IsCHEMiA in Vascular Epilepsy: Identifying Risks for Post Stroke Epilepsy

Commentary

The prevalence of poststroke survivors is increasing with modern stroke medicine and reperfusion therapy. ¹ Patients may recover from the acute stroke, make it through stroke rehabilitation, but then are hit with recurrent epileptic seizures after a latent period.

This is post-stroke epilepsy (PSE) a.k.a. vascular epilepsy and is defined as unprovoked seizures >7 days after the acute stroke, and has a prevalence of up to 8% at 5 years.^2,3

PSE is associated with significantly elevated morbidity and mortality. ⁴ There is a pressing need to identify those patients at highest risk to allow for potential interventions that may prevent PSE.

The authors used data from a prospective stroke registry at the Massachusetts General Hospital in Boston between 2016 and 2018 of 1436 adults admitted with acute ischemic stroke with an identifiable infarct on neuroimaging. A retrospective review of records was then performed to the end of study date of any post-stroke seizures or status epilepticus. PSE occurred in 218 patients (5.5%) over a median follow-up of 68 months. ⁵

Patients with previous epilepsy or on anti-seizure medications were excluded, including those who were put an antiseizure medication after a single early seizure.

The authors identified potential predictors of PSE based on a literature review to June 2024. Among others, five new variables not tested in existing scores were infarct size by volume, hemorrhagic transformation, age <65 years, anterior cerebral artery involvement, and thrombectomy.

Univariate analysis was followed by multivariate analysis using competing risk regression with all-cause mortality as a competing event. Competing risk regression models the probability of an event of interest occurring in the presence of events that prevent it, in this case early mortality, so reducing selection bias for severe stroke.

The model calculates a subdistribution hazard ratio (sHR) for each variable; values >1 increasing the cumulative incidence of an event.

The six variables significant in the multivariate model and included in the score (and the acronym) were Infarct size >5 cm (Is)—sHR 4.8, cortical involvement—sHR 2.1 (C), hemorrhagic transformation—sHR 4.3 (H), early seizure—sHR 4.6 (E), MCA involvement—sHR 2.4 (Mi), and age < 65 years—sHR 1.8 (A). Variables with a sHR <2 were given one point, and over 2 were given two points, adding up to a maximum of 9 points. The c-statistic, which measures the ability of a model to predict an outcome (a c-statistic of >0.8 is very good), was 0.87 in the US cohort and 0.84 overall.

Prediction estimates then showed that an IsCHEMiA score of ≥8 conferred a 67% risk at one year, and a 78% risk at 5 years. Scores less than 8 fell well below this, although longer follow-up may further elevate risk. An early seizure was a red flag for PSE if most other criteria are met.

The model was validated in two stroke center cohorts of 2534 patients in Hong Kong and Japan with similar patient characteristics and a median follow-up of 27–30 months. Public source data from the study is available for research.

Why should we use predictive models, and do we need a new one for PSE? The authors directly compared their model with the already established SeLECTS^2,6

Predictive models can allow uniform processing of an often complex collection of clinical variables to inform the physician on risk. One may argue that this is too rigid and that a clinician's expert view is more important. However, even the most experienced clinician could have unavoidable base rate neglect or heuristic bias—where we unconsciously ignore statistical information in favor of our intuition or experience of vivid cases. There are also limits to how many variables we can mentally juggle at one time. Many stroke physicians (usually neurologists in the USA but in many other countries may not be) may not have epileptology expertise—an objective tool based on clinical evidence may be a valuable asset, flagging higher risk patients for closer monitoring, an opinion my own non-neurology stroke colleagues strongly support.

Are there advantages of IsCHEMiA over SeLECTS? In IsCHEMiA, there is an emphasis on infarct size rather than NIHSS for stroke severity, which makes sense given a deep lacunar stroke with a low seizure risk will have a high NIHSS. The IsCHEMiA score is independent of stroke etiology, and can be done on clinical and imaging criteria alone. The inclusion of hemorrhagic transformation is fitting, given the accepted epileptogenicity of blood products (5% of patients in SeLECTS had “secondary hemorrhage” on imaging, but did not reach significance in their model). IsCHEMiA may be more relevant for US and Asian populations whereas SeLECTS was European-based. IsCHEMIA is from a more contemporary period, whereas SeLECTS being based on data collected between 2002 and 2008. IsCHEMiA had >500 patients under age 65 years in the US cohort, whereas SeLECTS had an older demographic so IsCHEMiA may be favored in younger age groups. SeLECTs had a shorter median follow-up of 28 months, and a lower c-statistic of 0.78.

However, the success of a tool depends on simplicity, accessibility, usability, and acceptance within target user groups. The SeLECTS tool has been around since 2018 and has its own website and app so it remains to be seen whether IsCHEMIA might displace it.

There are likely other relevant factors or biomarkers that could be incorporated in future models^1,7 or that are not readily available in many centers such as prolonged EEG. ⁸

The model is reassuring that most patients do not require long term anti-seizure medication. However, the risk is not zero even in low risk groups so we should still be vigilant for late seizures.

It is still early to tell whether predictive tools like IsCHEMiA or SeLECTS improves clinical outcomes in PSE.

How could a general stroke physician use the IsCHEMiA score? At the conclusion of the initial stroke assessment week, calculate the IsCHEMiA score. Magnetic susceptibility enhanced imaging should be done to identify hemosiderin in the post stroke period, particularly in patients at higher risk. A high IsCHEMiA score could signal a need for further testing such as with EEG or closer follow-up in the neurology clinic.

While it is currently not an accepted practice to use antiseizure medication for primary prophylaxis of PSE, a score of ≥8 may be high enough consider ASM, leading to a more personalized approach in the highest risk patients. A randomized controlled trial of anti-seizure medication after stroke with a high IsCHEMiA score to delay or prevent PSE is needed.

Of course, prevention of PSE starts with reducing stroke by vascular risk factor modification.