Attenuation of the 40-Hertz Auditory Steady State Response by Propofol Involves the Cortical and Subcortical Generators

After approval of the McGill University Health Center ethics board (Montreal, Quebec, Canada), we recruited 11 American Society of Anesthesiologists physical status I, paid volunteers (5 men and 6 women) aged 20–38 yr (mean, 29 yr). Subjects had no history of neurologic or hearing disorders. They gave written consent, underwent a comprehensive medical evaluation, and were screened for normal hearing with pure tone audiometry.

The auditory evoked responses were recorded during three periods: baseline (subject awake, lying quietly with eyes closed before any medication); propofol anesthesia (at the lowest concentration of propofol producing unconsciousness); and recovery (20 min after return of consciousness after discontinuation of propofol). Three blocks of 8 min were recorded for each period. Unconsciousness was defined as failure to respond to verbal commands.

Testing took place in the morning after an overnight fast. After placement of the electroencephalographic electrodes, we gave sodium citrate (30 ml orally) and inserted an 18-gauge cannula in a vein of the left forearm. Oxygen was given by facemask. Monitoring included three-lead electrocardiography, pulse oximetry, and inspired and expired concentrations of oxygen and carbon dioxide of gas sampled from nasal prongs. Propofol was administered with a pump (Harvard 22; Harvard Apparatus, Holliston, MA) driven by a personal computer running STANPUMP, version of May 11, 1996,#using the kinetic variables of Tackley et al.  36The effect site concentration was initially set to 1.5 μg/ml and was increased every 10 min by 0.5-μg/ml steps until the subject became quietly immobile and unresponsive to verbal commands. The target concentration required for unconsciousness was 2.0–4.5 μg/ml (mean, 3.0 μg/ml).

The recording protocol is based on that used by Makela and Hari.37The stimuli consisted of 1,000-Hz tone bursts (10 ms total duration with 2 ms rise/fall time, 100% amplitude modulation, 75 dB peak equivalent sound pressure level) delivered to right ear at the rate of 40/s in trains of 25 stimuli (625 ms). The interval between train onsets was 2.5 s. Stimuli were created using an ATHLON-XP–based (AMD, Sunnyvale, CA; 1.33 GHz) personal computer (Biostar M7MIA 2.0 RAID; BMA Industrial Inc., City of Industry, CA) using Matlab version 6.0 (MathWorks, Natick, MA) under Microsoft Windows 98 (Microsoft, Redmond, WA). They were delivered via  the integrated sound card connected to Sennheiser HDA 200 earphones (Wedemark, Germany). The Matlab routine also produced a brief (3-ms) synchronization marker coinciding with the onset of each tone burst using the digital-to-analog function of an AT-MIO-64E-3 board (National Instruments Corp., Austin, TX).

The electroencephalogram was recorded with gold disk cup electrodes from 32 scalp sites of the extended 10-20 system38(FP1/2, F9/10, F7/8, F3/4, FC5/6, FC1/2, T9/10, T7/8, C3/4, CP5/6, CP1/2, P9/10, P7/8, P3/4, O1/2, Fz, and Cz) referenced to A2. Ground was Fpz. Impedances were below 5 kΩ. The electroencephalogram was amplified (Isolated Bioelectric Amplifier; SA Instrumentation Corp., San Diego, CA; 0.01- to 300-Hz bandpass), digitized at 1,024 Hz and stored on disk with Harmonie software (version 5.1; Stellate Systems, Montreal, Quebec, Canada) using the above computer and the AT-MIO-64E-3 board for analog-to-digital conversion. An additional channel was used to record stimulus onset synchronization markers.

The electroencephalogram was averaged off-line using Matlab (version 7.3; MathWorks). The epoch for averaging was from −100 to 1,000 ms relative to onset of the first tone burst of each train. Epochs contaminated by artifacts (10% of points outside ±70 μV) were rejected automatically. The average waveform for each subject and period was based on approximately 500 epochs (24 min of recording).

