Best Practice & Research Clinical Anaesthesiology

Vol. 20, No. 1, pp. 147–159, 2006

doi:10.1016/j.bpa.2005.08.008
available online at http://www.sciencedirect.com

13

Classic electroencephalographic parameters:

Median frequency, spectral edge frequency etc

P.H. Tonner* MD
Professor of Anaesthesiology

B. Bein MD
Staff member
Department of Anaesthesiology and Intensive Care Medicine, University Hospital Schleswig-Holstein, Campus Kiel,
Schwanenweg 21, D–24105 Kiel, Germany

Even today many anaesthesiologists rely on parameters of the autonomic nervous system, such as
blood pressure and heart rate to decide if a patient is adequately anaesthetized. It is thought that
the electroencephalogram (EEG) may provide more information on the state of anaesthesia.
Because full EEG analysis is not possible in the operating room, processed EEG parameters have
been developed comprising complex information into a single value. Time and frequency domain
parameters are calculated. The power spectrum results from a Fourier analysis and can be
described by parameters such as median frequency, spectral edge frequency and others. It was
noted, however, that anaesthetics at low doses increase frequency of the EEG, whereas at high
doses the EEG is depressed. This biphasic response makes it difficult to clearly distinguish the
exact anaesthetic state of a patient. Median frequency and spectral edge frequency have been
studied in numerous studies. However, no sole indicator has been derived from the EEG that
could serve as a descriptor of anaesthetic depth.

Key words: depth of anaesthesia; electroencephalogram; median frequency; spectral edge

frequency; bispectral index.

INTRODUCTION

Even today many anaesthesiologists rely on parameters of the autonomic nervous

system, such as blood pressure and heart rate to decide if a patient is adequately
anaesthetized. However, the relationship between autonomic responses and cerebral
activity has not been well characterized during an operation especially when adjunct

* Corresponding author. Tel.: C49 431 597 2973; Fax: C49 431 597 2973.
E-mail address: [email protected] (P. H. Tonner).

1521-6896/$ - see front matter Q 2005 Elsevier Ltd. All rights reserved.
148 P. H. Tonner and B. Bein

drugs such as sympatholytics are used.1 The electroencephalogram (EEG) may provide
a better measure for intraoperative awareness than autonomic responses. Since
Berger’s classic studies in 1930 demonstrating a change in cerebral electrical activity
depending on eye opening and closure EEG was studied as a means to assess the
wakefulness of a subject.2 It was accepted early that a simple real-time descriptor of
near consciousness as determined by EEG would be a potential boon, especially for
avoiding awareness of intraoperative events.3
The information contained in an intraoperatively derived EEG signal contains
changes in the potential of a single frequency, however, it also consists of a multitude of
rapid and slow frequencies changing potentials. Thus, the information contained in an
EEG signal is very complex (Figure 1). This multitude of information in the EEG is very
hard to compute by a clinical anaesthesiologist. A registration and analysis of an EEG
throughout the entire intraoperative period is not feasible. Fortunately computing
power has increased dramatically during the last two decades (Moore’s law predicts a
doubling of processing power every 2 years) thus making it possible to perform
extensive calculations on the raw EEG in real time resulting in so called processed EEG
parameters. These parameters include a large amount of information in a compressed
form often presented as a single parameter. Intraoperatively a limit can be assigned to a
low or a high value allowing for continuous monitoring and alarming the
anaesthesiologist if preset values are not met.
However, one needs to be cautious with processed EEG parameters. With their
increasing clinical availability more and more anaesthesiologist use commercial
monitors without relating the values offered to the raw EEG. Most currently available
monitors use a simplified version of the EEG by utilizing a reduced number of
electrodes compared to a ‘full’ 16–20 lead EEG. In addition, the information is only
gathered from one hemisphere thus omitting information on the difference of
processing in both hemispheres. Even today the gold standard of EEG monitoring
remains the 16–20 lead EEG plotted on paper and evaluated by an experienced
electroencephalographer.4,5

AwakeEEG

Beta activation
Sedation/hypnosis

Slowing (delta shift)

Near burst suppression

Isoelectricity

Figure 1. Changes in the electroencephalogram (EEG) dependent on sedation/hypnosis.

