Neuroanatomical Biotypes Can Predict Parkinson Disease Progression

August 24, 2020
Marco Meglio
Marco Meglio

Marco Meglio, Associate Editor for NeurologyLive, has been with the team since October 2019. Follow him on Twitter @marcomeglio1 or email him at

Recognition of the distinct neuroanatomical patterns, symptoms, and rates of progression within different biotypes can play a pivotal role in the implementation of precision medicine for Parkinson disease.

Data collected from a cluster of patients with Parkinson disease (PD) from the Parkinson’s Progression Markers Initiative (PPMI) revealed that robust neuroanatomical biotypes exist in PD with distinct clinical and neuroanatomical patterns that can predict the course of longitudinal progression.

Lead author Lindo Wang, PhD, post-doctoral researcher, Institute of Science and Technology for Brain-inspired Intelligence, Fuden University, and colleagues identified 2 neuroanatomical biotypes through the PPMI. Biotype 1 comprised of 114 patients with PD who had subcortical brain volumes smaller than healthy controls, while biotype 2 consisted of 200 patients with PD that had subcortical brain volumes larger than healthy controls.

The investigators noted that biotype 1 had more severe motor impairment, autonomic dysfunction, and very much worse rapid eye movement (REM) sleep behavior disorder than biotype 2 at baseline (all P <.05; FDR correction). Additionally, these patients with PD and smaller subcortical brain volume had poorer prognosis, with more rapid decline in several clinical domains and in dopamine functional neuroimaging over an average of 5 years, despite disease duration being similar at initial visit and follow-up between the biotypes.

Wang and colleagues noted, “Neuroanatomical biotypes differed in symptomatology even at presentation before treatment, and thereafter progressed at different rates. The most striking baseline difference was the much higher REM Sleep Behavior Disorder screening questionnaire score in biotype 1.”

Voxel-level neuroanatomical features were estimated by the group using deformation-based morphometry (DBM) of T1-weighted magnetic resonance imaging (MRI), and revealed the distinct patterns that differentiated the 2 biotypes. Compared to controls, DBM values in the putamen, caudate, pallidum, lingual gyrus, temporal cortex, insula, amygdala, hippocampus, and orbital frontal cortex were significantly lower in biotype 1 patients.

READ MORE: Noninvasive Neuromodulation for Essential Tremor: An Important Tool for Maintaining Independence at Home

Lower values of hippocampus and orbital frontal cortex, which regulate motor-related, cognitive, and emotional related behavior were characterized in part by increased motor and non-motor symptom scores. In contrast, significantly higher DBM values in the brainstem, putamen, caudate, occipital lobe, lingual gyrus, olfactory, posterior cingulate cortex, and white matter areas were found in biotype 2 patients as opposed to controls.

“In summary, we hypothesize that the different rates of symptom progression relate to different brain reserves,” the investigators concluded. Brain reserves describe the differences in brain volume and structure that may support maintenance of function despite pathology. The study investigators noted that gross or regional brain volume reflects the quantity of neurons, neuronal integrity and synaptic densities, which can determine the brain’s ability to engage in compensatory activity.

Baseline differences in symptoms between biotypes revealed worse mentation, behavior and mood, indicated by higher Movement Disorder Society—sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) and REM Sleep Behavior Disorder Screening Questionnaire (RBDSQ) in biotype 1 compared to biotype 2.

Wang and colleagues reasoned that clustering could be used to train classifiers for the diagnosis of PD biotypes solely on the basis of structural MRI measures. They then trained classifiers to predict the PD biotype and found that support-vector machine (SVM) classifiers, using Gaussian kernel functions, yielded overall accuracy rates of 84.1% (sensitivity: 0.71; specificity: 0.89; area under curve [AUC]: 0.90).

After an average of 4 years of follow-up, biotype 1 had worse denervation of both left caudate and left putamen on dopaminergic single-photon emission computed tomography (SPECT) scanning. Notably, the right caudate and right putamen showed no significant difference in denervation between the 2 biotypes.

Investigators validated clustering results by investigating the differences in follow-up clinical symptoms and neuroanatomical patterns between sub-groups of patients, as well investigating the extent to which the analysis could reliably discriminate between different sub-groups of patients using a pattern classification approach.


Wang L, Cheng W, Rolls E, et al. Association of specific biotypes in patients with Parkinson disease and disease progression. Neurology. Published online August 14, 2020. doi: 10.1212/WNL.0000000000010498

Related Content:

Clinical | News | Movement Disorders