Postanoxic electrographic status epilepticus and serum biomarkers of brain injury

Aim: To explore if electrographic status epilepticus (ESE) after cardiac arrest causes additional secondary brain injury reflected by serum levels of two novel biomarkers of brain injury: neurofilament light chain (NfL) originating from neurons and glial fibrillary acidic protein (GFAP) from glial cells. Methods: Simplified continuous EEG (cEEG) and serum levels of NfL and GFAP, sampled at 24, 48 and 72 hours after cardiac arrest. All data were collected during the Target Temperature Management (TTM)-trial. Patients with and without ESE were matched for early predictors of poor neurological outcome. Results : 128 patients had available biomarkers and cEEG. Twenty-six (20%) patients developed ESE, the majority (69%) within 24 hours. ESE was an independent predictor of elevated serum NfL (p<0.001) but not of serum GFAP (p=0.16) at 72 hours after cardiac arrest. Compared to a control group matched for early predictors of poor neurological outcome, patients who developed ESE had higher levels of serum NfL (p=0.03) and GFAP (p=0.04) at 72 hours after cardiac arrest. Conclusion : ESE after cardiac arrest is associated with higher levels of serum NfL which may suggest increased secondary neuronal injury compared to matched patients without ESE but similar initial brain injury. Associations with GFAP reflecting glial injury are less clear. The study design cannot exclude imperfect matching or other mechanisms of secondary brain injury contributing to the higher levels of biomarkers of brain injury seen in the patients with ESE. Sedation age; first monitored rhythm; cardiac arrest


Introduction
Electrographic status epilepticus (ESE) is found in up to one third of comatose survivors of cardiac arrest and is associated with a poor neurological outcome 1 . A minority of patients with ESE recover with a good neurological outcome 2 . In these patients ESE typically has an onset after rewarming and other neuroprognostic markers do not indicate extensive brain injury 3 .
Further classification of ESE depending on discharge frequency appears not to affect the prognostic value of ESE after cardiac arrest 4 .
After status epilepticus of other aetiologies, several biomarkers of neuronal injury are increased in cerebrospinal fluid 5 , a sampling method often contraindicated after cardiac arrest due to anticoagulation. Elevated levels of neuron-specific enolase (NSE) in serum has been found in status epilepticus of mixed pathophysiology 6 .
Whether postanoxic ESE is simply a marker of severe encephalopathy or the cause of further secondary brain injury is controversial. Active treatment of seizures is recommended 7 since epileptic activity has the potential to increase the metabolic demand 8,9 and thereby inflict additional brain injury. These recommendations are based on expert advice awaiting evidence from randomized trials 10 .
Neurofilament light chain (NfL) is a novel biomarker of neuronal injury and a predictor of poor outcome after cardiac arrest 11 . After neuronal injury, serum NfL levels rise rapidly 12 and remain elevated for prolonged periods (weeks) 13 . Unlike NSE, NfL levels are not falsely elevated by hemolysis. Glial fibrillary acidic protein (GFAP) is a marker of glial cell injury with prognostic value after cardiac arrest 14 . Serum GFAP rises rapidly after cardiac arrest 15 and its half-life is long, up to 48 hours 13 .
The aim of the present study was to explore if serum levels of NfL and GFAP can provide any evidence whether or not postanoxic ESE causes additional secondary brain injury. Our hypotheses were: 1. ESE is an independent predictor of serum levels of NfL and GFAP at 72 hours after cardiac arrest.
2. After onset of ESE, patients have higher levels of serum NfL and GFAP compared to a control group without ESE, matched for early predictors of poor neurological outcome.

Methods
This study uses data collected during the Target Temperature Management after Out-of-Hospital Cardiac Arrest trial (2010-2013) 16,17 . Strict criteria for withdrawal of life-sustaining therapies (WLST) were applied. Sedation was mandatory during targeted temperature management (TTM) and continued after the intervention when indicated for medical reasons.
Choice of sedative and antiepileptic drugs was not protocolized. Neurological outcome was assessed at 180-days using the cerebral performance category (CPC) scale. Serum biomarkers were sampled at 24, 48 and 72 hours after cardiac arrest, and batch analyzed as previously described 11,14 . Laboratory methods used to measure GFAP (Banyan Biomarkers Inc) and NfL (Quanterix) have been described by the manufacturers.
Simplified cEEG was performed at six trial-sites using 4 electrodes (F3, P3, F4, P4), a reference in the Cz-position and ground in the Fz-position. All cEEG interpretations were performed by an EEG-specialist (EW) blinded to all clinical data 18 . The following patterns were considered to constitute ESE: • Regularly appearing (=periodic or rhythmic) epileptiform discharges at ≥1Hz continuously (≥90%) appearing during a 30-minute-period.
• Unequivocal electrographic seizure activity with ≥10 second duration defined as generalized rhythmic epileptiform discharges ≥3Hz or clearly evolving discharges of any type reaching >4Hz, according to the EEG criteria of the American Clinical Neurophysiology Society 19 .

