Effect of Treatment of Subclinical Neonatal Seizures Detected With aEEG: Randomized, Controlled Trial
OBJECTIVES: The goals were to investigate how many subclinical seizures in full-term neonates with hypoxic-ischemic encephalopathy (HIE) would be missed without continuous amplitude-integrated electroencephalography (aEEG) and whether immediate treatment of both clinical and subclinical seizures would result in a reduction in the total duration of seizures and a decrease in brain injury, as seen on MRI scans.
METHODS: In this multicenter, randomized, controlled trial, term infants with moderate to severe HIE and subclinical seizures were assigned randomly to either treatment of both clinical seizures and subclinical seizure patterns (group A) or blinding of the aEEG registration and treatment of clinical seizures only (group B). All recordings were reviewed with respect to the duration of seizure patterns and the use of antiepileptic drugs (AEDs). MRI scans were scored for the severity of brain injury.
RESULTS: Nineteen infants in group A and 14 infants in group B were available for comparison. The median duration of seizure patterns in group A was 196 minutes, compared with 503 minutes in group B (not statistically significant). No significant differences in the number of AEDs were seen. Five infants in group B received AEDs when no seizure discharges were seen on aEEG traces. Six of 19 infants in group A and 7 of 14 infants in group B died during the neonatal period. A significant correlation between the duration of seizure patterns and the severity of brain injury in the blinded group, as well as in the whole group, was found.
CONCLUSIONS: In this small group of infants with neonatal HIE and seizures, there was a trend for a reduction in seizure duration when clinical and subclinical seizures were treated. The severity of brain injury seen on MRI scans was associated with a longer duration of seizure patterns.
WHAT'S KNOWN ON THIS SUBJECT:
Seizures are common in full-term infants with HIE. A substantial portion of neonatal seizures are subclinical. There is concern about possible adverse effects of neonatal seizures on the immature brain.
WHAT THIS STUDY ADDS:
Immediate treatment of both clinical and subclinical seizures reduced the total duration of seizure patterns, which suggests a possible reduction of brain injury.
Neonatal seizures are common in full-term infants with hypoxic-ischemic encephalopathy (HIE) and pose a high risk for death or neurologic disability.1,–,4 Clinical recognition of neonatal seizures may be difficult, because manifestations may be subtle.4 Conventional electroencephalography (cEEG) is the standard method to confirm neonatal seizures. Unfortunately, this tool has its limitations; in most units, equipment, technicians, and experienced clinical neurophysiologists are not available 24 hours per day. In the past decade, amplitude-integrated electroencephalography (aEEG) has become a bedside tool that is now used routinely in many NICUs. Prolonged monitoring with aEEG as well as continuous video electroencephalography (EEG) has shown that a substantial proportion of neonatal seizures are subclinical, especially after administration of antiepileptic drugs (AEDs).5,–,9 Because clinical recognition of neonatal seizures is difficult, the presence of clinical seizures may be overestimated, resulting in unnecessary use of AEDs. Conversely, subclinical neonatal seizures may not be recognized without continuous monitoring, resulting in inadequate treatment.9,10 Although human data are scarce, studies suggest an adverse effect of both clinical and subclinical seizures on neurodevelopmental outcomes. Neonatal seizures have been reported to predispose patients to later problems with regard to cognition, behavior, and development of postneonatal epilepsy.1,2,11 Although there is potential harm of seizures in the immature brain, there also is concern about possible adverse effects of anticonvulsant medications on the developing brain.12,–,14 Previous studies showed that infants treated for clinical and subclinical seizures had a lower incidence of postneonatal epilepsy, compared with those treated only for clinical seizures.15,–,18 This was one of the reasons we performed this randomized, controlled trial to investigate whether immediate treatment of both clinical and subclinical seizures detected with continuous aEEG resulted in reduction of the total duration of seizures and less-severe brain injury seen on MRI scans.
We conducted a randomized, prospective, multicenter trial (Subclinical Seizure Question [Suseque] Study). Eleven perinatal centers in the Netherlands and Belgium participated between November 2003 and April 2008. All participating centers were trained in the application and interpretation of aEEG and followed a standardized study protocol. The institutional review board of every center approved the protocol. Written informed consent was obtained from parents before randomization.