 Classic electroencephalographic parameters 149

TIME DOMAIN AND FREQUENCY DOMAIN ANALYSIS

Complex information contained in a waveform such as the EEG or heart rate

tachogram may be analysed basically in two ways: in a time domain or in a frequency
domain. After derivation of the analogue EEG signal the signal is digitized for further
analysis by a computer program through an analogue-digital converter.
Already as early as 1950 a parameter for EEG analysis in the time domain has been
proposed for the control of ether anaesthesia.6 Since then several time domain based
parameters have been suggested such as the zero crossing frequency (ZXF). This
parameter describes the number of times the EEG curve crosses the baseline during a
pre-defined period of time. The number of zero crossings is high in awake subjects and
decreases with increasing anaesthetic depth. A correlation of the dose of an anaesthetic
administered and the zero crossing frequency has been described for propofol.7 A value
of more than 10 Hz has been shown to be indicative of only a light anaesthesia.
However, the method has some limitations. Not all waves contained in an EEG signal
actually cut the baseline thus some information may be missing. In addition, because the
EEG signal is very complex it consists of multiple frequencies. Thus, the zero crossing
frequency will shift from epoch to epoch and not always result in a meaningful
measurement.
A more advanced method of analysis is the aperiodic analysis of the EEG. The raw
EEG is separated in two bands from 0.5 to 7.9 Hz and from 8 to 29.9 Hz which are then
further analyzed. This method determines the time between two consecutive minima
of the waveform, allowing a frequency to be calculated. Although this method is limited
because of the simplistic nature of the analysis it was found to be a useful parameter in
some types of sedation.8,9

FOURIER TRANSFORMATION

Fourier transformation is also called spectral analysis because an EEG signal is

dissected in its component spectra. In principle, this method allows for the detection
of underlying sinus waves that in their sum result in the complex wave detected. The
result of this analysis is a power spectrum where the power is plotted versus
frequency (Figure 2). Early methods have used relatively simple parameters for the
analysis such as the peak frequency, the median frequency or the spectral edge
frequency (Figure 3).
For a processed EEG data are recorded for a preset period of time, then the
algorithm such as a Fourier transform is applied and the data are displayed in various
different ways. In monitoring depth of anaesthesia a fast update of information is
preferable for fast decision making of the clinical anaesthesiologist, however, one has to
keep in mind that shortening the data acquisition period (epoch) leads to a reduced
number of data points up to a value were Fourier transformation is meaningless. The
shorter the epoch the more differences between the epochs occur rendering
interpretation of the information increasingly difficult. In a study of different length of
epochs useful for monitoring depth of anaesthesia it was demonstrated that epoch
lengths of 2 s are feasible for this type of monitoring, today epochs of 4–8 s are more
common.10 During the epoch the EEG waveform is analyzed and dissected into its
component frequencies. For each frequency a correlation analysis is performed with
the EEG waveform and the covariance is calculated. The sum of each sinus wave yields
150 P. H. Tonner and B. Bein

Electroencephalogram Power spectrum

Fourier Transformation

Figure 2. After the raw EEG has been digitized a Fourier transformation is performed that is graphically
displayed as a power spectrum.

the amplitude and the phase length of the analyzed epoch. The graphic display of the
Fourier transformation will result in the power spectrum. Different means have been
derived to analyse power spectra. Thus, one may analyse the area under the curve
(AUC) for a specified frequency range. As a raw EEG is described as a sum of
frequencies some ranges have been defined for use in clinical applications. The alpha
band of the spectrum includes a range from 8 to 12 Hz, the beta band from 12 to 30 Hz,
the theta band from 4 to 8 Hz and the delta band from 0.5 to 4 Hz (Figure 3). Using
these frequency ranges an EEG can now be described by the AUC for the defined
frequency bands. However, the frequency bands have been developed for analysis of the
awake EEG. Changes in different frequency bands may thus be difficult to interpret
during anaesthesia. Quantitative derivation of depth of anaesthesia was initially

Median frequency Spectral edge frequency (SEF90)

Delta Theta Alpha Beta

Power

5 10 15 20 25 30
Frequency (Hz)

Figure 3. Schematic representation of a power spectrum. Traditionally four frequency bands have been
defined for the awake EEG (alpha, beta, theta, delta; see text). The median frequency is the frequency that
divides the area under the curve in half. The spectral edge frequency has several definitions: here the SEF90 is
shown, which divides the power into 90% and 10%.
 Classic electroencephalographic parameters 151

performed by applying the integrated total power.6,11–13 However, total power was
demonstrated not to be a reliable parameter of depth of anaesthesia in a range of
settings.14 With the development of computing power frequency dependent EEG
derivations were increasingly coming into focus such as the median frequency15–17 or
the spectral edge frequency.18–20
A simple Fourier analysis does not take into account phase relations between
component waveforms. The analysis of phase relations is nowadays possible with faster
processor technology and was implemented in the bispectral analysis. Although some
comparisons to this recent EEG measure as well as to others were undertaken for the
classical EEG parameters the bispectral index will not be discussed in detail here.

