This randomized controlled trial describes automated coadministration of propofol and remifentanil, guided by M-Entropy analysis of the electroencephalogram. The authors tested the hypothesis that a novel dual-loop controller with an M-Entropy monitor increases time spent within predetermined target entropy ranges.
Patients scheduled for elective surgery were randomly assigned in this single-blind study using a computer-generated list, to either dual-loop control using a proportional-integral-derivative controller or skilled manual control of propofol and remifentanil using target-controlled-infusion systems. In each group, propofol and remifentanil administration was titrated to a state entropy target of 50 and was subsequently targeted to values between 40 and 60. The primary outcome was the global score, which included the percentage of state entropy or response entropy in the range 40-60, the median absolute performance error and wobble. Data are presented as medians [interquartile range].
Thirty patients assigned to the dual-loop group and 31 assigned to the manual group completed the study. The dual-loop controller was able to provide induction and maintenance for all patients. The Global Score of State Entropy was better maintained with dual-loop than manual control (25 [19-53] vs. 44 [25-110], P = 0.043), and state entropy was more frequently maintained in the range of 40-60 (80 [60-85] vs. 60 [35-82]%, P = 0.046). Propofol (4.1 [2.9-4.9] vs. 4.5 [3.4-6.3] mg · kg(-1) · h(-1)) and remifentanil (0.18 [0.13-0.24] vs. 0.19 [0.15-0.26] μg · kg(-1) · min(-1)) consumptions and the incidence of somatic side effects were similar.
Intraoperative automated control of hypnosis and analgesia guided by M-Entropy is clinically feasible and more precise than skilled manual control.
What We Already Know about This Topic
The use of electroencephalography to guide anesthetic administration remains controversial
M-Entropy analysis provides dual analyses with differential sensitivities to hypnosis and analgesia
What This Article Tells Us That Is New
Intraoperative automated propofol-remifentanil anesthesia guided by M-Entropy is clinically feasible and more precise than skilled manual control
IN 1950, Mayo et al. reported that electrocortical activity could be used to automatically titrate ether administration.1Subsequently, various signals including the median frequency of the electroencephalogram power spectrum2and auditory evoked potentials3have been used to guide automated administration of methohexital,2propofol,3,4or alfentanil.5Currently, the Bispectral Index Monitor (BIS Monitor, Aspect Medical Systems, Newton, MA) is the best-studied electroencephalographic monitor for the single feedback control of propofol6,–,19or remifentanil.20
M-Entropy analysis of the electroencephalogram (GE Healthcare, Helsinki, Finland) calculates two parameters. The first is state entropy (SE), a measure of the irregularity of frontal electroencephalogram activity within the frequency range of 0.8–32 Hz. SE is a surrogate measure of hypnotic depth and can identify varying levels of sedation21and predict volunteers' ability to recall auditory stimuli.22The second parameter, response entropy (RE), measures electroencephalogram and facial muscle electromyogram activity with a frequency range of 0.8–47 Hz.23,24The difference between RE and SE represents frequencies within the band between 32 and 47 Hz, corresponding to facial muscle electromyogram activity. During painful stimuli, an increase in RE above SE facial muscle activation may thus indicate that analgesia is inadequate25because facial muscle activity is a surrogate measure of depth of antinociception.26The fact that M-Entropy analysis provides distinct indications of hypnotic and analgesic status makes it a potential output for a previously developed dual closed-loop proportional-integral-derivative controller.20
Therefore, we used our dual-loop system to control administration of propofol and remifentanil using M-Entropy SE and RE. We compared manual versus closed-loop control during general anesthesia induction and maintenance. Specifically, we tested the hypothesis that the dual-loop controller with M-Entropy as output increases time spent within predetermined target SE and RE ranges.
