| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
From the Institutes of *Anesthesiology, Neurology and Psychiatry, Ludwig Maximilians University; Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany; Department of Anesthesia and Critical Care, Johanniter Hospital, Stendal, Germany; ||Walther Straub Institute, Ludwig Maximilians University, Munich, Germany; ¶Department of Anesthesia and Critical Care, Martin Luther University, Halle, Germany; and #Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany.
Address correspondence and reprint requests to Dr. Alfred W. Doenicke, Institute for Anesthesiology, Ludwig Maximilians University, Pettenkoferstr. 8a, D – 80336 Munich, Germany. Address e-mail to doenickeaw{at}aol.com.
Abstract
BACKGROUND: The NarcotrendTM (NCT) is a one-channel electroencephalogram (EEG) monitor of the level of sedation. It is based on a visual EEG scoring system, which was developed by Loomis and modified by Kugler, to yield a visual expert classification (VEC) scheme for differentiation of six levels of sedation (A–F), which are subdivided into 16 substages. We designed the present study to test whether results of the automated classification of one-channel NCT input reflect those from VEC of five-channel EEG.
METHODS: Twelve healthy male volunteers received propofol using two different infusion regimens in a randomized, crossover design with concomitant NCT monitoring and VEC. Scoring results of NCT were compared with those of VEC.
RESULTS: During the infusion period, score differences of more than three substages were observed in 14 of 24 (= 58%) propofol administrations (4%–7% of total data). Often, the NCT indicated lighter sedation than VEC, which revealed more 
activity from nonfrontal leads. During recovery, NCT reported deeper sedation than VEC in 6 of 24 (= 25%) propofol administrations. Discordant trends (periods of at least five subsequent epochs with monotonic, but opposite trends for both NCT and VEC) were noted in 9 of 24 propofol administrations (37%). Furthermore, NCT had several periods when no staging information was displayed, varying from a few seconds to 10 min.
CONCLUSIONS: As the algorithm of NCT is proprietary and not accessible to the public, reasons for the observed differences between NCT and VEC cannot be analyzed and explanations must remain speculative.
Monitors of anesthesia depth have attracted considerable interest in recent years, not only because they may help to optimize drug administration, but also because they may be useful to detect inadequately light or excessively deep anesthesia, or to confirm adequate depth of anesthesia. Commercially available devices are based on electroencephalographic activity, but use proprietary signal processing methods to reduce artifacts and calculate (proprietary) (alpha)numerical index values. The NarcotrendTM (NCT) is an automated electroencephalogram (EEG) monitor that performs real-time classification using an alphanumeric scoring scheme (1) originally based on Kugler's sleep score with 16 substages (2). Over the years, the NCT underwent several modifications (1,3); and recently, the alphanumerical index has been translated into a numerical index (0 ... 100), which inversely correlates with the level of sedation and anesthesia (4).
Although conventional EEG analysis of sleep/ sedation requires input from at least two separate uni- or bipolar central electrodes (5) [usually C3 and C4 or C3/A2 and C4/A1, respectively, according to the international 10–20-positioning scheme (6)], the NCT monitor is based on a one-channel EEG with a frontal electrode at each hemisphere, thus relying almost exclusively on the frontal EEG when determining the depth of sedation (1,3,4,7–9).
The basis of NCT scoring is derived from a visual expert classification (VEC) of sleep/sedation as proposed by Kugler (2), which requires at least a five-channel EEG recording system with electrodes positioned at Fp2, F2, C4, P4, O2.
Clinical and experimental studies with the NCT monitor have been published in the past 5 yr (1,3,4,10), but its accuracy, when analyzing EEG activity from the forehead, when compared with VEC based on a multichannel EEG, has not been investigated systematically. Since both techniques share Kugler's scoring scheme for reporting, we aimed to compare the output of the NCT monitor with off-site VEC. We chose propofol as the sedative and exercised two administration protocols with different propofol concentrations (11) to assess the agreement of both techniques. We randomly assigned 12 volunteers to either protocol and then switched to the other protocol after a 14-day washout period. The results afforded some interesting differences that may be relevant in clinical settings.
METHODS
Subjects
Twelve healthy male volunteers (18–45 yr) were included in the study after approval by our ethics committee and individual written consent. Volunteers did not take any medications or drugs known to alter central nervous system (CNS) function and were free from diseases that affect CNS function.
