During moderate and deep levels of isoflurane anesthesia, the electroencephalogram (EEG) shows burst suppression patterns, i.e., high amplitude bursts of activity intermingled with periods of quiescence . Traditionally, clinical indicators of deep anesthesia have been reduced musculoskeletal and autonomic responses to intraoperative stimuli. The brain's reactivity to intraoperative stimuli is not monitored routinely. Although patients' reactions during general anesthesia are reflected in heart rate and movement it should be preferable to study these reactions from the brain itself. EEG provides a noninvasive method for studying basic cerebral neurophysiologic functions, such as arousal and reactivity during anesthesia. On the other hand EEG research during anesthesia provides a basis for developing neuromonitoring methods with which neuronal damage due to surgical trauma or ischemia may be detected or even avoided.
We studied the cortical reactivity to different stimulus modalities during moderate and deep isoflurane anesthesia in which the EEG shows burst suppression patterns. Previously auditory , somatosensory [3,4], and visual  stimulation have been shown to evoke bursts. We are not aware of any studies comparing the EEG reactivity and responses to different stimulus modalities during burst suppression. We wanted to determine whether bursts evoked by different stimulus modalities are merely nonspecific responses to activation or whether they have specific characteristics depending upon the stimulus. We were also interested to know how increasing anesthesia affects EEG reactivity at burst suppression level. Therefore we studied the reactivity to visual, auditory, and painful stimulation at two levels of burst suppression anesthesia.
The study was approved by the Ethical Committee of Tampere University Hospital Written, informed consent was obtained from 15 patients (ASA class I) scheduled for elective gynecologic surgery. We excluded patients with suspected drug or alcohol abuse or who smoked excessively, and those on medication known to alter the EEG.
All patients were anesthetized between 7:00 and 8:00 AM. They were premedicated with oxazepam (15 mg per os) 1 h before induction of anesthesia. Five hundred milliliters of 6% hydroxyethyl starch was infused during the first 20 min, followed by 1000 mL of lactated Ringer's solution. Anesthesia was induced with propofol 2.5 mg/kg and muscle relaxation was induced and maintained by vecuronium bromide 0.1 mg/kg. Mechanical ventilation was set at 6 breaths/min and end-tidal CO2 concentration was kept between 34 and 38 mm Hg, controlled with the aid of a capnometer (Cardiocap Registered Trademark; Datex, Espoo, Finland). Anesthesia was maintained with isoflurane in 35% oxygen/air. End-tidal isoflurane (ETisof) was monitored continuously at the distal end of the endotracheal tube using a Normac Registered Trademark monitor (Datex). Anesthesia was deepended until the appearance of the burst suppression pattern, then held at that level for 10 min to achieve steady-state burst suppression. This was maintained for 15 min and the patient was then given periods of visual, auditory, and median nerve stimulation, each lasting 5 min. Anesthesia was then increased by 0.3 ETisof and 15 min later the stimulation sequence was repeated. Surgery was delayed until after the study. Subsequently anesthesia was continued according to clinical needs. Systolic and diastolic arterial pressures and heart rate were measured every 5 min.
The EEG was recorded until the start of surgery from electrode pairs FZ-A1, CZ-A1, OZ-A1 (international 10-20 electrode system; passband 0.03-100 Hz) using silver-plated contact electrodes. The ground electrode was located on the forehead. The electrode impedance was <5 k Omega. A Neuropack Four mini evoked-potential measuring system (Nihon Kohden Corporation, Tokyo, Japan) was used for EEG amplification, monitoring, and stimulation. The data were digitized on-line at 200 samples/s to a PC-compatible microcomputer and stored on optical disks for off-line analysis. Electrocardiogram and the CO2 concentration of expired gas were also recorded.
Visual, auditory, and somatosensory stimuli consisted of 3-s episodes of flashing light (4 ms) via redlight-emitting diode goggles, eyelids closed; 3-s trains of clicks (80 dB, 0.1 ms) given with earphones; and 3-s episodes of electric pulses (20 mA, 0.2 ms) given to the left median nerve at the wrist, presented at a frequency of 20 Hz. The 3-s episodes of stimulation were given at irregular intervals ranging from 5 to 20 s (the time from the offset of the preceding stimulus to the onset of the next stimulus). The patient was not touched or stimulated in any other way during the experiment.
