Anaesth Intensive Care 2012; 40: 275-284

An observational study exploring amplitude-integratedelectroencephalogram and Spectral Edge Frequency duringpaediatric anaesthesiaS. McKEEVER*, L. JOHNSTONt, A. J. DAVIDSON^Department of Anaesthesia and Pain Management, Royal Children's Hospital; Murdoch Children's Research Institute andMelbourne School of Health Sciences, University of Melbourne, Melbourne, Victoria, Australia

SUMMARYProcessed electroencephalography is used in adults to guide anaesthesia, but the algorithms used may notapply to infants. Knowledge of infants' electroencephalogram (EEG) responses to anaesthetics is fragmentary.An earlier pilot study suggested amplitude-integrated EEG (aEEG) may be a useful measure of anaestheticeffect. The aim of this study was to determine how aEEG changes between awake and anaesthetisedchildren of varying ages and to compare the response to that seen with Spectral Edge Frequency 90% (SEF90).A prospective observational study of children receiving a general anaesthetic was conducted. Anaestheticregimen remained at the discretion of the treating anaesthetist. EEG data were collected using the BrainZReBrim™ monitor using forehead and biparietal montages. SEF90 and aEEG were compared across agegroups, EEG montage and between awake and anaesthetised states.

A total of 178 children (aged 24 days to 14 years) were recruited. All aEEGs were greater duringanaesthesia compared to when awake and this difference varied with age. Only children older than two yearsshowed lower SEF90 while anaesthetised compared to when awake. SEF90 from children younger thansix months was higher during anaesthesia compared to when awake. Analysis of parietal and foreheadEEG montages revealed age-related differences. These findings suggest that SEF90 and aEEG candiscriminate between awake and anaesthetised states in older children. In younger children aEEG changesare less pronounced and SEF90 either cannot discriminate between states or responds paradoxically. TheaEEG may be marginally better than other EEG parameters in measuring anaesthetic depth in children.

Key Words: brain, monitoring, anaesthesia, age factors

In adults, the use of processed electro-encephalogram (EEG) depth monitors, such as theBispectral Index* (Covidien Inc., Boulder, CO, USA)and Entropy* (GE Healthcare, Buckinghamshire,UK), have been shown in some studies to beassociated with a reduction in anaestheticconsumption, recovery times and postoperativenausea and vomiting'-. However, the effect of theiruse on the incidence of intraoperative awarenessremains controversial". It is less clear if the currentlyavailable depth of anaesthesia monitors offerbenefits in the paediatric population'. Studiesattempting to validate these devices in children

• RGN, RN (Child), DipTropNurse, BSc (Hons), PhD Candidate.t RN, PhD, Professor and Head of the School of Nursing & Midwifery,

Queen's tJniversity, Belfast, United Kingdom.t MB, BS, FANZCA, GradDipEdBiostat, MD, Group Leader.

Address for correspondence: Mr S. McKeever, Department of Anaesthesiaand Pain Management, Royal Children's Hospital, Flemington Road,Parkville, Vic. 3052. Email: stephen.mckeever@rch.org.au

Accepted for publication on November 20, 2011.

Anaesthesia and Intensive Care, VoL 40, No. 2, March 2012

have shown that in older children they usually haveperformance characteristics similar to adults, butin children younger than two years of age there isincreasing evidence that the monitors do not behaveas they do in adults or older children"". Thesefindings are not surprising considering thedevelopmental changes that occur in the EEG withincreasing age'-", such as the gradual increase indominant awake frequency with age'^ This changein dominant frequency would be expected to beparticularly relevant to processed EEG parametersthat use the frequency domain.

The amplitude-integrated electroencephalogram(aEEG) is a digitised version of the CerebralFunction Monitor developed by Maynard andcolleagues". As a time domain parameter, aEEGanalysis involves EEG information being subjectedto an envelope detection algorithm. This timecompression method permits assessment of EEGamplitude changes'". Initially used to monitorcritically ill adults", the Cerebral Function Monitor

276 S. MCKEEVER, L. JOHNSTON, A. J. DAVIDSON

has also been shown to detect anaesthetic-inducedchanges in adults". More recently, the aEEG hasbeen adopted into neonatal practice"' for seizuredetection"* and long-term monitoring after hypoxicischaemic events'"^".