Materials and Methods
Our prospective, randomized, single-blind clinical trial was approved by the Ethics Committee of our university (Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale, Hôpital A. Paré, N°06 06 59, Boulogne Billancourt, France) and the relevant French regulatory office (Agence Française de Sécurité Sanitaire des Produits de Santé). It was registered with ClinicalTrials.gov, file number NCT00391885. The study was conducted in a single university hospital (Hôpital Foch, Suresnes France). Patients were informed of the nature of the study and gave their written informed consent during the preoperative visit performed by the investigators. Inclusion criteria were patients age 18–80 yr presenting for elective surgical procedures (lung, vascular, orthopedic, gynecologic, urologic, and otolaryngologic). Exclusion criteria were combined regional/general anesthesia, a history of psychiatric and/or neurologic disorder, presence of a pacemaker, or planned intracranial surgery.
Protocol
An intravenous cannula dedicated to propofol and remifentanil infusion pumps was connected via a three-way Smartsite® Needle-Free System (Alaris Medical Systems, San Diego, CA) with a priming volume of 0.3 ml to the pumps. Routine monitoring included core temperature and neuromuscular function at the adductor pollicis. The Entropy electrode was positioned on the patient's forehead and connected to the M-Entropy Module (GE-Healthcare, Helsinki, Finland). Just before induction of general anesthesia, patients were randomly assigned to the dual-loop controller or to skilled manual control based on the M-Entropy Module. The sequence of treatments was determined in blocks of 10 (five dual-loop and five manual group) using a random number computer-generator in a 1:1 ratio. Assignments were kept in sequentially numbered opaque envelopes until just before surgery. Investigators had considerable clinical experience titrating intravenous anesthesia using the spectral entropy monitor and target-controlled infusion systems.
All patients received total intravenous anesthesia in target-controlled Infusion mode using the population pharmacokinetic sets of Schnider et al .27for propofol and Minto et al .28for remifentanil to target the effect-site concentration. Infusion Toolbox 95 version 4.11 software29(Université Libre de Bruxelles, Brussels, Belgium) implemented in a personal computer served as a platform for calculating effect-site concentrations of propofol and remifentanil; displaying effect-site concentration estimates in real time; providing a user interface that permits entry of patients' demographic data (sex, age, weight, and height) and modifications to target concentrations; controlling the propofol and remifentanil infusion pumps (Alaris Medical, Hampshire, United Kingdom); and recording calculated effect-site drug concentrations, entropy, and hemodynamic data from an AS/5 GE-Healthcare S/5™ monitor at 5-s intervals.
In the manual group, clinicians chose propofol and remifentanil effect-site concentrations for induction according to their clinical judgment. During maintenance, anesthetic agents were adjusted to maintain a SE value as close to 50 as practical. Remifentanil was given to maintain the RE value between 40 and 60 or to avoid a difference between RE and SE more than 5. As clinically necessary, anesthesiologists could modify the drug target concentration once 95% equilibration of the effect site compartment was reached without the constraint of upper or lower limits for the two agents.
For the dual-loop group, the controller has a cascade structure with a proportional-integral-derivative algorithm associated with a target-controlled infusion device as shown in figure 1. This controller was similar to a published controller.20Users entered the patient's sex, age, weight, and height. For this study, the user could modify the minimum or the maximum concentration of propofol or remifentanil targets, or switch between dual-loop and manual control. Clinicians chose the initial propofol effect-site concentrations for induction according to their clinical judgment. The controller decides the first remifentanil concentration related to initial propofol concentration. The controller includes several main elements:
Calculation of the SEerrordifference between the set point of 50 and the actual unfiltered SE value: it allows the titration until the target level of SE = 50 is obtained.
The “error” size, which determines which drug will be modified. If the SEerroris small, only the remifentanil is changed, and if the SEerroris higher than a threshold, the two drug concentrations are changed.
A proportional correction has been determined for each drug and for each SEerror. The controller continuously modifies the target concentration until a SEerror= 0 is obtained. The new target is modulated by the use of this accumulated error and provides the integral action of the controller.