The amplitude at Cz of the N1 was calculated as the mean amplitude over the 75- to 125-ms interval; that of the SP was calculated as the mean amplitude over the 275- to 675-ms interval. To search for residual N1 and SP activity during anesthesia, the mean amplitude at Cz over successive intervals of 50 ms centered at 50, 100, 150,… 900 ms was measured. This was done to determine subsequently whether the mean amplitude differed significantly from zero. This technique of using sequential means to look at a slowly changing waveform has long been used in event-related potential research.39 

The amplitude of the 40-Hz ASSR was obtained with a 512-point fast Fourier transform over a time window from 125 to 625 ms after the start of the stimulation on the Cz waveform of each subject for each period. The measurement was started at 125 ms to remove the effects of the transient γ band response and to allow build-up of the 40-Hz ASSR.40 

Source analysis will not be reported for N1 (and SP) because these responses were completely abolished by propofol. Therefore, no source analysis could be obtained for the anesthesia data.

The montage was first converted to average reference. The grand-average baseline data were used to obtain a source template with the program BESA (version 3.0; MEGIS Software GmbH, Gräfelfing, Germany) using a four-shell head model. The source model was similar to that of Herdman et al.  15with two exceptions. First, only one brainstem dipole (D5, vertically oriented) was used because the laterally oriented brainstem dipole in Herdman et al.  15is small and mostly reflects ear differences. Second, the radial and tangential cortical dipoles were allowed to separate in location (see Scherg et al.  32. The cortical dipoles were constrained to be symmetrical in location and mirror image in orientation. This template of five dipoles was then used to obtain the source waveforms of the 40-Hz ASSR of each subject and recording period. The period selected for this analysis was from 125 to 625 ms after train onset. The data were collapsed to two cycles (50 ms) and filtered 30–50 Hz before source analysis to increase signal-to-noise ratio. This procedure is called “within-epoch” averaging and is equivalent to conventional averaging using the first stimulus of each train and every other stimulus afterward. It was used by Herdman et al.  15and by others.41Dipole strength was assessed as the root-mean square of source waveform across the two cycles. The onset phase (i.e. , the section of the waveform that occurs at time 0) of the dipole was obtained by fast Fourier transform of the source waveform across the two cycles. This measurement can be converted to phase lag by subtracting from 360°. Phase lag is related to latency, but the measurement is ambiguous because of the unknown number of cycles preceding the measured cycle.

Unless indicated otherwise, all results are reported as mean ± SD and were analyzed with analyses of variance for repeated measures with Geisser-Greenhouse adjustment and Tukey honest significant difference for post hoc  tests.42The criterion for significance was 0.05. Procedures were performed with Statistica (version 4.1 for the Macintosh; Statsoft, Tulsa, OK). Power calculations were calculated with the function sampsizepwr from the Matlab statistical toolbox (version 5.3, MathWorks).

Additional procedures were performed to address specific needs as described below.

To establish the presence of response, one-sample t  tests were conducted to determine whether the mean Cz amplitude over successive intervals of 50 ms differed significantly from zero. The two-tailed criterion for significance was 0.0028 after Bonferroni correction (18 comparisons per recording period) for an overall error rate of 0.05 for each component.

To compare the effect of propofol on cortical versus  brainstem dipoles during anesthesia, dipole strength during anesthesia was divided by dipole strength during baseline to obtain a relative measure. The relative values for the four cortical dipoles were then combined in a weighted average to obtain a value for the cortical dipoles as a group. This was done for each subject separately. The weights reflected the relative contribution of each cortical dipole to total cortical dipole strength during baseline. This was necessary because Herdman et al.  15found 3.5-fold differences in strength among the cortical dipoles. Therefore, the global impact of the attenuation of a given cortical dipole by propofol depends on the contribution of the dipole to total cortical strength. A similar process was performed for recovery relative to baseline. Paired t  tests were used to compare relative strength of cortical dipoles as a group versus  the brainstem dipole during anesthesia and recovery. One-sample t  tests were used to determine whether the normalized dipole strength during anesthesia or recovery was different from baseline (i.e. , from one).

