Abstract
Background. Gamma (30–80 Hz) and high-gamma (80–200 Hz) thalamocortical EEG rhythms are involved in conscious processes and are attenuated by isoflurane and propofol. To explore the hypothesis that this attenuation is a correlate of anaesthetic action, we characterized the effect dexmedetomidine, a selective adrenergic α-2 agonist with lesser hypnotic potency, on these rhythms.
Methods. We recorded local field potentials from barrel cortex and ventroposteromedial thalamic nucleus in ten previously instrumented rats to measure spectral power (30–50 Hz, 51–75 Hz, 76–125 Hz, 126–200 Hz bands) during baseline, at four dexmedetomidine plasma concentrations obtained by i.v. target-controlled infusion (1.86, 3.75, 5.63 and 7.50 ng ml−1), and during recovery. Thalamocortical coherence over 0.3–200 Hz was also measured.
Results. Loss of righting reflex (LORR) occurred with 5.63 ng ml−1. Dexmedetomidine produced a linear concentration-dependent attenuation of cortical (P<0.04) and thalamic (P ≤ 0.0051) log power in all bands. Slopes for cortex and thalamus were similar. The slope for dexmedetomidine on thalamic power in the 76–200 Hz range was less than half that of the other agents (P<0.003). LORR was associated with an increase in delta band (0.3–4.0 Hz) thalamocortical coherence (P<0.001). Increased low-frequency coherence also occurred with propofol and isoflurane.
Conclusions. Dexmedetomidine attenuates high-frequency thalamocortical rhythms, but to a lesser degree than isoflurane and propofol. The main differences between dexmedetomidine and the other anaesthetics involved thalamic rhythms, further substantiating the link between impaired thalamic function and anaesthesia. Increased delta coherence likely reflects cyclic hyperpolarization of thalamocortical networks and may be a marker for loss of consciousness.
Editor’s key points
Anaesthetic agents are thought to induce unconsciousness by attenuating gamma and high gamma thalamocortical rhythms.
The authors studied the effect of dexmedetomidine on cortical and thalamic local field potentials.
Loss of righting reflex was associated with attenuation of high frequency rhythms, but increased delta coherence.
We have recently shown with recordings from the somatosensory (barrel) cortex and ventroposteromedial thalamus of the rat that both propofol and isoflurane cause a robust concentration-dependent attenuation of high-frequency (30–200 Hz) thalamocortical rhythms.12,13 Furthermore, the concentration-effect slopes for the thalamus were markedly steeper than for the cortex. These observations support the view that attenuation of high-frequency thalamocortical rhythms is a functional correlate of anaesthetic action. If similar effects could be demonstrated with other hypnotic drugs, the association between anaesthesia and impairment of high-frequency thalamocortical rhythms would be strengthened.
Our aim was to characterize the effects of dexmedetomidine on high frequency (30–200 Hz) thalamocortical rhythms. Dexmedetomidine is an adrenergic α-2 receptor agonist that causes sedation and hypnosis.14 We chose dexmedetomidine for two reasons. First, dexmedetomidine produces its sedative – hypnotic effect through different receptors than propofol and isoflurane, which enhance the action of the GABAA receptors.15 This allows the assessment of whether the diminution of high-frequency thalamocortical rhythms reflects anaesthetic action independently of the primary molecular target of the drug. Second, the sedative-hypnotic effect of dexmedetomidine is less intense than that of propofol or isoflurane, as revealed by the preserved ability to be awakened by verbal commands during deep sedation.14 If anaesthetic action depends on impairment of high-frequency rhythms, then the slope of the concentration-dependent attenuation of high-frequency thalamocortical rhythms by dexmedetomidine should be less than that of propofol or isoflurane.
