Secondary Logo

Journal Logo


State entropy and burst suppression ratio can show contradictory information

A retrospective study

Georgii, Marie-Therese; Pilge, Stefanie; Schneider, Gerhard; Kreuzer, Matthias

Author Information
European Journal of Anaesthesiology: December 2020 - Volume 37 - Issue 12 - p 1084-1092
doi: 10.1097/EJA.0000000000001312



Electroencephalographic (EEG) burst suppression reflects pathological and characteristic brain electrical activity of waxing and waning features that are most probably explained by very ‘deep’ levels of anaesthesia1 with common volatile or intravenous anaesthetics. In addition, experts have discussed further aspects associated with burst suppression such as cerebral hypoperfusion caused by phases of intra-operative hypotension.2 The occurrence of this EEG pattern seems to correlate with an increased risk for the patient to develop an adverse outcome such as postoperative delirium.3–6 Anaesthesia monitors that use the EEG to evaluate the level of anaesthesia usually calculate a specific index called burst suppression ratio (BSR) that reflects the amount of burst suppression within a certain time period. The bispectral index (BIS; Medtronic, Dublin, Ireland), the best known of these devices,1 linearly correlates with the amount of EEG suppression if the EEG is suppressed more than 40% in the observed time period.7 The monitor in our investigation was the EntropyTM Module (GE Healthcare, Helsinki, Finland).8 Its indices state entropy and response entropy do not show this linear relationship with the EEG suppression ratio (BSR), but rather present this information as an independent parameter.9,10 Hence, there is a possibility that contradictory state entropy/response entropy and BSR values can be displayed during surgery. This could mean a display of state entropy/response entropy representing a patient under ‘light’ or ‘adequate’ anaesthesia and the presence of burst suppression, or a patient with state entropy/response entropy indicating very ‘deep’ anaesthesia, but no burst suppression. Our analysis focused on the probability of such occurrences, based on a retrospective study of a large database, as well as one detailed and three supplementary case examples that help to explain these contradictory situations.

Materials and methods

Ethical approval for this study was provided by the Ethical Committee of the Medical Faculty at the Technical University of Munich, Germany on 6 November 2018 (Nr. 475/18 S-KK). It approved a retrospective investigation that included all patients of at least 60 years of age who had an intervention under general anaesthesia in our hospital. We focused on patients of 60 years or older because they seem more susceptible to EEG burst suppression11,12 and are at a higher risk of delirium after surgery.4,5

The relevant trend data, state entropy/response entropy and BSR from 8888 patients under general anaesthesia recorded between January and August 2018 were extracted from the patient data management system (PDMS). The PDMS is a software-based system routinely used in our department for anaesthesiology and intensive care. It is an online documentation tool that is used at the bedside or in the operating room. It collects information such as patient identification, age, height, body weight, vital parameters measured by the monitor, all parameters generated by the anaesthetic ventilator machine, all drugs that are given, laboratory findings and any other clinical marker or event during the procedure of anaesthesia. It enables continuous, error-reduced electronic data monitoring of each patient with immediate access to the information from every department in the facility. Regarding the resolution of the data and the EEG details, the EEG is recorded with the Entropy Module and the raw EEG is displayed on the patient monitor, together with the state entropy/response entropy and the BSR. Our PDMS does not allow for storage of the raw EEG. So, for the present analyses, we used the trend data provided by the Entropy Module. These data (state entropy/response entropy/BSR) were stored with a 10-s resolution. The haemodynamic parameters used to describe the patient are stored every 10 s (heart rate) or every 2 to 3 min (blood pressure). The information regarding the minimum alveolar concentration (MAC) of anaesthetic gases is also stored every 10 s.

For our investigation, the markers ‘clearance of case’ (start of procedure) and ‘end of case’ were of special interest to evaluate the occurrence of burst suppression during induction as well as intra-operatively. In order not to bias our analyses by invalid data, for example segments with no recorded trend data, we included only cases with more than 90% (arbitrarily chosen) valid data between the start and end of case. This left us with 2551 cases with good quality data and recorded burst suppression (BSR > 0). With respect to the surgical procedures, the 2551 cases included operations in various surgical disciplines except cardiac surgery (because this department is outsourced in another hospital). A histogram of the patient distribution in the different surgical disciplines is shown in supplemental Figure S1A, In all, 1503 patients received sevoflurane, 447 patients desflurane and one halothane as a maintenance anaesthetic. We used no gas in 85 patients and we could not obtain this information from our database in 515 patients. We conducted a short analysis to obtain primary results to establish whether the occurrence of burst suppression may differ between sevoflurane and desflurane. In order also to separate the induction drug from the maintenance drug, we defined the time point of induction along with 5 min as the (arbitrarily chosen) start for our analyses. This information is shown in supplemental Figure S1B, We also present the relationship between age and probability of burst suppression based in 3562 cases with good data quality and the maximum or median BSR relationship as supplemental data in Figure S2,

