Alpha Waves as a Neuromarker of Autism Spectrum Disorder: The Challenge of Reproducibility and Heterogeneity

Abstract

Background: There is no consensus in the literature concerning the presence of abnormal alpha wave profiles in patients with autism spectrum disorder (ASD). This may be due to phenotypic heterogeneity among patients as well as the limited sample sizes utilized. Here we present our results of alpha wave profile analysis based on a sample larger than most of those in the field, performed using a robust processing pipeline.Methods: We compared the alpha waves profiles at rest in children with ASD to those of age-, sex-, and IQ-matched control individuals. We used linear regression and non-parametric normative models using age as covariate forparsing the clinical heterogeneity. We explored the correlation between EEG profiles and the patient’s brain volumes, obtained from structural MRI. We automatized the detection of the alpha peak and visually quality controled our MRI measurements. We assessed the robustness of our results by running the EEG preprocessing with two different versions of Matlab as well as Python.Results: A simple linear regression between peak power or frequency of the alpha waves and the status or age of the participants did not allow to identify any statistically significant relationship. The non-parametric normative model (which took account the non-linear effect of age on the alpha profiles) suggested that participants with ASD displayed more variability than control participants for both frequency and amplitude of the alpha peak (p < 0.05). Independent of the status of the individual, we also observed weak associations (uncorrected p < 0.05) between the alpha frequency, and the volumes of several cortical and subcortical structures (in particular the striatum), but which did not survive correction for multiple testing and changed between analysis pelines.Discussions: Our study did not find evidence for abnormal alpha wave profiles in ASD. We propose, however, an analysis pipeline to perform standardized and automatized EEG analyses on large cohorts. These should help the community to address the challenge of clinical heterogeneity of ASD and to tackle the problems of reproducibility.

Autism spectrum disorder

“…Pioneering studies have revealed alterations in oscillatory activity during the resting state in people with autism, with more slow waves and less alpha waves, as well as less intrahemispheric and inter-hemispheric asymmetry compared to people without autism 94 . More recent work has demonstrated the presence of developmental trajectories as revealed through increasingly sophisticated spatio-spectral analyses, and has revealed how differences in the trajectories of EEG power in high-risk infants may represent an endophenotypes of autism 95,96 .…”

Toward Neurosubtypes in Autism

“…The present ASD neurosubtyping literature is in its infancy with a total of 12 studies in humans, 92% published since 2018 (60)(61)(62)(63)(64)(65)(66)(67)(68)(69)(70)(71). Studies vary in methodology (FIGURE 1, TABLE 1).…”

“…Although preliminary, initial results suggest that the identification of neurosubtypes may be promising to better explain brain-behavior relationships in regard to symptom severity (65) or diagnostic labels (62). Four studies used independent neuroimaging modalities (62,65,67,69). By validating neurosubtypes identified via regional structural features with specific brain connectivity metrics, these studies point towards neurosubtype-specific associations to both regional and large-scale brain organization.…”

“…The identification of the extreme cases in the distribution is used to find a subgroup of individuals that are more clearly separable from all others, and possibly more homogenous. To date, three ASD neuroimaging studies(66,69,70) have adopted this approach.Focusing on either cortical thickness(66,70) or alpha-wave EEG profiles deviation from normative curves derived from cross-sectional age data(69), their results showed that variability between ASD and NT is largely overlapping and only a subset of individuals shows a high degree of brain atypicalities. One disadvantage of this method is that it does not assume that there are multiple subgroups.…”

Normative age modelling of cortical thickness in autistic males

“…Normative modelling of age-related CT effects was done utilizing data from the typicallydeveloping group (TD) (see Figure 2 for a schematic overview). We used a local polynomial regression fitting procedure (LOESS) (Cleveland et al, 1988;Lefebvre et al, 2018), where the local width or smoothing kernel of the regression was determined by the model that provided the overall smallest sum of squared errors. Though we also assessed consistency of our output using centiles scoring (see Supplementary Materials).…”