The 30–200 Hz range was divided in four frequency bands as in our recent studies: 30–50, 51–75, 76–125, and 126–200 Hz.12,13 For each band, we characterized the concentration-effect relationship of dexmedetomidine on the spectral power of the local field potentials (LFPs)16 recorded from the somatosensory (barrel) cortex and ventroposteromedial thalamic nucleus (VPM) of the rats with previously implanted electrodes and we compared the steepness of the cortical and thalamic regression slopes. We also compared the potency of dexmedetomidine for attenuation high-frequency power with that of propofol and isoflurane as reported in our recently published work.12,13
The effects of general anaesthetics on EEG coherence is of current interest. It has been proposed that decreased EEG coherence observed during anaesthesia may reflect impaired communication between distinct areas of the cerebral cortex (see reference17 for recent animal evidence and background). We felt this topic of sufficient importance to justify an examination of the effects of dexmedetomidine on thalamocortical coherence. In addition to the four high-frequency bands listed above, we have also included in the analysis of coherence the standard EEG bands (delta: 0.3–4.0 Hz; theta 4.1–8.0 Hz, alpha: 8.1–12.0; Hz; sigma: 12.1–15.0; Hz; beta: 15.1–29.5 Hz).
Methods
The procedures are the same as those described in detail previously.12 We will present a summary of the essential points. The Animal Care Committee of the Montreal Neurological Institute approved the study and we adhered to the guidelines of the Canadian Council on Animal Care. We tested ten male Sprague Dawley rats (300–320 g). Animals were housed individually with normal light-cycle.
Surgery
Bipolar electrodes (twisted, teflon-coated stainless-steel wires; 125 μm diameter; vertical separation between tips: 0.5 mm; AM Systems, WA) were implanted sterotaxically under isoflurane anaesthesia in the ventroposteromedial thalamic nucleus (VPM; Paxinos atlas18 coordinates: A/P: -3.5, L/M: 2.7, V/D: -6.6, relative to bregma) and sensory (barrel) cortex (A/P: -2.3, L/M: 5.0, V/D: -3.6). The cortical electrode was placed to optimize the amplitude of the field excitatory post-synaptic potentials evoked by stimulation of the VPM stimulation. Thicker wires (280 μm diameter, AM Systems, WA) were anchored with screws in the contralateral parietal bone (reference) and the ipsilateral frontal bone (ground). The electrodes were inserted in a connector that was secured to the skull with dental cement. Animals received ketoprofen and buprenorphine for analgesia. One week was given for recovery.
Design
There was a single testing session with six conditions (baseline – target plasma dexmedetomidine concentrations plateaus of 1.86, 3.75, 5.63, 7.50 ng ml−1 and recovery after return of spontaneous ambulation). The concentrations were chosen on the basis of published reports19 and pilot tests to include both sub-hypnotic (1.86 and 3.75 ng ml−1) and hypnotic (5.63 and 7.50 ng ml−1) effects. LFPs were recorded during each condition, along with assessment of righting.
Dexmedetomidine (Dexdomitor, Pfizer Canada Animal Health) was administered in the right jugular vein (via a catheter previously inserted by the supplier, Charles River Laboratories, Senneville, Quebec) with a Harvard-22 syringe pump controlled by the Stanpump software (Steven L. Shafer, Standford University, CA) using the Bol and colleagues19 pharmacokinetic parameters. Fifteen min after reaching the desired plasma concentration, LFPs were recorded, spontaneous behaviour observed and righting reflex assessed. Unconsciousness was defined as absence of any righting attempts.
Electrophysiology
LFPs from the cortical and thalamic contacts were recorded with a common reference on the contralateral parietal bone (referential montage), amplified (0.1 Hz to 475 Hz band pass), digitized at 3030 Hz, and stored for offline analysis. For each period, two mins of high quality data devoid of artifacts was obtained. The recordings were subsequently reformatted offline to obtain a bipolar recording for each recording site (barrel cortex and VPM) consisting of the difference between the recordings of the two contacts (spaced 0.5 mm from each other) of each electrode. All analyses were conducted with the bipolar recordings because they are less subject to contamination from volume-conducted myogenic artifacts.
Spectral analysis was computed with two s long non-overlapping segments (Welch’s method20; Hamming window; psd.m function, Matlab Signal Processing toolbox, version 6, MathWorks Inc. Sherborn, MA). Power at 59–61 Hz, 119–121 Hz, and 179–181 Hz was excluded from analysis to avoid interference from external electrical sources. Normalization of the spectra is required to avoid an undue influence of animals with high baseline power. The spectra were normalized to ensure that the average power during baseline for each animal was equal to unity and the spectra for the other periods were scaled with the same parameter.
