| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
Waikato Clinical School, Waikato Hospital, Hamilton, New Zealand
Address correspondence and reprint requests to James Wallace Sleigh, MD, Waikato Clinical School, Waikato Hospital, Private Bag, Hamilton, New Zealand. Address e-mail to sleighj{at}hwl.co.nz
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
IMPLICATIONS: During the recovery phase from a remifentanil-based anesthetic, the bispectral index is not reliably predictive of the depth of consciousness, because of suppression ratio artifacts. Entropy measures of the electroencephalogram show promise, but there is still no gold standard to estimate anesthetic depth.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This study aims to examine whether there is a reliable correlation between the clinically assessed state of consciousness during recovery from almost pure remifentanil-based "opioid anesthesia" and various measures of EEG and midlatency AEPs. We also assessed whether significant implicit and explicit conceptual memory occurred in the patients in the period before they became responsive to verbal command.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
) were placed to produce two bipolar signals (Fp1 to F7 and Fp2 to F8, ground Fz). We used the Aspect A-1000 EEG monitor using software version 3.31 (Aspect Medical Systems, Newton, MA) and collected both the raw EEG signal (sampled at 128/s, filtered 0.5–70 Hz, 50-Hz notch) and the processed data (5-s updates of the BIS) into a personal computer for later analysis. Because of lack of space on the forehead, the AEP electrodes were placed slightly above the EEG electrodes—left mastoid (O1) to midline frontal (above Fz) montage, with the reference on the left frontal (above F7). We used the A-Line AEP monitor (Danmeter A/S, Odense, Denmark). The concurrent auditory input of the AEP monitor does not affect the BIS value (7). To prevent intraoperative awareness during surgery, the sevoflurane was adjusted to maintain a BIS value of <60, or as clinically indicated. Our study design included small amounts of sevoflurane before study commencement because 1) it would clearly be unethical to allow intraoperative awareness, 2) the problem of opioid-induced rigidity is eliminated, and 3) it most closely mimics a typical clinical situation, thus having direct applicability. At the end of surgery, the end-tidal sevoflurane concentration (FETsevo) was noted (median, 0.5%; range, 0.2%–1.1%), and the sevoflurane was turned off. This marked the beginning of a two-stage awakening phase. At this point, we began recording the AEP, BIS, and raw EEG data. During the first stage—the sevoflurane washout period—both sevoflurane (in decreasing concentrations) and remifentanil were present. The starting time point was designated as "sevo off." Because the object of the study was to record recovery from relatively pure "opioid anesthesia," we continued the remifentanil infusion until the sevoflurane was completely washed out. Once the measured FETsevo reached 0%, a further 2-min of washout continued before the remifentanil infusion stopped. From this point, it was assumed that the state of consciousness of the patient was almost completely determined by the (diminishing) remifentanil concentrations. Assuming the blood-brain effect compartment half-life of sevoflurane to be in the order of 2–3 min, the brain concentrations of sevoflurane at this point would be <0.1% (8). The sevoflurane washout phase lasted a median of 519 s (range, 140 to 1133 s). The remifentanil infusion was then turned off, a prerecorded tape was started, and the patient was allowed to awaken. The starting point of this second stage—the remifentanil washout period—was designated "remi off."
The tape instructed the patient every 30 s with the same command ("move your right hand") and a command to remember a series of simple word pairs (category/exemplar, e.g., "remember animal/mouse"). Patients were considered to be conscious when they responded to the loud repeated commands on the tape with appropriate responses or by opening their eyes; with the Observer Assessment of Alertness and Sedation (OAA/S) rating, this equates consciousness with a score of 
3 (9). This time point was designated "awake," the tape was stopped, and the patient was then tracheally extubated at the anesthesiologist’s discretion. In the postanesthetic care unit (PACU), they were given fentanyl (20–25 µg IV) if required for analgesia. Because we wanted to see whether any EEG measures could differentiate near-responsiveness/light anesthesia (OAA/S = 2) from responsiveness/sleep (OAA/S = 3), we designated a fourth point, 30 s before awakening, as "almost responsive." Because of the rapid pharmacokinetic characteristics of remifentanil, all patients but one were completely alert and cooperative by the time they got to PACU.
