Predicting antidepressant responsiveness in major depressive disorder patients via electroencephalography gamma-band dynamic functional connectivity in response to salient auditory stimuli.

Background: Heterogeneous pathophysiological characteristics in patients with major depressive disorder (MDD) lead to individually differentiated sensitivities to antidepressants. Based on the hypothesis that gamma-band dynamic fluctuations in cortical functional connectivity (FC) in response to salient stimuli are linked to pathophysiological characteristics, we conducted a classification analysis for antidepressant responsiveness prediction.

Methods: Biosignals and psychological measures were acquired from 47 patients with MDD prior to treatment. After 8 weeks of antidepressant therapy, patients were divided into non-remitted MDD (nrMDD; aged 42.55 ± 11.52 years; n = 20) and remitted MDD (rMDD; aged 47.22 ± 11.59 years; n = 27) groups based on their depressive symptom reduction. Electroencephalography (EEG) signals were acquired during the duration-variant auditory mismatch negativity paradigm. From the deviant condition, gamma-band weighted phase-lag index-based dynamic fluctuations were evaluated using a template generated from 21 demography-matched healthy control (aged 43.81 ± 14.10 years) data.

Results: Using these dynamic functional connectivity (dFC) features, a machine learning-based classification analysis was performed for nrMDD and rMDD. Using leave-one-out cross-validation, the linear discriminant analysis classifier achieved the best accuracy (82.98%) for classifying nrMDD and rMDD. Further simple effect analyses identified three core dFC features for nrMDD: (i) relatively intact time-dependent FC between the left frontal and right temporal regions; (ii) disrupted right frontoparietal FC; and (iii) disrupted left fronto-temporal FC. These dFC features commonly exhibit transient hyperconnections in patients with nrMDD.

Conclusions: We demonstrated that gamma-band dFC responses to salient stimuli could serve as potential biomarkers for antidepressant responsiveness prediction in patients with MDD.