PEAK FREQUENCY

Each power spectrum consist of several peaks and valleys. The peak frequency analysis
calculates the frequency which has the greatest power. Peak frequency analysis is a
simple measure of describing the power spectrum and can even be relatively validly
applied by visual control of the power spectrum. It was demonstrated that peak
frequencies shift from higher frequencies in awake subjects to lower frequencies in
subjects who are asleep or under anaesthesia. However, peak frequency is not always
reliable as the power spectrum may display several peaks of approximately similar
power and it may not be clear cut which frequency best describes the state of
consciousness subjects are in. Accordingly peak frequency was abandoned as a measure
for depth of anaesthesia.

MEDIAN FREQUENCY

The median frequency is defined as the frequency of a power spectrum at which 50% of
the power are at lower frequencies and 50% of the power are at higher frequencies
(Figure 3). Thus, the median frequency provides a simple means for analysing the power
spectrum and can easily be calculated. Accordingly, the median was one of the first
processed EEG measures that was introduced to clinical use. Typically median
frequencies of awake patients are at a level of 12 Hz. During anaesthesia the median
frequency shifts to lower levels and frequencies of about 2–3 Hz are found.21 The
median frequency was used by several investigators to quantify depth of anaesthesia
with various anaesthetics and it was found to perform better compared to spectral
edge frequency when etomidate was used.16 Schwilden et al. conceded for their studies
on EEG feedback controlled anaesthesia that a complex multimodal distribution like the
EEG power spectrum cannot be described by a single value, however, they also
concluded that gross shifts in the power spectrum can reasonably be determined
utilizing the median frequency.17

SPECTRAL EDGE FREQUENCY

Similar to the median frequency the spectral edge frequency is calculated from the
power in frequency ranges. The AUC of the power spectrum is calculated first.
Following the frequency is derived that separates 90% of the area from 10% of the
152 P. H. Tonner and B. Bein

area the so called SEF90 (Figure 3). Another measure, the SEF95 is derived in a
similar fashion by calculating the frequency at which power is separated into 95 and
5%. Hudson suggested to use this 95% quantile of the power spectrum as a more
stabile and reliable measure of the spectral edge.18 One method of determination of
the spectral edge frequency was introduced by Rampil.22 The power spectrum is
searched for the highest frequency contiguous 2 Hz wide band of activity. Each
frequency bin of the spectra within this band contains more than a threshold
quantity of power. The upper frequency limit of this band of activity is then
designated as the SEF.23 Both, the median frequency and the spectral edge frequency
rely heavily on filters that are used on the raw EEG signal. Although the SEF has
been suggested to be a reliable measure of depth of anaesthesia it is highly
dependent on minor activities in the high frequency bands and reflects poorly the
centre of the power spectrum distribution and the activities in the low frequency
bands.17 In addition, some of the early monitor systems filtered out the low
frequency bands because it was thought that most movement artefacts induce
artificial low frequency activity. However, since these bands comprise more than 60%
of the total activity during anaesthesia it may be disadvantageous to neglect these
frequencies altogether.17 Because the SEF compresses the EEG waveform into a
single number, information on the slower EEG frequencies is lost. However, SEF has
been demonstrated to correlate closely with serum concentrations of thiopental,
etomidate, and fentanyl as well as with the end-tidal concentration of halothane.16,18,19
One of the major problems with SEF is that major shifts of frequencies may occur
without a change in SEF. Thus, it can remain stable if 5% of the power of high
frequency components is stationary while shifts between the alpha band (8–13 Hz)
and lower frequencies occur.24 SEF can hide the shift of EEG activity from alpha band
to low frequency bands.

RELATIVE DELTA POWER

The relative delta power is derived from the power of a defined frequency range of the
power spectrum. As the delta band is located at the lower end of the power spectrum
between 0.5 and 4 Hz, the power of this range is calculated and displayed as a
percentage of the total power. However, similar to the spectral edge frequency several
different definitions of this parameter exist. Other ratios have also been utilized such as
the theta ratio.

