Cerebral microbleeds and hemorrhagic transformation improve late seizure prediction after ischemic stroke.
Abstract number :
3.31
Submission category :
4. Clinical Epilepsy / 4D. Prognosis
Year :
2025
Submission ID :
1048
Source :
www.aesnet.org
Presentation date :
12/8/2025 12:00:00 AM
Published date :
Authors :
Presenting Author: Alain Lekoubou Looti, MD, MS – Penn State University
Peter Ebasone, MD, PhD – Clinical Research Education Networking and Consultancy
Mariana Dejuk, BS – Penn State University
Bilal Siddiqui, BS – Penn State University
Maaz Ali, BS – Penn State University
Mahad Muhammad, BS – Penn State University
Syed Hussaini, BS – Penn State University
Jinpyo Hong, BS – Penn State University
Yunting Yu, BS – Penn State University
andre Pascal Kengne, MD, PhD – Clinical Research Education Networking and Consultancy
Vernon Chinchilli, MD, PhD – Penn State University
Bethany Jenkins, PhD – Penn State University
Rationale: Stroke accounts for a substantial proportion of epilepsy etiologies. The presence of blood products in the setting of ischemic stroke may increase the risk of late seizures. However, Models to separate patients with the highest risk of developing late seizures after ischemic stroke have not accounted for this important pathophysiologic aspect. We aimed to validate and update an existing model to predict late seizures one year after ischemic stroke.
Methods: Patients with imaging evidence of supra-tentorial (middle cerebral artery, anterior cerebral artery, or posterior cerebral artery) ischemic strokes at Penn State University and available follow-up at 1-year post-stroke were included in the current analysis. Incident late seizures were identified as reported by a physician (either as recorded in the medical record or any “yes” to the question “Have you ever been told by your doctor that you have seizures or epilepsy since your stroke?”). We applied a multivariable Cox proportional hazards model using 1) the predictors of the original SeLECT model, i.e., severity of stroke, large-artery atherosclerotic etiology-LAD, early seizures, cortical involvement, and territory of middle cerebral artery involvement-MCA, then replacing LAD and MCA territory by cerebral microbleeds and parenchymal hemorrhage. We then evaluated the performance of the original SeLECT model and the updated model (replacing LAD and MCA by cerebral microbleed defined as ≥10 microbleeds with ≥1 cortical microbleed and parenchymal hemorrhage) in our cohort of 963 ischemic stroke patients by estimating the discrimination and calibration parameters.
Results: Data on late seizures in one year were available in 99.9% of participants. At one year, 47 patients (4.88%) developed seizures. In the original SeLECT model, Hazard ratio (HR) and 95% confidence intervals were as follows: Stroke severity (NIHSS 4-10 vs. ≤4: 1.86, 0.82-4.32; NIHSS ≥11 vs. ≤4: 1.91, 0.87-4.22), early seizure (3.27: 1.14-9.36), cortical involvement (4.76, 1.13-20.1), large artery atherosclerosis (1.12, 0.6-2.09), and middle cerebral artery (2.06, 0.62-6.81). In the updated model, equivalent figures were as follows: Stroke severity (NIHSS 4-10 vs. ≤4: 1.86,18 0.83-4.18; NIHSS ≥ vs. ≤4: 1.64, 0.75-3.60), early seizure (3.32: 1.12, 9.85), cortical involvement (4.76, 1.21-21.7), cerebral microbleed (4.82, 1.43-16.2), and parenchymal hemorrhage (5.20, 2.19-12.3). The C-statistics for the original SeLECT model and the updated model were 0.683 (95% CI: 0.619-0.751) and 0.732 (0.664-0.801), respectively. The calibration slope for the original SeLECT model and the updated SeLECT models were respectively 0.739 and 0.876, indicating acceptable calibrations.
Conclusions: By replacing LAD and MCA territory with cerebral microbleed and parenchymal hemorrhage, we found that the updated SeLECT score had the potential to improve seizure prediction at one year in a selected US population. External validation of our model is warranted.
Funding: Sergievsky Award for Epilepsy Health Equity and Diversity from the American Epilepsy Society (AES)
Clinical Epilepsy