Free
Perioperative Medicine  |   April 2014
Monitoring Depth of Anesthesia Utilizing a Combination of Electroencephalographic and Standard Measures
Author Affiliations & Notes
  • Gerhard Schneider, M.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Denis Jordan, Ph.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Gerhard Schwarz, M.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Petra Bischoff, M.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Cornelis J. Kalkman, M.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Hermann Kuppe, M.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Ingrid Rundshagen, M.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Adem Omerovic, M. Sc.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Matthias Kreuzer, M.Sc.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Gudrun Stockmanns, Ph.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • Eberhard F. Kochs, M.D.
    From the Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany (G. Schneider); Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany (G. Schneider, D.J., A.O., M.K., and E.F.K.); Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria (G. Schwarz); Department of Anesthesiology, Knappschaftskrankenhaus Bochum Langendreer, Klinikum der Ruhr Universität Bochum, Bochum, Germany, and Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany (P.B.); Department of Anesthesiology F06-149, University Medical Center Utrecht, GA Utrecht, The Netherlands (C.J.K.); Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany (H.K.); Department of Anesthesiology and Operative Intensive Care, University Hospital Charité, Humboldt-University Berlin, Berlin, Germany (I.R.); Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany (Hans-Dieter Kochs, Ph.D.) (G. Stockmanns); and Institute of Pattern Recognition, Faculty of Electrical Engineering and Computer Science, University of Applied Sciences Hochschule Niederrhein, Krefeld, Germany (G. Stockmanns).
  • This article is featured in “This Month in Anesthesiology,” page 1A. Corresponding article on page 799. The first two authors contributed equally to this work.
    This article is featured in “This Month in Anesthesiology,” page 1A. Corresponding article on page 799. The first two authors contributed equally to this work.×
  • Submitted for publication March 9, 2013. Accepted for publication November 11, 2013.
    Submitted for publication March 9, 2013. Accepted for publication November 11, 2013.×
  • *Additional members of the European Multicenter EEG/AEP Anesthesia Monitoring Study Group are listed in the appendix.
    Additional members of the European Multicenter EEG/AEP Anesthesia Monitoring Study Group are listed in the appendix.×
  • Address correspondence to Dr. Kochs: Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Ismaningerstr. 22, 81675 Munich, Germany. e.f.kochs@lrz.tum.de. This article may be accessed for personal use at no charge through the Journal Web site, www.anesthesiology.org.
Article Information
Perioperative Medicine / Clinical Science / Central and Peripheral Nervous Systems
Perioperative Medicine   |   April 2014
Monitoring Depth of Anesthesia Utilizing a Combination of Electroencephalographic and Standard Measures
Anesthesiology 04 2014, Vol.120, 819-828. doi:10.1097/ALN.0000000000000151
Anesthesiology 04 2014, Vol.120, 819-828. doi:10.1097/ALN.0000000000000151
Abstract

Background:: For decades, monitoring depth of anesthesia was mainly based on unspecific effects of anesthetics, for example, blood pressure, heart rate, or drug concentrations. Today, electroencephalogram-based monitors promise a more specific assessment of the brain function. To date, most approaches were focused on a “head-to-head” comparison of either electroencephalogram- or standard parameter–based monitoring. In the current study, a multimodal indicator based on a combination of both electro encephalographic and standard anesthesia monitoring parameters is defined for quantification of “anesthesia depth.”

Methods:: Two hundred sixty-three adult patients from six European centers undergoing surgery with general anesthesia were assigned to 1 of 10 anesthetic combinations according to standards of the enrolling hospital. The anesthesia multimodal index of consciousness was developed using a data-driven approach, which maps standard monitoring and electroencephalographic parameters into an output indicator that separates different levels of anesthesia from awake to electroencephalographic burst suppression. Obtained results were compared with either a combination of standard monitoring parameters or the electroencephalogram-based bispectral index.

Results:: The anesthesia multimodal index of consciousness showed prediction probability (PK) of 0.96 (95% CI, 0.95 to 0.97) to separate different levels of anesthesia (wakefulness to burst suppression), whereas the bispectral index had significantly lower PK of 0.80 (0.76 to 0.81) at corrected threshold P value of less than 0.05. At the transition between consciousness and unconsciousness, anesthesia multimodal index of consciousness yielded a PK of 0.88 (0.85 to 0.91).

Conclusion:: A multimodal integration of both standard monitoring and electroencephalographic parameters may more precisely reflect the level of anesthesia compared with monitoring based on one of these aspects alone.

In a prospective study with 263 adult patients from 6 European centers undergoing surgery under general anesthesia, it was found that an index based on multimodal integration of electroencephalographic and standard monitoring parameters may more precisely assess depth of anesthesia than either standard monitoring or bispectral index monitoring alone.

What We Already Know about This Topic
  • Indices derived from the electroencephalogram may be useful but are not optimal for assessment of depth of anesthesia

  • It is unclear if an integration of measures derived from both standard and electroencephalogram monitoring may improve assessment of the hypnotic component of general anesthesia

What This Article Tells Us That Is New
  • In a prospective study with 263 adult patients from six European centers undergoing surgery under general anesthesia, it was found that an index based on multimodal integration of electroencephalographic and standard monitoring parameters may more precisely assess depth of anesthesia than either standard monitoring or bispectral index monitoring alone

FOR decades, adequacy of general anesthesia was estimated on the basis of drug concentrations and unspecific clinical parameters, that is, blood pressure, heart rate (HR), patient reactions, sweating, tearing, and many more. Although these physiological responses are affected by anesthesia, they may not reflect anesthetic effects on the main target organ of anesthesia, the brain. Today, an increasing number of monitors allow a more specific assessment of effects of general anesthesia on the brain. For this purpose, these monitors record and process spontaneous electrical activity of the brain (electroencephalogram) or auditory evoked potentials. On the basis of—mostly proprietary—algorithms, these devices calculate an index value. Usually this index value is a number between 0 and 100, which is inversely correlated to the hypnotic component of anesthesia or “depth” of anesthesia. These monitors provide insight into brain effects of general anesthesia, but also have limitations as they may not always adequately reflect the clinical status of a patient. Several studies showed that they may not reliably distinguish between responsive and unconscious patients.1–6  Despite of these limitations, data suggest that use of these monitors may improve clinical outcome. The Australian B-Aware trial showed that anesthetic administration guided by bispectral index (BIS) reduces the incidence of intraoperative awareness with recall in a high-risk population.7  These findings were challenged by the B-Unaware and the BAG-Recall trials. Both studies suggested that the incidence of awareness with recall is not reduced by a BIS-guided protocol when compared with a standard protocol based on a minimum end-tidal anesthetic gas concentration.8,9 
These studies compared an electroencephalogram-based index with standard monitoring parameters, that is, either unspecific anesthesia monitoring or specific (brain) monitoring only. From a clinical point of view, electroencephalogram-based monitoring should not compete with, but complete, current standard monitoring.
