The Narcotrend is a new electroencephalographic monitor designed to measure depth of anesthesia, based on a six-letter classification from A (awake) to F (increasing burst suppression) including 14 substages. This study was designed to investigate the impact of Narcotrend monitoring on recovery times and propofol consumption in comparison to Bispectral Index (BIS) monitoring or standard anesthetic practice.
With institutional review board approval and written informed consent, 120 adult patients scheduled to undergo minor orthopedic surgery were randomized to receive a propofol-remifentanil anesthetic controlled by Narcotrend, by BIS(R), or solely by clinical parameters. Anesthesia was induced with 0.4 micro x kg-1 x min-1 remifentanil and a propofol target-controlled infusion at 3.5 microg/ml. After intubation, remifentanil was reduced to 0.2 micro x kg-1 x min-1, whereas the propofol infusion was adjusted according to clinical parameters or to the following target values: during maintenance to D(0) (Narcotrend) or 50 (BIS); 15 min before the end of surgery to C(1) (Narcotrend) or 60 (BIS). Recovery times were recorded by a blinded investigator, and average normalized propofol consumption was calculated from induction and maintenance doses.
The groups were comparable for demographic data, duration of anesthesia, and mean remifentanil dosages. Compared with standard practice, patients with Narcotrend or BIS monitoring needed significantly less propofol (standard practice, 6.8 +/- 1.2 mg x kg-1 x h-1 vs. Narcotrend, 4.5 +/- 1.1 mg x kg-1 x h-1 or BIS(R), 4.8 +/- 1.0 mg x kg-1 x h-1; P < 0.001), opened their eyes earlier (9.3 +/- 5.2 vs. 3.4 +/- 2.2 or 3.5 +/- 2.9 min), and were extubated sooner (9.7 +/- 5.3 vs. 3.7 +/- 2.2 or 4.1 +/- 2.9 min).
The results indicate that Narcotrend and BIS monitoring are equally effective to facilitate a significant reduction of recovery times and propofol consumption when used for guidance of propofol titration during a propofol-remifentanil anesthetic.
A LARGE body of evidence suggests that Bispectral Index® (BIS®; Aspect Medical Systems Inc., Newton, MA) monitoring might help to assess the hypnotic component of anesthesia, 1–5reduce drug consumption, 6–12and shorten recovery times when compared to a standard practice protocol. 6–12
Recently, the Narcotrend (MonitorTechnik, Bad Bramstedt, Germany), another electroencephalographic monitor designed to measure depth of anesthesia, has become commercially available in Europe. The Narcotrend was developed by anesthesiologists and engineers at the University Medical School of Hannover, Hannover, Germany, and, like the BIS®, performs a computerized analysis of the raw electroencephalogram. In brief, two easily available electrocardiogram electrodes are placed on the patient's forehead; a third electrode on the forehead serves as a reference. After artifact detection, a multivariate statistical algorithm is used for raw electroencephalogram analysis, finally resulting in a six-letter-classification with 14 substages as originally described by Kugler 13in 1981. This classification was modified by Schultz et al. 14for the Narcotrend system (see Appendix). The results of the Narcotrend analysis have been reported to indicate the hypnotic component of anesthesia as 15:
A = awake
C0–2= light anesthesia
D0–2= general anesthesia
E0,1= general anesthesia with deep hypnosis
F0,1= general anesthesia with increasing burst suppression
A recent comparison between Narcotrend monitoring and BIS® analysis has shown corresponding results of both monitors during propofol-remifentanil anesthesia. 16In this study, Narcotrend stages D or E were equivalent to BIS values between 65 and 40 indicating general anesthesia.
The current clinical utility study was designed to investigate the impact of Narcotrend monitoring on recovery times and propofol consumption when compared to BIS® monitoring or a standard anesthetic practice protocol.
Materials and Methods
Before Induction of Anesthesia
With institutional review board approval (University of Saarland, Homburg/Saar, Germany) and written informed consent, 120 adult patients were randomized to receive a propofol–remifentanil anesthetic controlled by Narcotrend, by BIS®, or solely by clinical parameters. Men or women, aged 18–80 yr, with American Society of Anesthesiologists physical status classification of I, II, or III who were scheduled to undergo minor orthopedic surgery expected to last at least 1 h were studied prospectively. Exclusion criteria were a history of any disabling central nervous or cerebrovascular diseases, hypersensitivity to opioids or substance abuse, or treatment with opioids or any psychoactive medication. After enrollment patients were randomized by drawing lots from a closed box. All patients were anesthetized by the same anesthesiologist (S.K.).