The Entropy Module contains two different analytical approaches to evaluate the ‘level’ of consciousness represented by state entropy/response entropy and the amount of burst suppression as BSR.8 In short, state entropy/response entropy are derived by evaluating the shape of the EEG power spectrum using Shannon entropy applied to the power spectrum, an approach also termed spectral entropy. State entropy and response entropy are estimated by application of the spectral entropy approach on EEG segments of variable length and different filter settings.8 State entropy is designed to reflect the cortical state of the patient. EEG frequencies from 0.8 to 32 Hz are included to calculate state entropy. In contrast, response entropy also includes information from the EMG-dominant part of the spectrum by considering frequencies up to 47 Hz.8 The difference response entropy – state entropy may serve as an indicator for EMG activation.8

The BSR algorithm is based on a nonlinear energy operator (NLEO) that is calculated from two different EEG frequency bands for very short (0.05 s) EEG segments. The decision of suppression EEG is based on NLEO being below a fixed threshold for more than 0.5 s in the absence of artefacts.8,13 The BSR is the percentage of 0.05-s epochs within the last 60 s indicating suppression.8

We used MATLAB (R2017a; The Mathworks, Natick, Massachusetts, USA) for our analyses. This included extraction of the relevant trend data and time (event) information from the PDMS data that were stored in .csv format as well as analyses of the state entropy/BSR pairs. The trend data were recorded every 10 s for most of the time with a small chance of missing values in between.

For the analysis of the entire data set, that is from the 2551 patients, we extracted the following information from each trend data set:

The linear regression between state entropy and BSR for all cases with BSR ≥ 5, those with BSR at least 20 and those with BSR at least 30. We also calculated a quadratic fit between state entropy and BSR for the cases with BSR > 0 and BSR ≥ 5 to compare our results with published findings.10

The highest state entropy value with a BSR more than 0.

The duration of state entropy above a defined threshold (>40, >50 and > 60) with a BSR > 0 and BSR ≥ 5.

The lowest state entropy with a BSR = 0.

The duration of state entropy below a defined threshold (<40, <30 and <20) with a BSR = 0 and BSR ≥ 5.

We selected the BSR ≥ 5 threshold to compensate for spurious low BSR values caused by nonburst suppression events. This evaluation of burst suppression duration helps to evaluate whether (contradictory) state entropy/BSR pairs occur during a dynamic phase and different time delays in state entropy and BSR calculation,14 or if they are of longer duration. Although the threshold setting was arbitrary, a state entropy of 40 presents the border between ‘adequate’ (SE = 40 to 60) and ‘deep’ anaesthesia according to the manufacturer. We further calculated the difference between response entropy and state entropy for all maximum state entropy values with BSR > 0 as well as all minimum state entropy values with BSR = 0. We performed this analysis as a quality marker of the recording. An increased difference indicates a higher discrepancy in the EEG high frequency content,8 information that may implicate dynamic anaesthesia phases or an increased chance of high frequency artefacts. We report our results based on an arbitrary threshold of response entropy – state entropy = 10. Differences of response entropy – state entropy of 10 or less as a target corridor have been used to titrate opioids to prevent movement to noxious stimulation (i.e. if the difference is bigger, the opioid should be adjusted).15,16 We present these results as supplementary data.

Categories of exemplar trends

In order to highlight the different scenarios of state entropy and BSR trends with possibly contradicting results, we present four case examples. Within the manuscript, we present one case that combined both contradicting situations, that is high state entropy with BSR and low state entropy without BSR. In the supplement, we present three more cases that showed extreme manifestations of contradicting indices.

Statistical analysis

We decided to use descriptive statistical approaches to describe the associations between state entropy (response entropy) and BSR. We present median [IQR] or mean ± SD when appropriate. For the analyses of BSR > 0 or BSR ≥ 5 duration above a certain state entropy threshold, we used the concept of cumulative probability plots. We supplemented the histograms with box plots to present the median and quartile information. We calculated the linear and quadratic fits using the ‘fitlm’ and ‘fit’ function. We conducted all analyses and created all figures using MATLAB.


