Clinical effectiveness and cost-effectiveness of depth of anaesthesia monitoring (E-Entropy, Bispectral Index and Narcotrend): a systematic review and economic evaluation

Population

Studies were included if they included patients who received GA for surgery, including adults and children (over the age of 2 years) in whom the technology is licensed. Studies involving patients receiving sedation in intensive care or high-dependency units, studies carried out in healthy volunteers and studies of non-surgical anaesthesia were excluded.

Diagnostic technologies

The diagnostic technologies included were E-Entropy, BIS and Narcotrend.

Comparators

Comparators included standard clinical monitoring for monitoring delivery of anaesthesia, including one or more of the following clinical markers: end-tidal anaesthetic gas concentrations (for inhaled anaesthesia); pulse measurement; heart rhythm; blood pressure; lacrimation; and sweating.

Outcomes

Studies were included if at least one of the following outcomes was reported:

• probability of intraoperative awareness

• patient distress and other sequelae resulting from intraoperative awareness

• recovery status (e.g. Aldrete scoring system)

• time to emergence from anaesthesia

• time to extubation

• time to discharge from the recovery room

• consumption of anaesthetic agents

• morbidity and mortality including postoperative cognitive dysfunction (POCD) from anaesthetic agents, pain-relieving drugs, antibiotics, antisickness drugs and muscle relaxants.

Study design

The review was limited to prospective controlled trials (once studies had been included in the systematic review, priority was given to RCTs unless no RCT evidence for relevant parameters was available in which case non-RCT data would be considered). Systematic reviews that met the inclusion criteria were retrieved in order to check their reference lists for potentially relevant studies but were not themselves evaluated (except for the Cochrane systematic review of BIS technologies,34 which was considered in more detail when conducting data synthesis: see Data extraction and critical appraisal methods).

Base case

A base-case analysis is presented for a general surgical population (at average risk of intraoperative awareness) and for a population assumed to be at high risk of intraoperative awareness. In the general surgical population, additional potential benefits (in terms of reductions in anaesthetic dose and reduction in anaesthetic-related complications) that may be associated with depth of anaesthesia monitoring are included in the base-case analysis, based on data from our systematic review of patient outcomes. Where data from the systematic review of patient outcomes were insufficiently robust, or where no evidence specific to the technology being considered was identified, data derived for other included technologies were used to populate the model.

Deterministic sensitivity analysis

Uncertainties around the probability, resource use and cost estimates, as well as effect parameters derived in the systematic review of patient outcomes, were investigated by applying ranges around the point estimates used in the base-case analysis. Where possible the ranges used in the deterministic sensitivity analyses were based on 95% confidence intervals (CIs) estimated for each input parameter. The method adopted was univariate sensitivity analysis – that is, varying one parameter at a time, leaving all other variables unchanged. This is to highlight the impact, if any, of each selected parameter alone on the cost-effectiveness results.

Scenario analysis

Scenario analysis was used to address uncertainty associated with the choice of data source adopted for parameter values in the base case and for variables omitted from the model.

Commercial-in-confidence information

This report contains reference to confidential information provided as part of the NICE appraisal process. This information has been removed from the report, and the results, discussions and conclusions of the report do not include the confidential information. These sections are clearly marked in the report.

Risk of bias in Bispectral Index trials

Table 4 reports a summary of the risk of bias judgements for the trials of BIS included in this systematic review (NB. The risk of bias judgements for the 31 RCTs in the Cochrane BIS review are not tabulated in this report, but are summarised in the text below).

In many cases the risk of bias in the trials was unclear because of limitations in reporting of methodological details. Uncertainty was greatest in relation to concealment of the random allocation process, where details were unclear in all but two trials. In the Cochrane systematic review of BIS, 12 of the 31 (39%) trials were considered to have adequately concealed random allocation, with most of the remainder judged as unclear.

Details of blinding of participants and trial personnel to trial arm were also generally unclear, as was the case of blinding of outcome assessors. In the Cochrane BIS review34 just over half of the studies were judged to be of a low risk of bias because of blinding of outcome assessors (17/31; 55%).

Random sequence generation was one of the domains where risk of bias was lowest. However, although all studies were reported to be randomised trials, in six trials (46%) the method of randomisation was not given. In the Cochrane systematic review of BIS34 just under half of the included studies (15/31; 48%) were judged to be of a low risk of bias because of adequate random sequence generation. Most of the remainder were unclear because of lack of details given in trial publications.