All patients were premedicated with 0.15 mg/kg diazepam orally in the evening and on the morning before surgery. In the operating room, an intravenous catheter was inserted into a larger forearm vein, and standard monitors were applied. The electroencephalogram was continuously recorded using an A-2000 BIS® monitor (software version 3.0) and a Narcotrend monitor (software version 2.0 AF), simultaneously. After the skin of the forehead had been degreased with 70% isopropanol, both the BIS® (BIS® sensor; Aspect Medical Systems Inc.) and the Narcotrend (Blue sensor; Medicotest, Olstykke, Denmark) electrodes were positioned as recommended by the manufacturers: For the Narcotrend, two commercially available electrodes were placed on the patients forehead with a minimum distance of 8 cm, a third was positioned laterally, serving as a referential electrode. Finally, impedances were measured for each set of electrodes to ensure optimal electrode contact defined as 6 kΩ or less for the Narcotrend and 7.5 kΩ or less for the BIS® as required by the manufacturers.
Induction of Anesthesia
Induction of anesthesia was started with a remifentanil infusion at 0.4 μg · kg−1· min−1; 5 min later, propofol was given for hypnosis using a target-controlled infusion (TCI; Diprifusor; AstraZeneca, Wedel, Germany), initially started at 3.5 μg/ml. After loss of consciousness, oxygen was given by facemask ventilation; patients received 0.1 mg/kg cisatracurium. The trachea was intubated 3 min later, and the lungs were ventilated to an end-tidal carbon dioxide concentration of 35 mmHg. Immediately after intubation, remifentanil was reduced to a constant rate of 0.2 μg · kg−1· min−1, whereas propofol TCI was adjusted according to electroencephalographic target values or clinical parameters.
Maintenance of Anesthesia and Hemodynamic Control
Continuous monitoring included heart rate, systemic arterial pressure, respiratory rate, oxygen saturation, and end-tidal concentrations of carbon dioxide. Oxygen saturation was measured by pulse oximetry and maintained above 95%. Baseline systolic arterial pressure was defined as the lower of the two measurements obtained the day before surgery and immediately before induction of anesthesia. Heart rate and blood pressure were recorded every 5 min.
During maintenance, all patients were assessed for signs of inadequate anesthesia, hypotension, or bradycardia. Inadequate anesthesia was defined as hypertension, tachycardia, patient movement, eye opening, swallowing, grimacing, lacrimation, or sweating. The definition of adverse hemodynamic responses was adapted from Garrioch and Fitch 17: Responses were classified as hypertension (systolic arterial pressure > 40 mmHg from baseline), hypotension (systolic arterial pressure < 40 mmHg from baseline), tachycardia (heart rate > 100 beats/min), or bradycardia (heart rate < 45 beats/min).
In the standard practice group, if anesthesia was inadequate, the propofol target concentration was increased in steps of 0.5 μg/ml as necessary. If this was judged insufficient, the infusion rate of remifentanil was increased by 0.05 μg · kg−1· min−1. Hypotension was initially treated with intravenous fluid replacement; the propofol target concentration was then reduced in steps of 0.5 μg/ml; and finally, an intravenous vasopressor (Akrinor; AWD Pharma, Dresden, Germany; 1 ml contains 100 mg cafedrine and 5 mg theodrenaline) was given at a dose chosen by the investigator.
In the Narcotrend and BIS® groups, propofol TCI during maintenance of anesthesia was continuously adjusted according to a target value of D0for Narcotrend or 50 for BIS®. In case anesthesia was judged inadequate, although electroencephalographic target values were achieved, the infusion rate of remifentanil was increased by 0.05 μg · kg−1· min−1. Hypotension was initially treated with intravenous fluid replacement, and finally, the intravenous vasopressor was given. In all groups, bradycardia was treated with 0.5 mg atropine.
In all patients, irrespective of the individual group assignment, both BIS® values and Narcotrend levels were continuously recorded as data pairs in intervals of 5 min by a second independent investigator (S.A.). In the standard practice group, both monitors were covered behind a curtain and invisible for the attending anesthesiologist, whereas in the electroencephalographic groups, either the Narcotrend or the BIS® monitor was uncovered.