Study Protocols
Two different administration protocols were used to compare the results of the VEC versus NCT classification. We randomly assigned 12 volunteers to 1 of 2 administration regimens (Protocol 1, i.e., monophasic, or Protocol 2, i.e., biphasic sedation). After a washout period of 14 days, each volunteer was shifted to the other protocol in a crossover design. The study was performed in an air-conditioned operating room equipped with standard anesthesia monitors and critical care equipment. Propofol administrations were intentionally performed without any disturbing pain stimulus (e.g., incision, pinching), but repeated acoustic stimuli (6.38/s) for analysis of auditory evoked potentials (AEP) were applied.
Propofol (DiprivanTM 1%, Astra Zeneca, Wedel, Germany) was delivered by a programmed infusion pump (P6000/TIVA, AlarisTM Medical System, London, United Kingdom).
In Protocol 1 ("monophasic" sedation), an IV propofol dose of 2 mg/kg was administered over 4 min, followed by continuous infusion of the drug (83.3 µg · kg–1 · min–1 = 5 mg · kg–1 · h–1) for 10 min.
In protocol 2 ("biphasic" sedation), a propofol bolus (2 mg/kg) was given over 60 s. Twelve minutes later, a second bolus of propofol (1 mg/kg) was administered over 30 s, followed by propofol infusion (100 µg · kg–1 · min–1 = 6 mg · kg–1 · h–1) for 10 min.
Monitoring
A routine check of devices and wires in the study room was performed for each volunteer to verify that signals were free of interference and coupling artifacts. After proper placement of electrodes, monitors, and infusion lines, volunteers were allowed to rest for 10 min before the propofol administration was started. The NCT monitor, the conventional five-channel EEG recorder (for VEC), and the AEP recording device were synchronized with the ON/OFF switch of the infusion pump providing the START signal. A radio-controlled clock (Junghans Uhren GmbH, D-78701 Schramberg, Germany) served as a chronometer that was visible to all staff personnel.
Electrodes of the NCT monitor were positioned according to the manufacturer's recommendations near the F2 position of each hemisphere. The five-channel EEG electrodes at positions Fp2, F4, C4, P4, O2 [according to the international 10–20-standard positioning scheme (6)] were connected to reference electrode A2.
After impedance tests, monitoring was started 10 min before propofol was injected and continued until 35 min after the initial propofol injection.
EEG VEC
The visual EEG scoring system requires a multichannel EEG and is based on EEG patterns observed during sleep. It was initially developed by Loomis et al. (12), and later refined by Kugler (2) to yield 16 discernible stages (A0, A1, A2, B0, ..., E2, and one F stage). For visual evaluation, each recorded EEG was retrospectively divided into equal epochs of 20 s, beginning 1–4 min before the start of the infusion pump. This procedure transformed the space- and time-dependent EEG signals to the scalar 16-substage score for each 20-s epoch (Table 1). As shown in Table 1, VEC is based not only on an analysis of the frequency contents of the EEG signal, but also on so-called "grapho elements" (see description below), which were counted for each epoch with a focus on C4, P4, and O2 traces. Each epoch of the five-channel EEG was visually classified off-site according to published criteria (13) by two investigators, independently who were blinded to the clinical status and the output of the other monitor. In a few instances when both experts disagreed upon the substage, the region-of-interest (ROI) was presented again to each expert to reassess his/her decision (without knowing the peer's result). If this did not provide a result, a third expert (AWD) was involved and made his decision without knowing the peers' results, but from a paper copy of the region of dissent. As a last resort (one or two instances), the three experts sat together and discussed the ROI until an agreement was reached. Only then was the time axis of the offending epochs disclosed.
|
Baseline EEG Activity
To adjust for normal EEG variants, a baseline EEG (10 min) was recorded from all volunteers before the beginning of propofol administration. This is also useful for reference before induction of anesthesia (14).
Grapho Elements
Grapho elements are alternating periods of hypervoltage patterns (15), e.g., K complexes and vertex waves. K complexes have been associated with the "twilight zone" between wakefulness and sedation, but more specifically during the transition to wakefulness when external stimuli again begin to trigger brain activity (15). Vertex waves show in even lighter levels of sedation, but can only be seen in electrodes in the vertex position.
