The Role of Electroencephalography in Predicting Post-Stroke Seizures and an Updated Prognostic Model (SeLECT-EEG)

Abstract

Objective: Seizures negatively impact stroke outcomes, highlighting the need for reliable predictors of post-stroke epilepsy. Although acute symptomatic seizures are a known risk factor, most stroke survivors who develop epilepsy do not experience them. Early electroencephalography (EEG) findings may enhance risk prediction, particularly in patients without acute symptomatic seizures, aiding in patient management and counseling.

Methods: We conducted a multicenter cohort study using data from 1,105 stroke survivors (mean age 71 years, 54% male) with neuroimaging-confirmed ischemic stroke who underwent EEG within 7 days post-stroke. Electrographic biomarkers, including epileptiform activity and regional slowing, were analyzed for their association with post-stroke epilepsy using Cox proportional hazards regression and Fine-Gray subdistribution hazard models, adjusted for differences in EEG timing and patient characteristics.

Results: Post-stroke epilepsy developed in 119 patients (11%), whereas 233 (21%) had acute symptomatic seizures. The 5-year epilepsy risk was 42% (95% confidence interval [CI]: 30-49%) in patients with epileptiform activity versus 13% (95% CI: 9-16%) in those without. Regional slowing doubled the 5-year epilepsy risk (23%, 95% CI: 17-30% vs 11%, 95% CI: 7-16%). Epileptiform activity (subdistribution hazard ratio: 2.3, 95% CI: 1.5-3.4, p < 0.001) and regional slowing (subdistribution hazard ratio: 1.7, 95% CI: 1.1-2.7, p = 0.02) were independently associated with post-stroke epilepsy. A novel prognostic model, SeLECT-EEG (concordance statistic: 0.75, 95% CI: 0.71-0.80), outperformed the previous standard (SeLECT_2.0; 0.71, 95% CI: 0.65-0.76, p < 0.001).

Interpretation: Electrographic biomarkers improve post-stroke epilepsy prediction beyond clinical risk factors. The SeLECT-EEG model enhances early risk stratification, particularly in patients without acute symptomatic seizures, informing management strategies and patient counseling. ANN NEUROL 2025;98:814-825.

FIGURE 1

Forest plot of clinical and electrographic findings associated with post‐stroke epilepsy risk. This forest plot illustrates the hazard ratio (HR) and 95% confidence interval (CI) from Table 1 for various clinical and electrographic findings in relation to the development of post‐stroke epilepsy (PSE). Acute symptomatic seizures (ASyS) are shown alongside electrographic biomarkers such as regional slowing, interictal epileptiform discharges (IED), lateralized periodic discharges (LPD), and electrographic seizures (ESz). Each of these variables significantly increases the risk of PSE, with odds ratios (ORs) ranging from 2.6 to 6.6. Interestingly, ESz exhibited an HR comparable to clinical seizures (ASyS). [Color figure can be viewed at www.annalsofneurology.org]

FIGURE 2

Risk of epilepsy in individuals with or without regional slowing or any epileptiform activity. Cumulative incidence functions for the time to post‐stroke epilepsy are presented, comparing individuals with any epileptiform activity (A) or regional slowing (B). Separate survival curves are illustrated for those with (blue) and without (yellow) epileptiform activity (A) or regional slowing (B). After adjusting for age, sex, National Institutes of Health Stroke Scale (NIHSS) score at admission, cortical involvement, middle cerebral artery (MCA) territory involvement, large artery atherosclerosis, and generalized slowing, individuals exhibiting epileptiform activity or regional slowing demonstrated a significantly higher risk of developing post‐stroke epilepsy, indicating a higher likelihood of remote symptomatic seizures compared to those without these conditions (epileptiform activity: p < 0.001; regional slowing: p = 0.02). [Color figure can be viewed at www.annalsofneurology.org]

FIGURE 3

Predicted risk of post‐stroke epilepsy according to a new prognostic model in stroke survivors without acute symptomatic seizures including electrographic findings. A illustrates the predicted risk of unprovoked remote symptomatic seizures occurring 0–120 months post‐stroke. The red curves represent different SeLECT‐EEG values for stroke survivors including electrographic findings. Each curve corresponds to SeLECT‐EEG values ranging from 0 to 8 (in red). B and C present risk estimate charts for late seizures at 1 year and 10 years post‐stroke, respectively, based on the SeLECT‐EEG. Vertical lines denote the 95% confidence intervals. [Color figure can be viewed at www.annalsofneurology.org]