Unlike the aEEG, which uses the time domainof the EEG, the Spectral Edge Frequency (SEF)assesses EEG in the frequency domain. SEF is thefrequency in an epoch of EEG data below which aset percentage of the EEG power spectrum iscontained. Investigations of SEF during anaesthesiain the adult population have demonstrated consistentchanges-'.

A previous pilot study in infants--, found thatwhile there were no changes with emergence fromanaesthesia in the SEF 90% (SEF90), there was asuggestion that changes did occur in the aEEG. Thisstudy was limited, however, by its small sample size,only short periods of artefact-free recording andvery little recording of participants in the completelyawake state. Hence, further examination of theaEEG and SEF90 parameters was required. Wedecided that more evidence that aEEG doesrespond to anaesthesia in an observational setting isrequired before performing the more technicallyand ethically challenging studies to determinedose responses in children's aEEG and SEF90under controlled anaesthetic conditions. Thus, theaim of this prospective observational study was toinvestigate, in a number of different age groups, thechanges in aEEG between awake and anaesthetisedstates and compare these to changes in SEF90.A secondary aim was to examine potential EEGdifferences according to hemispheric location.

MATERIALS AND METHODSThe study was approved by the Royal Children's

Hospital Human Research Ethics Committee(Reference 28116A). Due to the exploratory natureof this observational study no power calculationwas undertaken prior to recruitment. Pragmatically,the number of cbildren chosen was the number tbatwas expected to be enrolled within one year. Asthe younger brain develops at a more rapid rate,recruitment of children was stratified into six agegroups and skewed toward younger children. Thesix age groups were: one day to six months, six to12 months, one to two years, two to four years,four to eight years and eight to 14 years. Inclusioncriteria were all children scheduled for generalanaesthesia. However, children were excluded ifthey had a known neurological condition or if thesensors would interfere with procedures such asfacial surgery or magnetic resonance imaging.

Informed, written consent was obtained froma parent of the child. Then if time and the childwould allow, EEG sensors were placed prior toinduction of anaesthesia. If this caused distress tothe child, the sensors were placed once the child wasanaesthetised. Sensor sites were C, C ,̂ P̂ P ,̂ andF J and F ^ according to the international 10-20system-\ and a reference electrode behind anear. BrainZ™ Hydrogel Sensors (Natus MedicalIncorporated, San Carlos, CA, USA) were appliedto the scalp after exfoliation of the site andapplication of a small amount of conductive paste.Sensors remained in situ until the child was awakeand about to leave the post anaesthesia care unit.

During this observational study, the choice ofanaesthetic was at the discretion of the treatinganaesthetist. Signal qualify data and raw EEGwere visible only to research staff for the durationof the anaesthetic. (Anaesthetic staff did not haveaccess to the EEG information as the monitor wasturned away from their line of view.) ProcessedEEG parameters were not available until lateranalysis. Collected demograpbic and perioperativedata included age of the child, procedure, any pre-existing medical illness, any regular medicationand use of sedative premedication. Once EEGrecording had commenced, the EEG tracingswere directly annotated. Relevant marked pointsincluded onset of anaesthesia, insertion of airwayand awakening.

SEF90 is obtained by subjecting raw EEG to aspectral analysis using a Fast Fourier Transform. Ina given epoch the SEF90 is the highest frequency(Hz) at which 90% of the power spectrum iscontained-^ Generating aEEG involves raw EEGbeing passed through an asymmetric filter-'. Thefiltered signal is then rectified to convert thenegative portion of the EEG signal to positive^\Utilising an envelope detection method, the rectifiedsignal is passed through a circuit that followsthe input peaks as they rise in amplitude-'. Thisprovides the maximum aEEG parameter. Once atthe top of the peak, the trace exponentially decaysuntil there is another peak in the rectified EEG.Tbe decay rate and the distance between EEGpeaks determine the minimum aEEG-''. Within thefrequency range 2 to 20 Hz the ReBrim'" monitor(BrainZ Instruments Limited, Auckland, NewZealand) calculates the aEEG and SEF90 fromfour second epochs-**.

