Share this article on:

Measuring depth of sedation in intensive care patients with the electroencephalographic Narcotrend Index

Weber, F.*,a; Steinberger, M.; Ritzka, M.¶,a; Prasser, C.; Bein, T.

European Journal of Anaesthesiology (EJA): February 2008 - Volume 25 - Issue 2 - p 123–128
doi: 10.1017/S0265021507001329
Original Article

Background and objective: The electroencephalographic Narcotrend Index was evaluated as a measure of sedation in mechanically ventilated intensive care unit patients. Narcotrend Index and conventional electroencephalography parameter values were compared to the Richmond Agitation-Sedation Scale and a simplified three-level sedation scale.

Methods: In all, 100 mechanically ventilated patients, admitted to the cardiac surgical intensive care unit after open-heart surgery, were enrolled in this prospective observational study. The Narcotrend Index was recorded while patients were either sedated by propofol infusion or without sedative medication while being weaned off the ventilator. Clinical assessment of the patients' level of sedation was performed by means of the Richmond Agitation-Sedation Scale by a single observer who was blinded to the Narcotrend Index.

Results: With the six-level Richmond Agitation-Sedation Scale, the prediction probability (P K) for the Narcotrend Index (0.81) was better than for all other electroencephalography parameters (P < 0.01) except for relative power in the β band (P K 0.75). Using the three-level sedation scale instead, P K values for the Narcotrend Index (0.88) and all electroencephalography parameters improved (P < 0.01), and the Narcotrend Index was now superior to all electroencephalography parameters. Narcotrend Index values were distributed among the various sedation levels with significant overlap.

Conclusion: When used for assessment of propofol sedation in postsurgical cardiac intensive care unit patients, the Narcotrend Index can distinguish between very light and deep sedation. With respect to differentiation between light and moderate or moderate and deep sedation, the Narcotrend Index was not reliable in this study.

*Sophia Children's Hospital, Erasmus University Medical Center Rotterdam, Department of Anesthesiology, Rotterdam, The Netherlands

University Hospital Regensburg, Department of Anaesthesia, Franz-JosefStrauss-Allee, Regensburg, Germany

University of Regensburg, Department of Neurology, UniversitätsstraΒe, Regensburg, Germany

Correspondence to: Frank Weber, Department of Anaesthesiology, Erasmus University Medical Center Rotterdam - Sophia Children's Hospital, P.O. Box 2060, 3000CB Rotterdam, The Netherlands. E-mail:; Tel: +31 463 1145; Fax: +31 463 6804

aFormerly: Department of Anaesthesia, University Hospital Regensburg, Franz-JosefStrauss-Allee 11, 93053 Regensburg, Germany.

Accepted for publication 19 June 2007

First published online 7 August 2007

Back to Top | Article Outline


Monitoring of sedation and guidance of hypnotic drug delivery in mechanically ventilated patients in the ICU in terms of a goal-directed therapy is critical for the success of the therapy [1]. The use of validated standardized scoring systems, usually involving the application of verbal or physical stimulation of the patient and observation of an arousal reaction [2,3], is gradually becoming the standard of care in many ICUs. However, arousal reactions to physical or verbal stimulation as a part of assessment of sedation with most of these scoring systems, among them the Richmond Agitation-Sedation Scale (RASS) [3], may lead the patient to distress or even result in deterioration of his actual physical status. Clinical scoring systems have via individual implementation and interpretation demonstrated significant inter-individual variability. In case of significant physical stimulation, as required to assign a patient to a level of very deep sedation, it remains uncertain whether a cortical reaction or a spinal cord reflex has been triggered. Ultimately, sedation-scoring systems provide only an intermittent assessment of the patient's hypnotic state.

The application of a monitoring device that continuously measures hypnotic depth without applying a stimulus could prevent the patient from assessment-induced arousal and continuously provide the caregivers in the ICU with the patient's current sedation level.

The Narcotrend® (MT MonitorTechnik, Bad Bramstedt, Germany) is an EEG monitor designed to measure the effects of anaesthetics and sedatives on the brain. The Narcotrend was developed at the University Medical School Hannover, Germany, and has received US Food and Drug Administration approval [4]. Narcotrend EEG processing leads to a variable called the Narcotrend Index (NI), a dimensionless number on a scale from 0 indicating very deep hypnosis to 100 indicating wakefulness (see Table 1). Detailed information about the Narcotrend algorithm has been given by Kreuer and colleagues [7]. Several recent studies either indicated the reliability of the Narcotrend for monitoring depth of hypnosis during general anaesthesia [8-10], or described the device as unreliable to detect consciousness [11] or the transition between awareness and unconsciousness [12].