Within 5 min of arrival in the PACU, the patients were questioned to test for explicit and implicit conceptual memory. Explicit memory was tested by asking patients to recall their last memory before anesthesia and their first memory after anesthesia. Any additional comments that were volunteered were noted. To test further for explicit and implicit memory, patients were questioned to see what they remembered from the tape. This involved reading to them the same list of categories from the tape. With each category (e.g., animal), they were also read 10 exemplars and were asked to pick one exemplar that they thought they might have dreamed about or heard (explicit recall) or, otherwise, the one exemplar that stood out to them (indicative of implicit conceptual memory).
Because we cannot assume that the exemplars selected follow a uniform probability distribution—some exemplars may intrinsically be more appealing, attractive, or noticeable—we also gave a questionnaire containing these same categories (10 exemplar lists) to 21 randomly selected healthy controls who had not been anesthetized or heard the tape. We instructed the controls to choose one exemplar (from the choice of 10 exemplars in each category) that stood out to them. An answer was defined as a "correct exemplar" if it was the same as the exemplar present on the tape. This allowed us to construct a probability distribution for correct exemplars from a group of people who had not previously heard the correct exemplars on the tape. Because not all patients got to the end of the word lists before they awoke, we randomly adjusted the length of the lists given to the controls so that they matched that achieved by the patients.
In addition to the Aspect-derived variables—BIS, 95% spectral edge frequency (SEF), and suppression ratio (SR)—four variables were calculated from the raw EEG signal. These were 1) the spectral entropy (SEN), 2) the entropy of the singular value decomposition (SVDEN), 3) the canonical univariate variable (CUP), and a subcomponent of the BIS, 4) the SynchFastSlow. The variables were calculated by use of the Matlab programming environment (Matlab 6.1; The Mathworks Inc., Natick, MA). The details and significance of the calculation of these EEG variables are discussed more fully in an on-line appendix to this article. The CUP and SEF were included because they have previously been recommended as good measures of opioid effect (2,10). The two entropies offer conceptually novel approaches to the quantification of consciousness that hitherto have not been applied to the estimation of opioid-induced sedation. They both measure the complexity, or irregularity, of the EEG signal. The SEN directly estimates the flatness of the power spectrum, whereas the SVDEN uses a lagged phase-space embedding of the time series similar to that used to calculate the approximate entropy. The EEG of an anesthetized patient is relatively simple because cortical function is constrained by the anesthetic drug. The EEG has a narrow power spectrum (SEN <0.7), and most of the information within the signal can be captured by a few dominant singular values (SVDEN <0.8). In contrast, the EEG of an awake patient is less constrained and has a broad power spectrum (SEN 
1). Because it is more complex (has higher dimensions), it requires many significant singular values to extract the information contained within the signal. The entropy of these singular values then approaches its maximum of 1. The SynchFastSlow and SR are subvariables of the BIS and were derived to aid in the understanding of the paradoxical changes in BIS that we observed. The A-Line ARX index (AAI) was used as a measure of AEP. This machine uses an autoregressive-exogenous input algorithm to rapidly extract midlatency AEPs in the window of 20 to 80 ms poststimulus and thence to derive an index based on the latency and amplitude of the evoked wave form. The advantage of this method is that it can detect the evoked potentials much more quickly than the traditional signal-averaging technique. The effect of opioids on AAI has not been studied in the context of recovery from anesthesia.