We used the Matlab function mscohere.m to compute the magnitude squared coherence between the cortical and thalamic recordings (two s segments; 50% overlap; Hamming window). Coherence values range from zero (completely independent signals) to unity (complete dependency).
Histology
Brains were obtained from eight animals. Accurate electrode placements were confirmed in six animals based on termination of the electrode tracks. In the other animals, tissue damage during retrieval or processing prevented identification of the electrode tracks or surrounding structures.
Statistical analysis
1- yijk is the spectral power of rat i (i = 1,2,…,10) at concentration j for site k
2- αik is the difference between the random intercept of rat i and β0k
3- β0k is the regression constant (intercept) for site k
4- β1k is the regression coefficient for the linear term (i.e. slope) for site k.
5- εijk is the error term, accounting for unexplained errors (e.g. measurement error).
To compare the regression coefficients for the linear term (β1) obtained for dexmedetomidine with those recently published for propofol12 and isoflurane,13 we assumed that a target concentration of 9 µg ml−1 propofol and 1.1% isoflurane were equivalent to 5.63 ng ml−1 of dexmedetomidine, because they are the lowest concentrations that abolished righting in all animals. The mean and standard error of the linear terms were multiplied by nine for propofol, by 1.1 for isoflurane and by 5.63 for dexmedetomidine. This is equivalent to computing the regression using 0, 0.33, 0.67, 1.00 and 1.33 for concentrations for the two i.v. agents and 0, 0.68, 1.00 and 1.36 for isoflurane. The linear terms were compared with Student’s t-tests for independent samples. All tests were adjusted for multiple comparisons (four tests for each site) with Hommel’s approach (SAS PROC MULTTEST).26
Coherence values were analysed with a repeated measures ANOVA combined with Tukey’s Honest Significant Difference Test for pair-wise post-hoc comparison (SAS PROC GLM). Tukey's test includes an adjustment for multiple comparisons.27
Results
Behavioural effects of dexmedetomidine
No noticeable changes in behaviour were observed with the 1.86 ng ml−1 concentration. At the 3.75 ng ml−1 concentration the animals continued to ambulate spontaneously but with grossly impaired coordination. At the 5.63 ng ml−1 concentration some animals moved slightly when righting was assessed but none made attempts at righting. At the 7.50 ng ml−1 concentration the animals remained completely immobile throughout.
Concentration-effect relationships
Figure 1 shows the effect of dexmedetomidine on 30-s EEG segments from one animal. The corresponding spectrograms reveal that dexmedetomidine reduced high-frequency power in a concentration-dependent manner. This is further illustrated by the line graphs to the right of the spectrograms. The concentration-dependent attenuation of power by dexmedetomidine is also readily noticeable in Figure 2 that shows the average spectra based on all animals (top panels) and on the data from a representative animal (bottom panels). The spectra reveal ordering on the basis of concentration with the baseline spectrum on top and the 7.50 ng ml−1 spectrum on the bottom, with a distance of about 0.2 log unit between baseline and the 7.50 ng ml−1 concentration. Figure 3 shows the data for each frequency range and the regression fit. It demonstrates that dexmedetomidine produced a linear decrease of logarithmic EEG power as a function concentration for both the cortex and thalamus in all frequency ranges. Between baseline and the 7.50 ng ml−1 plateau, power decreases by about 0.2 log unit (37% reduction). The attenuation of power by dexmedetomidine was statistically significant for both cortex (P<0.037) and thalamus (P<0.0051) (N=10) and the concordance correlation coefficients (CCC) were above 0.5 in all instances, indicating that the analysis yielded a good fit (Table 1). The effect of dexmedetomidine on cortex and thalamus was of same magnitude as revealed by similar slopes of the regression line (β1 parameter in Table 1) (P>0.46).