LIMITATIONS AND CLINICAL APPLICATIONS

As can readily be derived from the calculation of the frequency domain parameters all
measures rely heavily on the raw EEG. Thus, in order to provide correct numbers, the
raw EEG has to be absolutely free of artefacts and no shifting of the base line should occur.
Muscular activity will influence these measures because no filters are applied in order to
separate this electrical activity from cerebral electrical activity. Similarly electrical devices
may disturb the analysis by their inherent 50 or 60 Hz frequency. Electro-cautering by the
surgeons introduces artefacts that are almost impossible to filter. During clinical use of
EEG monitoring it is often not possible to completely exclude artefacts and faulty
measurements will occur. Even when the frequency range analyzed is kept relatively small
 Classic electroencephalographic parameters 153

(from 0.5 to 30 Hz) the descriptors will only provide a crude measure of the distribution
of the power in the spectrum and rapid changes may occur.
Apart from these technical difficulties there are also physiological difficulties with
these measures. At the most simple level an ideal parameter describing the conscious
state of a subject should be able to assign a single value to each state. However, this is
not the case with these simplistic measures. In general, anaesthetics change the
appearance of the raw EEG. When an anaesthetic is administered to a subject a phase of
cortical arousal is apparent at low doses (known as the state of excitation from the
Guedel scale for ether). The arousal is accompanied by an increase in the power of high
frequencies of the EEG. When the dose of the anaesthetic is increased the power of the
high frequencies decreases and lower frequency ranges increase in power. With
increasing depth of anaesthesia one will find a decrease of alpha activity in occipital
regions whereas the beta activity in frontal regions increases. Concurrently the
amplitude of the signal increases and the frequencies shift to lower values. During
general anaesthesia alpha activity is mainly found in frontal regions and no longer in the
occipital regions. At very deep stages of anaesthesia the frequencies slow down further
until no more activity can be detected. Thus, the median and the spectral edge
frequency may increase during light anaesthesia and then decrease at deeper levels, at
times showing similar values in anaesthetised subjects compared to awake subjects.
The initial increase leading to a biphasic effect in cerebral electrical activity has for
example been demonstrated in a study on the intravenous anaesthetic thiopental.9,25 At
low concentrations thiopental induced an activation of the EEG, with an increase in the
number of waves per second. This increase was followed by a decrease in the number of
waves with increasing concentrations of thiopental. At high concentrations of
thiopental an isoelectric EEG was seen.25 Although it seems that the biphasic response
on the EEG also seen with other anaesthetic drugs prohibits the use of EEG derived
parameters in determination of anaesthetic depth, it was demonstrated that once a
sufficient depth of anaesthesia was reached, simple parameters such as the number of
waves per second in an aperiodic waveform analysis may help to quantitate depth of
anaesthesia.25 However, it was also demanded early on that an EEG parameter needs to
be found that displays a monophasic response to anaesthesia.
When very deep levels of anaesthesia are reached a so called burst suppression
occurs in the EEG. Bust suppression is characterised by periods of electrical silence of
the EEG interspersed by periods of high activity (bursts). Some anaesthetics such as
barbiturates are able to suppress these bursts at high doses until the EEG is completely
suppressed. As long as bursts are present the resulting median and spectral edge
frequency may increase thus indicating a higher level of consciousness although very
deep levels of anaesthesia have been attained. Administration of higher doses of
anaesthetics in order to reduce the median or the spectral edge frequency to ‘less
awake’ levels may result in overdosing of patients.26,27 These changes apply more or less
to all general anaesthetics, however, one has to be aware that different anaesthetics
have different effects on the EEG activity making it impossible to relate directly the
effect of an anaesthetic on the EEG to a physiological effect.
In order to establish a valid measure of depth of anaesthesia the obtained values
need to be compared to patient responses to stimuli at constant drug concentrations.
The classic parameter for determination of anaesthetic depth is the minimal alveolar
concentration (MAC).28 Using the MAC as a physiological measure of anaesthesia has
been criticised, because movements are caused by spinal rather than cortical
mechanisms. Movements of a patient may also occur when a peripheral stimulus
such as laryngoscopy or intubation induces spinal or brain stem reflexes. In contrast
154 P. H. Tonner and B. Bein

