British Journal of Anaesthesia, 119 (4): 664–73 (2017)

doi: 10.1093/bja/aex170
Advance Access Publication Date: 9 August 2017
Neurosciences and Neuroanaesthesia

NEUROSCIENCES AND NEUROANAESTHESIA

Frontal alpha-delta EEG does not preclude volitional

response during anaesthesia: prospective cohort
study of the isolated forearm technique
A. L. Gaskell1,2,*, D. F. Hight2, J. Winders2, G. Tran3, A. Defresne3,4,
V. Bonhomme3,4,5, A. Raz6,7, J. W. Sleigh1,2 and R. D. Sanders6
1
Department of Anaesthesia, Waikato Hospital, Hamilton, New Zealand, 2Department of Anaesthesiology,
Waikato Clinical School, Waikato Clinical Campus, University of Auckland, New Zealand, 3Coma Science
Group, GIGA research, University and CHU University Hospital of Liege, Belgium, 4Department of Anesthesia
and Intensive Care Medicine, CHU University Hospital of Liege, Belgium, 5University Department of
Anesthesia and Intensive Care Medicine, CHR Citadelle, Liege, Belgium, 6Department of Anesthesiology,
University of Wisconsin School of Medicine and Public Health University of Wisconsin, Madison, WI, USA and
7
Department of Anesthesiology, Rambam Health Care Campus, Haifa, Israel

*Corresponding author. E-mail: [email protected]

Abstract
Background: The isolated forearm test (IFT) is the gold standard test of connected consciousness (awareness of the environment)
during anaesthesia. The frontal alpha-delta EEG pattern (seen in slow wave sleep) is widely held to indicate anaesthetic-induced
unconsciousness. A priori we proposed that one responder with the frontal alpha-delta EEG pattern would falsify this concept.
Methods: Frontal EEG was recorded in a subset of patients from three centres participating in an international multicentre study
of IFT responsiveness following tracheal intubation. Raw EEG waveforms were analysed for power–frequency spectra, depth-of-
anaesthesia indices, permutation entropy, slow wave activity saturation and alpha-delta amplitude-phase coupling.
Results: Volitional responses to verbal command occurred in six out of 90 patients. Three responses occurred immediately
following intubation in patients (from Sites 1 and 2) exhibiting an alpha-delta dominant (delta power >20 dB, alpha power
>10 dB) EEG pattern. The power–frequency spectra obtained during these responses were similar to those of non-responders
(P>0.05) at those sites. A further three responses occurred in (Site 3) patients not exhibiting the classic alpha-delta EEG pat-
tern; these responses occurred later relative to intubation, and in patients had been co-administered ketamine and less vol-
atile anaesthetic compared with Site 1 and 2 patients. None of the derived depth-of-anaesthesia indices could robustly dis-
crimate IFT responders and non-responders.
Conclusions: Connected consciousness can occur in the presence of the frontal alpha-delta EEG pattern during anaesthesia.
Frontal EEG parameters do not readily discriminate volitional responsiveness (a marker of connected consciousness) and
unresponsiveness during anaesthesia.
Clinical trial registration: NCT02248623

Key words: anaesthesia general; anaesthesia awareness; EEG; consciousness

Editorial decision: April 23, 2017; Accepted: May 9, 2017

C The Author 2017. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved.
V
For Permissions, please email: [email protected]