Fifteen minutes before the expected end of surgery, propofol TCI was reduced in all groups to facilitate rapid emergence from anesthesia, whereas the remifentanil infusion rate remained unchanged throughout the end of the procedure. In the Narcotrend and BIS® groups, propofol TCI was adjusted to a value of C1for Narcotrend or 60 for BIS®, whereas in the standard protocol group propofol TCI was reduced as much as was clinically judged possible without allowing for intraoperative awakening. Simultaneously, complete neuromuscular recovery was ensured by neuromuscular monitoring. All patients received a 100-ml infusion of 0.9% NaCl containing 25 mg/kg metamizol for postoperative pain relief.
The end of surgery was defined as the final surgical suture, when anesthetic delivery was stopped. Emergence from anesthesia was assessed by measuring the times to spontaneous opening of eyes, extubation, and arrival in the postanesthesia care unit. Recovery times and propofol consumption were recorded by a blinded investigator (A.B.). Finally, all patients were visited in the postanesthesia care unit and on the first and third postoperative day and were interviewed about intraoperative recall.
Endpoints and Statistical Analysis
The primary endpoint of this study was defined as the time to spontaneous opening of the eyes. Applying an a priori power analysis, at least 26 patients had to be enrolled in each treatment group to provide 90% power to detect a difference of 3 min at α= 0.05. Statistical analysis included demographic data, duration of anesthesia, recovery times, and consumption of remifentanil and propofol. In addition, a mean propofol infusion rate normalized to weight was calculated from induction and maintenance doses. For nominal data, statistical analysis was performed by means of a chi-square test; for numerical data statistical analysis was performed by means of a Student t test, Mann–Whitney U test, or one-way analysis of variance with Student-Newman-Keuls test for multiple comparisons as appropriate. All tests were two-tailed with statistical significance defined as P < 0.05; data are presented as mean and SD. The recovery time to opening of the eyes was also compared using a Kaplan-Meier survival analysis.
Statistical calculations were planned and performed in collaboration with a statistician of the Institute of Medical Biometrics, Epidemiology and Informatics, University of Saarland, using SigmaStat 2.03 and SigmaPlot 2000 computer software (SPSS GmbH, Erkrath, NRW, Germany).
One hundred twenty patients were enrolled in this study with 40 patients in each group; the groups were similar with respect to age, weight, height, American Society of Anesthesiologists physical status, and duration of surgery (table 1). Problems with skin adherence of the electrodes were not observed; none of the electrodes fell off. No patient reported intraoperative recall.
Recovery times were significantly shorter in the Narcotrend and BIS® groups when compared to standard practice, e.g. , the time to opening of the eyes was 3.4 ± 2.2 min for Narcotrend, 3.5 ± 2.9 min for BIS®, and 9.3 ± 5.2 min for standard practice (P < 0.001;table 2). Both monitors facilitated a similar reduction of recovery times of approximately 60% for the times to opening of the eyes and to extubation and of approximately 45% for the time to arrival at the postanesthesia care unit when compared to standard practice (table 2). In figure 1, a Kaplan-Meier survival analysis shows the percentage of patients remaining unconscious after termination of propofol and remifentanil infusion, indicating a similar course of awakening for both monitor groups.
Bispectral Index Values
Bispectral Index values were obtained in all three treatment groups, irrespective of the individual group assignment, and are displayed in figure 2. Although BIS values were comparable for patients in the Narcotrend and BIS® group, significantly lower BIS values were observed in the standard practice group at various time points of anesthesia.
In addition, the time fractions of actual (vs. targeted) BIS® and Narcotrend values obtained during maintenance of anesthesia were analyzed: Actual BIS values within a range of 45–55 were observed during 52% of the investigation time. During 36.6% of the time, BIS values were found to be lower but not below a value of 35, whereas during 11.4% of the time, BIS values were higher than 55 but not above a value of 70. In the Narcotrend group, stages D0and D1were reached during 52.9% of the study time. During 46.3% of the time, Narcotrend stages were found to indicate a deeper level of anesthesia but not deeper than stage E1, whereas during 0.8% of the time, Narcotrend stage C2was observed.
Average normalized propofol and remifentanil infusion rates were calculated from induction and maintenance doses. Propofol consumption was significantly lower in the Narcotrend (4.5 ± 1.1 mg · kg−1· h−1) and BIS® (4.8 ± 1.0 mg · kg−1· h−1) groups when compared to standard practice (6.8 ± 1.2 mg · kg−1· h−1;P < 0.001). In both monitor groups, the reduction of propofol consumption was of similar extent (table 3). In a parallel manner, propofol TCI concentrations as calculated by the Diprifusor data set were significantly lower with Narcotrend and BIS® monitoring than with standard practice, during both maintenance and emergence from anesthesia (fig. 3). At the same time, average normalized remifentanil infusion rates were similar for Narcotrend (0.21 ± 0.07 μg · kg−1· min−1), BIS® (0.22 ± 0.07 μg · kg−1· min−1), and standard practice (0.20 ± 0.07 μg · kg−1· min−1).