Analysis of grapho elements are essential for VEC. Each of the two VEC experts separately browsed the fanfold paper strip epoch by epoch looking for these elements.
NCT
The NCT is an automated monitor that samples EEG signals from the forebrain via a single pair of frontal EEG electrodes and a third lead positioned over the right forehead versus ground (4). After an initial impedance check, further recordings of the EEG signal were automatically interrupted at regular intervals for impedance checks. The data stream is subjected to a Fast Fourier Transformation to yield proportions of the signal in four relevant frequency bands for each 5-s epoch (4). Signals of 5 s epochs are analyzed almost in real-time and, on the basis of a proprietary algorithm, an index of anesthetic depth is calculated. After summarizing four consecutive epochs and applying multivariate comparative statistics with stored profiles from an internal EEG database in real-time mode, the NCT calculates a simple one-letter-one-digit score for each 20-s epoch. A more recent version provides a numerical index ("100" = wakefulness; "0" = electrical silence) (3). In the present study, NCT version 2.0AF/F was applied.
The NCT score deviates from Kugler's scheme (2) by a single "A" stage (for awareness), but differentiates two F stages, F0 and F1, based on the duration of cortical suppression (1,3,8). The NCT manufacturer suggests that any D stage (D0, D1, D2) provides sufficient hypnosis for surgical anesthesia (4).
Signals that did not pass the NCT algorithm caused an error, and no score was displayed for such epochs.
Baseline EEG and Grapho Elements
The NCT does not require a baseline exercise before performing its analysis, nor does it seem to identify grapho elements, but apparently subjects them to ordinary frequency analysis.
AEP
To probe the depth of sedation, a continuous binaural auditory stimulation was performed with rarefaction clicks (0.4 ms; 70 dB above hearing level) and a repeat frequency of 6.38/s, using insert earphones (AW180, Oticon, Strandvejen, Denmark). In addition to sampling AEP, EEG activity was recorded from two electrodes placed at the vertex and the right mastoid using a specially designed amplifier described previously (16). The obtained AEP were analyzed separately and are not the subject of this article.
Data Analysis and Statistics
Measured variables were expressed as median with quartiles or mean ± sd where appropriate.
Propofol Effect-Site Concentrations
Based on the data set of Schnider et al. (17), propofol effect-site concentrations were calculated for both protocols to provide an additional estimate of propofol's effects.
Classification
For statistical comparison of visual and automated EEG classification, the scored data were transformed into an integer scale running from 0 (awareness) to 15 (comatose sleep). Table 1 shows the juxtaposed stages and their numerical equivalents that were used for statistical comparison. The substages of Kugler's score were translated to a simple integer, i.e., A0 = 0, A1 = 1, A2 = 2, B0 = 3, ..., E2 = 14, F = 15. Likewise, the NCT output was transformed into an integer score running from 1 (A) to 16 (F1). The NCT uses only one "A" stage that was assigned an integer value of "1," corresponding to "A1" of Kugler's scale.
Bland–Altman Test
Since both NCT and VEC classification were originally based on the EEG and on Kugler's scoring system (2,4), a Bland–Altman analysis was performed to show agreement of the two methods (18).
Trend Analysis
In addition to the comparison of absolute values, trends of parameter values were analyzed.
For two methods that claim to measure depth of sedation based on a common score, one may or may not expect that the identical index value is calculated by both methods. Irrespective of potential differences between the absolute index values, a clear sign of discrepancy would be if one technique indicates deepening of anesthesia whereas the other indicates lightening of sedation. To identify these extremes, trend analysis of sequential scores was performed offline for the output from one-channel and five-channel EEG.
A "trend" was defined as a sequence of five or more consecutive epochs that featured a monotonic increase or decrease of sedation. We arbitrarily set a sequence of at least five consecutive epochs (100 s) as the ROI, because automated controllers of sedation would not tolerate a deviation over 100 s without readjusting drug delivery. For example, a score sequence of C2–D1–D2–D2–E0 was recognized as a (monotonic) downward trend (towards deeper sedation), whereas the sequence C0–B2–B0–A0–A0 was recognized as an upward trend (indicating "arousal"). Similar (i.e., same end-points), but not monotonic, sequences like C2–D2–D1–D2–E0 (towards sedation) or C0–B2–A2– B1–A0 (towards arousal) would not qualify as monotonic trends.
