ABSTRACT NUMBER: ESOC2026A2148 EARLY BRAIN NETWORK FAILURE IDENTIFIES STROKE PATIENTS AT HIGH RISK FOR POST STROKE DELIRIUM

Abstract Background and aims Delirium is a frequent and disabling complication in the acute phase after stroke, strongly associated with poor neurological and functional outcomes. Despite its clinical relevance, reliable early prediction remains challenging. Increasing evidence suggests that post-stroke delirium (PSD) reflects a breakdown of large-scale brain network dynamics rather than isolated lesion effects. Transcranial magnetic stimulation combined with electroencephalography (TMS–EEG) allows causal interrogation of cortical network responsiveness and information integration. Methods In this prospective observational study, we investigated whether the fast perturbational complexity index (PCIST), a quantitative measure of TMS-evoked cortical response complexity, predicts the subsequent development of PSD. Within 48 hours of stroke onset and prior to delirium emergence, TMS–EEG was applied over the dorsolateral prefrontal cortex and superior parietal lobule of both hemispheres in 34 acute stroke patients. Delirium was assessed every 8 hours using the ICDSC and DSM-5 criteria. Resting-state EEG measures, lesion volume, and NIHSS were included as covariates. Results Thirteen patients (38%) developed PSD. PCIST was significantly reduced across all stimulation sites in patients who later developed PSD (p < 0.001), independent of lesion laterality. A two-way ANOVA revealed a strong group effect without site interaction, indicating a global disruption of cortical network complexity. PCIST correlated with stroke severity and lesion volume but remained an independent predictor of PSD. The highest predictive accuracy was observed for ipsilesional parietal stimulation (AUC = 0.87). Conclusions Early global degradation of TMS-evoked brain network complexity precedes delirium and captures a systems-level vulnerability beyond focal injury. Conflict of interest Annerose Mengel, Gongfei Li, Katharina Feil, Maria Ioanna Stefanou, Lena Geiger-Primo, Ulf Ziemann has nothing to disclose in relationship to this project