ARTICLE

Accuracy of Bedside Electroencephalographic

Monitoring in Comparison With Simultaneous
Continuous Conventional Electroencephalography
for Seizure Detection in Term Infants
Divyen K. Shah, MB, ChBa,b, Mark T. Mackay, MB, BSb,c, Shelly Lavery, RNb, Susan Watson, Dipc, A. Simon Harvey, MDb,c,
John Zempel, MD, PhDa,d, Amit Mathur, MDa, Terrie E. Inder, MDa,b,d

Departments of aPediatrics and dNeurology, Washington University, St Louis, Missouri; bMurdoch Children’s Research Institute, Melbourne, Australia;
cDepartment of Neurology, Children’s Neuroscience Centre, Royal Children’s Hospital, Melbourne, Australia

Financial Disclosure: Dr Shah was the recipient of a PhD scholarship from BrainZ Instruments, and Ms Lavery is a research nurse cocoordinator whose salary was partly funded by BrainZ Instruments; the other
authors have indicated they have no financial relationships relevant to this article to disclose.

What’s Known on This Subject What This Study Adds

To date, there have been no studies on the accuracy of digital bedside EEG monitors for Compared with continuous conventional EEG, amplitude-integrated EEG plus 2-chan-
electrical seizure detection in at-risk term infants. nel EEG has acceptable sensitivity and specificity for electrical seizure detection in criti-
cally ill term newborns, when used off-line by experienced users.

ABSTRACT
OBJECTIVE. Our goals were to compare (1) single-channel amplitude-integrated electro-
encephalography alone, (2) 2-channel amplitude-integrated electroencephalography
alone, and (3) amplitude-integrated electroencephalography plus 2-channel electro- www.pediatrics.org/cgi/doi/10.1542/
peds.2007-1839
encephalography with simultaneous continuous conventional electroencephalogra-
phy for seizure detection in term infants to check the accuracy of limited channels doi:10.1542/peds.2007-1839
and compare the different modalities of bedside electroencephalography monitoring. Key Words
neonate, seizure, amplitude-integrated
METHODS. Infants referred to a tertiary center with clinical seizures underwent simul- electroencephalography
taneous continuous conventional electroencephalography and 2-channel (C3-P3 Abbreviations
and C4-P4) bedside monitoring. Off-line analysis of the continuous conventional EEG— electroencephalography
ccEEG— continuous conventional
electroencephalographic results was performed independently by 2 neurologists. electroencephalography
Two experienced neonatal readers reviewed results obtained with amplitude-inte- aEEG—amplitude-integrated
grated electroencephalography and 2-channel electroencephalography combined electroencephalography
and single-channel and 2-channel amplitude-integrated electroencephalography. All Accepted for publication Sep 27, 2007

readings were performed independently and then compared. Address correspondence to Divyen K. Shah,
MB, ChB, Washington University, Department
RESULTS. Twenty-one term newborns were monitored. Seizures were detected in 7 of Pediatrics, One Children’s Place, St Louis,
MO 63110. E-mail: shah㛭[email protected]
patients who had up to 12 electrical seizures, with 1 infant in status epilepticus.
PEDIATRICS (ISSN Numbers: Print, 0031-4005;
Seizures were identified correctly in 6 of 7 patients with amplitude-integrated Online, 1098-4275). Copyright © 2008 by the
electroencephalography plus 2-channel electroencephalography. The missed infant American Academy of Pediatrics
had an isolated 12-second seizure. With amplitude-integrated electroencephalogra-
phy plus 2-channel electroencephalography, 31 of 41 non–status epilepticus seizures were correctly identified
(sensitivity, 76%; specificity, 78%; positive predictive value, 78%; negative predictive value, 78%), with a substantial
level of interrater agreement. The seizures missed were predominantly slow sharp waves of occipital origin from a
single patient (7 of 10 seizures). Nine false-positive results were obtained in 351 hours of recording (1 false-positive
result per 39 hours). These were thought to be related to muscle, electrode, and patting artifacts. Use of amplitude-
integrated electroencephalography alone (1 or 2 channel) provided low sensitivity (27%–56%) and low interob-
server agreement.

CONCLUSIONS. Limited-channel bedside electroencephalography combining amplitude-integrated electroencephalogra-

phy with 2-channel electroencephalography, interpreted by experienced neonatal readers, detected the majority of
electrical seizures in at-risk newborn infants. Pediatrics 2008;121:1146–1154

T HE INCIDENCE OF seizures in the term newborn population is estimated at 2 to 5 cases per 1000 infants.1,2 The
management of seizures in newborns remains controversial, but experimental evidence in animal models
suggests that seizures may exacerbate ischemic cerebral injury.2,3 In addition, there is preliminary evidence that
seizures may be associated with a detrimental neurologic impact in human newborn infants.4,5

