Stratifying Seizure Risk with a Rapid EEG

By David Chuang, MD

SYNOPSIS: In a retrospective, large, multicenter trial, rapid response electroencephalogram (EEG) was found to be non-inferior to conventional EEG when incorporated into the 2HELPS2B score to guide how long patients should stay on EEG.

SOURCE: Kalkach-Aparicio M, Fatima S, Selte A, et al. Seizure assessment and forecasting with efficient rapid-EEG: A retrospective multicenter comparative effectiveness study. Neurology 2024;103:e209621.

It is recognized that many critically ill patients have seizures that are not clinically apparent, and that electroencephalogram (EEG) is needed to identify these seizures. A continuous EEG (cEEG), which is an EEG lasting at least two hours, often is needed to capture such seizures. Putting all critically ill patients on cEEG is resource-intensive and hard to accomplish on a timely basis. Thus, a tool to identify who would benefit from cEEG, and for how long, was developed (2HELPS2B score) to optimize the use of hospital EEG resources.¹ The 2HELPS2B score estimates how long a patient should be on EEG based on whether they have a history of seizure and the epileptiform findings present on a one-hour EEG. Even doing a one-hour EEG requires significant resources. The investigators aimed to see if a simplified EEG, with limited cerebral coverage designed to be done quickly without a need for extensive training — known as rapid response EEG (rrEEG) — can take the place of a conventional EEG.

The 2HELPS2B score assigns one point for each of the following: history of seizure, sporadic epileptiform discharge on the one-hour EEG, features associated with the epileptiform discharges, and epileptiform discharges being > 2 Hz. In addition, two points are given if brief, potentially ictal rhythmic discharge is seen. A one-hour EEG study is recommended if the 2HELPS2B score is 0. A 12-hour study is recommended for a score of 1. A 24-hour study is recommended for a score of 2 or higher.

The authors conducted a retrospective, multicenter study at four tertiary care centers: Yale University, Massachusetts General Hospital, the University of New Mexico, and the University of Wisconsin-Madison (UWM). The study consecutively sampled records of patients age 18 years or older from Jan. 1, 2018, to June 20, 2022, who had rrEEG for at least one hour and a cEEG of at least four hours. Exclusion criteria were uninterpretable EEG and duplicate or incomplete data. Reported EEG findings needed to calculate the 2HELPS2B score were verified from EEG data files. Data from patients who had cEEG without a prior rrEEG were used for comparison. Since UWM does not use rrEEG, they served as a comparator sample for cEEG data.

A total of 240 patients who had rrEEG and then cEEG were included in the study. They were compared to 650 unmatched and 240 matched patients who had cEEG but not rrEEG. The median time between rrEEG end and cEEG start was six minutes. The rrEEG cohort, when compared to the unmatched cEEG cohort, was more likely to be older in age, have a shorter cEEG duration, have a history of epilepsy, be suspected of having seizures, be taking anti-seizure medication (ASM), and have acute brain injury not related to ischemic stroke. These differences were not significant with the cEEG matched cohort. Twenty-eight percent of the rrEEG group went on to have electrographic seizure on EEG vs. 30% in the unmatched cEEG group and 27% of the matched cEEG group. This was found not to be significantly different.

The authors found that one-hour rrEEG was non-inferior to predict seizure risk using the 2HELPS2B score by using the original 2HELPS2B validation study. On secondary outcomes, there was no statistically significant difference in survival curves, epileptiform EEG finding, or seizure risk based on 2HELPS2B seizure risk stratification between all groups.

Commentary

This study addresses an important issue in the efficient use of hospital EEG resources. As the study noted, many hospitals do not have the EEG technical coverage, or even expertise, to rapidly perform EEGs. They also may have limitations in terms of the number of EEG machines available for cEEG. Thus, the 2HELPS2B score was created to address this problem.

By replacing a one-hour EEG with rrEEG, the authors speed up how fast a medical decision using the 2HELPS2B score can be made. As the authors stated, this could rapidly guide allocation of EEG resources. For instance, if there are two patients with suspected non-clinically obvious seizures and only one cEEG machine available, the rrEEG would allow the clinician to prioritize who needs the cEEG and who does not. Unfortunately, the study design did not allow the authors to establish if using rrEEG improves resource use, although it can be reasoned that it would in a real-life scenario.

Of note, on secondary analysis, there is no improvement nor worsening of survival outcome with the use of rrEEG over traditional EEG. This is likely because all the research sites are tertiary centers where cEEG resources likely are not lacking. This can be inferred from the reported median time between the end of rrEEG to the start of cEEG being only six minutes, especially when considering it takes 20-40 minutes just to set up a conventional EEG. Most likely, the cEEG was already being placed while the rrEEG was still recording. Conducting a similar study in an EEG resource-limited hospital may provide real-world data on how rrEEG can improve clinical outcome and decrease resource use.

The advantage that rrEEG provides in being able to do EEG rapidly needs to be balanced with its disadvantages. rrEEG has a limited electrode placement. Seizures near the vertex of the head and midline temporal region theoretically can be missed if the seizure was restricted to those areas. If the patient is post-craniotomy or has a skull injury, this may limit the usefulness of rrEEG, since the surgical site may prevent it from being placed, whereas, since a conventional EEG’s electrode is individually placed by a technician, the placement can be tailored to avoid surgical sites. The disposable rrEEG device is also expensive, another factor to consider if working with limited resources. Overall, the finding has the potential to improve patient care and resource use, which is especially necessary more so in hospitals without robust EEG resources.

REFERENCE

1. Struck AF, Tabaeizadeh M, Schmitt SE, et al. Assessment of the validity of the 2HELPS2B score for inpatient seizure risk prediction. JAMA Neurol 2020;77:500-507.

David Chuang, MD, is Assistant Professor of Clinical Neurology, Weill Cornell Medical College.