We analysed 544 210 state entropy index values with BSR > 0 from 2551 patients. The median [IQR] age of our patients was 71 [66 to 77] years, 1015 patients were female and 1536 were male. Median anaesthesia duration was 145 [60 to 227] min. Figure 1a displays the distribution of the recorded state entropy and BSR combinations. The supplemental Figure S3A, and Figure S3B, are histograms showing the distribution of age and anaesthesia duration. The occurrence of burst suppression was dependent on age. We found a strong linear relationship (probability of BSUPP = 0.22 + 0.07∗age, r2 = 0.59, P < 0.001) between age and the probability of the patient showing burst suppression. We also found a significant relationship between age and maximum BSR (max. BSR = 35.22 + 0.26∗age, r2 < 0.004, P < 0.001)), but not for age and median BSR, as presented in supplemental Figure S2,

Fig. 1:
Heat map and box plots of (a) all (544 210) SE/BSR pairs and (b) the maximum SE with BSR > 0 for each of the 2551 cases. For low BSR, there is a wide range of corresponding maximum SE values. BSR, burst suppression ratio; SE, state entropy.

Regression between state entropy and burst suppression ratio

The linear relationship between state entropy and BSR can be described by the following equations.BSR5:SE=0.46*BSR+38.7,r2=0.66,n=423340;BSR20:SE=0.39*BSR+34.6,r2=0.75,n=216726;BSR40:SE=0.33*BSR+30.6,r2=0.81,n=100105.

The quadratic fit for the BSR ≥ 5 cases returned the coefficient SE = 41.8–0.7163∗BSR + 0.0031∗BSR2, r2 = 0.64 and for BSR> 0, SE = 42-0.7254∗BSR + 0.0032∗BSR2, r2 = 0.64.

Maximum (high) state entropy values with burst suppression

We found a wide range of maximum state entropy values with BSR > 0 or BSR ≥ 5 among the patients. The median [IQR] of the maximum state entropy values observed among the patients with BSR ≥ 5 was 48 [41 to 56] and with BSR > 0, it was 53 [45 to 61]. The corresponding BSR at maximum state entropy was 6 [5 to 9] for the maximum with BSR ≥ 5 analysis and 3 [2 to 5] for the maximum with BSR > 0. The heat maps in Fig. 1a and b show this distribution in detail. The analysis of our 2551 cases further revealed 2205 (86%) patients with positive BSR at state entropy of 40 or more, that is the presence of burst suppression at an index value that suggests ‘adequate’ level of anaesthesia. In 79% of the patients, the BSR was greater than 5; 29% (17%) of the patients had state entropy more than 60 indicating a light level of general anaesthesia together with a positive BSR (BSR ≥ 5) as depicted in Fig. 2a for BSR ≥ 5 or supplementary Figure S4, for BSR > 0. Despite these contradicting situations, a large number of SE/BSR pairs followed a linear trend of lower state entropy with higher BSR.

Fig. 2:
Description of the high SE (SE>40) characteristics with a BSR ≥ 5. (a) Over 85% of the patients showe situations with SE at least 40 with BSR ≥ 5. The fraction of patients decreased with increasing the SE threshold. (b) The median duration of SE at least 40 with BSR > 0 was 1.7 min. This duration decreased for higher SE values with BSR ≥ 5. The solid black line indicates the median, the dark grey area indicates the second and third quartile and the light grey area the 10th to 90th percentile. (c) Cumulative probability plots of episode duration, pooled for all patients, when SE was above a defined threshold (20, 30, 40 or 50) and BSR ≥ 5. BSR, burst suppression ratio; SE, state entropy.

Duration of high state entropy values in presence of burst suppression ratio more than 0

The median duration for each patient of state entropy at least 40 and positive BSR situations was 3.5 [1.0 to 8.7] min for BSR > 0 and 1.7 [0.7 to 4.5] min for BSR ≥ 5; 10% of the patients had these contradicting situations for 21.7 min or longer. The duration also decreased with increasing state entropy. For state entropy more than 60, the median was 0.3 [0.2 to 1] min for BSR > 0 and 0.2 [0.2 to 0.5] min for BSR ≥ 5, with 10% being longer than 3 min. Figure 2b and supplemental Figure S4B, contain the detailed information. The median episode duration, that is the consecutive time of contradiction, was 30 s (20 s for BSR ≥ 5) for a threshold of state entropy = 40. Further, 20% of these episodes were longer than 100 s (70 s for BSR ≥ 5) and every 10th episode was longer than 190 s (130 s for BSR ≥ 5). These episode durations decreased with increasing state entropy threshold as depicted in Fig. 2c and supplemental Figure S4C, Our analyses show that contradicting episodes, for instance state entropy at least 40 indicating ‘adequate’ anaesthesia but BSR > 0 or BSR ≥ 5, are frequent and can last up to several minutes.