Therefore, the current study was designed to analyze whether a multimodal integration of standard and electroencephalographic monitoring may provide reliable information about the level of anesthesia. For this purpose, an indicator was combined from electroencephalographic and standard monitoring parameters (anesthesia multimodal index of consciousness [AMIC]). It was tested whether this indicator separates consciousness from unconsciousness at the transition of both states and indicates different levels of anesthesia from wakefulness to deep anesthesia. In addition, AMIC was compared with a model from standard parameters only and with the electroencephalogram-based BIS.
Materials and Methods
Study Overview
In a European multicenter study, electroencephalogram and a variety of standard parameters were recorded synchronously during clinical anesthesia. The AMIC was designed using a combination of standard and electroencephalographic parameters. The ability of AMIC to quantify the hypnotic component of anesthesia was assessed. AMIC was compared with an index calculated from standard monitoring parameters only. In addition, stored electroencephalographic data were replayed to the BIS® (BIS A2000; Aspect Medical Systems, Norwood, MA) monitor to calculate the BIS index, and this electroencephalogram-based approach was also analyzed.
Selection of Patients
In a total of six European centers, 263 adult patients undergoing elective surgery with general anesthesia were included. Patients were recruited from August 2004 to July 2005. There was no randomization procedure, and at the time the study was performed, no formal registration was required. Still, the study was registered later at clinicaltrials.gov. After obtaining approval from the ethics committees, patients with American Society of Anesthesiologists physical status I to III gave informed written consent. Patients with contraindications to the study drugs, a history of psychiatric or neurologic diseases, drug abuse or medication known to affect the central nervous system, pregnancy, or indication for rapid-sequence induction were excluded from the study.
Study Design and Clinical Protocol
According to the hospital standards, patients were assigned to receive 1 of 10 different anesthetic regimens, combining opioids (fentanyl, remifentanil, and sufentanil), hypnotic drugs for induction (etomidate, propofol, and thiopental), and maintenance (propofol, desflurane, isoflurane, and sevoflurane), as presented in table 1. After enrollment, one patient in group 9 (induction: etomidate; maintenance: propofol–sevoflurane) did not receive propofol, but only etomidate and sevoflurane. As per intention to treat, he was included in group 9. Each patient obtained a specific combination of opioids and hypnotic drugs according to the standard clinical practice in the according hospital. Each study center used at least two anesthetic regimens and each anesthetic regimen was used in at least two different study centers. All anesthetic and surgical interventions were recorded in a standardized manner. During induction, patients were asked every 15 s to squeeze the investigator’s hand. A response was verified by an immediate repetition to differentiate involuntary movement from a response to command. Loss of consciousness (LOC) was defined as a missing response to a repeated verbal command. When neuromuscular-blocking agents were given, Tunstall’s10  isolated forearm technique was used to preserve the patient’s ability to follow command. For this purpose, a tourniquet was inflated on the right forearm before a muscle relaxant was injected into a vein either proximal to the tourniquet or on the opposite arm. Intact neuromuscular transmission was not verified to avoid additional painful stimuli. After skin incision, anesthetic concentrations were increased until the electroencephalogram showed burst suppression (BS) with a minimum of 1 s suppression in the electroencephalogram (“silent second”). Subsequently, drug concentrations were decreased and general anesthesia was maintained according to standard clinical practice until end of surgery. At the end of surgery, drug administration was stopped and patients were asked to squeeze the investigator’s hand again. The first verified squeeze of hand marked return of consciousness (ROC). Study design and selected data points used in the analysis are shown in figure 1. Once recovered from anesthesia, patients were assessed for awareness with recall using a structured interview.11  This interview was repeated within 3 days on the ward.
Table 1.
Anesthetic Combinations for Induction and Maintenance
Anesthetic Combinations for Induction and Maintenance×
Image Not Available
Table 1.
Anesthetic Combinations for Induction and Maintenance
Anesthetic Combinations for Induction and Maintenance×
×
Fig. 1.
Study diagram showing time points loss of consciousness (LOC), skin incision, burst suppression (BS), and return of consciousness (ROC) and phases of concentration increase and decrease. We performed data analysis at selected time points reflecting (A) the transition between consciousness (phase 1, time points LOC −15 s and ROC + 30 s) and unconsciousness (phase 0, LOC + 30 s and ROC −15 s,) and (B) levels awake (phase 2, LOC −180 s, ROC + 30 s), anesthesia according to clinical practice (phase −1, skin incision −60 s, end decrease +30 s) and BS (phase −3, end increase +30 s, start decrease −60 s).
Image Not Available
Fig. 1.
Study diagram showing time points loss of consciousness (LOC), skin incision, burst suppression (BS), and return of consciousness (ROC) and phases of concentration increase and decrease. We performed data analysis at selected time points reflecting (A) the transition between consciousness (phase 1, time points LOC −15 s and ROC + 30 s) and unconsciousness (phase 0, LOC + 30 s and ROC −15 s,) and (B) levels awake (phase 2, LOC −180 s, ROC + 30 s), anesthesia according to clinical practice (phase −1, skin incision −60 s, end decrease +30 s) and BS (phase −3, end increase +30 s, start decrease −60 s).
