Identification of the subgroups of nonmotor symptoms in patients with early Parkinson's disease using cluster analysis

Abstract

Introduction: Parkinsons disease is a clinically and pathologically heterogeneous disorder. Classical classification and data-driven classification of Parkinsons disease subtypes were both analyzed primarily based on motor symptoms and little attention has been paid to the clustering of nonmotor symptoms. Methods: Clinical data on demographic, motor and nonmotor features including the Korean version of sniffin stick test results and the responses to the nonmotor symptoms screening questionnaire were collected from 56 Parkinsons disease patients with disease onset within three years. Nonmotor symptom subgroups were classified using unsupervised hierarchical cluster analysis. Multiscale bootstrap resampling was conducted to validate the confidence of the hierarchical clustering. Results: Forty nine (87.5%) patients indicated hyposmia. Dream-enactment behavior was higher in patients with lower olfactory score which implies worse olfactory function. In whole Parkinsons disease patients, cluster analysis of nine nonmotor symptoms gave three clusters of symptoms. Cluster analysis in de novo subjects revealed the two main clusters without a priori assumptions about the relatedness. The clustering stability was assessed by comparing the results of different methods of measuring similarity and different measures of intergroup distance. We obtain the same clustering results concluded that the group structure stable. Conclusion: Unsupervised clustering using hierarchical approach suggests three nonmotor symptom clusters in whole subjects and two clusters in de novo Parkinsons disease patients. This study is the first report about the identification of subtypes in multiple nonmotor symptom constellations.