Regression results. CCC: concordance correlation coefficient. β1: regression coefficient for the linear term (slope). SE: standard error of the regression coefficient. P: probability that the regression coefficient is different from zero (t-distribution). P corrected: P values after correction for multiple (4) comparisons
| CCC | β1 | SE β1 | P β1 | P β1 corrected | |
|---|---|---|---|---|---|
| Cortex | |||||
| 30–50 Hz | 0.71 | −0.032 | 0.012 | 0.0105 | 0.0249 |
| 51–75 Hz | 0.73 | −0.035 | 0.013 | 0.0089 | 0.0249 |
| 76–125 Hz | 0.71 | −0.026 | 0.010 | 0.0166 | 0.0332 |
| 126–200 Hz | 0.66 | −0.021 | 0.010 | 0.0365 | 0.0365 |
| Thalamus | |||||
| 30–50 Hz | 0.81 | −0.031 | 0.009 | 0.0008 | 0.0016 |
| 51–75 Hz | 0.73 | −0.032 | 0.009 | 0.0005 | 0.0012 |
| 76–125 Hz | 0.82 | −0.037 | 0.009 | <0.0001 | 0.0004 |
| 126–200 Hz | 0.71 | −0.031 | 0.011 | 0.0051 | 0.0051 |
| CCC | β1 | SE β1 | P β1 | P β1 corrected | |
|---|---|---|---|---|---|
| Cortex | |||||
| 30–50 Hz | 0.71 | −0.032 | 0.012 | 0.0105 | 0.0249 |
| 51–75 Hz | 0.73 | −0.035 | 0.013 | 0.0089 | 0.0249 |
| 76–125 Hz | 0.71 | −0.026 | 0.010 | 0.0166 | 0.0332 |
| 126–200 Hz | 0.66 | −0.021 | 0.010 | 0.0365 | 0.0365 |
| Thalamus | |||||
| 30–50 Hz | 0.81 | −0.031 | 0.009 | 0.0008 | 0.0016 |
| 51–75 Hz | 0.73 | −0.032 | 0.009 | 0.0005 | 0.0012 |
| 76–125 Hz | 0.82 | −0.037 | 0.009 | <0.0001 | 0.0004 |
| 126–200 Hz | 0.71 | −0.031 | 0.011 | 0.0051 | 0.0051 |
Regression results. CCC: concordance correlation coefficient. β1: regression coefficient for the linear term (slope). SE: standard error of the regression coefficient. P: probability that the regression coefficient is different from zero (t-distribution). P corrected: P values after correction for multiple (4) comparisons
| CCC | β1 | SE β1 | P β1 | P β1 corrected | |
|---|---|---|---|---|---|
| Cortex | |||||
| 30–50 Hz | 0.71 | −0.032 | 0.012 | 0.0105 | 0.0249 |
| 51–75 Hz | 0.73 | −0.035 | 0.013 | 0.0089 | 0.0249 |
| 76–125 Hz | 0.71 | −0.026 | 0.010 | 0.0166 | 0.0332 |
| 126–200 Hz | 0.66 | −0.021 | 0.010 | 0.0365 | 0.0365 |
| Thalamus | |||||
| 30–50 Hz | 0.81 | −0.031 | 0.009 | 0.0008 | 0.0016 |
| 51–75 Hz | 0.73 | −0.032 | 0.009 | 0.0005 | 0.0012 |
| 76–125 Hz | 0.82 | −0.037 | 0.009 | <0.0001 | 0.0004 |
| 126–200 Hz | 0.71 | −0.031 | 0.011 | 0.0051 | 0.0051 |
| CCC | β1 | SE β1 | P β1 | P β1 corrected | |
|---|---|---|---|---|---|
| Cortex | |||||
| 30–50 Hz | 0.71 | −0.032 | 0.012 | 0.0105 | 0.0249 |
| 51–75 Hz | 0.73 | −0.035 | 0.013 | 0.0089 | 0.0249 |
| 76–125 Hz | 0.71 | −0.026 | 0.010 | 0.0166 | 0.0332 |
| 126–200 Hz | 0.66 | −0.021 | 0.010 | 0.0365 | 0.0365 |
| Thalamus | |||||
| 30–50 Hz | 0.81 | −0.031 | 0.009 | 0.0008 | 0.0016 |
| 51–75 Hz | 0.73 | −0.032 | 0.009 | 0.0005 | 0.0012 |
| 76–125 Hz | 0.82 | −0.037 | 0.009 | <0.0001 | 0.0004 |
| 126–200 Hz | 0.71 | −0.031 | 0.011 | 0.0051 | 0.0051 |
Unprocessed EEG traces and spectrograms from one animal. Top panels: On the left EEG traces (30 s duration) for each period and recording site (cortex in black, thalamus in blue). On the right is an expanded view of a 3 s segment. For better visibility the vertical scale is adjusted to ensure that each segment occupies the entire plotting area and the number to the left of each EEG segment indicates the scale of the vertical bar in µV. Middle panels – cortex: On the left is shown the spectrogram for the 30 s EEG segment recorded from the cortex for