Minimal (low) state entropy values without burst suppression

In 70% of patients, we observed situations with state entropy of 20 or less and no burst suppression indicated and in 46% we observed situations with state entropy of 10 or less and no burst suppression indicated (Fig. 3a). The median of the minimum state entropy without burst suppression from each patient was 21 [15 to 26] (Fig. 3b).

Fig. 3:
Description of the low state entropy characteristics without burst suppression indicated. (a) Ninety-six percent of patients showed situations with SE = 40 without BS, but 70% of patients also had situations with SE = 20 and no BS indicated, and 46% situations with SE = 10 and no BS indicated. (b) Median of minimum SE without BS was 21.0 (first and third quartile: 15 and 26). (c) The duration of low SE values without indicated BS decreased with increasing SE. At SE = 10, the median duration was 3.5 min [IQR 1.2 to 7]. It was 5.8 min [2.5 to 12] for SE = 20. The solid black line indicates the median, the dark grey area indicates the second and third quartile and the light grey area the 10th to 90th percentile. (d) Cumulative probability plots of episode duration, pooled for all patients, when SE was below a defined threshold (20, 30, 40 or 50) and BSR was 0. BS, burst suppression; SE, state entropy.

Duration of low state entropy with no indication of burst suppression

The duration of possibly contradicting low state entropy values without burst suppression per patient decreased with decreasing state entropy. The median duration for state entropy of 30 or less without burst suppression was 5.8 [1.2 to 28.0] min. In 10% of the patients, these situations occurred for 78.9 min or more. For a state entropy of 20 or less without burst suppression, the median duration was 0.8 [0.3 to 3.0] min and it was 9.8 min or more for 10% of the patients. Figure 3c highlights the relationship between low state entropy without burst suppression and duration. The episode length of these low state entropy and no burst suppression combinations also decreased with lower state entropy as depicted in Fig. 3d. The median episode length for state entropy of 30 and lower without burst suppression was 20 s. The episodes lasted longer than 60 s in 20% of the patients and longer than 120 s in 10% of cases. For the state entropy = 20 threshold, 50% of episodes were longer than 20 s and 20% were longer than 50 s. Every 10th episode was longer than 90 s.

Case scenario

Looking at some plots in more detail, we identified various interesting cases that demonstrate the paradox of some measurements. A 77-year-old male patient (height 172 cm, weight 83 kg, BMI 23 kg m−2, American Society of Anesthesiologists’ physical status 4) underwent a urological procedure under general anaesthesia induced with propofol and maintained with sevoflurane and sufentanil. This procedure lasted 156 min. Figure 4 shows the trend of the indices, drug concentration and haemodynamic parameters. After induction of anaesthesia and at the beginning of surgery, the state entropy showed low index values around 20 without BSR for a long duration (>1000 s). Towards the second half of anaesthesia, state entropy values higher than 40, indicating a level of ‘adequate’ anaesthesia with BSR more than 10 occurred.

Fig. 4:
Case scenario of a 77-year old patient with (very) low state entropy during induction without burst suppression and a state entropy indicating adequate anaesthesia (SE >40) during maintenance with positive burst suppression ratio at the same time. The blue lines represent the SE (light blue) and RE (dark blue). The orange area depicts positive BSR. The green trend reflects the heart rate and the grey dots the blood pressure. The black dots show the MAC (multiplied by 100 to match the y-axis: 1.2 MAC = 120 on the y-axis). BS, burst suppression; BSR, burst suppression ratio; MAC, minimum alveolar concentration of anaesthetic gas; RE, response entropy; SE, state entropy.

Anaesthesia was induced with propofol 100 mg and sufentanil 50 μg. In accordance with the amount of delivered anaesthetics, the state entropy reached very low indices (down to 15). During this induction period with state entropy indicating ‘deep’ anaesthesia, the BSR remained at 0. Later in the procedure, a rather high dose of sevoflurane (MAC up to 1.3) was delivered. Hence, the positive BSR may be valid but inconsistent with the state entropy indices ranging between 40 and 50. Heart rate and blood pressure were stable, ruling out a haemodynamic effect on cerebral perfusion and electrical activity. For three other examples of paradoxical state entropy and BSR measurements, please see the supplemental data including detailed medical background (Table S1, & S2, and supplemental Figures S6 to S8,


Our analyses allowed us to describe the relationship between state entropy and BSR in close detail. We focused on patients who were 60 years of age or older. Older patients seem at a higher risk for burst suppression than younger patients.11,12 In general, we observed a strong linear relationship between state entropy and BSR for BSR more than 40 with r2 value more than 0.8. This finding supports and strengthens the result presented previously.17 Further, we found a similar quadratic fit for the BSR > 0 and state entropy relationship as presented by Vanluchene et al.10 Apart from these strong relationships, we observed situations with contradicting BSR and state entropy indices.

