Neuroprediction of Treatment Effectiveness in Young Children with Autism Spectrum Disorder

Thursday, May 12, 2016: 11:30 AM-1:30 PM
Hall A (Baltimore Convention Center)
D. Yang, K. Pelphrey, C. A. Paisley, M. L. Braconnier, S. M. Abdullahi, D. G. Sukhodolsky, M. J. Crowley and P. E. Ventola, Yale Child Study Center, Yale School of Medicine, New Haven, CT
Background: Advances in genetics, molecular biology, and cognitive neuroscience offer promise towards personalized treatment to improve outcomes in individuals with ASD. Recent clinical trials have shown favorable results; however, the promise of precision medicine is hindered by a lack of sensitive, objective measures to identify subgroups likely to respond to specific treatments. Instead, our field relies on availability of service, trial-and-error, and clinical judgment to make treatment decisions. Here we built upon our prior research characterizing the neural-systems-level basis of core social communication symptoms in ASD to identify a potential stratification biomarker.

Objectives: We investigated the degree to which fMRI neurobiomarkers predict treatment response in a sample (N = 17) of children (4-7 years; 5 girls, 12 boys) with ASD (Mean IQ=102.82, SD=16.72) who participated in a 16-week trial of Pivotal Response Treatment (PRT), a behavioral treatment focused on social communication skill development.

Methods: Treatment included 6 hours per week of individual work with the child plus parent training. Primary clinical outcome was the SRS-2 Total Raw Score. During a 5-minute fMRI scan at 3 Tesla, conducted at baseline, the participants viewed well-validated neuroimaging stimuli depicting point light displays of coherent biological (BIO) or scrambled biological (SCRAM) motion. We evaluated the extent to which activation in response to viewing BIO vs. SCRAM at baseline predicted the change in the SRS score from baseline to the treatment endpoint, while controlling for the baseline SRS score.

Results: Using a whole-brain group analysis (mixed-effects modeling using FSL’s FLAME1+2, voxel-level thresholding Z > 2.33, cluster-level thresholding p < .05, controlling for sex), we estimated the correlation between change in SRS total raw score from baseline to treatment endpoint and magnitude of pre-PRT brain response to BIO vs. SCRAM. This revealed three clusters of neuropredictive activity (Figure 1). Cluster 1 (green) was centered in the right ventrolateral prefrontal cortex, orbitofrontal cortex, anterior insula, and temporal pole. Cluster 2 (blue) was centered in the right fusiform gyrus, inferior and middle temporal gyri, and superior temporal sulcus. Cluster 3 (red) was centered in the left putamen, pallidum, hippocampus, amygdala, and ventral striatum/nucleus accumbens. Figure 2 showed the scatter plot of the improvement in social communication skills (y-axis) vs. pre-PRT BIO > SCRAM activity (x-axis) for each of the three clusters. Strikingly, none of the demographic (age, IQ, sex) or baseline behavioral (ADOS, ADI-R, SRS, Vineland-II, CELF) variables predicted response to treatment.

Conclusions: We discovered a neuroimaging-based biologically informed stratification biomarker that predicts magnitude of response to an evidence-based behavioral treatment in young children with ASD. Neurosynth results suggest that baseline levels of activity in well-known emotional regulation (cluster 1), social perception and face recognition (cluster 2), and social reward/motivation and emotion (cluster 3) networks predicted the magnitude of clinical response to PRT. Importantly, these biomarkers outperformed pre-treatment behavioral measures of social functioning, language level, and cognitive abilities. Our results provide the first-ever clear evidence of a neuroimaging-derived stratification biomarker in ASD and help the field progress to the goal of targeted, personalized treatment for individuals with ASD.