You are on page 1of 5

British Journal of Anaesthesia 2015, i27i31

doi: 10.1093/bja/aev212
Review Article

REVIEW ARTICLE

Changes in the electroencephalogram during


anaesthesia and their physiological basis
S. Hagihira*
Department of Anesthesiology and Intensive Care Medicine, Osaka University Graduate School of Medicine,
2-2 Yamadaoka, Suita City, Osaka 565-0871, Japan
*E-mail: hagihira@anes.med.osaka-u.ac-jp

Abstract
The use of EEG monitors to assess the level of hypnosis during anaesthesia has become widespread. Anaesthetists, however,
do not usually observe the raw EEG data: they generally pay attention only to the Bispectral Index (BIS) and other indices
calculated by EEG monitors. This abstracted information only partially characterizes EEG features. To properly appreciate the
availability and reliability of EEG-derived indices, it is necessary to understand how raw EEG changes during anaesthesia.
With hemi-frontal lead EEGs obtained under volatile anaesthesia or propofol anaesthesia, the dominant EEG frequency
decreases and the amplitude increases with increasing concentrations of anaesthetic. Looking more closely, the EEG
changes are more complicated. At surgical concentrations of anaesthesia, spindle waves (alpha range) become dominant. At
deeper levels, this activity decreases, and theta and delta waves predominate. At even deeper levels, EEG waveform changes
into a burst and suppression pattern, and nally becomes at. EEG waveforms vary in the presence of noxious stimuli
(surgical skin incision), which is not always reected in BIS, or other processed EEG indices. Spindle waves are adequately
sensitive, however, to noxious stimuli: under surgical anaesthesia they disappear when noxious stimuli are applied, and
reappear when adequate analgesia is obtained. To prevent awareness during anaesthesia, I speculate that the most effective
strategy is to administer anaesthetic agents in such a way as to maintain anaesthesia at a level where spindle waves
predominate.

Key words: electroencephalogram; reticulate nuclei of thalamus; spindle; thalamus

During the past few decades, EEG monitors, such as the BIS
Editors key points
monitor (Covidien, Boulder, CO USA), have become widely used.
Anaesthetists commonly use processed electroencephalo- These devices calculate proprietary indices that purportedly
graphic data to guide anaesthesia, but the raw EEG can be show levels of hypnosis as numerical values. Most anaesthetists
more informative. normally pay attention only to the indices and rarely take interest
The raw EEG changes in a characteristic dose-dependent in the raw EEG. While it is true that EEG indices are convenient to
manner for both volatile and i.v. anaesthetics, probably as use, they reveal only some aspects of raw EEG. Indeed, raw EEG,
a result of enhanced inhibitory effects. presenting richly complete data, initially seem more difcult to
Processed EEG indices such as BIS are not as sensitive to interpret. It is easy, however, to observe dramatic changes in
noxious stimuli as the raw EEG; use of the raw EEG is a more the raw EEG as the administered concentration of anaesthetic in-
sensitive way to monitor anaesthesia during surgery. creases.1 While raw EEG patterns are anaesthetic-specic, con-
centration-related changes in EEG waveforms are quite similar

This Article is accompanied by Editorial Aev223.


Accepted: May 15, 2015
The Author 2015. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved.
For Permissions, please email: journals.permissions@oup.com

i27
i28 | Hagihira

among different agents that potentiate gamma-aminobutyric


acid type-A (GABAA) receptors. After just one or two weeks train-
(mV)
ing, it seems likely that most anaesthetists should be able to reli-
+50
ably judge the effects of anaesthetic by observing raw EEG. In fact,
Barnard and colleagues1 have reported that most anaesthetists in 0
their study were able to differentiate EEGs from anaesthetized 4
(se
(sec)
and conscious states after only a short educational presentation. 50
Iso=0.3%
Furthermore, Bottros and colleagues2 have reported that anaes-
(mV)
thetists, after brief structured education and with access to data
+50
from a frontal EEG and relevant clinical data, can make accurate
estimates of the processed bispectral index (BIS). 0
Many factors, such as noxious stimuli, hypercapnoea, hypo- 4
capnoea, and hypothermia, also affect changes in raw EEG wave- 50 (se
s
(sec)
forms. Among these, noxious stimuli are quite important during Iso=0.6%
(mV)
surgery, so the inuence of noxious stimuli on the EEG waveform
+50
is discussed.
If anaesthetists have knowledge of changes in the raw EEG 0
during anaesthesia, it could help them judge the adequacy of 4
EEG indices and enable them to respond more rapidly and con- (s
(se
(sec)
50
dently in circumstances where equipment algorithms provide
misleading indications. EEG waveforms usually depend on elec- (mV) Iso=0.9%
trode position. Here, I describe EEG changes and their neuro- +50
physiological background obtained using from hemi-frontal
0
leads such as FP1A1 or FPzAt1, which are commonly used in 4
EEG-based anaesthesia monitors. 50 (se
(sec)

