Electrographic Neonatal Seizures after Infant Heart Surgery


Address correspondence and reprint requests to Dr. R.R. Clancy at Division of Neurology, 6 Wood, The Children's Hospital of Philadelphia, Philadelphia, PA 19105, U.S.A. E-mail: Clancy@email.chop.edu


Summary: Purpose: Neonatal seizures are relatively common and an important early sign of acute encephalopathy in those who survive infant heart surgery. The contemporary occurrence of seizures in this setting is not fully known, and their electrographic characteristics are incompletely described. This study describes the characteristics of electrographic neonatal seizures (ENSs) in contemporary infants with congenital heart disease (CHD) surgically repaired by using cardiopulmonary bypass, with or without deep hypothermic circulatory arrest.

Methods: Consecutive infants undergoing heart surgery were monitored by video-EEG for 48 h postoperatively to establish the time of first seizure, total number of ENSs, site(s) of ENS(s) origin and other characteristics.

Results: ENSs occurred in 21 (11.5%) of 183 infants. None had clinically visible seizures. The mean time to the first ENS was 21 h (range, 10–36 h). The total number of ENSs among the entire cohort was 1,429. Mean total number of ENSs per patient over a 48-h period was 72 (range, 1–217). Phenobarbital administration was associated with a ≥50% reduction in seizure counts in five (41.7%) of 12 subjects.

Conclusions: ENSs were relatively common in a large, contemporary cohort of infants after infant heart surgery. A wide variation was noted in seizure burden, but many experienced numerous seizures. Electrographic neonatal seizures are a candidate outcome end point in future neuroprotection trials in this patient population.

Seizures are a common and ominous sign in the newborn infant and are associated with a substantial risk of morbidity and mortality (1–5). A growing concern exists that seizures per se may actively contribute to some part of the infant's ultimate unfavorable outcome (6–8). Infants who undergo heart surgery have an especially high risk of seizures. In the earliest era of congenital heart defect (CHD) surgery, postoperative seizures were reported in ≤50% of patients (9,10). In the time period examined by the Boston circulatory arrest trial (1988–1992), clinical seizures were reported in 11 (6.4%) of 170, and electrographic seizures, in 27 (19.8%) of 136 infants who underwent surgical repair of transposition of the great arteries (TGA) by using a randomly assigned intraoperative support strategy of either cardiopulmonary bypass (CPB) or deep hypothermic circulatory arrest (DHCA) (11). A more recent Boston study between 1992 and 1996 evaluated two types of blood gas–management strategies during CPB and documented EEG seizures in six (5%) of 116 neonates and infants undergoing repair of a variety of cardiac defects, except for “single ventricle” lesions (12). In a study from this institution (1992–1997), infants who survived infant heart surgery using DHCA for a wide variety of CHDs displayed clinical seizures in 17.7% of the whole group (13). However, subgroups were noted with particularly high percentages of seizures depending on their specific CHD anatomy, the duration of DHCA, and the presence of comorbid genetic conditions.

The visual diagnosis of seizures in the newborn is fraught with errors of overdiagnosis and underdiagnosis. The use of neuromuscular blockade, for example, renders clinical recognition of neonatal seizures nearly impossible. Consequently, it is appreciated that the EEG is the gold standard for the detection and quantification of electrographic neonatal seizures (ENSs) (14). This study was conducted to measure the contemporary occurrence of postoperative ENSs for infants with a wide variety of CHDs who undergo infant heart surgery by using CPB with or without DHCA. The specific characteristics of the ENSs, associations with magnetic resonance imaging (MRI) findings, and response to antiepileptic drug (AED) treatment are emphasized in this report.


Patient population

A long-term study of the association of apolipoprotein E (apoE) polymorphisms and neurodevelopmental outcome in survivors of infant heart surgery was initiated at this institution in 1998 and currently includes 550 subjects. Enrolled infants had CHD requiring surgery by using CPB with or without DHCA, were aged newborn to 6 months at the time of surgery, lacked known genetic syndromes except for the 22Q11 deletion, and English was the primary language spoken in the household. A subset population of the larger apoE study was invited to be enrolled into this video-EEG study.

Demographics and clinical characteristics of the EEG study population were available from the database of the apoE polymorphism study. Signed, informed consent was obtained from the parents, and the study was approved by the Institutional Review Board.

Timing of the video-EEG examinations

A brief, 15-min preoperative baseline video-EEG was recorded just before the surgery and then resumed for 48 h after surgery while the patient remained in the Cardiac Intensive Care Unit (CICU). Studies were terminated early only for death or at parental request. The 48 h of postoperative recording were classified into four periods: epoch one (first 12 h postoperatively), epoch two (second 12 h postoperatively), and so forth. Some patients required continuation of video-EEG monitoring for clinical indications to monitor seizures that had persisted beyond the 48-h study period.

