Speech subtypes are associated with worsened tremor and axial symptoms in Parkinson’s disease patients

Abstract

Background

Parkinson’s disease (PD) is a heterogeneous disorder, suggesting the presence of distinct subtypes. Speech data, though easy to collect, remains underutilized in subtyping PD.

Methods

Cross-sectional study with PD patients recruited from the Movement Disorders Outpatient Clinic of the Neurology Service at the University Hospital in Porto Alegre, Brazil. We included participants diagnosed with idiopathic PD and excluded participants with other disorders that could affect speech. Clinical and sociodemographic data were collected alongside MDS-UPDRS Parts II and III motor assessments. Tremor and gait-posture scores were derived from specific MDS-UPDRS items, with additional data on Deep Brain Stimulation (DBS) status and Levodopa Equivalent Daily Dose (LEDD). The tasks diadochokinesis (DDK) and monologue were recorded and acoustically analyzed using software. We compared our identified clusters using clinical data through an analysis of covariance adjusted for age, sex, and disease duration.

Results

Ninety individuals with PD were included, with 61.2 (± 9.4) years old, 13.6 (± 6.6) disease duration, and 47.6 (± 10) age at onset. We identified three speech groups with strong separation between them, comprising 49 (mild), 13 (moderate), and 29 (severe) patients. Tremor and postural-gait stability scores differed significantly across the three clusters, with cluster 3 exhibiting higher tremor (13.42 ± 10.66 vs. 7.09 ± 6.62, p = 0.020) and greater postural-gait instability (10.25 ± 6.69 vs. 5.46 ± 4.91, p = 0.009) than cluster 1. These differences weren’t explainable by distinct age, sex, or disease duration.

Conclusion

Our speech-based clustering algorithm effectively differentiated Parkinson’s disease subtypes in this sample, identifying distinct groups based on tremor and axial symptoms.