Changes in the raw electroencephalogram (mV) Iso=1.2%


during anaesthesia +50

Isourane, sevourane, thiopental, and propofol produce anaes- 0


thesia in part by potentiating inhibitory GABAA receptors.3 4
50 (se
se
e
(sec)
Although each of these agents has its own characteristic EEG
waveform, these waveforms undergo similar changes as the con- Iso=1.5%
centration of administered anaesthetic increases. Figure 1 shows
raw EEG waveforms during isourane anaesthesia. During light
Fig 1 EEG waveform (4 s duration) during 0.3, 0.6, 0.9, 1.2, and 1.5% isourane
anaesthesia, amplitude is shallow and frequency is high. When
anaesthesia in a 55-year-old ovarian tumor patient undergoing resection.
a higher concentration is administered, amplitude deepens and
EEG frequency slows. During deep anaesthesia, a burst and sup-
pression pattern becomes apparent, characterized by extreme ac-
tivity, represented by high-frequency, large-amplitude waves
Changes of electroencephalogram-derived
(bursts), alternating with at traces (suppression). This pattern,
excluding brain ischaemia or other factors, indicates that anaes-
parameters during anaesthesia
thesia is too deep. Beyond this, at traces become dominant and, To clarify anaesthesia-related changes in EEG, it is necessary to
eventually waveforms are no longer apparent. During isourane, consider how EEG data is analysed and presented by monitoring
sevourane or propofol anaesthesia, this sequence of changes in devices. Several analytic methods have been applied to construct
pattern is almost identical. The major difference in EEG between EEG indices or EEG parameters. For example, time domain ana-
the volatile agents (isourane or sevourane) and propofol is ap- lysis is used for detection of suppression or calculation of ampli-
parent in power in the theta range. During propofol anaesthesia, tude and to provide burst suppression ratio (BSR) information.
theta power remains low regardless of concentration, but during Power spectral analysis is the most commonly used tool, and is
isourane or sevourane anaesthesia, it increases at surgical applied to calculate several parameters including spectral edge
concentrations of anaesthesia. frequency (SEF), median frequency (MF), and relative ratio
It is well known that alpha oscillation (around 10 Hz) are often (RBR). Spectral entropies are also calculated from power spec-
observed when a person is awake with eyes closed. Such alpha trum data. Bispectral analysis, the core technology of the BIS
oscillation is mostly observed in occipital regions.4 On the other monitor, quanties phase relations between the frequency com-
hand, alpha rhythms observed during anaesthesia or natural ponents of EEG signals, and supplies SynchFastSlow (SFS) infor-
sleep are mostly observed in the frontal region. This alpha mation.7 Bispectral analysis will be discussed later in greater
rhythm shift is referred to as anteriorization of the alpha rhythm, detail. It should be noted that BSR, RBR, and SFS results are
and is common during isourane, sevourane or propofol anaes- used as sub-parameters for BIS calculation.7
thesia.5 Usually, anteriorization of the alpha rhythm predomi- Generally speaking, as isourane is administered in greater
nates when anaesthesia is adequate for surgery, and is not concentrations, EEG amplitude increases and SEF95 decreases.
clearly apparent just after loss of responsiveness. For example, In other words, the EEG waveform changes from fast wave,
Blain-Moraes and colleagues6 have reported that sevourane small amplitude to slow wave, large amplitude. But this is just
did not result in consistent anteriorization of the alpha rhythm a broad view: Figure 2 shows power spectrum changes during iso-
at around 0.8%, which was sufcient for loss of consciousness. urane anaesthesia. At isourane 0.3%, frequency is in the beta
Changes in EEG during anaesthesia | i29

25 0.0

0.5
0.3%

Relative b ratio
15
0.5% 1.0
0.7%
(m V2)

10 0.9% 1.5
1.1%
2.0
5
2.5
Awake 0.3 0.5 0.7 0.9 1.1 1.3 1.5
Isoflurane (%)
0 10 20 30
(Hz)
Fig 3 Relative ratio (RBR) changes during isourane anaesthesia. RBR
values were calculated from stored data used in a previous report.
Fig 2 Power spectrum changes during isourane anaesthesia in a 43-
(Source: ref.8).
year-old female patient undergoing resection for colon cancer.