Video-EEG examinations

Video-EEGs were recorded on one of three identical portable Telefactor “millennium Beehive” machines (Conshohocken, PA, U.S.A.), capturing time-synchronized video images and digital EEG data. In all subjects, samples of EEG background for all four epochs of the 48-h examination and all examples of ENSs were transferred to compact discs for later review and detailed EEG analyses.

Recording technique

Surface electrodes were applied according to the international 10-20 system, modified for neonates. Tracings were recorded at 15 mm/s with a recording sensitivity of 7 μV/mm. Surface electrodes were applied with collodion glue, and records were frequently reevaluated for technical adequacy and electrodes reapplied or regelled if needed.

Recognition of electrographic neonatal seizures

Each record was visually reviewed in its entirety every 24 h by the recording EEG technologist and independently by one of the authors (R.R.C.). An ENS was recognized by the appearance of a sudden, repetitive, evolving, stereotyped ictal pattern with a definite beginning, middle, and end, with a minimum duration of 10 s and a minimal amplitude of 2 μV (15). An example of an ENS is shown in Fig. 1. If an ENS appeared soon after a prior one, a minimum period of 10 s was required between the two events to consider them separate seizures. Rhythmic EEG artifact mimicking ENSs due to “patting the baby” or chest physiotherapy was easily determined by review of the coincident video files.

Figure 1.

Example of an electrographic neonatal seizure, displayed with the restricted montage of the five regions of interest.

After the electroencephalographer confirmed the presence of ENSs in the daily review of the prior 24 h of study, the CICU attending physician was directly informed of their presence. Treatment decisions were exercised at the discretion of the attending physician.

Quantification of ENSs

After the last patient was enrolled into the 48-h video-EEG study, detailed analyses of ENS characteristics were conducted. The time of onset of ENSs, the number of ENSs per hour for the 48-h study period and the spatial distribution of ENS onset were examined. To simplify the geographic analysis, the full array of the neonatal montage was reduced to five nonoverlapping regions of interest (ROIs) (Fig. 2). ROI-1 was represented by the bipolar derivation FP3→ T3. ROI-2 was represented by the bipolar derivation FP4→ T4. ROI-3 was represented by the bipolar derivation C3→ O1. ROI-4 was represented by the bipolar derivation C4→ O2, and ROI-5, by the bipolar derivation FZ→ CZ. If an individual seizure appeared to arise simultaneously in two ROIs, each region was assigned a value of “one-half,” indicating that the onset of the seizures was equally shared by the two regions. Analogous designations were assigned if an ENS appeared to originate simultaneously from three, four, or all five regions. The durations of ENSs were not measured.

Figure 2.

Modified neonatal montage showing the five nonoverlapping regions of interest (ROIs).

Postoperative neuroimaging examinations

Many of the infants with postoperative ENSs underwent subsequent neuroimaging examinations as part of their clinically warranted evaluations. These were rereviewed to examine possible associations between the geographic distributions of the onsets of their seizures, measured by the distribution of ENSs among the ROIs, and the locations of acute lesions on neuroimaging.


Patient population

ENSs were identified in 21 (11.5%) of 183 patients studied from September 2001 to March 2003. The risk-of-seizure models in this population have been reported elsewhere (16). Affected infants were predominantly male (52%) and term, with a mean estimated gestational age (EGA) of 38.3 ± 1.9 weeks (range, 36–41 weeks). The mean conceptional age (determined by adding the legal age to the EGA) when surgery was performed was 38.4 ± 1.9 weeks. All infants underwent CPB, and 66% also required DHCA. A variety of CHDs were represented (Table 1), including both cyanotic and acyanotic defects. Extracorporeal membrane oxygenation (ECMO) was used postoperatively on three infants, of whom two died. The mean time from the end of anesthesia to the start of the postoperative video-EEG monitoring was 2 ± 2.5 h.

Table 1. Types of congenital heart defects in infants with electrographic neonatal seizures
Congenital heart disease diagnosesNumber of patients
Hypoplastic left heart syndrome8
Atrioventricular canal defects3
Coarctation of the aorta2
Tetralogy of Fallot3
Transposition of the great arteries2

Characteristics of the ENSs

No ENS was detected in the preoperative baseline, and all postoperative ENSs initially arose subclinically. In the entire cohort of 21 infants, a total of 1,429 ENSs was recorded. The first ENS arose between 10 and 36 h postoperatively (mean time of onset, 21 ± 6 h). The total number of hours in which at least one ENS was detected ranged from 1 to 33 h of the 48-h monitoring period.