664
 Frontal EEG findings during isolated forearm technique responses | 665

have examined the raw EEG recording, which to the best of our
Editor’s key points knowledge has not been done previously. Our aim was to ascer-
tain if the archetypal alpha-delta EEG pattern was present in
• The frontal alpha-delta EEG pattern is an index of
any of the patients who responded appropriately to command
anaesthetic-induced unconsciousness, and the isolated
during the IFT; any such patients would disprove the proposi-
forearm test is a validated method of detecting con-
tion that the frontal alpha-delta pattern reliably indicates
sciousness under general anaesthesia. anaesthetic unconsciousness.
• In a prospective multicentre study, positive responses
to the isolated forearm test following induction of gen-
eral anaesthesia and tracheal intubation were observed Methods
in several patients despite frontal alpha-delta EEG
Study design
patterns.
• The alpha-delta-dominant EEG pattern does not preclude This is a substudy of EEG recordings taking at three of the sites
connected consciousness under general anaesthesia. (Waikato Hospital, New Zealand, University of Wisconsin,
Madison, USA, and CHR Citadelle Hospital of Liège, Belgium)
involved in a larger prospective observational study.16
Institutional Review Board or Local Ethics Committee approval
was obtained at each site and the study was placed on a clinical
Connected consciousness can be defined as ‘subjective experi- trials registry (NCT02248623).
ence of the external environment.’1 Reports of awareness with Participants were adults (16 yr and over) undergoing laryngo-
recall show that connected consciousness does occur occasion- scopy and tracheal intubation as part of their routine anaesthetic
ally during intended general anaesthesia. However, as anaes- care. Informed consent was obtained from all patients. Exclusion
thetic agents are amnesic at subhypnotic doses,2 lack of explicit criteria were inability to complete commands preoperatively,
recall cannot be regarded as evidence of absence of experience. inaccessibility of the dominant forearm or contraindication to
The isolated forearm test (IFT)3 is a practical means of overcom- tourniquet (e.g. lymphoedema risk or operative site) and cardiac
ing the amnesia problem, and hence of assessing consciousness or intracranial neurosurgical procedures. Recruitment took place
during anaesthesia in real-time when muscle relaxants are used. between December 2014 and August 2015.
The common-sense interpretation of a timely volitional response
to a command is that the patient had some form of connected
consciousness. We can assume that subjective experience of the
Study conduct
command occurs with the execution of an appropriate directed Conduct of anaesthesia was at the discretion of the responsible
response. However, it should be noted that the IFT responsive anaesthetist, who was independent from the study, with no
state is not necessarily equivalent to the normal alert state. stipulations upon which agents were administered. Standard
General anaesthesia with >0.5 minimum alveolar concentra- monitoring (non-invasive blood pressure, ECG, pulse oximetry,
tion (MAC) of a volatile anaesthetic or propofol >2 mg ml1 is end-tidal CO2, O2 and volatile agent concentrations) was per-
almost always associated with the appearance of a high voltage formed on all patients. Age, gender, BMI, comorbidities and reg-
slow wave (alpha-delta) pattern in the frontal EEG.4 5 This pat- ular medications were recorded. Details of all medications
tern is a sign of thalamo-cortical hyperpolarization and is very administered and laryngoscopy difficulty and duration were
similar to that seen in slow wave sleep. Consequently, it has recorded, along with heart rate and blood pressure readings
been suggested by many authors4–10 to be indicative of anaes- during the study period. A modified Brice questionnaire was
thetic unconsciousness. If this were so, it should not be possible administered within 24 h of the anaesthetic episode to screen
for a patient to make an appropriate volitional response in the for awareness with explicit recall.
presence of this EEG pattern.
Karl Popper captured the weak link in scientific inductive
reasoning with the famous quote: ‘. . .no matter how many instan-
Isolated forearm testing
ces of white swans we may have observed, this does not justify the Following induction of anaesthesia, a tourniquet was inflated to
conclusion that all swans are white.’ We have followed this logic— 50 mm Hg above systolic blood pressure on the forearm prior to
that a single episode of an event (e.g. a black swan) may invalid- administration of muscle relaxant. Prior to and following lar-
ate an ‘always’ notion (e.g. all swans are white).11 Proposed yngoscopy and intubation, the patient was given a command:
markers of consciousness under anaesthesia typically take this ‘(name), if you can hear me, squeeze my hand.’ A volitional
black and white view and thus may be challenged along response (IFT positive) was defined as an appropriate squeeze
Popperian lines. of the investigator’s hand immediately after the verbal com-
Commercially available processed EEG monitors have com- mand; this was witnessed where possible by a second observer.
monly failed to predict or detect volitional behavioural Other movements that occurred during and following laryngo-
responses revealed by isolated forearm testing during general scopy and intubation, independent of command, were not con-
anaesthesia.12–15 However, as only processed dimensionless sidered to be cognitive volitional responses.16 Neuromuscular
EEG values derived from proprietary algorithms were reported transmission in the forearm was verified by train-of-four count
in these studies, it has been unclear why the monitors failed to by stimulation of the ulnar nerve. At one site (Site 3), commands
detect such events. We recently conducted an international were given repeatedly and volitional responses that occurred
multicentre prospective cohort study to determine the inci- >1 min following intubation were marked as late responses. Site
dence of IFT responsiveness following laryngoscopy and tra- 3 also differed from the other two sites in that it was routine
cheal intubation during routine anaesthetic clinical practice.16 local practice to co-administer ketamine during induction of
We present our findings from a subset of these patients in anaesthesia and volatile anaesthesia was not typically com-
whom simultaneous EEG recordings were taken. Specifically, we menced until after intubation.
 666 | Gaskell et al.

