| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
From the *Département d'Anesthésie, Institut Gustave Roussy, Villejuif; Département d'Anesthésie-Réanimation, Hôpital Bellevue, Saint-Etienne; and AP-HP, Service d'Anesthésie-Réanimation Chirurgicale, Hôpital Bichat, Paris, France.
Address correspondence and reprint requests to Dr. Grégoire Weil, Département d'Anesthésie, Institut Gustave Roussy, 39, rue Camille Desmoulins, 94805 Villejuif, France. Address e-mail to weil{at}igr.fr.
 |
Abstract |
|---|
 Top Abstract IntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
METHODS: In this prospective, open, multicenter study, 105 patients received propofol then remifentanil target-controlled infusion for induction of anesthesia, with randomized remifentanil targets ranging from 2 to 8 ng/mL. Half of the patients received neuromuscular blockade. Intubation and incision were used as standard noxious stimulations, motor or hemodynamic responses were recorded, and spectral entropy values before and after stimulations were compared between responders and nonresponders.
RESULTS: No difference was found in response entropy (RE), state entropy (SE), or (RE – SE) between patients with or without hemodynamic response to stimulations. Patients with motor response to intubation had higher values of RE, SE, and (RE – SE) both before and after the intubation than patients with no response. These results were confirmed by a prediction probability analysis, showing a significant but weak predictive value of entropy for motor response only.
CONCLUSIONS: Entropy predicted a motor response to noxious stimulations but not a hemodynamic response, which limits its usefulness for assessing the analgesic component of anesthesia in paralyzed patients. High values (RE >55) before the stimulation should be avoided in order to decrease the risk of motor response, but lower values might not prevent this response when the opioid concentration is insufficient, despite an adequate hypnosis.
|
Introduction |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
As all anesthetic drugs also induce dose-dependant changes in the electroencephalogram (EEG), EEG-derived parameters have been proposed for several years to assess depth of anesthesia.1 The first monitor designed specifically to assess depth of anesthesia was the Bispectral Index Monitor (BISTM). The BIS is related to hypnotic drug concentration and the probability of unconsciousness2 but does not predict the response to a noxious stimulation before it occurs.3
More recently, other EEG analysis techniques based on entropy have been proposed. Initially defined for thermodynamic analysis, then used in information theory by Shannon and Weaver, entropy expresses the amount of disorder and unpredictability in a system. When applied to EEG waveforms, entropy decreases at deeper planes of anesthesia. EEG entropy was first calculated in the time domain and showed a good correlation with volatile anesthetic concentrations.4 Entropy calculated in the frequency domain is the basis for a commercially available device for monitoring anesthetic depth (EntropyTM Module for S/5 monitor, Datex-Ohmeda, Helsinki, Finland).
Compared with the time domain prototypes, frequency domain entropy calculation, as proposed in this device, has several theoretical advantages. The time epoch used to calculate the spectrum is not fixed but decreases from 60 s for low frequencies to 1.96 s for the highest frequencies. This short epoch should allow immediately catching any change in fast frequencies and might therefore be useful to detect an EEG response to a noxious stimulation more rapidly in case of insufficient analgesia.5
Moreover, it is not limited to EEG analysis but also includes electromyogram (EMG) as a part of the signal rather than as an artifact since it may reflect a patient response to some external stimulus, such as a painful stimulus. The entropy module computes two parameters, state entropy (SE) and response entropy (RE). SE is calculated in a frequency band ranging from 0.8 to 32 Hz (dominated by the EEG signal) and is expected to reflect mainly the hypnotic component of anesthesia, similar to the BIS monitor. RE extends the frequency response from 0.8 to 47 Hz, and is assumed to reflect both hypnosis and analgesia, the additional spectrum being dominated by EMG activity. Both parameters are displayed on the monitor, as well as the difference (RE – SE), which is considered to be a marker of nociception, and tends to 0 when there is no EMG.5
Although the EMG appearance may reflect insufficient analgesia, the sensitivity and specificity of EMG as a marker of pain are unknown. Further, EMG activity can be totally eliminated by neuromuscular blockade (NMB), and the behavior of entropy in this case needs to be described.
The aim of this study was to assess the correlation between entropy parameters, analgesic drug concentration and clinical response to a noxious stimulation, as well as to examine the influence of NMB on this relationship.
|
METHODS |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Exclusion criteria were age <18 or >70 yr, neurological disease or treatment liable to affect EEG, preoperative administration of sedative or opioid drugs, foreseeable difficult intubation, and known anesthetic drug allergy.