bicoherence shows high values in rather restricted regions in


range and amplitude is comparatively low. When isourane is the bi-frequency space.9 When isourane concentration in-
increased, peak frequency moves into the alpha range and amp- creases, two gradually growing peaks of EEG bicoherence emerge
litude increases. When concentration is further increased, amp- around the diagonal line representing f1=f2, one at around 4 Hz
litude decreases. In contrast, the power spectrum below 4 Hz and another at around 10 Hz. Designating the 4 Hz peak height
increases as isourane concentration increased. Thus, power as pBIClow and the 10 Hz peak as pBIChigh. At 0.3% isourane,
spectrum changes are more complicated than they supercially both pBICs are low. As isourane concentration increases, pBIC
appear. These more subtle changes might be conveyed by low increases until it reaches a plateau at 0.9% isourane. Mean-
changes in the RBR. According to Rampil,7 RBR is dened as the while, pBIChigh also clearly increases to reach a somewhat
logarithm of the power ratios from the 3047 Hz and 1120 Hz lower plateau at isourane 0.9%, whereupon it slightly decreases
ranges. RBR is mainly used to calculate BIS values from wakeful- at isourane concentrations of more than 1.1%. Furthermore, the
ness to light anaesthesia, that is, BIS 60 to 100. Figure 3 shows the frequency of pBIChigh gradually slows as isourane concentra-
changes of RBR during isourane anaesthesia. Unlike SEF95 va- tion increases.8
lues, RBR shows bi-phasic changes, which would be caused by
the transient increase of power spectrum in alpha range. Minimal
RBR values correspond with maximal alpha activity.
Physiological basis of the electroencephalogram
As alpha activity is most apparent at surgical concentrations
of anaesthesia, I speculated that anaesthetic-concentration-
during anaesthesia
related changes in alpha activity might hold the key to assessing Alpha activity observed during the stage II of slow wave sleep
anaesthetic effects during surgery. As suggested above, it is (SWS) is called spindle activity, a pattern characterized by wax-
possible to estimate alpha activity from the RBR. ing and waning of waveforms in the 714 Hz range. Alpha activity
Bispectral analysis is an advanced signal-processing tech- observed during anaesthesia appears to be generated by
nique. The concept is briey outlined as follows. Assuming that the same mechanism as sleep spindles. Spindles are usually ob-
a nucleus receives two waves with respective frequencies f1 and served, but only occasionally, in SWS. During surgical anaesthe-
f2 and phase angles 1 and 2, a wave with frequency of f1+f2 sia, however, they appear almost continuously. The usefulness
would be generated. of observing spindle activity can be understood if we consider
If the phase angles of original waves (1, 2) are inherited to the why the two pBICs changed as described in the previous section.
phase angle of the output signal as 1+2, those signal compo- At 714 Hz, the frequency of pBIChigh corresponds to the fre-
nents are regarded as phase coupled. Such a phenomenon is quency of spindle patterns; the frequency of pBIClow corre-
often observed in nonlinear systems. Bispectral analysis statistic- sponds to delta waves. Using EEG combined with simultaneous
ally quanties the degree of phase coupling among the frequency intracellular recordings, Steriade and colleagues.10 11 extensively
components of a signal. Using a proper method for bispectral investigated the drivers of these two EEG waves. They concluded
analysis of EEG data during anaesthesia, the direct indicator of that pacemaking for spindle patterns originated in reticulate nu-
the degree of phase coupling is bicoherence, the normalized bis- clei (RE) of thalamus and thalamo-cortico-thalamic circuits, that
pectrum parameter.9 While the bispectrum value is inuenced the rhythm of delta waves was the intrinsic rhythm of thalamo-
by the magnitude of the original components and by the degree cortical neurons, and that these EEG rhythms were determined
of phase-coupling, bicoherence values indicate only the phase by the membrane potential of thalamocortical relay (TC) neu-
relation. To investigate the relationship between EEG bicoher- rones. Nuez and colleagues12 reported that the EEG rhythm
ence and anaesthetic concentration, Bispectrum Analyzer (BSA) transits to a spindle pattern when the TC membrane potential
software was developed for real time bispectral analysis of EEG was in the range of 55 mV to 65 mV, subsequently changing
data.9 As bispectrum or bicoherence is determined by two fre- to delta waveforms when the membrane potential decreases
quencies, plots of the values become three dimensional. The below 65 mV. Figure 4 schematically presents these neuronal
current version of BSA can gather raw EEG data and EEG-derived circuits. RE of thalamus have GABAergic inhibitory input to TC
parameters such as BIS values from a BIS monitor. EEG neurones. Thus anaesthetic agents might concentration-
i30 | Hagihira

