• 
• 

Three Part Question

In [term neonates] Is [CFM or amplitude EEG (aEEG) as accurate as conventional EEG (cEEG)] in [detecting seizures]?

Clinical Scenario

A term baby girl is admitted to the neonatal unit following concerns regarding abnormal movements noted on the postnatal ward. She is reported to have had tonic posturing of the limbs and back-arching at 10 h of age. As the neonatal registrar on the night shift, you are asked to review the baby and you wonder whether cerebral function monitoring (CFM) of the baby would be an effective and accurate method of detecting seizure activity.

Search Strategy

A search of the EMBASE and MEDLINE healthcare databases (limits: publication year 1974–current and English language) identified 42 articles (search performed March 2010).

No systematic reviews were identified from a search of the Cochrane Library.
The following search keywords were used: [(neonat* OR newborn* OR infant*) AND (cerebral function monitor* OR electroencephalogra* OR amplitude EEG OR conventional EEG OR aEEG OR cEEG) AND (seizure* OR convulsion*)].ti

Search Outcome

5 relevant articles were identified. These studies included sample groups, investigations and outcome measures which were relevant to our original question. All five studies were diagnostic test comparison studies which statistically compared the sensitivity of seizure detection and interobserver agreement of aEEG with cEEG in term or near-term (at least 34 weeks' gestation) neonates with hypoxic ischaemic encephalopathy, suspected clinical seizures or at risk of seizures due to hypoxic events, meningitis or known cerebral pathology. Studies which dealt with the prognostic value of CFM after perinatal asphyxia, the use of combined methods to monitor neurological status, the use of aEEG in different medical conditions and post-anticonvulsant therapy, other aspects of cerebral monitoring in neurological conditions, the effects of antiepileptic drug treatment of subclinical seizures as compared to clinical seizures and non-human subjects, were excluded from the analysis.

Relevant Paper(s)

Author, date and country	Patient group	Study type (level of evidence)	Outcomes	Key results	Study Weaknesses
Toet et al, 2002	36 neonates (gestational age 36 weeks onwards) who had HIE or suspected seizures 33 traces were analysed (3 were excluded due to movement artefacts) Device used: analogue, CFM 4640 (Lectromed Devices, Letchworth, UK)	Diagnostic test study (level 1b)	Value and limitations of aEEG compared to cEEG by expert observers	Results for seizure detection on CFM: κ value 0.7 Sensitivity 80% Specificity 100% PPV 100% NPV 92% No false positives on CFM	Occipital, focal, low amplitude, short duration seizures may be missed on CFM Prolonged recordings may improve the rates of seizure detection CFM seems to be a reliable monitoring tool with cEEG being used for confirmation if there is suspected seizure activity
Rennie et al, 2004	19 neonates (mean gestational age 35 weeks) who were at risk of seizures. 19 seizure events and 21 non-seizures events noted Device used: analogue, Telefactor Beehive Video-EEG system (Telefactor, West Conshohocken, Pennsylvania, USA).	Diagnostic test study (level 1b)	Accuracy of seizure detection in non-experts (inexperienced aEEG raters who received 3–5 h of training) using CFM compared to video-EEG	κ value 0.01–0.39 for interobserver association on single-channel CFM p<0.001 for likelihood of misclassification of seizures, especially at lower (usual) speeds Babies without seizures also more likely to be misclassified	Single channel CFM is not a suitable tool for diagnosis of seizures, especially by non-experts Neonatal seizures need to be detected using EEG first. CFM may be useful in long-term monitoring
Shellhaas et al, 2007	851 neonatal seizure events were identified from 125 cEEG tracings available from 121 neonates (gestational ages 34 weeks onwards) Device used: not stated	Diagnostic test study (level 1b)	Sensitivity of single-channel aEEG in seizure detection by neonatologists (with varying aEEG interpretation expertise and unavailability of the raw EEG) compared to the cEEG	Seizures were detected in 22–57% of aEEG records but only 12–38% of individual seizures were detected No false positives on aEEG	Sensitivity of seizure detection on aEEG depends on expertise of interpreter; seizure appearance, amplitude, duration and frequency per hour (95% CI and p values were <0.001) aEEG is a useful supplemental tool but cannot be used alone in the diagnosis and quantification of seizures
Shah et al, 2008	21 term infants who had clinical seizures Each digital aEEG tracing was (1) single channel, (2) two-channel and (3) two-channel with raw EEG and was compared with cEEG Device used: digital, BRM2 brain monitor (BrainZ Instruments, Auckland, New Zealand)	Diagnostic test study (level 1b	Accuracy of limited channel aEEG compared to different modalities of bedside EEG monitoring with cEEG by expert observers	aEEG with two-channel EEG: Sensitivity 76% Specificity 78% PPV 78% NPV 78% with a substantial level of inter-rater agreement (κ value 0.67) aEEG (one or two channel alone): low sensitivity (27–56%) and low interobserver agreement (κ value 0.29–0.31).	Combined aEEG with simultaneously recorded EEG improves seizure detection rate among experienced raters
Bourez-Swart et al, 2009	12 term neonates (121 seizure events) out of a total of 125 babies admitted with HIE and suspected seizures, underwent cEEG and single-channel aEEG (which was derived from the cEEG signals) Device used: digital, Micromed Systemplus EEG system (Micromed, Mogliano, Veneto, Italy)	Diagnostic test study (level 1b	Seizure pattern detection with single-channel and multi-channel aEEG compared to cEEG by expert observers	Sensitivity of seizure detection rate in aEEG: (1) in relation to localisation: multi-channel: 39% (95% CI 0.31 to 0.48) and single-channel: 30% (95% CI 0.22 to 0.38); (2) in relation to identification of patients with neonatal seizures: multi-channel: 100% (95% CI 0.76 to 1) and single-channel: 92% (95% CI 0.65 to 0.99)	Did not study seizures shorter than 10 s Multichannel aEEG is better than single channel EEG for the detection of seizure patterns However, the sensitivity of detection of individual seizures was unsatisfactory as only 39% were detected