1146 SHAH et al

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user
 Bedside electroencephalography (EEG) using ampli- Off-line Analyses
tude-integrated EEG (aEEG) for newborns has been
Criteria for Seizures
used to assist in prognostication,6,7 to recruit infants for
The diagnostic criteria for electrical seizure activity in
neuroprotection intervention studies,8,9 and to monitor
ccEEG were rhythmic, repetitive, stereotypic waveforms
for electrical seizures and responses to treatment.10,11
with an evolution of morphologic features, amplitude,
However, there are few data on the accuracy of limited-
or electric field,14,15 lasting ⱖ10 seconds. Status epilepti-
channel bedside monitoring with aEEG for seizure de-
cus was defined as recurrent seizure activity lasting
tection in newborns.12
ⱖ50% of the time for a minimal duration of 1 hour.14
Modern, digital, bedside EEG monitors incorporate
aEEG results with unprocessed EEG signals from 1 or 2
channels. To date, there are no studies making a direct ccEEG
comparison of the utility of digital monitors and concur- The ccEEG recordings were reviewed independently off-
rent continuous conventional EEG (ccEEG) for seizure line by 2 neurologists with expertise in EEG (Drs Mackay
detection. Hellström-Westas12 showed previously that, and Harvey).
with the use of aEEG alone, short seizures (⬍30 seconds)
were more likely to be missed and some seizures were aEEG Plus 2-Channel EEG
difficult to detect because of external artifacts. The aim of Two neonatal raters (Dr Shah and Ms Lavery), who had
the present study was to compare the accuracy of digital, 3 years of experience each but no specific training, in-
bedside, 2-channel aEEG and simultaneous ccEEG for dependently reviewed the 2-channel aEEG recordings in
off-line seizure detection in term newborns. combination with the raw traces (aEEG plus 2-channel
EEG) off-line for electrical seizures, in a blinded fashion.
When there was disagreement between the 2 raters, a
METHODS consensus was reached, so that a meaningful compari-
son could be made between the 2-channel bedside EEG
Study Group
tool and multichannel conventional EEG.
Infants with clinical seizures who were referred to a
tertiary center (Royal Children’s Hospital, Melbourne,
Australia) were recruited prospectively between April aEEG
2004 and October 2005. Infant recruitment and the Two aEEG users (Drs Mathur and Zempel), who also had
duration of recording depended on the availability of 3 years of experience each but no specific training, in-
ccEEG. The study had received approval from the insti- dependently reviewed the cross-cerebral (P3-P4) aEEG
tutional research and ethics committee, and informed traces and the 2-channel (C3-P3 and C4-P4) aEEG
consent was obtained from the parents of each patient. traces, with a 3-month interval between reviews. The
users were blinded to patient information and the raw
EEG traces. They were asked to point out areas of “def-
Bedside EEG and ccEEG inite” electrical seizure activity on each aEEG trace, as
Infants underwent ccEEG concurrently with bedside would be constituted by a distinct rise in the lower and
aEEG by using a BRM2 brain monitor (BrainZ Instru- upper margins of the aEEG trace above baseline.
ments, Auckland, New Zealand). EEG-video monitoring
was not available for this study. The BRM2 monitor Data Analyses
provided an aEEG trace as well as a raw EEG trace from To allow reporting of specificity and negative predictive
1 channel from each hemisphere (C3-P3 and C4-P4, ie, values, a model was used with the assumption that the
central and parietal electrode positions).13 The aEEG number of ictal periods was equal to the number of
trace resulted from smoothing, rectifying, and filtering interictal periods (ie, the period immediately adjacent to
the raw EEG signal and attenuating frequencies of ⬍2 an electrical seizure). The ictal periods consist of true-
and ⬎15 Hz. The 1-channel aEEG trace was obtained as positive episodes plus false-negative episodes. The inter-
the voltage potential difference between the P3 and P4 ictal periods consist of false-positive episodes plus true-
electrodes, and the 2-channel traces were obtained as negative episodes. Therefore, with known values for
the voltage potential differences between the C3-P3 and true-positive, false-positive, and false-negative episodes,
C4-P4 electrodes. it was possible to calculate values for true-negative epi-
Infants underwent ccEEG with 11 standard, gold disk sodes for each patient. Seizure-free epochs were not
electrodes applied by using the 10 –20 international sys- used to model true-negative episodes because the pro-
tem, at the following locations: Fp2, Fp1, T4, T3, C4, Cz, longed periods of recording provided very high ␬ scores
C3, P4, P3, O2, and O1, with 1 ground electrode, 1 for all modalities of monitoring, limiting the discrimina-
reference electrode, and 1 electrocardiographic elec- tion between tests.
trode. The C3, P3, C4, and P4 electrodes were used for The ␬ statistic was obtained for each pair of indepen-
both conventional EEG (Siesta; Compumedics, Mel- dent reviewers for each modality of bedside monitoring
bourne, Australia) and digital BRM2 monitoring, with assessed. Sensitivity, specificity, positive predictive
split leads. BRM2 and ccEEG traces were recorded simul- value, and negative predictive values were calculated for
taneously. The input signal gain was not altered when all modalities with respect to ccEEG. Analyses were con-
electrodes were shared between the 2 systems. ducted by using SPSS 15 (SPSS, Chicago, IL).