movement may also be caused centrally during light anaesthesia when the central
nervous system is not adequately suppressed. In addition, the MAC concept only
applies to inhalational anaesthetics, a similar physiological measure of anaesthetic depth
is lacking for intravenous anaesthetics. Determination of the concentration of
intravenous anaesthetics is not readily available during routine anaesthesia but only
under experimental conditions. The quest for the right physiological response to
compare and validate EEG parameters is not over.
Currently, in routine anaesthesia practice a mixture of various substances is used
such as sedative/hypnotics, opioids, and muscle relaxants. Concurrent medication or
perioperative medication often includes the use of beta-blockers, or alpha2-adrenergic
compounds. Consequently, the clinical signs of depth of anaesthesia such as movements
or hemodynamic responses are not very reliable. However, the EEG response to
anaesthetics is less affected by drug mixtures and EEG parameters may be better
indicators of anaesthetic depth compared to classical clinical parameters. Intraopera-
tive monitoring of the EEG has been proposed for the assessment of depth of
anaesthesia because the EEG may provide continuous measures of changes in
pharmacodynamics.23,24,29–31
Rampil et al. found that a correlation exists between the EEG and the hemodynamic
response to laryngoscopy. When the SEF was held below 14 Hz under thiopental
anaesthesia the blood pressure response to direct laryngoscopy was minimal, whereas
at a SEF of more than 14 Hz the blood pressure response may be exaggerated.23
However, it was cautioned, that because other anaesthetics cause different EEG
patterns, SEF criteria different from those used with thiopental may be required. In
contrast to the findings by Rampil et al. White and Boyle1 did not observe consistent
changes in SEF associated with hemodynamic changes after induction of anaesthesia.
In a study of several EEG parameters including the power spectrum, SEF95, median
frequency, and the ratio of power in the 8–20 Hz frequency range to the power in the
1–4 Hz frequency range, the so called delta ratio, it was demonstrated that no one
descriptor of EEG activity was superior in anticipating when patients would respond by
opening their eyes.27 However, it was shown that awakening was always presaged by an
abrupt and marked increase in the delta ratio. This reaction was much less marked with
fentanyl and a more graded change was described.17,19,27 Compared to the delta ratio
the SEF95 and the median frequency were seen to change gradually during emergence
from anaesthesia when isoflurane was discontinued.27
Using the median frequency as the main parameter to guide anaesthesia delivery it
was demonstrated that under isoflurane-nitrous oxide a median frequency of !5 Hz
indicated anaesthesia whereas with median frequencies of more than 5 Hz patients
were awake.17 In a different study on feedback control of methohexital anaesthesia a
median frequency of 2–3 Hz was chosen as the desired set-point.24
Similar to work on median frequency performed by Schwilden and Stoeckel17 it was
found that a threshold value of 5.2–6.2 Hz exists for median frequency at which
awareness recurs although different equipment and a slightly different study design was
used.32 In addition, spectral edge frequency was able to differentiate intraoperative and
prearousal states at a threshold of 13.5 Hz reasonably well.32 Confirming previous data
it was also seen, that a delta shift occurred before arousal and during emergence of
anaesthesia, although this shift was not seen in all patients and was at times only gradual
in the later study.27,32
Interestingly it was observed during a closed loop feedback control study with
methohexital that in some cases the median frequency stayed at low median
 Classic electroencephalographic parameters 155

frequencies for about 15–30 min without methohexital being administered.24 Thus, it
has to be concluded that the interaction between methohexital concentration, which
was assumed to change on a much more rapid time scale, and median frequency is at
least not in all situations very close.
Schwilden and Stoeckel also observed a biphasic response of median frequency and
SEF95 during isoflurane and nitrous oxide anaesthesia.17 Both measures decreased with
increasing depth of anaesthesia, only at low anaesthetic concentrations during
induction and recovery an activation of high frequency bands occurred leading to
increases of median frequency and SEF9517 (Figure 4). Although there was a tendency
that values for both parameters were lower at 1.5 MAC than at 1.3 MAC there was a
substantial overlap due to individual variability making it impossible to predict for a
given EEG epoch whether the value was determined at the higher or the lower
concentration of the anaesthetic.17 However, when median frequency and SEF95 were
compared the median frequency displayed a greater degree of discrimination between
the intraoperative period and the recovery phase.17 The authors concluded that there
was a threshold for median frequency of 5 Hz above which awareness recurs. This
threshold was also observed for other drugs.15,33
There was controversy which EEG changes are characteristic responses to surgical
stimuli in anaesthetized humans. Both, shifts of the power spectrum to faster waves as
well as to slower waves the so called paradoxical arousal have been observed.34 Surgical
stimulation induced EEG patterns comparable to those seen with increases in
anaesthetic depth. The physiologic basis of the shift to slower frequencies after surgical
stimulation remains unclear.
Comparing different EEG parameters (median frequency, spectral edge frequency
(SEF90), total power, frequency band power etc.) in patients undergoing isoflurane/ni-
trous oxide anaesthesia it was demonstrated that even when threshold values for
predicting intraoperative arousal were defined post hoc there was no sole completely
reliable predictor.32 However, when looking at emergence from anaesthesia median
frequency, SEF90, total power, and frequency band power ratio underwent similar
consistent changes and these EEG descriptors may be utilized as trend monitors
regarding a change of depth of anaesthesia.