EEG collection and analysis 2. Slow Wave Activity Saturation (SWAS)20—the degree to
which each individual has approached their maximum pos-
Frontal EEG recordings were obtained using the Bispectral Index
sible 1 Hz power; which has been linked to functional brain
monitor (BIS Vista, Covidien, Medtronic, Minneapolis, MN, USA)
imaging evidence of loss of thalamic and cortical activation
at Sites 1 and 2 and the NeuroSENSE monitor (NeuroWave
to auditory and noxious stimuli during propofol administra-
Systems Inc., Cleveland Heights, OH, USA) at Site 3. Standard
tion. This was calculated as the slow wave amplitude
commercial electrode strips, montages and impedance checks
(0.1–1.5 Hz) at the time of interest divided by the maximum
were used. The EEG recordings were commenced at the start of
slow wave amplitude observed in that patient during the
anaesthesia, and times of events were noted using a stopwatch,
whole induction period. This does assume that the true
taps or monitor event markers.
maximum possible slow wave amplitude was in fact reached
Raw EEG waveforms for each patient were uploaded from the
at some time during the induction process, due to relative
monitors and analysed using MATLAB software (The Mathworks,
overdosing of the induction bolus. We calculated alpha satu-
Natick, MA, USA). High-pass 0.2 Hz and low-pass 45 Hz third-order
ration in a similar manner using peak alpha amplitudes.
Butterworth filters were applied sequentially with the phase-
3. The alpha-slow wave amplitude-phase coupling—commonly
preserving MATLAB ‘filtfilt.m’ function. EEG signals acquired from
called ‘Peak/TroughMAX’(PmaxI)—a measure of the tendency
the NeuroSENSE monitors were then downsampled to match the
of the maximum alpha amplitude to occur at the peaks of
acquired BIS sampling frequency of 128 s–1 (note both monitors
slow wave oscillations.21 22 We used a slightly modified ver-
record at considerably higher sampling frequencies than are avail-
sion of the algorithm. Initially, 60 s segments of the raw EEG
able as the waveform output). Artifact-free epochs of 20 s duration
signal were bandpass filtered between 0.1 and 45 Hz using
were selected for each patient, which were typically 20–40 s prior
fifth-order Butterworth filters and the phase-preserving ‘filt-
to the onset of laryngoscopy and 20–40 s after tracheal intubation,
filt.m’ MATLAB function. Sections of EEG signal >3 SD were
except in the cases of the late responses where a 20 s artifact-free
rejected as artifacts. We then extracted the two frequency
epoch centred on the recorded response was used.
bands of interest by applying fifth-order phase-preserving
Burst suppression was identified as alternating episodes of
Butterworth bandpass filters for 0.1–1 Hz [slow wave oscilla-
electrical quiescence and waveforms by review of raw EEG and
tion (SWO), and 8–14 Hz (alpha)]. Using an artificial white
spectrograms by consensus of three researchers experienced in
noise test signal, this showed about 20 dB Hz–1 drop off either
EEG interpretation. Where burst suppression was present, no
side of the frequency band of interest. The instantaneous
further spectral analysis was employed.
phase and amplitude for each of these signals was then
The typical EEG pattern seen during general anaesthesia
obtained using the Hilbert transformation. A purpose written
with volatile anaesthetics is marked by an increase in amplitude
MATLAB function then obtained the envelope of the ampli-
and a shift to low frequencies, the so-called alpha-delta pattern
tude in the alpha frequency band, and produced the 18 bin
seen in the power–frequency spectrum.5 Therefore, power–fre-
modulogram linking the alpha envelope amplitude to the
quency spectra were obtained by a modified Fast Fourier
phase of the SWO. A peak-max pattern of coupling should
Transform using the Chronux17 toolbox ‘mtspectrumc.m’ multi-
show maximum alpha amplitude when the slow wave phase
taper function in MATLAB with a frequency resolution of 0.4 Hz
is around zero radians; whereas the trough-max pattern
(with 15 tapers) followed by logarithmic transformation of
shows maximum alpha amplitude when the slow wave
power to decibels (dB, referenced to 1 mV). Spectrograms were
phase is close to p and p radians. Thus, the ratio of mean
created using the Chronux ’mtspecgramc.m’ function with a
alpha amplitude in bins 7–12, divided by the mean alpha
moving window of 4 s duration with 1 s overlap.
amplitude in bins 1–3 and 16–18 will give an index (PmaxI). A
Delta power was defined as the peak power observed in
PmaxI value greater than one indicates tendency to peak-max,
the delta frequency band (0.5–4 Hz). Alpha power was defined as
and less than one indicates trough-max. The original papers
the peak power in the extended alpha frequency band (7–17 Hz);
suggest that at least 120 s of EEG is required to accurately esti-
we designate the frequency of this peak as the alpha peak
mate the amplitude-phase coupling index; however, it is not
frequency. Oscillatory alpha power is the power at the alpha
possible to achieve a stationary signal of this length in the
peak frequency over and above the background broadband
rapidly changing situation around the clinical induction of
noise.18 Theta, beta and low gamma powers reported were
anaesthesia and intubation.
the average power in these frequency bands (4–7 Hz, 17–32 Hz
and 32–45 Hz, respectively). Mean BIS or NeuroSense and
EMG values corresponding to the selected epochs are also Statistical analysis
reported.
The study was not statistically powered to show a difference in
the mean values of EEG parameters between IFT responders
Putative EEG indices of consciousness and non-responders. A priori we did not know the incidence of
IFT responsiveness, and, given the low rate of IFT responsive-
In addition to the usual spectral analysis of the EEG, there are a
ness encountered, a study size many orders of magnitude
number of recently published, more nuanced, single-channel EEG
greater would be required to do so with reasonable certainty. In
analysis methods that have been proposed to be reliable bio-
any case, showing a difference in means is meaningless and
markers of anaesthetic-induced unconsciousness. We therefore
unrelated to the primary scientific question, namely ‘Is voli-
applied these algorithms to our data, modified where needed for
tional responsiveness (our surrogate of connected conscious-
shorter EEG, epochs appropriate to the fluctuating clinical situa-
ness) possible when the frontal EEG exhibits the classic alpha-
tion around induction of anaesthesia, to see if they could reliably
delta signature during general anaesthesia?’ This requires
distinguish volitional responders from non-responders.
examination of outliers, not the measure of central tendency.
We examined:
The underlying philosophy of the analysis was therefore that of
1. Permutation Entropy (PE)19—a robust measure of the com- direct falsification. If we could find any instances of patients
plexity inherent in the EEG waveform. showing an unequivocal volitional IFT response when their EEG
 Frontal EEG findings during isolated forearm technique responses | 667

was suggestive of unconsciousness, then we could claim that occur to further commands at approximately 120 (Responder D),
that particular EEG pattern (or EEG index) was not logically nec- 80 (Responder E) and 160 s (Responder F) after intubation.
essary for unconsciousness.
One of the critiques of this approach is that there might be
EEG spectra of responders
auxiliary questions related to quantitative indices of thalamo-
cortical hyperpolarization. Perhaps ‘Is volitional unresponsive- No responses occurred with burst suppression evident on the
ness only seen when we have achieved a maximal amount of EEG
delta power (SWAS), or the correct alpha-delta phase relation-
ship (PmaxI)?’ Our secondary aim was therefore to determine Sites 1 and 2
whether any existing quantitative frontal alpha-delta EEG
All three IFT responses (Responders A, B and C) occurred in the
measurements showed potential clinical utility in the detection
presence of a clear frontal alpha-delta dominant EEG pattern—
of IFT responsiveness. We were therefore not interested in usual
both prior to laryngoscopy and at the time of the response after
statistical recipes of comparing mean values between different
intubation. We term these ‘alpha-delta responders’. Figure 2
groups, but instead examined how the indices for individual
shows the spectrogram of one such responder; the alpha-delta
responders compared with the distribution [inter-quartile range
pattern was present throughout the period of interest. The spec-
(IQR)] of non-responders: any values lying within the IQR of
tra of responders are displayed in colour against non-
non-responders suggesting poor potential discriminatory
responders from epochs prior to laryngoscopy (Fig. 1A) and
ability.
around the IFT command (Fig. 1C). No consistent distinguishing
Although we do not have large numbers of IFT positive data
features were apparent from these spectra, and alpha and delta
points, it is possible to derive a likelihood function (‘normli-
powers of these responders lie comfortably above the lower
ke.m’) to estimate the negative log-likelihood that our res-
quartile of non-responders (Table 1).
ponder data points came from a range of possible probability
distributions, and thus get an idea of whether the putative index
has achieved much separation between responders and non- Site 3
responders. The discriminatory capability of an EEG index The spectra of these responders (D, E and F), differed signifi-
requires there to be minimal overlap between IFT responsive- cantly from the alpha-delta responders, showing reduced power
ness and non-responsiveness. in the lower frequency ranges (Fig. 1 and Table 2). The pre-
laryngoscopy spectra for these responders were comparable
with the non-responders from the same site, with similar clini-
Results cal regimes, but the spectra during responses were at the
Complete EEG data were analysed from 86 of 90 participants extreme of the Site 3 spectra, particularly showing reduced
from three sites. Four EEGs were excluded from analysis power in the delta frequency range (Fig. 1D). Note that these
because the timings of IFT commands were uncertain (none of epochs were later relative to intubation (and therefore also later
these participants responded to command). Train-of-four test- relative to propofol induction of anaesthesia) than the non-
ing verified intact neuromuscular transmission in the isolated responder epochs. Figure 3 is a typical spectrogram of a res-
forearm in all cases. On postoperative questioning there were ponder from this site showing a gradual reduction in lower fre-
no episodes of explicit recall of surgery or the IFT commands. quency activity and increasing higher frequency activity leading
up to the recorded responses.

