Conditioning on the Brain: Electroencephalography Derived Non-Linear Indices of Brain Dynamics and the Association with Oral Ketamine Treatment Response

Jules Mitchell

doi:10.25907/00974

Low-dose ketamine shows promise as a novel therapeutic for treatment resistant Major Depressive Disorder (MDD) and suicidality, with rapid clinical improvements observed within 24-hours following single or repeated doses. However, average response rates vary considerably (30-70%) and effects are often not sustained long-term. This variability underscores the need for biomarkers to guide treatment and better understand the brain dynamic changes in responders and non-responders over the short and long-term. The value of neuroimaging biomarkers for optimising treatment outcomes, such as those obtained from Magnetic Resonance Imaging (MRI) and Electroencephalography (EEG), has been demonstrated for antidepressants and transcranial magnetic stimulation, with emerging evidence for biomarkers associated with intravenous ketamine response. To date, research into oral ketamine biomarkers has been limited, with most studies utilising MRI. In contrast, EEG, given its low cost, portability, and millisecond-level temporal resolution, offers a practical option for identifying biomarkers that can guide ketamine treatment in everyday clinical practice. This thesis draws on predictive processing and dynamic systems theory as core conceptual and methodological frameworks to investigate whether pre-treatment EEG dynamics (i.e., initial conditions) and treatment associated changes (i.e., trajectory) differed between individuals who went on to achieve significant clinical improvement at the post-treatment timepoint (responders) as compared to those who did not (non-responders) during the open-label Oral Ketamine Trial on Suicidality (OKTOS). Preceding the empirical chapters (IV-VI), chapter I outlines the current challenges facing psychiatry, and potential solutions, chapter II sets out the theoretical and empirical framework, and chapter III details the study design, analytical and statistical methods applied herein. OKTOS participants presented with treatment resistant MDD and a self-reported Beck Scale for Suicide Ideation (BSS) total score ≥ 6. They received sub-anaesthetic titrated doses (0.5-3.0 milligram (mg)/ kilogram (kg)) of racemic oral ketamine once per week for six weeks. Resting-state (eyes-open and closed) and task-based (AX-Continuous Performance Task) EEG was collected at three timepoints: pre-treatment (Baseline; BAS), post-treatment (POST; week 6), and at follow-up (FUP; week 10). These data were analysed using non-linear measures of activity (complexity), effective connectivity (multivariate transfer entropy (mTE)), and network topology (graph theory). The overarching hypothesis was that EEG-derived non-linear indices would differ between responders and non-responders prior to and following treatment. Utilising Bayesian mixed effect modelling, I identify distinct BAS brain dynamics in those who went on responders at POST (N = 22; 42.6 years (SD = 13.8), 50% female) compared to those who went on to be non-responders (N = 9; 51.4 years (SD = 13.5), 55% female)). Study one (Chapter IV) examined the complexity of eyes-closed and open resting-state EEG dynamics using Lempel-Ziv complexity (LZC) and multi-scale entropy (MSE). At BAS, those who went on to be responders at POST had greater eyes-open complexity. A post-hoc exploratory channel-level analysis revealed this was localised to four left-frontal channels (Fp1, AF3, F3, FC1). From BAS to POST, LZC at these four channels decreased for responders, but increased for non-responders. Study two (Chapter V) examined the network effective connectivity (mTE) and topology (global and local graph theory metrics) of eyes-closed and open resting-state EEG to determine if the observed elevation in left-frontal complexity was indicative of hyper-function or dys-function. At BAS, those who went on to be responders at POST displayed spatially non-uniform increases (compared to non-responders) in clustering coefficient, in-degree and out-degree across the eyes-closed and open conditions, which was predominantly localised to left-side fronto-central and parieto-occipital regions. Notably, these channels included those with elevated BAS LZC from study one. Oral ketamine differentially impacted these metrics in responders and non-responders, increasing for non-responders following treatment, whilst decreasing for responders. Study three (Chapter VI) examined the complexity (sample entropy), network effective connectivity (mTE) and topology (global and local graph theory metrics) during the AX-CPT A-B trial, which indexes inhibitory and attentional control. Informed by the resting-state findings, it was hypothesised that left-fronto-parieto-occipital channels would similarly dominate information processing at BAS in responders when under cognitive task, as indicated by increased complexity, global efficiency, clustering coefficient, in- and out-degree. At BAS, those who went on to be non-responders at POST displayed spatially non-uniform increases in sample entropy at midline and para-sagittal frontal, central, and occipital channels, and greater Fz P3a mean amplitude, suggesting they required more processing to inhibit their response. Oral ketamine differentially impacted sample entropy, increasing it for non-responders from BAS to POST at primarily fronto-central and parieto-occipital channels, yet decreasing it for responders from BAS to POST at primarily fronto-central and parieto-occipital channels. This thesis demonstrates the potential of EED-derived biomarkers and non-linear analysis methods in predicting oral ketamine treatment response, highlighting unique brain activity and connectivity dynamics for different brain states. More broadly, these findings reinforce the idea that pre-treatment brain dynamics shape oral ketamine treatment outcomes; aligning with the theory that neuropsychiatric dysfunction is an epiphenomenon of the brain’s information processing dynamics, and, consequently the nature of these dynamics should influence the therapeutic response. As the first study to investigate oral ketamine’s interaction with non-linear EEG-derived measure of brain dynamics, these findings represent a significant step-towards personalised mental health treatment, wherein, individuals are stratified to the treatment option with the greatest chances of symptom improvement. In the case of oral ketamine, activity and connectivity dynamics at fronto-parieto-occipital channels may prove particularly useful in this effort, as these channels repeatedly differentiated those who went on to be responders at POST across the resting and task-based EEG. Whether a similar profile of BAS brain dynamics differentiates those who went on to be prolonged (i.e., FUP) responders remains an open question for future research. Furthermore, out-of-sample applications of the methods applied herein are needed to scrutinise the generality, reproducibility, validity and clinical applicability of these findings. In conclusion, this thesis identifies distinct EEG-derived biomarkers that distinguish oral ketamine treatment responders and non-responders, offering insights into how initial brain dynamics influence therapeutic outcomes.

Conditioning on the Brain: Electroencephalography Derived Non-Linear Indices of Brain Dynamics and the Association with Oral Ketamine Treatment Response

Files and links (1)

Abstract

Details

Metrics