Seizure burden is independently associated with short term outcome in critically ill children

Abstract

Seizures are common among critically ill children, but their relationship to outcome remains unclear. We sought to quantify the relationship between electrographic seizure burden and short-term neurological outcome, while controlling for diagnosis and illness severity. Furthermore, we sought to determine whether there is a seizure burden threshold above which there is an increased probability of neurological decline. We prospectively evaluated all infants and children admitted to our paediatric and cardiac intensive care units who underwent clinically ordered continuous video-electroencephalography monitoring over a 3-year period. Seizure burden was quantified by calculating the maximum percentage of any hour that was occupied by electrographic seizures. Outcome measures included neurological decline, defined as a worsening Paediatric Cerebral Performance Category score between hospital admission and discharge, and in-hospital mortality. Two hundred and fifty-nine subjects were evaluated (51% male) with a median age of 2.2 years (interquartile range: 0.3 days-9.7 years). The median duration of continuous video-electroencephalography monitoring was 37 h (interquartile range: 21-56 h). Seizures occurred in 93 subjects (36%, 95% confidence interval = 30-42%), with 23 (9%, 95% confidence interval = 5-12%) experiencing status epilepticus. Neurological decline was observed in 174 subjects (67%), who had a mean maximum seizure burden of 15.7% per hour, compared to 1.8% per hour for those without neurological decline (P < 0.0001). Above a maximum seizure burden threshold of 20% per hour (12 min), both the probability and magnitude of neurological decline rose sharply (P < 0.0001) across all diagnostic categories. On multivariable analysis adjusting for diagnosis and illness severity, the odds of neurological decline increased by 1.13 (95% confidence interval = 1.05-1.21, P = 0.0016) for every 1% increase in maximum hourly seizure burden. Seizure burden was not associated with mortality (odds ratio: 1.003, 95% confidence interval: 0.99-1.02, P = 0.613). We conclude that in this cohort of critically ill children, increasing seizure burden was independently associated with a greater probability and magnitude of neurological decline. Our observation that a seizure burden of more than 12 min in a given hour was strongly associated with neurological decline suggests that early antiepileptic drug management is warranted in this population, and identifies this seizure burden threshold as a potential therapeutic target. These findings support the hypothesis that electrographic seizures independently contribute to brain injury and worsen outcome. Our results motivate and inform the design of future studies to determine whether more aggressive seizure treatment can improve outcome.

Keywords: child; continuous video-EEG monitoring; critical illness; outcome; seizure burden.

Figures

Figure 1

Maximum hourly seizure burden of 20% (12 min) is associated with neurological decline.Comparisons performed using Fisher’s exact test. The single subject with a seizure burden ≥20% per hour who did not experience neurological decline had a baseline PCPC score of 3. *Comparison of the ‘no seizures’ and ‘50% per hour’ groups combined.

Figure 2

Probability of neurological decline rises with increasing seizure burden across all diagnostic categories. Prediction plot is based on the unadjusted multivariable model presented in Table 2, stratified by diagnostic category. The shaded areas represent the 95% confidence intervals.

Figure 3

Higher seizure burden is associated with a greater magnitude of neurological decline. Small or no neurological decline is defined as ≤1 category increase in PCPC score; large neurological decline is defined as ≥2 category increase in PCPC score. Comparison performed using Fisher’s exact test.

Figure 4

Magnitude of neurological decline rises with increasing seizure burden across all diagnostic categories. The magnitude of neurological decline was categorized as no decline, small decline (a worsening by one PCPC category), and large decline (a worsening by two or more PCPC categories). Multinomial logistic regression with a generalized logit model was fitted and prediction plots were generated to illustrate the probability of neurological decline by seizure burden among the three primary diagnostic categories. The shaded areas represent the 95% confidence intervals. Across all diagnostic categories, as seizure burden increased, the probability of a large decline rose and the probability of no decline fell. The probability of a small decline peaked with moderate seizure burden (30–35% per hour), except among subjects with acute brain injury, where it declined monotonically.