PEDIATRICS Volume 121, Number 6, June 2008 1147

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user
 TABLE 1 Patient Characteristics Preterm infants were included in the study if they were
No. of infants studied 21 of term-corrected postconceptional age. Table 1 shows
Gender, n the characteristics of the infants entered into the study.
Male 13 The most common diagnosis was hypoxic-ischemic
Female 8 encephalopathy. The majority of the infants (19 of 21
Gestational age at monitoring, median 40 (37–43) infants) had been treated with anticonvulsants before
(range), wk study entry, with up to 5 different anticonvulsants being
Birth weight, median (range), g 3300 (1300–4320) used. Phenobarbitone and phenytoin were the most
Apgar score, median (range)
commonly used. Nineteen infants (90%) required me-
1 min 7 (0–9)
5 min 9 (1–9)
chanical endotracheal ventilation, in relation to the se-
Diagnoses, n verity of seizures with a need for airway protection.
Hypoxia-ischemia 6
Vascular, intraventricular hemorrhage 2 (1 subdural hemorrhage, 1
extradural hemorrhage) ccEEG Seizures
Seizures (unknown cause) 4 Infants were monitored for a median period of 18.6
Sepsis 3 (including 1 meningitis) hours, with 351 hours of data obtained. From the group,
Cardiorespiratory 3 7 infants had electrical seizures captured with ccEEG
Inborn error of metabolism 1
(Table 2). One infant (patient E) had status epilepticus. A
Infants treated with anticonvulsants, n/N (%) 19/21 (90)
No. of anticonvulsants, median (range) 2 (0–5)
total of 41 non–status epilepticus seizures were detected.
Infants needing inotropic support, n/N (%) 3/21 (14) The interobserver agreement ␬ for the ccEEG review was
Infants requiring mechanical ventilation, 19/21 (90) 0.84 (P ⬍ .001), consistent with a high level of agree-
n/N (%) ment.
Age at monitoring, median (range), d 4.3 (0.9–71)
Period of monitoring, median (range), h 18.6 (4.3–42.1)
Known outcomes to date aEEG Plus 2-Channel EEG
Death, n infants 7 Seizures were correctly identified for 6 of 7 patients with
Cerebral palsy 3 (2 hemiplegia, 1 quadraparesis) ccEEG electrical seizures by using aEEG in combination
Isolated cognitive delay 1
with the raw trace on 2 channels of the bedside monitor
Developmental delay at 2 y 2
No abnormalities at 2 y 1 (Table 2). Status epilepticus in patient E was also cor-
Awaiting 2-y outcome 7 rectly identified. The duration of seizures identified
ranged from 12 to 300 seconds (median: 89 seconds). A
short single seizure (11-second duration) in patient G
was not detected by using aEEG plus 2-channel EEG.
RESULTS The seizures were recorded within 5 to 24 hours after
Study Group commencement of monitoring. Using aEEG plus 2-chan-
Between April 2004 and October 2005, 21 term infants nel EEG, 31 (76%) of 41 non–status epilepticus seizures
referred with clinical seizures were enrolled. During this were identified (sensitivity, 76%; specificity, 78%; pos-
time period, a total of 53 infants were admitted to the itive predictive value, 78%; negative predictive value,
NICU with a diagnosis of seizures. Three families refused 78%) (Table 3). The interobserver agreement (Cohen’s
consent, and the remaining infants could not be enrolled ␬) was 0.67 (P ⬍ .001), consistent with a substantial
in the study because of restricted availability of ccEEG. level of agreement.

TABLE 2 Characteristics of Detected and Missed Electrical Seizures

Patient Diagnosis ccEEG Seizures aEEG Background No. Detected Detected No. Missed Comment About Seizures
Duration, Missed Duration, s
Range, s
A Intraventricular hemorrhage 8 Discontinuous 7 55–120 1 18 Low-amplitude, sharp waves
B Seizures (UC) 12 Normal SWC 11 45–300 1 290 Right hemispheric PLEDs consisting of
low-amplitude sharp waves
C Seizures (UC) 2 Normal SWC 2 34, 85 0 Bilateral seizures
D Extradural hemorrhage 10 Normal SWC 3 89–128 7 27–753 Detected: left temporal, fast spiking;
missed: left occipital, slow, sharp
waves
E Inborn error of metabolism 3, then SE Normal SWC, SE 43–49 0 Predominantly right-sided seizures
discontinuous, BS
F Seizures (UC) 5 Discontinuous 5 12–18 0 3 T3 rhythmic sharp waves, 2 occipital,
low-voltage, sharp waves
G HIE, stage 2 1 Discontinuous 0 1 11 Left temporal, low-amplitude,
rhythmic, sharp waves
UC indicates unknown cause; HIE, hypoxic-ischemic encephalopathy; SWC, sleep/wake cycle; BS, burst suppression; PLEDs, periodic lateralized epileptiform discharges; SE, status epilepticus.

