Prognostic Value of aEEG in HIE: Noncooled Situation

The value of the background pattern in the prediction of neurodevelopmental outcome in infants with HIE was already well established with the use of the standard EEG. A poor background pattern, which persists beyond the first 12 to 24 hours after birth (BS, low voltage, and FT) are well known to carry a poor prognosis. A more recent study by Murray et al. described the evolution of EEG changes after a hypoxic insult. They recorded continuous, multichannel, video EEG from 6 hours to 72 hours after delivery, and neurologic outcome was assessed at 24 months in 44 infants. Of those, 20 (45%) had an abnormal outcome. The best predictive ability was seen at 6 hours of age (area under the receiver operator characteristic curve: 0.958 (95% confidence interval [CI] 0.88-1.04; P .001). EEG features associated with an abnormal outcome were background amplitude less than 30 μV, interburst intervals (IBIs) of more than 30 seconds, electrographic seizures, and absence of sleep-wake cycling (SWC) at 48 hours after birth.

The prognostic value of early aEEG in HIE is described in the meta-analysis of 8 studies by Spitzmiller et al. A sensitivity of 91% (CI 87–95) and a negative likelihood ratio of 0.09 for aEEG tracings were found to accurately predict poor outcome. The relationship among aEEG amplitude measures, Sarnat grades, and MRI abnormality scores has also been reported. The relationship was strongest for the minimum amplitude measures in both hemispheres. A minimum amplitude of less than 4 μV was useful in predicting severe MRI abnormalities.

Both positive and negative predictive values were slightly lower when aEEG was assessed at 3 instead of 6 hours after birth, but they were still considered sufficiently high to use this technique for early selection in hypothermia or other intervention studies. Combining a neurologic examination with aEEG performed less than 12 hours after birth further increased predictive accuracy from 75% to 85%.

Recovery of the background pattern within 24 hours after perinatal asphyxia with a poor background activity (BS, FT, and CLV) has been reported in 20% of the cases. Of these infants, 60% survived with a mild disability or were normal at follow-up. The patients who did not recover either died in the neonatal period or survived with a severe disability.

Another way of looking at recovery of the background pattern is to assess the presence, quality, and time of onset of SWC (see also Figs. 14.1A, 14.8A , and 14.9C ). The presence, time of onset, and quality of SWC reflect the severity of the hypoxic-ischemic insult to which newborns have been exposed. The time of onset of SWC was shown to predict neurodevelopmental outcome based on whether SWC returns before 36 hours (good outcome) or after 36 hours (bad outcome). Therefore we recommend continuous monitoring for at least 48 hours or until a normal SWC pattern is established.

Seizure Detection

A multichannel video EEG study by Murray et al. showed that only one-third of neonatal EEG seizures display clinical signs on simultaneous video recordings. Two-thirds of these clinical manifestations were not recognized or were misinterpreted by experienced neonatal staff with very low interobserver agreement. These findings show that clinical diagnosis is not sufficient for the recognition and management of neonatal seizures and underline the importance of EEG monitoring in infants at risk of developing seizures.

A rapid rise of both the lower and the upper margins of the aEEG tracing is suggestive of an ictal discharge ( Fig. 14.3B ). Seizures can be recognized as single seizures, repetitive seizures, and as status epilepticus (see also Fig. 14.10C ). The latter usually resembles a sawtooth pattern. Correct interpretation is greatly improved by simultaneous reading of the raw EEG, which is now available on most digital aEEG monitors (see Fig. 14.9A ).

The patient was born at 40 weeks’ gestation and weighed 3150 g at birth. An emergency cesarean section was performed for suspected fetal compromise with bradycardia on the cardiotocogram because of a nuchal cord. Apgar scores were 1, 0, 0 at 1, 5, and 10 minutes. The infant was resuscitated for 15 minutes with intravenous adrenalin given once. The umbilical cord pH was 7.25, and base excess was –3.5. the first arterial lactate concentration was 26.4 mmol/L. The infant was cooled for 72 hours at 33.5°C. Seizures were treated with phenobarbital, midazolam, and lidocaine. Magnetic resonance imaging on day 4 showed severe abnormalities in the basal ganglia and thalami. The infant died on day 5 after redirection of care. A, Cooling was started at 3 hours after birth. B, Amplitude-integrated electroencephalogram (aEEG) at 18 hours after birth during cooling shows a flat trace pattern with seizures (indicated by the pink bars ). 10, care; 11, x-ray; 12 + 13, midazolam is given. C, A sparse burst suppression pattern is seen 36 hours after birth (during cooling). D, A dense burst suppression is seen 48 hours after birth. E, A discontinuous normal voltage pattern is seen 72 hours after birth.