Table 1

Table 1

Since there are no published data available concerning the performance of the Narcotrend in the ICU setting, the aim of this prospective observational study was to assess the utility of the NI as a measure of sedation in mechanically ventilated ICU patients.

Back to Top | Article Outline


In all, 100 consecutive adult, tracheally intubated and mechanically ventilated patients, admitted to the cardiac surgical ICU of the University Hospital Regensburg, Germany, after open-heart surgery, were enrolled in this study. Patients were excluded from the study if they were unable to communicate with the investigator due to hearing difficulties or because they did not understand German, if they suffered from significant neurological disease or if they were taking chronic medication known to affect the central nervous system.

According to departmental standards for the immediate postoperative course, patients either received continuous propofol infusion or were without sedative medication in order to be weaned off the ventilator within several hours after surgery. At the time of this study, no formal protocol for titration of propofol application existed in our cardiac surgical ICU, therefore propofol infusion rates were based on the clinical judgement of the intensivist in charge and the nursing staff. In daily clinical practice, sedation scales such as the RASS are not routinely used in our ICU. For postoperative pain relief, single doses of opioid piritramide (Dipidolor®; JANSSEN-CILAG, Neuss, Germany) were given as necessary.

This strictly observational study was approved by the institutional Ethics Committee of the University of Regensburg, Germany. Due to the fact that both EEG and clinical assessment of sedation are commonly performed non-invasive procedures in the ICU, the Ethics Committee waived the need to obtain written informed consent from the patients.

The EEG of each patient was recorded at 128 samples per second with 12-bit resolution, a 0.5 Hz high-pass filter, a 45 Hz low-pass filter and a 50 Hz notch filter [5], using the Narcotrend EEG monitor. Software version 4.0 was used in patients no. 1-50, and version 4.3 in patients no. 51-100. The only difference between the two software versions relates to the artefact handling of the system, the algorithm itself remained unchanged. Thus, we only compared the numbers of data sets being rejected by the two software versions.

After the skin was prepared with alcohol and abraded with gauze, two silver-silver chloride electrodes (Medicotest A/S; Olstykke, Denmark) were positioned on the left and right lateral parts of the patient's forehead with the maximum achievable distance and a third one on the mid-forehead, serving as a referential electrode. Electrode placement and skin preparation were performed until the electrodes' impedance was less than 6000 Ω. Besides EEG processing and computation of the NI, the Narcotrend records the ‘classical EEG parameters' (cEEG) total EEG Power (TP), relative power (RP) in the beta band (12.5-30 Hz) (RP β), the α band (7.5-12.5 Hz) (RP α), the theta band (3.5-7.5 Hz) (RP γ), the delta band (0.5-3.5 Hz) (RP δ), spectral edge frequency 95% (SEF) and median frequency (MF). All Narcotrend recordings were stored on disk for subsequent analysis.

Narcotrend EEG recording was started either after initial stabilization of the patients on the ICU after being transferred from the operation theatre or after propofol infusion was stopped for weaning. Assessment of sedation was performed by a single investigator (MS) who was blinded to the screen of the Narcotrend monitor, using the RASS. As described by Sessler and colleagues [3], all patients with RASS levels >−1 are awake. Positive RASS levels (+1 to +5) describe various degrees of restlessness, agitation and aggressive behaviour in awake patients. RASS level 0 describes a state of calm alertness, whereas RASS levels −1 to −5 describe the flow of hypnotic states from drowsiness to deep hypnosis. Patients being awake and restless or agitated (RASS levels 1-5) were a priori excluded from the study, because the Narcotrend was designed to measure degrees of sedation, but not to categorize agitation.

EEG data from three case milestones (Baseline; RASS-assessment; RASS-assessment +5 min) were analysed. In order to obtain information about the background sedation level, one single sample of mean NI and cEEG values from a 15 s artefact-free recording period immediately before RASS assessment, approximately 5 min after the beginning of Narcotrend monitoring (Baseline), was chosen for P K analysis. At this time, patients were untouched and undisturbed from acoustic input exceeding the usual sound-level of the ICU. In order to detect possible EEG alterations due to physical stimulation, two additional 15 s EEG data sets, recorded while performing RASS assessment, and 5 min later, were compared to the EEG data set chosen for P K analysis.