Circular statistics43were used to compute the mean and SD of the phase obtained by fast Fourier transform. Phase is reported in degrees (0–360°), counterclockwise from horizontal and relative to a cosine. Pairwise comparisons (between dipoles during baseline; between baseline and anesthesia) were obtained with a Hotelling T²test44implemented in Matlab (version 7.3). To limit the number of tests and the risk of spurious significance when amplitudes were small, tests of phase differences were obtained only for the three main dipoles (D1, D3, and D5).

Inspection of the raw electroencephalographic tracings revealed increased amplitude of background rhythms, mainly affecting the α and β bands. This pattern is similar to the stage 1 of light anesthesia reported by Reddy et al.  45 

Figure 1shows the original waveforms recorded at Cz from each subject as well as the grand-average waveforms across all the subjects. The 0- to 20-Hz filtered data reveal clear N1 and SP during baseline and recovery. During anesthesia, there is no clear response. The amplitude of N1 (mean ± SD) was −1.18 ± 0.36, 0.15 ± 1.20, and −0.49 ± 0.38 μV during baseline, anesthesia, and recovery, respectively (P < 0.01 for anesthesia vs.  baseline and recovery). The amplitude of the SP was −1.46 ± 0.63, −0.26 ± 0.98, and −1.46 ± 0.78 during baseline, anesthesia, and recovery, respectively (P < 0.01 for anesthesia vs.  baseline and recovery).

The mean amplitude of the intervals corresponding to the N1 (100 ms) and to the SP (intervals centered at 300–650 ms) differed significantly (one-sample t  tests; P < 0.0028) from zero during baseline and recovery. By contrast, the mean amplitude at Cz over successive 50 ms-intervals (table 1) did not differ significantly from zero during anesthesia, indicating that the N1 and SP were no longer recordable.

The 40-Hz ASSR is visible as the 30- to 50-Hz filtered traces in figure 1and in figure 2, which shows the grand-average waveforms (all subjects pooled) used for source analysis after collapsing the period from 125 to 625 ms after train onset to a two-cycle (50 ms). Inspection of figure 2reveals that propofol reduced the amplitude of the 40-Hz ASSR on all channels. Measures from individual subjects showed that the amplitude of the 40-Hz ASSR at Cz was (μV) 0.17 ± 0.07 during baseline, 0.08 ± 0.03 during propofol, and 0.23 ± 0.09 during recovery (P < 0.001 for propofol vs.  baseline or recovery; P < 0.01 for baseline vs.  recovery). Figure 2shows that propofol also changed the phase of response on most channels.

Residual variance (%) for the dipole fit on the grand-average waveforms was 4, 22, and 3 during baseline, anesthesia, and recovery, respectively. Residual variance (%) for the fit when applied to individual subjects was 12 ± 7, 30 ± 13, and 12 ± 8 during baseline, anesthesia, and recovery, respectively. The specific contribution of each dipole to the reduction of the variance during baseline was measured by removing from the model each source, one at a time. The contributions were 12, 5, 9, 6, and 8% for dipoles 1–5, respectively. (The sum of these values is less than the total variance explained [96%] because, when one source is removed, the increase in residual variance is absorbed in part by the remaining sources.)

Figure 3Ashows the dipole diagrams and figure 3Bshows the activity of the source waveforms for each period. Inspection of this figure reveals the following. During baseline, the strength of the brainstem dipole (D5) is largest, followed by that of the two tangential cortical dipoles (D1 and D3). The two radial dipoles (D2 and D4) have the lowest strength. During baseline, the phase of the source waveforms are such that the latency of the first positive peak is shortest for the brainstem dipole (D5, arrow), followed by that of the two tangential cortical dipoles (D1 and D3) and, last, by that of the two radial dipoles (D2 and D4). This pattern is consistent with ascending activation in the auditory pathways. The figure also reveals that propofol reduced the strength of all dipoles and changed the phase of the cortical dipoles (D1–D4). The source strength during recovery appears slightly larger than during baseline. Propofol did not alter the phase of the brainstem dipole (D5).