We estimated remifentanil effect-site concentrations for each patient during remifentanil recovery period by using published pharmacokinetic variables (2,11). Statistical analysis was performed with SyStat Version 10 (SPSS Inc, Chicago, IL). The Pearson correlation coefficient (r) and 
2 were used where appropriate. We used the 2 test of goodness of fit to compare the distribution of correct words between the control group and the patient group. Within-subject changes in EEG variables at each time point were compared by use of the paired Student’s t-test with Bonferroni correction for three comparisons: 1) "sevo off" to "remi off" (the effect of sevoflurane), 2) "remi off" to "almost responsive" (the effect of remifentanil), and 3) "almost responsive" to "awake" (the effect of recovery of consciousness). Statistical significance was set at P < 0.05. The performance of each EEG variable was further compared by using the prediction probability statistic (Pk) as described by Smith et al. (12). It was calculated with the Somer’s dxy method. A Pk value of 1 indicates perfect concordance between the EEG variable and the clinical level of consciousness. Values of <0.5 suggest more discordance than concordance. First, we tested the discrimination of each EEG variable by using data from the "sevo off" (unresponsive) to "awake" (responsive) time points. This would be a measure of combined sevoflurane and remifentanil effects. Second, we compared EEG variables between "remi off" (unresponsive except for two patients who woke up at the full remifentanil dose) and "awake" (responsive) time points. This estimates the contribution of remifentanil effects only.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
2). At this point, remifentanil was still being infused at 0.4 µg · kg-1 · min-1 (just before the beginning of the remifentanil washout phase). Two of the patients were not unconscious at this point: one patient woke (opened eyes and obeyed command, FETsevo = 0%,SEN = 0.95, BIS = 96, AAI = 50) with the first command on the audio tape, and the other patient woke (opened eyes, FETsevo = 0.13%, SEN = 0.94, BIS = 82, AAI = 40) before the remifentanil stopped and the audio tape started.
For the implicit memory testing, the 2 goodness-of-fit test demonstrated that the probability of picking correct exemplars in the patient group was no better (P = 0.24) than for the control volunteers who had not heard the tape (Table one). In fact, there was a trend toward proportionately more correct exemplars in the control group. Eight of the patients had some explicit memory of the tape at the moment of awakening (eye opening or obeying command, BIS range 21–97). Of these, one had explicit memory from 60 s before awakening (she remembered "being told to move my right arm and something about poplar and dog"; BIS = 15, AEP = 19), and one had memory from 30 s before awakening (she remembered "dream being broken into and hearing the word food"; BIS = 60, AEP = 84).
|
|
|
The Pk values for each EEG variable are shown in Table 2. With the caveat of the very small numbers studied, the results indicate some superiority of the entropy measures over the AAI and the BIS. However, there is a high predictive error for all the EEG variables of 15% (SEN) to >40% (BIS). The calculated mean (SD) effect-site concentrations of remifentanil at the "remi off" time point were 11.5 ng/mL (2.9 ng/mL), and at the point of awakening these were 5.8 ng/mL (2.8 ng/mL).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Like Barr et al. (13), we demonstrated that BIS is not useful as a measure of depth of anesthesia or decrease in opioid concentration when remifentanil is used at clinical doses. This implies that extrapolated isobolograms for combinations of hypnotics and opioids constructed by using the BIS may not be reliable at very small concentrations of hypnotic (14). As part of the clinical management of the patients during surgery, we adjusted the sevoflurane concentrations to achieve BIS values in the recommended range. However, in retrospect, this study suggests that this practice may not be wise when an anesthetic technique is used with predominantly remifentanil and minimal sevoflurane. With this technique, the BIS definitely does not always show a monotonic relationship to the level of anesthesia. In approximately a quarter of the patients, the BIS plummets during the remifentanil washout phase. In this phase, the EEG is of low amplitude, which the BIS interprets as burst suppression, and the SR dominates the BIS, giving a misleadingly small value. We have observed this phenomenon subsequently when using the Aspect A2000 monitor.
Except for the BIS, the bias was toward lack of awareness; i.e., the EEG/AEP variables indicated that the patients should have been awake, but they were unresponsive to verbal command. A subgroup of five patients had EEG/AEP variables indicative of being awake (BIS >70, AAI >70, SEN >0.85, and SVDEN >0.92) for longer than 60 seconds (sometimes several minutes; see Fig. 1) before becoming responsive. As evidenced by these desynchronized EEG patterns, it is clear that they were in a different state of unresponsiveness than the usual anesthetic state induced by GABAergic hypnotic anesthetics. This phenomenon limits the use of any variable as an indicator of unconsciousness—mainly because a group of patients are clinically unresponsive, but exhibit an alert-looking EEG.
