Delirium is associated with frequency band specific dysconnectivity in intrinsic connectivity networks: preliminary...

Pathophysiological concepts in delirium are not sufficient to define objective biomarkers suited to improve clinical approaches. Advances in neuroimaging have revalued electroencephalography (EEG) as a tool to assess oscillatory network activity in neuropsychiatric disease. Yet, research in the field is limited to small populations and largely confined to postoperative delirium, which impedes generalizability of findings and planning of prospective studies in other populations. This study aimed to assess effect sizes of connectivity measures in a large mixed population to demonstrate that there are measurable EEG differences between delirium and control patients.

This retrospective pilot study investigated EEG measures as biomarkers in delirium using a case-control design including patients diagnosed with delirium (DSM-5 criteria) and age-/gender-matched controls drawn from a database of 9980 patients (n = 129 and 414, respectively). Assessors were not blinded for groups. Power spectra and connectivity estimates, using the weighted phase log index, of continuous EEG data were compared between conditions. Alterations of information flow through nodes of intrinsic connectivity networks (ICN; default mode, salience, and executive control network) were evaluated in source space using betweenness centrality. This was done frequency specific and network nodes were defined by the multimodal human cerebral cortex parcellation based on human connectome project data.

Delirium and control patients exhibited distinct EEG power, connectivity, and network characteristics (F _(72,540) = 70.3, p < .001; F _(493,1079) = 2.69, p < .001; and F _(718,2159) = 1.14, p = .007, respectively). Connectivity analyses revealed global alpha and regional beta band disconnectivity that was accompanied by theta band hyperconnectivity in delirious patients. Source and network analyses yielded that these changes are not specific to single intrinsic connectivity networks but affect multiple nodes of networks engaged in level of consciousness, attention, working memory, executive control, and salience detection. Effect sizes were medium to strong in this mixed population of delirious patients.

We quantified effect sizes for EEG connectivity and network analyses to be expected in delirium. This study implicates that theta band hyperconnectivity and alpha band disconnectivity may be essential mechanisms in the pathophysiology of delirium. Upcoming prospective studies will build upon these results and evaluate the clinical utility of identified EEG measures as therapeutic and prognostic biomarkers.