British Journal of Anaesthesia 1994; 72: 397^02
Respiratory sinus arrhythmia: comparison with EEG indices during
isoflurane anaesthesia at 0.65 and 1.2 MAC
C. J. D. POMFRETT, J. R. SNEYD, J. R. BARRIE AND T. E. J. HEALY
SUMMARY
KEY WORDS
Anaesthesia: depth. Heart: heart rate variability. Brain: electro encephalography.
Respiratory sinus arrhythmia (RSA) is a nonpathological change in instantaneous heart rate under
the control of a parasympathetic respiratory reflex.
The level of RSA may be used as an index of vagal
tone and, therefore, as a window on autonomic
activity processed in the nucleus solitarius of the
brainstem [1, 2]. Most research on RSA has focused
on using it as a method of screening diabetics
PATIENTS AND METHODS
We studied 10 ASA I or II patients (mean age
42.5 yr, range 32-59 yr; mean weight 69.9 (SD 15.6)
kg, range 45-97 kg) undergoing elective breast or
gynaecological surgery; the patients gave informed
consent to a study design approved by the local
Ethics Committee. Anaesthesia was induced with
propofol 2.5 mg kg"1 and fentanyl 100 ug and maintained with 66 % nitrous oxide in oxygen and
alternating 10-min periods of 0.85% or 1.7% endtidal isoflurane, measured using a calibrated anaesthetic gas analyser (Datex, Normac or Ultima;
steps corresponding to 0.65 and 1.2 MAC). Vecuronium 0.1 mg kg"1 was given at induction and the
lungs were ventilated mechanically via a tracheal
tube. Bradycardia during anaesthesia was treated
with atropine 0.2-0.6 mg. Neuromuscular block was
antagonized with neostigmine 2.5 mg and atropine
600 ug or glycopyrronium 0.5 mg.
Anaesthetic and surgical events were logged using
a free-text database and the event data were timelocked to the appropriate ECG, EEG and RSA data
files in order to facilitate off-line analysis. Additional,
standard theatre monitors supplied data regarding
oxygen saturation and arterial pressure.
Data acquisition methods have been described
elsewhere [5,6,9]. RSA was quantified using a
C. J. D. POMFRETT, BSC, PH.D., J. R. SNEYD, M.D., M.A., F.R.C.A.,
J. R. BARRIE, B.SC., M.B., CH.B., F.R.C.A., T. E. J. HEALY, M.SC.,
M.D., F.R.C.A., Department of Anaesthesia, University of Manchester, Withington Hospital, Manchester M20 8LR. Accepted
for Publication: November 11, 1993.
Correspondence to C.J.D.P.
Downloaded from http://bja.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016
Respiratory sinus arrhythmia (RSA) is a cyclical
variation in heart rate during breathing, where the
heart rate increases during inspiration and decreases
during expiration. RSA and the electroencephalogram (EEG) were monitored in 10 patients undergoing elective surgery with isoflurane and nitrous
oxide in oxygen anaesthesia after induction with
propofol. All patients were subject to controlled
ventilation and recovery from competitive neuromuscular block was facilitated by neostigmine and
glycopyrronium (seven patients) or atropine (three
patients). Median and spectral edge (95%) frequencies of the raw EEG were derived off-line. RSA
and EEG indices were obtained during preinduction
(baseline), induction, incision, 0.65 and 1.2 MAC
of isoflurane maintenance during surgery and
recovery. Significant decreases in the level of RSA,
median and spectral edge frequencies were observed during induction and significant increases in
all indices were observed at recovery in all patients.
Significant decreases in the median and spectral
edge EEG frequencies occurred in patients treated
with atropine both to counteract bradycardia after
propofol induction and at antagonism of neuromuscular block (n = 3), compared with patients
treated with glycopyrronium (n = 7). In contrast,
the level of RSA did not decrease significantly with
atropine. It is concluded that measurements of RSA
could form the basis of a useful index of anaesthetic
depth during isoflurane anaesthesia, even during
the use of pharmacologically appropriate doses of
atropine. However, any effects of atropine on the
raw EEG and on indices derived from the EEG,
should be characterized further so that these effects
are not confused with changes in anaesthetic depth.