Delay between each new modification of propofol or remifentanil concentration: it is determined by the time necessary for equilibration of the previous effect site compartment given by the pharmacokinetic models27,28(fig. 1).
A derivative term of the controller: check the profile every 5 s and decide on a rapid concentration correction.
Interaction rule between propofol and remifentanil: if the controller increases the remifentanil concentration successively more than three times then the controller increases the propofol concentration.
Safety feature: the system automatically maintains the calculated drug concentrations in the case of controller or Entropy dysfunction or low signal quality index less than 50%.
In both groups, the induction phase was defined from the start of propofol and remifentanil administration to SE <60 for 30 s; subsequent times until the end of surgery were considered to be the maintenance phase. Neuromuscular blockers were given to facilitate tracheal intubation followed by continuous or bolus administrations at the discretion of the anesthesiologist during the maintenance phase. Hemodynamic modifications were managed by administration of fluids, vasopressors, or antihypertensives.
Patients were ventilated with an air/oxygen mixture without nitrous oxide. Cardiovascular management, premedication, duration of anesthesia, blood loss, and somatic events (movements, grimacing, eye opening) were recorded. Surgery was classified as minor or major. Approximately 45 min before the scheduled end of surgery, intravenous analgesics were given to provide postoperative pain relief: morphine 0.05–0.15 mg/kg−1, proparacetamol (1 g), nefopam (20 mg), and nonsteroidal antiinflammatory drugs at the discretion of the clinician. A neuromuscular antagonist was given if indicated. At the end of surgery, propofol and remifentanil infusions were stopped. Time to tracheal extubation was defined as the time from discontinuation of the infusions until tracheal extubation. All patients were visited and interviewed about intraoperative recall in the postanesthesia care unit and on the second or third postoperative day.30
The primary outcome was the Global Score11(SEGlobal Score), which characterized the overall performance of the controller including the percentage of adequate anesthesia, defined as SE between 40 and 60 (SE40–60) and oscillation of the SE determined by the median absolute performance error and the wobble.31The controller performance and the parameters were calculated according to the following equations:
The performance error, or PE, calculated as the difference between actual and desired values (set point): PEij =((SEmeasuredij– SEset point)/SEset point) × 100; the bias or median performance error (MDPE): MDPEi= median [PEij,j = 1, …., Ni]; the inaccuracy or median absolute performance error (MDAPE): MDAPEi= median [|PEij|, j = 1, …., Ni]; the wobble, which measures the intraindividual variability in performance error: Wobblei= [|Peij− MDPEi|, j = 1, …., Ni], where: i = subject number, j = jth(one) measurement of observation period, n = total number of measurements during observation period. The Global Score was calculated according to the following equation:11SEGlobal Score= (MDAPE + Wobble)/SE40–60.
Excellent controller performance is characterized by a low median absolute performance error and wobble and high percentage of SE value in the range 40 and 60. Consequently, a low Global Score indicated superior performance.
Secondary outcomes included the percentage of adequate anesthesia, overshoot (SE<40) and undershoot (SE>60) periods, occurrence of suppression ratio (SR) defined as SR more than 10% lasting at least 1 min, Varvel parameters (PE, median performance error, median absolute performance error, wobble).31Secondary outcomes included also clinical data: drug consumption, number of somatic events (i.e. , movements, grimacing), time to tracheal extubation (i.e. , time from discontinuation of propofol and remifentanil infusions until extubation), and recall of intraoperative events as determined by a standardized interview performed in the postanesthesia care unit and on the second or third postoperative day.30
Data Analysis
A preliminary study indicated a mean Global Score for SE was roughly 50 ± 19 using manual propofol and remifentanil target-controlled infusion; we expected an improvement of at least 30% with our dual-loop controller. Based on these values, we estimated that a total of 54 patients (27 per group) would provide an 80% power at 5% two-sided type I error. We thus planned to recruit 70 patients to allow for dropouts. This trial was not overseen by an independent data safety monitoring board.