EEG findings prior to laryngoscopy and intubation Further putative EEG indices
There was significant variance in anaesthetic practice between As regards to frontal SWAS, no responders were fully saturated
Sites 1 and 2 (n¼63) and Site 3 (n¼23) where ketamine was rou- (defined as SWAS values >95%), although one response
tinely used for co-induction and volatile anaesthetic was rarely occurred at 92% saturation. SWAS values for the alpha-delta res-
administered prior to intubation. In addition, Site 3 used the ponders were all above the lower quartile; thus—with the pro-
NeuroSENSE monitor to acquire the EEG signal whereas the viso that peak SWA may not have been achieved in all cases
other sites used BIS. Thus, EEG patterns differed markedly by around induction—SWAS did not show strong discriminatory
site (Fig. 1), even prior to laryngoscopy and intubation [e.g. delta capability for these responders.
power 26.6 (3.5) dB at Sites 1 and 2 vs 15.1 (2.9) dB at Site 3, Overall the PmaxI was significantly >1 (P<0.0001) in non-
P<0.0001]. Burst suppression also occurred more frequently at responders, indicating that most patients were showing a ten-
Sites 1 and 2, compared with Site 3. For these reasons the spec- dency to a peak-max pattern. The six responders had a neu-
tra and indices have been presented in two respective groups in tral or trough-max pattern at the time of IFT; five of six of
Figure 1 and Tables 1 and 2. these values were outside the IQRs of the respective non-res-
ponders; however, they were not at the extremes of values
observed in non-responders, and 30 of the non-responders
Volitional response to verbal command had values <1.
A clear volitional (IFT positive) response following laryngoscopy Most of the derived EEG indices of the alpha-delta respond-
and intubation was witnessed in six patients (6.7%). Their char- ers lay within the middle quartiles (or even towards the unex-
acteristics and anaesthetic doses and regimens are shown in pected end of the distribution), so none showed an unequivocal
Table 3. All of the responses at Sites 1 and 2 (Responders A, B delineation between responders and non-responders.
and C) occurred immediately after the initial verbal command Conversely, many of the derived indices for the Site 3 respond-
(within 1 min of intubation). At Site 3, commands were repeated ers were outside the IQRs of non-responders (Table 2), consis-
around every 10 s for a longer period after intubation. There tent with the spectral findings observed. BIS values for the
were no responses to the initial command, but responses did alpha-delta responders were all below 45. NeuroSense values
 668 | Gaskell et al.

A C
30 30

20 20

Power (dB)
10 10

0 0

–10 –10

–20 –20

0 10 20 30 0 10 20 30

B D
30 30

20 20
Power (dB)

10 10

0 0

–10 –10

–20 –20

0 10 20 30 0 10 20 30
Frequency (Hz) Frequency (Hz)

Fig 1 Frequency–power spectra for isolated forearm test (IFT) responders (colour) and non-responders (grey). Left-hand figures are from Sites 1 and 2. (A ) Prior to
laryngoscopy and intubation. (B) Following laryngoscopy and intubation. Right-hand figures are from Site 3. (C) Prior to laryngoscopy and intubation.
(D) Following laryngoscopy and intubation (at time of response for late responders). Cases with burst suppression omitted. Blue, green and pink lines represent
responders A, B and C, respectively, and gold, orange and light blue lines represent responders D, E and F, respectively.

Table 1. EEG parameters of responders vs summary statistics of non-responders at Sites 1 and 2. *Burst suppression cases excluded from
†
spectral-based parameters. Responder parameter values lying outside the IQR of non-responder parameters from the same group
(towards the extreme of the distribution typically associated with wakefulness)

Sites 1 and 2 Responder A Responder B Responder C Non-responders (n¼60)

Median (IQR)

Delta power (dB)* 23.7 25.6 24.6 25.1 (22.2–27.7)

SWAS ratio* 0.78† 0.92 0.77† 0.89 (0.82–0.96)
Theta power (dB)* 10.3 8.1 1.4† 5.9 (3.5–8.4)
Alpha peak frequency (Hz)* 9.5 9.9 10.5 9.3 (8.3–10.1)
Alpha power (dB)* 16.4 17.6 11.2 9.4 (7.0–14.4)
Alpha saturation* 0.78 0.76 1 0.77 (0.62–0.83)
Oscillatory alpha (dB)* 8.5 11.9 12.5 6.6 (4.7–9.6)
Beta power dB* –5.6† –8.3 –12.6 –9.0 (–11.1 to –7.5)
Low gamma power dB* –16.4† –19.3 –23.1 –19.0 (–21.3 to –17.2)
Permutation entropy 0.71 0.68 0.70 0.73 (0.70–0.76)
PmaxI 0.966 0.819† 0.942† 1.077 (0.959–1.171)
BIS value 44 37 44 37 (30–44)
EMG 31.2 30.6 26.9 29.6 (27.9–31.9)
Burst suppression Absent Absent Absent 15/60
 Frontal EEG findings during isolated forearm technique responses | 669