(Br. J. Anaesth. 1994; 72: 397-402)
exhibiting autonomic neuropathy [3]. However,
interest in RSA changes during anaesthesia began
when an off-line study suggested that on-line analysis
of RSA could provide a method of measuring the
depth of anaesthesia in patients anaesthetized with
isoflurane [4]. We have previously described changes
in RSA and median EEG frequency during propofol
anaesthesia [5] and RSA during propofol sedation
[6]. The objective of this study was to test if RSA
may be used as an index of anaesthetic depth during
two standardized levels of isoflurane anaesthesia. We
measured RSA in real-time and compared it with
median and spectral edge frequencies of the EEG,
which have been presented as indices of anaesthetic
depth [7]. Some of these data have been reported
briefly in conference proceedings [8].
BRITISH JOURNAL OF ANAESTHESIA
398
0.12i
0.081
12
3 4
5
6
7
8
6000
1011
Elapsed time (s)
2000
1
2
3
4
5
4000
6000
6 78 9
10 1113151617
1214
8000
Elapsed time (s)
FIG. 1. Off-line record from one patient depicting the time course
of RSA, EEG median frequency (MF) and mean (SD) ECG R-R
interval. Events: 1 = propofol; 2 = intubation; 3 = 0.85% isoflurane; 4 = 1 . 7 % isoflurane; 5 = 0.85% isoflurane; 6 = disconnect circuit; 7 = restart in theatre; 8 = incision; 9 = 1.7%
isoflurane; 10 = 0.85% isoflurane; 11 = 1.7% isoflurane; 12 =
surgery discontinued; 13 = stitching; 14 <= neostigmine; 15 •=
isoflurane off; 16 = spontaneous respiration; 17 = disconnect.
( • ) = 1.2, ( 0 ) = 0.65 MAC of isoflurane.
= interruption in
isoflurane. The 5 Hz line across median frequency denotes the
empirical guideline at which anaesthesia has been considered to
be adequate.
modified version of a technique used for screening
diabetics [3], which has been described in detail
elsewhere [5,6]. The EEG was obtained from
silver-silver chloride electrodes fixed to the scalp at
frontal and mastoid locations, with electrode impedance maintained at less than 7 kQ. EEG waveforms were analogue bandpass filtered from 0.1 to
40 Hz (-3-dB points, Neurolog NL125) with a
50-Hz (mains line frequency, 20 Hz width at — 3-dB
points) notch filter. The EEG, ECG and respiratory
waveforms were digitized (12 bits; 1 kHz) and stored
using a microcomputer-based data logging system
[9]. Each patient record exceeded 20 MBytes of
computer storage capacity. All raw EEG waveforms
were reviewed off-line in 6-s epochs to exclude
artefacts and periods of burst suppression, which
FIG. 2. Off-line record from one patient depicting the time course
of RSA, EEG median frequency (MF) and mean (SD) ECG R-R
interval, and systolic arterial pressure (SAP). Events: 1 =
propofol and atropine; 2 = 1 . 7 % isoflurane; 3 = 0.85% isoflurane^ = incision; 5 = 1.7% isoflurane; 6 = 0.85% isoflurane;
7 = 1.7% isoflurane; 8 = 0.85% isoflurane; 9 = isoflurane off,
antagonism of paralysis and atropine; 10 = coughing; 11 =
responding. ( • ) = 1.2, ( 0 ) = 0.65 MAC of isoflurane.
have been noted as adversely affecting power spectral
analysis [10]. Raw EEG records were checked also
for contamination by ECG waveforms, and rejected
if any were seen. Failure to reject such records could
have led to changes in heart rate affecting spectral
analysis of the EEG.
Fast Fourier transformation (Hanning window;
2.048-s intervals, extracted from 6-s samples of
EEG; 2048 points) and power spectral analysis were
performed on the EEG to determine the median and
spectral edge (95%) frequencies using software
written for the study incorporating standard algorithms [11]. R, median and spectral edge frequency
time series were subjected to exponential smoothing
with appropriate non-seasonal models. Intra-patient
predictive autoregressive integrated moving averages
(ARIMA) were determined using standard software
[12]. ARIMA analysis describes disturbances or
shocks in a time series and mathematically defines
them within models. These models were used to
forecast the future behaviour of R based on its
behaviour over the preceding 15-min interval.