Multichannel video EEG is the gold standard for neonatal seizure detection, but it is not always readily available or feasible in the NICU. The advantages of limited channel aEEG compared with multichannel EEG are the easy application and interpretation that can be done in real time by NICU personnel. This can reduce the time to diagnosis and treatment of seizures significantly. However, owing to the nature of the aEEG technique it is not surprising that very brief seizure activity as well as focal seizure activity may be missed. This was also shown by Shellhaas et al. in a large dataset of 125 multichannel EEGs with 851 neonatal seizures. They found that 94% of the multichannel EEGs detected one or more seizures on the C3–C4 channel, which is often used for aEEG. Thus multichannel EEG remains the gold standard for quantification of seizure burden. Rakshasbhuvankar and colleagues recently reviewed 10 studies to answer the question of whether aEEG is as accurate as multichannel EEG in the detection of seizures.

Detection of individual seizures using aEEG was difficult (12%–38%) without access to the real EEG, especially when the seizures were infrequent, brief, or of low amplitude. There were no false-positive findings among control records. Infants with focal seizures, however, usually develop more widespread ictal discharges, which will be identified by limited channel aEEG. In addition, 81% of the neonatal seizures originated from central temporal or midline vertex electrodes, which can potentially be picked up by the aEEG electrodes. An important study is the one from Shah et al., showing that the combination of two-channel aEEG with the real EEG signal detected the majority (76%) of electrical seizures in at-risk newborns. Use of aEEG in combination with the real EEG signal was clearly better in seizure detection than use of aEEG alone (27%–44%). In a review by Evans and colleagues, 44 studies with aEEG and simultaneous multichannel EEG were analyzed. Sensitivity for the presence of seizures by aEEG was 80% and specificity was 50%. Several seizures were overdiagnosed by aEEG as well as by standard EEG (63% vs. 45.5%, respectively). The specificity of aEEG for seizure detection was higher in neonates undergoing EEG for suspected seizures. On the other hand, monitoring for seizures with limited channel aEEG (two channels with access to the real EEG) can be accurately interpreted, compares favorably to multichannel EEG, and is associated with a trend toward reduced seizure burden. It is important to use at least two channels, especially in infants with suspected unilateral brain lesions.

Sometimes the aEEG shows a seizure pattern, but the two-channel real EEG is not conclusive. This could be due to multifocal epileptiform activity. The only way to check this is to perform a multichannel EEG ( Fig. 14.4 ).

Labor was induced at 42 weeks gestational age. The infant’s birth weight was 4100 g. There was shoulder dystocia during labor. Apgar scores were 2 and 5 after 1 and 5 minutes, respectively. The infant was resuscitated for 4 minutes. Umbilical pH was 6.98 and base excess was –18. The infant was not cooled but started to have seizures within 12 hours after birth. He was treated initially with phenobarbital and was transferred to the neonatal intensive care unit. He subsequently needed ventilatory support and was treated with phenobarbital, lidocaine, and midazolam. Cranial ultrasound performed within 24 hours after birth showed diffuse echogenicities in the subcortical white matter and basal ganglia. Magnetic resonance imaging showed extensive cortical gray matter and subcortical white matter abnormalities in the entire left hemisphere and in large areas of the right hemisphere together with the thalami. A, Amplitude-integrated electroencephalogram at 25 hours after birth was performed for suspected seizures, but findings from the two-channel aEEG and real electroencephalogram (B) were not conclusive. C, Multichannel electroencephalogram showed epileptic activity over the vertex (red arrows) . D, Axial T2-weighted image. E, Axial diffusion-weighted image.

Since the increased use of continuous monitoring, it has become apparent that subclinical seizures are common and occur especially following administration of the first AED. This so-called uncoupling or electroclinical dissociation has been reported by several groups and was found in 50% to 60% of the children studied. The aEEG can play an important role in the detection of these subclinical seizures.

Even status epilepticus is not uncommon and occurred in 18% of 56 full-term infants admitted with neonatal seizures recorded with aEEG. The duration of status epilepticus may influence prognosis as well. In a group of 48 infants with HIE and aEEG-detected status epilepticus, there was a significant difference in background pattern, as well as in duration of the status epilepticus between infants with a poor outcome, compared with those with a good outcome. The background pattern at the onset of status epilepticus appeared to be the main predictor of outcome in all neonates with status epilepticus. The background pattern also proved to be an independent predictor of seizures in a group of 90 infants with HIE treated with therapeutic hypothermia. The incidence of seizures did not change after the introduction of therapeutic hypothermia, but the overall seizure burden was reduced. However, status epilepticus is not uncommon in infants with HIE treated with hypothermia; status epilepticus was observed in 10% to 23% of infants. A high seizure burden and status epilepticus have been related to more severe brain injury on MRI in hypothermia-treated infants.