15

12
Median freq. (Hz)

0
0 30 60 90 120 150
Time (min)

Figure 4. Trend display of median frequency during anaesthesia. The biphasic response at induction and
emergence can clearly be distinguished. Modified according to (21).
156 P. H. Tonner and B. Bein

COMPARISON OF CLASSIC ENCEPHALOGRAPHIC PARAMETERS

TO MODERN TECHNIQUES

Looking at the predictive performance of several processed EEG parameters such as

approximate entropy, bispectral index and SEF95 it was found that all parameters
performed similarly on predicting consciousness vs. unconsciousness as determined by
Observers Assessment of Alertness/Sedation (OAAS), however, there was only a weak
correlation on the prediction of response to airway manipulation.35 These results
indicate that as long as relatively simple endpoints are considered even the spectral
edge frequency may perform well, however, when tested in a more complex clinical
setting the performance was not satisfying.
In a study trying to determine which parameters constitute the BIS value it was
demonstrated that SEF95 and BIS correlated well at BIS values between 30 and 80, thus
showing that classic parameters of EEG analysis may perform well if anaesthesia is deep
enough.36 However, it was pointed out that SEF response during the course of
anaesthesia is biphasic and that SEF may not be reliable at lighter states of anaesthesia.37
The authors concluded that BIS uses other parameters at light anaesthetic states for
example the beta ratio, in order to compensate for this phenomenon.36
A value to determine the accuracy of a parameter for distinguishing between two or
more states of anaesthesia is the so-called prediction probability (Pk). If the Pk is 0.5 for
an EEG parameter this parameter determines anaesthetic depth by chance. At a Pk
value of 1.0 a clear distinction can be made of the anaesthetic state with 100% accuracy
and no overlap.38 Schmidt et al determined Pk values for several parameters of
anaesthetic depth during propofol/remifentanil anaesthesia.39 When the two recent

1.00

0.75

0.50
Pk

0.25

0.00
cy

cy

nd
de
en

en

tre
in
qu

qu

co
l
tra
fre

fre

ar
ec

N
n

sp
ia

dg
ed

Bi
le
M

tra
ec
Sp

Figure 5. Comparison of the awake state with: loss of response, loss of eyelash reflex and steady state
anaesthesia. Whereas median frequency and spectral edge frequency are less reliable measures of anaesthesia
the bispectral index and the Narcotrend perform well during propofol/remifentanil anaesthesia. Modified
according to (39).
 Classic electroencephalographic parameters 157

parameters BIS and Narcotrend were compared to the classic parameters median
frequency and SEF it was found that the newer parameters were more reliable in
determination the state of anaesthesia39 (Figure 5). Similar results were obtained in a
study of patients undergoing anaesthesia with propofol/remifentanil while the airway
was maintained with a laryngeal mask40 and three different monitors were compared:
BIS, processed EEG values (SEF90), and the Alaris Auditory Evoked Potentials (AAEP).
The success rate was greater in the order of BIS, AAEP, SEF90.40 In a study looking at
approximate entropy, BIS, and SEF95 it was found that the prediction probability of
wakeup stimuli and airway manipulation was higher for approximate entropy than for
SEF95 in volunteers. It was also demonstrated that approximate entropy was essentially
equivalent to BIS35.

CONCLUSION

Various studies have been performed investigating the effects of anaesthetics on

cerebral electrical activity. It is well accepted that different anaesthetics may exert
different effects on the EEG. Although in some instances classical electroencephalo-
graphic parameters may predict the response to a stimulus correctly, they suffer from
their simplistic approach to analyze the EEG. With the abundance of literature
documenting success and mishaps these measures may nowadays only be used under
very circumscribed conditions. In an editorial on depth of anaesthesia it was concluded
that there are practically no EEG patterns that prove consciousness to be present.41 In
addition, it was stated that the EEG pattern seen during most anaesthetics do not
provide clear information about the level of consciousness.42 It has to be taken into
account that the EEG changes during emergence from anaesthesia are in part caused by
artefacts from increased muscle tone.27
Yli-Hankala et al. concluded in 1989 that up to that time no sole indicator had been
derived from the EEG that could serve as a descriptor of anaesthetic depth.14 However,
the results from recent studies comparing classical with modern parameters
demonstrate that significant steps have been taken in finding a monitor for
consciousness. Unfortunately, we are not there, yet. The ultimate monitor working
under different conditions, with all anaesthetic drugs, and with all patients is one of the
holy grails of anaesthesia.

