|Home | About | Journals | Submit | Contact Us | Français|
To describe the association between electrographic seizures and brain injury judged from magnetic resonance imaging (MRI) in newborns treated with hypothermia.
56 newborns treated with hypothermia were monitored using video-EEG through cooling and rewarming, and imaged at a median of 5 days. EEGs were reviewed for seizures and status epilepticus. Moderate-severe injury on MRI was measured using a classification similar to one predicting abnormal outcome in an analogous population.
Seizures were recorded in 17 newborns, five withstatus epilepticus. Moderate-severe injury was more common in newborns with seizures (RR 2.9; 95%CI 1.2-4.5; P=0.02), and present in all with status epilepticus. Children with moderate-severe injury had seizures that were multifocal, later onset, and more likely to have ongoing seizures following 20mg/kg phenobarbital. Newborns with only subclinical seizures were as likely to have injury as compared with those whose seizures had a clinical correlate (57% vs. 60%).
Seizures remain a risk factor for brain injury in the setting of therapeutic hypothermia, especially in neonates with status epilepticus, multifocal onset seizures, and need for multiple medications.However, 40% were spared from brain injury, suggesting that the outcome following seizures is not uniformly poor in children treated with therapeutic hypothermia.
Neonatal seizures are an important risk factor for impairedneurodevelopment in the setting of hypoxic-ischemic encephalopathy[1-4][[5, 6]. Althoughthe adverse outcome is thought to be primarily related to brain injury caused by hypoxia-ischemia, there is also increasing evidence from animal models and human studies that seizures themselves may be harmful to the developing brain, especially in the setting of underlying injury[7-12]. Most studies of neonatal seizures have relied on clinical observation. However, increased use of video-electroencephalogram (EEG) monitoring has highlighted the fact that seizures in newborns are often subclinical[13-15], and that clinicians are poor at clinically distinguishing seizures from non-seizure spells[16, 17]. There has been debate as to the importance of seizures that are electrographic only as compared with those that manifest obvious clinical signs.
Previous studies have demonstrated a clear relationship between seizure burden, and injury on MRI. However, little is known regarding seizure frequency, localization, timing of onset and clinical expression of seizures and their relationship to MRI injury, especially in the setting of therapeutic hypothermia.
The objective of this study was to examine the association between electrographic seizures, both with and without clinical signs, and brain injury as detected by MRI in a cohort of term newborns treated with therapeutic hypothermia.
We examined a cohort of newborns who were treated with therapeutic hypothermia using whole body cooling at University of California, San Francisco, between November 2007 and May 2010. A subset of these subjects were reported by Nash et al. Inclusion criteria for hypothermia at UCSF were based on those used in randomized controlled trials[1, 20-22] and have been reported in detail previously. Clinical data were extracted from electronic medical records and bedside charts. A pediatric neurologist measured the maximal degree of encephalopathy over the first 3 days of life on a scale of 0-6 that considers feeding, alertness, tone, respiratory status, reflexes, and seizures. Clinical and electrographic seizures were treated with lorazepam, phenobarbital, fosphenytoinand/or levetiracetam according to institutional guidelines. The Committee on Human Research at the University of California, San Francisco, approved the study.
Conventional video-EEG using the NicoletOne monitoring system was performed in all subjects. Surface electrodes were applied by a trained technologist according to the international 10-20 system, modified for neonates. Monitoring was initiated as part of routine clinical care as soon as possible after admission to the intensive care nursery (mean 10 ± 3 hours of life), and was maintained throughout the duration of therapeutic hypothermia and rewarming according to institutional clinical guidelines in all but two of the surviving newborns (mean duration of EEG recording 89 ± 25 hours). A neurophysiologist (MRC or JS) reviewed the archived full-length video-EEG recordings to evaluate for the presence of electrographic seizures and status epilepticus, as well as seizure localization and clinical correlates. An electrographic seizure was defined as a repetitive, evolving, and stereotyped waveform, with a definite beginning and end, with a minimum duration of 10 seconds and minimum amplitude of 2 microvolts[25, 26]. Status epilepticus was defined as continuous electrographic seizure activity for at least 30 minutes, or recurrent electrographic seizures for at least 50% of 1-3 hours of recording time[26, 27].EEG background at the onset of recording was classified into one of 5 patterns as previously described[19, 28]: 1) normal for gestational age, including recordings with transient discontinuityfor less than 50% of the recording, with presence of distinct state changes; 2) excessively discontinuous, with persistence of discontinuous activity occupying more than 50% of the recording, and consisting of bursts of normal activity separated by abnormally long, interburst intervals of more than 6 seconds duration, and amplitude >5μV and<25μV, with poor state changes; 3) depressed and undifferentiated, with persistently low-voltage background activity with amplitude between 5μV and 15μV and without normal features; 4) burst suppression, invariant and unreactive pattern of bursts of paroxysmal activity with mixed features but no age-appropriate activity, lasting less than 10 seconds, and alternating with periods of marked voltage attenuation with amplitude ≤5μV; 5) extremely low voltage, invariant and unreactive pattern, with amplitude <5μV or with no discernible cerebral activity.