EEG data were analysed offiine followingconversion to BrainZ Rescue Monitor™ format usingFile Converter version 1.03 (BrainZ InstrumentsLimited, Auckland, New Zealand). Once in BrainZ

Anaesthesia and Intensive Care, Vol. 40, No. 2, March 2012

SEF90 AND A E E G DURING PAEDIATRIC ANAESTHESIA 277

Rescue Monitor format, files were reviewed usingAnalyze Research 1.7 (BrainZ Instruments Limited,Auckland, New Zealand). SEF90, maximum andminimum aEEG, Mains Hum (HUM), Impedance(IMP) and electromyography (EMG) valuesalong with event marks contained in one minuteepochs were then exported to Microsoft* Excel2002 (Microsoft Corporation, Washington, USA).Combined data was then imported into Matlab*Version 7.7.0 (R2008b) (The MathWorks'", Inc.Natick, MA, USA) for filtering and analysis.Filtering consisted of removing periods of highIMP HUM and EMG artefact. A bespoke Matlabprogram was used to identify EEG data withperiods of IMP >10 kOhm, HUM >50 /xVpp andEMG > 10 IJ.V-. Once a period of excessive artefactwas identified this line of EEG data was removed.In addition, to reduce the impact of interferenceon adjacent lines of data, the program alsoremoved the minute before and after. Tbis filteringremoved contamination with artefact such aselectrocautery and movement.

EEG parameters were analysed when childrenwere 'awake' and 'during anaesthesia'. To obtain tbe'during anaesthesia' EEG data, a Matlab programwas written that examined the availability ofindividual patient's EEG data 10 minutes after theestablisbment of the artificial airway. If this portionof data had been removed, due to excessive artefact,the program then searched, at increasing minutes,for usable data until tbe time of airway removal.The presence of anaesthetic stability or surgicalstimuli was not ascertained.

To identify suitable 'awake' EEG data, the filteredEEG parameters were plotted against time forindividual patient files. These revealed periods ofpost-filtered EEG available for analysis. The rawEEG was then visually inspected, around theidentified timepoint, to identify one minute ofsuitable data. The data was rejected if it hadexcessive artefact, events marks indicating atypicalactivity such as the patient touching the sensors,or indicating the patient was asleep.

STATISTICAL ANALYSISAnalysis of children's data was undertaken

overall and within their six predefined age groups.Statistical software package STATA*, Release 11.0(STATA Corporation, College station, Texas, USA)was utilised for the statistical analysis. Assessmentof awake and anaesthetised EEG data was made onone minute epochs.

Assessment of the influence of montage onaEEG and SEF90 recordings comparisons of F ,-F ,Anaesthesia and Intensive Care, Vol. 40, No. 2, March 2012

(forehead), C.-P, (left) and Ĉ -P̂ (right) montageswere made using a non-paired Student's t-test. AStudent's t-test was also conducted on the age-grouped data.

To identify the correlation coefficient betweenEEG parameters and age, a linear regressionanalysis was used. This analysis was applied onvalues obtained while awake and during anaesthesia.Initially regression analysis was conducted on allage groups then post hoc on children under andover two years of age.

EEG parameters obtained during awake andanaesthesia for the different ages were comparedwith a non-paired Student's t-test. The influenceof age and timepoint was analysed using a two-wayanalysis of variance. Throughout the analyses aP value of

278 S. MCKEEVER, L. JOHNSTON, A. J. DAVIDSON

HUM >50fiVpp and EMG >10/xV"^ demonstratedan acceptable preservation of data,

EEG recordings obtained while the children'sbreathing was maintained with an artificial airwayhad a median duration of 33 minutes (range 6 to153 minutes). Total anaesthetised samples usedwere forehead n=118, and parietal n=122. Giventhe varied time duration of the procedures the postfiltered data available for analysis reduced withtime. Utilisation of the timepoint 10 minutes afterthe establishment of the artificial airway to start toretrieve data for subsequent analysis correspondedwith peak availability of post filtered data. Thus,a majority of the data was obtained from tbis

10 minute timepoint (forehead 110 samples andparietal 106 samples).

Hemispheric comparisons of EEG parameters whenawake and anaesthetised

SEF90 and aEEG values obtained from left andright hemispheres during anaesthesia or awake forall age groups showed no significant differencebetween left and right montages {P >0.4). Sub-sequently, the mean of the left and right EEG datawas used for analysis as a parietal value.