Trend analysis was performed by feeding the scoring information of NCT or VEC output into a customized module of the proFit software package (v6.0.4; Quantum Software, Zurich, CH). The software begins with the first epoch and compares its score with the next four epochs. If it sees a monotonic trend in either direction ("more" or "less" sedation), it extends its selection to the next epoch until a violation of the "current trend" is observed. Sign and length of the monotonic trend are marked, and the program recommences its analysis beginning with the next epoch. Uniform trends of <5 epochs were ignored, as were periods with no change of classification ("zero" trend). The trend lists of NCT and VEC were compared and searched for occurrences where the signs of concomitant trends were contradictory. Discordant trends were verified by reexamining the pertinent raw data. To keep this approach conservative, we also required that trend mismatches comprised changes of more than two substages for both the NCT and the VEC, due to an assumed interobserver variation of ±1 substage for both methods and a potential variability of ±1 substage during stage "A," which is divided into three substages for VEC, but does not contain substages for NCT.
RESULTS
Twelve healthy male volunteers (age, 23–44 yr) were enrolled and completed the study. From all participants, valid sets of NCT and VEC data were obtained and evaluated. Including epochs of 35 min from the beginning of drug administration for each of the 12 volunteers, we expected score data from 1260 epochs for each protocol and scoring method. The NCT did not display score information for 51 epochs of the monophasic and for 114 epochs of the biphasic protocol (blank screen). Only two procedures had no score drop-outs at all, whereas the longest sequence with a blank screen was 43 epochs (Volunteer 5 during biphasic protocol). For expert classification, 12 epochs (derived from three procedures) were excluded because no classification was obtained because of electrical noise and artifacts.
Classification of NCT and VEC
Protocol 1 (Monophasic Sedation, n = 12)
The total dose of propofol per volunteer was 2.8 mg/kg. The NCT (Fig. 1a, top panel) indicated a maximum sedative effect after 4–5 min with a median level of D2 in agreement with VEC (median level E0/E1; Fig. 1a, bottom panel). Thereafter, NCT returned rapidly to C0/B2, whereas VEC remained D2/D0. After the end of the drug administration, both methods showed considerable between-subject variation. Two minutes before the recording period was terminated, NCT rapidly increased to stage B0 (subvigilance) in agreement with VEC, which showed a slow increase to B0 with reduced incidence of K-complexes with increasing vigilance.
|
|
|
Protocol 2 (Biphasic Sedation, n = 12)
The total dose of propofol per volunteer was 4 mg/kg. VEC and NCT classification were in close agreement during the first 4 min (Fig. 1b, top and bottom panel). According to the NCT, the level of sedation rapidly returned to B1, whereas the return of VEC (to B2) occurred more slowly. After the second bolus, the median NCT value hardly exceeded stage D2, and returned rapidly to B2 during the infusion; VEC indicated deep sedation (D2–E0) and after the end of the propofol infusion exhibited increasing amounts of K-complexes (when compared with Protocol 1; cf. Figs. 1a and b, bottom panels).
Comparison of Protocol 1 vs 2
All volunteers spontaneously reported more drowsiness at the end of the study period of Protocol 2 compared with Protocol 1. Regardless of the protocol, we observed matching results (score differences <3 substages) of both NCT classification and VEC for 10 of 24 procedures (= 42%). For 14 of 24 propofol sedations (= 58%) at least one major discrepancy between visual and automated scoring occurred during or shortly after propofol administration. We counted 90 epochs with major discrepancies for Protocol 1 and 58 epochs for Protocol 2, i.e., 4%–7% of all epochs were conflicting. The discrepancies were mostly seen at stages of light sedation (B2–C2) in the NCT (Figs. 1a and b), when VEC often indicated deeper levels of sedation (C2–E0).
Statistical Comparison of NCT Versus VEC
Baseline recordings were arbitrarily excluded from statistical comparison. Since the NCT did not calculate an index value in 51 epochs of Protocol 1 and no VEC could be obtained for six epochs, (12 x 35 x 3) – 57 = 1203 data pairs were used for statistical testing. With Protocol 2, the NCT did not report a score for 114 epochs, and for six epochs no classification was obtained by the VEC, yielding 1260 – 120 = 1140 data pairs. To compare VEC and NCT performance, a Bland–Altman analysis was performed on the basis of 2343 data pairs of NCT and VEC. Arithmetic differences of data pairs were plotted against calculated averages for each protocol (Fig. 2).
|
Protocol 1: the mean difference ("bias") was –0.8 (Fig. 2a), i.e., with this more simple protocol the NCT provided an EEG measure that was nearly one stage different to the VEC EEG measure in the direction of a less hypnotic state, but with averages between five and nine the spread of scoring differences was large (95% confidence interval (CI), –6.2 to 4.5).