dependently hyperpolarize the membrane potential of TC neu- the EEG pattern changed to show a predominance of large delta
rones, decreasing the frequency of pBIC-high. The changing waves, so called paradoxical arousal14 marked by dips in BIS or
power spectrum pattern in Figure 2 also seems to reect a transi- SEF or other EEG index values. Kochs and colleagues15 have re-
ent increase in spindle activity and concentration-related ported that noxious stimuli promptly decreased alpha activity
changes in delta activity. in 53% and increased delta activity in 44% of instances during
1.2% isourane with 66% nitrous oxide. Such EEG power spec-
trum changes seem to caused BIS or SEF values to decrease.
Inuence of noxious stimuli on the EEG The situation is further complicated in that the EEG involved a
mixed pattern of these two patterns in some patients. Currently,
So far, only the effects on the EEG of varying anaesthetic concen-
there is little information about what situations give rise to
trations have been discussed. EEG is also inuenced, however, by
paradoxical arousal. Using an anaesthetized cat model, Kaada
noxious stimuli.
and colleagues16 have reported on the effect on EEG of high-
Changes in EEG bicoherence and BIS or other EEG-derived
frequency electrical stimulation to the midbrain reticular forma-
parameters have been reported in 12 subjects at incision and at
tion. Their results suggest that EEG changes depend on both on
5 min after administration of 3 g kg1 of fentanyl during admin-
the concentration of anaesthetic and the intensity of stimulus,
istration of 1.0% isourane.13 After initial skin incision (noxious
even when stimulation was intensied at the same neurone.
stimulus), the EEG frequency increased and the amplitude de-
When noxious stimulation is applied, BIS shows variable
creased. In other words, the EEG pattern was desynchronized.
changes: in some patients, the BIS value increases after incision;
This pattern was different, however, to those obtained with iso-
in others, it decreases; and in others, BIS remains unchanged.
urane at lower concentrations. Furthermore, in some patients
Interestingly, when BIS changed, administration of fentanyl
restored BIS values to those before incision in a sample of 12 sub-
jects.13 SEF95 values show a tendency to change in the same way
as BIS values. These results indicate that it is imprudent to assess
analgesia using BIS or SEF95 values. Going further, it is rst ne-
Pyramidal neuron GABA cessary to ensure adequate analgesia before assessing the level
Glu of hypnosis using BIS or SEF95 values. Changes in EEG after nox-
Ach Ach ious stimuli during sevourane anaesthesia have also been in-
Excitatory synapse vestigated and the results were quite similar to those obtained
Inhibitory synapse
during isourane anaesthesia.13 Furthermore, similarities are
Glu also apparent during propofol anaesthesia (data not published).
Ach
On the other hand, pBIC-high decreased after incision and, in
Basal ganglia Glu all patients, rose to previous values after administration of
GABA TC neuron
fentanyl (Fig. 5). Meanwhile, spindle activity disappeared after
RE neuron incision and, in all patients, re-appeared after fentanyl adminis-
Ach Ach tration. This correspondence indicates that pBIC-high is likely to
be a good indicator of adequate analgesia. As pBIC-high seems
PPT/LDT to indicate spindle activity, spindle patterns observed in raw
EEG or in the power spectrum might also provide a basis for as-
sessing adequacy of analgesia. This hypothesis warrants further
Fig 4 Scheme of neuronal circuits related to spindle generation GABA,
gamma-amino butyric acid; Ach, acetylcholine; Glu, glutamate; PPT/LDT,
study.
pedunculopontine tegmental/laterodorsal tegmental nuclei. (Source: During more than 15 years of clinical practice, I have observed
compiled from refs.10 11). the raw EEG during anaesthesia in more than 10 000 surgeries,
and I have managed many patients using pBIC-high as an indica-
tor of adequacy of analgesia. A limitation of my strategy is inter-
individual variation in the raw EEG during anaesthesia. Some
patients do not manifest high spindle activity at any level of
anaesthesia. Generally speaking, this absence is more frequent
50 in elderly patients but, even so, some patients aged more than
1.0% isoflurane 80 or 85 still show predominant spindle activity. At the same
40 time, some younger patients, even in the 30 to 40 year range, do
not show predominant spindle activity. The variation cannot be
30 solely attributed to aging. What we can conclude is that absence
of spindle activity does not necessarily indicate insufcient
(%)

analgesia. On the other hand, after assessing spindle activity


20 after induction of anaesthesia and immediately before the start
of surgery, a shift to predominant spindle activity is likely to be
10 a good indicator that analgesia is adequate during surgery.