Measures from individual subjects were subjected to statistical analysis to validate the observations described above. There were significant differences in dipole magnitude (fig. 4) during baseline: D1 (tangential left cortical) was larger than D2 (radial left cortical) (P < 0.05) and D4 (radial right cortical) (P < 0.01); D3 (tangential right cortical) was larger than D2 (radial left cortical) (P < 0.001) and D4 (radial right cortical) (P < 0.01); D5 (brainstem) was larger than all other dipoles (P < 0.001). The magnitude of all dipoles was significantly less during anesthesia than during baseline and recovery except for D2 (left cortex radial) compared with baseline (table 2and fig. 4). The magnitude of all dipoles during recovery was higher than during baseline (fig. 4; see normalized measures below for statistical comparisons).

The relative strength of the cortical dipoles during anesthesia (baseline = 1) was 0.41 ± 0.26 for D1, 0.68 ± 0.58 for D2, 0.61 ± 0.36 for D3, and 0.51 ± 0.40 for D4. There were no significant differences between these values, which were all significantly less than baseline (P < 0.01; one-sample t  test). The pooled relative strength during anesthesia of the four cortical dipoles as a group was 0.54 ± 0.24, and it did not differ significantly from the relative strength of the brainstem dipole (D5), 0.53 ± 0.40 (paired t  test; t = 0.12, P = 0.91; fig. 5). These two values were significantly smaller than baseline (P < 0.01; one-sample t  test). Therefore, the cortical dipoles as a group and the brainstem dipole were attenuated to the same extent during anesthesia. The difference between the relative strength of the pooled cortical dipoles and of the brainstem dipole that could have been detected with a power of 0.8 at P < 0.05 (one-tailed) was 0.37 (i.e. , cortical attenuation exceeding brainstem attenuation by 0.37).

During recovery, the relative value for the pooled cortical dipoles and the brainstem dipole was 1.3 ± 0.3 and 1.2 ± 0.0.8, respectively (fig. 5). The cortical value was significantly higher than baseline (P < 0.01; one-sample t  test). The difference between the two values was not significant (paired t  test; t =−0.71, P = 0.50).

Phase comparisons were limited to the three dipoles with the largest magnitude: the two tangential cortical dipoles (D1 and D3) and the brainstem dipole (D5). During baseline, the onset phase of D5 (195 ± 18, in degrees) was significantly (P < 0.05) greater than that of the two tangential dipoles: D1 (131 ± 25) and D3 (81 ± 29), which were also significantly (P < 0.001) different from each other. In terms of phase lag, the D1 and D3 responses occurred later than the brainstem responses. There were significant phase differences between baseline and anesthesia for the two tangential dipoles (D1: 131 ± 25 during baseline vs.  325 ± 45 during anesthesia; D3: 81 ± 29 vs.  34 ± 29). Both of these can be interpreted as increased phase lag. The phase of the brainstem dipole did not change significantly (195 ± 18 vs.  218 ± 40).

The main finding of this study is that the cortical and subcortical generators of the 40-Hz ASSR are affected to the same extent during propofol anesthesia This observation clearly supports the reciprocal excitation hypothesis , whereby the 40-Hz ASSR involves recurrent excitation between cortical and brainstem sources24and appears at first glance at variance with the prevalent view that impairment of (nonnociceptive) sensory processing by propofol involves mostly cortical neurons, with only minimal, if any, alterations at the level of the brainstem. It must be noted, however, that these observations23supporting the cortical hypothesis  concern the somesthetic system.

It is likely that the attenuation of the cortical generator results from a direct effect of propofol given the prominent inhibitory effect of this drug on the cerebral cortex.23The reduced activity of the brainstem generator of the 40-Hz ASSR may result from a direct or indirect effect of propofol. Brainstem neurons, most notably in the inferior colliculus, which is a probable generator of the 40-Hz ASSR based on animal studies,17,19,21,22are clearly affected by general anesthetics.46,47There is, however, no data for propofol. Thus, there is partial support from literature for proposing a direct anesthetic effect at the brainstem level. An indirect effect on the brainstem is also possible. The reduction of cortical activity by propofol could indirectly attenuate the activity of brainstem source by decreasing the excitatory cortical input and thus attenuating the recurrent excitation between the brainstem and cortex.24Proper evaluation of the relative contributions of the direct and indirect actions of propofol will require animal experimentation.