In general, there appeared to be low risk of bias in terms of selective reporting of outcomes, as could be judged from the details reported in the trial publications. This was also the case in the Cochrane BIS review. 34 Bias associated with incomplete outcome data was judged low in around half of the trials (and in just under half in the Cochrane BIS review,34 15/31; 48%). In the remainder it was unclear, and in one trial it was judged to be high because of an imbalance in the percentage of patients excluded from the analysis between trial arms. 62 In general, it was not considered that risk of other forms of bias were present. However, in one trial the risk was considered high because of the study being funded in part by the BIS module manufacturer. 45

The trials varied in terms of their sample sizes, from as low as 20 patients to over 6000. There were seven (46%)45,46,48,52,53,59,62 that included fewer than 100 patients and five (33%)48,51,61,63,64 that had between 101 and 200 patients. One trial included 921 patients,47 another included 530940 and another, the largest, included 6041 patients. 44 In the Cochrane BIS review34 the majority of trials included fewer than 100 patients (21/31; 68%). Seven trials (23%) included between 101 and 200 patients. Another study – the B-Unaware trial by Avidan and colleagues 2008 – included 1941 patients,27 and the largest included 2463 patients. 79 (NB. The Cochrane BIS review appears to count two publications relating to this single trial as two separate studies. One publication reports the main trial results,79 and a second publication focuses on recovery outcomes from the trial. 74)

Six (55%)40,44,46,49,51,62 of the 11 BIS trials reported a statistical sample size calculation based on a nominated primary outcome, although one of these trials reported that the number of patients chosen was arbitrary rather than being based on a statistical calculation. 49 The Cochrane BIS review34 did not comment on sample size power calculations in the studies included.

Six (55%)40,44,46,48,49,62 of the 11 BIS trials reported patient attrition. The attrition rate varied from 1.5%40 to 15% 49 of the total number of patients enrolled. Most of the studies reported the reasons for attrition, generally comprising exclusions from the analyses as a result of deviations from the study protocol. Given the nature of the procedure and the relatively short follow-up duration, loss to follow-up was rarely reported. In five (45%) studies it was reported by the authors that there was no attrition, or there did not appear to be any attrition. 45,47,51–53 Whether or not an intention-to-treat (ITT) analysis had been employed was rarely mentioned in the trial reports. Only two trials mentioned that patients had been analysed according to the procedure to which they had been randomised. 44,46

Five of the BIS trials disclosed information about funding. 40,44,45,49,51 Funding for two of these trials was provided by medical research funding organisations and/or hospital departmental grants. 44,51 The other three trials reported varying financial associations with BIS manufacturers. 40,44,49 The trial by Bannister and colleagues45 stated that Aspect Medical Systems supplied the BIS monitor, and that one author was employed by Aspect Medical Systems and another author was a paid consultant to Aspect Medical Systems. This funding therefore represents a conflict of interest. The trial by Kerssens and colleagues49 reported that Aspect Medical Systems did not financially support the study, but that the lead author had received an educational grant in support of her salary from Aspect Medical Systems, and one co-author was a paid consultant to Aspect Medical Systems. In the trial by Zhang and colleagues,40 Aspect Medical Systems provided BIS electrodes, but no further detail on funding was given. None of the other BIS trials stated or appeared to have any major conflicts of interest. The Cochrane BIS review34 did not report funding details of the included trials, or whether or not any of the trials had conflicts of interests.

Risk of bias in E-Entropy trials

Table 5 reports a summary of the risk of bias judgements for the trials of E-Entropy included in this systematic review.

The risk of bias in the E-Entropy trials was unclear in many cases because of limitations in the reporting of methodological details. Uncertainty was greatest concerning allocation concealment and the blinding of participants and personnel, which were not adequately reported in any of the seven E-Entropy trials.

Risk of bias because of random sequence generation was considered low in four of the trials, in which sequences were generated either by computer56,57,61 or by drawing lots. 62 Risk of bias because of random sequence generation was deemed unclear in the remaining three trials, which provided no information on the method of sequence generation.

The method of allocation concealment was considered to pose unclear risk of bias in all seven of the trials, either because no relevant information was reported58,61,62 or sealed envelopes were used for allocation codes, but it was not stated whether or not the envelopes were opaque. 54–57

Anaesthetists who administered anaesthesia according to standard clinical monitoring were blinded to E-Entropy values. However, none of the studies unequivocally reported that study participants and personnel were blinded to the study groups. The risk of bias because of inadequate blinding in each of the E-Entropy studies was therefore judged to be unclear.

In three of the seven E-Entropy trials, the risk of attrition bias because of analysis of incomplete outcome data was considered low, as exclusions were a minor proportion of the sample size,54 or were generally balanced between groups with generally similar reasons given,61 or the analysis was conducted by ITT with no discernible attrition. 56 Two trials were considered at high risk of attrition bias because the rate of attrition was ≥ 10% in at least one of the study arms, and not balanced across the arms. 58,62 The remaining two trials were judged to have unclear risk of attrition bias because of incomplete outcome data, either because attrition was not reported at all55 or it was not reported separately by study arm. 57

The risk of bias because of selective reporting of outcomes was judged to be low for six of the seven E-Entropy trials, as there was no indication within the primary publications that more outcomes had been measured than were subsequently reported (in general, there was concordance between the outcomes specified in the methods and results sections of the publications). In the remaining trial,57 risk of bias from selective reporting was considered unclear as several outcomes were reported narratively without any supporting quantitative data that could be checked by the reviewers.

