Although Huntington's disease is characterized by motor onset, psychiatric disturbances may present years prior and affect functioning. However, there is inter-individual variability in psychiatric expression and progression. This study therefore strives to stratify longitudinal psychiatric signatures that may inform Huntington's disease prognosis, with potential clinical applications. Forty-six Huntington's disease gene carriers (21 premanifest, 25 manifest; 31 female; age range 25-69) underwent short-Problem Behavior Assessment for depression, irritability, apathy, and dysexecutive behaviors for up to six longitudinal visits. The Disease Trajectories software, a machine-learning approach, was employed to perform unsupervised clustering of psychiatric trajectories. Linear fits were calculated for each cluster. Lastly, the main clusters of shared trajectories were assessed for group differences in demographic and clinical characteristics. The Disease Trajectories analysis software identified two main psychiatric patterns comprising premanifest and manifest patients that explained 54% of the sample. These two clusters evinced a dissociation in the development of depression and irritability; the first cluster was defined by increasing irritability with no depression and the second by a rise-and-fall in depression with no irritability. Both clusters showed a longitudinal increase in clinically relevant apathy and dysexecutive behaviors. Ultimately, through the detection of individual-level psychiatric trajectories with machine-learning, this exploratory study reveals that a dissociation of depression and irritability is apparent even in premanifest stages. These findings underscore individual differences in the severity of longitudinal multivariate clinical characteristics for real-world patient stratification, with implications for precision medicine.
Read full abstract