The beginning and ending of bursts, abrupt periods of high amplitude, mixed frequency EEG activity, separated by periods of relative quiescence (suppression) were marked manually. The bursts were classified into three groups: stimulus-onset, stimulus-offset, and spontaneous bursts. A burst was regarded as a stimulus-onset burst if it began within 1 s after the stimulus onset, i.e., the first flash, click, or electric pulse in a 3-s episode of stimuli. The bursts beginning within 1 s after the stimulus offset, i.e., the last flash, click, or electric pulse in a 3-s episode of stimuli, were classified as offset bursts and the bursts that appeared later than 1 s after the stimulus offset as spontaneous. The latency of bursts was measured visually. The latency of onset burst is the time interval from the first stimulus in the episode of stimuli to the onset of a burst in the EEG. Correspondingly latency of offset burst is the time from the the last stimulus in the episode of stimuli to the onset of a burst.
The effect of stimulation on the burst suppression pattern was tested by comparing the EEG during stimulation with a 100-s control period preceding stimulation. The durations of the bursts and suppressions with respect to the control period were analyzed using analysis of covariance for repeated measures, where we used the control periods as covariates. Total burst duration was analyzed in the same manner, with the control periods used as covariates. The effect of increasing anesthesia on stimulus evoked bursts was analyzed using repeated-measures analysis of variance (ANOVA). The latencies of bursts evoked by different stimulus modalities were compared using ANOVA where the equality of variances was tested using Levene's test . The Brown-Forsythe approximation was then applied when necessary. The equality of variances is assumed in the ANOVA. If this assumption does not hold, we adjust the degrees of freedom using the Brown-Forsythe approximation . Post-hoc comparisons used the Bonferroni test.
In order to test statistically whether there is a predominance of offset responses at a deeper level of anesthesia we compared the number of onset bursts and offset bursts within the same subject at the two different anesthetic levels. We compared the number of subjects who had a predominance of onset to offset bursts at the lighter level of anesthesia to the number of subjects who had a predominance of onset to offset bursts at the deeper level of anesthesia using a two-tailed Fisher's exact test. We also calculated the odds ratio for the occurrence of offset responses at the deeper level of anesthesia compared to the lighter level of anesthesia. The odds ratio equals one when there is no association between factors .
In 10 patients we obtained two distinct stages of steady burst suppression anesthesia at 1.5 +/- 0.1 ETisof and 1.8 +/- 0.1 ETisof. In the other five subjects, the burst suppression pattern was very unstable and occurred within a very narrow range; small decreases in isoflurane concentration or stimulation could switch the EEG from burst suppression to continuous, while small increases in isoflurane concentration could switch the EEG to total suppression. In one subject there was continuous suppression at a deeper level of anesthesia with no responses to stimulation. In another subject, the EEG switched directly from continuous EEG to total nonreactive suppression at the same isoflurane concentration (ETisof 2.5 vol%). In three patients the EEG switched from a burst suppression pattern to continuous EEG during somatosensory stimulation. These five subjects, who did not show burst suppression pattern at both anesthetic levels, were excluded from the statistical analysis.
All stimulus modalities evoked bursts Figure 1. Auditory evoked bursts were related only to the offset of the 3-s episode of clicks Figure 1A. Visual and somatosensory evoked bursts were related to both onset and offset of the 3-s episode of flashes and electric pulses Figure 1 B, C. During lighter anesthesia almost all stimuli given during EEG suppression evoked either onset or offset bursts: 100% of visual stimuli, 98% +/- 4% of electric stimuli, and 94% +/- 9% of auditory stimuli. During deeper burst suppression anesthesia, 78% +/- 37% of visual stimuli, 72% +/- 31% of electric stimuli, and 64% +/- 40% of auditory stimuli evoked bursts. There was, however, no significant difference in the efficacy of different stimulus modalities to evoke bursts. There was a significant decrease in the percentage of stimuli that evoked bursts during deeper anesthesia compared to lighter anesthesia Table 1 and Table 2.