Although for the noncontradictory situations decision making is easy, it is not as straightforward if state entropy and BSR contradict each other. If BSR = 0 and state entropy is within the recommended range, there is no need to change the anaesthetic level. In case of a BSR > 0 and low state entropy values, the patient most probably is in EEG burst suppression. Because this excessively deep level of anaesthesia seems associated with adverse outcomes,3,4 the anaesthetic level should be reduced. For contradictory situations in their most extreme manifestation, the state entropy was higher than 60, indicating a sedated or awake patient with BSR > 0. In most instances, the most probable explanation is the time delay of state entropy and BSR calculation.14 However, it cannot explain why state entropy more than 40 or more than 50 and BSR > 0 could persist for up to a few minutes. Low BSR values indicating only a very short duration of suppressed EEG may occur due to artefacts, but we also observed these situations for BSR more than 20, as presented in one of the example cases. Silent EEG periods of short duration could also reflect an EEG feature that may not be directly associated with burst suppression, but precede its occurrence18 and may cause low BSR values.19 Further, we observed periods of very low index values (state entropy < 30 or <20) without burst suppression being indicated. These cases may represent a state with dominant oscillatory activity in one frequency range, most likely in the delta range, as they are a marker for deep anaesthesia. A first step to correctly interpret the contradicting situation would be to check the raw EEG or the density spectral array as well as other parameters, such as the patient's haemodynamics.

Few publications have previously analysed the association between state entropy/response entropy values and BSR.9,17,20 On the basis of a study with 10 patients between 18 and 45 years of age who received propofol anaesthesia, the authors describe a good correlation of response entropy/state entropy with BSR at increasing levels of anaesthesia. The highest state entropy values with BSR > 0 were around 50.20 Another study, based on 70 patients aged between 18 and 65 years under propofol, thiopentone or sevoflurane anaesthesia, presented a very similar association between state entropy/response entropy and BSR. Maximum state entropy with BSR > 0 was lower than 50.9 Thus, our findings add more detail to these observations and present some more extreme cases, probably caused by the delays in state entropy and BSR calculation.14

Finally, we found a wide range of state entropy and BSR value combinations. The maximum state entropy value at which the BSR was positive could be as high as 89. Although these combinations are rare and of short duration, we still found numerous state entropy and BSR pairs wherein the state entropy was above 40, indicating ‘adequate’ anaesthesia as specified by the manufacturer. The exemplar trends are presented in case scenarios in the manuscript as well as in the supplementary files that highlight and describe contradicting situations within the clinical setting.

At the same time, we also observed a number of low state entropy values without BSR as derived from our cases that all contained burst suppression as detected by the monitor at some point.

In general, a comparison of BSR calculated by the Entropy Module and BSR calculated by another monitor is difficult, because the underlying algorithms leading to the BSR are different. In order to further understand the strengths and weaknesses of the single BSR indices to detect EEG suppression, different devices could be simultaneously placed on a patient's forehead. Also, simulated suppression episodes with different noise levels or underlying oscillatory activity could help to understand the limitations of these algorithms in more detail. Nevertheless, the clinical impact of their results should not differ. Burst suppression is a clinically relevant parameter and should be reliably detected.

Automatic detection of burst suppression – technical background

Although burst suppression presents an EEG signature that can be visually identified quite easily, automatic detection of this pattern and a distinction from artefacts may not be straightforward. The Entropy Module uses information derived by the NLEO algorithm from short windows,8,13 but the BIS uses an EEG amplitude threshold to detect suppression episodes.1 This can lead to episodes of contradictory results regarding the presence of burst suppression, as presented by Aho et al.21