ARIMA models were constructed at interventions in
the time series caused by changes in the level of
Downloaded from http://bja.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016
4000
2000
RESPIRATORY SINUS ARRHYTHMIA AND EEG INDICES
anaesthesia and control charts were prepared showing the 99 % confidence limits for a change from the
predicted behaviour of R. Inter-patient grouped
analysis was performed on 2-min data samples
obtained from the time series of each patient, 1 min
after events such as step changes in isoflurane
concentration. Grouped data sets were checked for
normality and equivalence of variance (Levene's
test) and inter-group comparisons were performed
with Student's two-tailed t tests and one-way
analysis of variance (ANOVA). All statistical analysis
was performed using standard software [12] running
on an Intel 486-based PC. Groups were tested
against the RSA or EEG indices seen at recovery, as
avoidance of awakening during anaesthesia is the
main objective of any putative monitor of anaesthetic
depth.
All patients showed changes in the level of RSA,
median and spectral edge indices as a result of
anaesthesia. Figure 1 shows a typical time course for
some indices in one patient. This record was of
particular interest because, as a result of unexpected
delay in the operating room, the patient was anaesthetized in an adjacent anaesthetic room 1 h before
surgery commenced. Marked reductions in the level
of RSA and both EEG indices were observed at
induction of anaesthesia with propofol. Large increases were observed in both RSA and median EEG
frequencies as the propofol bolus dose for induction
wore off, although the patient was breathing nitrous
oxide and oxygen. Isoflurane caused the level of RSA
and median frequency to diminish. Step changes
between 0.65 and 1.2 MAC of isoflurane resulted in
step changes in the level of RSA and similar changes
in median EEG frequency. Recovery from anaesthesia, in the presence of neostigmine and glycopyrronium, was accompanied by an increase in
the level of RSA and increases in EEG median
frequency.
Figure 2 shows the effect of administering atropine
during the procedure in another patient, to antagonize bradycardia after propofol induction, and at
the end of the procedure with antagonism of residual
neuromuscular block with neostigmine. Levels of
RSA increased progressively as a result of stimulation caused by intubation, as did the level of
median EEG frequency. However, even after treatment with atropine 100 ug to counteract bradycardia
associated with induction, the level of RSA changed
in response to changes in levels of isoflurane.
Recovery was accompanied by a marked increase in
the level of RSA, occurring before a corresponding
increase in median EEG frequency.
Figure 3 shows the results from analysis of
grouped data from all 10 patients. One 2-min sample
(four, 30-s data points) was taken from the time
series data, with a 1-min delay (two, 30-s data points)
after a change in anaesthetic level, to allow the
anaesthetic to take effect. The grouped data show
that RSA and the EEG indices were reduced during
increasing levels of anaesthesia. RSA decreased
significantly after propofol induction (P < 0.05) and
during isoflurane anaesthesia (0.65 MAC, P < 0.05,
1.2 MAC, P<0.01). The level of RSA increased
during incision and was not significantly different
from the level seen at recovery. The EEG spectral
edge frequency before anaesthesia was significantly
different from that after recovery (P < 0.01). Spectral edge frequency was not significantly different
from that at recovery after propofol induction or
incision, but was significantly different during
isoflurane anaesthesia (P < 0.001). The median frequency of the EEG was significantly different from
that at recovery after propofol induction (P < 0.05),
during incision (P < 0.01) and isoflurane anaesthesia
(P < 0.001). Neither the EEG median frequency nor
the spectral edge frequency increased during
incision.
Figure 4 shows the mean data for the three
patients given atropine after propofol induction and
at antagonism of competitive neuromuscular block
0.07-
r40
0.06-
-35
-30 0
_ 0.05-
-10
ll
-25 x
£ 0.04-
CD
••
1-20 -g"
c
< 0.03w
oc
CD
-5
|
o
CD
"10 £
0.02-
-5
0.01L
0-
15
Incis.
