New Models Predict Seizures in Brain-Injured Children

Pauline Anderson

December 11, 2014

SEATTLE, Washington — With dwindling hospital resources and no sure-fire automated means of reading electroencephalography (EEG) results, interest in developing a model for seizure detection in critically ill children is growing at centers across the United States.

Such models could help pinpoint which patients are at highest risk for subclinical seizures and should be prioritized for receiving continuous EEG monitoring (CEEG).

Two such promising models were discussed here at a press briefing during the American Epilepsy Society (AES) 68th Annual Meeting.

An estimated 10% to 30% of patients with an acute brain condition, including traumatic brain injury (TBI), brain infection, and postcardiac hypoxic ischemia, develop seizures. Mounting evidence suggests that these seizures are associated with worse functional outcomes and quality of life and higher risk for subsequent epilepsy.

In many cases, it's impossible, even for experts, to know just by observing patients whether they are seizing, said Nicholas Abend, MD, a pediatric neurologist at Children's Hospital of Philadelphia (CHOP), Pennsylvania, who is involved with the Critical Care EEG Monitoring Research Consortium.

"In the majority of these cases, even if you're looking right at the patient during the seizure you wouldn't see anything; it's entirely on the EEG that you see it and there's no clinical manifestation."

"Incredibly Time-Consuming"

The problem is that EEG monitoring is "incredibly time-consuming," said Dr Abend. "If we started hooking up every patient in every ICU [intensive care unit] who might meet criteria, we would need way more EEG doctors and readers."

At present, there's no perfect automated system to read EEG results, partly because seizures are so highly variable, he noted. "Someone has to actually sit there and scroll on a computer through every second of EEG that's been recorded looking for seizures."

So models that bedside physicians can use would be extremely useful, added Rajsekar Rajaraman, MD, a pediatric neurology fellow at the University of California at Los Angeles (UCLA). "If we could figure out which kids are going to have subclinical seizures, we can tailor our EEG monitoring for these kids, capture the subclinical seizures sooner and treat them more quickly. It's also more cost-effective."

Dr Rajaraman and Dr Abend helped develop one such model that identified three variables that increase the likelihood that a pediatric patient will develop subclinical seizures: being under 2 years of age, having intradural bleeding, and being exposed to child abuse (abusive head trauma).

To create the model, they studied consecutive patients with TBI admitted to the pediatric intensive care unit at UCLA Medical Center who were put on CEEG monitoring since 2008, as well as children admitted to Children's Hospital of Colorado from July 2009 to May 2010. This group comprised 135 children. Also included in the study were 44 patients admitted to CHOP.

Child Abuse

The investigators looked at the mechanism of injury: whether it was from a fall, blunt trauma, motor vehicle accident, or abusive head injury (child abuse). They also looked at seizure type and at computed tomographic scans to identify such things as fractures and epidural, intradural, or subdural bleeding.

The CHOP cohort was a more critical group. For example, 32% of these patients had subclinical seizures compared with 12.6% of those in the UCLA group.

A "really worrisome" finding of the study was that 87.1% of the patients with subclinical seizures went on to have subclinical status epilepticus, a prolonged seizure with a risk for morbidity and mortality, said Dr Rajaraman.

From these findings, the researchers recommend that all TBI patients under the age of two years who have intradural blood receive CEEG. If they have a history that suggests child abuse, their chances are even higher of having sub-clinical seizures, said Dr Rajaraman.

It's not clear why abusive head trauma would increase the odds for subclinical seizures. Many of these children are younger than age 6 months, so it could be related to age; rat studies show that the developing brain is more prone to seizures. The repetitive and chronic shaking involved in many of these cases may also contribute, as could the fact that children with abusive head trauma may not arrive at the emergency department as quickly as, for example, a child whose head trauma was due to a motor vehicle accident.

Most large centers in the United States see about one case of pediatric TBI per week. For all acute neurologic problems, these centers might see five to 10 children a week, according to Dr Abend.

The second predictive model discussed at the AES meeting was developed by using data from the Critical Care EEG Monitoring Research Consortium, which includes some 35 different sites, including 10 pediatric centers that share data related to seizures in critically ill patients. Each site contributed retrospective data on consecutive children with acute encephalopathy without an epilepsy diagnosis who underwent clinically indicated CEEG.

The researchers looked at various predictors of seizures. "We tried to keep it to things that a clinician would know pretty easily right at admission, not detailed data that might not be available until much later because they have to decide at the beginning whether to do EEG monitoring," said Dr Abend.

The five data points or variables were age, etiology category, clinical seizures before CEEG, initial EEG background category, and inter-ictal discharge category.

The researchers took the multicenter data and tested the model at a single center, CHOP. "We wanted to see if large data acquired by a consortium where each site is slightly different and has different practice patterns and different referral patterns can actually be useful when applied at a single center," explained Dr Abend.

The model worked "reasonably well" in that it's possible to cut the amount of monitoring in half but with the "tradeoff" that some patients who will experience seizures are missed, said Dr Abend. "I take this as pretty positive in that given the variability of going multicenter to single center with only 5 pretty basic data points, we can cut the monitoring in half and only miss 10%."

With "cleaner data and clearer criteria" for which patients are included, and perhaps the inclusion of more data points, "we may be able to do a lot better" and perhaps cut the amount of missed seizures to 5%, said Dr Abend.

Individual centers can determine how strict they want the cutoffs to be, according to available resources. This approach could be advantageous for centers with limited EEG machines and readers, said Dr Abend. "You could at least have a logical strategy for who gets monitored and when you disconnect one patient and attach another patient," he said.

The plan now is to see whether the model translates to other centers across the country that are involved in the Critical Care EEG Monitoring Research Consortium.

The authors have disclosed no relevant financial relationships.

American Epilepsy Society (AES) 68th Annual Meeting. Platform Session A.09, Abstract 1.182. Presented December 8 and December 6, 2014, respectively.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.