2024 American Academy of Neurology Abstract Website

Objective:

To identify new clinical and magnetic resonance imaging (MRI) characteristics predictive of post-stroke epilepsy (PSE).

Background:

Clinical prediction tools for PSE (e.g. SeLECT score) are largely derived from clinical and computed tomography (CT) parameters. With the increasing use of MRI, we aimed to identify new clinical and MRI characteristics to develop a novel risk score for PSE.

Design/Methods:

This was a single-centre, retrospective study, consisting of 852 consecutive patients admitted to Queen Mary Hospital for acute ischemic stroke from August 2018-December 2020. All patients received a CT brain, whilst 70.1% had an MRI brain. We determined the risk associations of individual components of SeLECT (Severity, Large-artery atherosclerosis, Early seizures (≤7 days post-stroke), Cortical, Middle cerebral artery (MCA) Territory) with PSE, and further included new parameters including infarct size and haemorrhagic transformation (HT). Predictive values were illustrated by calculating area-under-curve (AUC) of receiver operating characteristic (ROC) curves.

Results:

At median follow-up of 893 days, PSE occurred in 5.3% of patients (mean time-to-first seizure: 311.2±367.0 days). Severe stroke (NIHSS ≥11), MCA territory involvement, cortical involvement, and large infarct size (≥5cm or ≥1/3 of MCA territory) were associated with PSE.

Multivariate regression further showed that HT of any severity was independently associated with PSE (sHR 4.72, 95% CI 2.36-9.41, p<0.001). “H-SeLECT” was obtained by adding 2 marks to SeLECT for presence of any HT, which was calculated by dividing its sub-distribution hazard ratio (6.99) with the median of the lowest three values (3.43) in univariate analysis and rounding to the nearest integer. The AUC of ROC curve of H-SeLECT was significantly increased when compared with SeLECT using Z test (0.813 vs. 0.774, Z=2.67; p=0.008).

Conclusions:

The newly proposed H-SeLECT score significantly increases the predictive value in detecting PSE compared to existing risk scores. Further studies are required to externally validate its utility beyond our local population.