Choosing wisely: who really benefits from critical care EEG monitoring?

Seizures and status epilepticus are a common complication of critical illness in neonates, children and adults with acute brain injury and, less frequently, in those with systemic illness. Most seizures occur without accompanying clinical manifestations, and thus require electroencephalogram (EEG) for their detection, motivating a widespread increase in the use of critical care continuous EEG monitoring (cEEG) [1]. However, access to cEEG remains limited—even in high-income countries—due to high capital equipment costs and a shortage of qualified EEG professionals, raising concerns about cost-effectiveness [2]. Interestingly, two recent randomized controlled trials (RCTs) allocating critically ill adults to cEEG vs. repeated routine EEG showed comparable outcomes across groups [3, 4]. Why might this be? In both RCTs, subjects who received cEEG had more seizures detected and received more anti-seizure treatment; however, these trials did not assess whether these interventions led to a reduction in seizure burden, which might have been important to be able to achieve improved outcomes.

What is the evidence that the EEG seizures accompanying critical illness cause harm—should they be viewed more as a ‘Silent Killer’ or an ‘Innocent Bystander’? Several observational studies in critically ill neonates, children and adults have consistently demonstrated that a higher seizure burden (e.g., peak seizure intensity during a given hour, and overall duration of seizures) is associated with greater illness severity and worse functional outcome [5–7]. However, despite attempts to adjust for the underlying etiologies and illness severity, their observational design and potential residual confounding prevents them from refuting the ‘Innocent Bystander’ hypothesis: that seizures might represent an epiphenomenon of critical illness rather than independent contributors to worse outcome. Well-designed RCTs have the potential to establish a causal link between seizures and outcome, and several such trials have been attempted. Three RCTs in critically ill neonates with hypoxic ischemic encephalopathy showed that cEEG-guided anti-seizure therapy can successfully reduce seizure burden; however, these trials faced recruitment challenges, limiting their power to detect an outcome difference [8–10]. Similarly, an RCT in comatose adults post-cardiac arrest did not find any outcome benefit from treatment (including anesthetics) of rhythmic and periodic EEG patterns; however, inclusion of patients with highly malignant EEG patterns and likely very poor prognosis may have diluted a possible benefit of cEEG-guided therapy [11]. In a prespecified small subgroup analysis among patients with definite EEG seizures, 30% of those randomized to treatment had a favorable outcome vs. 0% of controls, illustrating the importance of patient selection [12, 13]. In summary, both a causal relationship between seizure burden and outcome, and a benefit of cEEG-guided therapy to reduce seizure burden remain to be proven.

If seizures indeed represent a modifiable risk factor for worse outcome in critically ill patients, this could justify the widespread adoption and considerable costs of cEEG. Conversely, if seizures represent an epiphenomenon of critical illness, then cEEG for seizure management is unlikely to be cost-effective. Given the current uncertainty about whether, or more likely which, critically ill patients stand to benefit from cEEG-guided anti-seizure therapy, clinicians face significant challenges in defining the optimal allocation of potentially limited resources. Decisions about which patients receive cEEG are often influenced by local practice preferences, medico-legal-considerations, and financial limitations or motivations.

Closing the evidence gap

An RCT remains the ideal study design to answer these key questions given that effective randomization should eliminate both known and unknown confounders between the study groups. However, an RCT would also come with several potential challenges. First, it would need to include a large sample of patients (potentially thousands) [3] from a population that would most likely benefit from a seizure burden reduction, specifically excluding patients with extremely severe brain injuries. Second, sites would need to rapidly recruit, obtain informed consent, initiate cEEG, diagnose seizures and administer anti-seizure therapy around-the-clock. These requirements would likely limit participation to highly specialized centers that already have well-resourced cEEG programs. However, such centers, where cEEG represents the standard of care, may be reluctant to adjust their practices in a trial context. Furthermore, potential subjects and their families may be unwilling to participate if there is a widespread perception that more intensive monitoring and aggressive therapy represents better care.

An alternative to an RCT would be to adopt a comparative effectiveness approach in the form of a “natural experiment” taking advantage of the substantial variability in current cEEG practice across centers. Challenges would also include selection of subjects likely to benefit from cEEG-guided therapy and enrolling many patients who are relatively homogeneous in terms of illness severity and etiology given the lack of randomization and high likelihood of confounding. Equally important would be the detailed quantification of seizure burden, timeliness and intensity of anti-seizure therapy, therapeutic endpoints achieved (e.g., seizure cessation vs. burst suppression), and a thorough ascertainment of potential confounders, including the timing and type of EEG interpretation (e.g., 24/7 versus 2–3 times/day; by certified neurophysiologists or registered EEG technologists using raw EEG data, versus intensive care unit (ICU) personnel using quantitative EEG devices) [14].

Although such large trials would come at a considerable cost, the resulting optimization of cEEG monitoring resources and improvements in cEEG-guided therapy could ultimately result in considerable cost savings.

Choosing wisely

Given the current uncertainty, what strategies should clinicians employ when allocating their available cEEG resources? Fig. 1 illustrates suggested approaches depending on a center’s resources and the patient profile. A high-resource center with ample, around-the-clock cEEG availability could take a liberal approach, choosing to monitor even those patients who are less likely to have seizures or benefit from anti-seizure therapy. The advantage might be an improved outcome in patients benefitting from the intervention at the cost of a greater misallocation of resources. On the other hand, a center with very limited cEEG availability could choose to restrict cEEG use to titration of therapy in patients with established seizures, while deploying intermittent routine EEGs in other at-risk groups, perhaps in combination with prophylactic anti-seizure therapy. The advantage would be the more parsimonious use of resources, while the risks may include undertreatment and potential worsening of brain injury and outcomes. A center with moderate cEEG resources could choose to take an in-between approach.

Fig. 1.

Proposed resource-based prioritization of critical care EEG monitoring. For each patient profile, the position of the lozenge indicates the proposed priority for cEEG and its width indicates the range of potential priorities based on individual patient factors (e.g., history of prior clinical seizures, routine EEG findings, illness severity). Dashed lines indicate potential triage thresholds for consideration in high, moderate and limited resource settings

In conclusion, the current uncertainty about the contribution of seizures to outcomes and the benefits of cEEG-guided therapy precludes taking a dogmatic approach. Until we have better evidence, each center will need to establish its own cEEG guidelines based on their available resources, local practice preferences, and patient profiles.

Funding

Open access funding provided by University of Lausanne.

Data availability

Not applicable, as this is a commentary.

Declarations

Conflict of interest

Cecil Hahn has received consulting income from Ceribell, Inc. unrelated to this work. Cecil Hahn serves as an Associate Editor for the Journal of Clinical Neurophysiology. Cecil Hahn served as Past President and serves as co-chair of the Development Committee of the American Clinical Neurophysiology Society. Andrea Rossetti has no potential conflicts of interest to declare.