each concentration. Y axis: frequency; X axis: time; colour scale: power (dB). The decrease in high-frequency power is revealed by the predominance of the blue colour for the higher concentrations of dexmedetomidine. On the right, line plots showing the average value in dB for each frequency band and period. The grey rectangle shows the spectrogram data corresponding to the 30–50 Hz band during baseline and the average of these data (∼ 33 dB) is shown by the dot circled in grey. The line plots show that power decreased with increasing concentrations. Bottom panels – thalamus: Same as middle panels but for thalamic recordings.
Changes in EEG power spectra during dexmedetomidine anaesthesia. The top panels show average power spectra for the group of 10 animals for cortex and thalamus. Bars show standard error. The blanks at 60, 120 and 180 Hz represent data excluded from analysis to minimize interference from external electrical sources. Note the log-log scales. Bottom panels show power spectra obtained from one animal for cortex and thalamus. Recovery data not shown to avoid distracting overlap. Abbreviations: base: baseline.
EEG power as a function of dexmedetomidine concentration for each frequency band for cortex and thalamus. Thin lines represent data from each animal. The thick blue line shows the mean. The dashed red line shows the regression fit. Data from the recovery period were not included in the regression. The significance of the regression model was P<0.05 in all instances. See Table 1 for exact P values and regression parameter estimates.
The inclusion of all data points in Figure 3 draws attention to the variability of the data. Conventional representation of these data as mean and standard error (Supplementary Fig. S1) revealed normal variability. Furthermore, analysis of the residuals as part of the regression procedure revealed no anomalies. Finally, comparison with the raw data of our recent reports12,13 indicated that the variances were similar across the three studies (data not shown).
Figure 4 shows the comparison of the normalized attenuation slope for dexmedetomidine with those recently reported for propofol12 and isoflurane.13 The slope for dexmedetomidine in the cortex was less than half that of isoflurane in the 76–125 Hz (P=0.0014) and 126–200 Hz (P<0.0001) bands. The effect of dexmedetomidine on the cortex for frequencies above 75 Hz is thus less pronounced than that of isoflurane, but similar to that of propofol. For the thalamus, the slope for dexmedetomidine in the 51–75 Hz, 76–125 Hz and 126–200 Hz bands was less than half that of propofol (P=0.018, P=0.0008 and P=0.003, respectively) and isoflurane (P<0.0001 for all three bands). The effects dexmedetomidine on the thalamus for frequencies above 50 Hz is thus much less pronounced than that of propofol and isoflurane.
Normalized regression slopes for dexmedetomidine, propofol and isoflurane for cortex and thalamus (mean std err). The P values (adjusted for multiple comparisons) indicate the significance of the slope differences between isoflurane or propofol and dexmedetomidine for each frequency band. For the cortex the slope for dexmedetomidine is significantly less steep than that of isoflurane for the 76–125 Hz and 126–200 Hz bands. For the thalamus, the slope for dexmedetomidine is significantly less steep than that of propofol and isoflurane for all frequency bands except 30–50 Hz.