Infants were imaged using a specialized neonatal head coil on a 1.5 Tesla SignaEchoSpeed system (GE Medical Systems) at a median age of 5 days (range 4 to 14 days). All infants were normothermic at the time of imaging. Imaging sequences, including spoiled gradient echo volumetric T1, axial spin echo T2 MR imaging, and diffusion-weighted imaging, were optimized for the neonatal brain, as previously described [23, 29, 30]. The severity of brain injury was measured using conventional T1 and T2 (short and long echo) MRI, and diffusion weighted-imaging. A pediatric neuroradiologist who was blinded to the clinical history prospectively evaluated MRI images. Injury to the basal ganglia/thalamus and the watershed areas were scored independently using a system that is strongly predictive of neurodevelopmental outcome following neonatal encephalopathy[29, 31]. Brain injury was categorized as “none-mild” or “moderate-severe” using a classification similar to one that was highly predictive for neurologic disability at 18 months of age in newborns treated with hypothermia. Moderate-severe injury was considered present when there was injury to the basal ganglia/thalamus (abnormal signal in the thalamus and lentiform nucleus, or more extensive injury) and/or the watershed areas (abnormal signal in anterior or posterior watershed cortex and white matter, or more extensive involvement). Newborns were also grouped into “basal ganglia/thalamus” or “watershed” patterns of injury on the basis of the predominant site of injury on MRI, as previously described. Newborns were considered to have a predominant watershed pattern of injury when the watershed scores were higher than the basal ganglia/thalamus scores, and a basal ganglia/thalamus predominant injury when the basal ganglia/thalamus scores were as high as or higher than the watershed scores. Newborns were considered to have “near total brain injury” when injury scores were maximal in both the basal ganglia/thalamus and watershed regions.
Statistical analyses were performed using Stata 10.1 software (Stata Corp, College Station, Texas). Chi-squared and Fisher exact tests were used to compare dichotomous variables and t-test was used for continuous variables. Wilcoxon Rank Sum test (Mann-Whitney U) was used to compare nonparametric data. We considered a p value ≤0.05 significant.
During the study period, 61 newborns were treated with hypothermia and monitored with conventional video-EEG. Of these, 56 were evaluated with MRI and were included in this study. The clinical characteristics of the subjects with and without seizures are presented in Table I.
Eight newborns were treated with a phenobarbital loading dose of 20mg/kg prior to the initiation of monitoring: six for suspected clinical seizures and two for seizures on amplitude-integrated EEG (aEEG). Two of these infants (one with suspected clinical seizures and one with aEEG seizures) went on to have electrographic seizures recorded onconventional video-EEG. Overall, electrographic seizures were recorded in 17 newborns (30%), five of whom had status epilepticus(9%). Four newborns had electrographic seizures evident at the onset of monitoring (two with status epilepticus). Of those with electrographic seizures, 7 (41%) never had seizures with a clinical correlate that was detectible on the video recording. Of those with status epilepticus, three(60%) never had recognizable clinical signs of seizure. Subjects with seizures were more likely to have an abnormal EEG background at the onset of recording (94% vs 66%, p=0.0002)(Table II).
Moderate-severe brain injury was present in 59% of newborns with seizures as compared with 26% of those without seizures (RR 2.3; 95%CI 1.2-4.5; P=0.02)(Table III). There was no apparent difference in the risk of moderate-severe injury between newborns with clinically apparent seizures (60%) and those with only subclinical seizures (57%, RR 1.1; 95%CI 0.5-2.4, P=0.9). Of the six children who were treated for suspected clinical seizures prior to arrival at our center, two (33%) had moderate-severe brain injury, which was similar to those without early suspected clinical seizures (36%, RR 0.93; 95% CI 0.3-3.0). All five newborns with status epilepticus had moderate-severe injury (P=0.04).
Seizure characteristics in the newborns with moderate-severe injury differed from those in the children with no-mild injury. Children with moderate-severe injury had seizures that were, on average, later onset (median 18.3 hours vs. 12.1 hours of life, P=0.001), and less likely to be recorded within the first 12 hours of life (0% vs. 43%, P=0.05). The two newborns that were treated with antiepileptic drugs (AEDs) prior to EEG monitoring had later onset of seizures as compared with those who were not treated prior to monitoring (22.9 vs 15.3 hours), though the difference was not significant (P = 0.2). Newborns with moderate-severe injury had seizures that were more difficult to control (30% had seizure resolution with a single loading dose of 20mg/kg of phenobarbital vs. 100% of those with no-mild injury, P=0.004).
Of the newborns with seizures, ten had moderate-severe injury, whereas seven had either no injury (four newborns) or only mild injury. Imaging characteristics of the newborns with and without seizures are presented in Table IV. Newborns with seizures were more likely to have cortical injury (59% vs 31%, p = 0.05) and near total brain injury (29% vs 5%, P=0.01).