Infants were eligible for the study on the basis of gestational age of ≥37 weeks, admission to 1 of the NICUs <24 hours after birth, and diagnosis of HIE and neonatal seizures. HIE was defined on the basis of meeting ≥3 of the following criteria: (1) signs of intrauterine asphyxia (ie, late decelerations on fetal electrocardiograms or meconium-stained liquor), (2) arterial cord blood pH of <7.10, (3) delayed onset of spontaneous respiration, (4) Apgar score of ≤5 at 5 minutes, or (5) multiorgan failure (elevated liver enzyme levels, reduced diuresis, and cardiovascular problems).
Exclusion criteria included the presence of congenital or chromosomal abnormalities, maternal use of narcotics or sedatives, treatment with phenytoin before referral, and administration of muscle-relaxing drugs. Infants who demonstrated subclinical status epilepticus at the beginning of the aEEG registration also were excluded, because immediate treatment with AEDs was considered indicated.
For infants who met the entry criteria, aEEG was started immediately after admission. If infants demonstrated clinical seizures, then they were treated with AEDs. When infants showed their first subclinical seizure, as confirmed with aEEG, they were assigned randomly to either group A (treatment of both clinical and subclinical seizure patterns) or group B (blinding of the aEEG registration and treatment of only clinical seizures). We stratified randomization according to center with a randomized block design with a block size of 6. Randomization codes for every center were supplied in numbered sealed envelopes.
aEEG was performed with an Olympic 6000 cerebral function monitor (Natus, Seattle, WA). Single-channel aEEG signals were recorded from 2 parietal needle electrodes (corresponding to P3 and P4 in the International 10–20 System). The EEG signal was filtered, rectified, and smoothed before it was printed out at slow speed (6 cm/hour). A second tracing recorded the electrode impedance continuously. The Olympic 6000 monitor gave access to the raw EEG data in the review mode and, with the latest software version, the raw EEG signal ran continuously during recording. For blinding of the screen, we used special software (Olympic Medical, Seattle, WA). During blinding, the impedance recording was visible and events such as care procedures, medications, and clinical seizures could be marked.
Table 1 shows the treatment protocol. In the first years (November 2003 to June 2005), lidocaine was given as a second-line drug. Because of concerns about potential cardiovascular side effects and the fact that midazolam was sometimes used around the time of intubation, we changed our treatment protocol,19 with midazolam being given as the second AED. Twenty-six infants received midazolam as a second drug.
Grading of HIE
HIE was classified as moderate (grade II) or severe (grade III) according to the criteria described by Sarnat and Sarnat.20 Evaluation of HIE took place 24 and 48 hours after birth.
Seizure patterns (characteristic pattern, with a sudden increase of both minimal and maximal amplitudes of the recorded signal and a decrease in amplitude in the postictal period) were classified21 as a single seizure pattern, repetitive seizures (≥3 seizure patterns during a 30-minute period), or status epilepticus (continuous seizure pattern for ≥30 minutes, presenting as a “sawtooth pattern” or as continuous increases of the lower and upper margins). All aEEG recordings were analyzed off-line after the completion of enrollment. Analysis was performed independently by 2 aEEG experts (Drs de Vries and Toet), with respect to seizure patterns and background pattern. The readers had full access to all marked events, for differentiation between true ictal discharges and artifacts. When there was disagreement about seizure discharges or background patterns between the 2 raters, a third rater (Dr van Rooij) was involved and consensus was reached.
For each infant, the total duration of seizures was calculated (in minutes) by using the raw EEG data. When status epilepticus was seen, this was taken as 1 period. Only seizure patterns that could be confirmed with the raw EEG data were selected. When AEDs were given, we assessed whether treatment was appropriate, meaning that AEDs were given within 2 hours after the onset of clinical and/or subclinical seizures and another AED was administered if no effect was seen within 1 to 2 hours. For infants in group B, who received AEDs for clinical seizures, we analyzed the aEEG findings for the presence of EEG seizure patterns at the time when clinical seizures were observed.
