Exploring EEG resting state differences in autism: sparse findings from a large cohort

BACKGROUND: Autism is a complex neurodevelopmental condition, the precise neurobiological underpinnings of which remain elusive. Here, we focus on group differences in resting state EEG (rsEEG). Although many previous reports have pointed to differences between autistic and neurotypical participants in rsEEG, results have failed to replicate, sample sizes have typically been small, and only a small number of variables are reported in each study. METHODS: Here, we combined five datasets to create a large sample of autistic and neurotypical individuals (n = 776) and extracted 726 variables from each participant's data. We computed effect sizes and split-half replication rate for group differences between autistic and neurotypical individuals for each EEG variable while accounting for age, sex and IQ. Bootstrapping analysis with different sample sizes was done to establish how effect size and replicability varied with sample size. RESULTS: Despite the broad and exploratory approach, very few EEG measures varied with autism diagnosis, and when larger effects were found, the majority were not replicable under split-half testing. In the bootstrap analysis, smaller sample sizes were associated with larger effect sizes but lower replication rates. LIMITATIONS: Although we extracted a comprehensive set of EEG signal components from the data, there is the possibility that measures more sensitive to group differences may exist outside the set that we tested. The combination of data from different laboratories may have obscured group differences. However, our harmonisation process was sufficient to reveal several expected maturational changes in the EEG (e.g. delta power reduction with age), providing reassurance regarding both the integrity of the data and the validity of our data-handling and analysis approaches. CONCLUSIONS: Taken together, these data do not produce compelling evidence for a clear neurobiological signature that can be identified in autism. Instead, our results are consistent with heterogeneity in autism, and caution against studies that use autism diagnosis alone as a method to categorise complex and varied neurobiological profiles.