Back to Top | Article Outline


Statistical analyses were performed using SigmaStat, version 3.0 (Systat Software Inc., San Jose, CA, USA). As revealed by the Kolmogorov-Smirnov method, most data sets significantly varied from a pattern expected if they were drawn from a population with a normal distribution. Non-parametric procedures (Mann-Whithney rank sum test; Friedman repeated measures analysis of variance on ranks) were therefore applied as appropriate. In case of significant differences, especially in additional post hoc analysis, all pairwise multiple comparison procedures (Dunn's method) were applied. Comparison between the two software versions with respect to artefact rejection of data sets (number of patients without NI-calculation) was performed by means of Fisher's exact test. Results were considered significant when P < 0.05. Data are presented as mean ± SD or median (interquartile range, IQR) as appropriate.

The ability and accuracy of the NI and cEEG to distinguish between different hypnotic states were evaluated using the prediction probability (P K), which compares the performance of indicators having different units of measurements, as described by Smith and colleagues [13]. P K was calculated using a custom spreadsheet macro, the P K-MACRO, described and provided by Smith and colleagues [13]. We used the jackknife method to compute the standard error (SE) of the estimate. A P K value of 1.0 means that the parameter (e.g. the NI) predicts the clinical states (RASS classification) correctly 100% of the time. A P K value of 0.5 means that the prediction is no better than chance alone. A P K value of less than 0.5 indicates discordance more than concordance. P K values for the NI were calculated based on the RASS (6 levels) and on a more clinical-based three-level sedation scale, for which RASS levels were condensed as follows: levels −5 and −4 = deep sedation; levels −3 and −2 = moderate sedation; and levels −1 and 0 = awake.

Back to Top | Article Outline


A total of 100 patients (female/male-ratio 36/64; median (IQR) age 69 (62-74) yr), admitted to the cardiac surgical ICU after open-heart surgery, all tracheally intubated and mechanically ventilated, were enrolled. EEG data sets from three patients (6%) studied with the older Narcotrend software version 4.0 were classified by the system to be excessively artefact contaminated, which resulted in rejection from NI calculation. Software version 4.3 with its modified artefact handling calculated NI values for all 50 patients (Fisher's exact test n.s.).

Thus, EEG data from a total of 97 patients, under propofol sedation (n = 49, dosage 2.3 ± 1.1 mg kg−1 h−1) or without sedative medication while being weaned off the ventilator (n = 48), could be analysed.

Using the RASS with six levels as the clinical sedation scale, the overall P K value for the NI (0.81, SE 0.03) was significantly better than for all other cEEG parameters (P < 0.01) except for RP in the β band (P K 0.75, SE 0.03). Using the three-level sedation scale (awake-moderate-deep), P K values for the NI (0.88, 0.03) and all cEEG parameters improved significantly (P < 0.01), and the NI was now superior to all cEEG parameters. P K values are given in Table 2. NI values were distributed among the various RASS and sedation levels with significant overlap (see Fig. 1). The time course of EEG data over the three case milestones (Baseline, RASS assessment, RASS assessment + 5 min) is displayed in Table 3.

Table 2

Table 2

Figure 1.

Figure 1.

Table 3

Table 3

Patients under propofol sedation showed significant differences from patients in the weaning period with respect to NI values (propofol: 32(26-48); weaning: 77(37-97); P < 0.001), RASS scores (propofol: −5(−5 to −3); weaning: −1(−4 to 0); P < 0.001), P K values vs. RASS (propofol: 0.76 (SE 0.07); weaning: 0.85 (SE 0.04); P < 0.001), and P K values vs. the three-level sedation scale (propofol: 0.87 (SE 0.07); weaning: 0.89(SE0.04); P < 0.001).

Back to Top | Article Outline


This is the first study to evaluate the performance of the NI as a measure of sedation in mechanically ventilated ICU patients. P K values of 0.81 and 0.88, depending on fineness of the clinical sedation scale applied, suggest that the NI, with several limitations, is indicative of the patient's level of sedation. The results of this study furthermore show the superiority of the NI over conventional EEG parameters (see Table 2). However, there were at least two unexpected low NI values in awake patients: One patient with RASS level −1 had an NI value of 47, and a second patient with RASS level 0 had an NI value of 32 (see Fig. 1). In both patients, the Narcotrend system classified the EEG samples as artefact free. Thus we have no explanation for these findings. Except from these two outliers, all NI values in awake/drowsy patients were within the expected range [14,15]. NI values in deeply sedated patients were, with no exception, within a range indicating general anaesthesia. In the group of patients who were moderately sedated, we found a wide distribution of NI values, which significantly overlapped both the awake and the deep sedation range. Thus, we conclude that the NI can help to differentiate between drowsiness and deep sedation, but fails to identify patients who are moderately sedated.