×
Monitoring and Data Recording
Anesthesia standard monitoring parameters (noninvasive blood pressure, HR, oxygen saturation, end-tidal carbon dioxide concentration, anesthetic gas concentrations, and respiratory variables; table 2) were continuously measured with a VisiCon M211 patient monitor (BBraun, Melsungen, Germany) and stored together with relevant patient data (e.g., age, weight, and American Society of Anesthesiologists status) on disk. In addition, two electroencephalographic channels were recorded continuously and stored using an amplifier developed for intraoperative electroencephalographic recordings (total bandwidth 0.5 to 400 Hz using analog filters, sampling rate 1,000 Hz, 12-bit amplitude resolution). This device had been developed for a European multicenter study.12,13  It is based on a digital signal processing board in a personal computer and an isolated preamplifier unit. The preamplifier is powered by two AA-cells and has integrated auditory and somatosensory stimulators, automatic calibration, and impedance checking. A detailed description of the equipment has been published before.14,15  Four self-adhesive electroencephalographic/electrocardiographic electrodes (BlueSensor N; Ambu A/S, Ballerup, Denmark) were positioned at AT1, M2, Fz (reference), and Fpz (ground) according to the international 10 to 20 system. At each electrode position, the skin was degreased with alcohol to maintain electrode impedances below 5 kΩ. Automated artifact rejection was used to discard signal segments of 10 s length with amplitudes greater than 250 μV, slopes greater than 140 μV/ms, or zero-signals of defined minimal length.16  The electroencephalogram was preprocessed using a digital low-pass filter at 30 Hz (subsequent downsampling at 200 Hz) to improve signal to noise ratio and to reduce the influence of high-frequency content (with potential electromyogram influence) on the electroencephalogram. Electroencephalographic parameters were calculated from signal segments of 10 s length with a 5 s overlap. In addition, recorded electroencephalogram was replayed using a play-back device17  for off-line evaluation of the BIS.
Table 2.
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data×
Image Not Available
Table 2.
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data×
×
Selection of AMIC Variables
The AMIC design integrates variables derived from the electroencephalogram, standard monitoring parameters, individual patient data, and drug information (table 2). Electroencephalographic parameters for the AMIC model were selected on the basis of previous studies.5,16,18–20  They consist of the weighted spectral median frequency (particularly designed to separate consciousness and unconsciousness at the transition between both states),20  the approximate entropy (ApEn; nonlinear electroencephalographic characteristics, monotonic behavior during deepening of anesthesia),18,19  permutation entropy (PeEn; nonlinear electroencephalographic characteristics, robust against artifacts, nonlinear signal characteristics, reliably separates consciousness and unconsciousness),16  and suppression ratio (SuppRatio; amplitude based detection of detection of BS patterns).19,21  Selection of standard monitoring parameters and patient and drug information was based on clinical expert’s opinion. HR and mean arterial blood pressure (MAP) were selected to provide information about cardiovascular status. The additional parameters HR trend (ΔHR) and BP MP trend (ΔMAP) were calculated as the difference between two subsequent measurements. Respiratory parameters were the calculated difference between in- and expiratory oxygen concentration (Gas O2 IE), end-expiratory carbon dioxide concentration (Gas CO2 exp), peak inspiratory pressure (SPIRO Ppeak), end-expiratory gas concentration (monitored anesthesia care [MAC] exp; MAC values calculated according to Nickalls et al.22 ), differences between inspiratory and end-expiratory volatile anesthetic concentration (MAC IE; MAC values calculated according to Nickalls et al.22 ), which may contain information about the current dynamics of the anesthetic level (i.e., increase or decrease of the volatile concentration), propofol plasma concentrations according to the Marsh23  model (Cp), induction and maintenance drug (Drug Ind/Main), age, sex, and body mass index. Parameters were recorded with respect to specific sampling rates, for example, 10 s for electroencephalographic parameters, 5 min for blood pressure, and time-constant demographic patient data (table 2). Trend values ΔHR and ΔMAP are indicated by the time derivative over two consecutive measurements. Finally, different time scales were resampled using a sliding window technique to provide synchronous parameter updates every 5 s.
Design of the AMIC
An Adaptive Neuro Fuzzy Inference System24–26  was used to generate the AMIC. Standard monitoring parameters were preselected on the basis of clinical expert knowledge and electroencephalographic parameters according to preliminary analysis (table 2). Knowledge based on the current data was not used for parameter preselection for design of the AMIC model. In a stepwise approach, the model was designed to reach: A. “distinction between consciousness and unconsciousness at the transition between these conditions” and B. “recognition of wakefulness, anesthesia according to clinical practice, and excessively deep anesthesia(electroencephalographic BS).” Subsequently, the AMIC model was developed, integrating indicators for A. and B. into an index. Figure 2 shows the workflow for real-time AMIC processing from data acquisition to the two-step classification procedure.
Fig. 2.
Workflow for (real-time) computation of the anesthesia multimodal index of consciousness (AMIC). After data acquisition (including preprocessing, e.g., artifact detection), parameters (according to table 2) are extracted for each modality. Independent of the sampling rate of specific parameters (e.g., 10 s for electroencephalographic parameters and constant in case of patient age), each parameter is resampled at 5 s and synchronized in time to obtain a multivariate time series. In a first step, classification is performed in two independent models (model A trained for separation of consciousness and unconsciousness and model B trained for separation of different hypnotic states from wakefulness to burst suppression) using an Adaptive Neuro Fuzzy Inference System. Both models were trained to provide best fit of input data to paradigm A and B, leading to data reduction for further processing. In a second step, resulting indices A and B are combined by a third Adaptive Neuro Fuzzy Inference System model resulting in AMIC. ApEn = approximate entropy; BMI = body mass index; BS = burst suppression; HR = heart rate; MAC IE = monitored anesthesia care equivalent of difference between in- and end-expiratory gas concentration; MAP = mean arterial blood pressure; PeEn = permutation entropy; WSMF = weighted spectral median frequency.
Image Not Available
Fig. 2.
Workflow for (real-time) computation of the anesthesia multimodal index of consciousness (AMIC). After data acquisition (including preprocessing, e.g., artifact detection), parameters (according to table 2) are extracted for each modality. Independent of the sampling rate of specific parameters (e.g., 10 s for electroencephalographic parameters and constant in case of patient age), each parameter is resampled at 5 s and synchronized in time to obtain a multivariate time series. In a first step, classification is performed in two independent models (model A trained for separation of consciousness and unconsciousness and model B trained for separation of different hypnotic states from wakefulness to burst suppression) using an Adaptive Neuro Fuzzy Inference System. Both models were trained to provide best fit of input data to paradigm A and B, leading to data reduction for further processing. In a second step, resulting indices A and B are combined by a third Adaptive Neuro Fuzzy Inference System model resulting in AMIC. ApEn = approximate entropy; BMI = body mass index; BS = burst suppression; HR = heart rate; MAC IE = monitored anesthesia care equivalent of difference between in- and end-expiratory gas concentration; MAP = mean arterial blood pressure; PeEn = permutation entropy; WSMF = weighted spectral median frequency.
