Biological Psychiatry, 2020, 87(9), S402-S403. [Presented at the 75th Annual Meeting of the Society of Biological Psychiatry (SoBP) in New York, NY, April 30 - May 2, 2020.]
Identifying a frontal theta network that differentially predicts outcomes of treatment with SSRI or placebo in depression: A CSD-PCA approach to analyze sensor-level EEG connectivity
Ezra E. Smith1, Diego A. Pizzagalli2, Patricia J. Deldin3, Melvin G. McInnis3, Madhukar H. Trivedi4, Myrna M. Weissman1,5, Gerard E. Bruder5, Jürgen Kayser1,5
1New York State Psychiatric Institute; 2McLean Hospital/Harvard Medical School; 3University of Michigan; 4University of Texas Southwestern Dallas; 5Columbia University, New York, New York, USA
Abstract
Background: Improving patient-treatment matching can relieve patient suffering, lower healthcare costs, and also reveal psychiatric subtypes. Methods: Resting pretreatment EEG was obtained in a multisite randomized clinical trial (EMBARC) from 212 patients with major depressive disorder (MDD) and 35 healthy controls. Patients were administered placebo or the selective serotonin reuptake inhibitor (SSRI) sertraline for eight weeks. We assessed functional connectivity (EEG synchronization via debiased weighted phase lag index) in theta and alpha frequency range between 69 recording sites as hypothesized predictors of posttreatment depression symptoms. EEG epochs were transformed to current source density (CSD) estimates representing reference-free radial current flow at scalp with improved spatial resolution. Two-step principal components analysis (PCA) was used to systematically reduce the high-dimensionality of functional connectivity graphs into distinct spectral-spatial connectivity components. Results: Several high-variance connectivity patterns were extracted with theta and alpha frequency peaks involving frontal and/or posterior regions. These connectivity components were highly consistent between patients and controls. A priori hypotheses regarding theta and alpha spectra and their regional specificity as predictors of antidepressant response guided component selection for further analysis. A midfrontal theta connectivity component was differentially related to treatment outcome (linear regression; p=.009). Strong pretreatment theta connections predicted symptom improvement for SSRI but deterioration for placebo. Moreover, this theta component interacted with two posterior alpha connectivity components (p=.03) that accounted for additional variance of treatment outcomes. Conclusions: Results suggest that parallel analyses of theta and alpha networks may help to improve prediction of MDD treatment outcomes and optimize treatment selection.
Key Words: Depression; EEG Theta/Alpha; Functional Connectivity; Treatment Response; CSD-PCA
[Supported by National Institute of Mental Health (MH115299; MH092221; MH092250)].
doi:10.1016/j.biopsych.2020.02.1028