Comment(s)

About 0.35% of newborn infants show seizure activity (Volpe)which may be subtle, short-lived or status epilepticus. There is increasing demand for the identification of seizure activity in the neonatal brain as it is associated with high mortality and morbidity (Scher 1989, McBride). Subclinical seizures may also occur after administration of the first antiepileptic drug, a phenomenon known as uncoupling (Scher 2003)or electro-clinical dissociation (de Vries). Interobserver variation in the clinical detection of seizures and difficulty differentiating them from non-seizure movements are common recognised problems.

The reference standard for the detection of neonatal seizures is the conventional continuous multi-channel EEG (cEEG). This requires application by trained personnel and expert interpretation by neurophysiologists. Apart from the non-availability of this service out of hours, the continuous nature of the recording and the complexities of electrode placement may also interfere with regular neonatal intensive care nursing procedures. The limitations of cEEG led to the use of alternative devices which were easy to apply, were easily available and could be used at the bedside immediately after a baby's admission to the neonatal unit (de vriers).CFM or aEEG fulfils these requirements and has been increasingly used as a bedside tool for seizure detection in neonatal units, but there are few data on accuracy (Hellström-Westas). The term aEEG is preferred as CFM refers to a particular monitor. Modern devices display simplified digital recordings from single or multiple channels and may also have simultaneous ‘raw’ EEG recordings. The raw EEG is usually a single-channel recording of waveforms which represent the brain's spontaneous electrical activity. The simultaneous recording of the raw EEG may help to differentiate artefacts arising from electrical or movement interference (Hagmann) and are felt to be more amenable to interpretation by trained non-experts.

The literature search identified five relevant studies.Methodology details were clear and all numbers were accounted for. The aEEG and cEEG tracings were subjected to off-line, blinded and independent analysis by multiple observers. The accuracy and sensitivity of seizure detection in high-risk neonates was determined by statistical analysis.

However, these studies also had some limitations. All studies were conducted in single centres. Sample numbers were variable, ranging from 19 to 121 babies and only one study described a power calculation (Shellhaas). The individual devices varied between the studies. Some babies were treated with antiepileptic medication in three studies (Toet, Shah, Bourez-Swart) and this may have influenced seizure discharge patterns. One study (Shellhaas) commented that the EEG recordings may have been brief but were felt to be representative of the short-term aEEG monitoring that is often used in clinical practice.

The studies consistently showed that aEEG was not highly sensitive to individual seizure detection. The sensitivity (95% confidence interval and a significant p value of <0.001) of neonatal seizure detection by aEEG is multifactorial (Shellhaas). However, recurrent seizure patterns or epilepsy can be diagnosed due to the high frequency of seizures which are picked up on the aEEG. The sensitivity of seizure detection can be optimised by longer durations of multi-channel aEEG recording with time compression and simultaneous raw EEG recordings. Although 81% of all seizures originate in the central, temporal or midline vertex electrodes (Shellhaas) a small group of patients with seizures originating in the frontal, occipital and temporal areas (which may have low amplitudes) may be missed (Toet, Bourez-Swart). The κ value or coefficient is a statistical measure of interobserver agreement for qualitative (categorical) items. The κ score is good among those with expertise in interpreting CFM traces, especially if simultaneous raw EEG tracings are performed. There is more variability among non-experts. There was poor correlation among non-expert users and one study (Shellhaas 2007)showed that neonatal seizures may still be difficult to detect for physicians with extensive experience.