×
During the procedure of model generation, specific rules are constructed for incomplete input parameter values. Therefore, the Adaptive Neuro Fuzzy Inference System model allows the generation of valid AMIC values even if single subparameters are missing. For model generation (training and test) of the AMIC study, data were retrospectively analyzed (see Statistical Analysis).
Because it has been questioned whether addition of electroencephalographic parameters may improve standard clinical monitoring,27  it was analyzed whether a combination of standard monitoring may reach similar results. Therefore, a data-driven model was designed which only includes standard parameters. For this purpose, a basis parameter set including HR, MAP, Cp, and MAC exp was supplemented by all permutations of remaining standard parameters, drug and patient information. The data-driven model leading to best prediction performance was identified and used for comparison.
Statistical Analysis
The aim of the current study was to evaluate the ability of AMIC to separate different anesthesia states. In addition, the performance of the indicator including standard monitoring parameters (reflecting only indirect anesthetic brain effects) and of BIS (only electroencephalographic parameters) was evaluated for the same clinical questions. Therefore, indicators were calculated from data pairs before and after LOC (time points LOC −15 s, LOC +30 s) and ROC (time points ROC −15 s, ROC +30 s) to analyze their ability to separate consciousness from unconsciousness at the transition between these states (fig. 1A). The interval of 15 s in asking the patient to squeeze hand allows a clear definition of the clinical endpoints LOC/ROC, and thus allows the conclusion that 15 s before ROC (LOC) the patient was unconscious (conscious). To compensate for time delays of the electroencephalographic monitor,4  a computing time of 30 s after LOC/ROC was added. Second, differentiation of levels awake (LOC −180 s, ROC +30 s), anesthesia according to clinical practice (skin incision −60 s, end of anesthetic decrease +30 s), and BS (end of anesthetic increase +30 s, start anesthetic decrease −60 s) was performed for indicator assessment (according to fig. 1B).
Prediction probability (PK) was used to assess performance of the indicators.28,29  PK is a dimensionless number between 0 and 1, where 1 reflects a perfect separation of different clinical states, 0.5 reflects a random result, and 0 reflects a 100% disagreement between monitor and clinical status. Estimation of uncertainty was based on 95% percentile bootstrap CIs.28,30  PK was calculated at predefined time points as described above.
A three-fold cross-validation was performed for the AMIC model generation to avoid overfitting through overlapping training and test data: A number of patients as equal as possible was randomly assigned to each of the three groups (I: 90 patients, II: 88, III: 85) by keeping all anesthetic combinations in the three groups to provide training and test of the AMIC on different patients but based on analogous anesthetic procedures. The AMIC was consecutively trained on data of two groups (I + II, I + III, II + III) to identify three models leading to maximum overall PK (awake/consciousness, unconsciousness, clinical practice, BS). These models were evaluated on data of the remaining third group (III, II, I) to provide AMIC values in data of all 263 patients for final model evaluation.
Comparison of AMIC, standard parameter indicator, and of BIS was performed using percentile bootstrap CIs of PK (P < 0.05, Bonferroni correction for multiple comparisons).28  Patient characteristics and demographic data in both groups were calculated and compared with t tests and chi-square test (two-sided tests with unequal variances).
Results
Data of 263 study patients were included in the analysis, 111 were men (42%) and 152 women (58%). Age (mean ± SD) was 50.1 ± 13.7 yr, height was 172 ± 9 cm, and weight was 77 ± 16 kg. American Society of Anesthesiologists physical status was I in 101 patients, II in 115 patients, and III in 47 patients. Demographic data did not show significant differences between groups with different anesthetic regimens (table 1). No patient reported awareness with recall.
All parameters listed in table 2 were used for the AMIC. Model generation was based on a three-fold cross-validation according to the data of groups I, II, and III in table 3. The table indicates the number of patients and the percentage of anesthetic regimens included in each group. The model of an indicator which was based on standard monitoring parameters leading to a highest overall PK (awake/consciousness, unconsciousness, clinical practice/BS) included the following clinical parameters: HR, MAP, Cp, MAC exp, Gas carbon dioxide exp, SPIRO Ppeak, and ΔMAP (table 2). On the basis of standard monitoring parameters, a PK of 0.81 (0.77 to 0.84) was obtained during the transition between consciousness and unconsciousness, which is significantly lower than a PK of 0.88 (0.85 to 0.91) as obtained for the AMIC (all anesthetic combinations, P < 0.05 corrected). In contrast, separation of levels awake and clinical practice yields to equal PK of 1.00 for AMIC and for the indicator of standard monitoring parameters.
Table 3.
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group×
Image Not Available
Table 3.
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group×
×
Results of the AMIC and BIS are shown in table 4. A comparison of AMIC with BIS shows a better discrimination of clinical states for the new indicator. Figure 3 illustrates the distribution of indicator values of AMIC (fig. 3A) and BIS (fig. 3B) for selected anesthetic levels and all anesthetic combinations. In AMIC model generation, BS was used as an electroencephalographic endpoint representing deep anesthesia. The AMIC provides a similar PK for separation of levels awake/consciousness, unconsciousness, clinical practice without inclusion of data from the level BS (0.95; 0.94 to 0.98) compared with the PK from data including BS (0.96; 0.95 to 0.97).
Table 4.
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B×
Image Not Available
Table 4.
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B×
×
Fig. 3.
Indicator value distribution of the anesthesia multimodal index of consciousness (AMIC; A) and bispectral index (BIS; B) for the different anesthetic levels (central line: median, dots: 5% and 95% percentiles, whiskers: 10% and 90% percentiles, boxes: 50% quartile). Analyzed data consist of 742 data points for awake and consciousness, 526 data points for unconsciousness, 336 data points for anesthesia according to clinical practice, and 335 data points for burst suppression (BS). Horizontal dashed lines indicate optimum threshold (based on maximum sum of sensitivity and specificity) for separation of consciousness versus unconsciousness (AMIC: 80, BIS: 80) and for separation of awake/consciousness versus unconsciousness/clinical practice/BS (AMIC: 79, BIS 72).
Image Not Available
Fig. 3.
Indicator value distribution of the anesthesia multimodal index of consciousness (AMIC; A) and bispectral index (BIS; B) for the different anesthetic levels (central line: median, dots: 5% and 95% percentiles, whiskers: 10% and 90% percentiles, boxes: 50% quartile). Analyzed data consist of 742 data points for awake and consciousness, 526 data points for unconsciousness, 336 data points for anesthesia according to clinical practice, and 335 data points for burst suppression (BS). Horizontal dashed lines indicate optimum threshold (based on maximum sum of sensitivity and specificity) for separation of consciousness versus unconsciousness (AMIC: 80, BIS: 80) and for separation of awake/consciousness versus unconsciousness/clinical practice/BS (AMIC: 79, BIS 72).