Protocol 2: the offset was –1.9. In comparison to VEC the NCT under-estimated depth of hypnosis by almost two stages (Fig. 2b). The variation of data points was smaller (CI: –6.3 to 2.5), presumably because the increasing spread of score differences among volunteers was observed only after the propofol infusion was terminated at toff = 22 min (in contrast to toff = 14 min with Protocol 1).
Equally for the different protocols, both differences and spread of differences were most prominent during levels indicating sedation (average classification of 5–8, Fig. 2).
Trend Observations in VEC and NCT Classifications
Table 2 summarizes the incidence and duration of all major trend conflicts that were detected in the 24 procedures. We identified nine instances (Protocol 1: 6; Protocol 2: 3) in nine procedures when the trends of five or more consecutive epochs indicated changes in the level of sedation in the opposite direction: either the NCT indicated return from sedation towards arousal whereas VEC indicated an increasing level of sedation or vice versa. The shaded entries in Table 2 denote instances when indicated trends of one (or both) of the scoring systems were in contrast to calculated propofol effect-site concentrations (Fig. 3).
|
|
Five of nine discordant trends were noted in the recovery period of Protocol 1 (Volunteers 3, 8, 10, 11, and 12). With Protocol 2, two of the three trend deviations occurred immediately before the second administration of propofol.
Figure 4a shows the situation of Volunteer 6 when NCT and VEC both indicate the onset of propofol's action, but shortly thereafter the NCT indicates decreasing sedation with continuing propofol infusion. Eventually, with further continuing infusion, NCT decreased to levels indicating deep sedation. After 28 min when a single K complex signaled "arousal" and the VEC increased to B0, 5 min elapsed before the NCT indicated return to "wakefulness."
|
In Volunteer 3 (Fig. 4b), responses of NCT and VEC to the first propofol injection were similar, but the NCT indicated faster recovery than VEC, indicating moderate sedation almost until the second propofol bolus was administered. Before the onset of the second propofol bolus, VEC indicated a decreasing level of sedation, whereas NCT showed increasing sedation in this volunteer, followed by a decrease of anesthetic effect, until the effects of the second bolus were indicated. Soon after the end of the second injection, with continuing infusion, NCT values decreased indicating "light sleep" again, whereas VEC indicated continuous deep sedation for almost 30 min. NCT continued indicating light sedation, but at the end of propofol infusion, the index slowly decreased towards "C1" followed by a blank screen for approximately 5 min, which prevented further comparison. When the display resumed operation, the score was D1 until the recording period was almost over. Two K complexes were seen in the five-channel EEG at 0:34:20, immediately before both VEC and NCT score returned to wakefulness.
These two examples clearly show that the two methods not only disagreed on the absolute depth of sedation, but also presented contradictory responses to the action of propofol.
DISCUSSION
The present study compared depth of anesthesia classification of propofol monoanesthesia by two scoring systems, the automated NCT and the nonautomated VEC. This comparison was performed, because, according to the manufacturer, VEC is the basis of NCT classification. As the results of the comparison show, NCT classification does not always reflect VEC, and in some instances, considerable disagreement was found. Differences occurred mainly during sedation levels, which may be particularly critical because decreases of the anesthetic concentration would bear the risk of inadvertent consciousness, whereas increases may result in excessively deep levels of anesthesia. In total, the number of conflicting results was only 6.3% of the periods where NCT values were calculated. Nevertheless, even smaller numbers of conflicting results may lead to major differences in general anesthesia if an index is used to guide the administration of anesthetics. To obtain an estimate for this aspect, trend analysis was performed, which identified conflicting trends when one index would suggest a decrease of the anesthetic dose and the other index would suggest an increase. Results of this analysis showed that, during 9 of 24 propofol infusions, there was a conflict between NCT-based scores and VEC-based assessments, which would translate into a different course of anesthesia.