Data are presented as medians [interquartile ranges], percentages, or number of cases. Continuous data were compared by Mann–Whitney U tests. Categorical data were compared with Fisher exact tests. Difference in heart rate, systolic blood pressure, SE, and RE-SE were evaluated with repeated-measures ANOVA. Speed to tracheal extubation was compared using a Kaplan–Meier survival analysis, followed by a log-rank test. Significance was defined by P values less than 0.05 using a two-tailed test. Data analysis was performed using SPSS version 11.0 (SPSS Science Inc., Chicago, Il).
Results
Seventy patients were recruited between November 2006 and April 2007. Thirty-one patients in the manual group and 30 in the dual-loop group completed the study with useable data (fig. 2). Baseline characteristics were similar in the two groups (table 1). More than a third of the subjects took at least one cardiovascular medication preoperatively. One investigator (F.B.-L.) managed 80% of the patients in the manual group; N.L. and T.C. managed the others. The median SE and RE values from induction to discontinuation of propofol and remifentanil infusion are presented in figure 3.
Anesthetic induction was similar with dual-loop and manual induction (table 2). The dual-loop system maintained anesthesia for a total of 68 h, during which time 1,250 target modifications were made automatically (519 for the propofol and 731 for the remifentanil). In the dual-loop patients, no manual modifications were made. However, the remifentanil upper limit was increased (the maximum default concentration was 12 ng/ml−1) for two patients and the propofol lower limit was decreased in three patients (the minimum default concentration was 1.3 μg/ml−1). Time to tracheal extubation was comparable in the two groups (table 3).
The fraction of time during which SE was adequate, SE40–60, was significantly greater in the Dual-loop group 80%[60–85] than in the Manual group 60%[35–82]. Periods of excessive anesthesia (SE<40) were shorter in the Dual-loop group, and the Global Score of SE was better (fig. 3and table 4). In contrast, RE responses were similar in each group (fig. 3and table 5). The occurrence of burst suppression ratio (table 4) and the amount administered and the median effect site-concentrations of propofol and remifentanil were similar between the two groups (fig. 3and table 4).
Heart rate, noninvasive systolic blood pressure, SE, and RE-SE changes in response to stimulating events such as laryngoscopy or skin incision were similar in each group (fig. 4). There were no significant differences in the incidence of somatic events or in treatment of hypotension or hypertension (table 3). No cases of awareness with recall were detected.
Discussion
Despite numerous potential advantages of automated drug delivery, no fully automated system is currently approved or routinely used in anesthesia.32In addition, the only commercial automated propofol administration device nearing Food and Drug Administration approval was designed for endoscopy with sedation provided by nonanesthesiologists.33
Numerous control algorithms have nonetheless been proposed, each using a single output to control administration of a single drug. For example, more than 400 patients have been anesthetized since 1998 with a closed-loop controller using the BIS Monitor as output.6,–,20In contrast, fully automated coadministration of a hypnotic and an opioid, such as in our study, remains rare. A closed-loop controller was described with the BIS Monitor and a fixed mixture of propofol and alfentanil.7Use of an isoboller controller has also been described, in which propofol and fentanyl administration was based on the midlatency auditory evoked potential and a fuzzy-logic algorithm that combined heart rate and mean arterial pressure.34
The dual-loop controller we used with M-Entropy is an extension of the controller developed in a previous study that used BIS as the output.20Propofol and remifentanil administration was based on the assumption that intraoperative BIS fluctuation is related to noxious stimuli intensity changes. The controller thus changed remifentanil more frequently than propofol, depending on the difference between the measured BIS and a designated setpoint. The trial demonstrated that the BIS can automatically guide propofol and remifentanil administration; specifically, BIS was more frequently between 40 and 60 with the dual-loop controller than with manual target control, and time to tracheal extubation was shorter.