This lack of identifying moderately sedated patients has also been reported for another EEG-derived index of hypnosis, the Bispectral Index (BIS; Aspect Medical, Natick, MA, USA) [16]. In a recent editorial, Nasraway even concluded that the BIS has failed to demonstrate consistent reproducibility as a monitor of sedation in heterogeneous populations of ICU patients [17]. Moreover, there is still an ongoing controversial debate about whether or not the BIS is a useful tool for sedation monitoring in the ICU.

The EEG Patient State Index (PSI; Physiometrix, North Billerica, MA, USA) has been investigated as a measure of sedation in ICU patients by Schneider and colleagues [18]. In their study, a P K value of 0.92 in patients receiving propofol/sufentanil sedation was interpreted to be indicative for the applicability of the PSI to quantify sedation in this particular pharmacologic setting. In relation to these data and implications, a study by Chisholm and colleagues [19] on comparison of the PSI and the BIS with clinical assessment of the level of sedation is noteworthy. The authors found a significant lack of both monitors with respect to their ability to reliably distinguish between light and deep sedation.

ICU nurses, who are responsible for the adjustment of sedative medication in most of the ICU's in Germany, usually allocate their patients to one of the following three categories:

  1. awake/drowsy;
  2. moderate sedation, with patient being responsive to verbal or mild physical stimulation;
  3. deep/excessive sedation, with patient being responsive to significant (painful) stimulation only or even being unresponsive at all.

Apart from specific rare cases, in which very deep sedation is mandatory, moderate sedation is the most common target in intubated ICU patients. This corresponds to RASS levels −2 and −3. With respect to the Prediction Probability, the number of levels of a sedation scale also has a significant impact. The more the sedation levels, the lower the P K value [20]. We calculated P K values for both the RASS and the simplified three-level sedation scale, and not unexpectedly [20], the P K value for the three-level sedation scale (0.88) was significantly higher than that for the RASS (0.81).

As revealed by the RASS, almost 50% of our study patients had received deep sedation. This is comparable to the data presented by Roustan and colleagues [16], who interpreted this as oversedation in 60% of their patients. Whether or not 47 of our patients were oversedated at the particular moment of NI monitoring is speculative, but there can be no doubt that without sedation assessment this condition would have remained unnoticed. However, we still have not found the ideal tool to assess our patients' hypnotic state. Subjective clinical scoring systems are only indicative for a specific moment, and by their nature introduce potential error via individual implementation and interpretation. Furthermore, depending on the intensity of potentially distressing painful physical stimulation, they are not necessarily only a measure of cortical activity but also of spinal cord reflexes. Therefore we cannot expect total agreement between the RASS and an EEG-based index of hypnotic depth. Interestingly, in this study NI and RP (%) β values, which were superior to all other EEG parameters with respect to P K values, remained unaffected by RASS assessment.

Unexpectedly, P K values for the NI were significantly better in patients during the weaning period than in patients during the propofol sedation period. During the weaning period, RASS values were significantly higher than during the sedation period. Whether or not this contributes to the improvement of the performance of the Narcotrend remains unclear. However, together with the evaluation of the continuous performance of the NI in ICU patients, this aspect should be urgently focused in future studies.

In contrast to clinical scoring systems, reliable EEG-based monitoring systems - which are not available at the moment - would offer the opportunity to provide the ICU caregivers a continuous measure of the patient's conscious state without the need for physical stimulation.

The NI, like the BIS and all other EEG-derived indices, requires significant improvement, especially with respect to its ability to identify moderately sedated patients.

This observational study was not designed to show that the Narcotrend could probably replace clinical scoring systems in the ICU. It was our intention to investigate whether the Narcotrend works in the ICU setting, which differs significantly from the operation theatre environment. Further studies are required to focus on aspects of long-term application of the EEG sensors, the impact of the clinical course of the patient (pyrexia, delirium, etc.) and co-medication, particularly muscle relaxants.

Despite the limitations mentioned above, it appears that the NI might be a useful tool in the ICU setting, particularly to prevent patients from potentially dangerous oversedation.