Figure 4 also reveals another interesting difference between dexmedetomidine and the other two agents. For propofol and isoflurane, the regression slopes become steeper (more negative) for the higher frequency ranges. For example, for propofol and cortical power, the slope for the 30–50 Hz, 51–75 Hz, 76–125 Hz and 126–200 Hz bands is 0.06, -0.12, -0.24, and -0.34, respectively. We did not comment explicitly on this observation in our previous papers12,13 although this was a robust effect for both agents and recording sites, as revealed by significant interactions between factors concentration and band frequency (repeated measures ANOVA implemented with SAS PROC MIXED; P<0.004). By contrast, the slopes for dexmedetomidine are about the same for all frequency bands. The same analysis with SAS PROC MIXED for the present dexmedetomidine data revealed no significant interaction (P=0.39 for cortex and 0.976 for thalamus).
Recovery
Valid recordings were obtained from seven animals. There was a significant increase in power for all frequency bands for both cortex (P=0.023) and thalamus (P=0.0186) during recovery compared with the previous period (7.50 ng ml−1 concentration) (data shown in Fig. 3). Recovery recordings were obtained to exclude unrelated causes for the observed decrease in power such as technical fault.
Thalamocortical coherence
Figure 5 shows the average coherence for baseline alone (left column) and for all periods (right column) for the four high-frequency bands and for 0.3–30 Hz range. A modest increase in coherence in the delta band (shown by the grey rectangle) was observed at the two dexmedetomidine concentrations causing unconsciousness (5.63 and 7.50 ng ml−1). The mean and standard error of delta band coherence for each period is shown on the bar graph. Coherence during the two periods causing unconsciousness was significantly higher than during all other periods when the animals were conscious except for the 5.63 ng ml−1vs 3.75 ng ml−1 comparison (repeated measures ANOVA and pair-wise comparisons with Tukey’s HSD test). This was a robust effect with all P values except one below 0.01.
Thalamocortical coherence. The left column shows the average coherence with 95% confidence interval for the group of 10 animals for each frequency band during baseline. The right column shows the average coherence for the same frequency bands and for each period. The grey rectangle show the delta (0.3–4.0 Hz) band where coherence was increased at the two highest dexmedetomidine concentrations, when the animals were unconscious. The blanks at 60, 120 and 180 Hz represent data excluded from analysis to minimize interference from external electrical sources. The bar graph on the right shows the mean and standard error of delta coherence measures obtained from each animal. The grey bars correspond to periods when the animals were unconscious and show a modest, but highly significant, increase in coherence compared with the black bars that correspond to periods when consciousness was present.
A similar analysis of coherence for the previously published propofol data12 also revealed an increase in delta coherence at the two propofol concentrations causing unconsciousness (9 µg ml−1 and 12 µg ml−1) (data not shown). Coherence during these two periods was significantly higher than during baseline and recovery (P<0.05). Analysis of the previously published isoflurane data13 also showed an increase in low frequency coherence but it was limited to the 0.3–1.0 Hz range and observed only at the 1.5% concentration (data not shown). Coherence at this concentration was significantly higher than during baseline and recovery (P<0.05).
Discussion
The main finding of this study is that dexmedetomidine produces a concentration-dependent reduction of the power of high-frequency thalamocortical (30–200 Hz) rhythms in both cortex and thalamus. Thus, despite its different molecular mechanism action dexmedetomidine shares with propofol12 and isoflurane13 the ability to attenuate high-frequency thalamocortical rhythms. The findings also reveal noteworthy differences between dexmedetomidine and the other agents. The most interesting in our view is that the effect of dexmedetomidine on the thalamus is more modest than that of propofol and isoflurane. The slope of the concentration-dependent attenuation of thalamic rhythms for power above 50 Hz by dexmedetomidine is less half than that of propofol or isoflurane. Furthermore, the effect of dexmedetomidine on cortex and thalamus is of similar magnitude, in contrast to both isoflurane and propofol, which exerted a stronger effect on the thalamus.