Two groups of patients were considered: those requiring NMB for surgery had a bolus of NMB at LOC to facilitate tracheal intubation, whereas the other patients were intubated without NMB. In each group, patients were randomly allocated to receive remifentanil via target-controlled infusion (TCI) to a remifentanil effect-site concentration at intubation of 2, 4, 6, or 8 ng/mL in order to explore the range of concentrations usually recommended for intubation.
Study Design
After arrival in the operating room, an IV line was inserted into a forearm vein for fluid and drug infusion. Standard monitoring was established to monitor both hemodynamics and ventilation using an S/5 Anesthesia Monitor (GE Health Care, Helsinki, Finland).
The forehead skin was prepared with 70% isopropanol and Red Dot® tape (3M) and an Entropy sensor was positioned as recommended by the manufacturer and connected to the Entropy Module of an S/5 Anesthesia Monitor.
Anesthesia was induced using propofol and remifentanil TCIs targeting the effect-site (Base PrimeaTM Infusion System, Fresenius, Brezins, France). The system used Schnider et al.'s and Minto et al.'s pharmacokinetic models for propofol and remifentanil, respectively.6,7
The propofol infusion was started at an initial target concentration (TC) of 4 to 5 µg/mL, which was increased if the patient had not lost consciousness after 4 min.
When appropriate, an NMB bolus (atracurium 0.5 mg/kg or cisatracurium 0.2 mg/kg) was given after LOC.
Remifentanil TCI was started at least 5 min after propofol (or 5 min after the last TC change) at the randomized effect-site concentration, which was then kept constant for intubation and until 5 min after incision.
Laryngoscopy and tracheal intubation were performed at least 5 min after starting remifentanil, and were always performed by the same physicians in each center (i.e., the coauthors).
Data Recording
Heart rate (HR) and SE and RE were recorded at 5-s intervals with the S/5 Collect software (GE Healthcare, Finland) onto a computer hard drive for off-line analysis.
HR, mean arterial blood pressure (MBP), SE, RE, (RE – SE), and predicted propofol and remifentanil effect-site concentrations values were related to the following clinical end points identified in the record by event markers:
Motor response to intubation or incision was defined by a movement or cough within 5 min after the stimulus. Hemodynamic response was defined as maximal HR or MBP increase >20% from prestimulation values within 5 min after the stimulation.
Statistical Analysis
The number of patients to include was based on the means and standard deviations of MBP and BIS response to intubation described by Guignard et al.3
The responses to intubation or incision were analyzed regardless of the predicted remifentanil TC (as may be in clinical practice without TCI), but considering the NMB status of the patients.
Demographic data, entropy values (RE, SE, and [RE – SE]) before and after intubation or incision were compared between NMB and no NMB, responders and nonresponders using Student's t-test. Motor response and hemodynamic response to both stimulations were considered separately.
The correlation of RE, SE, and RE – SE with the type of response to the stimulation was also assessed using the prediction probability (PK) described by Smith et al.8 and ranging from 0 to 1. A value of 0.5 means that the parameter predicts the response not better than chance. A value of 1 means that the parameter predicts the response correctly each time it is used. The prestimulation entropy values were used to assess the ability of this parameter to predict a response to stimulation. The poststimulation values were used to consider the relationship between entropy and the type of response. The jackknife method was used to compute the standard error of the estimate 
Pk. The value ([PK – 0.5]/Pk) was tested using Student's t-test and the PK value was judged to be better than random if P < 0.05.
Then, the sensitivity, specificity, and predictive value of prestimulation entropy indexes to predict the response was calculated for various entropy thresholds, in order to determine the threshold to recommend in clinical practice.
Analyses were performed with Excel (Microsoft, Richmond) and PK was calculated with the PK MACRO datasheet for Excel developed by Smith and corrected by Desmet. An 
level of 5% was used to establish statistical significance. Quantitative data are given as mean ± sd.
|
RESULTS |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Demographic data are shown in Table 1 without significant difference related to the use of NMB.
|
All but four patients were intubated at the first attempt and the other four at the second attempt in less than 5 min. The incidence of response to intubation or incision is shown in Table 2. Movements or hemodynamic responses dissipated after a few minutes. Three patients needed ephedrine sulfate injection (6 mg single bolus) associated with a reduction in propofol TC to correct a hypotensive episode. No other adverse effect was observed throughout the study period.
|
The hemodynamic response to intubation or incision did not significantly differ with NMB allowing the entropy values to be examined versus hemodynamic response for the whole population (paralyzed and unparalyzed, n = 105).
