Autism spectrum disorder (ASD) is a range of neurodevelopmental problems without certain causes. Conventional diagnostic or screening tools for ASD rely on the observation of children's behavioral presentations. Novel methods are focused on the alterations of some important biochemical matters in ASD patients, which are applicable in the screening for ASD. This study investigated and compared amino acids in the first morning urine from age and sex matched ASD and non-ASD children using high performance liquid chromatography. Significantly lower urinary free methionine, phenylalanine, valine, tryptophan, and leucine plus isoleucine were observed in ASD children. The effects of using urinary free amino acids (UFAAs) singly or conjointly to classify participants into ASD or control group were analyzed and compared. ROC curves on these UFAAs singly in classification performed the sensitivity of 0.593-0.889 and the specificity of 0.704-0.963. Binary-logistic regression analysis of these UFAAs obtained a final regression model comprised of urinary free valine and tryptophan. The ROC curve established by the linear combination of the two amino acids achieved a sensitivity of 0.926 and a specificity of 0.889, which showed superiority to single UFAA and comparability to existing diagnostic or screening tools. It was suggested that the multivariate model based on UFAAs was possibly applicable in screening for children at higher risk of ASD.
Keywords: Amino acid; Autism spectrum disorder; Logistic regression; Multivariate model.
Copyright © 2018 Elsevier Ltd. All rights reserved.