Depending on their clinical condition, infants underwent MRI 4 to 10 days after birth. MRI scans were reviewed retrospectively by 2 investigators (Drs Groenendaal and de Vries), who were blinded to aEEG results. The severity of brain injury was assessed by using conventional T1- and T2-weighted spin echo sequences, with diffusion-weighted imaging and apparent diffusion coefficient maps when available. Injury was scored for the basal ganglia and thalami in combination with cortical involvement, the watershed areas, and the posterior limb of the internal capsule, by using systems described previously as being predictive for neurodevelopmental outcomes after HIE22,–,24 (Table 2).
Before the study was started, a power analysis was performed by using a power (1 − β) of .80 and a significance level (α) of .05. This resulted in a sample size of 65 infants in both groups. Statistical analysis was performed by using SPSS 12.0 for Windows (SPSS Inc, Chicago, IL). Comparisons of baseline and aEEG characteristics between groups were made with Fisher's exact test or χ2 tests for categorical variables and with t tests for logarithmically transformed continuous variables. Univariate linear regression models were used to test differences in the duration of seizure patterns between groups and to evaluate the association between seizure duration and MRI scores. The level of significance was set at .05.
During the study period, a total of 138 infants met the inclusion criteria (Fig 1). Neonatal baseline characteristics are summarized in Table 3. There were no substantial differences between groups regarding clinical characteristics. One infant also had a right-sided middle cerebral artery infarction and a left-sided anterior cerebral artery infarction. None of the infants received hypothermia as treatment for HIE.
aEEG characteristics are summarized in Table 4. These characteristics were not statistically different between treatment arms.
Duration of Seizure Patterns
For the 33 infants, we calculated a total duration of seizure patterns of 19 378 minutes (10.8% of total registration time). For 12 of the 33 infants, there was ≥1 episode of disagreement between the 2 reviewers regarding seizure discharges, but consensus was reached in all cases.
In both groups, there was a wide distribution of seizure discharges (Fig 2). The duration (median ± SD) of seizure patterns was 196 ± 340 minutes in group A, compared with 503 ± 1084 minutes in group B (Fig 3). No significant difference in duration was found between the groups by using linear regression. In both groups, a longer duration of seizure activity was noted for infants with grade III HIE, compared with infants with grade II HIE, although this difference was not significant (P = .8) (Fig 4).
Twelve infants (63%) in group A and 9 infants (64%) in group B received phenobarbitone in the referral hospital (given for treatment of suspected clinical seizures and not as prophylaxis). Fourteen (74%) of the 19 infants in group A received ≥3 AEDs, compared with 7 infants (50%) in group B. These differences were not statistically significant.
aEEG Analysis and AED Treatment in Group A
In a review of data for the 19 infants in group A, treatment was appropriate for only 8 infants. For the other 11 infants, seizures existed for ≥2 hours before treatment was started or a second- or third-line AED was given or treatment was not effective but no other AED was given. In a comparison of the duration of seizure patterns for the infants who were treated appropriately (n = 8) and those who were not (n = 11), we found a significant difference in duration (37 vs 248 minutes; P = .02).
aEEG Analysis and AED Treatment in Group B
In a review of aEEG registration for the infants in group B, clinically suspected seizures (treated with 1 or 2 AEDs) could not be confirmed with aEEG for 6 of the 14 infants. All infants showed EEG seizure patterns several hours before (n = 3) or after (n = 3) treatment of clinically suspected seizures. For 7 infants, clinical manifestations were confirmed with aEEG. For 4 of those infants, seizure patterns existed for a longer time before clinical signs were seen; for 5 of them, clinical symptoms resolved after AED administration but EEG seizure patterns persisted. One infant, who was assigned randomly after a subclinical seizure pattern was observed, received no AEDs. No clinical signs were noted after blinding of the registration, and no new EEG seizures were seen in a review of the recording.
MRI was performed for 26 (79%) of the 33 infants (15 in group A and 11 in group B). Infants in both groups underwent MRI at a mean age of 5.5 days (range: 3–9 days). We excluded 1 infant with bilateral perinatal arterial stroke from the analysis. In both groups, the median MRI score was 4. Five infants in group A and 4 in group B had MRI scores of <4. When data for the whole group of 25 infants were assessed, there was a significant relationship between the duration of seizure patterns and MRI scores in linear regression analyses (Fig 5). When analyses were performed for both groups, a significant relationship was found only in the blinded group (Fig 5).