The cEEG remains the mainstay for quantification of seizure burden. Two studies (Rennie, Bourez-Swart) concluded that single-channel CFM alone was not suitable for the diagnosis of neonatal seizures and Shah et al demonstrated that simultaneous recording of the raw EEG improved seizure detection rates. One study (Bourez-Swart) recommended that at least one cEEG should be performed whenever aEEG monitoring is performed, in order to supplement negative aEEGs, select appropriate aEEG channels and detect other unusual EEG patterns that may be mistaken for seizures. Therefore, aEEG may be useful as an adjunct and monitoring tool in suspected seizures and for monitoring the effects of anticonvulsant therapy after seizure diagnosis by cEEG. However, practical implementation of this strategy is not straightforward as cEEG may not be readily available in all neonatal units and the interpretation of the aEEG may be performed by non-experts.

Although simultaneous recording of the raw EEG provides an acceptable level of sensitivity of seizure detection among those with expertise in aEEG interpretation, larger studies involving less experienced neonatal clinicians are required to assess the feasibility and impact of monitoring for seizures using bedside aEEG techniques. Further research into the use of validated algorithms as a method of objective, computer-based seizure detection may help to improve seizure detection rates.

Editor Comment

The κ value is a statistical measure of inter-rater agreement for qualitative items. κ value <0.40 indicates low association, 0.40–0.75 indicates medium association, >0.75 indicates high association. Levels of evidence according to the Centre for Evidence Based Medicine, Oxford (www.cebm.net).

aEEG, amplitude EEG; cEEG, conventional EEG; CFM, cerebral function monitor; HIE, hypoxic ischaemic encephalopathy; NPV, negative predictive value; PPV, positive predictive value.

Clinical Bottom Line

aEEG is not as accurate as cEEG in the detection of neonatal seizures and further evaluation of possible seizure activity by cEEG is required. (Grade A)

aEEG can be used as a monitoring tool for seizures on the neonatal unit but caution must be exercised in its interpretation by non-experts. (Grade A)

References

1. Toet MC, van der Meij W, de Vries LS, et al . Comparison between simultaneously recorded amplitude integrated electroencephalogram (cerebral function monitor) and standard electroencephalogram in neonates. Pediatrics 2002;109:772–9.
2. Rennie JM, Chorley G, Boylan GB, et al . Non-expert use of the cerebral function monitor for neonatal seizure detection. Arch Dis Child Fetal Neonatal Ed 2004;89:F37–40.
3. Shellhaas RA, Soaita AI, Clancy RR . Sensitivity of amplitude-integrated electroencephalography for neonatal seizure detection. Pediatrics 2007;120:770–7.
4. Shah DK, Mackay MT, Lavery S, et al . Accuracy of bedside electroencephalographic monitoring in comparison with simultaneous continuous conventional electroencephalography for seizure detection in term infants. Pediatrics 2008;121:1146–54.
5. Bourez-Swart MD, van Rooij L, Rizzo C, et al . Detection of subclinical electroencephalographic seizure patterns with multichannel amplitude-integrated EEG in full-term neonates. Clin Neurophysiol 2009;120:1916–22.
6. Volpe JJ . Perinatal brain injury: from pathogenesis to neuroprotection. Ment Retard Dev Disabil Res Rev 2001;7:56–64.
7. Scher MS, Painter MJ, Bergman I, et al . EEG diagnoses of neonatal seizures: clinical correlations and outcome. Pediatr Neurol 1989;5:17–24.
8. McBride MC, Laroia N, Guillet R . Electrographic seizures in neonates correlate with poor neurodevelopmental outcome. Neurology 2000;55:506–13.
9. Scher MS, Alvin J, Gaus L, et al . Uncoupling of EEG-clinical neonatal seizures after antiepileptic drug use. Pediatr Neurol 2003;28:277–80.
10. de Vries LS, Hellström-Westas L . Role of cerebral function monitoring in the newborn. Arch Dis Child Fetal Neonatal Ed 2005;90:F201–7.
11. Hellström-Westas L. Comparison between tape-recorded and amplitude-integrated EEG monitoring in sick newborn infants. Acta Paediatr 1992;81:812–19.
12. Hagmann CF, Robertson NJ, Azzopardi D. Artifacts on electroencephalograms may influence the amplitude-integrated EEG classification: a qualitative analysis in neonatal encephalopathy. Pediatrics 2006;118:2552–4.
13. Shellhaas RA, Clancy RR . Characterization of neonatal seizures by conventional EEG and single-channel EEG. Clin Neurophysiol 2007;118:2156–61.