Post-stroke seizure risk prediction models: a systematic review and meta-analysis

Area of Science:

• Neurology
• Epidemiology
• Clinical Prediction Models

Background:

• Epileptic seizures are a common complication of stroke, particularly in older adults.
• While risk factors for post-stroke seizures (PSS) are known, accurate prediction remains a clinical challenge.
• Developing reliable PSS risk prediction models is crucial for personalized patient management.

Purpose of the Study:

• To systematically review and evaluate the predictive accuracy of existing PSS risk prediction models.
• To identify the most accurate models for predicting PSS in different stroke types (ischemic and hemorrhagic).

Main Methods:

• A systematic review and meta-analysis of studies from MEDLINE and EMBASE (inception to December 2020).
• Inclusion of peer-reviewed articles on PSS risk prediction model development or validation.
• Utilized random-effects meta-analysis and receiver operating characteristic curves; quality appraisal with PROBAST.

Main Results:

• 13 studies with 182,673 patients reported 15 PSS risk prediction models.
• Incidence of early PSS (≤1 week) was 4.5%; late PSS (>1 week) was 2.1%.
• SeLECT (AUC 0.77) and CAVE (AUC 0.81) showed the highest predictive accuracy for late PSS in ischemic and hemorrhagic stroke, respectively. However, 14/15 studies had high risk of bias due to lack of validation and reporting.

Conclusions:

• Few validated, low-bias multivariate models exist for predicting individual PSS risk.
• Accurate models are needed to personalize clinical management and identify high-risk patients for anti-epileptic drug prophylaxis trials.
• Further research is required to develop and validate robust PSS prediction tools.