Table 2 EEG parameters of responders vs summary statistics of non-responders at Site 3. *Burst suppression cases excluded from spec-
tral-based parameters. †Responder parameter values lying outside the IQR of non-responder parameters from the same group (towards
the extreme of the distribution typically associated with wakefulness)

Site 3 Responder D Responder E Responder F Non-responders (n520)

Median (IQR)

Delta power (dB)* 5.1† 6.4† 2.8† 11.3 (8.4–14.3)

SWAS ratio* 0.29† 0.25† 0.21† 0.62 (0.53–0.73)
Theta power (dB) 1.5 –0.7† –1.6† 1.6 (0.4–3.8)
Alpha peak frequency (Hz)* 7.2 12.2 15.5 10.9 (9.7–12.9)
Alpha power (dB)* 7.4 4.4† 6.9 8.4 (4.9–11.7)
Alpha saturation* 0.53† 0.60† 0.82 0.90 (0.79–0.99)
Oscillatory alpha (dB)* 3.2† 4.7† 8.6 7.5 (5.8–8.5)
Beta power (dB)* –2.6† –7.5 –6.5 –7.1 (–9.4 to –4.9)
Low gamma power (dB)* –13.6† –18.0 –17.2 –16.9 (–18.8 to –15.7)
Permutation entropy 0.92† 0.90 0.90 0.89 (0.87–0.90)
PmaxI 0.927† 0.873† 0.943† 1.040 (0.970–1.137)
NeuroSENSE index 90† 80† 72 65 (50–73)
EMG 30.2 27.6 27.8 30.2 (27.0–33.3)
Burst suppression Absent Absent Absent 1/20

Table 3. Patient characteristics and anaesthetic regimens administered to isolated forearm test responders. OSA, obstructive sleep
apnoea; ET, end-tidal volatile anaesthetic concentration

Age Sex Weight BMI ASA Comorbidities Propofol Opioid (mg) Adjuncts (mg) Volatile prior to
(kg) (mg) laryngoscopy

A 27 F 48 21 3 VATER syndrome, end- 100 Fentanyl 100 No ET sevoflurane 2.1%

stage renal failure,
anxiety, depression
B 23 M 74 21 2 Gastro-oesphageal 200 Fentanyl 50 Midazolam 2 ET sevoflurane 1.3%
reflux, depression, Lidocaine 80
sleep disorder
C 32 F 73 28 3 Cerebral aneurysm, 150 Fentanyl 250 Midazolam 2 No
anxiety, migraine Lidocaine 100
D 40 M 95 26 1 Nil 190 Sufentanil 15 Ketamine 25 No
E 40 F 110 37 2 OSA, asthma 200 Sufentanil 10 Ketamine 25 ET desflurane 1.5%
F 38 M 115 35 2 OSA, diabetes 200 Sufentanil 10 Ketamine 25 No

for the Site 3 responders were 90, 80 and 72. PE, like BIS, delta change the state of the individuals so as to become able to make
power and SWAS, was unable to differentiate the alpha-delta a volitional response to command. This arousal was not associ-
responders from the non-responders. Log-likelihood curve anal- ated with the loss of low frequency activity and return of high
ysis suggested that the alpha-delta responders data came from frequency activity typically associated with return of respon-
probability distributions very similar to those of the non- siveness following anaesthesia (and seen in our Site 3 respond-
responding patients. ers) and was essentially not visible to the frontal EEG. In fact,
our data suggest that the only frontal EEG pattern that precludes
volitional response is burst suppression; however, slow wave
Discussion saturation or peak trough max index >1 cannot be completely
Falsifiability is a philosophy of science, articulated by Karl dismissed. Notably, and in contrast to many of the other indi-
Popper,11 that a single event is sufficient to refute some hypoth- ces, the non-responder PmaxI values from the different sites
eses: the observation of a single black swan is sufficient to dis- were comparable. This index seems to be resilient to different
prove the notion that all swans are white. Our most striking anaesthetic techniques including lower doses of volatile anaes-
finding was the occurrence of volitional responses in the pres- thetics and the disorganizing electroencephalographic effects of
ence of strong frontal alpha-delta activity; essentially we dis- ketamine.
covered three black swans. These establish that the frontal The frontal alpha-delta EEG signature has been proposed as
alpha-delta EEG pattern is not causally linked to unconscious- a plausible biomarker of anaesthetic unconsciousness, as an
ness. In these three responders, the noxious stimulus of lar- indicator of a quiescent hyperpolarized thalamo-cortical system
yngoscopy and intubation resulted in sufficient arousal to with diminished capacity for sensory information transfer to
 670 | Gaskell et al.

A
Frequency (Hz) 50 30
40 20

Power (dB)
30 10

20 0
–10
10
–20
0
100 150 200 250 300 350

B 40

30
Power (dB)

20

10

0
100 150 200 250 300 350
Time (s)

Fig 2 Spectrogram (A) of Responder A and (B) alpha (gold) and delta (blue) power demonstrating clear alpha-delta pattern at times of intubation (vertical black
line) and response to verbal command (vertical green lines). Alpha and delta power are maintained throughout the period of interest.

