Article Text


Electrographic seizures are associated with brain injury in newborns undergoing therapeutic hypothermia
  1. Divyen K Shah1,2,
  2. Courtney J Wusthoff3,
  3. Paul Clarke4,
  4. John S Wyatt5,
  5. Sridhar M Ramaiah4,
  6. Ryan J Dias1,
  7. Julie-Clare Becher6,
  8. Olga Kapellou7,
  9. James P Boardman6,8
  1. 1Royal London Hospital, London, UK
  2. 2Barts and the London Medical School, London, UK
  3. 3Stanford University School of Medicine, Palo Alto, California, UK
  4. 4Norfolk and Norwich University Hospitals, Norwich, UK
  5. 5University College London, London, UK
  6. 6Royal Infirmary of Edinburgh, Edinburgh, UK
  7. 7Homerton University Hospital, London, UK
  8. 8MRC/University of Edinburgh Centre for Reproductive Health, Edinburgh, UK
  1. Correspondence to Dr Divyen K Shah, Barts and the London Childrens Hospital, Neonatology, Royal London Hospital, Whitechapel, London E1 1BB, UK; divyen.shah{at}


Objective Seizures are common among newborns with hypoxic-ischaemic encephalopathy (HIE) but the relationship between seizure burden and severity of brain injury among neonates receiving therapeutic hypothermia (TH) for HIE is unclear. We tested the hypothesis that seizure burden is associated with cerebral tissue injury independent of amplitude-integrated EEG (aEEG) background activity.

Study design Term neonates undergoing 72 h of TH at four centres were selected for study if they had continuous aEEG and MRI. The aEEG with corresponding 2-channel raw EEG (aEEG/EEG), was classified by severity of background and seizure burden; MR images were classified by the severity of tissue injury.

Results Of 85 neonates, 52% had seizures on aEEG/EEG. Overall, 35% had high seizure burden, 49% had abnormal aEEG background in the first 24 h and 36% had severe injury on MRI. Seizures were most common on the first day, with significant recurrence during and after rewarming. Factors associated with severe injury on MRI were high seizure burden, poor aEEG background, 10 min Apgar and the need for more than one anticonvulsant. In multivariate logistic regression, high seizure burden was independently associated with greater injury on MRI (OR 5.00, 95% CI 1.47 to 17.05 p=0.01). Neither aEEG background, nor 10 min Apgar score were significant.

Conclusions Electrographic seizure burden is associated with severity of brain injury on MRI in newborns with HIE undergoing TH, independent of degree of abnormality on aEEG background. Seizures are common during cooling, particularly on day 1, with a significant rebound on day 4.

Statistics from

What is already known on this topic

  • Electrographic seizures are common in term newborns undergoing therapeutic hypothermia for hypoxic-ischaemic encephalopathy. It is unclear whether electrographic seizure burden is a risk factor for cerebral injury seen on MRI, independent of perinatal variables and aEEG background.

What this study adds

  • Electrographic seizures as captured on aEEG with concurrent 2-channel EEG are associated with cerebral injury on MRI independent of aEEG background and Apgar at 10 min. Seizures are most common on the first day after birth with a significant rebound during rewarming.


Hypoxic-ischaemic encephalopathy (HIE) is the most common cause of seizures in term newborns.1 ,2 Full array EEG shows that electrographic seizures are common during treatment with therapeutic hypothermia (TH),2–4 and animal and human studies suggest that neonatal seizures due to hypoxia-ischaemia are associated with long-term neurodevelopmental deficits.5–8 However, the extent to which seizures exacerbate injury is unclear,9 and this contributes to equipoise among clinicians regarding thresholds for treating clinical or electrographic seizures in neonates with HIE.10

TH is standard care for infants with HIE because of its favourable effect on death and disability in resource-rich settings.11 Amplitude-integrated EEG (aEEG) is a standard monitoring tool for assessing cortical electrical activity trends and detecting electrographic seizures in neonates undergoing TH: aEEG with two channels of raw EEG reliably identifies up to three-quarters of electrographic seizures.12

MRI is the gold standard test for characterising the pattern and severity of brain injury in HIE, and is a reliable prognostic marker for subsequent death or cerebral palsy.13–15 MRI's predictive value in the era of TH is established: moderate or severe lesions in the basal ganglia and thalami, severe white matter injury, or an abnormal posterior limb of the internal capsule (PLIC) are predictive of death or severe disability at 18 months of age.16

The role of seizures in the pathogenesis of brain injury is controversial, and uncertainties remain about whether seizure activity, per se, is injurious.3 ,8 ,9 ,17 ,18 We aimed to test the hypothesis that electrographic seizure burden in the first 96 h after birth is associated with severe brain injury on MRI in term infants with HIE treated with TH, independent of clinical variables and severity of aEEG background.