×
In these analyses, AMIC was reconstructed in all data of the 263 patients based on three independent models obtained from the cross-validation. The individual models which were trained on data of two groups (I + II, I + III, II + III) and tested on data of the groups III, II, I lead to overall PK ranging between 0.94 and 0.96, which is comparable with the PK of 0.96 of the model used in evaluation of all 263 patients (separation of all levels).
Discussion
The current study shows that a combination of standard monitoring and electroencephalographic parameters can distinguish different levels of anesthesia better than either standard monitoring or electroencephalogram-based BIS monitoring alone.
For this purpose, the AMIC was developed and evaluated (three-fold cross-validation) on the basis of data of a European multicenter study. The AMIC combines “state-of-the-art” electroencephalographic parameters, standard monitoring, drug and patient information. This approach was chosen to combine the additional information from the electroencephalogram with standard monitoring practice rather than to perform a “head-to-head” comparison between two approaches. In our point of view, this reflects a clinical approach to electroencephalogram-based monitoring of anesthesia. Compared with standard (nonelectroencephalographic) monitoring, information about anesthetic effects on the main target organ of anesthesia, the brain, is added. It seems unlikely that this additional information can replace (unspecific) standard monitoring. Therefore, we chose a monitoring approach which reflects the aspect that information about cortical effects is added. This point of view is supported by the results at the transition between consciousness and unconsciousness, where PK = 0.88 of the AMIC (electroencephalographic and standard monitoring parameters) is significantly higher than the PK = 0.81 of the standard parameter model alone.
The study was designed to test whether the combined indicator reliably separates consciousness from unconsciousness and indicates increase and decrease of the anesthetic level over the entire range from light sedation to deep anesthesia.
As a challenging approach, the transition from consciousness to unconsciousness and vice versa was used to assess the anesthesia indices. (Un)consciousness was defined as (un)responsiveness to command. This approach may have limitations, for example, it requires an active response of the patient.31  Still, it provides one of the few clinical endpoints with a marked change in patient status. In clinical anesthesia, LOC represents only one endpoint of the entire range of general anesthesia. “Deeper” anesthesia can be achieved with increasing drug concentrations. This “deepening” of the level of anesthesia may not only be reflected by cortical effects.32  As a consequence, AMIC includes both electroencephalographic and nonelectroencephalographic parameters.
On the basis of the data of the current study, the AMIC outperforms both the “electroencephalographic-only” BIS and the standard monitoring parameters approach. Several paradigms were used to assess the AMIC, with a focus on the dynamic transition between consciousness and unconsciousness at LOC and ROC.
The patient individual model design used here includes data from different sources: target electroencephalogram with high temporal resolution, nonspecific standard monitoring data which provide additional trend information with lower time resolution, and individual patient data and drug protocols. The indicator AMIC integrates these modalities by sophisticated multiscale data-driven fuzzy inference including “state-of-the-art” parameters for quantifying anesthetic-induced effects in the electroencephalogram. The proposed AMIC approach additionally stabilizes the integration of multimodal and multiscalar data through the “bridging” of missing values that may occur during clinical anesthesia, for example, as a result of artifacts or delayed data acquisition. This is reached by a flexible adaptation of the model’s built-in set of rules, weighting each integrated parameter individually. Thus, an AMIC value is calculated even if the number of input variables is incomplete.
Results show significant advantages of this multimodal approach. In particular, at the transition between consciousness and unconsciousness, the AMIC provides PK of 0.88, which is significantly higher than PK of 0.74 for BIS (corrected threshold P < 0.05). In the detection of anesthetic-induced transitions from wakefulness to BS (including onset, respectively fading of unconsciousness and changes in anesthetic levels according to clinical practice), we found a PK of 0.96 for the AMIC, which is significantly higher than PK of BIS (0.80) at corrected thresholds P value of less than 0.05. This indicates that electroencephalographic information alone only reflects one aspect of the clinical status of the patient. However, in a calculation scheme without electroencephalographic information, that is, a combination from standard monitoring parameters, the indicator reaches a lower PK value than the AMIC (incorporating both, electroencephalographic and physiologic measures). In concordance with previous findings, this shows that monitoring on the basis of standard clinical practice may also result in an incomplete picture. Our findings are supported by results of the BAG-Recall trial, where the electroencephalogram-based parameter BIS does not outperform an approach based on a standard monitoring parameter.
The AMIC fundamentally differs from existing electroencephalographic monitors of the hypnotic component. It can be considered as an expert system that delivers probabilistic decision support regarding depth of anesthesia. Accurate detection of consciousness and unconsciousness at the transition between both states indicates that the multimodal indicator benefits from analysis methods allowing short periods of 10 s for adequate electroencephalographic signal processing: On the basis of previous studies, a generalized spectral parameter (weighted spectral median frequency) has been developed to provide reliable separation of consciousness and unconsciousness at the transition between both states.20  Additional information that may be related to the dynamics of the cerebral cortex is reflected by nonlinear electroencephalographic characteristics. Therefore, the nonlinear parameter entropy seems to be an appropriate approach to quantify the hypnotic component of anesthesia, because LOC is related to a change in information processing in neuronal networks.33,34  The AMIC includes the ordinal approach of PeEn quantifying characteristics of the electroencephalogram related to the information content of the signal.16,35  It allows reliable estimates using electroencephalographic signals of a few seconds length ensuring very short time delays for indicator calculation.