The combined observations of the effects of these three agents on thalamic high-frequency (30–200 Hz) rhythms indicate that anaesthetic-induced hypnosis is linked with impairment of thalamic function and support earlier reports of thalamic involvement.28–31 We propose that these alterations of thalamic fast rhythms could be considered an electrophysiological signature of the anaesthetized state because they likely reflect changes in thalamic neuronal activity that have also been repeatedly documented by functional brain imaging studies.28,29,32 High-frequency LFP power correlates closely with regional cerebral blood flow and the precision of synchronous firing of neurones within a brain volume contributes to the magnitude of both the haemodynamic response and of the fast neuronal oscillations.33 Decreased thalamic blood flow during anaesthetic-induced unconsciousness has been one of the most robust functional brain imaging findings. Thus electrophysiological and blood flow studies provide cohesive evidence of decreased thalamic activity during anaesthesia.
The fact that some degree of impaired thalamic function was observed with all anaesthetic agents should not let us ignore the differences in the electrophysiological effects of these drugs. This is a further reminder that anaesthetic-induced hypnosis may be associated with different patterns of thalamocortical EEG alterations.
The picture for cortical rhythms is more complex. Cortical power in the 30–200 Hz range is attenuated in a concentration-dependent manner by the three agents tested, except for propofol which had no effect on cortical power in the 30–75 Hz range. Thus impairment of cortical function, as revealed by attenuation of high-frequency rhythms above 75 Hz, may provide a correlate of hypnotic effect. Comparison between agents revealed a stronger attenuation of cortical power by isoflurane compared with dexmedetomidine (over the 76–200 Hz range) or propofol (over the 30–200 Hz range).13 The most likely explanation for the steeper concentration-effect slope of isoflurane is that this agent, like other volatile anaesthetics, has additional molecular targets, notably the two-pore potassium channels and the NMDA receptor.34
The slope of the concentration-dependent attenuation of high-frequency power by dexmedetomidine was of similar magnitude for all frequency bands, in contrast to isoflurane and propofol, for which the regression slope was steeper for the higher frequency bands. A possible explanation for the influence of the frequency band for isoflurane and propofol is the pivotal role that inhibitory GABAergic neuronal networks play in the synchronization of fast rhythms.35 As the frequency of the rhythm increases, so does its dependence on inhibitory currents with fast time constants. Isoflurane and propofol modulate the action of GABAA receptors and prolong the inhibitory currents, thereby disrupting normal physiologic circuits that require precise timing of GABAergic input.15 Dexmedetomidine, by contrast, exerts no direct effect on GABAA receptors.
The comparison between dexmedetomidine, propofol and isoflurane require careful interpretation. The concentration ranges were chosen by aiming for equivalent effects on the loss of the righting reflex. Thus, this comparison of potency may not apply to the other effects of these drugs, such as suppression of movement in response to pain. We can nevertheless affirm that, in the concentration range relevant loss of the righting reflex, the attenuation of fast rhythms by dexmedetomidine is in most instances less pronounced than that of propofol and isoflurane. Whether this could account for the differences of clinical profile between dexmedetomidine and the other drugs (such as the preserved ability to be awakened by verbal commands during deep sedation with dexmedetomidine14) remains to be ascertained.
The neurophysiological role of cortical fast rhythms in attentional and cognitive processes is well documented.35 There is also evidence that these rhythms may be involved in conscious perception. In a study of conscious and nonconscious processing of briefly flashed words using a visual masking procedure while recording intracranial EEG in ten epileptic patients, Gaillard and colleagues36 reported that conscious processing was associated with an large increase in high-gamma (50–100 Hz) power occurring 200–500 ms post-stimulus. This increase was much larger and more sustained than for words that were not consciously perceived. If the neural events leading to conscious perception involve sustained increases in the power of fast rhythms, it is plausible that the impairment of fast rhythms by general anaesthetics could reflect an interference with neural processes required for conscious access.
We proposed12,13 that the attenuation of power 80 Hz and above by propofol and isoflurane reflects a decrease of neuronal firing rates based on the evidence that high-gamma power (80–150 Hz) in LFPs is strongly correlated with average firing rate11,37 and that both drugs decrease spontaneous rate of neuronal action potentials in the both cortex and thalamus.38–41 It is plausible that this may also be the case for dexmedetomidine, although we found no information on its effects on cortical and thalamic firing rates. Simultaneous recordings of action potentials and LFPs with the same electrode could resolve this issue.