The total number of ENSs for 48 h in individuals ranged from 1 to 217 (Fig. 3), with a mean total seizure count of 72 ± 76. The maximum number of seizures per hour ranged from one to 16 (mean maximal rate of seizures, 7 ± 5 per hour). Clinically indicated video-EEG monitoring was continued beyond 48 h in 15 (71%) of 21, of whom six continued to have ENSs.

Figure 3.

The distribution of the total electrographic neonatal seizure counts during the 48-h monitoring period.

An example of one study subject's ENSs spatial-temporal profile is shown in Fig. 4. Note that the first ENS was detected in hour 12, peaked to a value of 13 ENSs in hour 24, and continued at a variable rate through hour 48 of the study, after which a clinically requested video-EEG examination commenced. Most ENSs arose from the right hemisphere (ROIs 2 and 4).

Figure 4.

An example of the temporal and spatial distribution of electrographic neonatal seizures from one study subject. The administration of the first dose of phenobarbital also is shown, which has no visible effect on seizure recurrence.

Response to AED treatment

The time of administration of phenobarbital (PB) was determined from retrospective review of the medical records. The size of each dose was not chosen through a specified protocol, and individual “milligram per kilogram” doses varied considerably. The initial PB “loading” doses ranged from 2.4 to 20 (median, 11.2) mg/kg/dose. One patient was administered PB before the onset of ENSs for the purpose of sedation. Two patients were administered PB after the apparently spontaneous remission of ENSs. Four patients were administered a benzodiazepine (BZD) drug. Twelve infants received PB treatment at a time when the ENSs were actively ongoing.

To assess the possible response of ENSs to PB, a pretreatment baseline was determined by counting the mean number of seizures per hour in the 3-h period before treatment and compared with the number of seizures per hour in the 1-h period after treatment. Figure 5 shows the seizure counts in the baseline and posttreatment periods. PB administration while ENSs were actively in progress was associated with a ≤50% seizure reduction in five (41.7%) of 12 subjects. All patients were eventually discharged from the hospital taking PB.

Figure 5.

Comparison of the mean number of electrographic neonatal seizures per hour in a 3-h baseline before phenobarbital administration and 1 h after treatment.

Correlation of locations of ENSs onset with neuroimaging lesions

Table 2 shows the distribution of onset of ENSs by the five ROIs. Most ENSs arose in the right hemisphere (ROIs 2 and 4), followed by the left hemisphere (ROIs 1 and 3) and the midline (ROI 5). The asymmetry between the left and right hemispheric onset was attributed to the occurrence of strictly right-sided focal anatomic lesions in some of this study cohort.

Table 2. Distribution of geographic sites of origin of ENSs among the total (N = 1,429)
Number and percentage of ENSs by ROI54 (3.7%)302 (21.1%)388 (27.2%)439 (30.7%)246 (17.2%)
  1. ENSs, electro-graphic neonatal seizure; ROI, region of interest.

 Left hemisphereMidline (ROI-5)Right hemisphere
 (ROI-1 + ROI-3)(ROI-5)(ROI-2 + ROI-4)
Number and percentage of ENSs by region442 (30.9%)246 (17.2%)741 (51.9%)

A consistent association was found between the laterality of ENSs onset and major findings on MRI examinations. A normal postoperative computed tomography (CT) scan was obtained on one infant. MRIs were obtained in 16 of 20 cases and showed abnormalities in 15. The sites of ENSs origin were bilateral or multifocal in 10 infants with diffuse periventricular leukomalacia (PVL) or global hypoxic ischemic injury. In the five infants with single localized imaging abnormalities, the origins of ENSs were spatially restricted and correlated with the site of focal infarcts or hemorrhages, all of which were right-sided. Figure 6 illustrates these points in two subjects with localized or diffuse parenchymal injuries.

Figure 6.

The relation between site(s) of onset of electrographic neonatal seizures (ENSs) and location(s) of acute lesions detected by postoperative magnetic resonance imaging. Patient A has a parenchymal hemorrhage in the right occipital pole, and all ENSs originate from ROI-4 (C4 → O2). In contrast, patient B has bilateral periventricular leukomalacia (arrows), and ENSs arise from all five ROIs.


Neonatal seizures remain a classic sign of high risk for neonates, a situation that has not changed during the past 30 years (1–3,5,17–20). Specifically, the occurrence of neonatal seizures after infant heart surgery has long been recognized (9,13), but the incidence of clinical seizures appears to be waning. Postoperative seizures correlate with poorer neurodevelopmental scores on subsequent short-term testing (21), but their long-term significance is less clear (22,23).