1148 SHAH et al

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user
 TABLE 3 Sensitivity, Specificity, and Predictive Values of Bedside rate, and abnormal respiratory pattern. The seventh pa-
Monitoring With Respect to ccEEG tient (patient G), who had had a single short seizure, did
not have a clinical correlate and was also missed with all
Sensitivity, Specificity, PPV, NPV, ␬ P
% % % % modes of bedside monitoring.
Of the 9 infants in Table 4, 8 had received anticon-
aEEG plus raw signal, 76 78 78 78 0.67 ⬍.001
vulsants before commencement of monitoring. For 7
agreed seizures
1-channel aEEG infants, ccEEG seizures occurred within 1 to 12 hours
Rater A 56 85 79 85 0.29 .03 after a dose of anticonvulsant. For 2 infants, seizures
Rater B 41 66 55 66 were noted within 1 hour after anticonvulsant adminis-
2-channel aEEG tration. For 1 infant (patient E), who was thought to
Rater A 27 98 92 98 0.31 .01 have an undiagnosed inborn error of metabolism, an
Rater B 44 83 72 83 increase in electrical seizures frequency (with clinical
PPV indicates positive predictive value; NPV, negative predictive value. correlates) was noted to coincide with the administra-
tion of phenytoin.

Seizures Not Detected With aEEG Plus 2-Channel EEG

“Error” Patients
The duration of seizures missed by using aEEG plus
Seven of 21 infants had either false-positive episodes or
2-channel EEG ranged from 11 to 753 seconds (median:
seizures not detected with aEEG plus 2-channel EEG.
170 seconds). The majority of the undetected seizures (7
Three of the 4 infants with missed seizures had some
of 10 seizures) were from patient D. Six of those seizures
true-positive electrical seizures detected by using aEEG
consisted of unilateral, occipital, slow, sharp wave activ-
plus 2-channel EEG (7 of 8, 11 of 12, and 3 of 10
ity (Fig 1).
seizures) (Table 2). The fourth infant (patient G) had an
isolated short (12-second) seizure that was not detected
False-Positive Results with aEEG plus 2-channel EEG. For all 4 patients, clin-
There were 9 false-positive results in 351 hours of re- ical management was not changed as a result of conven-
cording (1 false-positive result per 39 hours). These were tional EEG findings.
obtained from 4 patients. Seven episodes were obtained Four of these 7 error patients had false-positive sei-
from 2 patients who had had no electrical seizures on zure episodes on aEEG plus 2-channel EEG. Two infants
ccEEG. These were thought to be related to electrode, (patients D and F) had single false-positive episodes but
patting, and muscle artifact (Fig 2). also had true-positive episodes (5 and 10 electrical sei-
zures on ccEEG). Patients H and I had 5 and 2 false-
Clinical Course of Infants in Relation to Monitoring and positive episodes, respectively, and neither had had any
Anticonvulsant Administration true-positive episodes. One such episode in patient H
Table 4 shows the clinical course of the infants who had coincided with an episode of apnea for which the patient
ccEEG seizures (patients A to G) and the “error” patients. received bag-and-mask ventilation. In the absence of
Six of the 7 patients who had seizures on ccEEG (pa- EEG-video monitoring, it could not ascertained whether
tients A to F) had clinical correlates with the electrical the electrode artifact was related to the apnea or to the
seizure activity, although those occurred in only 20% to resuscitation process.
50% of ccEEG-confirmed seizures. All infants with clin-
ical seizures had true-positive findings noted with aEEG
plus 2-channel EEG. In addition, 23 of 38 non–status Patients With No ccEEG Seizure Activity
epilepticus seizures on ccEEG and 16 of 31 non–status Of the 14 patients who did not have seizures on ccEEG,
epilepticus seizures on aEEG plus 2-channel EEG did not 2 infants were noted to have apnea, desaturation, and
have a clinical correlate noted. abnormal limb movements. Both infants had already
The most common clinical correlate was apnea and received loading doses of anticonvulsant treatment. An-
desaturation, which was noted for all 6 infants. Other other 2 infants were noted to have apnea and desatura-
signs included clonic jerking, posturing, increase in heart tion only.

FIGURE 1
ccEEG (left) and bedside monitor (right) images of slow,
sharp wave seizure, predominantly in the left occipital area
(arrow), that was not clearly detected by the bedside mon-
itor (patient D).

PEDIATRICS Volume 121, Number 6, June 2008 1149

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user
 FIGURE 2
Examples of false-positive results on the bedside monitor
(left), as seen with ccEEG (right, arrows), related to elec-
trode artifacts.

