A Sparse Panel of Biomarkers In Blood Distinguishes Children with Autism Spectrum Disorder From Typically Developing Children

Background:  Despite being a clear neurobiological disorder, no medical tests to detect autism in infants and toddlers exist. Currently, the disorder is identified based solely on the presence of clinical symptoms, and considerable expertise is required to identify autism during the first 1-3 years of life. Blood tests are routine in medical offices, and as such, hold promise as a method to identify autism prior to the onset of clinical symptoms. Evidence of an immune system dysfunction in autism has risen rapidly to the research fore and markers such as those in the interleukin family have been implicated in autism. However, biomarkers may play different roles at different times in development. For example, sFas and TNF-α are involved in early brain development, although they are commonly thought of as involved in immune function in the mature brain. As such, it may be possible to develop a combination of biomarkers that discriminates toddlers with an autism spectrum disorder (ASD) from those that are typically developing (TD) or developmentally delayed (DD). 

Objectives: To determine if biomarker levels can accurately classify toddlers with an ASD from those that are TD or DD during the first 3 years of life. 

Methods: Nine biomarkers (TNF-α, IL-6, IL-10, IP-10, sIL-6R, sFas, VEGF, sVEGFR-1 and tPAI-1) were measured in the blood plasma of two separate, independent samples of toddlers. In sample 1, the training sample, 142 toddlers between the ages of 12 and 48 months participated (49 ASD, 66 TD and 27 DD;  mean age 23.5 months). In sample 2, the test sample, 78 toddlers participated (39 ASD and 39 TD). Biomarkers were measured in duplicate using multiplex bead suspension array assays with a Luminex 100 platform (Bio-Rad, Hercules, CA). Classification and regression tree analyses were performed on both the training and test samples. The markers that most clearly discriminated between groups in the training sample were used as predictor variables to classify subjects in the test sample. 

Results: The training sample identified that lower levels of IL-6 and TNF-α and higher levels of sFas discriminated ASD from TD toddlers. Overall, the classifiers achieved 80% accuracy on the training dataset. Using bootstrap methods and evaluating the classifier on the the test set, the classification accuracy was 60%. The accuracy of prediction on the test dataset was 63%. Using a permuation test, the classification accuracy was significantly better than chance on the test dataset (p=0.004).

Conclusions: Abnormalities in biomarker profiles may signify abnormal brain growth and function commonly found in autism by disrupting the balance between the immune system, neural growth factors, neural stem cells, and neurotransmitters in the developing brain. Results indicate that it may be possible to detect these abnormalities in blood and to develop a simple and inexpensive early diagnostic test for ASD.