There was no consistent association between seizure localization and location of MRI injury (Table V; available at www.jpeds.com). Some children had seizures that localized to a clear focus of injury andothers had no apparent injury in the location of the seizure onset. The two newborns with multifocal onset seizures had maximal injury in both the basal ganglia/thalamus and watershed areas (i.e. near total brain injury).
In this cohort of 56 newborns with moderate-severe HIE who were cooled using whole body hypothermia,status epilepticus, multifocal seizures, and seizures that were resistant to a single loading dose of 20mg/kg of phenobarbital were associated with moderate-severe injury on neonatal MRI. Newborns whose electrographic seizures never had an obvious clinical correlate were as likely to have moderate-severe injury as those who had electrographic seizures with a clinical correlate.
Previous studies have demonstrated that electrographic seizures are a risk factor for adverse outcome in the setting of HIE, with rates of death and disability ranging from 60-100%[5, 6]. We show that seizures remain a risk factor for adverse outcome in the setting of therapeutic hypothermia, However, children with seizures that are easily treated with a first line agent may be spared moderate or severe brain injury, suggesting that the outcome following seizures is not uniformly poor in children treated with therapeutic hypothermia. This finding is consistent with preclinical models of therapeutic hypothermia that showneuroprotectionin spite of alack of reduction in the number of delayed seizures.
The finding that a greater burden of seizures was important corroborates results by Van Rooijet al, who reported a relationship between seizure duration detected with aEEGand degree of MRI injury.Pisaniet al also found that status epilepticuswas a stronger risk factor than recurrent seizures for adverse neurodevelopmental outcome,and McBride et al, showed that having >75 seizures was associated with microcephaly, severe CP, and failure to thrive[5, 18, 27].
In this study, newborns with subclinical seizures diagnosed with video-EEG monitoring were as likely to have MRI injury as those whose seizures had clinical signs. This finding supports data fromBjorkmanet al, who examined newborn piglets subjected to a hypoxic-ischemic insult, and found that animals with seizures had worse MRI, histological injury and neurobehavior when compared withthose without seizures, even if the seizures had no obvious clinical signs. In addition, Connell et al showed that seizures detected using a two-channel recorder were equally likely to be associated with adverse neurodevelopmental outcome whether or not they were associated with clinical signs.
The finding that seizure onset beyond the first 12 hours of life is not only common in newborns with HIE, but is associated with moderate-severe injury on MRI challenges conventional clinical thinking about the relationship between timing of onset of neonatal seizures and severity of insult. Rafayet al found that seizures within the first 12 hours of life were more likely to indicate HIE thanstroke. Only 14% of newborns in that cohort had seizure onset after 12 hours of life, compared with 82% of our cohort. However, the later onset of seizures in the more severely affected newborns is in keeping with the fetal sheep model, where longer duration of ischemia was associated with more prolonged EEG suppression, and delayed transition to epileptiform activity. The reasons for the difference may include the following: (1) EEG versus clinical diagnosis ofseizures; (2) use of continuous EEG monitoring; or (3) differences in the evolution of energy failure, injury and seizures in newborns who are treated with therapeutic hypothermia.
Though our study is strengthened by high quality data acquisition, it is not without limitations. First, monitoring was initiated, on average, at 10 hours of life, and so we may have missed very early onset seizures in some subjects, including those with moderate-severe brain injury. EEG recording in the first hours of life was not possible due to factors including delayed recognition of neurological signs, need for transport from a referring center, patient access, and the time required to place a full montage. Nonetheless, monitoringwas initiated within 12 hours of life in 75% of subjects. Furthermore there was no difference in timing of onset of recording among those with and without brain injury. Given the usual course for neonatal seizures, we believe that it is unlikely that neonates with moderate-severe injury had early seizures that spontaneously resolved prior to the onset of monitoring only to recur at a later time point. However, this possibility does not change the clinically relevant findingthat late emergence of seizures is associated with injury.Second, the long-term clinical outcome data for this cohort arenot yet available. Rutherford et al reported that MRI remains highly predictive of outcome in newborns treated with hypothermia, suggesting that the MRI results are likely to remain predictive as this cohort matures into childhood. Finally, the small number of subjects with seizures prohibits multivariable analysis.
The results of this study have implications for diagnosis, treatment and counseling for neonatal seizures that occur in the setting of therapeutic hypothermia. The high rate of electrographic seizures with onset after the first 12 hours of life suggests that short-term, intermittentEEG recording - as is routine in many centers –is inadequate for detecting the full burden of electrographic seizures in cooled newborns. Further, the similar severity of brain injury in newborns with electroclinical seizures as compared with those with seizures withoutclinical correlate supports the need for randomized controlled trials to evaluate the practice of treating only infants with convulsive seizures. Finally, these data may aid counseling for families regarding prognosis in newborns with seizures in the setting of therapeutic hypothermia.
Supported by the NIH/NCRR UCSF-CTSI (grant UL1 RR024131) and NIH/NINDS (grants5P50NS035902 and NS40117).H.G. is supported by NIH/NINDS K23NS066137 and the Neonatal Brain Research Institute at UCSF.The contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
The authors declare no conflicts of interest.