If frequencies of the electroencephalographic γ-band are included in index calculation, detection of consciousness may also be affected by frontal muscle activity. In particular with frontal electrodes, there is an overlap in frequency ranges of electroencephalogram and electromyogram, with usually higher amplitudes of frontal electromyographic activity.21,36  The AMIC algorithm is independent from the electroencephalographic γ-band, which improves signal to noise ratio of target electroencephalographic analysis,37  that is, it reduces the influence of overlapping frontal muscle electromyogram to frontal electroencephalogram. Methods of PeEn, ApEn, and weighted spectral median frequency have been integrated into the model. In previous studies, PeEn and weighted spectral median frequency provided significantly higher discrimination power in separating consciousness from unconsciousness at the transition between both states compared with the discrimination power provided by BIS, even if only electroencephalographic frequencies up to 30 Hz are included in the calculation.16,20  In contrast, ApEn reflects changes in low-dimensional dynamical processes, which can be used for reliable indication of the hypnotic state from light to deep anesthesia with electroencephalographic BS.19 
There are some potential limitations of the current study: Even if the evaluation of the AMIC was based on a cross-validation, the comparison of AMIC and BIS may be biased in favor of the AMIC, because data from the same clinical study were used for the AMIC model generation, but of course unknown to BIS. Nevertheless, results suggest that overfitting to the data represents a minor effect, as we obtained similar PK of 0.94 to 0.96 for AMIC evaluation based on the three separate test data sets and a PK of 0.96 for evaluation performed on the whole data (separation of levels awake/consciousness, unconsciousness, clinical practice, BS). An unbiased analysis of AMIC and other traditionally used monitoring paradigms should be based on independent data from a prospective evaluation study.
AMIC was designed to separate unconsciousness from wakefulness, and to reflect the entire range from awake to deep anesthesia. For this purpose, “deep anesthesia” was defined as electroencephalographic BS with a minimum of 1 s suppression in the electroencephalogram (“silent second”). This may induce a bias toward electroencephalogram in the selection of parameters. This potential limitation of this approach was accepted, because it allows a definition of “deep anesthesia” based on effects in the main target organ of anesthesia, the brain, whereas other clinically used parameters, for example, bradycardia, hypotension, would focus on side effects rather than the main effect of anesthetics. PK obtained from data with BS (PK = 0.96) or without BS (PK = 0.95) were similar, indicating that inclusion of BS in the model generation does not strongly bias the behavior of the AMIC.
In the current study, there were no data from patients awake during neuromuscular blockade. Therefore, it remains unclear whether such a patient would be detected by AMIC. The main effect of neuromuscular blockade is a reduction of muscle activity in the electroencephalographic signal. Muscle activity is mainly prominent in the range of high frequencies. Therefore, a selection of electroencephalographic parameters which are not based on high-frequency components of the signal for the AMIC suggests that detection of awake patients should also be reached under neuromuscular blockade.
Furthermore, unevenly distributed sample sizes and different surgical procedures for included anesthetic combinations may bias results if the drug protocol has an influence on indicator values. Therefore, additional studies are required to validate results of the current study.
In summary, AMIC combines traditional parameters used to assess the hypnotic component of anesthesia with a measurement from the main target organ, the brain. This combination allows a better separation of different anesthetic levels than standard monitoring parameters or electroencephalographic monitors alone.
Acknowledgments
Supported by departmental sources and a grant (POM4) from B. Braun AG, Melsungen, Germany.
Competing Interests
E. F. Kochs, G. Schneider, D. Jordan, A. Omerovic, and M. Kreuzer have applied for a European (11178111.8; filing date: August 19, 2011, priority date: June 17, 2011) and international (PCT/EP2012/002 158; filing date: May 21, 2012) patent. All other authors declare no competing interests.
References
Schneider, G, Kochs, EF, Horn, B, Kreuzer, M, Ningler, M Narcotrend does not adequately detect the transition between awareness and unconsciousness in surgical patients.. Anesthesiology. (2004). 101 1105–11 [Article] [PubMed]
Hoymork, SC, Hval, K, Jensen, EW, Raeder, J Can the cerebral state monitor replace the bispectral index in monitoring hypnotic effect during propofol/remifentanil anaesthesia?. Acta Anaesthesiol Scand. (2007). 51 210–6 [Article] [PubMed]
Messner, M, Beese, U, Romstöck, J, Dinkel, M, Tschaikowsky, K The bispectral index declines during neuromuscular block in fully awake persons.. Anesth Analg. (2003). 97 488–91 [Article] [PubMed]
Pilge, S, Zanner, R, Schneider, G, Blum, J, Kreuzer, M, Kochs, EF Time delay of index calculation: Analysis of cerebral state, bispectral, and narcotrend indices.. Anesthesiology. (2006). 104 488–94 [Article] [PubMed]
Schneider, G, Hollweck, R, Ningler, M, Stockmanns, G, Kochs, EF Detection of consciousness by electroencephalogram and auditory evoked potentials.. Anesthesiology. (2005). 103 934–43 [Article] [PubMed]
Zanner, R, Pilge, S, Kochs, EF, Kreuzer, M, Schneider, G Time delay of electroencephalogram index calculation: Analysis of cerebral state, bispectral, and Narcotrend indices using perioperatively recorded electroencephalographic signals.. Br J Anaesth. (2009). 103 394–9 [Article] [PubMed]
Myles, PS, Leslie, K, McNeil, J, Forbes, A, Chan, MT Bispectral index monitoring to prevent awareness during anaesthesia: The B-Aware randomised controlled trial.. Lancet. (2004). 363 1757–63 [Article] [PubMed]
Avidan, MS, Zhang, L, Burnside, BA, Finkel, KJ, Searleman, AC, Selvidge, JA, Saager, L, Turner, MS, Rao, S, Bottros, M, Hantler, C, Jacobsohn, E, Evers, AS Anesthesia awareness and the bispectral index.. N Engl J Med. (2008). 358 1097–108 [Article] [PubMed]
Avidan, MS, Jacobsohn, E, Glick, D, Burnside, BA, Zhang, L, Villafranca, A, Karl, L, Kamal, S, Torres, B, O’Connor, M, Evers, AS, Gradwohl, S, Lin, N, Palanca, BJ, Mashour, GA BAG-RECALL Research Group, Prevention of intraoperative awareness in a high-risk surgical population.. N Engl J Med. (2011). 365 591–600 [Article] [PubMed]