Duration of Seizures on the Bedside Monitor, Compared With (range: 27–753 seconds; median: 184 seconds). All 5
Duration on ccEEG seizures missed in another patient (patient F) (Table 2)
The duration of electrical seizure activity observed on were short (range: 10 –18 seconds; median: 17 seconds).
the bedside monitor using aEEG with 2-channel EEG These were all diagnosed with aEEG in combination
correlated strongly with the duration of the seizures with the raw traces. On posthoc review, it was noted
observed with conventional EEG (Pearson’s correlation that there was no distinct rise in the lower and upper
coefficient: 0.95; P ⬍ .001) (Fig 3). aEEG band margins for 8 of 13 of the missed seizures. It
was also observed that 6 of those had a discontinuous
aEEG Tracing Alone background and 7 had a normal but “spiky” background.
With single-channel aEEG, 23 (56%) of 41 seizures were
identified in 4 of 7 infants with electrical seizures on Infant Outcomes
ccEEG by 1 or both raters. Between the 2 raters, 14 To date, 2-year outcome information is available for 12
false-positive seizures were identified. With 2-channel of 21 infants (Table 1). Seven (33%) of the 21 infants
aEEG alone, 18 (44%) of 41 seizures were correctly have died and 3 infants have cerebral palsy. Of the 7
identified in 5 of 7 infants by 1 or both raters. Between infants who died, 3 infants had hypoxic-ischemic en-
the 2 raters, 10 false-positive results were obtained. The cephalopathy, 2 had cardiorespiratory problems, 1 had
interobserver agreement ␬ for aEEG tracing alone was intraventricular hemorrhage, and 1 had seizures of un-
0.29 to 0.31 (P ⬍ .05), consistent with a fair degree of determined cause. Only 2 of those infants had electrical
agreement. seizures detected with ccEEG.

Seizures Not Detected With Single-Channel or 2-Channel aEEG DISCUSSION

Thirteen seizures were missed by both raters by using Bedside EEG monitors are being used increasingly for
both single-channel and 2-channel aEEG alone. These seizure detection and treatment monitoring in sick new-
included the same 6 seizures from patient D that were borns in the neonatal intensive care setting. However,
missed by using aEEG plus 2-channel EEG (ie, unilateral, there are few data on their accuracy. From this study,
occipital, slow, sharp waves that were moderately long 2-channel bedside monitoring using aEEG and raw sig-

1150 SHAH et al

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user
by National Guard Health Affairs user
TABLE 4 Clinical Course of Infants in Relation to EEG Monitoring
Patient Diagnosis ccEEG Seizures Clinical Seizure Correlates Signs of Clinical Seizures AEDs Before Monitoring Time of EEG Seizure After AED Clinical Course No. of Infant Outcome
AEDs
A Intraventricular 8 4 A&D, jerking, increases in HR PB, 80 mg/kg; PHY, 40 mg/kg; Within 1 h after PB loading; additional Encephalopathic, intubated during 3 Died at 29 d of age
hemorrhage CLZ, 0.5 mg/kg loading dose improved control monitoring; AED reduced frequency of
electrical seizures
B Seizures (UC) 12 4 A&D, focal limb jerking, PB, 20 mg/kg 5 h; infant given loading PHY dose, Intubated because of apnea 2 Mild developmental
contralateral to EEG with improvement in electrical delay at 2 y
seizure seizure activity
C Seizures (UC) 2 1 A&D, clonic limb jerking, PB, 40 mg/kg 12 h after maintenance PB dose Not intubated, sucking, oral feeding during 3 Normal neurologic
mouthing, eye flickering hospital course findings at 19 mo
D Extradural hemorrhage 10 2 A&D PB, 20 mg/kg Within 1 h after maintenance PB dose Extubated immediately before monitoring; 1 No neurologic
neurologic features normalized and abnormalities

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

sucking soon after noted at 1 y
E Inborn error of 3, then status Multiple A&D, clonic limb jerking PB, 40 mg/kg; PHY, 20 mg/kg Increased frequency to status Very difficult seizure control; intubated 3 Developmental
metabolism epilepticus epilepticus after commencement during course of monitoring delay at 2 y
of PHY load; required additional PB
load for better control
F Seizures (UC) 5 1 A&D, posturing, abnormal PB, 6 mg/kg; PHY, 20 mg/kg Within 4 h after maintenance PHY Seizure control worsened after monitoring 4 Died at 4 mo of age
respiration pattern, dose ceased, requiring intubation and
increases in HR ventilation
G HIE, stage 2 1 No No PB, 40 mg/kg; PHY, 15 mg/kg; 6 h Pretreated with AEDs, monitored on third 3 Diplegic cerebral
CLZ, 0.5 mg/kg day of life palsy at 2 y
H HIE, stage 3 No 1 apnea correlated with A&D requiring bag-and- PB, 40 mg/kg Did not require intubation and ventilation 1 Normal
false-positive episode mask ventilation for A&D
I Bronchiolitis No No No None Intubated and ventilated for 1 Alive
cardiorespiratory support
UC indicates unknown cause; HIE, hypoxic-ischemic encephalopathy; A&D, apnea and desaturation; AED, antiepileptic drug; HR, heart rate; PB, phenobarbitone; PHY, phenytoin; CLZ, clonazepam.