Consecutive term infants with HIE who received TH at four tertiary Neonatal Intensive Care Units (NICUs) from the Royal London Hospital, Homerton University Hospital, Norfolk and Norwich University Hospital, and the Royal Infirmary of Edinburgh in the UK between October 2007 and October 2011 were studied. Data were obtained as part of usual care and the study did not require National Health Service (NHS) ethical review under the terms of the Governance Arrangements for Research Ethics Committees in the UK (advice from South East Scotland Research Ethics Committee).


Neonates were eligible if born at ≥36 weeks’ gestation, were treated with whole-body TH for moderate to severe HIE defined using standard criteria,19 had aEEG monitoring within 12 h and had brain MRI within six weeks. Exclusion criteria: death prior to MR imaging; images degraded by motion artifact; infants with a congenital anomaly or primary diagnosis of an inborn error of metabolism.

Clinical features

Infants received intensive care and TH according to local guidelines, which were informed by the UK TOBY Cooling Register Clinician's Handbook.20 Seizures were treated with anticonvulsant medication at the discretion of the attending neonatologist. Phenobarbitone was the first-line treatment, with phenytoin and/or benzodiazepine and/or lidocaine used if seizures were refractory. Sentinel events were defined as a sustained (preterminal) fetal bradycardia necessitating delivery of the infant, antepartum haemorrhage secondary to placental abruption, placenta previa, cord prolapse or rupture, uterine rupture and shoulder dystocia.

Electrographic seizures

Three channels (C3-P3, C4-P4 and C3-C4) of aEEG and raw source EEG trace were recorded using digital aEEG monitors (BRM2/3 BrainZ Instruments, NZ and Olympic BrainZ Monitor, Natus Medical, USA). Recordings were reviewed off-line by DKS and CJW independently; both were blinded to clinical details and MRI outcomes. The aEEG was reviewed in conjunction with the corresponding raw EEG (aEEG/EEG) for up to 96 h for each patient. A seizure was defined as evolving rhythmic activity lasting for at least 10 s on raw EEG in the absence of artefact. For each 24 h epoch, seizure burden was classified into categories defined a priori: (1) high seizure burden, if seizures were present for > 15 min in total in any 1 h period (frequent seizures) or >30 min in any 1 h period (status epilepticus); or (2) low seizure burden if seizures totalled <15 min of the hour (sporadic seizures), or no seizures (figure 1). For seizures commencing within the first 24 h after birth, the 6 h epoch during which seizures commenced was recorded. Inter-rater agreement was calculated and disagreement was resolved by consensus.

Figure 1

(A) (no seizure) and (B) (short sporadic seizure lasting 5 min) show a low seizure burden; (C) (frequent seizures >15 min, but <30 min) and (D) (status epilepticus seizure lasting more than half an hour) represent a high seizure burden.

aEEG background

The aEEG background was rated independently for the first 24 h after birth by the same blinded raters (DKS and CJW); the most abnormal aEEG background present for at least 1 h on the cross-cerebral channel C3-C4 was identified. The aEEG background was then rated at 48 h after birth. A combination of aEEG pattern recognition and the corresponding raw EEG trace were analysed using the method described by Thoresen et al.21 The aEEG background was classified dichotomously as ‘normal’ for continuous normal voltage and discontinuous normal voltage traces and ‘abnormal’ for burst suppression, continuous low voltage and flat trace recordings.