Only pre-remifentanil RE and after intubation (RE – SE) showed a small but statistically significant difference in patients who had a hemodynamic response to intubation compared with nonresponders (max [RE – SE] = 5 ± 6 vs 3 ± 3, Fig. 1).
|
Entropy values did not change markedly with incision (SE: 34 ± 11 before vs 37 ± 12 after, RE: 36 ± 13 before vs 37 ± 13), similarly in patients who had or not a hemodynamic response.
No patients having received NMB had a motor response to intubation or to incision.
Without NMB, 28 patients showed a motor response to intubation. For these patients, SE, RE, and (RE – SE) were significantly higher than in nonmovers after intubation (1 and 2 min), but also before starting remifentanil, i.e., under propofol TCI alone (Fig. 2).
|
Without NMB, six patients had a motor response to incision. SE and RE values were higher in these patients compared with the patients who did not move, but only RE preincision reached statistical significance (Fig. 3).
|
The prediction probability was consistent with these results, showing no significant predictive value of entropy parameters for hemodynamic response to intubation or incision, and a significant but weak predictive value of both RE and SE for motor response to intubation or incision (Table 3).
|
The best positive predictive value (i.e., RE above the threshold predicting a motor response to intubation) was obtained for RE = 55 with a predictive value of 86%. However, 22 patients did show a motor response with a preintubation RE value <55 as expressed by a negative predictive value of 51% only. In other words, RE <55 appeared to be necessary before any attempt of intubation without NMB to decrease the risk of motor response but it is not sufficient to be sure that this response will not occur.
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
In our study, all three parameters, RE, SE, and (RE – SE) values were higher after intubation in patients who had a motor response, the difference being slightly more significant for RE. They were also higher in patients who moved after incision but the difference did not reach statistical significance because of the high interindividual variability and the small number of patients moving (n = 6) due to the study design (no change in TC between intubation and incision). Similar results have been observed in three recent studies where RE was the best entropy parameter to reflect the response to skin incision11,12 or insertion of an arterial catheter.13 But, conversely to our findings, these studies found (RE – SE) to be a better measure than SE.
The small difference among the three parameters and the poor performance of (RE – SE) may be explained by the fact that 32 Hz is not an absolute lower bound for EMG: most of the EMG activity is above this threshold, but some EMG activities have lower frequencies and may be captured by both RE and SE, making (RE – SE) display useless. To support this assumption, a recent study found that a RE increase during surgery was associated with both BIS and EMG increases, which confirms that some EMG activity was present below 32 Hz.14
The difference in (RE – SE) behavior among studies may also have been due to the hypnotic used (volatile for Seitsonen et al.12 or Wheeler et al.13 and propofol for Rantanen et al.11 as in our study), which might have depressed muscular tone differently, or to the stimulation studied which was different. To examine the influence of the stimulation on entropy response, two stimulations were considered in our study: tracheal intubation and skin incision. As already described in previous studies,15,16 intubation appeared to be a stronger stimulation than incision, since 54% of patients had a motor response to intubation (vs 12% for incision) and 65% had a hemodynamic response with the criteria chosen (vs 25% for incision) with similar TCs. The entropy response was quite similar, the max RE after stimulation being 15 U higher in responders versus nonresponders for both intubation and incision. In other words, entropy parameters reacted when patients showed a motor response, but the amplitude of the entropy response might not be correlated to the stimulation or to the probability of clinical response. Nevertheless, RE was, in all studies, the parameter response that was the most significant and clinically visible.
Even in the 32–47 Hz frequency band, this entropy response observed after noxious stimulation may not only be related to EMG but also to fast frequencies in the EEG, called 
waves.17 However, waves have only been correlated to waking up or near waking up processes,18 but not with response to stimulation, except when analgesia was provided by ketamine.19 The lack of entropy response to noxious stimulation in fully paralyzed patients, observed in our study, as in a article by Hans et al.,20 rather suggests that this entropy response is purely EMG related because waves should not be influenced by NMB. Consequently, entropy is useless for reflecting a response to noxious stimulations in fully paralyzed patients but might be reliable in incomplete blockade (as during recovery from a single bolus), because frontal muscles are among the most resistant muscles and EMG response might precede a motor response and help to prevent it.9,21
This EMG-related mechanism is supported by Liu et al.22 who found that a NMB bolus without noxious stimulation decreased entropy values. This result was not observed in our study where entropy values in nonmovers without NMB were not lower than the values with NMB. This discrepancy might have been due to a very light hypnotic component of anesthesia in Liu et al.'s study (propofol 2.7 µg/mL) inducing a near awakening state with EMG activation.23
The second original result of our study was the lack of correlation between entropy changes and hemodynamic response to tracheal intubation, observed in both paralyzed and nonparalyzed patients. The relationship between EEG changes and hemodynamic response to noxious stimulation is still controversial. Entropy has been unable to discriminate if hemodynamic stability was obtained by adding an opioid or by a cardiovascular treatment such as with esmolol24 or dexmedetomidine.14 Similarly, BIS has been described to increase with hemodynamic responses in several studies,3,25 but some others did not reproduce this result.26,27 Our results suggest that, conversely with motor response, a hemodynamic response is not reliably associated with EMG activation and therefore is poorly captured by entropy parameters. Hemodynamic responses should be monitored and treated directly without using surrogate measures such as EEG-derived parameters.
