Refined Subtyping of Autistic Patients Based on Pathogenetic Components

Background:  Using principal component analysis and cluster analysis on 245 patients, we have recently described four patient clusters: (a) patients with prominent immune abnormalities accompanied by some circadian and sensory issues; (b) individuals displaying major circadian and sensory dysfunction, with little or no immune symptoms; (c) a third group of patients characterized by prominent stereotypic behaviors, and (d) a residual group showing a mixture of all four components, with slightly greater developmental delay.

Objectives:  To replicate and extend this initial clustering of ASD patients using a larger and complete data set.

Methods:  We performed 1) hierarchical cluster analysis using a dendrogram and 2) k-means clustering of 286 patients with complete clinical reports, using regression-based factors, each representing one cumulative component score preliminarily obtained by principal component analysis. Hierarchical clustering is a method of cluster analysis which seeks to build a hierarchy of clusters through a "top down" dendrogram approach, where all observations start in one cluster and splits are performed recursively as one moves down the hierarchy. The K-means method is an unsupervised learning algorithm that assumes k clusters fixed a priori and defines k centroids, one for each cluster.

Results:  ASD patients could be categorized into seven subtypes according to cluster dendrogram: three clusters, collectively including 72 patients (25.0%), are characterized by intense immune-related symptoms, accompanied by either normal neurodevelopment, isolated language delay or global developmental delay. A forth cluster displays frequent mental retardation with prominent motor/verbal stereotypies, but little or no immune dysfunction (N=64 patients, 22.0%). A fifth cluster shows a predominance of circadian and sensory-related symptoms in the absence of other relevant features (N=67 patients, 23.3%). A rare sixth cluster displays prominent neurodevelopmental delay (N=2, 0.7%). Finally, 81 (29.0%) patients show a mixed pattern.

Conclusions:  These results confirm and extend our previous 4-cluster definition, by splitting our original “immune” cluster into three “immune” subgroups based on neurodevelopmental delay and by separating out the neurodevelopmental delay subtype which is quite rare in our sample. Despite the long-recognized interindividual variability in clinical phenotype, it seems increasingly possible to dissect clusters of autistic patients based on clinical, patient and family history variables. We shall now proceed to replicate and extend these results using external validators, such as genetic variants, immune underpinnings, developmental trajectories, biomarkers and response to treatment.