Discriminant Diagnostic Quality of Resting-State EEG for Fragile X Syndrome, Autism Spectrum Disorders, and Typical Development

Background: Fragile X chromosome syndrome (FXS) is a neurodevelopmental condition, caused by a mutation of the FMR1 gene. It is the second most common single-gene cause of inherited intellectual disability and the most common genetic cause of autism. While similar in several behavioral manifestations and probably neurobiologically related, FXS is different from idiopathic autism spectrum disorders (ASD) in intervention strategies, prognosis, and relevance of genetic counselling. A reliable and accessible biomarker, such as EEG, could enable timely differential diagnosis. The findings of the resting-state EEG changes in FXS are relatively consistent: most authors report elevated theta power, reduced alpha power as well as increased epileptiform abnormalities. Some EEG changes in ASD follow the same pattern but in general they are more variable, probably reflecting the higher heterogeneity of autism. Most resting-state EEG studies have had limitations associated with the experimental sample size.

Objectives: The spectral resting-state EEG data were used to determine typical changes in FXS and ASD in larger samples and individual spectral features with good discriminant properties were chosen to predict assignment of subject to FXS or ASD groups based on EEG.

Methods: 47 children with FXS (age 4 to 18) and 51 children with ASD (age 3 to 18) participated in the study. All subjects with FXS had a full FMR1 mutation; the subjects with ASD had a clinical diagnosis within the autism spectrum (F84.0, F84.1 or F84.5) according to ICD-10, they also scored 15 or higher on Social Communication Questionnaire. The resting-state EEG (with closed eyes) was collected from 14 standard 10/20% electrode locations with A1+A2 reference and recorded with a Neuro-KM EEG system. Spectral power was calculated for standard and narrow – 1, 1.5 and 2 Hz – frequency bands. The values were compared to 70 EEG records of typically developing (TD) children from the normative database of 700 TD records organized in one-year age groups. Linear discriminators were chosen with forward stepwise analysis among normalized power, relative power, coherency, asymmetry, and power ratio predictors. Linear discriminant functions were identified with leave-one-out cross-validation method.

Results: The FXS showed a particular pattern of EEG changes with higher theta and beta-2 power and lower alpha power (p<0.01 FXS vs.TD). In ASD alpha power was lower and beta power was higher (p<0.01 ASD vs.TD). Models for optimal discrimination of groups were identified. They included four predictors for the FXS-TD comparison, three predictors for the FXS-ASD comparison, and four predictors for ASD-TD comparison. Sensitivity ranged from 90% for FXS vs.ASD to 100% for FXS vs.TD (93% for ASD vs. TD); sensitivity ranged from 90% for FXS vs.ASD and ASD vs.TD to 92% for FXS vs.TD.

Conclusions: We identified the distinct pattern of resting-state EEG changes in FXS, manifesting in higher theta and beta 2 power and lower alpha power, in a larger sample of children. Discriminant analysis assigns subjects to ASD, FXS, and TD groups with excellent specificity and sensitivity.