Generally, there are two scenarios for false burst suppression detection: no burst suppression in the EEG, but burst suppression detected by the monitor; and burst suppression in the EEG, but not detected by the monitor. We present state entropy and BSR combinations as well as exemplar cases that support both scenarios. Because of the unavailability of raw EEG traces for these cases, we cannot draw final conclusions for the misleading values. However, for the BIS and the M-Entropy, certain cases have been described. Low EEG amplitudes during general anaesthesia may be misinterpreted as suppression episodes by the BSR algorithm of the BIS, as described in previous publications.22,23 Further, burst suppression may not be detected, although present, because of limitations in the detection algorithm. This may lead to drug overdosing.24,25 In general, automatic detection of burst suppression seems to present a challenge. Numerous different approaches to detect burst suppression automatically based on suppression EEG amplitude,1,26,27 on signal variance,28,29 on higher order spectra3 or time-domain parameters have been described. But so far, no absolutely reliable approach seems to exist and current systems probably underestimate the real rate of burst suppression.30 A potential approach to improve burst suppression detection could be to include information from the bursts and not just try to detect suppression. Parameters reflecting the burst activity could help to supplement the suppression information. The knowledge that different substances induce specific burst patterns31,32 could help to develop reliable burst and suppression detection. For instance, information from the EEG alpha band could help to predict the occurrence of suppression episodes, as described recently by Cartailler et al.18

Clinical implication of implausible state entropy/burst suppression combinations

Our findings show that, in contrast to the BIS, state entropy/response entropy indices – indicating a level of anaesthesia adequate for surgical intervention – can simultaneously occur with BSR > 0. This is not the first report of such cases. Hart et al.25 and Hagihira et al.33 also presented cases with contradicting combinations of Entropy and BIS indices.

Our example describes the combination of ongoing ‘high’ state entropy and BSR > 0. If such paradox combinations are detected, the anaesthesiologist should firstly consult the raw EEG to evaluate whether burst suppression patterns can actually be identified visually as well as rule out any EEG artefacts. Analysing the raw EEG in more detail may help to explain the origin of such values. In addition, haemodynamic trends such as the heart rate and blood pressure should be taken into account. Nevertheless, these special situations present the anaesthesiologist clinically with a remarkable challenge regarding appropriate management of the anaesthetic dosage. The fourth case in the supplement, contrary to the other three cases presented ‘very low’ state entropy indices without BSR being indicated. In the case of BIS, an index cannot be below 30 without a positive suppression ratio.7 However, for state entropy, this scenario is possible as highlighted in that example. Again, the anaesthesiologist should check the raw EEG in order to get a clearer picture of the situation. In this case, decreasing the depth of anaesthesia may not be necessary.

Burst suppression patterns seem to be associated with a potential risk of postoperative cognitive disorders such as a delirium.4,6 Negative outcomes like these are associated with further complications such as long-term neurocognitive decline, and higher rates of mortality.34,35 However, patients who are at risk of developing these outcomes may also develop these burst suppression patterns at quite low anaesthetic concentrations,2,12 whereas young and healthy patients with a ‘robust’ brain may be able to tolerate high doses without developing burst suppression, that is EEG patterns that may cause very low state entropy values with a BSR = 0. In very preliminary results, we were able to show that for both volatile anaesthetic gases, sevoflurane and desflurane, high BSR values can occur at low age-adjusted MAC concentrations. The corresponding heat maps are shown in supplementary Figure S5, Further research is required to establish whether these are patients with a ‘frail’ brain. We also confirmed previous results describing an increased occurrence of burst suppression with age.11

Implications for practical neurological monitoring

The use of devices for intra-operative neurological monitoring has become more popular over the last two decades. The principal focus was put on intra-operative events, such as awareness and burst suppression patterns, and their concomitant negative postoperative outcomes. Most of the results showed that, despite various (technical) limitations, EEG-based monitoring offers an easy and very helpful tool for anaesthesiologists to optimise anaesthetic management. Consequently, it is important that each junior doctor as well as anaesthetic specialist is trained professionally in the correct interpretation of raw EEG traces.


A number of limitations of our analyses must be considered. First, we used a pragmatic or simplistic approach to assess the general state entropy to BSR relationships without considering factors such as medical history, haemodynamic instability, accuracy of the applied electrodes or possible artefacts not detected by the monitors. As we did not have any EEG recordings, only the trend data, we cannot shed any light on the underlying EEG (or artefact) features that led to implausible state entropy/BSR pairs. However, we are confident that our approach helps to highlight issues which anaesthesiologists occasionally face in their daily practice, such as a wide dynamic range of possible state entropy/BSR pairs, especially for situations of high state entropy with burst suppression as well as low state entropy without burst suppression. The analysis of the duration of state entropy above a threshold and BSR > 0 also revealed that, at state entropy between 40 and 50, burst suppression can be indicated over a long period of time. Further, we hope that our exemplar cases show the possible clinical relevance of our findings.


