Objective: To compare machine-learning models to predict the progression of PD using gut microbiota and clinical features.

Background: A total of 21 studies including ours has been reported on gut microbiota in PD. We showed that genus Akkermansia is increased and genera Roseburia and Faecalibacterium are decreased in PD across five countries. Previous studies showed that aging, male, advanced Hoehn & Yahr (H&Y) stages, cognitive impairment, and the PIGD-PD (postural instability gait difficulty-PD) subtype are predictive markers for poor prognosis in PD. However, no prognosis-predicting model has been reported with gut microbiota in PD.

Method: (1) Clinical features were followed up in 223 PD patients for two years. Sequencing of 16S rRNA V3-V4 regions was performed at years 0 and 2.
(2) PD patients were divided into the stable and deteriorated groups by reviewing the advancement of the H&Y stage.
(3) Random-forest models to differentiate the stable and deteriorated groups were generated using bacterial features and 32 clinical features (age, sex, MMSE, and others).
(4) Abundances of four intestinal genera that were changed in PD in our dataset (Akkermansia, Faecalibacterium, Fusicatenibacter, and Blautia) were analyzed at years 0 and 2 in controls, the stable group, and the deteriorated group.
(5) These four genera were also analyzed as a function of H&Y stage.

Results: (1) Two-year prognoses of PD patients at H&Y stages 1-3, as well as at all H&Y stages, were predicted better with gut microbiota than with clinical features. In patients with H&Y stage 1, genus Faecalibacterium alone efficiently predicted disease progression (AUC of ROC curve = 0.86).
(2) Increase of Akkermansia or decrease of Faecalibacterium and Blautia well predicted disease progression, but these genera remained unchanged in two years in both the stable and deteriorated groups.
(3) Genus Akkermansia was increased, whereas genera Faecalibacterium, Fusicatenibacter, and Blautia were decreased with progression of H&Y stage.

Conclusion: The microbiota-based models predicted disease progression in PD better than the clinical feature-based models. PD patients with high Akkermansia or low Faecalibacterium progressed faster, but their abundances remained unchanged even in the deteriorated group in two years. High Akkermansia and low Faecalibacterium are unlikely to be the consequence of PD progression, but are likely to cause PD progression.

« Back to MDS Virtual Congress 2021

MDS Abstracts - https://www.mdsabstracts.org/abstract/gut-microbiota-in-patients-with-parkinsons-disease-pd-predicts-disease-progression-better-than-clinical-features/