Thirteen infants died during the neonatal period, including 6 in group A and 7 in group B. All except 1 infant had grade III HIE; for most of the infants, intensive care treatment was withdrawn because of an expected poor prognosis. This decision was based on multiple examinations, including clinical assessment, EEG background patterns and persistence of seizure patterns, and neuroimaging data (cranial ultrasound and/or MRI data). Infants who died during the neonatal period had a longer duration of seizure patterns than did survivors (428 vs 164 minutes). Infants who died had shorter recording times because of withdrawal of intensive care, which probably restricted the total duration of seizure patterns. In the group of survivors, no significant differences between the groups with respect to time on a ventilator, time of stay in the NICU, and time to discharge were found.
This is the first randomized, controlled trial studying the effect of treatment of subclinical seizures. In this small group of infants, we found a trend for reduction of seizure duration when clinical and subclinical seizure patterns detected with aEEG were treated, although this trend was not statistically significant.
There was no statistically significant difference between groups in the number of AEDs used, although 74% of the infants in the active treatment group received ≥3 drugs, compared with 50% in the other group (treatment of clinical seizures only). It was noted that infants who were treated appropriately for both clinical and subclinical seizure patterns received more AEDs within a relatively short period, which resulted in a significantly shorter duration of seizure patterns, compared with those who were not treated appropriately. This suggests that treatment is more likely to be effective when it is initiated without delay. It is possible that the differences between groups would have been stronger and perhaps even statistically significant if treatment of the infants in the active treatment group had been optimal for all infants.
The mortality rate was lower in group A (32%) than in group B (50%). Although the 2-year assessment of the infants is still awaited, MRI results now are often considered as markers for short-term outcomes. A significant association was found between seizure pattern duration and higher MRI scores (more-severe brain injury), especially for infants in the blinded group who had a longer duration of seizure patterns. This finding supports the assumption that prolonged clinical and subclinical seizures induce or enhance already-existing brain injury.1,2,25 Wirrell et al25 reported that seizures superimposed on hypoxia-ischemia significantly exacerbated brain injury in rat pups. Translation of these findings into clinical practice is complicated. In human studies, it is difficult to measure the effect of seizures on neuronal injury and to distinguish this from the underlying pathogenesis of brain damage and possible effects of AED treatment. The few available studies in neonates do suggest that seizures are likely to increase neuronal injury.1,2,26,–,28
Only neonates with HIE were included in this study, to yield a homogeneous group. Only 1 infant also showed bilateral perinatal arterial stroke. The median duration of seizure patterns in this group of infants with HIE was longer than expected. One half of the patients had grade II HIE and, because most of the infants were undergoing mechanical ventilation and showed some signs of multiorgan failure, it is likely that we included more-severely affected infants within the moderate HIE spectrum. All infants developed their first seizure and were admitted within 24 hours after birth. Infants with mild grade II HIE would by definition show clinical seizures.20 When their seizures were controlled with a loading dose of phenobarbitone and the infants were otherwise faring well, they usually would not be referred to a NICU and therefore would not be eligible for our study.
Four infants were excluded from analyses because they showed no EEG seizures retrospectively. Three of them were assigned randomly to group A. Only 2 of them received phenobarbitone in the referral hospital. In all cases, the suspected seizure discharge was restricted to a single short episode and the decision was made not to treat the infants. All 4 infants showed normal background patterns and survived the neonatal period.