Because of the uncertainties associated with clinical neonatal seizure detection and quantification, the use of continuous video-EEG monitoring provides a more accurate method to detect and quantify ENSs in this vulnerable patient population (14). Although this study was not intended to measure the occurrence of clinical seizures, it is noteworthy that bedside caretakers detected no clinical seizures. It is well known that ENSs may sometimes be associated with paroxysmal fluctuations in vital signs such as sudden periods of tachycardia or blood pressure instability, especially in those with neuromuscular blockade (24–26). Indeed, in this post–cardiac surgery population, caregivers occasionally activated the “event” button for a concern of seizures with pure autonomic signs. However, not one of these indexed events was accompanied by an electrographic seizure in this study population. Still, electrographic seizures have occasionally been discovered in the past during routine patient care when fluctuations of vital signs have led to an EEG examination. Because of the difficulties with the clinical recognition of seizures, video-EEG monitoring is the ideal method to detect the presence and number of seizures. The availability of relatively inexpensive equipment and the efficient review of long periods of EEG records can now be accomplished in an acceptably brief period.

In the contemporary practice of infant heart surgery, postoperative ENSs are relatively common, occurring in 11.5% of 183 monitored infants. Their “burden” of seizures varied widely, with total seizure counts ranging from one to 217, but most patients had numerous seizures. Prior reports of electrographic seizures after newborn heart surgery were provided by Newburger (11) and du Plessis (12). The details of the electrographic seizure characteristics in the original Boston circulatory arrest trial were described by Helmers et al. (27), who described multiple seizures, mostly clinically silent, with first onset between 4 and 36 h after surgery. Although the total seizure counts per patient were not specified, the combined duration of electrographic seizures per patient ranged from 6 s to 980 min.

A glimpse of the “natural history” of neonatal seizures can be appreciated in this study because AED treatment was not possible for hours after subclinical onset of ENSs. Clearly, some appeared to enjoy spontaneous cessation of ENSs without any AED administration. For many, the first dose of PB had little impact on hourly seizure counts. However, these results must be interpreted cautiously because a wide range of initial drug dosing occurred. Nevertheless, 15 (71%) of 21 required video-EEG monitoring beyond the end of the 48-h study period because seizures persisted despite multiple AEDs. A true efficacy study of AEDs in the treatment of neonatal seizures has never been performed. Several small series reported a variable percentage of seizure reduction after PB administration, but none was placebo controlled (28–33). This is a major limitation because, as this study demonstrates, some infants have an apparently spontaneous remission of seizures before drug administration. Painter's study (34) compared PB with phenytoin (PHT), but, lacking a placebo-control arm, absolute efficacy could not be measured.

Reasonable anatomic correlations between the location(s) of acute injury on MRI and site(s) of onset of ENSs were found. The high occurrence of imaging abnormalities suggests that MRI examinations should be considered in the evaluation of those in whom postoperative seizures developed.

It is generally assumed that the major force behind the mortality and morbidity of neonatal seizures is the underlying etiology that triggered them in the first place, such as hypoxic ischemic encephalopathy, meningitis, or trauma. However, growing evidence in animal models suggests that seizures per se may leave a lasting mark on the immature central nervous system (35–37). In Holmes's study (6), a series of 25 brief seizures induced by fluorothyl led to subsequent measurable impairments in visual spatial memory and increased seizure susceptibility. Indeed, even the profile of postnatal expression of γ-aminobutyric acid (GABA)A receptor subunits is altered in animals after neonatal seizures (7,8).

Newborns with CHD who face infant heart surgery by using CPB with or without DHCA face an exaggerated risk for both postoperative seizures and neurodevelopmental abnormalities and constitute an ideal population for future neuroprotection trials (38). More generally, neonatal seizures are a good target end point for other neonatal neuroprotection trials, including hypothermia trials for neonatal hypoxic ischemic encephalopathy (39). Future neuroprotection studies in the infant heart surgery population could consider the prevention of onset of electrographic neonatal seizures, detected by continuous video-EEG monitoring. Furthermore, those in whom subclinical ENSs develop may be an appropriate group with which to explore definitive, placebo-controlled efficacy trials of PB, BZDs, or other AEDs.


Acknowledgment:  This study was supported by grants from the Fannie E. Rippel Foundation and the American Heart Association. We gratefully acknowledge the contribution of our EEG technologists: Maureen Corcoran, REEGT; Carol Deighan, REEGT, and Edmund Wyeth, REEGT