Mean arterial blood pressure values at various time points during anesthesia were comparable for all treatment groups (table 4). However, intervention with a vasopressor was necessary in significantly more patients (n = 27) with standard practice than in the Narcotrend (n = 14) or in the BIS® group (n = 17;P < 0.05). Furthermore, the mean drug amount in those patients treated was significantly higher with standard practice (45 ± 49 mg cafedrine and 2.25 ± 2.45 mg theodrenaline) than with Narcotrend (18 ± 34 mg cafedrine and 0.9 ± 1.7 mg theodrenaline) or BIS® (18 ± 28 mg cafedrine and 0.9 ± 1.4 mg theodrenaline;P < 0.05). Five patients in each group needed 0.5 mg atropine for treatment of bradycardia.
Because of the sex-balanced study design, 60 male and 60 female patients completed the study. With standard practice, male and female patients received comparable amounts of propofol, but recovery times were significantly shorter in women than in men (table 5). However, in the electroencephalogram-monitored groups, propofol consumption was significantly lower for men with BIS® (P < 0.05) and tended to be lower with Narcotrend (P = 0.09).
In the current clinical study, the Narcotrend, a new electroencephalographic monitor designed to measure depth of anesthesia, was investigated during a propofol–remifentanil anesthetic. Our results demonstrate that Narcotrend guidance for the titration of propofol infusion resulted in a significant reduction of recovery times and propofol consumption, which was equally effective as BIS® monitoring when compared to a standard practice protocol. Furthermore, hemodynamic stability was obviously superior in patients monitored with Narcotrend or BIS® when expressed as the need for vasopressor support according to a predefined hemodynamic protocol.
These results are of current interest because a number of different monitor systems have recently been developed that promise to optimize the control of depth of anesthesia. The BIS scale has been studied in a number of investigations 18and must now be considered as the main competitor for all new devices developed in this field. In several studies, a correlation between BIS and anesthetic concentrations or anesthetic effect was shown for midazolam, 1,5propofol, 1,4,19propofol and alfentanil, 2and the volatile anesthetics isoflurane, 1desflurane, 20and sevoflurane. 3
Different to the BIS scale from 100 to 0, the Narcotrend provides a six-letter classification including 14 substages of the depth of anesthesia from A (awake) to F (general anesthesia with increasing burst suppression). 14,15In a preceding investigation in 50 orthopedic patients, we assessed in a parallel manner the results of Narcotrend and BIS® monitoring during a propofol–remifentanil general anesthetic. 16We found that changes in the depth of anesthesia as indicated by BIS® were accompanied by typical corresponding effects as displayed by the Narcotrend: If BIS values were measured between 100 and 85, 95.5% of all data pairs indicated a Narcotrend stage of A or B. In case the BIS value was found to be 65 to 40, the corresponding Narcotrend stages were predominantly measured as D (52.4%) or E (41.1%). In detail, BIS values of 50 and 60 were matched to Narcotrend stages of D0and C1, respectively, and therefore, these results were used as target values for the guidance of propofol titration in the current investigation.
The impact of electroencephalogram-derived parameters on clinical practice has recently been discussed in detail. 18Although a reduction of the incidence of intraoperative awareness is still questionable and difficult to study, 21some investigations have addressed the potential benefit of this monitoring for the guidance of anesthetic titration. 6–12,22,23Gan et al. 7investigated the influence of BIS® monitoring on recovery times and propofol consumption in individuals receiving a propofol–alfentanil–nitrous oxide anesthetic. Compared to patients in whom propofol was solely titrated according to clinical parameters, BIS® guided propofol titration facilitated an approximate 35% reduction of the times to opening of the eyes, responding to command, and extubation. In a parallel manner, mean propofol consumption per patient was reduced by 23%. In the current investigation, it was for the first time demonstrated that Narcotrend monitoring may be equally effective as BIS® monitoring in reducing recovery times and propofol consumption if these monitoring systems are used for the guidance of propofol titration: In both groups, we observed a diminution of the times to opening of the eyes and extubation of 58–63% and a reduction of propofol consumption of 26%.