0.8% I
1.7% I
Rec.
0
FIG. 3. Mean (SEM) grouped data obtained from 2-min samples of EEG median frequency (MF), RSA indices and
mean spectral edge frequency, 1 min after a change in the level of isoflurane anaesthesia for all 10 patients. Significance
was tested using Student's t test, comparing mean responses against those seen at recovery: *P < 0.05, **P < 0.01,
***P < 0.001. The bar across median frequency represents the 5 Hz empirical guideline suggested to denote the
level at which anaesthesia is adequate [7]. Pre. I = Before induction; Ind. = induction. Incis. = incision; 0.8% I,
1.7% I = concentrations of isoflurane; Rec. = recovery.
Downloaded from http://bja.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016
RESULTS
399
BRITISH JOURNAL OF ANAESTHESIA
400
-10
0.05n
r 30
DISCUSSION
D
a
'M
X
—
M
-isf
ral
X
LU
LLI
s
Q.
Lo
Glycopyrronium
LQ
Atropine
compared with the seven patients given glycopyrronium only at antagonism. There was no significant
difference (ANOVA, P = 0.47) between grouped
levels of RSA on treatment with glycopyrronium or
atropine. However, both median and spectral edge
frequencies were significantly lower after treatment
with atropine (P < 0.01) compared with glycopyrronium.
ARIMA intervention analysis at changes in the
level of isoflurane maintenance showed that increasing isoflurane decreased the level of RSA and
vice versa. Figure 5 shows the effect of decreasing
the level of isoflurane in one patient. Within 5 min of
decreasing isoflurane from 1.2 to 0.65 MAC, the
level of RSA increased significantly (P < 0.01) and
remained high.
0.02
6027
6121
6209
6271
6338
6462
6557
6619
6854
6916
Elapsed time (s)
FIG. 5. ARIMA control chart demonstrating the effect of decreasing isoflurane anaesthesia (1.2-0.65 MAC) on RSA
in one patient.
= R, • • • = predicted behaviour of R, • • • = 99 % confidence limits.
Downloaded from http://bja.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016
FIG. 4. Mean (SEM) grouped data of RSA, EEG median
frequency (MF) and mean spectral edge frequency from all
patients after induction of anaesthesia until recovery, comparing
the effects of atropine administered after induction and at
antagonism of neuromuscular block (three patients) with glycopyrronium given only at antagonism (** P < 0.01; ANOVA).
Measurements of vagal tone using studies of heart
rate variability have been described as an important
window into the activity of the nervous system [2].
Anaesthesia suppresses activity progressively in the
central nervous system and it has been proposed that
indices of heart rate variability reflect the depth of
anaesthesia [4, 5]. However, in common with the
EEG, vagal tone is prone to several confounding
influences which may not result from the level of
anaesthetic depth. It is important, therefore, to
isolate specific components from both the EEG and
heart rate variability which minimize these artefacts.
However, regardless of the number of data processing steps, the raw EEG, and any spectral
frequency derived from it, represents the summed
potential from only the upper 2 mm of cerebral
cortex. Only evoked potentials represent activity
from deeper centres, transmitted to the scalp by
volume conduction. The cerebral cortex can reflect
cognitive awareness under anaesthesia, but may not
reflect the first signs of wakening, if those signs
originate from deeper centres of the brain with an
influence on the brainstem, or from the brainstem
itself. RSA is an ideal method of detecting waking
arousal in the central nervous system, as it is a
component of heart rate variability which reflects
parasympathetic activity originating in the brainstem.
Several groups, using fundamentally different
methods compared with the present study, have
reported that changes in the level of RSA mirror
changes in anaesthetic level. One study [13] demonstrated a marked reduction in high frequency power
of the cardiac spectral distribution on increasing the
level of isoflurane anaesthesia progressively from no
anaesthesia to 1.0, 1.5 and 2.0 MAC. The high
RESPIRATORY SINUS ARRHYTHMIA AND EEG INDICES
We observed significant reductions in both median
and spectral edge frequencies of the EEG in patients
treated with atropine (figs 2, 4). This reduction may
explain the reason why the EEG median frequencies
of many of our patients in both this and an earlier
study [5] were generally greater than the 5 Hz
empirical guideline for adequate surgical anaesthesia
adopted by Schwilden and co-workers [7, 10].