A 50 30
Frequency (Hz)

40 20

Power (dB)
30 10

20 0
–10
10
–20
0
400 450 500 550 600 650 700 750 800 850 900
B 40
30
Power (dB)

20

10

0
400 450 500 550 600 650 700 750 800 850 900
Time (s)

Fig 3 (A) Spectrogram of Responder E. Note different time scale from Figure 2 and artifact at 750–780 s during laryngoscopy (intubation is marked by black vertical
line). (B) Alpha (gold) and delta (blue) power. There was less lower frequency activity prior to laryngoscopy compared with Responder A but also a further loss of
low frequency activity leading up to and at the time of response (vertical green line).

the cortex, with the resultant synchronous alpha oscillations its disappearance prior to return of responsiveness during
and slow waves impairing integration of information.4–10 emergence from anaesthesia.4 20 However, unresponsive-
Multiple studies in volunteers and patients have associated the ness 6¼ unconsciousness,1 as we all understand from our experi-
appearance of this pattern with loss of responsiveness during ences of dreaming during sleep where experience is driven by
induction of propofol or volatile-agent-based anaesthesia and internal sources. In this context, while external stimuli are
 Frontal EEG findings during isolated forearm technique responses | 671

occasionally perceived during dreaming, the dream is not gener- anaesthesia monitoring should measure something that is cau-
ated by external sensory stimuli (a state of disconnected sally linked to consciousness. As such we should record from
consciousness while unresponsive to the environment). Here, sites other than, or in addition to, the frontal cortex.
we extend these observations to show that experience of, and It is reassuring that the burst response pattern was not asso-
interaction with, our environment can occur even in the ciated with connected consciousness. Unlike local sleep, local
presence of profound suppression of the frontal cortical EEG. burst responses at the EEG level have not been noted. Hence, it
Neuroscientists might find this less surprising as: (1) local sleep is likely that frontal burst suppression indicates a global sup-
occurs during wakefulness (e.g. after prolonged wakefulness pression of cortical activity that would suppress the integration
local slow waves appear in the EEG);23–25 (2) large bilateral fron- of information and therefore consciousness.
tal slow waves can be induced with muscimol injections in ani- Our sample size is small, and with only six responders we
mal models with no apparent change in behaviour;26 (3) lesion can make only limited statistical inferences from our data.
studies show that the frontal cortex is not necessary for con- However, in this case, applying the same EEG parameters to
sciousness;27 (4) frontal seizures are associated with loss of larger samples would not advance matters, it would simply
executive functions but not consciousness;28 (5) data from equate to looking harder in the wrong place—finding another 20
dreaming in sleep suggest disconnected consciousness can black swans does not change the conclusion (although it does
occur in the absence of frontal cortical activity;29 and (6) propo- tell you how common black swans might be).
fol sedation diminishes frontal, but not temporal, cortical activ- This study was a substudy of a larger observational study of
ity to ambiguous words, implying that higher-order cortical routine anaesthetic practice, and there was significant heteroge-
function can be blunted but sound perception maintained.30 31 neity in administered anaesthetic regimes. This resulted in post
Interestingly, reducing frontal activation during sedation corre- hoc analysis of one site separately. Comparing the preintubation
lated with reduced implicit memory after the event, and so it is spectra from Sites 1 and 2, it appears that ketamine masks
possible that frontal slowing may be associated with poor mem- appearance of the alpha-delta signature somewhat, despite the
ory consolidation. However, the relationship between the use of comparable induction doses of propofol. Site 3 spectra had
alpha-delta pattern and awareness with explicit recall remains lower delta and alpha activity and the alpha peak occurred at a
unknown. Recent studies of disconnected consciousness have higher frequency. These findings are consistent with a study
demonstrated that posterior cortical activity correlates closely specifically designed to investigate the propofol/ketamine inter-
with dream reports when woken from sleep.29 32 Extension of action34 and with the known effects of ketamine at higher doses
these findings to our experiment, where responders appear to in decreasing alpha and increasing gamma activity.35 Thus, we
be roused from anaesthesia by the intense noxious stimulation attribute the differences between Site 3 and Sites 1 and 2 spectra
of laryngoscopy and intubation, suggests that a similar posterior predominantly to the different anaesthetic regime typically used
region could mediate some episodes of connected conscious- at Site 3. However, it is also possible that some of the differences
ness during anaesthesia. The consistency of our findings with in spectral findings between sites were in part due to the differ-
human lesion data, demonstrating clearly that consciousness ent EEG acquisition systems used (e.g. electrodes systems, inbuilt
does not require frontal cortical activity, is most compelling. noise reduction and filtering).
Future work must consider more broadly the brain regions These differences have made interpretation of the EEG find-
monitored during anaesthesia. ings somewhat problematic. Additionally, the alpha-delta
Our findings are consistent with previous studies in which responses at Sites 1 and 2 were clearly temporally related to the
behavioural responsiveness has been observed in patients with noxious stimuli of intubation whereas two of the Site 3
processed EEG depth-of-anaesthesia values lying well within responses occurred significantly later (around 120 and 160 s
manufacturers’ recommended target range for general anaes- after intubation), and only after repeated commands (which
thesia.13–15 Indeed in 154 patients in our full study dataset, were not executed at the other sites), when some decay in pro-
numerical values for depth of anaesthesia monitors were avail- pofol concentrations might have occurred. Our ‘black swans’
able.16 At the population level, monitor values were higher in from Sites 1 and 2 suggest that the alpha-delta pattern might be
IFT responders than non-responders; however, receiver-operat- better regarded as a sign of gamma-amino-butyric-acid(GABA)-
ing curve analysis demonstrated that processed depth of anaes- ergic-induced hyperpolarization than unconsciousness per se.
thesia values were not able to reliably discriminate between IFT The alpha-delta signature might still be considered evidence of
responders and non-responders. We performed spectral analy- GABA-ergic agent delivery, but the return of responsiveness
sis around the time of the IFT response to test whether we could induced by the stimulus of laryngscopy and intubation was not
identify an EEG pattern suggestive of unconsciousness, or if an accompanied by loss of this pattern. However, it should be
issue such as loss of EMG, noise, artifact, or algorithm or equip- noted that GABA-ergic drug delivery is not always accompanied
ment malfunction affected the monitor output values. The by the classic alpha-delta signature, particularly in the presence
alpha-delta responders we observed had frontal EEG spectral of ketamine, which has a wide variety of effects on synaptic
appearances typical of anaesthetized patients, indicating that function and neuromodulator activity,36 which might partially
the low depth of anaesthesia values, and failure of the monitor explain this observation. In contrast, at the time of their
to detect the IFT response, were probably not due to muscle- responses the Site 3 responders probably had low GABA-ergic
relaxant-induced loss of EMG, which has been identified as a activity, and this might explain why their frontal EEG responses
limitation of the BIS monitor.33 The Site 3 responders had spec- were more consistent with those traditionally associated with
tral findings more consistent with cortical depolarization, which normal wakefulness.
might have been caused by the ketamine, low or absent volatile The existence of the three patients from Sites 1 and 2 who
anaesthetic doses, or by declining propofol concentrations. This were responsive in the presence of a strong alpha-delta pattern
was detected appropriately by the NeuroSENSE monitor in the means this pattern is not sufficient for unconsciousness.
presence of muscle relaxants. However, the proprietary indices Conversely, the existence of many patients from Site 3 who
would have ‘missed’ half of the IFT responders in this study. We were unresponsive in the absence of an alpha-delta pattern
argue that any measure that is proposed to advance depth of shows that this pattern is not necessary for unconsciousness.
 672 | Gaskell et al.