We were surprised at the frequent prevalence of explicit and implicit amnesia that we found. Opioids are generally not considered to be powerful amnesic drugs (15). Most previous work dealing with opioids and amnesia has concentrated on measuring explicit recall. We could find no articles dealing with implicit memory with an opioid-only anesthetic technique. We used a measure of conceptual implicit memory. It implies that for an unconscious memory to form, some degree of semantic processing must take place; i.e., the exemplar (e.g., "apple") must be linked to and understood as an example of the generic category ("fruit"). It is likely that some form of (unconscious) memory may be established under general anesthesia that does not require the higher semantic processing. This may be termed "perceptual" implicit memory and is often tested for by using a word-stem completion test. It is likely that implicit/unconscious perceptual memory formation is much more common under general anesthesia than implicit conceptual memory formation. However, it could be argued that alteration of the central nervous system ("memory") without the attachment of any meaning is of little importance. In any case, this "memory-like" process of imprinting is known to occur at all levels of the sensory pathway (16). In contrast to the lack of amnesic effects of opioids, it has been well documented that volatile anesthetics have a profound effect on memory (17). It is therefore probable that most of the amnesia seen in our study can be attributed to the residual effects of the very small brain concentrations of sevoflurane whose amnesic effects were potentiated by the remifentanil. Like those of other authors (18, 19), our data support the lack of inhibitory effect of remifentanil on AEPs. Among our patients, almost half had high AAI values (>50) once the sevoflurane washout had finished, but they were unresponsive.
Jhaveri et al.’s (1) findings and those of Lang et al. (20) suggest that remifentanil is not suitable as a sole anesthetic, either as an induction drug or as a maintenance drug. Nevertheless, we found that—even at the more modest dose of 0.4 µg · kg-1 · min-1—only extremely small amounts of sevoflurane are needed for unconsciousness. The two patients who woke when the sevoflurane was withdrawn did have EEG variables with very high values that would not normally be indicative of unconsciousness. Conscious recall would appear to be unlikely if the sevoflurane is titrated on top of the remifentanil infusion to keep the SEN <0.7 or the SVDEN <0.8. This needs to be confirmed with a larger study. The SVDEN may be a useful variable for assessing anesthetic and analgesic effects on the central nervous system.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. Jeleazcov, G. Schneider, M. Daunderer, B. Scheller, J. Schuttler, and H. Schwilden The Discriminant Power of Simultaneous Monitoring of Spontaneous Electroencephalogram and Evoked Potentials as a Predictor of Different Clinical States of General Anesthesia Anesth. Analg., October 1, 2006; 103(4): 894 - 901. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C.-C. Wu, M. S. Mok, J.-Y. Chen, G.-J. Wu, Y.-R. Wen, and C.-S. Lin Doxapram shortens recovery following sevoflurane anesthesia: [Le doxapram hate la recuperation apres une anesthesie au sevoflurane]. Can J Anesth, May 1, 2006; 53(5): 456 - 460. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. F. White, J. Tang, G. F. Romero, R. H. Wender, R. Naruse, A. Sloninsky, and R. Kariger A Comparison of State and Response Entropy Versus Bispectral Index Values During the Perioperative Period Anesth. Analg., January 1, 2006; 102(1): 160 - 167. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. A. Dahaba Different Conditions That Could Result in the Bispectral Index Indicating an Incorrect Hypnotic State Anesth. Analg., September 1, 2005; 101(3): 765 - 773. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Hagihira, K. Okitsu, and M. Kawaguchi Unusually Low Bispectral Index Values During Emergence from Anesthesia Anesth. Analg., April 1, 2004; 98(4): 1036 - 1038. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. E. Anderson and J. G. Jakobsson Entropy of EEG during anaesthetic induction: a comparative study with propofol or nitrous oxide as sole agent{dagger} Br. J. Anaesth., February 1, 2004; 92(2): 167 - 170. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. J. Manberg, A. R. G. Muncaster, J. W. Sleigh, and M. Williams BIS Monitoring Requires Proper Electrode Placement for Optimum Performance * Response Anesth. Analg., October 1, 2003; 97(4): 1206 - 1206. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
