Blood biomarkers predictive of epilepsy after an acute stroke event.

Blood biomarkers have not been widely investigated in poststroke epilepsy. In this study, we aimed to describe clinical factors and biomarkers present during acute stroke and analyze their association with the development of epilepsy at long term. A panel of 14 blood biomarkers was evaluated in patients with ischemic and hemorrhagic stroke. Biomarkers were normalized and standardized using Z-scores. Stroke and epilepsy-related variables were also assessed: stroke severity, determined by National Institutes of Health Stroke Scale (NIHSS) score, stroke type and cause, time from stroke to onset of late seizures, and type of seizure. Multiple Cox regression models were used to identify clinical variables and biomarkers independently associated with epilepsy. From a cohort of 1115 patients, 895 patients were included. Mean ± standard deviation (SD) age was 72.0 ± 13.1 years, and 57.8% of patients were men. Fifty-one patients (5.7%) developed late seizures, with a median time to onset of 232 days (interquartile range [IQR] 86-491). NIHSS score ≥8 (P < .001, hazard ratio [HR] 4.013, 95% confidence interval [CI] 2.123-7.586) and a history of early onset seizures (P < .001, HR 4.038, 95% CI 1.802-9.045) were factors independently associated with a risk of developing epilepsy. Independent blood biomarkers predictive of epilepsy were high endostatin levels >1.203 (P = .046, HR 4.300, 95% CI 1.028-17.996) and low levels of heat shock 70 kDa protein-8 (Hsc70) <2.496 (P = .006, HR 3.795, 95% CI 1.476-9.760) and S100B <1.364 (P = .001, HR 2.955, 95% CI 1.534-5.491). The risk of epilepsy when these biomarkers were combined increased to 17%. The area under the receiver-operating characteristic (ROC) curve of the predictive model was stronger when clinical variables were combined with blood biomarkers (74.3%, 95% CI 65.2%-83.3%) than when they were used alone (68.9%, 95% CI 60.3%-77.6%). Downregulated S100B and Hsc70 and upregulated endostatin may assist in prediction of poststroke epilepsy and may provide additional information to clinical risk factors. In addition, these data are hypothesis-generating for the epileptogenic process.