Practice points

† classical encephalographic parameters have several limitations rendering them

interesting only for historical reasons
† modern measures of anaesthetic depth are monophasic, thus being more
reliable parameters that the classical ones

REFERENCES

1. White PF & Boyle WA. Relationship between hemodynamic and electroencephalographic changes during
general anesthesia. Anesthesia & Analgesia 1989; 68: 177–181.
2. Berger H. Über das Elektroenkephalogramm des Menschen. Arch Psychiatr 1931; 101: 452–459.
3. Blacher RS. Awareness during surgery. Anesthesiology 1984; 61: 1–2.
158 P. H. Tonner and B. Bein

4. Martin JT & Faulconer A. Electroencephalography in anesthesiology. Anesthesiology 1959; 20: 359–

362.
5. Sharbrough FW, Messick Jr. JM & Sundt Jr. TM. Correlation of continuous electroencephalograms with
cerebral blood flow measurements during carotid endarterectomy. Stroke 1973; 4: 674–683.
6. Bickford RG. Automatic encephalographic control of general anesthesia. Electroencephalogr Clin
Neurophysiol 1950; 2: 93–96.
7. Herregods L, Rolly G, Mortier E et al. EEG and SEMG monitoring during induction and maintenance of
anesthesia with propofol. Int J Clin Monit Comput 1989; 6: 67–73.
8. Ghignone M, Quintin L, Duke PC et al. Effects of clonidine on narcotic requirements and hemodynamic
response during induction of fentanyl anesthesia and endotracheal intubation. Anesthesiology 1986; 64:
36–42.
9. Buhrer M, Maitre PO, Hung O & Stanski DR. Electroencephalographic effects of benzodiazepines. I.
Choosing an electroencephalographic parameter to measure the effect of midazolam on the central
nervous system. Clin Pharmacol Ther 1990; 48: 544–554.
10. Levy WJ. Effect of epoch length on power spectrum analysis of the EEG. Anesthesiology 1987; 66: 489–495.
11. Bickford RG. Use of frequency discrimination in the automatic electroencephalographic control of
anesthesia (servo-anesthesia). Electroencephalogr Clin Neurophysiol 1951; 3: 83–86.
12. Belleville JW & Attura GM. Servo control of general anesthesia. Science 1957; 126: 827–830.
13. Belleville JW, Fennel PJ, Murphy T & Howland WS. The relative potencies of methohexital and thiopental.
J Pharmacol Exp Ther 1960; 129: 108–114.
14. Yli-Hankala A, Eskola H & Kaukinen S. EEG spectral power during halothane anaesthesia. A comparison of
spectral bands in the monitoring of anaesthesia level. Acta Anaesthesiol Scand 1989; 33: 304–308.
15. Schwilden H, Schuttler J & Stoeckel H. Quantitation of the EEG and pharmacodynamic modelling of
hypnotic drugs: etomidate as an example. Eur J Anaesthesiol 1985; 2: 121–131.
16. Arden JR, Holley FO & Stanski DR. Increased sensitivity to etomidate in the elderly: initial distribution
versus altered brain response. Anesthesiology 1986; 65: 19–27.
17. Schwilden H & Stoeckel H. Quantitative EEG analysis during anaesthesia with isoflurane in nitrous oxide
at 1.3 and 1.5 MAC. Br J Anaesth 1987; 59: 738–745.
18. Hudson RJ, Stanski DR, Saidman LJ & Meathe E. A model for studying depth of anesthesia and acute
tolerance to thiopental. Anesthesiology 1983; 59: 301–308.
19. Scott JC, Ponganis KV & Stanski DR. EEG quantitation of narcotic effect: the comparative
pharmacodynamics of fentanyl and alfentanil. Anesthesiology 1985; 62: 234–241.
20. Homer TD & Stanski DR. The effect of increasing age on thiopental disposition and anesthetic
requirement. Anesthesiology 1985; 62: 714–724.
21. Schwilden H & Stoeckel H. Closed-loop feedback controlled administration of alfentanil during alfentanil-
nitrous oxide anaesthesia. Br J Anaesth 1993; 70: 389–393.
22. Rampil IJ, Holzer JA, Quest DO et al. Prognostic value of computerized EEG analysis during carotid
endarterectomy. Anesth Analg 1983; 62: 186–192.
23. Rampil IJ & Matteo RS. Changes in EEG spectral edge frequency correlate with the hemodynamic
response to laryngoscopy and intubation. Anesthesiology 1987; 67: 139–142.
24. Schwilden H, Schuttler J & Stoeckel H. Closed-loop feedback control of methohexital anesthesia by
quantitative EEG analysis in humans. Anesthesiology 1987; 67: 341–347.
25. Hung OR, Varvel JR, Shafer SL & Stanski DR. Thiopental pharmacodynamics. II. Quantitation of clinical and
electroencephalographic depth of anesthesia. Anesthesiology 1992; 77: 237–244.
26. Ghouri AF, Monk TG & White PF. Electroencephalogram spectral edge frequency, lower
esophageal contractility, and autonomic responsiveness during general anesthesia. J Clin Monit 1993;
9: 176–185.
27. Long CW, Shah NK, Loughlin C et al. A comparison of EEG determinants of near-awakening from
isoflurane and fentanyl anesthesia. Spectral edge, median power frequency, and delta ratio. Anesth Analg
1989; 69: 169–173.
28. Eger 2nd. EI, Saidman LJ & Brandstater B. Minimal alveolar anesthetic concentration: a standard of
anesthetic potency. Anesthesiology 1965; 26: 756–763.
29. Kiersey DK, Bickford RG & Faulconer A. Electro-encephalographic patterns produced by thiopental
sodium during surgical operations: description and classification. Br J Anaesth 1951; 19: 141–152.
 Classic electroencephalographic parameters 159