The difference in the reduction of recovery times observed in the current investigation and in the study by Gan et al. 7may primarily be explained by the different anesthetic regimen including remifentanil instead of alfentanil. In addition, our results demonstrate that both Narcotrend and BIS® guidance obviously allow for a better individual titration of propofol infusion, apparently resulting in more stable hemodynamics when expressed as the need for vasopressor support for the prevention of hypotension.
Interestingly, although BIS values observed in the BIS® and Narcotrend groups were significantly different from those recorded in the standard practice group, this difference was not as substantial as might be expected (fig. 2). At the same time, patients in the standard practice group received substantially higher propofol TCI concentrations (fig. 3). This observation is best explained as a “BIS plateau phenomenon.” Such a plateau phenomenon at BIS values similar to ours has already been described by Olofsen and Dahan 24for sevoflurane and isoflurane: BIS values seem to reach a plateau or show only minimal changes although propofol, sevoflurane, or isoflurane concentrations are substantially increasing.
Another device currently available and developed to measure hypnosis during anesthesia delivery is the Patient State Analyser PSA 4000 monitor (Physiometrix Inc., N. Billerica, MA). This monitor provides the Patient State Index, and it was recently demonstrated by Drover et al. 22that Patient State Index guidance of propofol titration may also facilitate a reduction of recovery times and propofol consumption during a propofol–alfentanil–nitrous oxide anesthetic.
A potential shortcoming of the current study and of all other investigations with a similar study design is the question of investigator bias in the standard practice group. On the one hand, “learning contamination” bias 25must be addressed as a problem of unintended improvement of standard clinical practice patterns occurring with the introduction of a new monitor device, thereby reducing the difference of results in a randomized device trial. In the current study, however, the investigator responsible for the titration of propofol infusion was experienced in BIS® and Narcotrend monitoring and was involved in a preceding investigation with both types of monitors using an identical propofol–remifentanil general anesthetic technique. 16On the other hand, results of the standard practice group may be negatively influenced by subtle investigator bias, i.e. , leading to an overestimation of the difference between standard practice and the device monitored groups. However, as clearly outlined in figure 3, propofol infusion rates used in the standard practice group of the current investigation were well within the recommended therapeutic range, and remifentanil infusion rates were identical in all three groups according to the study design. In addition and consistent with good clinical practice, propofol infusions were reduced during maintenance of anesthesia with a further marked reduction during the last 15 min before the expected end of surgery. Therefore, significant investigator bias can obviously be excluded as a confounding factor for the explanation of the current results.
Investigations on sex differences have attracted growing interest in recent years, and differences have been reported for various aspects of anesthesia. 26,27In an a posteriori analysis of sex effect, Gan et al. 28reported that women emerged faster than men from a propofol–alfentanil–nitrous oxide anesthetic. In another study, Glass et al. 1were able to demonstrate that women had significantly higher BIS values than men, whereas the measured propofol plasma concentrations were similar in both groups. In the current investigation, a sex-balanced study design was chosen a priori , and 60 male and 60 female patients were enrolled. We were able to show that with comparable amounts of propofol in the standard practice group, recovery times were significantly shorter in women than in men, whereas in the electroencephalogram-monitored groups, propofol consumption was significantly lower for men with BIS® and tended to be lower for men with Narcotrend monitoring. Therefore, our data underline that studies on devices developed for the guidance of anesthetic titration should consider sex differences as a relevant cofactor for the design of the investigation or for the interpretation of its results.
In conclusion, we evaluated the influence of Narcotrend guidance on the titration of propofol infusion in comparison to BIS® monitoring and a standard practice group. During propofol–remifentanil anesthesia, Narcotrend and BIS® monitoring were equally effective in reducing recovery times and propofol consumption when compared to standard clinical practice. Further clinical utility studies are warranted to investigate the impact of Narcotrend guidance on the titration of inhaled anesthetics.
Appendix: Narcotrend Algorithm
The Narcotrend monitor performs an automatic analysis of the electroencephalogram during anesthesia. The methods for the automatic classification were developed on the basis of a visual assessment of the electroencephalogram, which, in its origins, is related to sleep classification.
In 1937, Loomis et al. 29described systematic changes of the electroencephalogram during human sleep and defined five stages, A–E, to distinguish different electroencephalographic patterns. Subsequently, the scale was extended, refined by the definition of substages, 13and applied to the classification of electroencephalograms recorded during anesthesia (stages A = awake to F = very deep level of anesthesia).