MR Imaging

MRI was performed at local centres with conventional T1-weighted and T2-weighted sequences at 1.5 Tesla. Images were rated independently by two investigators (OK and JPB), who were blind to aEEG and clinical data. The pattern of MRI injury was classified into two groups using the system described by Rutherford et al,16 which has prognostic value in the era of TH. Group 1 had a severe pattern of injury including: reversed or abnormal signal intensity bilaterally on T1-weighted and/or T2-weighted sequences in the PLIC; multifocal or widespread abnormal signal intensity in the basal ganglia and thalami; or severe widespread white matter lesions including infarction, haemorrhage and long T1 and T2. Group 2 had either normal images or less severe patterns of injury that are associated with normal or mildly abnormal neurodevelopmental outcome (see figure 2 and see online supplementary appendix 1). Inter-rater agreement of categorisation of MRI injury into group 1 or 2 was calculated and disagreements were resolved by consensus.

Figure 2

MRI categorisation by severity of injury among infants with hypoxic-ischaemic encephalopathy (HIE) (T2 weighted images shown). Images shown in (A) and (B) were not categorised as severe: (A) shows normal signal intensity throughout the white matter, occasional foci of abnormal signal intensity in basal ganglia and thalami, and evidence of myelin in the posterior limb of the internal capsule (PLIC); (B) shows moderate white matter abnormality (long T2 extending to subcortical frontal and posterior white matter (short arrows), normal basal ganglia and thalami and normal myelination in the PLIC (thick arrow). (C)–(F) illustrate patterns that were classified as severe. (C) shows severe abnormal signal intensity throughout the basal ganglia and thalami, and there was cortical highlighting (not shown), with no myelin in the PLIC and relative sparing of white matter. (D) and (E) are acquired from the same infant, and show right intraparenchymal haematoma (arrow) with surrounding long T2 that extends superiorly to the level of the centrum semiovale (E). (F) shows severe widespread white matter injury, severe basal ganglia lesions and an absence of myelin in the PLIC bilaterally.

Statistical analysis

Data were analysed using SPSS V.20 (IBM). Clinical characteristics between groups were compared by χ2 for categorical data and t test or Mann–Whitney for continuous data, as appropriate. We investigated the relationship between the following risk factors and severe MRI outcome in bivariate analyses: seizure burden, aEEG background in first 24 h, aEEG background at 48 h, cord or first pH, 10 min Apgar score and the need for one anticonvulsant compared to more than one. To test the association between independent factors and MRI outcome, Pearson's χ2 statistic was used for categorical variables and analysis of variance (ANOVA) was used for continuous variables. Significant factors in the bivariate analyses were included in a multivariate logistic regression model to calculate adjusted odds ratios (OR) to assess the effect of seizure burden on MRI outcome. Two-sided p values <0·05 were regarded as significant.



Eighty-seven patients were eligible. Two were excluded from analysis: one did not receive 72 h of TH due to clinical instability and one had a metabolic encephalopathy. The characteristics of the 85 included cases are shown in table 1.

Table 1

Perinatal characteristics (total subjects=85)

Electrographic seizures

Forty-four (52%) neonates had electrographic seizures; 30 (35%) of these had high seizure burden. In 33 cases, seizures were first identified within 24 h of birth, in 5 cases between 24 h and 48 h, in 2 cases between 48 h and 72 h, and in four cases seizures were first identified after >72 h during/after rewarming (figure 3). For infants seizing in the first 24 h, 10 commenced within the first 6 h after birth, 14 at 6–12 h, eight at between 12 h and 18 h and one at 18–24 h. The prevalence of seizures declined on the second and third days, with a significant increase on the fourth day as compared with the third (χ2=44.41, p<0.001) (figure 3). Rebound seizures, as defined by worsening seizure severity on day 4 compared with day 3 or the appearance of seizures on day 4 de novo, was noted in 12/73 (16%) of newborns (aEEG recording not being available on day 4 for 12 patients). The two raters agreed on the seizure burden for 266/318 (84%) 24 h epochs, Cohen's κ=0.69. They agreed about the presence/absence of seizures over the whole recording in 71/85 (84%), Cohen's κ=0.68.

Figure 3

Electrographic seizure frequency from day 1 to 4 after birth. Seizure frequency was classified according to sporadic (black), frequent (white) and status epilepticus (grey).