30. Galla SJ, Rocco AG & Vandam LG. Evaluation of the traditional signs and stages of anesthesia: an
electroencephalographic and clinical study. Anesthesiology 1958; 19: 328–338.
31. Berezowskyj JL, McEwen JA, Anderson GB & Jenkins LC. A study of anaesthesia depth by power spectral
analysis of the electroencephalogram (EEG). Can Anaesth Soc J 1976; 23: 1–8.
32. Drummond JC, Brann CA, Perkins DE & Wolfe DE. A comparison of median frequency, spectral edge
frequency, a frequency band power ratio, total power, and dominance shift in the determination of depth
of anesthesia. Acta Anaesthesiol Scand 1991; 35: 693–699.
33. Schwilden H & Stoeckel H. Investigations on several EEG-parameters as indicators of the state of
anaesthesia the median—a quantitative measure of the depth of anaesthesia (author’s transl). Anasth
Intensivther Notfallmed 1980; 15: 279–286.
34. Kochs E, Bischoff P, Pichlmeier U & Schulte am Esch J. Surgical stimulation induces changes in brain
electrical activity during isoflurane/nitrous oxide anesthesia. A topographic electroencephalographic
analysis. Anesthesiology 1994; 80: 1026–1034.
35. Bruhn J, Bouillon TW, Radulescu L et al. Correlation of approximate entropy, bispectral index, and
spectral edge frequency 95 (SEF95) with clinical signs of ‘anesthetic depth’ during coadministration of
propofol and remifentanil. Anesthesiology 2003; 98: 621–627.
36. Morimoto Y, Hagihira S, Koizumi Y et al. The relationship between bispectral index and
electroencephalographic parameters during isoflurane anesthesia. Anesth Analg 2004; 98: 1336–
1340. (table of contents).
37. Katoh T, Suzuki A & Ikeda K. Electroencephalographic derivatives as a tool for predicting the depth of
sedation and anesthesia induced by sevoflurane. Anesthesiology 1998; 88: 642–650.
38. Smith WD, Dutton RC & Smith NT. Measuring the performance of anesthetic depth indicators.
Anesthesiology 1996; 84: 38–51.
39. Schmidt GN, Bischoff P, Standl T et al. Comparative evaluation of narcotrend, bispectral index, and
classical electroencephalographic variables during induction, maintenance, and emergence of a
propofol/remifentanil anesthesia. Anesth Analg 2004; 98: 1346–1353. (table of contents).
40. Nishiyama T, Matsukawa T & Hanaoka K. A comparison of the clinical usefulness of three different
electroencephalogram monitors: bispectral index, processed electroencephalogram, and alaris auditory
evoked potentials. Anesth Analg 2004; 98: 1341–1345. (table of contents).
41. Plourde G. Depth of anaesthesia. Can J Anaesth 1991; 38: 270–274.
42. Marshall M, Longley BP & Stanton WH. Electroencephalography in anaesthetic practice. Br J Anaesth 1965;
37: 845–857.