A Predictive Biomarker Model Using Quantitative Electroencephalography in Adolescent Major Depressive Disorder

Background: With evolving understanding of psychiatric diagnosis and treatment, demand for biomarkers for psychiatric disorders in children and adolescents has grown dramatically. This study utilized quantitative electroencephalography (qEEG) to develop a predictive model for adolescent major depressive disorder (MDD). We hypothesized that youth with MDD compared to healthy controls (HCs) could be differentiated using a singular logistic regression model that utilized qEEG data alone. Methods: qEEG data and psychometric measures were obtained in adolescents aged 14-17 years with MDD (n = 35) and age- and gender-matched HCs (n = 14). qEEG in four frequency bands (alpha, beta, theta, and delta) was collected and coherence, cross-correlation, and power data streams obtained. A two-stage analytical framework was then used to develop the final logistic regression model, which was then evaluated using a receiver-operating characteristic curve (ROC) analysis. Results: Within the initial analysis, six qEEG dyads (all coherence) had significant predictive values. Within the final biomarkers, just four predictors, including F3-C3 (R frontal) alpha coherence, P3-O1 (R parietal) theta coherence, CZ-PZ (central) beta coherence, and P8-O2 (L parietal occipital) theta power were used in the final model, which yielded an ROC area of 0.8226. Conclusions: We replicated our previous findings of qEEG differences between adolescents and HCs and successfully developed a single-value predictive model with a robust ROC area. Furthermore, the brain areas involved in behavioral disinhibition and resting state/default mode networks were again shown to be involved in the observed differences. Thus, qEEG appears to be a potential low-cost and effective intermediate biomarker for MDD in youth.