Future studies may show that atropine reduces
median and spectral edge frequencies in a dosedependent manner. Few previous studies have
mentioned any effect of atropine on the EEG.
Atropine 0.5 mg in two awake subjects did not show
large changes in the compressed spectral array (CSA)
of the EEG [17]. That study did not, however,
measure median or spectral edge frequencies and
close visual inspection of both the CSA and raw
EEG does indeed show some differences before and
after administration of atropine. A subsequent study
of EEG indices in volunteers treated with atropine
for protection against the effects of cholinesteraseinhibiting nerve agents, showed significant increases
in delta power, decreases in alpha power and
reductions in beta and theta frequency after treatment with atropine [18]. We suggest, therefore,
that further studies on the effect of atropine and
other premedicants on the EEG must be performed
so that any changes in the spontaneous EEG, or
derived indices such as the auditory evoked response,
caused by these drugs may be differentiated from
changes in EEG frequency as a result of the depth of
anaesthesia. It is also important that potential
contamination of the EEG by ECG artefacts is
eliminated before spectral analysis, otherwise
changes in heart rate during recovery may falsely raise
the median and, to a lesser extent, the spectral edge
frequency of the EEG.
There are several advantages to using RSA
analysis to determine the depth of anaesthesia. These
include the use of standard ECG and respiratory
signals and the inherently fast, artefact-free signal
analysis. RSA also shows a more marked response to
surgical stimulation than either of the EEG indices
examined. This is a potentially useful feature, giving
die anaesthetist the ability to increase the level of
anaesthesia for a short period of severe surgical
stimulation, which would otherwise not affect EEG
indices due to their relatively long latency of response
to lightening anaesthesia. The only potential exclusion criteria for the use of RSA are severe
autonomic neuropathy and after heart transplantation. In both of these instances, components of the
parasympathetic reflex loop governing RSA are
disrupted.
In conclusion, we suggest that RSA can be used to
stage the level of anaesthesia during isoflurane
maintenance. In addition, RSA is responsive to
surgical stimulation during anaesthesia, allowing the
anaesthetist to gauge anaesthesia not only by the
need for an empirical baseline, but also on the needs
of the patient for increased anaesthesia based on the
physiological response to noxious stimuli, even
during competitive neuromuscular block.
REFERENCES
1. Genovely H, Pfeifer MA. R-R variation: The autonomic test
of choice in diabetes. Diabetes/Metabolism Reviews 1988; 4:
255-271.
2. Porges SW. Vagal mediation of respiratory sinus arrhythmia.
Implications for drug delivery. Annals of the New York
Academy of Sciences 1991; 618: 57-66.
3. Weinberg CR, Pfeifer MA. An improved method for
measuring heart-rate variability: assessment of cardiac autonomic function. Biometrics 1984; 40: 855-861.
4. Donchin Y, Feld JM, Porges SW. Respiratory sinus
arrhythmia during recovery from isoflurane-nitrous oxide
anesthesia. Anesthesia and Analgesia 1985; 64: 811-815.
5. Pomfrett CJD, Barrie JR, Healy TEJ. Respiratory sinus
arrhythmia: An index of light anaesthesia. British Journal of
Anaesthesia 1993; 71: 212-217.
6. Wang DY, Pomfrett CJD, Healy TEJ. Respiratory sinus
Downloaded from http://bja.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016
frequency component of the cardiac spectral distribution was attributed to RSA and the results
showed the same trend as we have observed.
However, the previous study was off-line and was
performed on 256-s epochs of ECG R-wave timings.