The relationship between electrographic and clinical seizures

Of the 44 infants with seizures on aEEG/EEG, 36 (82%) had clinical seizures documented in the medical record. Fifty-eight infants (68% of the total group) were diagnosed with clinical seizures and 56/58 received anticonvulsants. Twenty-seven received phenobarbitone alone, 15 received two anticonvulsants, 12 received three anticonvulsants and two received four drugs. Conversely, 36/58 (62%) with clinically suspected seizures had electrographic confirmation on aEEG/EEG.

aEEG background

There were 42/85 (49%) infants who had burst suppression, continuous low voltage or flat trace on aEEG background for at least 1 h in the first 24 h. Of these, 14/82 (17%) had a persistent abnormal background at 48 h of age. The two raters agreed on the aEEG background in 236/291 (81%) of the 6 h epochs reviewed over the first 24 h, Cohen's κ=0.69.

MRI outcomes

The median age at MRI was 9 (range 4–35) days. Thirty-one infants (36%) had a severe pattern of injury on MRI (group 1). In this group, 26 (30%) had moderate to severe basal ganglia /thalamic lesions; 13 (15%) had severe white matter injury and 14 had a severely abnormal PLIC (16%). 38 subjects (45%) had evidence of cortical injury (mild in 14 cases, moderate in 15 cases and severe in 9 cases), and 21 (25%) had subdural haemorrhage (all minor and none requiring neurosurgical intervention). Four infants had arterial ischaemic stroke in an middle cerebral artery territory.

The two raters agreed on the primary MRI assignment to group 1 or 2 in 76/85 (89%) cases (Cohen's κ=0.69).

Association between perinatal variables, electrographic seizures and MRI outcomes

Two-thirds (20/30) of infants with high seizure burden had severe brain injury. Features significantly associated with severe MRI injury on bivariate analysis included the presence of any electrographic seizures, high seizure burden, abnormal aEEG background in first 24 h, abnormal aEEG background at 48 h age, and Apgar score at 10 min age (table 2 and figure 4). The degree of metabolic acidosis, the age at MRI, the use of anticonvulsants and the presence of rebound seizures were not associated with more severe patterns of injury on MRI. Of babies who were treated with anticonvulsants, the need for more than one anticonvulsant was also significantly associated with cerebral injury. In the multivariate analysis high seizure burden was independently associated with severe injury on MR (p=0.01) (table 2). There was no significant relationship between the timing of the onset of seizures and the MRI outcomes. First-base deficit was significantly associated with seizure burden (F=5.04, p=0.03) as was the 10 min Apgar (F=7.85, p=0.006), but the first pH was not (F=1.32, p=0.25).

Table 2

Crude bivariate analyses for association between risk factors and severe MRI outcome (upper grid) and multivariate regression model assessing the effect of factors significant in bivariate analyses on MRI outcomes (lower grid)

Figure 4

Numbers of infants with good (grey) and poor (black) MRI outcomes in seizure categories broken down into no seizures, sporadic seizures, frequent seizures and status epilepticus. Infants with severe (group 1 black) and non-severe (group 2 grey) patterns of cerebral injury on MRI.


These data demonstrate that electrographic seizure burden is associated with tissue injury on MRI, independent of brain dysfunction reflected in the aEEG background or markers of illness severity at birth. We observed electrographic seizures in 52% of cases, which is comparable with other series,2 ,3 and the seizure burden was high in 35% of the whole cohort. Seizures were most frequent 6–12 h after birth, followed by a diminution in seizure frequency over the following 2 days with a significant increase in the incidence of seizures on the fourth day, coinciding with rewarming period. The phenomenon of ‘rebound seizures’ is well recognised in animal and human studies although the mechanisms are unclear.22–24 TH has direct anticonvulsant effects in adult and paediatric populations with status epilepticus,25 ,26 and there is evidence that TH attenuates the release and accumulation of excitatory neurotransmitters and raises the threshold for seizures and status epilepticus under conditions of neonatal hypoxia-ischaemia.27–29 As expected, there was an association between the need for two or more anticonvulsants and cerebral tissue injury on MRI as compared with the need for one, which is likely to be explained by the extent of cerebral dysfunction in infants who require two or more agents. This study was not designed to assess the direct effects of anticonvulsants on cerebral tissue injury.