PEDIATRICS Volume 121, Number 6, June 2008

1151
 Seizure duration on bedside monitor (s) focal occipital seizures. Limited channels also limit ob-
5 0.0 0 1 00 .0 0 1 50 .0 0 2 00 .0 0 2 50 .0 0
servation of spatial evolution, compared with multiple
channels. Our findings suggest that, in addition to the
3 00 duration and focus of the seizures, other important
factors that may contribute to seizure detection with
Seizure duration on ccEEG (s)

limited-channel bedside EEG monitoring include the

amplitude, frequency, and morphologic features of the
seizure waves.
2 00
Over 351 hours, 9 false-positive results were ob-
tained by using the combination of aEEG with 2-chan-
nel EEG. These false-positive results were thought to
be related to cup electrode artifacts, but the exact
nature could not be defined without direct observa-
tion or EEG-video monitoring. Of the 7 error patients,
1 00 4 patients had true seizures on ccEEG. Of the remain-
ing 3, 1 had a short, isolated, electrical seizure that
was not detected by using aEEG with 2-channel EEG
and the other 2 infants had 7 false-positive seizure
episodes between them. These 3 patients had been
receiving anticonvulsant treatment before monitor-
50 1 00 1 50 2 00 2 50 ing, and no changes were made in clinical decision-
Seizure duration on bedside monitor (s) making on the basis of monitoring findings.
With the use of aEEG alone, there was no substan-
FIGURE 3 tial difference between 1 or 2 channels for seizure
Duration of seizures on the raw trace of the bedside EEG monitor, compared with dura-
tion on ccEEG. detection, although using aEEG alone was clearly less
accurate than using the combination of aEEG plus
2-channel EEG. With the use of aEEG alone, all 5
nals identified electrical seizures off-line in 6 of 7 infants seizures in 1 infant were missed. On posthoc review of
with seizures on ccEEG. The same 6 patients had clinical the aEEG trace, there was no distinct discernible rise
seizures that correlated with electrical seizures, as ob- in the lower and upper aEEG margins and these sei-
served at the bedside, although clinical correlates were zures were all ⬍20 seconds in duration and were
noted for only 37% of all non–status epilepticus electri- predominantly from infants with normal or discontin-
cal seizures. The single infant who was missed had a uous aEEG backgrounds. The time compression in
short isolated seizure, the clinical significance of which aEEG facilitates monitoring of the background and its
might be questionable. Seventy-six percent of all non– evolution after the onset of encephalopathy.6,16 How-
status epilepticus seizures were identified. The duration ever, for seizure detection with aEEG alone, time com-
of seizures detected with aEEG plus 2-channel EEG cor- pression necessitates a longer duration of seizure.
related strongly with the duration of electrical seizures Therefore, the ability to identify seizures by using
on ccEEG. aEEG alone not only relies on the experience of the
Ten seizure episodes were not detected with aEEG users17 but also seems to be related to the duration of
plus 2-channel EEG. Seven were in 1 patient who had the seizures12 and the aEEG background.
focal, occipital, slow, sharp wave seizure activity. The The present study included critically ill infants with
missed seizures were not brief, with a longer median a heterogeneous group of diagnoses who had already
duration than the detected electrical seizures. Electrical been treated with substantial doses of multiple anti-
seizures in newborn infants tend to be predominantly convulsants. This may account for why only one third
focal or multifocal.2 Limited-channel bedside EEG mon- of this high-risk group had electrical seizures. Anti-
itoring may miss focal seizures remote from the cen- convulsants may bring about a quantifiable reduction
troparietal region, as demonstrated in our study with in EEG amplitude13 and thus suppress aEEG back-

TABLE 5 Review of Studies That Compared the Use of Bedside Monitoring With Conventional EEG for Seizure Detection in Newborn Infants
Authors Total Time Monitored, Infants With Electrical Sensitivity Comment
Median (Range), h Seizures, n/N
Hellström-Westas12 47.7 (0.55–11.2) 6/10 15 (31%) of 48 seizures Seizures of 5–30 s not detected
Toet et al19 10/33 8 (80%) of 10 infants Comparison made with 30-min conventional EEG
recording
Rennie et al17 19 38%–55% of infants Selected traces; 3 speeds of aEEG recording;
inexperienced raters received 3–5 h of training
Present study 351 (4–42) 7/21 31 (67%) of 41 seizures aEEG with 2-channel EEG better than aEEG alone; 1
false-positive result in 39 h of recording