This implies that a temporal resolution of only
approximately 4.5 min could be obtained. In addition, the fast Fourier transform (FFT) based
spectral analysis used in the previous study depends
on an epoch being a representative sample of a
theoretically infinite data continuum. This implies
that artefacts in the ECG record are not rejected
easily on-line. This contrasts with the circular
statistical analysis which we have used in this study,
which is particularly selective and robust against
contamination from cardiac artefacts other than
RSA, such as ectopic beats and changes in heart rate
[3]. We are also able to resolve changes in the RSA
characteristics within 30 s, making our technique
very suitable for incorporation into a usable patient
monitor.
We were particularly interested in assessing the
effect of pharmacologically suitable doses of atropine
on RSA, as the abolition of RSA with high-dose
atropine is a demonstration of its parasympathetic
origin. Previous work in unanaesthetized subjects
[14] and the results shown here during isoflurane
anaesthesia, suggest that RSA is attenuated, but not
abolished, by standard doses of atropine, as used
during clinical anaesthesia. The three atropinetreated patients demonstrated visually obvious increases in the level of RSA on recovery, but these
changes did not achieve statistical significance because of the small sample size (n = 3). We feel that
our findings are not in conflict with established
literature, as our technique is capable of determining
RSA at very low levels, certainly below the classic
definitions of RSA, for example as a variation in the
ECG P-P interval of greater than 120 ms or variation
in sinus cycle length of 10% or more [15, 16]. We
observed that RSA, median and spectral edge EEG
frequencies did not change markedly in the grouped
data when atropine-treated patients were subjected
to isoflurane concentrations of 0.65-1.2 MAC,
although individual records still showed that the
responses did change (fig. 2). We suggest that
atropine in pharmacologically relevant doses serves
to attenuate, but not abolish, changes in RSA and
EEG indices in response to changing anaesthetic
depth.
401
402
7.
8.
9.
10.
11.
arrhythmia: A new, objective sedation score. British Journal
of Anaesthesia 1993; 71: 345-358.
Schwilden H. Use of the median EEG frequency and
phannacokinetics in determining depth of anaesthesia.
Bailieres Clinical Anaesthesiology 1989; 3: 603-621.
Pomfrert CJD, Sneyd JR., HeaJy TEJ. Both respiratory sinus
arrhythmia and the EEG median frequency reflect step
changes in the level of isoflurane anaesthesia. European
Journal of Anaesthesiology 1993; 10: 137-138P.
Sneyd JR., Pomfrett CJD, Healy TEJ. Continuous acquisition
and real-time analysis of EEG, ECG and respiratory data in
the anaesthetized human. Journal of Physiology {London)
1992; 446: 76P.
Schwilden H, Stoeckel H. Quantitative EEG analysis during
anaesthesia with isoflurane in nitrous oxide at 1.3 and
1.5 MAC. British Journal of Anaesthesia 1987; 59: 738-745.
Press WH, Flannery BP, Teukolsky SA, Vetterling WT.
Numerical Recipes in C. Cambridge: Cambridge University
Press, 1988.
BRITISH JOURNAL OF ANAESTHESIA
12. SPSS/PC + Trends v4, MS-DOS Version. Chicago: SPSS
Inc.
13. Kato M, Komatsu T, Kimura T, Sugiyama F, Nakashima K,
Shimada Y. Spectral analysis of heart rate variability during
isoflurane anesthesia. Anesthesiology 1992; 77: 669-674.
14. Alimelkkila T, Kaila T, Antila K, Halkola L, Iisalo E. Effects
of glycopyrrolate and atropine on heart rate variability. Ada
Anaesthesiologica Scandinavica 1991; 35: 436—441.
15. Mandel J. Cardiac Arrhythmias, their Mechanisms, Diagnosis
and Management. Philadelphia, Toronto: J. B. Lippincort
Co., 1980; 112.
16. Rowlands DJ. Understanding the electrocardiogram: III
Rhythm abnormalities. ICI, 1987.
17. Pichmayer I. EEG Atlas for Anaesthesiologists. Berlin:
Springer-Verlag, 1987.
18. Pickworth WB, Herning RI, Koeppl B, Henningfield JE.
Dose-dependent atropine-induced changes in spontaneous
electroencephalogram in human volunteers.
Military
Medicine 1990; 155: 166-170.
Downloaded from http://bja.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016