Neural Sensitivity to Social Stimuli at 8 Months Predicts Social Behaviour at Two Years and ASD-Outcome at Three Years.

Background: Autism Spectrum Disorder (ASD) is typically diagnosed at age two or three, and is characterised by difficulties in social communication and interaction. Atypical neural responses to social stimuli have been associated with difficulties in socialisation in toddlerhood (Webb et al. 2011; Dawson et al., 2012). Early evidence suggests that such brain-related changes emerge earlier than clinical manifestations, raising the possibility of early identification (Elsabbagh et al., 2012; Jones et al., 2016). In order to have clinical potential, such effects must be replicated and extended to individual-level prediction.

Objectives: We first tested whether differences in neural correlates of attention engagement to social stimuli at 8 months are associated with atypical social behaviour at two years and autism outcome at three years (study 1). Additionally, we investigated whether a specific subset of brain responses to the same stimuli can predict individual ASD outcome (study 2).

Methods: EEG activity was recorded from 67 low-risk (LR) infants and 145 high-risk (HR) siblings while looking at faces with direct and averted gaze and visual noise (Figure 1). Clinical outcome of HR infants was established at 36 months (typical development [HR-TD; n=77]; atypical development [HR-Aty; n=37]; ASD [HR-ASD; n=31]). The Vineland Adaptive Behavior Score (VABS) questionnaire was used to assess social behaviour at 24 months.

Analysis 1 examined whether groups showed differences in mean amplitude of the Nc (attention-related) component when looking at face versus noise. In study 2, a Support Vector Machine (SVM) classifier discriminated HR-ASD from other siblings using ERP features selected by a genetic algorithm.

Results: Analysis 1: LR and HR groups showed significantly different Nc responses to the face-noise contrast (p=.04), with HR showing a significantly less negative early Nc component when looking at faces vs noise (p<.000). HR-ASD infants showed a larger difference in mean amplitude of the early Nc over the right frontal region with respect to the other groups (t(340)=-2.09, p=0.037). The amplitude difference in the face-noise contrast in infancy significantly predicted VABS socialization scores at two years (beta=-0.55, p=.007, Figure 2). Analysis 2: classification of HR-ASD at 8 months was possible with 80.4% Area Under the Curve (AUC; 95% confidence interval, CI [72.4, 87.6]). The most relevant features included differences in P4 (latency) and P1 for the gaze shift contrast; and P1 (amplitude), N290 (latency), and P4 for the face-noise contrast. Finally, responses to faces and the face-noise contrast showed high predictive power for ASD (80.7% AUC; 95% CI [72.5, 88.0]).

Conclusions: Differences in neural responses to faces vs. noise at 8 months can predict socialisation skills at 2 years, and individual ASD outcome at 3 years. Early disruptions to social attention engagement can be critical to later development of ASD traits such as social interaction difficulties, and have important clinical implications for early risk assessment and targeted intervention.