The effect of seizures on clinical outcomes is uncertain; some authors have shown that clinical seizures are associated with a worse neurodevelopmental outcome,8 ,30 ,31 while others have not found an association between clinical seizures and worse outcome among infants treated with TH.9 Glass et al showed that infants with electrographic seizures on video EEG are more likely to have moderate-severe cerebral injury on MRI, but the study left some uncertainty about the contribution of global cerebral dysfunction reflected in the EEG background on tissue injury.3 Our data are consistent with those of Glass et al in showing that electrographic seizures are closely associated with severity of brain injury, and they provide clarification that the association between electrographic seizures and brain injury is independent of aEEG background and several markers of illness severity at birth. The distribution of cerebral lesions in the present cohort is comparable with those noted by others.16 ,32 Involvement of the PLIC was noted in 33% of babies compared with 21% and 47% as noted by Cheong et al32 and Rutherford et al,16 respectively.

There are plausible mechanisms by which seizures could amplify brain injury under conditions of hypoxia-ischaemia. Miller et al33 used magnetic resonance spectroscopy to study cerebral energetics in infants with HIE and found that clinical seizure severity was associated with abnormal cerebral metabolism (increased lactate to choline ratio) and markers of neuronal injury (reduced N-acetyl-aspartate to choline ratio). It has been suggested that prolonged or frequent seizures increase cerebral metabolic demands that are challenging to meet in the asphyxial state; that seizures cause local alterations to cerebral haemodynamics that compromise delivery of oxygen and nutrients to other regions of the brain; and that the surge in excitatory neurotransmitters that accompanies seizures exacerbates ischaemic injury.6

Our study has some limitations. There are potential sources of error in the use of aEEG to assess seizure activity, and in the qualitative assessment of aEEG and MRI data. The gold standard for diagnosing electrographic seizures is multichannel video EEG telemetry34; our study sites used aEEG in conjunction with limited channel raw EEG, which is standard practice in the UK because of its usefulness in assessing encephalopathy and predicting outcome.21 ,35 ,36 Our classification system for seizure burden was pragmatic based on a two-channel read-out, and in the absence of an established system for quantifying this measure, using aEEG/EEG. Although it is possible that in this study true electrographic seizure burden may have been underestimated because only limited channels of EEG were available, the overall seizure burden of 52% is similar to other series of infants undergoing TH for HIE.2 ,3 In future studies, use of continuous full-array EEG may improve the precision of seizure burden measurements. Furthermore, blinded raters agreed on the classification of seizure burden and tissue injury in 81% and 89% of cases respectively, with final classification reached through consensus. Improved automated seizure detection algorithms and the use of computational tools to analyse objective measures of cerebral tissue injury on MRI could overcome these sources of error.37 We used tissue injury on MR imaging as a surrogate marker of long-term neurodevelopmental outcome. The system that we used to categorise brain injury is strongly predictive of outcome,16 but it remains important to follow-up large cohorts of infants treated with TH for HIE to determine the effect of perinatal exposures on long-term outcomes.

In conclusion, seizure burden is associated with cerebral tissue injury after controlling for 10 min Apgar, and abnormal aEEG background (burst suppression, continuous low voltage or flat trace) in the first 24 h and at 48 h. Further investigation of the effect of stringent electrographic seizure control versus more permissive approaches for improving long-term clinical outcome is warranted.


JPB would like to acknowledge funding support received from Theirworld.


View Abstract
  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors DKS conceptualised and helped design the study, rated the aEEGs, carried out the data analysis, drafted the initial manuscript, and approved the final manuscript. CJW helped design the study, rated the aEEGs, assisted in writing the manuscript, and approved the final manuscript. PC helped design the study, assisted in writing the manuscript, and approved the final manuscript as submitted. JSW made important intellectual contributions to the manuscript and approved the final manuscript. SMR assisted in aEEG recordings, data collection and approved the final manuscript. RJD assisted in aEEG recordings, data collection and approved the final manuscript. J-CB assisted in aEEG recordings, data collection, writing the manuscript and approved the final manuscript. OK helped conceptualise and design the study, rated the MRIs and approved the final manuscript. JPB helped design the study, rated the MRIs, assisted in the data analysis, assisted in writing the manuscript, and approved the final manuscript.

  • Funding Theirworld Charity provided some funding for Dr Boardman's involvement in this study.

  • Competing interests None.

  • Ethics approval On discussing this study with the South East Scotland Research Ethics Committee, the authors were advised that because the data were obtained as part of usual care NHS ethical review was not required under the terms of the Governance Arrangements for Research Ethics Committees in the UK. The authors were provided with a statement in writing.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.