Background: Autistic children and adolescents frequently exhibit motor abnormalities, including toe walking (TW). Previous research has found persistent TW (i.e., TW that would not resolve spontaneously by 3-7 years of age) associated with greater cognitive impairment, language delay, and orthopedic complications, while evidence of abnormal sensory processing is less consistent.

Objectives: The goal of this study is to outline the overall schema of natural relationships between clinical variables and Short Sensory Profile (SSP) scores in a sample of autistic children and adolescents with and without TW, by applying artificial neural networks

Methods: 256 clinical variables across four domains (prenatal-perinatal issues, medical and neurological features, developmental characteristics, family history) were retrieved for analysis from a cohort of 112 autistic children and adolescents assessed for TW (see Camia et al., 2024, doi: 10.1016/j.rasd.2024.102457). Each domain was processed using Auto-CM, a fourth-generation auto-associative neural network. After training, this system developed weights that reflect the strength of associations between all variables. A minimum spanning tree filter was then applied to the resulting weight matrix, allowing for a graphical representation of the strongest connections between variables, in other words their semantic connectivity map. Preliminary analyses performed on each domain identified 31 clinical variables with up to one degree of separation from “TW-present”/”TW-absent”. Six variables were then dichotomized yielding a total of 41 clinical variables. Final analyses were then performed on complete data sets applying Auto-CM: (a) to these 41 clinical variables on a sample of 79 patients (N=46 with and 33 without TW), and (b) by adding also 24 variables, three for each SSP domain (clinical vs borderline vs normal) and for its total score, available for 61 patients (36 with TW and 25 without TW ).

Results: The semantic connectivity map generated by Auto-CM revealed a clear separation between presence/absence of TW: (a) Analyzing only clinical variables, “TW-present” was identified as a hub node, linked to motor coordination disorder, abnormal cranial conformation, dysmorphisms, reduced pain sensitivity, hypersensitivity of other senses, sensory self-stimulation, hyperactivity, stereotypic behaviors, and severe autism. TW-absence was linked to presence of imitation/pretend play, mild autism severity, and better language skills, both receptive and expressive. (b) Including also SSP variables, auto-CM generated a map also connecting TW with a string of SSP scores in the clinical range (hyporeactivity, auditory filtering, tactile sensitivity, etc), counteracted by normal SSP scores linked to “TW-absent”.

Conclusions: These results confirm the greater severity of motor, cognitive, social, and verbal impairment associated with TW. Sensory abnormalities, linked to TW here but not in our prior study, are likely part of a constellation of signs and symptoms expressing a more pervasive and profound derangement of neurodevelopmental processes, rather than a direct cause of TW. More research will be necessary to understand the neural substrates of TW in autism. Advanced computational tools, like Auto-CM, are indeed helpful in overviewing and partitioning complex behavioral phenotypes.

Notes: 

(1)Dept. of Biomedical, Metabolic and Neural Sciences, University of Modena and Reggio Emilia, Modena, MO, Italy, (2)Service for Neurodevelopmental Disorders, “Campus Bio-Medico” University Hospital, Rome, RM, Italy, (3)IRCCS Centro Neurolesi Bonino Pulejo, Messina, Italy, (4)Center for Autism “Dopo di noi”, Barcellona P.G., Messina, Italy, (5)Cantonal Psychiatric Clinic, Cantonal Socio-psychiatric Organization (O.S.C.), Mendrisio, Repubblica e Cantone Ticino, Switzerland, (6)Institute for Biomedical Research and Innovation (I.R.I.B.), National Research Council of Italy (C.N.R.), Messina, ME, Italy, (7)Child Neuropsychiatry Unit, "G. Martino" University Hospital, Messina, ME, Italy, (8)Autism Research Unit, Villa S. Maria Foundation, Tavernerio, CO, Italy