The last interesting result of our study was that both RE and SE, but not (RE – SE), were significantly higher after LOC and before administering remifentanil in the patients who were about to move. As the propofol TC was similar (4.8 ± 0.9 µg/mL in nonmovers vs 5.1 ± 0.9 in movers), this result suggests that the movers were relatively resistant to propofol and might have required higher targets to achieve the same level of hypnosis. In our study, the propofol TC was adjusted to ensure clinical LOC, which was achieved in all patients with targets similar to the usually recommended values.3,28 No further adjustment was made on entropy values because the difference between movers and nonmovers was a retrospective statistical finding, with individual entropy values being in the recommended range for all patients (SE: 39 ± 11, RE: 42 ± 12, mean ± sd). The response to stimulation was expected to be mainly prevented by the opioid, which was confirmed by our results (Table 2).
However, a response to noxious stimulation is influenced by both hypnotic and opioid concentrations, with a synergistic interaction.29–31 For the same amount of opioid (as in our study where the remifentanil target was randomized), a lighter hypnotic component of anesthesia, detected by higher entropy values, might increase the risk of the clinical response to stimulation. In the literature, the data examining this assumption are conflicting. Some studies found lower entropy32 or BIS values33,34 in nonmovers compared with movers whereas some others did not find any difference before stimulation.3,10 One might notice that all investigations who found no predictive values have used relevant concentrations of opioids, whereas the investigators who found a difference used no or very low opioid concentrations.
Our results suggest that a deep hypnotic concentration, inducing low entropy values, is necessary to prevent a response to stimulation with no or few opioids, but becomes unreliable in the presence of high opioid concentrations because EEG parameters are poorly sensitive to the opioid concentration. This poor EEG performance may be due to several factors.
Remifentanil, as all µ agonists, induces dose-dependant EEG changes at very high concentrations (C50 using spectral analysis: 19 ng/mL).35 But much lower concentrations (4–8 ng/mL) are sufficient to induce analgesic3 or even sedative36 clinical effects without EEG change.
Our results suggest that this gap between opioid concentrations inducing clinical versus EEG effects is also true with entropy, and support the option of adjusting propofol depending on clinical effects.
The poor correlation between remifentanil concentrations and entropy values may also be explained by the fact that opioids' analgesic effects are not limited to the brain cortex, but also work also on deep brain structures and at the spinal cord level.37–39 As these structures have different opioid receptor content and opioids distribute to the central nervous system with different kinetics than to the brain cortex, changes in electrical activity are unlikely to reflect the opioid's effect.
In conclusion, the motor response to a noxious stimulation could be detected by EMG-mediated spectral entropy increase predominating in RE. Both movement and EMG are blocked by appropriate NMB but are favored, in unparalyzed patients, by a light hypnotic component of anesthesia, which induced high entropy values before the stimulation and should be corrected.
The hemodynamic response, defined in our study by MBP or HR changes after starting the stimulation, is not associated with EMG activation or entropy changes. As a consequence, entropy parameters are useless for predicting the hemodynamic response to noxious stimulations in patients receiving a balanced anesthesia with propofol and opioid.
|
Footnotes |
|---|
Financial support and equipment for data collection for this study were provided by Datex-Ohmeda, Helsinki, Finland.
Presented in part at the ASA Annual Meeting in Atlanta, October 2005.
|
REFERENCES |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Kawaguchi, I. Takamatsu, K. Masui, and T. Kazama Effect of landiolol on bispectral index and spectral entropy responses to tracheal intubation during propofol anaesthesia Br. J. Anaesth., August 1, 2008; 101(2): 273 - 278. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