1152 SHAH et al

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user
 ground.18 This may affect electrical seizure detection rithms21 have been developed and validated for use in
in this group, compared with infants with no previous newborns.
anticonvulsant treatment.
Previous studies have made direct comparisons be- ACKNOWLEDGMENTS
tween analog, not digital, aEEG and conventional EEG
We acknowledge BrainZ Instruments for research grant
for seizure detection in newborns12,17,19 (Table 5). Hell-
support for this study.
ström-Westas12 showed that seizures of short duration
We are grateful to Prof Lex W. Doyle for expert advice
may be missed by aEEG. Toet et al19 demonstrated high
and Dr Peter Anderson and Kelly Howard for neurode-
interobserver agreement, sensitivity, specificity, and pre-
velopmental assessments. We are indebted to the fami-
dictive values for aEEG, compared with conventional
lies involved in the study.
EEG recordings of 30-minute duration. Rennie et al17
found poor interobserver agreement and sensitivity for
seizure detection for selected aEEG traces, in comparison REFERENCES
with EEG-video monitoring. However, Rennie et al17 1. Lanska MJ, Lanska DJ. Neonatal seizures in the United States:
results of the National Hospital Discharge Survey, 1980 –1991.
used variable aEEG speeds and inexperienced aEEG rat-
Neuroepidemiology. 1996;15(3):117–125
ers who had received 3 to 5 hours of training.
2. Volpe JJ. Neurology of the Newborn. 4th ed. Philadelphia, PA:
Our findings confirm that the current practice of us- Saunders; 2001
ing aEEG in combination with the unprocessed EEG 3. Wirrell EC, Armstrong EA, Osman LD, Yager JY. Prolonged
signal in modern bedside EEG monitors is clearly better seizures exacerbate perinatal hypoxic-ischemic brain damage.
for seizure detection, compared with aEEG alone, in Pediatr Res. 2001;50(4):445– 454
newborn infants. The interobserver agreement for the 4. McBride MC, Laroia N, Guillet R. Electrographic seizures in
use of the aEEG with 2-channel EEG was substantially neonates correlate with poor neurodevelopmental outcome.
better than that for either single-channel aEEG alone or Neurology. 2000;55(4):506 –513
2-channel aEEG alone, when conducted by experienced 5. Miller SP, Weiss J, Barnwell A, et al. Seizure-associated brain
injury in term newborns with perinatal asphyxia. Neurology.
reviewers. By using the former, a greater proportion of
2002;58(4):542–548
seizures were detected correctly and greater sensitivity
6. Toet MC, Hellström-Westas L, Groenendaal F, Eken P, de Vries
and specificity were obtained. LS. Amplitude integrated EEG 3 and 6 hours after birth in full
The significance and management of electrical sei- term neonates with hypoxic-ischaemic encephalopathy. Arch
zures in newborns remain controversial. However, Dis Child Fetal Neonatal Ed. 1999;81(1):F19 –F23
repeated seizures may exacerbate ischemic injury or 7. al Naqeeb N, Edwards AD, Cowan FM, Azzopardi D. Assess-
lead to direct injury from the seizures.2,3,5 Given that a ment of neonatal encephalopathy by amplitude-integrated
substantial proportion of seizures in newborns are electroencephalography. Pediatrics. 1999;103(6):1263–1271
subclinical,20 accurate methods for detecting electrical 8. Gluckman PD, Wyatt JS, Azzopardi D, et al. Selective head
seizures are required. Continuous, multiple-channel, cooling with mild systemic hypothermia after neonatal
encephalopathy: multicentre randomised trial. Lancet. 2005;
EEG-video monitoring would be the ideal tool for
365(9460):663– 670
monitoring seizure activity. It allows better distinction
9. Azzopardi D, Robertson NJ, Cowan FM, Rutherford MA, Ram-
between seizure activity and artifact and detects focal pling M, Edwards AD. Pilot study of treatment with whole
seizures in cortical locations remote from bedside body hypothermia for neonatal encephalopathy. Pediatrics.
monitor electrode locations. However, it requires sub- 2000;106(4):684 – 694
stantially more resources and trained personnel, par- 10. Hellström-Westas L, Rosen I, Swenningsen NW. Silent seizures
ticularly to give online, 24-hour feedback for seizure in sick infants in early life: diagnosis by continuous cerebral
management. function monitoring. Acta Paediatr Scand. 1985;74(5):741–748
Our study shows that bedside EEG monitoring by 11. Shany E, Khvatskin S, Golan A, Karplus M. Amplitude-
using aEEG in combination with the unprocessed EEG integrated electroencephalography: a tool for monitoring silent
seizures in neonates. Pediatr Neurol. 2006;34(3):194 –199
signal, reviewed off-line by skilled operators, provides
12. Hellström-Westas L. Comparison between tape-recorded and
acceptable sensitivity, specificity, and positive and neg-
amplitude-integrated EEG monitoring in sick newborn infants.
ative predictive values, compared with simultaneous Acta Paediatr. 1992;81(10):812– 819
ccEEG. Our preliminary data confirm that the use of 13. Shah DK, Lavery S, Doyle LW, Wong C, McDougall P, Inder
limited-channel, bedside EEG monitoring may be effec- TE. Use of 2-channel bedside electroencephalogram monitor-
tive in screening at-risk term infants for electrical seizure ing in term-born encephalopathic infants related to cerebral
activity. Electrical seizures detected on the bedside mon- injury defined by magnetic resonance imaging. Pediatrics. 2006;
itor should be confirmed with conventional EEG or 118(1):47–55
EEG-video monitoring, where available. 14. Scher MS, Hamid MY, Steppe DA, Beggarly ME, Painter MJ.
A larger study with around-the-clock bedside EEG Ictal and interictal electrographic seizure durations in preterm
and term neonates. Epilepsia. 1993;34(2):284 –288
monitoring used online by less-experienced neonatal
15. Clancy RR, Legido A. The exact ictal and interictal duration of
clinicians is required to assess the feasibility and im-
electroencephalographic neonatal seizures. Epilepsia. 1987;
pact of monitoring for electrical seizures in the NICU. 28(5):537–541
Real-time seizure detection with limited-channel bed- 16. ter Horst HJ, Sommer C, Bergman KA, Fock JM, van Weerden
side monitors may become a more practical alternative TW, Bos AF. Prognostic significance of amplitude-integrated
to EEG-video and multichannel EEG monitoring EEG during the first 72 hours after birth in severely asphyxi-
when reliable, computerized, seizure detection algo- ated neonates. Pediatr Res. 2004;55(6):1026 –1033