We compared state entropy and BSR values during general anaesthesia in patients aged 60 years or older, and our exemplar cases describe the clinical relevance of the findings. In general, the state entropy does not show a strict relationship with the BSR as BIS does. Hence, a number of implausible BSR and state entropy combinations can occur. This does not mean that the Entropy Module performs worse than the BIS, but it requires more allowance for evaluation and interpretation. The precision of both monitors to detect burst suppression has to be assessed in further investigations. Our findings highlight the necessity for the anaesthesiologist to be trained professionally to check and interpret the raw EEG when using these monitoring systems to be able to react in situations of contradicting state entropy and BSR information. The take-home message of our work is that, when using these EEG-based indices, the anaesthesiologist should check the raw EEG to exclude artefact contamination and estimate the anaesthetic level, and if possible, also evaluate the spectral representation of the EEG.

Acknowledgements relating to this article

Assistance with the study: we would like to thank Dierk Lehmann for his assistance with the collection and editing of the patient data from the study.

Financial support and sponsorship: this work was not supported by external funding.

Conflicts of interest: none.

Presentation: preliminary data from this retrospective analysis were presented at the ‘Hauptstadtkongress für Anästhesiologie und Intensivmedizin - HAI’, 20 to 22 September 2018, Berlin.


1. Rampil IJ. A primer for EEG signal processing in anesthesia. Anesthesiology 1998; 89:980–1002.
2. Sessler DI, Sigl JC, Kelley SD, et al. Hospital stay and mortality are increased in patients having a ‘triple low’ of low blood pressure, low bispectral index, and low minimum alveolar concentration of volatile anesthesia. Anesthesiology 2012; 116:1195–1203.
3. Soehle M, Dittmann A, Ellerkmann RK, et al. Intraoperative burst suppression is associated with postoperative delirium following cardiac surgery: a prospective, observational study. BMC Anesthesiol 2015; 15:61.
4. Fritz BA, Kalarickal PL, Maybrier HR, et al. Intraoperative electroencephalogram suppression predicts postoperative delirium. Anesth Analg 2016; 122:234–242.
5. Radtke F, Franck M, Lendner J, et al. Monitoring depth of anaesthesia in a randomized trial decreases the rate of postoperative delirium but not postoperative cognitive dysfunction. Br J Anaesth 2013; 110: (Suppl 1): i98–i105.
6. Hesse S, Kreuzer M, Hight D, et al. Association of electroencephalogram trajectories during emergence from anaesthesia with delirium in the postanaesthesia care unit: an early sign of postoperative complications. Br J Anaesth 2019; 122:622–634.
7. Bruhn J, Bouillon T, Shafer S. Bispectral index (BIS) and burst suppression: revealing a part of the BIS algorithm. J Clin Monit Comput 2000; 16:593–596.
8. Viertio-Oja H, Maja V, Sarkela M, et al. Description of the Entropy algorithm as applied in the Datex-Ohmeda S/5 Entropy Module. Acta Anaesthesiol Scand 2004; 48:154–161.
9. Vakkuri A, Yli-Hankala A, Talja P, et al. Time-frequency balanced spectral entropy as a measure of anesthetic drug effect in central nervous system during sevoflurane, propofol, and thiopental anesthesia. Acta Anaesthesiol Scand 2004; 48:145–153.
10. Vanluchene AL, Vereecke H, Thas O, et al. Spectral entropy as an electroencephalographic measure of anesthetic drug effecta comparison with bispectral index and processed midlatency auditory evoked response. Anesthesiology 2004; 101:34–42.
11. Purdon P, Pavone K, Akeju O, et al. The ageing brain: age-dependent changes in the electroencephalogram during propofol and sevoflurane general anaesthesia. Br J Anaesth 2015; 115: (Suppl 1): i46–i57.
12. Besch G, Liu N, Samain E, et al. Occurrence of and risk factors for electroencephalogram burst suppression during propofol-remifentanil anaesthesia. Br J Anaesth 2011; 107:749–756.
13. Särkelä M, Mustola S, Seppänen T, et al. Automatic analysis and monitoring of burst suppression in anesthesia. J Clin Monit Comput 2002; 17:125–134.
