Lack of univariate, clinically-relevant biomarkers of autism in resting state EEG: a study of 776 participants

Abstract

Abstract Neurodevelopmental disorders are difficult to diagnose, requiring expert clinicians and subjective judgements. Consequently, there has been interest in finding quantitative biomarkers for some disorders using resting state electroencephalogram (EEG) data. Here, we focus on resting state EEG biomarkers of autism. Although many previous reports have pointed to EEG-based differences between autistic and neurotypical participants, results have often failed to replicate and sample sizes have typically been small. Taking a big-data, open-science approach, we combined data from five studies to create a large sample of autistic and neurotypical individuals (n = 776) and used high-power computing to extract 942 variables from each participant’s data. Using a systematic, preregistered analysis pipeline, we failed to identify even a single EEG-based variable that could serve as a practically useful biomarker of autism clinical diagnosis. Our results highlight that a biomarker for autism drawn from resting state EEG data is an elusive construct that may not exist.