Scoring systems for assessment of anesthetic depth are attracting anesthesiologists who are looking for simple means to verify and communicate the uneventful course of successful anesthesia. As the brain is the main target of hypnotic effects of anesthetic drugs, the assessment of brain activity may provide the means to assess these anesthetic effects. Therefore, measurement of electrical activity of the brain cortex, i.e., the EEG, may be the basis of such a scoring system. In 1962, Brechner et al. (19) already felt it important to use two leads for EEG measurements, and recommended either fronto-occipital or fronto-coronal bipolar placement. Similar requirements were devised by Wauquier (20), Hansen and Claassen (14), Rechtschaffen and Kales (21), and many others (11,12,15,22–31).
Based on EEG analysis, Kugler's VEC score was published in 1981 as a refinement of Loomis' 5-stage score for the assessment of sleep (2). Since 1966, it has been successfully used to assess depth of sedation during anesthesia (32), with a variety of anesthetic drugs, e.g., barbiturates, propanidid, benzodiazepines, opioid analgesics, ketamine, etomidate, and propofol (13,32–34).
The NCT monitor is the first automated monitor to be based on Kugler's score for assessment of depth of anesthesia or sedation in the operating room (1). In contrast to Kugler's original EEG scoring system, which required at least five bipolar electrodes at different regions of the brain, the NCT score is based on only two frontal electrodes and a ground electrode. On the basis of one EEG-channel, the NCT monitor calculates the NCT index.
Both, the manufacturer of the NCT (3) and other authors (10) refer to an abstract published in 1999 (35), which indicated an excellent agreement (Spearman coefficient, 0.91) between NCT and VEC. However, this abstract was based only on the analysis of a rapid transition from wakefulness to deep sedation. Detailed analysis of the entire study period, including other epochs, had already revealed considerable differences between the two classification methods reflected by a linear regression coefficient of r2 = 0.632 for the full observation period (36).
In the present volunteer study, one simple and one more complex protocol for administration of propofol was applied. All 12 volunteers underwent both procedures in a randomized, crossover design in two sessions that were 2 wk apart. In addition to the two EEG devices, we also used an AEP monitor as a third measure of the CNS effects of propofol. In addition to data acquisition, we reasoned that the regular AEP clicks would provide a repetitive stimulus for wakefulness to prevent "after sleep" and other nondrug related effects that would increase sedation.
Analysis of the 24 propofol administrations showed that the two scoring systems NCT and VEC often agreed in their assessment of the level of sedation, but did not match in every instance. To qualify these differences, two approaches were attempted:
Comparison of Pooled Scoring Results
As shown in Figure 1, pooled data clearly reflect the differences between the two protocols, but failed to discriminate specific events. Even though propofol was administered on milligram per kilogram basis, wide ranges of scoring results (Figs. 1a and b) suggest a wide variance of propofol's pharmacodynamic effects. For both the monophasic and biphasic procedures, the NCT score indicated a faster recovery towards light sedation shortly after the injection was complete and the infusion started. Results of VEC seem to be more closely related to calculated propofol effect-site concentrations. One may suspect that arousal stimuli may lead to a lighter level of anesthesia than that indicated by propofol concentrations. Nevertheless, NCT results are surprising, since the infused drug was expected to provide sufficient sedation against nonpainful stimuli such as AEP clicks. Although NCT indicated rapid emergence from sedation, there were almost no K complexes in the simultaneously recorded five-channel EEG. K complexes belong to the so-called class of grapho elements, which may not be picked-up by automated EEG analysis (without specific pattern recognition algorithms) but may give additional information about the functional state of the brain. In contrast, emergence, as indicated by VEC, was accompanied by numerous K complexes.
Another prominent feature of the pooled presentation was the gain in variance as the level of sedation decreased. One might expect that at the end of drug administration all volunteers would gradually recover from sedation, but after the study period, when volunteers were transferred to the poststudy observation area, clear signs of prolonged sedation were noted in some of them, which explains the observed broadening of range bars in Figure 1. Consistent with calculated propofol effect-site concentration for the monophasic protocol (Fig. 1a), VEC indicated recovery towards light sedation after the end of propofol infusion, whereas according to the NCT, recovery to light levels of sedation already occurred shortly after the end of the propofol injection. During the second propofol administration of the biphasic protocol, NCT, consistent with calculated propofol concentrations, indicated a lighter level of anesthesia than after the initial bolus, whereas the VEC scored a second phase of deep sedation that was characterized by the complete absence of K complexes and by a smooth return to wakefulness after the drug infusion was terminated. When K complexes started to reappear, the NCT was already displaying "wakefulness," which is difficult to explain with regard to the context-sensitive half-life time of 2–3 min reported for propofol. These observations indicated that an individual epoch-by-epoch comparison was necessary to fully analyze the discrepancies between the two methods.
