Predicting the Prognosis of Multiple System Atrophy Using Cluster and Principal Component Analysis

Abstract

Background: Multiple system atrophy (MSA) is an intractable neurodegenerative disorder with poorly understanding of prognostic factors. Objective: The purpose of this retrospective longitudinal study was to explore the main predictors of survival of MSA patients with new clinical subtypes based on cluster analysis. Methods: A total of 153 Chinese MSA patients were recruited in our study. The basic demographic data and motor and nonmotor symptoms were assessed. Cluster and principal component analysis (PCA) were used to eliminate collinearity and search for new clinical subtypes. The multivariable Cox regression was used to find factors associated with survival in MSA patients. Results: The median survival time from symptom onset to death (estimated using data from all patients by Kaplan-Meier analysis) was 6.3 (95% CI = 6.1–6.7) years. The survival model showed that a shorter survival time was associated with motor principal component (PC)1 (HR = 1.71, 95% CI: 1.26–2.30, p < 0.001) and nonmotor PC3 (HR = 1.68, 95% CI: 1.31–2.10, p < 0.001) through PCA. Four clusters were identified: Cluster 1 (mild), Cluster 2 (mood disorder-dominant), Cluster 3 (axial symptoms and cognitive impairment-dominant), and Cluster 4 (autonomic failure-dominant). Multivariate Cox regression indicated that Cluster 3 (HR = 4.15, 95% CI: 1.73–9.90, p = 0.001) and Cluster 4 (HR = 4.18, 95% CI: 1.73–10.1, p = 0.002) were independently associated with shorter survival time. Conclusion: More serious motor symptoms, axial symptoms such as falls and dysphagia, orthostatic hypotension, and cognitive impairment were associated with poor survival in MSA via PCA and cluster analysis.