Electroencephalograms from the states of sleep and anesthesia have in common that a progressive “deepening” of either condition is accompanied by a slowing, i.e. , an increasing proportion of waves in the lower frequency bands. Apart from this, differences are well known, e.g. , the burst suppression pattern of anesthesia (stage F) is not found in normal sleep of adults, and typical graphoelements of the sleep electroencephalogram, e.g. , saw-tooth waves, seem to be missing during anesthesia with hypnotically acting drugs.
Schultz et al. 14used a scale with the substages A, B0–2, C0–2, D0–2, E0,1and F0,1for visually characterizing and classifying electroencephalographic patterns observed during anesthesia with different volatile and intravenous drugs (fig. 4).
The first step in the development of the automatic classification algorithm of the Narcotrend was to build up a database of typical examples for all substages from A to F. For this purpose, artifact-free electroencephalographic epochs from a large database of one-channel electroencephalographic recordings collected from the electrode positions C3–P3during anesthesia with thiopental–enflurane or propofol were selected. Initially, more than 1,000 artifact-free epochs with a length of 20 s were visually classified to form the basis for the development of automatic classification algorithms. From these electroencephalographic epochs, numerous quantitative features from the time and the frequency domain were extracted, e.g. , spectral parameters, entropy measures, and autoregressive parameters. 30The extracted parameters were statistically analyzed to identify a subset of electroencephalographic parameters that were best suited to discriminate between the different visually determined electroencephalographic substages. Age-related changes in the electroencephalogram were incorporated. The resulting parameters were combined in multivariate discriminant functions to classify an electroencephalographic epoch into one of the substages between A and E. The discriminant analysis yields probabilities for the degree of similarity of an electroencephalographic epoch with the typical stages A–E during anesthesia. In addition, algorithms for the classification of stage F were developed that are based on the proportion and intensity of very flat electroencephalographic segments. The algorithms were revised continuously including extensions for the standard use of frontal electrode positions.
The clinical application of the algorithms covers anesthesia with volatile anesthetics (desflurane, sevoflurane, isoflurane, enflurane, halothane), nitrous oxide, and intravenous agents (propofol, etomidate, methohexital, thiopental, benzodiazepines) in combination with different analgetics (fentanyl, alfentanil, sufentanil, remifentanil).
One of the most recent validations of the classification functions was performed on an independent data set containing 1,148 epochs from one-channel, frontal-lead routine clinical electroencephalographic recordings with inhalational anesthetics (isoflurane, enflurane, and sevoflurane) and with propofol. The correlation between visual and automatic classification was high with a prediction probability of PK= 0.90. 15
Algorithm development also included the evaluation of characteristic artifacts in the electroencephalogram, e.g. , caused by muscle activity, eye movements, and electrocautery, as well as the development of functions for automatic identification and exclusion of artifacts from subsequent data processing.
The basic steps in the computation of the Narcotrend stages are shown in figure 5. The electroencephalogram is recorded from a frontal channel using self-adhesive pre-gelled standard electrocardiographic electrodes. In principle, other electrode positions and types can be chosen, and a two-channel recording is also possible.
The signals are sampled at 128 samples per second with a 12-bit resolution and bandpass filtered to 0.5–45 Hz. During the recording, impedances and electrode potentials are tested at defined time intervals to ensure a constantly high quality of the electroencephalographic signal. Epochs with a length of 20 s are the units of classification. To provide a smooth and continuous assessment of the electroencephalographic signal, the preceding 20-s epoch is classified every 5 s, yielding an overlap of 75% of the recorded data.
First, extensive algorithms for artifact detection are applied to prevent artifact-loaded epochs from disturbing subsequent data analysis. Thereafter, the electroencephalographic parameters that are relevant for suppression detection and that contribute to the discriminant functions are calculated, and the epoch is classified into one of the substages. A sufficient similarity of the epoch to one of the typical electroencephalographic stages is required for a classification to be made.
“Background” parameters are calculated and updated during the course of an electroencephalographic recording. These parameters are sensitive to electroencephalographic patterns untypical for general anesthesia, e.g. , distinct epileptiform activity or K complexes, and are used for plausibility checks of the classification results. For example, a range of plausible amplitude values is calculated for each recording. Values outside these individually adjusted intervals are not likely to represent normal artifact-free electroencephalographic activity during anesthesia.
Finally, a smoothed value of the Narcotrend classification is calculated as a weighted mean (weights depend on the background parameters) and depicted as trend in the “cerebrogram.”