PEDIATRICS Volume 121, Number 6, June 2008 1153

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user
 17. Rennie JM, Chorley G, Boylan GB, Pressler R, Nguyen Y, monitor) and standard electroencephalogram in neonates. Pe-
Hooper R. Non-expert use of the cerebral function monitor for diatrics. 2002;109(5):772–779
neonatal seizure detection. Arch Dis Child Fetal Neonatal Ed. 20. Boylan GB, Rennie JM, Chorley G, et al. Second-line anticon-
2004;89(1):F37–F40 vulsant treatment of neonatal seizures: a video-EEG monitor-
18. Shany E. The influence of phenobarbital overdose on aEEG ing study. Neurology. 2004;62(3):486 – 488
recording. Eur J Paediatr Neurol. 2004;8(6):323–325 21. Navakatikyan MA, Colditz PB, Burke CJ, Inder TE, Richmond
19. Toet MC, van der Meij W, de Vries LS, Uiterwaal CS, van J, Williams CE. Seizure detection algorithm for neonates based
Huffelen KC. Comparison between simultaneously recorded on wave-sequence analysis. Clin Neurophysiol. 2006;117(6):
amplitude integrated electroencephalogram (cerebral function 1190 –1203

THE HOSPITAL FOR SICK CHILDREN HANDBOOK OF PEDIATRIC EMERGENCY MEDICINE

Editors: Lalani, Amina and Schneeweiss, Suzan

Publisher: Jones and Bartlett Publishers
List Price: $41.95
Reviewer: Rebecca Hutchings, MD (Ochsner Clinic Foundation)
Description: The focus of this general handbook for the evidence-based
practice of pediatric emergency medicine is on diagnosis and management
with minimal discussion of pathophysiology. The guidelines, based on current
literature, are laid out in a bulleted format.
Purpose: The purpose is to provide an evidence-based handbook for practi-
tioners of pediatric emergency medicine. It is a successful survey of the
current literature presented in an easily accessible form. With the present
trends in medicine now focusing heavily on evidence-based guidelines, this is
a much needed resource for physicians. The concise format makes it easy to
use in day-to-day situations.
Audience: The authors of this book direct this book at emergency medicine
practitioners. The bulleted format and basic management guidelines make
this book best suited for resident physicians in emergency medicine, family
practice, and pediatrics. It is an excellent management guide, but lacks some
detail about the underlying causes of disease. For this reason, it might not be
ideal for medical students who need to learn about the pathophysiology of
disease before they begin to manage patients.
Features: The first section covers the management of critical illness in
children, which includes topics such as pediatric advanced life support, ad-
vanced trauma life support, neonatal resuscitation, apparent life-threatening
events, and shock. Subsequent chapters are organized in a systems-based
approach. Particularly good are the chapters covering infectious, hematolog-
ic/oncologic, and neurologic emergencies.
Assessment: This is a great pocket guide for residents rotating through the
emergency department, or taking call overnight on the wards. It does not
fully explain all disease processes, but it answers the question “what should
I do?” based on evidence-based principals from the current literature. Overall,
I found this book very helpful. In emergency medicine, there often isn’t time
to peruse all the literature, or reference one of the major authoritative texts
such as Tintinalli (Emergency Medicine: A Comprehensive Study Guide, 6th edi-
tion, McGraw-Hill, 2004) or Rosen (Rosen’s Emergency Medicine: Concepts and
Clinical Practice, 6th edition, Marx et al. (Elsevier, 2006). The chapter on
dermatology and the accompanying images are not comprehensive enough.
Dermatology questions are better addressed in a dedicated textbook with
color illustrations such as Fitzpatrick’s Dermatology in General Medicine, 7th
edition, Wolff et al. (McGraw-Hill, 2007). The section reviewing common
emergency procedures is cursory; a better reference for this topic is Roberts’
Clinical Procedures in Emergency Medicine, 4th edition (Elsevier, 2004).
Review provided by ©Doody’s Review Service™. www.medinfonow.com

1154 SHAH et al

Downloaded from http://publications.aap.org/pediatrics/article-pdf/121/6/1146/1122589/zpe00608001146.pdf

by National Guard Health Affairs user