Brain causality alterations in major depressive disorder treatment

Major depressive disorder (MDD) remains a leading cause of disability worldwide, with limited objective biomarkers to guide treatment selection and monitor early therapeutic effects. This study examines whether pharmacological and neurostimulation treatments produce distinct alterations in directional brain connectivity patterns within the first week of treatment and whether baseline connectivity patterns differ between treatment responders and nonresponders assessed at 4-6 weeks.

We perform an information theory-based causality analysis on instantaneous phase time series derived from electroencephalography recordings of 176 MDD patients. Patients received either pharmacological treatment or neurostimulation and were recorded at baseline (visit 1) and one week post-treatment initiation (visit 2). We quantified directional information flow across 19 EEG channels in five frequency bands ( $\delta$ : 1.5-3.5Hz, $\theta$ : 4-8Hz, $\alpha$ : 8-12Hz, ${\beta }_1$ : 12.5-17.5Hz, ${\beta }_2$ : 18-25.5Hz) using global FLOW metrics and local INFLOW/OUTFLOW metrics per channel.

Treatment modalities produced distinct frequency-specific alterations in brain causality. Pharmacological treatment significantly increased global and local information transmission in the β ₂ band from visit 1 to visit 2, with widespread effects across brain regions and larger increases in right hemisphere inflow. Neurostimulation treatment increased information flow primarily in the δ band, with the strongest effects originating from left hemisphere to the whole brain. At baseline, pharmacological nonresponders exhibited significantly higher α band information flow than responders, particularly for right-to-left hemisphere transmission and bilateral inflow. No significant baseline differences emerged between neurostimulation responders and nonresponders, though large effect sizes in δ band metrics suggest findings may achieve significance with larger samples.

Early-phase directional brain connectivity analysis reveals that pharmacological and neurostimulation treatments engage distinct neural oscillatory mechanisms within the first week. Elevated baseline right-dominant α-band causal transmission identifies patients unlikely to respond to pharmacological treatment, enabling alternative interventions before weeks of ineffective therapy. These phase-based causality metrics offer promising biomarkers for early treatment stratification and monitoring. The ability to distinguish treatment responders from nonresponders at baseline and detect treatment-specific changes within one week, before clinical response manifestation, suggests these causality measures capture early mechanistic alterations that could inform timely treatment optimization decisions. (EUDRACT Nos. 2005-000826-22 and 2015-001639-19, registered via www.clinicaltrialsregister.eu).