An alternative, and perhaps controversial, explanation of the was supported by a scholarship from the Australian and New
considerable overlap we have observed in frontal EEG patterns Zealand College of Anaesthetists.
in behaviourally distinct (responsive and unresponsive) states,
is that many of the non-responding patients were, in fact, con-
scious. It is probable that isolated forearm testing underesti- References
mates the incidence of real-time intraoperative awareness due 1. Sanders RD, Tononi G, Laureys S, Sleigh JW. Unresponsiveness
to effects of anaesthetic agents on motivation to respond, cogni- 6 unconsciousness. Anesthesiology 2012; 116: 946–59
¼
tive processing and the motor response.1 For instance ketamine 2. Nordstrom O, Sandin R. Recall during intermittent propofol
might have contributed to the lack of responsiveness in some of anaesthesia. Br J Anaesth 1996; 76: 699–701
the Site 3 non-responders, who had EEG findings more tradi- 3. Tunstall ME. Detecting wakefulness during general anaes-
tionally associated with wakefulness. Furthermore, any poten- thesia for caesarean section. Br Med J 1977; 1: 1321
tially conscious but disconnected (dreaming) individuals who 4. Purdon PL, Pierce ET, Mukamel EA, et al. Electroencephalo-
were unaware of their environment, would also be defined as gram signatures of loss and recovery of consciousness from
IFT negative. Thus, it is possible that many more patients who propofol. Proc Natl Acad Sci USA 2013; 110: E1142–51
showed a typical alpha-delta EEG pattern were actually in a 5. Purdon PL, Sampson A, Pavone KJ, Brown EN.Clinical
state of at least partial consciousness. Conversely, we also electroencephalography for anesthesiologists: Part I: back-
assume that an appropriate response is evidence of conscious- ground and basic signatures. Anesthesiology 2015; 123: 937–60
ness (i.e. the subjective experience of the auditory command). 6. Chander D, Garcia PS, MacColl JN, Illing S, Sleigh JW.
We cannot completely exclude the possibility that the informed Electroencephalographic variation during end maintenance and
consent process conditioned participants for a somewhat auto- emergence from surgical anesthesia. PLoS One 2014; 9: e106291
mated response to the sound of their name. In addition, as no 7. Brown EN, Lydic R, Schiff ND. General anesthesia, sleep, and
recall was reported in our responders, it is possible that the coma. N Engl J Med 2010; 363: 2638–50
alpha-delta signature is associated with (but not a requisite for) 8. Feshchenko VA, Veselis RA, Reinsel RA. Propofol-induced
anaesthetic-induced amnesia. Given the prominent role of the alpha rhythm. Neuropsychobiology 2004; 50: 257–66
frontal cortex in higher-order cognition it may be that the 9. Ching S, Cimenser A, Purdon PL, Brown EN, Kopell NJ.
alpha-delta responders still had a degree of cognitive impair- Thalamocortical model for a propofol-induced alpha-
ment not detected by the simple command we used—the quali- rhythm associated with loss of consciousness. Proc Natl Acad
tative natures of potential IFT responsive states remain to be Sci USA 2010; 107: 22665–70
fully elucidated and IFT responsiveness should not necessarily 10. Sleigh J, Scheib C, Sanders R. General anaesthesia and elec-
be considered equivalent to the normal awake state. troencephalographic spindles. Trends Anaesth Crit Care 2011;
1: 263–9
11. Popper KR. A survey of some fundamental problems. In: The
Conclusions
Logic of Scientific Discovery. London: Routledge, 1959; 3–27
Frontal EEG parameters do not readily discriminate volitional 12. Russell IF. The Narcotrend ‘depth of anaesthesia’ monitor
responsiveness (a marker of connected consciousness) and cannot reliably detect consciousness during general anaes-
unresponsiveness during anaesthesia. The alpha-delta domi- thesia: an investigation using the isolated forearm techni-
nant frontal EEG signature is not sufficient to ensure uncon- que. Br J Anaesth 2006; 96: 346–52
sciousness during intended general anaesthesia. 13. Russell IF. The ability of bispectral index to detect intra-
operative wakefulness during isoflurane/air anaesthesia,
compared with the isolated forearm technique. Anaesthesia
Authors’ contributions 2013; 68: 1010–20
Data collection, data analysis, writing of manuscript: A.L.G. 14. Russell IF. The ability of bispectral index to detect intra-
Data collection, data analysis, review of manuscript: D.F.H. operative wakefulness during total intravenous anaesthesia
Data collection, review of manuscript: J.W., G.T., A.D., V.B. compared with the isolated forearm technique. Anaesthesia
Data collection, revision of manuscript: A.R. 2013; 68: 502–11
Study design, data analysis, writing of manuscript: J.W.S. 15. Schneider G, Wagner K, Reeker W, Hanel F, Werner C, Kochs
Study design, data collection, writing of manuscript: R.D.S. E. Bispectral Index (BIS) may not predict awareness reaction
Reviewed and approved the final draft of manuscript: all to intubation in surgical patients. J Neurosurg Anesthesiol
authors. 2002; 14: 7–11
16. Sanders RD, Gaskell A, Raz A, et al. Incidence of connected
Declaration of interest consciousness after tracheal intubation: a prospective, inter-
national, multicenter cohort study of the isolated forearm
V.B. unrestricted grant Euros from Orion Pharma (Espoo, technique. Anesthesiology 2017; 126: 214–22
Findland) for another study. Editor-in-Chief of Acta 17. Bokil H, Andrews P, Kulkarni JE, Mehta S, Mitra PP. Chronux:
Anaesthesiologica Belgica. a platform for analyzing neural signals. J Neurosci Methods
R.D.S. is a member of the Editorial Board of the British Journal of 2010; 192: 146–51
Anaesthesia. J.W.S. is the associate editor of Anesthesiology. 18. Leslie K, Sleigh J, Paech MJ, Voss L, Lim CW, Sleigh C.
A.L.G., D.F.H., J.W., G.T., A.D., A.R.: no interests to declare. Dreaming and electroencephalographic changes during
anesthesia maintained with propofol or desflurane.
Anesthesiology 2009; 111: 547–55
Funding 19. Olofsen E, Sleigh JW, Dahan A. Permutation entropy of the
This study was funded from departmental resources. D.F.H. electroencephalogram: a measure of anaesthetic drug effect.
was supported by the James S McDonnell Foundation and A.L.G. Br J Anaesth 2008; 101: 810–21
 Frontal EEG findings during isolated forearm technique responses | 673