An important issue to consider is the reliability of the BESA modeling applied to the current data. Reliability can be assessed by the comparing residual variance and the pattern of source activity with published work. The level of residual variance (a measure of the signal-to-noise ratio, i.e. , quality of the recordings) during baseline and recovery is almost identical to that reported by Herdman et al.  15in waking subjects, and this applies to both grand-average and individual data. As expected, the residual variance during anesthesia was higher than during baseline and recovery. The most likely explanation is a reduction of signal-to-noise ratio because propofol reduces the amplitude of the evoked response10and increases the amplitude of background electroencephalogram.45 

The pattern of 40-Hz ASSR source activity obtained during baseline (greater magnitude of the brainstem source compared with cortical sources; greater magnitude of the tangential cortical sources compared with the radial cortical sources) is similar to that described by Herdman et al.  15The similarity between our findings during wakefulness and those of Herdman et al.  15despite different recording environments provides a strong assurance of reliability.

The higher strength of the midbrain source is unusual when the organization of the assumed generator areas are considered. Evoked potentials probably originate from postsynaptic potentials in parallel oriented neural structures, which are sparse in the midbrain–brainstem and thalamus but abundant in the cortex. Herdman et al.  48also reported higher brainstem source strength. Responses generated earlier in the pathway may be more synchronous, and this may compensate for the geometric advantage of the cortical sources.

A possible explanation for the strong midbrain source is that this source may have captured some activity from the cortical sources. The extent of misallocation can be assessed by how much the removal of the brainstem source increases the source strength of the cortical activations. We accordingly obtained a revised source analysis after excluding the brainstem source (D5) from the model which we applied to the baseline grand-averaged waveforms. Source strength was unchanged for D1 and was increased by 19, 9, and 5% for D2, D3, and D4, respectively. The mean increase for the four dipoles was 8%. This modest increase indicates that the capture of cortical activity by the brainstem dipole (D5) had little impact on its strength. Furthermore, the cortical and brainstem sources were relatively independent in their topographies (data not shown). This provides additional evidence against possible misallocation of cortical activity to the brainstem source (D5).

The significant differences in the phase of the dipole source waveforms provides additional support for the validity of the model. The phase difference between the brainstem dipole and the two main (tangential) cortical dipoles during baseline likely reflects the earlier activation of the brainstem source. Likewise, the phase difference between the two tangential cortical dipoles likely reflects the earlier activation of the contralateral cortex. Finally, the change in phase between baseline and anesthesia noted with the two main tangential cortical dipoles but not with the brainstem dipole is analogous to the observation that general anesthetics affect the latency of cortical evoked potentials much more than that of the brainstem responses.10 

Another concern is the validity of electrophysiologic source modeling in general. As noted introduction, there is no known method to unequivocally identify the cerebral generators of sensory evoked potentials recorded at the scalp. Despite this limitation, it is reasonable to accept the solution obtained by source modeling if one can demonstrate that the source model (1) explains the observed data as well as can be expected from the level of residual noise, (2) is reproducible (e.g. , baseline vs.  recovery), and (3) makes sense physiologically (e.g. , see the previous paragraph on phase changes).12 

The right tangential cortical dipole (D3) seems to have been less affected by propofol than the other cortical dipoles. Although there was no significant difference between cortical dipoles in the extent of attenuation, the seemingly lesser impact of propofol on D3 remains puzzling. We have no clear explanation, but we think that this finding may be related to the asymmetry of stimulus input (right ear).

