelerated volume loss compared to the control group and the Motor Only and Psychiatric & Motor subtypes; this was notable for left cerebellar cortex (p=<0.0001, 0.015, and 0.07 (not significant), respectively), right cerebellar cortex (p=<0.0001, 0.0008, and 0.04, respectively), left hippocampus (p=<0.0001, 0.003, and 0.04, respectively), and right hippocampus (p=<0.0001, 0.003, and 0.06, respectively). Supplementary Table 4 lists all pairwise comparisons across subtypes.

Fig. 3.. Assessing if baseline brain volume and thickness measures predict longitudinal behavioral change across PD clinical subtypes

After selecting MRI measures that were significantly different across PD subtypes at baseline or longitudinally, growth models were employed to assess whether those MRI measures predict intra-individual behavioral change within each subtype. Additionally, interactions involving behavioral measures that did not significantly change over time within the specific subtype were excluded from the results. Top three rows focus on the Cognitive & Motor subtype. Top row: Baseline lateral ventricular volume predicts worse cognitive performance globally and in individual domains and higher PIGD score overtime. Second row: Thinner cortex (as measured by mean cortical thickness) predicts worse visuospatial scores and worse motor scores. Third row: Examples of other regional cortical and subcortical measures. Bottom row provides examples of significant interactions in the other two subtypes. See Supplementary Table 7 for a full account on all significant interactions including estimates and p-values. R: right; L: left. The ‘mean’, ‘low’, and ‘high’ labels denote the mean, +1 standard deviation, and −1 standard deviation of the MRI measure distribution within the subtype group. PIGD: Postural Instability and Gait Disorder; UPDRS-III: Unified Parkinson’s Disease Rating Scale part 3.