14. Kreuzer M, Zanner R, Pilge S, et al. Time delay of monitors of the hypnotic component of anesthesia: analysis of state entropy and index of consciousness. Anesth Analg 2012; 115:315–319.
15. Mathews D, Cirullo P, Struys M, et al. Feasibility study for the administration of remifentanil based on the difference between response entropy and state entropy. Br J Anaesth 2007; 98:785–791.
16. Vakkuri A, Yli-Hankala A, Sandin R, et al. Spectral entropy monitoring is associated with reduced propofol use and faster emergence in propofol–nitrous oxide–alfentanil anesthesia. Anesthesiology 2005; 103:274–279.
17. Baulig W, Seifert B, Schmid ER, Schwarz U. Comparison of spectral entropy and bispectral index electroencephalography in coronary artery bypass graft surgery. J Cardiothorac Vasc Anesth 2010; 24:544–549.
18. Cartailler J, Parutto P, Touchard C, et al. Alpha rhythm collapse predicts iso-electric suppressions during anesthesia. Commun Biol 2019; 2:327.
19. Bresson J, Gayat E, Agrawal G, et al. A randomized controlled trial comparison of neuroSENSE and bispectral brain monitors during propofol-based versus sevoflurane-based general anesthesia. Anesth Analg 2015; 121:1194–1201.
20. Vanluchene AL, Struys MM, Heyse BE, Mortier EP. Spectral entropy measurement of patient responsiveness during propofol and remifentanil. A comparison with the bispectral index. Br J Anaesth 2004; 93:645–654.
21. Aho AJ, Kamata K, Jantti V, et al. Comparison of bispectral index and entropy values with electroencephalogram during surgical anaesthesia with sevoflurane. Br J Anaesth 2015; 115:258–266.
22. Hagihira S, Takashina M, Mori T, et al. Electroencephalographic bicoherence is sensitive to noxious stimuli during isoflurane or sevoflurane anesthesia. Anesthesiology 2004; 100:818–825.
23. Schnider TW, Minto CF, Gambus PL, et al. The influence of method of administration and covariates on the pharmacokinetics of propofol in adult volunteers. Anesthesiology 1998; 88:1170–1182.
24. Liang Z, Wang Y, Ren Y, et al. Detection of burst suppression patterns in EEG using recurrence rate. ScientificWorldJournal 2014; 2014:295070.
25. Hart SM, Buchannan CR, Sleigh JW. A failure of M-Entropy to correctly detect burst suppression leading to sevoflurane overdosage. Anaesth Intensive Care 2009; 37:1002–1004.
26. Jensen EW, Litvan H, Revuelta M, et al. Cerebral state index during propofol anesthesia: a comparison with the bispectral index and the A-Line ARX Index. Anesthesiology 2006; 105:28–36.
27. Chemali J, Ching S, Purdon PL, et al. Burst suppression probability algorithms: state-space methods for tracking EEG burst suppression. J Neural Eng 2013; 10:056017–156017.
28. Westover MB, Shafi MM, Ching S, et al. Real-time segmentation of burst suppression patterns in critical care EEG monitoring. J Neurosci Methods 2013; 219:131–141.
29. An J, Jonnalagadda D, Moura V, et al. Spatial variation in automated burst suppression detection in pharmacologically induced coma. Conf Proc IEEE Eng Med Biol Soc 2015; 7430–7433. doi:10.1109/EMBC.2015.7320109.
30. Muhlhofer W, Zak R, Kamal T, et al. Burst-suppression ratio underestimates absolute duration of electroencephalogram suppression compared with visual analysis of intraoperative electroencephalogram. Br J Anaesth 2017; 118:755–761.
31. Kenny JD, Westover MB, Ching S, et al. Propofol and sevoflurane induce distinct burst suppression patterns in rats. Front Syst Neurosci 2014; 8:237.
32. Fleischmann A, Pilge S, Kiel T, et al. Substance-specific differences in human electroencephalographic burst suppression patterns. Front Hum Neurosci 2018; 12:368.
33. Hagihira S, Okitsu K, Kawaguchi M. Unusually low bispectral index values during emergence from anesthesia. Anesth Analg 2004; 98:1036–1038.
34. Bellelli G, Mazzola P, Morandi A, et al. Duration of postoperative delirium is an independent predictor of 6-month mortality in older adults after hip fracture. J Am Geriatr Soc 2014; 62:1335–1340.
35. Saczynski JS, Marcantonio ER, Quach L, et al. Cognitive trajectories after postoperative delirium. N Engl J Med 2012; 367:30–39.


Supplemental Digital Content

Copyright © 2020 European Society of Anaesthesiology and Intensive Care. Unauthorized reproduction of this article is prohibited.