When aEEG values from all cbildren wereanalysed, whether awake or anaesthetised, thoseobtained from the forehead montage were lower

Age in days, tnean (SD)

Gender, M/F

Procedure

Gastrointestinal endoscopy

Penile/genital surgery/eystoscopy

Hernia repair

Orthopaedic

Plastic surgery

Ophthalmic

ENT

Intra-abdominal stirgery

Bone marrow harvest

Midazolam premedication

Anaesthetic variables

Propofol

N,O

Airway: facemask/LMA/ETT

Inhaled anaesthetics

Sevo only

Iso only

Sevo to Iso

Sevo to Iso to Sevo

Opioids

Neuromuscular blockage

Blocks

Caudal/femoral/nerve/umbilical

SEF90 AND A E E G DURING PAEDIATRIC ANAESTHESIA 279

than those from the parietal area (P

280 S. MCKEEVER, L. JOHNSTON, A. J. DAVIDSON

to 14 years. While anaesthetised, the differencebetween the parietal and forehead aEEG ofchildren changed with increasing age, as shownin Table 2. Children younger than four yearsdemonstrated higher aEEG values from the parietalcompared to forehead montages {P

SEF90 AND A E E G DURING PAEDIATRIC ANAESTHESIA 281

between SEF90 and age in the forehead montage(r-'=0.17,P=O.0O03).

Changes in aEEG and SEF90 while awake accordingto age

EEG sensors were applied pre-anaesthesia to 168children. These recordings had a median durationof 48 minutes (range 4 to 330 minutes). Awakerecordings showed a large amount of contaminationfrom artefact such as electromyogram, blinkingand movement. This was particularly evident inforehead regions with only 21 (13%) awake datapoints available for analysis. From the parietal

recordings there were 93 (55%) useble awake datapoints.

Overall, aEEG from awake cbildren showed apoor degree of correlation with age in the forehead(r- 0.4) and parietal (r̂ 0.1)montages as shown in Figure 2A and B. Whenthe children younger than two years of age wereanalysed independently there was a consistentlyhigher degree of correlation between aEEGparameter and age in the forehead (r- >0.3, P 0.28, P < 0.0001) aEEG values.

SEF90 of all children showed a moderate to largedegree of correlation with age in the samples taken

A)aEEG (max) Parietal

o.in

o -

è

24d-6m 6-12 m 1-2 y 2-4 yAge group

4-8 y 8-14 y

aEEG (min) ParietalB)

o ,o •

O-r

o - f . * . * . *24d-6m 6-12 m 1-2 y 2-4 y

Age group4-8 y 8-14 y

C) o .

N O .

O

24d-6m 6-12 m 1-2 y 2-4 y 4-8 y 8-14 yAge group

Awake Anaesthetised

FIGURE 3: Box-and-whisker diagrams showing the parietal amplitude-integrated electroencephalogram (aEEG)and Spectral Edge Frequency 90% (SEF90) at awake and anaesthetised according to age group.

Anaesthesia and Intensive Care, VoL 40, No. 2, March 2012

282 S. MCKEEVER, L. JOHNSTON, A. J. DAVIDSON

from parietal montages (r-=0.5, P

SEF90 AND A E E G DURING PAEDIATRIC ANAESTHESIA 283

The lack of evidence for a difference in SEF90between awake and anaesthetised state in infantsis consistent with the poor performance, in this agegroup, of depth monitors such as Bispectral Indexthat predominantly rely on analysis of the frequencydomain. Unlike SEF90, aEEG did show someevidence for a difference between anaesthesia andawake state in infants. The difference was marginalbut this finding does suggest further more controlledstudies should be done to explore the possibilitythat aEEG eould be used to measure anaesthesiadepth in infants. It should be noted that thefull complexities of EEG changes that oecurduring anaesthesia ' are not accounted for in thisalgorithm. At higher doses of some anaesthetics,burst suppression occurs, while others provokehigher frequency beta activity". These changes willnot be accounted for in single dimension algorithmssuch as SEF90 or aEEG. Future investigations willbe required to assess age-related changes to burstsuppression patterns or beta activation.