The deeper level of burst suppression anesthesia was associated with a predominance of stimulus-offset evoked bursts over stimulus-onset evoked bursts. There was significant predominance of visually evoked offset bursts over onset bursts in the deeper level of anesthesia (P = 0.023) Figure 1B and a similar predominance of offset bursts with electric stimulation (P = 0.057) Figure 1C. The odds ratio for offset bursts during deeper anesthesia was 16.0 for visual stimuli and 13.5 for electric stimuli.
Lighter (1.5 +/- 0.1 ETisof) and deeper (1.8 +/- 0.1 ETisof) levels of burst suppression anesthesia differed significantly in duration of suppressions, duration of bursts, and in total burst duration Table 1 and Table 2. The stimulation decreased the mean duration of suppression from 18 +/- 15 s to 7 +/- 2 s during lighter anesthesia and from 50 +/- 40 s to 19 +/- 16 during deeper anesthesia. The stimulation increased the total percentage of burst duration from 43% +/- 22% to 66% +/- 13% at the lighter level and from 22% +/- 22% to 36% +/- 20% at the deeper level of anesthesia. The mean duration of bursts was 9 +/- 4 s at the lighter level of anesthesia and 9 +/- 10 s at the deeper level of anesthesia before stimulation. The stimulation increased the mean duration of bursts to 15 +/- 32 s at 1.5 ETisof and decreased it to 8 +/- 3 s at 1.8 ETisof.
Visually evoked bursts had shorter latency (330 +/- 85 ms) than auditory (480 +/- 100 ms) or electrically (490 +/- 110 ms) evoked bursts (ANOVA, P < 0.001, pairwise significance level 0.01). There was a significant difference in the latencies of onset and offset bursts (P < 0.05).
Within the same subject different stimulus modalities evoked bursts with different waveforms Figure 2. Identical stimuli in the same subject at identical anesthetic levels evoked bursts with similar latencies and waveforms, a DC shift exceeding 100 mu V with superimposed mixed frequency activity Figure 2. Painful stimuli were associated with increases in heart rate Figure 3.
This study shows that, at moderate and deep levels of isoflurane anesthesia, external stimuli evoke high amplitude responses that have a constant waveform and show only slight variability in latency, that is, the waveforms and the latencies of the evoked bursts differ very little when the same stimulus is applied many times within the same subject at the same anesthetic level. The latency of these evoked bursts is close to that of long latency evoked potentials such as P300 . The stimulus evoked bursts can in fact be called evoked potentials. Due to the high amplitude of the responses, averaging is not needed to reveal the evoked potentials as is the case with classical evoked potentials.
Burst suppression during anesthesia was first described in cats by Derbyshire et al. . They stimulated the sciatic nerve with electric pulses, which evoked bursts with a latency of 40-60 ms. Hand clapping has been used to test whether bursts or even epileptic discharges can be provoked during anesthesia . The first systematic study on evoked bursts was by Yli-Hankala et al.  who used vibration stimulation to provoke bursts. Recently we have shown that visual stimulation can evoke bursts during isoflurane-induced burst suppression anesthesia  as well as during barbiturate anesthesia . Little is known about the neuronal mechanisms producing burst suppression. The first study on the cellular events during burst suppression was done by Steriade et al.  who studied burst suppression with intracellular and multi-site simultaneous extracellular recordings from the cat neocortical and thalamic neurons. They showed that there is a close correspondence between neocortical and thalamic activities at the single cell level and mass electrical events recorded from the cortex, thalamus, and upper brainstem. They also showed that, during cortical suppression, there was still activity in the thalamus, and that some neurons in the dorsal lateral geniculate could be driven by light stimulation during EEG suppression. However, in their experiments they used a combination of several drugs. As we have shown in our recent paper , the waveforms of bursts differ depending on the drug, which may reflect different cellular events under different anesthetics.