Tunstall, ME Detecting wakefulness during general anaesthesia for caesarean section.. Br Med J. (1977). 1 1321 [Article] [PubMed]
Brice, DD, Hetherington, RR, Utting, JE A simple study of awareness and dreaming during anaesthesia.. Br J Anaesth. (1970). 42 535–42 [Article] [PubMed]
Kochs, E, Kalkman, CJ, Thornton, C, Newton, D, Bischoff, P, Kuppe, H, Abke, J, Konecny, E, Nahm, W, Stockmanns, G Middle latency auditory evoked responses and electroencephalographic derived variables do not predict movement to noxious stimulation during 1 minimum alveolar anesthetic concentration isoflurane/nitrous oxide anesthesia.. Anesth Analg. (1999). 88 1412–7 [PubMed]
Kochs, E, Stockmanns, G, Thornton, C, Nahm, W, Kalkman, CJ Wavelet analysis of middle latency auditory evoked responses: Calculation of an index for detection of awareness during propofol administration.. Anesthesiology. (2001). 95 1141–50 [Article] [PubMed]
Jordan, C, Weller, C, Thornton, C, Newton, DE Monitoring evoked potentials during surgery to assess the level of anaesthesia.. J Med Eng Technol. (1995). 19 77–9 [Article] [PubMed]
Stockmanns, G, Ningler, M, Omerovic, A, Kochs, E NeuMonD: A tool for the development of new indicators of anaesthetic effect.. Biomed Tech. (2007). 52 96–101 [Article]
Jordan, D, Stockmanns, G, Kochs, EF, Pilge, S, Schneider, G Electroencephalographic order pattern analysis for the separation of consciousness and unconsciousness: An analysis of approximate entropy, permutation entropy, recurrence rate, and phase coupling of order recurrence plots.. Anesthesiology. (2008). 109 1014–22 [Article] [PubMed]
Kreuzer, M, Kochs, EF, Pilge, S, Stockmanns, G, Schneider, G Construction of the electroencephalogram player: A device to present electroencephalogram data to electroencephalogram-based anesthesia monitors.. Anesth Analg. (2007). 104 135–9 [Article] [PubMed]
Bruhn, J, Röpcke, H, Hoeft, A Approximate entropy as an electroencephalographic measure of anesthetic drug effect during desflurane anesthesia.. Anesthesiology. (2000). 92 715–26 [Article] [PubMed]
Horn, B, Pilge, S, Kochs, EF, Stockmanns, G, Hock, A, Schneider, G A combination of electroencephalogram and auditory evoked potentials separates different levels of anesthesia in volunteers.. Anesth Analg. (2009). 108 1512–21 [Article] [PubMed]
Jordan, D, Stockmanns, G, Kochs, EF, Schneider, G Median frequency revisited: An approach to improve a classic spectral electroencephalographic parameter for the separation of consciousness from unconsciousness.. Anesthesiology. (2007). 107 397–405 [Article] [PubMed]
Rampil, IJ A primer for EEG signal processing in anesthesia.. Anesthesiology. (1998). 89 980–1002 [Article] [PubMed]
Nickalls, RW, Mapleson, WW Age-related iso-MAC charts for isoflurane, sevoflurane and desflurane in man.. Br J Anaesth. (2003). 91 170–4 [Article] [PubMed]
Marsh, B, White, M, Morton, N, Kenny, GN Pharmacokinetic model driven infusion of propofol in children.. Br J Anaesth. (1991). 67 41–8 [Article] [PubMed]
Mamdani, EH, Assilian, S An experiment in linguistic synthesis with a fuzzy logic controller.. Int J Man Mach Stud. (1975). 7 1–13 [Article]
Takagi, T, Sugeno, M Fuzzy identification of systems and its application to modeling and control.. IEEE Trans Syst Man Cybern. (1985). 15 116–32 [Article]
Jang, J-SR ANFIS: Adaptive-network-based fuzzy inference systems.. IEEE Trans Syst Man Cybern. (1993). 23 665–84 [Article]
Avidan, MS, Mashour, GA Prevention of intraoperative awareness with explicit recall: Making sense of the evidence.. Anesthesiology. (2013). 118 449–56 [Article] [PubMed]
Jordan, D, Steiner, M, Kochs, EF, Schneider, G A program for computing the prediction probability and the related receiver operating characteristic graph.. Anesth Analg. (2010). 111 1416–21 [Article] [PubMed]
Smith, WD, Dutton, RC, Smith, NT Measuring the performance of anesthetic depth indicators.. Anesthesiology. (1996). 84 38–51 [Article] [PubMed]
Davison, AC, Hinkley, DV Bootstrap methods and their application.. Cambridge Series in Statistical and Probabilistic Mathematics. (1997). Cambridge Cambridge University Press 191–55
Sanders, RD, Tononi, G, Laureys, S, Sleigh, JW Unresponsiveness ≠ unconsciousness.. Anesthesiology. (2012). 116 946–59 [Article] [PubMed]
Sleigh, JW Depth of anesthesia: Perhaps the patient isn’t a submarine.. Anesthesiology. (2011). 115 1149–50 [PubMed]
Alkire, MT, Hudetz, AG, Tononi, G Consciousness and anesthesia.. Science. (2008). 322 876–80 [Article] [PubMed]
Imas, OA, Ropella, KM, Ward, BD, Wood, JD, Hudetz, AG Volatile anesthetics disrupt frontal-posterior recurrent information transfer at gamma frequencies in rat.. Neurosci Lett. (2005). 387 145–50 [Article] [PubMed]
Bandt, C, Pompe, B Permutation entropy: A natural complexity measure for time series.. Phys Rev Lett. (2002). 88 174102 [Article] [PubMed]
Viertiö-Oja, H, Maja, V, Särkelä, M, Talja, P, Tenkanen, N, Tolvanen-Laakso, H, Paloheimo, M, Vakkuri, A, Yli-Hankala, A, Meriläinen, P Description of the Entropy algorithm as applied in the Datex-Ohmeda S/5 Entropy Module.. Acta Anaesthesiol Scand. (2004). 48 154–61 [Article] [PubMed]
Bonhomme, V, Hans, P Muscle relaxation and depth of anaesthesia: Where is the missing link?. Br J Anaesth. (2007). 99 456–60 [Article] [PubMed]
Appendix
Additional members of the European Multicenter EEG/AEP Anesthesia Monitoring Study Group are (in alphabetical order): Jasmin Blum, M.D., Andreas Hock, Ph.D., Bettina Horn, M.D., Sybille Jochum, M.D. (all from Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany), Gerhard Litscher, M.Sc., Ph.D., M.D.sc. (Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria), Jakob Müller, M.D. (Department of Anesthesiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany), Michael Ningler, Ph.D. (Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany), Eva-Christine Pichlmaier, M.D., Stefanie Pilge, M.D. (both from Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany, and Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany), Argid Rutenberg, R.N., Raimund Rutenberg, R.N. (Department of Anesthesiology, German Heart Center Berlin, Berlin, Germany), Andreas Schöpfer, M.D. (Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria), and Gisela Untergehrer, M.D. (Department of Anesthesiology I, Witten/Herdecke University, Helios Clinic Wuppertal, Wuppertal, Germany, and Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany). Additional members of the Research Group Knowledge-Based Signal Processing are (in alphabetical order): Henry Arenbeck, M.Sc., Michael Gallinat, M.Sc., Thomas Hensel, M.Sc., and Daniela Lücke-Janssen, Ph.D. (all from Institute of Information Logistics, Department of Computer Science and Applied Cognitive Science, University of Duisburg-Essen, Germany).