Our study is an aEEG study, and it is well recognized that this technique has limitations. Previous studies showed that short and focal seizures may be missed.29,30 Shellhaas et al31 reported that aEEG alone has significant limitations in the diagnosis and quantification of neonatal seizures. In that study, however, the interpreters had no access to the raw EEG data when evaluating aEEG recordings. New digital aEEG devices do have access to a simultaneous display of the original raw EEG signal, and a novel seizure-detection algorithm has been developed to improve seizure detection with aEEG.23,32,–,34 This probably will help to increase expertise in the recognition of seizures. For our study, we included only seizures that were confirmed by the raw 1-channel EEG tracing. In a recent study by Shah et al,33 76% of seizures seen on full cEEG recordings were identified by using 2-channel aEEG with access to the raw 2-channel EEG tracings. The long duration of the aEEG registration seems to outweigh the limitations of obtaining detailed information during a much shorter, 30-minute cEEG registration. aEEG does not replace cEEG, and ≥1 cEEG study should be performed for every child presenting with moderate or severe encephalopathy.
A limitation of our study is the study population. Before the study was started, a sample size of 130 infants was calculated with power analysis. Obtaining informed parental consent was more difficult than anticipated, because the parents were reluctant to allow blinding of the monitoring device. Also, parents were asked to give permission shortly after an unexpected complicated delivery of a critically ill infant who required acute neonatal transfer to a level 3 NICU. Another limitation is that we did not use video registration; therefore, we cannot exclude the possibility that some subclinical seizures had some subtle clinical symptoms.4,10 However, Murray et al10 showed that only 27% of clinically suspected seizures were subsequently confirmed to be ictal discharges with cEEG. This was also seen in our study, because infants in group B were sometimes treated for clinical seizures that could not be confirmed with aEEG or EEG. These movements have been described as “motor automatism” of uncertain origin.8 One could imagine that these movements were indeed seizures but, being focal, were not identified with single-channel aEEG or EEG recording.
Furthermore, there was a discrepancy in the level of expertise at the participating centers, with 2 centers having >10 years of experience and other centers having started using the technique only recently. This lack of experience probably led to a delay in seizure treatment of 11 infants in group A.
We report the results of the first randomized, controlled trial of treatment of subclinical seizures. In this small group of infants with HIE, a trend was found for a reduction in the duration of seizure patterns when clinical seizures and subclinical seizure patterns were treated. This trend, as well as the significant association of seizure duration and severity of brain injury found on MRI scans, which was seen for infants who received treatment for clinical seizures only, suggests that recognition and treatment of neonatal seizures in infants with HIE can reduce brain injury.
Dr van Rooij was supported by the Dutch Epilepsy Foundation (grant NEF 3–15).
- Accepted August 3, 2009.
- Address correspondence to Linda S. de Vries, MD, PhD, Wilhelmina Children's Hospital, Department of Neonatology, KE 04.123.1, PO Box 85090, 3508 AB Utrecht, Netherlands. E-mail:
This trial has been registered at the International Standard Randomized Controlled Trial Number Registry (identifier ISRCTN61541169).
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- AED =
- antiepileptic drug •
- aEEG =
- amplitude-integrated electroencephalography •
- cEEG =
- conventional electroencephalography •
- HIE =
- hypoxic-ischemic encephalopathy •
- EEG =
- McBride MC,
- Laroia N,
- Guillet R
- Miller SP,
- Weiss J,
- Barnwell A,
- et al
- Volpe JJ
- Boylan GB,
- Rennie JM,
- Pressler RM,
- et al
- Mizrahi EM,
- Kellaway P
- Murray DM,
- Boylan GB,
- Ali I,
- Ryan CA,
- Murphy BP,
- Connolly S
- Levene M
- Bittigau P,
- Sifringer M,
- Genz K,
- et al
- Hellström-Westas L,
- Blennow G,
- Lindroth M,
- Rosen I,
- Svenningsen NW
- Barkovich AJ,
- Hajnal BL,
- Vigneron D,
- et al
- Shah DK,
- Lavery S,
- Doyle LW,
- et al
- Rutherford MA,
- Pennock JM,
- Counsell SJ,
- et al
- Legido A,
- Clancy RR,
- Berman PH
- van Rooij LG,
- de Vries LS,
- Handryastuti S,
- et al
- Toet MC,
- van der Meij W,
- de Vries LS,
- Uiterwaal CSPM,
- van Huffelen KC
- Shellhaas RA,
- Soaita AI,
- Clancy RR
- Shah DK,
- Mackay MT,
- Lavery S,
- et al
- Copyright © 2010 by the American Academy of Pediatrics