Previous studies of the effect of general anesthetics on the generators of the 40-Hz ASSR used pentobarbital or isoflurane. Sem-Jacobsen et al.  49recorded the response from the temporal cortex of patients with implanted electrodes to a train of clicks presented at rates near 40/s (based on close examination of fig. 6 after magnification). The authors reported that the cortical response decreased as the depth of thiopental anesthesia increased. Under extremely deep levels of anesthesia, the response could not be elicited. These studies were published in 1956, well before recognition of the 40-Hz ASSR as a separate entity.50Makela et al.  20reported that pentobarbital anesthesia (35–40 mg/kg intraperitoneal) in cats causes a near complete suppression of the cortical 40-Hz ASSR recorded from the auditory cortex and approximately a 35% attenuation of the 40-Hz ASSR recorded from the medial geniculate nucleus. Szalda et al.  22showed that pentobarbital anesthesia (40 mg/kg intramuscular) predominantly affects the cortical generator of the ASSR**in the chinchilla, with minimal effect on the inferior colliculus. Kuwada et al.  24suggested, based on the effect of pentobarbital (12–25 mg intravenous) on the amplitude modulation rate transfer, that the drug reduces the ASSR in both the auditory cortex and inferior colliculus but that the cortical response was slightly more attenuated. A higher effective concentration of pentobarbital in the study of Kuwada et al.  24may explain why they observed an effect on the inferior colliculus whereas Szalda et al.  22did not. Santarelli et al. ,51using subdural electrodes overlying the auditory cortex of the rat, showed that the amplitude of the 40-Hz ASSR was decreased by half by 0.4–1.1% isoflurane.

This is the first study to examine the effect of a general anesthetic on the SP, which was no longer recordable during anesthesia. This suggests that the SP is a sensitive marker of anesthetic effect. The SP has features that could be advantageous for clinical monitoring the level of consciousness in clinical anesthesia. Its plateau shape facilitates its extraction from background noise, and its duration depends on the length of the stimulus train, which could be varied to facilitate its recognition. Before further consideration, however, the effect of sedative concentrations of propofol on the SP must be assessed.

The current study shows that the N1 is abolished during propofol anesthesia, in accord with the observations of Simpson et al.  31and in contrast to those of Ypparila et al.  29and van Hooff et al. ,30who reported the persistence of the N1. This discrepancy may be explained by the influence of surgery and the mode of administration of propofol. The study of van Hooff et al.  30was conducted during cardiac surgery, and that of Ypparila et al.  29was conducted immediately after cardiac surgery. By contrast, Simpson et al.  31collected their data before surgery, and our subjects did not undergo surgery. In the two studies involving surgery, surgical pain may have increased arousal, and the patients may have had difficulty signaling their consciousness because of the other drugs used (opiates, muscle relaxants). Furthermore, propofol was not infused in a target-controlled mode in these two studies, in contrast to the study of Simpson et al.  31and the current study.

Why are the N1 and SP more affected than the cortical generator of the 40-Hz ASSR? The most parsimonious explanation is that the cortical ASSR generator mainly arises from a cortical response (the middle latency [15- to 35-ms] auditory response), which precedes the N1 and SP response and is more directly connected to the brainstem input than later responses. The N1 and SP involve more synapses52and depend more on corticocortical activation, which is highly susceptible to propofol-induced interference.23 

We conclude that (1) the decrease of the amplitude of the 40-Hz ASSR during propofol anesthesia results from an attenuation of similar magnitude of the activity of the cortical and brainstem generators, and (2) the N1 and SP are so profoundly attenuated during propofol anesthesia that they are no longer recordable from the scalp.

The authors thank Chantale Porlier (Electrophysiology Technician, Department of Anesthesia, Royal Victoria Hospital, Montreal, Quebec, Canada) and Louise Ullyatt, R.N. (Department of Anesthesia, Royal Victoria Hospital), for their help; Jean Gotman, Ph.D. (President and CEO, Stellate Systems, Montreal, Quebec, Canada), for allowing the authors to use the Harmonie software for electroencephalogram recording; the staff of the Recovery Room of the Royal Victoria Hospital for their cooperation; Janet Macdonald (Secretary, Department of Anesthesia, Montreal Neurological Hospital, Montreal, Quebec, Canada) for editorial assistance; and the anonymous reviewers for many useful suggestions.