Quantitative EEG-Based Detection of Poststroke Delirium Endotypes and Their Association With Biomarkers of Inflammation.

BACKGROUND AND PURPOSE: Poststroke delirium (PSD) is common and prognostically relevant, yet under-detected. Mechanistic and biomarker research is constrained by reliance on an intermittently observed binary phenotype, while delirium reflects one severity level on a continuum of delirium-related encephalopathy. Quantitative EEG (qEEG) may capture encephalopathy endotypes more directly and enable severity-spectrum characterization. METHODS: In this prospective, single-center, observational cohort study, 87 consecutive patients with acute ischemic stroke or transient ischemic attack were assessed within 48 h using the Confusion Assessment Method (CAM). A 64-channel EEG was recorded at enrollment; spectral power (delta/theta/alpha/beta) and functional connectivity metrics (phase lag index; amplitude envelope correlation corrected [AECc]) were computed. Neuroinflammatory and systemic biomarkers were quantified from routine serum sampling in an exploratory, add-on subcohort (n = 31). RESULTS: PSD occurred in 28 of 87 (32%). PSD was characterized by spectral slowing and altered AECc. A multivariable qEEG model discriminated PSD with AUC = 0.892 (p < 0.001) and overall accuracy of 81.4% (non-delirium 89.8%, delirium 63.0%). In the paired EEG plus serum subset, nominal exploratory correlations between qEEG metrics and selected biomarkers suggested links between network dysfunction and inflammatory signaling, including inverse correlations of theta-band AECc with VILIP-1 (r = -0.454, p = 0.045) and CX3CL1 (r = -0.604, p = 0.005), whereas biomarker-phenotype associations were less consistent. CONCLUSIONS: qEEG connectivity provides an encephalopathy-proximal readout that can detect PSD and may characterize delirium-related endotypes beyond the intermittently observed clinical phenotype. Findings require validation in larger, multicenter cohorts.