Monitoring of anesthetic depth using M-Entropy can reduce propofol35or sevoflurane36consumption, speed emergence from general anesthesia,35guide anesthetic management in pediatric patients,37and reduce the risk of recall.22Our study demonstrates that coadministration of propofol and remifentanil can be guided by spectral M-Entropy during anesthesia induction and maintenance. Dual-loop and manual control performed similarly during induction of general anesthesia, but the controller improved the global score of SE, the fraction of adequate anesthesia (i.e. , SE40–60), and simultaneously reduced the amount of excessive anesthesia (i.e. , SE<40).
We currently have closed-loop controllers that use electrocortical activity provided by M-Entropy or BIS.20A potential advantage of M-Entropy is that it provides distinct signals (RE and SE) that distinguish need for a hypnotic or opioid. A further potential advantage of M-Entropy is that it may react more quickly to rapid changes of anesthesia depth23,24and a potential drawback to BIS is the inconstant time delay between electroencephalogram change and the index calculation.38However, neither system is linear or time-invariant during rapid changes in cortical activity39and comparably distinguish anesthesia depth.21Finally, how each monitor compares during electrocautery artifacts and clinical transients remains to be determined. It will thus be of obvious interest to directly compare the performance of each system in realistic clinical situations.
Mathews et al. used feed-back control based on the RE-SE difference to titrate remifentanil in 20 patients having orthopedic surgery.26A difficulty is that the RE-SE difference or muscle facial activity25response to noxious stimuli was altered40or suppressed by the use of a neuromuscular blocking agent.41In contrast, the SE arousal response after noxious stimuli remains intact during paralysis.42In the current study, the controller adjusted the hypnotic and antinociceptive medications using the SEerror; specifically, the RE-SE difference was used to amplify the correction thus preventing neuromuscular block from limiting controller function.
A limitation of any opioid control system is that there is no specific measure of intraoperative analgesia. Determining whether remifentanil administration was adequate with our closed-loop controller was thus challenging. For example, hemodynamic changes are not specific to analgesic need, but have nonetheless been used in closed-loop systems to titrate alfentanil43and remifentanil.44In the current study, which used electroencephalogram changes to titrate the deficit of antinociception, hemodynamic characteristics before and after painful stimuli suggest acceptable hemodynamic stability. But before being widely adopted, our M-Entropy closed-loop controller should be tested clinically under physiologically challenging conditions45including hypertension, hypotension, morbid obesity, in pediatric patients, and during major surgery such as cardiac surgery16or lung transplantation.14Several reasons can be advanced to explain why a closed-loop controller outperforms manual control to maintain the SE in the desired range: the investigators were influenced by nonelectroencephalogram components such as hemodynamic (e.g ., tachycardia, hypertension, hypotension) or intraoperative events (e.g ., painful stimuli, vascular clamping, blood loss) and investigators modified the target of anesthesia depth; the investigators' vigilance decreased with the duration of surgery. The titration was very active initially, then the number of concentration adjustments decreased over time, and anesthesia was too deep in the manual group despite a decrease of calculated drug concentrations compared with the initial period of anesthesia. But, after 3 h, we had only three cases in the manual group and we could not come to a definite conclusion (fig. 3). Finally, the current study demonstrated the interest of automated technology for monotonous and repetitive tasks. The human brain is more efficient in making complex decisions.46
We make the assumption that tight control of SE or RE is preferable to greater variability; however, there is currently no evidence that tight control actually improves patient outcomes. Moreover, depth of analgesia can be evaluated by other methods such as fixed-order autoregressive moving average analysis of frontal electroencephalogram47or the Surgical State Index,48which are probably among the more specific measures of remifentanil effect. Whether these measures are comparable (or perhaps superior) outputs for closed-loop controllers remains unknown.
In conclusion, electrocortical activity, as provided by the Spectral M-Entropy monitor, is a surrogate measure of hypnosis and analgesia that allows simultaneous automated titration of propofol and remifentanil during induction and maintenance of general anesthesia. The Dual closed-loop controller outperforms manual control during maintenance of general anesthesia.