20. Ni Mhuircheartaigh R, Warnaby C, Rogers R, Jbabdi S, Tracey constituent functions: a video-EEG study in patients with
I. Slow-wave activity saturation and thalamocortical isola- focal epilepsy. Brain 2002; 125: 2691–8
tion during propofol anesthesia in humans. Sci Transl Med 29. Koch C, Massimini M, Boly M, Tononi G. Neural correlates of
2013; 5: 208ra148 consciousness: progress and problems. Nat Rev Neurosci 2016;
21. Tort AB, Komorowski R, Eichenbaum H, Kopell N. Measuring 17: 307–21
phase-amplitude coupling between neuronal oscillations of 30. Davis MH, Coleman MR, Absalom AR, et al. Dissociating
different frequencies. J Neurophysiol 2010; 104: 1195–210 speech perception and comprehension at reduced levels of
22. Mukamel EA, Pirondini E, Babadi B, et al. A transition in brain awareness. Proc Natl Acad Sci USA 2007; 104: 16032–7
state during propofol-induced unconsciousness. J Neurosci 31. Adapa RM, Davis MH, Stamatakis EA, Absalom AR, Menon
2014; 34: 839–45 DK. Neural correlates of successful semantic proce
23. Vyazovskiy VV, Olcese U, Hanlon EC, Nir Y, Cirelli C, Tononi ssing during propofol sedation. Hum Brain Mapp 2014; 35:
G. Local sleep in awake rats. Nature 2011; 472: 443–7 2935–49
24. Funk CM, Honjoh S, Rodriguez AV, Cirelli C, Tononi G. Local 32. Siclari F, LaRocque JJ, Bernardi G, Postle BR, Tononi G. The
slow waves in superficial layers of primary cortical areas neural correlates of consciousness in sleep: a no-task,
during REM sleep. Curr Biol 2016; 26: 396–403 within-state paradigm. bioRxiv 2014; 012443
25. Bernardi G, Siclari F, Yu X, et al. Neural and behavioral corre- 33. Schuller PJ, Newell S, Strickland PA, Barry JJ. Response of bis-
lates of extended training during sleep deprivation in pectral index to neuromuscular block in awake volunteers.
humans: evidence for local, task-specific effects. J Neurosci Br J Anaesth 2015; 115: i95–i103
2015; 35: 4487–500 34. Hayashi K, Tsuda N, Sawa T, Hagihira S. Ketamine increases
26. Leung LS, Petropoulos S, Shen B, et al. Lesion of cholinergic the frequency of electroencephalographic bicoherence peak
neurons in nucleus basalis enhances response to general on the alpha spindle area induced with propofol. Br J Anaesth
anesthetics. Exp Neurol 2011; 228: 259–69 2007; 99: 389–95
27. Mataro M, Jurado MA, Garcia-Sanchez C, Barraquer L, Costa- 35. Lee U, Ku S, Noh G, Baek S, Choi B, Mashour GA. Disruption of
Jussa FR, Junque C. Long-term effects of bilateral frontal frontal-parietal communication by ketamine, propofol, and
brain lesion: 60 years after injury with an iron bar. Arch sevoflurane. Anesthesiology 2013; 118: 1264–75
Neurol 2001; 58: 1139–42 36. Garcia P, Sleigh J. Ketamine: a drug at war with itself.
28. Lux S, Kurthen M, Helmstaedter C, Hartje W, Reuber M, Elger Anesthesiology 2017; 126: 371–2
CE. The localizing value of ictal consciousness and its
Handling editor: Hugh C Hemmings Jr