The increase in delta band thalamocortical coherence during unconsciousness induced by dexmedetomidine and the absence of changes in the other bands is surprising given the evidence that general anaesthesia is associated with decreased corticocortical EEG coherence for both low and high-frequency ranges.17,42 Furthermore, an analysis of our previously published data12,13 indicates that an increase in low frequency coherence also occurs with propofol and isoflurane.
Decreased EEG coherence between distinct areas of the cerebral cortex during anaesthesia is interpreted as a sign of impaired neural communication, one of the proposed mechanisms by which general anaesthetics impair consciousness.1 How then, can the present finding of increased thalamocortical coherence in the delta band be reconciled with the substantial literature reporting decreased EEG coherence during anaesthesia?17 We see three explanatory elements. Most of the literature or anaesthetics and EEG coherence involves coherence between different areas of the cortex (corticocortical coherence), is based on recordings with macro-electrodes (scalp EEG or skull screws) that sample extended volume of brain tissue and use referential recording (i.e. the electrodes are all referred to one single electrode, the reference, that hopefully has little brain activity in it). The present study is about coherence between cortex and thalamus, is based on microelectrode recordings that sample the activity of much smaller volume of brain tissue and uses bipolar recording (i.e. between the two exposed tips of the electrodes). These factors may possibly account for the apparent discrepancy between the literature and the present findings.43,44
It should be noted that the use of referential recordings to assess EEG coherence is considered problematic and often overestimates coherence because the reference electrode is rarely completely devoid of EEG activity.43,44 This caveat does not affect the validity of observations that report decreases in coherence during anaesthesia on the basis of referential recordings. It does however suggest that the interpretation of such findings is not as simple as previously assumed.
The increase in low frequency coherence caused by dexmedetomidine, propofol and isoflurane is compatible with the proposal that the slow waves observed during physiologic slow-wave sleep and anaesthesia reflect cyclic hyperpolarization of thalamocortical networks, a phenomenon leading to a phenomenon now called UP and DOWN states.45–47 It is plausible that the present increase in low frequency coherence results from the large slow waves overriding the low amplitude neural activity that sustains neural communication during wakefulness. For dexmedetomidine, the hyperpolarization of thalamocortical networks likely results from its action on endogenous sleep pathways.48 For propofol and isoflurane, potentiation of GABA acting on GABAA receptors and other mechanisms could explain the hyperpolarization.49,50
We conclude that dexmedetomidine, like propofol and isoflurane, produces a concentration-dependent attenuation of the spectral power of thalamocortical rhythms in the 30–200 Hz range but that there are two noteworthy differences. The effect of dexmedetomidine is weaker than that of the other agents and of similar magnitude for thalamus and cortex, in contrast to the other agents that exerted a stronger effect on thalamic rhythms. While these correlational observations support the hypothesis that anaesthetic-induced unconsciousness is associated with alterations of high-frequency thalamocortical rhythms, they also indicate that the magnitude of these alterations depends on the anaesthetic agent. More work is required to determine how and to what extent the physiological changes revealed by attenuation of high-frequency spectral power contribute to the hypnotic action of different general anaesthetics.
Authors’ contributions
Study design/planning: G.P.
Study conduct: G.P., F.A.
Data analysis: G.P., F.A.
Writing paper: G.P.
Revising paper: all authors
Supplementary material
Supplementary material is available at British Journal of Anaesthesia online.
Acknowledgements
We thank Steven Shafer, MD (Department of Anesthesiology, Columbia University, New York, NY) for help with the Stanpump software; René Deroussent for technical assistance; Xianming Tan (Biostatistics Core Facilities, McGill University Health Center Research Institute) for advice on regression analysis and SAS programming; C. Andrew Chapman (Concordia University, Montreal, QC) for comments on earlier version of the manuscript.
Declaration of interest
None declared.
Funding
Supported in part by grants from the Canadian Anesthesiologists' Society, and Fondation d'Anesthésiologie et Réanimation du Québec and from departmental funds.