Fig. 1.
Study diagram showing time points loss of consciousness (LOC), skin incision, burst suppression (BS), and return of consciousness (ROC) and phases of concentration increase and decrease. We performed data analysis at selected time points reflecting (A) the transition between consciousness (phase 1, time points LOC −15 s and ROC + 30 s) and unconsciousness (phase 0, LOC + 30 s and ROC −15 s,) and (B) levels awake (phase 2, LOC −180 s, ROC + 30 s), anesthesia according to clinical practice (phase −1, skin incision −60 s, end decrease +30 s) and BS (phase −3, end increase +30 s, start decrease −60 s).
Image Not Available
Fig. 1.
Study diagram showing time points loss of consciousness (LOC), skin incision, burst suppression (BS), and return of consciousness (ROC) and phases of concentration increase and decrease. We performed data analysis at selected time points reflecting (A) the transition between consciousness (phase 1, time points LOC −15 s and ROC + 30 s) and unconsciousness (phase 0, LOC + 30 s and ROC −15 s,) and (B) levels awake (phase 2, LOC −180 s, ROC + 30 s), anesthesia according to clinical practice (phase −1, skin incision −60 s, end decrease +30 s) and BS (phase −3, end increase +30 s, start decrease −60 s).
×
Fig. 2.
Workflow for (real-time) computation of the anesthesia multimodal index of consciousness (AMIC). After data acquisition (including preprocessing, e.g., artifact detection), parameters (according to table 2) are extracted for each modality. Independent of the sampling rate of specific parameters (e.g., 10 s for electroencephalographic parameters and constant in case of patient age), each parameter is resampled at 5 s and synchronized in time to obtain a multivariate time series. In a first step, classification is performed in two independent models (model A trained for separation of consciousness and unconsciousness and model B trained for separation of different hypnotic states from wakefulness to burst suppression) using an Adaptive Neuro Fuzzy Inference System. Both models were trained to provide best fit of input data to paradigm A and B, leading to data reduction for further processing. In a second step, resulting indices A and B are combined by a third Adaptive Neuro Fuzzy Inference System model resulting in AMIC. ApEn = approximate entropy; BMI = body mass index; BS = burst suppression; HR = heart rate; MAC IE = monitored anesthesia care equivalent of difference between in- and end-expiratory gas concentration; MAP = mean arterial blood pressure; PeEn = permutation entropy; WSMF = weighted spectral median frequency.
Image Not Available
Fig. 2.
Workflow for (real-time) computation of the anesthesia multimodal index of consciousness (AMIC). After data acquisition (including preprocessing, e.g., artifact detection), parameters (according to table 2) are extracted for each modality. Independent of the sampling rate of specific parameters (e.g., 10 s for electroencephalographic parameters and constant in case of patient age), each parameter is resampled at 5 s and synchronized in time to obtain a multivariate time series. In a first step, classification is performed in two independent models (model A trained for separation of consciousness and unconsciousness and model B trained for separation of different hypnotic states from wakefulness to burst suppression) using an Adaptive Neuro Fuzzy Inference System. Both models were trained to provide best fit of input data to paradigm A and B, leading to data reduction for further processing. In a second step, resulting indices A and B are combined by a third Adaptive Neuro Fuzzy Inference System model resulting in AMIC. ApEn = approximate entropy; BMI = body mass index; BS = burst suppression; HR = heart rate; MAC IE = monitored anesthesia care equivalent of difference between in- and end-expiratory gas concentration; MAP = mean arterial blood pressure; PeEn = permutation entropy; WSMF = weighted spectral median frequency.
×
Fig. 3.
Indicator value distribution of the anesthesia multimodal index of consciousness (AMIC; A) and bispectral index (BIS; B) for the different anesthetic levels (central line: median, dots: 5% and 95% percentiles, whiskers: 10% and 90% percentiles, boxes: 50% quartile). Analyzed data consist of 742 data points for awake and consciousness, 526 data points for unconsciousness, 336 data points for anesthesia according to clinical practice, and 335 data points for burst suppression (BS). Horizontal dashed lines indicate optimum threshold (based on maximum sum of sensitivity and specificity) for separation of consciousness versus unconsciousness (AMIC: 80, BIS: 80) and for separation of awake/consciousness versus unconsciousness/clinical practice/BS (AMIC: 79, BIS 72).
Image Not Available
Fig. 3.
Indicator value distribution of the anesthesia multimodal index of consciousness (AMIC; A) and bispectral index (BIS; B) for the different anesthetic levels (central line: median, dots: 5% and 95% percentiles, whiskers: 10% and 90% percentiles, boxes: 50% quartile). Analyzed data consist of 742 data points for awake and consciousness, 526 data points for unconsciousness, 336 data points for anesthesia according to clinical practice, and 335 data points for burst suppression (BS). Horizontal dashed lines indicate optimum threshold (based on maximum sum of sensitivity and specificity) for separation of consciousness versus unconsciousness (AMIC: 80, BIS: 80) and for separation of awake/consciousness versus unconsciousness/clinical practice/BS (AMIC: 79, BIS 72).
×
Table 1.
Anesthetic Combinations for Induction and Maintenance
Anesthetic Combinations for Induction and Maintenance×
Image Not Available
Table 1.
Anesthetic Combinations for Induction and Maintenance
Anesthetic Combinations for Induction and Maintenance×
×
Table 2.
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data×
Image Not Available
Table 2.
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data
Overview of the Parameters Used in the Anesthesia Multimodal Index of Consciousness Divided in Electroencephalographic Parameters, Parameters of Standard Monitoring, Drug Protocol, and Individual Patient Data×
×
Table 3.
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group×
Image Not Available
Table 3.
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group
Number of Patients and Percentage of Included Anesthetic Regimen in Each Cross-validation Group×
×
Table 4.
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B×
Image Not Available
Table 4.
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B
PK Statistics of the AMIC and of the BIS with Respect to Different Anesthetic Combinations and to Different Levels Formed from the Anesthesia States Contained in the Data sets A, B, and A and B×
×