For the duration of this study the choice ofanaesthetic remained at the anaesthetist's discretion.It is recognised that this introduces a hetero-geneity which might reduce the power of thisexplorative study. A more controlled anaestheticregimen would be better suited to determine theexact relationship between anaesthesia and EEG atdifferent stages. However, an anaesthesia protocolwhich preseribes the anaesthetic is more difficultto perform and raises ethical issues if the protocolis not in accordance with what the treating doctorconsiders is in the child's best interest. Thus, beforedesigning such a study it was decided to performthis exploratory observational study where theanaesthetic was not prescribed, to establish directionsfor future investigations. There is a possibility thatour comparisons between ages may be infiuencedby factors not controlled for in this study such as useof premedication, varying dose of agents, differentagents and varying surgical stimuli. Future studieswill need to control for these variables.

In addition, due to the exploratory nature of thisstudy, no sample size calculation was undertakenprior to recruitment. The lack of available research,regarding the EEG of children during anaesthesia,meant that relevant differences in the EEG werenot available to complete a power calculation. Thus,there will be uncertainty regarding the true scientificsignificance of the reported findings. However, thisstudy provides preliminary information that willhelp to inform future investigation into the EEGof children during anaesthesia.

Aniicsthvsiu nut! Intensive Care. Vol. 4tK No. 2. Manh 2012

Another limitaticMi of this study is theasymmetrical filter used by the ReBrim monitorprior to processing EEG processing. This filterstrongly attenuates activity below 2 Hz and above20 Hz-\ EEG activity relevant to the monitoringof EEG during anaesthesia may oecur outsidethese ranges such as information in the delta,higher beta'' and gamma' frequencies. However,examination of the EEG within the 2 to 20 Hzrange reduces artefact contamination and reducesthe signal-to-noise ratio". Future studies will needto establisb tbe clinical relevance of paediatrieEEG information outside the 2 to 20 Hz range andovercome the challenges of ensuring EEG data isnot contaminated with artefact.

This study also found that during anaesthesia,children younger than four years of age have loweraEEG from forehead regions compared to theparietal recordings, and in older children theforehead SEF90 was significantly higher thanparietal reeordings. The findings of this study implythat choice of montage is important when choosingpotential EEG parameters to measure anaesthesiadepth in children. However, the use of three EEGmontages is also a limitation of this study. Attemptsto obtain larger montage recordings may bave led toincreasing distress to awake children (entertainingchildren, and attaching seven sensors, then keepingthe sensors in situ while the child was awake waschallenging). Much of the awake EEG data wascontaminated with artefaet and removed by thepost hoc filtering. This artefact was particularlyevident in the forehead region due to blinking andfacial movements. To improve the quality of theawake data providing an area where the childrencould be quiet with eyes closed may have revealedcleaner EEG data,

CONCLUSIONThis exploratory study of a heterogeneous

clinical population found that in older ehildienthere are clear differences in aEEG and SEF9Übetween awake and anaesthetised states. Ininfants the difference in aEEG between awakeand anaesthetised is less clear, while SEF9() is anineonsistent discriminator of anaesthetic state.These findings support further investigation ofaEEG as a measure of anaesthesia depth in infants.

ACKNOWLEDGEMENTSThis project was supported by the Victorian

Government's Operational Infrastructure SupportProgram. BrainZ instrtiments/Natus provided salary

284 S. MCKEEVER, L. JOHNSTON, A. J. DAVIDSON

support for data collection, loan of ReBrimmonitor and consumables. S. McKeever receivedPostgraduate Health Research Scholarships fromthe Murdoch Childrcns Research Institute,Melbourne and a Trust Fund Faculty ResearehScholarship from the Faculty of Medieine. Dentistryand Health Sciences at The University of Melbourne.

REFERENCES1. Punjasawadwong Y, Boonjeungmonknl N. Phongchiewboon

A. Bispcctral Index for improving anaesthetic delivery andpostoperative recovery. Coehrane Database .Syst Rev 2(107:

10.

II

2. Bein B. Entropy. Best Pract Res Clin Anaesthesiol 2006:20:101-109.

3. Avidan MS. Jacobsohn E. (jlick D. Burnside BA. Zhang L,Villalranca A et al. Prevention of ¡ntraopcrative awareness in ahigh-risk surgical population. N EngI J Med 2011: 365:591-600.