Trend Analysis
Trend analysis of EEG patterns may be used to detect changes of patterns. There are two possible changes: (a) an increase of sedative effects (as recognized by an increase of slow EEG activity) and (b) a decrease of sedative effects (as indicated by the decrease of waves and the (re)appearance of faster 
and eventually ß and 
activity). A third combination featuring a steady pattern without changes was of special interest because some instances were noted when no trend could be expected because of the apparent near steady-state conditions during continuous infusion (Figs. 4a and b; Table 2, shaded entries). It is not clear why at the same time one method reported unchanged sedation, the other indicated continuous changes in the level of sedation. The two scoring techniques claim to measure similar parameters and, according to the manufacturer of NCT, used comparable criteria for assessment. One would naively expect that whenever one technique indicates a sedation (transition from wakefulness to sleep) or an arousal reaction (transition from sleep/sedation to wakefulness) the other method should report the same phenomenon, at least within a certain time frame. We arbitrarily set this timeframe to at least five consecutive epochs. By that time, any device that would be used to control depth of sedation would have triggered an action in response to the measured change (trend). We deliberately restricted the detection of discordant trends even further by ignoring any discordant trends when the overall amplitude would not exceed a full stage (=3 substages). When applying these strict criteria, we still identified nine discordant trends observed in nine different procedures (Table 2). Therefore, we conclude that NCT does not always reflect VEC (which is, according to the manufacturer, the basis of NCT classification). Not only are classification results different, but there are contrasting results in trend analysis. During clinical application of both methods, this would cause serious problems. As a consequence of the discordant trends, on the basis of one score the propofol rate would be increased for "deepening" of anesthesia, whereas on the basis of the other it would be decreased for "lightening" of anesthesia.
The observed differences may have been based on problems of the algorithm causing misclassification, or by the influence of artifacts. Unfortunately, in contrast to the VEC algorithm, the algorithm for NCT calculation is proprietary. Therefore, the search for mechanisms behind conflicting results must remain speculative. Despite this limitation, the study revealed that NCT does not always reproduce the results of VEC, which, according to the manufacturer, is the basis of NCT classification.
Artifacts
In some instances, the NCT did not display a classification result for >5 min. This problem has already been mentioned by Raymondos et al. (7) and Schneider et al. (37), who described it for the period of transition between wakefulness and unconsciousness. The time delay to the next calculated index value was between 0.3 and 9 min. This seemed to be an intrinsic problem of NCT, which may have been due to the presence of noise and signal artifacts. The current study was performed in volunteers; therefore, one may expect that the problem of artifacts may be even more pronounced during surgery, e.g., in the presence of electrocautery.
In conclusion, automated processing of EEG signals from a single frontal lead and the VEC of a multichannel EEG recording did not always match. In some instances, the differences were slight, but one method sometimes indicated deep sleep during propofol administration whereas the other reported light sleep that would prompt a dose adjustment. More importantly, trend analysis revealed nine situations within 24 propofol administrations when NCT and VEC, the basis of automated NCT scoring, suggested opposite trends (i.e., one scoring system indicated deepening, the other lightening of anesthesia). Observed scoring differences may have been in part because NCT is calculated from a single frontal channel, whereas VEC is based on analysis of a multichannel EEG. Unfortunately, the algorithm of NCT has not been published. Therefore, the search for reasons behind the differences remains elusive.
Footnotes
Accepted for publication July 5, 2007.
Supported, in part, by Astra-Zeneca, Wedel, Germany; Institute for Anesthesiology, Ludwig Maximilians University of Munich; Institute for Anesthesiology, Martin Luther University, Halle, Germany; Department of Anesthesiology, Klinikum rechts der Isar, Technische Universität München, Munich, Germany.
REFERENCES
C, eds. New aspects of high technology in medicine 2000. Bologna: Monduzzi Editore, 2000:285–91
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
