First evaluation of N170 latency as an EEG biomarker of treatment response in a randomised controlled trial of arbaclofen for autism

Background Autism clinical trials are hampered by the lack of treatment response biomarkers. We investigated N170 latency (N170L), an electrophysiological (EEG) index of face processing speed, as a candidate biomarker of treatment response to arbaclofen, a selective GABA-B agonist. We previously reported that N170L is, on average, slower in autism, and linked to genetic, neurobiological, and socialisation variability. Here, we explored whether: 1) arbaclofen modulates N170L over 16 weeks; 2) baseline N170L or change in N170L (ΔN170L) predicts change in social functioning over 16 weeks; 3) N170L improves estimation of treatment effects. Methods In this phase II, randomised, double-blind, placebo-controlled trial, 122 autistic children and adolescents (aged 5–17) were administered arbaclofen ( N = 59) or placebo ( N = 63) for 16 weeks. N170L was extracted from EEG recorded during passive viewing of upright faces at week 0 (baseline; N = 96) and week 16 (endpoint; N = 87). Treatment efficacy on social functioning was assessed using the Socialization domain of the Vineland Adaptive Behavior Scales, 3rd edition, alongside parent/clinician-rated outcomes. Pre-registered hypotheses were tested using linear mixed-effects models. Results N170L remained stable over 16 weeks, with no significant change overall ( p = 0.371), and arbaclofen did not modulate N170L relative to placebo ( p = 0.737). Baseline N170L did not predict change in any clinical outcome (all p > 0.472). However, individual differences in ΔN170L were associated with selected clinical outcomes. In the placebo group, slowing of N170L was associated with greater peer interaction difficulties ( p = 0.032) and improved global clinical status ( p = 0.002). For social cognition, the associations differed by treatment group ( p = 0.022), with N170L slowing linked to greater social-cognitive difficulties under placebo, but fewer social-cognitive difficulties under arbaclofen. Baseline N170L and ΔN170L minimally impacted treatment effects estimation as covariates. Limitations Interpretation of N170L–clinical associations is limited by short trial duration and reliance on broad, informant-based social measures that, while clinically meaningful, indirectly index face-processing abilities. Conclusions N170L showed limited utility as treatment response biomarker in this trial, although its relevance in other therapeutic contexts remains to be determined. We also demonstrated feasibility of integrating EEG into multi-site autism clinical trials, supporting future biomarker-enabled pharmacological research. Trial registration ClinicalTrials.gov, NCT03682978. Registered September 25, 2018.