Matched control group:
For patients with ESE, matched controls without ESE were identified by propensity score matching, using early (on admission) independent predictors of poor neurological outcome in patients without missing data: age; first monitored rhythm; cardiac arrest at home; time to ROSC; treatment with adrenaline 20 .

Statistics:
Continuous data are reported as median and interquartile range. Predictors of 72-hour serum levels were assessed by univariate and multivariate linear regression analysis with logistic transformation. Matched data were compared using Wilcoxon signed rank test. A two-sided p-value of <0.05 was considered significant. Expert statistical advice was sought for all analyses. Propensity score matching was performed using R with library Match It and optimal matching. For all other analyses IBM SPSS Statistics version 24 was used.

Results
At the six trial sites, 302 patients were included and 134/302 were monitored with cEEG (figure 1). In patients monitored with cEEG, clinical seizures were more common (40%) than in those who were not (26%), other patient characteristics were similar (supplementary table   1

Discussion
ESE was an independent predictor of serum NfL levels at 72-hours after cardiac arrest.
Patients with ESE had higher levels of serum NfL at 72-hours compared to a control group matched for early predictors of poor neurological outcome. These results may suggest additional neuronal injury in patients with ESE and are consistent with an earlier study where ESE was found to be an independent predictor of death 1 .
ESE was not an independent predictor of serum GFAP at 72-hours, although GFAP levels in patients with ESE were higher at 72 hours compared with matched controls. The shorter halflife of GFAP compared with NfL (ref: https://www.ncbi.nlm.nih.gov/pmc/articles/pmid/28717351/) may have attenuated differences between groups. However, the difference in results between the two biomarkers of brain injury may potentially also be explained by ESE predominantly causing injury to neurons as opposed to glial cells.
In an attempt to explore the question of causality and what comes first, the ESE or the secondary brain injury, a control group matched for early predictors of poor neurological outcome was identified among patients without ESE. The variables used for matching were cardiac arrest variables suggesting similar severity of the initial primary brain injury and therefore similar risk factors for developing secondary brain injury (due to reperfusion injury, fever, hyperglycemia, hypoperfusion, seizures, etc.). The biomarkers of brain injury NfL and GFAP were chosen due to their quick release and prolonged presence in serum after brain injury 13 , making increasing levels in subsequent samples more likely to represent additional secondary injury compared to biomarkers of brain injury with shorter half-lives. At 72 hours after cardiac arrest, serum NfL and GFAP levels were significantly higher in patients with ESE compared to matched controls (levels at 24 and 48 hours did not reach significance), suggesting that ESE may cause additional secondary brain injury.

Strengths and limitations
Our results are strengthened by the multicentre study design, prospective data collection and strict criteria for WLST. We used available data to perform matching but acknowledge that matching based on pre-hospital data is imperfect by default. Conclusions are hampered by missing data. Due to the explorative nature and small number of data correlations of repeated measurements were not accounted for. The TTM-trial was not primarily designed to investigate ESE, e.g., serum sampling was not timed according to ESE onset. Additionally, the study design cannot exclude other mechanisms of secondary brain injury as a cause for the higher levels of biomarkers of brain injury seen in ESE. The attempt to match the groups based on cardiac arrest characteristics may be too blunt, creating groups with important differences in severity in brain injury. Therefore, our data should be regarded as hypothesisgenerating and need to be tested prospectively. Some patients received antiepileptic drugs due to clinical and/or electrographic seizures. Due to variability in treatment with antiepileptics, its effect of serum biomarkers could not be assessed.

Conclusions
After cardiac arrest, ESE is associated with higher levels of serum NfL which may indicate a potential secondary neuronal injury caused by ESE. Associations with GFAP, a marker of astrocytic activation/injury, are less clear.