0 Conclusion
Control After incision After fentanyl
Although the raw EEG varies widely from individual to individual,
the way the EEG patterns change when differing concentrations
Fig 5 Effect of surgical incision (noxious stimuli) on pBIC-high in 12 patients
during isourane anaesthesia (Source: Plotter data from ref.13). of anaesthetic are administered is quite similar among patients.
While it is easier to refer to EEG indices, they might not adequately
Changes in EEG during anaesthesia | i31

convey the information necessary to control anaesthesia. In par- 7. Rampil IJ. A primer for EEG signal processing in anesthesia.
ticular, adequate analgesia is required before the indices can be Anesthesiology 1998; 89: 9801002
used as indicators of level of hypnosis. As it is relatively easy to 8. Hagihira S, Takashina M, Mori T, Mashimo T, Yoshiya I.
interpret raw EEG changes during anaesthesia, all anaesthetists Changes of electroencephalographic bicoherence during iso-
should acquire the ability to interpret raw EEG. urane anesthesia combined with epidural anesthesia.
Anesthesiology 2002; 97: 140915
9. Hagihira S, Takashina M, Mori T, Mashimo T, Yoshiya I. Prac-
Authors contribution
tical issues in bispectral analysis of electroencephalographic
S.H. is the sole author of this paper, accountable for accuracy and signals. Anesth Analg 2001; 93: 96670
integrity of the paper. 10. Steriade M, Nuez A, Amzica F. Intracellular analysis of rela-
tions between the slow (<1 Hz) neocortical oscillation and
Declaration of interest other sleep rhythms of the electroencephalogram. J Neurosci
1993; 13: 326683
None declared.
11. Steriade M, Contreras D, Dossi RC, Nuez A. The slow (<1 Hz)
oscillation in reticular thalamic and thalamocortical neu-
Funding rons: Scenario of sleep rhythm generation in interacting
This study was funded by the Department of Anesthesiology and thalamic and neocortical networks. J Neurosci 1993; 13:
Intensive Care Medicine, Osaka University Graduate School of 328499
Medicine, Suita, Osaka, Japan. 12. Nuez A, Dossi CC, Contreras D, Steriade M. Intracellular evi-
dence for incompatibility between spindle and delta oscilla-
tions in thalamocortical neurons of cat. Neuroscience 1992; 48:
References 7585
1. Barnard JP, Bennett C, Voss LJ, Sleigh JW. Can anaesthetists be 13. Hagihira S, Takashina M, Mori T, Ueyama H, Mashimo T.
taught to interpret the effects of general anaesthesia on the Electroencephalographic bicoherence is sensitive to noxious
electroencephalogram? Comparison of performance with stimuli during isourane or sevourane anesthesia.
the BIS and spectral entropy. Br J Anaesth 2007; 99: 5327 Anesthesiology 2004; 100: 81825
2. Bottros MM, Palanca BJA, Mashour GA, et al. Estimation of the 14. Bischoff P, Kochs E, Droese D, Meyer-Moldenhauer WH,
Bispectral Index by anesthesiologists. An inverse turing test. Schulte am Esch J. Topographic-quantitative EEG-analysis
Anesthesiology 2011; 114: 1093101 of the paradoxical arousal reaction. EEG changes during uro-
3. Hemmings HC Jr, Akabas MH, Goldstein PA, Trudell JR, logic surgery using isourane/N2O anesthesia. Anaesthesist
Orser BA, Harrison NL. Emerging molecular mechanisms of 1993; 42: 1428
general anesthetic action. Trends Pharmacol Sci 2005; 26: 15. Kochs E, Bischoff P, Pichlmeier U, Schulte am Esch J. Surgical
50310 stimulation induces changes in brain electrical activity
4. Hall JE, Guyton AC. Textbook of Medical Physiology, 12th Edn. during isourane/nitrous oxide anesthesia. a topographic
Philadelphia: Saunders Elsevier, 2011 electroencephalographic analysis. Anesthesiology 1994; 80:
5. Akeju O, Westover B, Pavone KJ, et al. Effects of sevourane 102634
and propofol on frontal electroencephalogram power and co- 16. Kaada BR, Thomas F, Alnaes E, Wester K. EEG synchroniza-
herence. Anesthesiology 2014; 121: 9908 tion induced by high frequency midbrain reticular stimula-
6. Blain-Moares S, Tarnal V, Vanini G, et al. Neurophysiological tion in anesthetized cats. Electroenceph Clin Neurophysiol
correlates of sevourane-induced unconsciousness. Anes- 1967; 22: 22030
thesiology 2015; 122: 30716

Handling editor: H. C. Hemmings

You might also like