Identification of Neuropsychological Phenotypes in Pediatric Multiple Sclerosis Using Unsupervised Machine Learning
Damiano Mistri1, Monica Margoni2, Paolo Preziosa4, Alessandro Meani1, Carmen Vizzino1, Lucia Moiola3, Massimo Filippi5, Maria Rocca6
1Neuroimaging Research Unit, Division of Neuroscience, 2Neuroimaging Research Unit, Division of Neuroscience, and Neurology Unit, 3Neurology Unit, IRCCS San Raffaele Scientific Institute, 4Neuroimaging Research Unit, Division of Neuroscience, IRCCS San Raffaele Scientific Institute; Neurology Unit, IRCCS San Raffaele Scientific Institute; Vita-Salute San Raffaele University, 5Neuroimaging Research Unit, Division of Neuroscience, Neurology Unit, Neurorehabilitation Unit, and Neurophysiology Service, 6Neuroimaging Research Unit, Division of Neuroscience, and Neurology Unit, IRCCS San Raffaele Scientific Institute; Vita-Salute San Raffaele University
Objective:
To identify cognitive phenotypes in pediatric multiple sclerosis (MS) using an unsupervised machine learning approach.
Background:
Recent studies identified distinct cognitive phenotypes in adult MS patients using unsupervised machine learning approaches.
Design/Methods:
Fifty-three pediatric MS patients underwent clinical examination including Expanded Disability Status Scale (EDSS) to assess global disability. A comprehensive neuropsychological battery (Selective Reminding Test, Spatial Recall Test, Trail Making Test, Symbol Digit Modalities Test, Semantic verbal fluency test and Phonemic verbal fluency test) was administered to all patients. Raw scores of each test were converted to z-scores according to normative data. Cognitive tests z-scores were included in a k-means cluster analysis to identify cognitive phenotypes. Between-group comparisons of demographic, clinical and cognitive variables were computed using linear regression models or nonparametric tests as appropriate. Statistical significance was corrected for multiple comparisons (Bonferroni method).
Results:
In our sample, median age was 15.9 years (interquartile range [IQR] 13.6 – 16.8), 68% (36 patients) were girls and the mean disease duration was 1.8±1.7 years. Cluster analysis identified three cognitive phenotypes: preserved cognition (27 patients [51%]), mild verbal memory/ semantic fluency involvement (19 patients [36%]) and severe multidomain involvement (7 patients [13%]). Across groups, there were no significant differences in age, sex and disease duration. Lower EDSS scores were found in patients with preserved cognition (median EDSS 1.0 [IQR 1.0 – 1.0]) than patients with mild verbal memory/ semantic fluency involvement (median EDSS 1.5 [IQR 1.0 – 2.0]; p=0.01) and patients with severe multidomain involvement (median EDSS 2.0 [IQR 1.25 – 2.75]; p=0.004).
Conclusions:
Pediatric MS patients present with distinct cognitive phenotypes ranging from preserved cognition to severe multidomain involvement. Cognitive phenotypes are not associated with age, sex and disease duration. Patients with preserved cognition are characterized by low global disability.