4. Myles PS, Leslie K. McNeil .1. Forbes A. Chan MTV. Bispectralindex monitoring to prevent awareness during anaesthesia: theB-Aware randomised controlled trial. Lancet 2004: 363:1757-1763.

5. Avidan MS. Zhang L. Burnside BA. Finkel KJ. SearlemanAC. Sclvidge JA et al. Anesthesia awareness and the bispectralindex. N EngI J Med 2008: 358:1097-1108.

6. Davidson AJ. Monitoring the anaesthetic depth in children -an update. Curr Opin Anaesthesiol 2007: 20:236-243.

7. Edwards JJ. Soto RG. Bedford RF. Bispectral Index values arehigher during h:ilothane vs. sevotlurane anesthesia in children,but not in infants. Acta Anaesthesiol Scand 2005: 49:1084-1087.

8. Murat I. Constant 1. Bispectral index in pediatrics: fashion or anew tool? Paediatr Anaesth 2005; 15:177-180.

9. Watcha MF. Investigations of the bispectral index monitorin pédiatrie anesthesia: first things first. Anesth Analg 2001:92:805-807.Shandcr A. Lobel G. Caveats of bispectral index monitoring inthe pediiitric population. Chest 2005: 128:14-16.Rodriguez RA. Hall LE. Duggan S. Splinter WM. The bispec-tral index docs not correlate with clinical signs of inhalationalanesthesia during sevoflurane induction and arousal in chil-dren. Can J Anesth 2004; 51:472-480.

12. Bell MA. The Ontogeny of the EEG During Infancy andChildhood: Implications for Cognitive Development. In:Garrean B. editor. Neuroimaging in Child NeuropsychiatrieDisorders. Berlin. Springer-Verlag 1998. p. 97-111.

13. Bell MA. Fox NA. Brain Development Over the First Year ofLife: Relations Between EEG Frequency & Coherence andCognitive & Alfeetive Behaviors. In: Dawson G. Fischer KW.eds. Human Behavior and the Developing Brain. New York.Guilford. 1994. p. 314-345.

14. Matousek M. Petersén I. Frequency Analysis of the EEG inNormal Children and in Normal Adolescents. In: Kellaway P.Petersén I. eds. Automation of Clinical Electroencephalography.New York. Raven Press. 1973. p. 75-102.Maynard D. Prior PF. Scott DF. Device for continuous moni-toring of cerebral aetivity in resuscitated patients. Br Med J1969;4:54.'i-546.Ilellstriini-Westas L. de Vries LS. Rosen 1. Atlas of Amplitude-Integrated EEGs in the Newborn. 2nd ed. London. InformaHealtheare 2006.

17. Prior P. Monitoring in Anaesthesia. In: Prior P. ed. MonitoringCerebral Function: Long-term Recordings of CerebralElectrical Activity. Amsterdam. Elscvier 1979. p. 69-1II.

15

16

18. Shany E. Khvatskin S. Golan A. Kurplus M. Amplitude-integrated clectroeneephiilogriiphy: a tool for monitoring silentseizures in nconates. Pediatr Neuiol 20(16: 34:194-199.

19. Hellstroni-Westas L. Rosen I. Svenningsen NW. Predictivevalue of early continuous amplitude integrated EEG record-ings on outcome after severe birth asphyxia in full term infants.Arch Dis Child Fetal Neonatal Ed 1995: 72:F34-F38.

20. Wertheim D. Mercuri E. Faundez JC. Rutherford M. Acolet D.Dubiiwitz L. Prognostic value of continuous clectroencephalo-graphic recording in lull term infants with hypoxic isehaemicencephalopathy. Areh Dis Child 1994: 71:F97-FI(I2.

21. Tonner PH. Bein B. Classic electroeneephalographie param-eters: median frequency, spectral edge frequency etc. BestPract Res Clin Anaesthesiol 2006: 20:147-159.

22. Davidson AJ, Sale SM, Wong C, McKeever S. Sheppard S.Chan Z et al. The electroencephalcigraph during anesthesiaand emergence in infants and ehildren. Paediatr Anaesth 200

Copyright of Anaesthesia & Intensive Care is the property of Australian Society of Anaesthetists and its content

may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express

written permission. However, users may print, download, or email articles for individual use.