Should We Treat Subclinical Seizures?

There is no consensus whether clinical events without an EEG correlate should be treated or how aggressively to treat electrographic-only seizures. Although human data are scarce, several studies do suggest an adverse effect of both clinical and subclinical seizures on neurodevelopmental outcome. Neonatal seizures have been reported to predispose patients to later problems with regard to cognition, behavior, and development of postneonatal epilepsy. Two previous aEEG studies have shown that infants treated for both clinical and subclinical seizures had a lower incidence of postneonatal epilepsy (8%–9%) compared with those treated only for clinical seizures (20%–50%). It was of interest that in the study by Brunquell the infants with subtle seizures and generalized tonic seizures had a significantly higher prevalence of postneonatal epilepsy ( P = .04 and P = .01, respectively), cognitive impairment ( P = .02; P = .007), and cerebral palsy ( P = .03; P = .002) compared with patients with other seizure types. Subtle seizures were, in addition, more likely to be associated with abnormalities on neurologic examination at follow-up ( P = .03). Prolonged seizures can increase brain temperature and thus increase metabolic demands. Prolonged seizures cause progressive cerebral hypoxia and increase local cerebral blood flow and may steal perfusion from injured brain regions. In a study by Miller et al. of term newborns with HIE, brain injury was independently associated with the severity of seizures. They performed MRI and proton magnetic resonance spectroscopy (MRS) in 90 full-term infants. EEG-confirmed seizures developed in 33 (37%), and the seizures were scored based on frequency and severity, EEG findings, and AED use. Multivariable linear regression tested the independent association of seizure severity with impaired cerebral metabolism measured by lactate/choline and compromised neuronal integrity measured by N -acetylaspartate/choline in the basal nuclei and the intervascular boundary zones. Seizure severity was associated with increased lactate/choline in both the intervascular boundary zone ( P < .001) and the basal nuclei ( P = .011) when controlling for potential confounders of MRI abnormalities and extent of resuscitation at birth. Seizure severity was independently associated with diminished N -acetylaspartate/choline in the intervascular boundary zone ( P = .034).

The same group was able to show an independent effect of clinical neonatal seizures and their treatment on neurodevelopment in 77 children who were born at term and were at risk for hypoxic-ischemic brain injury, and had brain MRI performed in the newborn period. About one-third of the children (25/77) had clinically detected neonatal seizures. Neonatal MRIs were classified for anatomic distribution and severity of acute injury. The severity of brain injury was assessed using high-resolution newborn MRI, and outcome was assessed at age 4 years, using the Full-Scale Intelligence Quotient (FSIQ) of the Wechsler Preschool and Primary Scale of Intelligence–Revised, as well as a neuromotor score. After controlling for severity of injury on MRI, the children with neonatal seizures had worse motor and cognitive outcomes compared with those without seizures. The children with severe seizures had a lower FSIQ than those with mild or moderate seizures ( P < .001). The major limitation of these two important studies is the reliance on clinical evaluation for classification of seizure diagnosis and severity.

In the randomized controlled trial of van Rooij et al. the seizure burden was very high in both groups (treatment of subclinical seizures vs. treatment of clinical seizures only), but the burden was higher in the treatment of clinical seizures only group. It was of interest to see that there was a significant correlation between the duration of seizure patterns and the severity of brain injury on MRI in the blinded group, but not in the group treated for clinical as well as subclinical seizures. More recently, another randomized controlled trial showed that infants treated for electrographic seizures alone had fewer seizures, a lower seizure burden. and a shorter time to treatment than infants treated for clinical seizures only. These authors also confirmed that seizure burden was associated with brain injury scores and showed in addition that high seizure burden was associated with poorer outcome at 18 to 24 months.

Adequate and fast detection of electrographic seizures is important to reduction of the seizure burden. However, it is difficult to constantly monitor real-time EEG. On some digital aEEG machines seizure detection algorithms are available, which may help in detecting seizures. More recently, attempts have been made to develop neonatal seizure detection algorithms for multichannel EEG. The performance of this seizure detection algorithm was not altered by the presence of a depressed EEG background activity caused by phenobarbital administration.