All stimulus modalities readily evoked bursts. Visual and painful stimulation evoked bursts both at onset and offset of the 3-s stimulus episodes. Interestingly, auditory evoked bursts were associated only with the offset of the stimulus episode, i.e., the bursts were probably evoked by the missing click instead of the actual clicks in the episode of auditory stimuli. In this study there was no statistically significant difference between different stimuli in their ability to evoke bursts or in the duration of suppressions or bursts. There may be differences in reactivity to different stimuli that did not reach statistical significance due to the small number of subjects. However, noxious somatosensory stimulation seemed to be most efficient in activating the EEG. In three patients, noxious stimulation switched burst suppression to continuous EEG. Similar phenomena have been described by Steriade et al.  who reported that thalamic volleys delivered during cortical electrical silence could restore the EEG activity.
Deepening anesthesia at burst suppression level was associated with increases in the duration of suppressions and decreases in the total duration of bursts. There was a significant decrease in reactivity to external stimuli during deeper anesthesia. However, during deeper burst suppression anesthesia, where the mean duration of suppressions before stimulation was 50 s, 71% of all stimuli still evoked bursts. In deeper burst suppression anesthesia, there was a predominance of offset evoked bursts over onset evoked bursts. Deepening anesthesia is probably associated with both increasing inhibition and increasing neuronal excitability, and these two stages oscillate in a nonlinear, on-off manner. External stimuli can move the system from one stage to another. More frequent offset bursts could reflect the stronger inhibition during deeper anesthesia. We hypothesize that during repetitive stimuli there is active inhibition and the release from inhibition at stimulus offset is reflected in increased cortical excitability and off-responses. However, more studies at the EEG and cellular levels are needed before we can determine why the deepening anesthesia increases the probability that, instead of the actual physical stimuli, the burst is evoked by disappearance of the stimuli.
Burst suppression stimulation studies should prove useful in determining the mechanisms and sites of action of new anesthetics. Whether stimulus-induced bursts could be used to test the integrity of visual, auditory, or somatosensory pathways during surgery remains to be studied. The immediate and clear response to single stimulation without need for averaging might prove to have advantages in comparison to classical evoked potentials in neuromonitoring during surgery.
1. Homi J, Konchigeri HN, Eckenhoff JE. A new anesthetic agent Forane: preliminary observations in man. Anesth Analg 1972;51:439-77.
2. Joas TA, Stevens WC, Eger EI II. Electroencephalographic seizure activity in dogs during anaesthesia. Br J Anaesth 1971;43:739-45.
3. Derbyshire AJ, Rempel B, Forbes A, Lambert EF. The effect of anesthetics on action potentials in the cerebral cortex of the cat. Am J Physiol 1936;116:577-96.
4. Yli Hankala A, Jantti V, Pyykko I, Lindgren L. Vibration stimulus induced EEG bursts in isoflurane anaesthesia. Electroencephalogr Clin Neurophysiol 1993;87:215-20.
5. Hartikainen K, Rorarius M, Makela K, et al. Visually evoked bursts during isoflurane anaesthesia. Br J Anaesth 1995;74:681-5.
6. Levene H. Robust tests for the equality of variance. In: Olkin I, ed. Contributions to probability and statistics. Palo Alto: Stanford University Press, 1960.
7. Brown MB, Forsythe AB. Robust tests for the equality of variances. J Am Stat Assoc 1974;69:364-7.
8. Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York: Wiley, 1981.
9. Altenmuller EO. Psychophysiology and EEG. In: Niedermeyer E, da Silva FL, eds. Electroencephalography: basic principles, clinical applications, and related fields. Baltimore: Williams & Wilkins, 1993;597-613.
10. Rampil IJ, Lockhart SH, Eger EI, et al. The electroencephalographic effects of desflurane in humans. Anesthesiology 1991;74:434-9.
11. Jantti V, Eriksson K, Hartikainen K, Baer GA. Epileptic EEG discharges during burst suppression. Neuropediatrics 1994;25:271-3.
12. Steriade M, Amiza F, Contreras D. Cortical and thalamic cellular correlates of electroencephalographic burst-suppression. Electroencephalogr Clin Neurophysiol 1994;90:1-16.
© 1995 International Anesthesia Research Society
13. Hartikainen K, Rorarius M, Makela K, Jantti V. Propofol and isoflurane induced EEG burst suppression patterns in rabbits. Acta Anaesthesiol Scand 1995;39:814-8.