Back to Top | Article Outline


This study was solely supported by departmental funding (Department of Anaesthesia, University Hospital Regensburg, Germany). The Narcotrend monitor was provided on loan by the manufacturer, MT MonitorTechnik, Bad Bramstedt, Germany. MT MonitorTechnik was not involved in the design, conduct or analysis of this study.

Back to Top | Article Outline


1. Nasraway SA Jr, Jacobi J, Murray MJ et al. Sedation, analgesia, and neuromuscular blockade of the critically ill adult: revised clinical practice guidelines for 2002. Crit Care Med 2002; 30: 117-118.
2. Riker RR, Fraser GL, Simmons LE et al. Validating the Sedation-Agitation Scale with the Bispectral Index and Visual Analog Scale in adult ICU patients after cardiac surgery. Intensive Care Med 2001; 27: 853-858.
3. Sessler CN, Gosnell MS, Grap MJ et al. The Richmond Agitation-Sedation Scale: validity and reliability in adult intensive care unit patients. Am J Respir and Crit Care Med 2002; 166: 1338-1344.
4. Kreuer S, Wilhelm W. The Narcotrend monitor. Best Pract Res Clin Anaesthesiol 2006; 20: 111-119.
5. Schultz A, Grouven U, Beger FA et al. The Narcotrend Index: classification algorithm, correlation with propofol effect-site concentrations, and comparison with spectral parameters. Biomed Tech 2004; 49: 38-42.
6. Kreuer S, Bruhn J, Larsen R et al. Comparability of Narcotrend Index and bispectral index during propofol anaesthesia. Br J Anaesth 2004; 93: 235-240.
    7. Kreuer S, Biedler A, Larsen R et al. Narcotrend monitoring allows faster emergence and a reduction of drug consumption in propofol-remifentanil anesthesia. Anesthesiology 2003; 99: 34-41.
    8. Kreuer S, Biedler A, Larsen R et al. The Narcotrend - a new EEG monitor designed to measure the depth of anaesthesia. A comparison with bispectral index monitoring during propofol-remifentanil anaesthesia [in German]. Anaesthesist 2001; 50: 921-925.
    9. Kreuer S, Molter G, Biedler A et al. Narcotrend stages and end-tidal desflurane concentrations. An investigation during recovery from desflurane/remifentanil anaesthesia [in German]. Anaesthesist 2002; 51: 800-804.
      10. Wilhelm W, Kreuer S Larsen R. Narcotrend EEG monitoring during total intravenous anaesthesia in 4630 patients [in German]. Anaesthesist 2002; 51: 980-988.
      11. Russell IF. The Narcotrend ‘depth of anaesthesia' monitor cannot reliably detect consciousness during general anaesthesia: an investigation using the isolated forearm technique. Br J Anaesth 2006; 96: 346-352.
      12. Schneider G, Kochs EF, Horn B et al. Narcotrend does not adequately detect the transition between awareness and unconsciousness in surgical patients. Anesthesiology 2004; 101: 1105-1111.
      13. Smith WD, Dutton RC Smith NT. Measuring the performance of anesthetic depth indicators. Anesthesiology 1996; 84: 38-51.
      14. Schultz B, Grouven U Schultz A. Automatic classification algorithms of the EEG monitor Narcotrend for routinely recorded EEG data from general anaesthesia: a validation study. Biomed Tech 2002; 47: 9-13.
      15. Schultz B, Kreuer S, Wilhelm W et al. The Narcotrend Monitor: development and interpretation algorithms [in German]. Anaesthesist 2003; 52: 1143-1148.
      16. Roustan JP, Valette S, Aubas P et al. Can electroencephalographic analysis be used to determine sedation levels in critically ill patients? Anesth Analg 2005; 101: 1141-1151.
      17. Nasraway SA Jr. The Bispectral Index: expanded performance for everyday use in the intensive care unit? Crit Care Med 2005; 33: 685-687.
      18. Schneider G, Heglmeier S, Schneider J et al. Patient State Index (PSI) measures depth of sedation in intensive care patients. Intensive Care Med 2004; 30: 213-216.
      19. Chisholm CJ, Zurica J, Mironov D et al. Comparison of electrophysiologic monitors with clinical assessment of level of sedation. Mayo Clin Proc 2006; 81: 46-52.
      20. Weber Jensen E, Rodriguez B, Litvan H. P K value does depend on the fineness of the observer scale [Letter]. Acta Anaesthesiol Scand 2005; 49: 427.


      © 2008 European Society of Anaesthesiology