One of the E-Entropy trials56 reported that no external funding was used, and one trial62 did not report whether or how the work was funded. Two trials were funded by non-commercial sponsors, which were a university54 and a national science organisation. 58 The remaining three E-Entropy trials were supported by the E-Entropy device manufacturer (GE Healthcare; formerly Datex-Ohmeda), either through provision of equipment alone55,61 or through provision of equipment, funding and also technical support. 57 The authors of this latter trial57 included a research engineer, research scientist and chief scientist of the device manufacturer and two medical advisors to the device manufacturer. These three trials that involved support from the device manufacturer could be at risk of bias because of conflict of interests. The study that involved the most extensive links with the manufacturer57 was deemed by the reviewers to be at high risk of bias because of a high likelihood of conflicting interests. In the four E-Entropy trials that were not supported by the E-Entropy device manufacturer, three did not refer to conflict of interests54,58,62 and one stated that no conflicts were disclosed. 56

The seven E-Entropy studies were published during 2005 to 2010 and ranged in their total sample size from 50 to 335 patients. 54–58,61,62 Five of the trials involved a two-arm comparison of E-Entropy against standard clinical monitoring. 54–58 One trial involved a three-arm comparison of BIS, E-Entropy and standard clinical monitoring. 61 The remaining trial was a three-arm comparison of E-Entropy, E-Entropy and BIS, and standard practice. 62 The number of patients randomised per arm ranged from 25 to 40 in six trials. 54–56,58,61,62 In the seventh (largest) trial, only the number per arm after attrition (160 patients) was reported. 57

Only one of the E-Entropy trials did not report a sample size calculation. 58 Three trials calculated the sample size needed to detect a specified percentage difference in anaesthetic consumption for sevoflurane54,61 or propofol. 62 The remaining three trials calculated the sample size needed to detect differences in patient recovery from anaesthesia, namely the time to eye opening,55 time to awakening (not defined)56 or the time to response to a verbal command. 57

Overall, the range of attrition in the trials was 0–11% of the total population per trial, or 0–17% of the population per study arm. Attrition appeared to be zero in one trial,56 and was not reported in one trial. 55 Among the remaining five trials, reasons for attrition were clearly reported separately by study group in two trials;54,61 were reported only for aggregated data across study groups in one trial;57 were vaguely specified as resulting from ‘technical problems’ in one trial;54 and were not specified in the remaining trial. 58

An analysis by ITT was explicitly reported in one trial and appears valid as no attrition was discernible in the study report. 56 Another trial55 did not explicitly mention ITT analysis but appeared to have used an ITT approach, as it was stated that all patients were included in the final analysis, although attrition was not reported. A third trial54 analysed nearly all the randomised patients [only 1/40 per group (2.5%) were excluded], which may be considered close to an ITT approach. The remaining four trials57,58,61,62 did not follow the ITT principle as their analyses excluded from 4% to 17% of the randomised patients per study arm.

Risk of bias in Narcotrend trials

Table 6 reports a summary of the risk of bias judgements for the trials of Narcotrend included in this systematic review.

In many cases the risk of bias in the trials was unclear because of limitations in reporting of methodological details. Uncertainty was greatest in relation to concealment of the random allocation process and blinding of participants and personnel, where details were unclear in all four trials. 59,60,63,64

Both the method of random sequence generation and blinding of outcome assessment were unclear in two trials,59,60 with low risk of bias for these domains in the other two trials. 63,64 Risk of bias because of incomplete outcome data was low in all but one trial in which details were unclear. 60

In general, there appeared to be low risk of bias in terms of selective reporting of outcomes, as could be judged from the details reported in the trial publications. Other sources of bias were reported for only one study where the paper was translated from Chinese to English prior to publication and it is unclear whether or not any checks were made to ensure fidelity of the published version to the original work. 59

The trials were conducted between 2003 and 2010 and trial sizes ranged from 120 patients63,64 to 48 patients60 and 40 patients. 59 All but the smallest study reported the use of a sample size calculation. No attrition was reported for three trials59,63,64 and these studies conducted ITT analyses. The fourth trial60 reported attrition although not by study group, and analyses did not include all patients who started but it is unclear whether or not attrition happened pre or post randomisation. All four trials59,60,63,64 did not report any conflict of interest. Two studies63,64 stated that the study was solely supported by departmental funding, one59 did not report any details of the sponsor and the fourth60 reported that the study was supported by a pharmaceutical company and a university institutional research grant.

Study populations

Five of the 11 BIS trials were conducted in children, with mean ages of between 4 and 6 years, and age ranges from 2 to 18 years. 45,46,51–53 The remaining six studies were conducted in adults,40,44,47–49,62 with mean ages ranging across the studies from 43 to 64 years. One study was conducted to investigate POCD in an elderly population, defined as > 60 years (no further age information given) (conference abstract). 47 All of the trials included in the Cochrane BIS review34 studied adult patients (the review's inclusion criteria specified adults over the age of 18 years).

All of the trials included mixed-sex populations. Generally, there was an even mix of males and females in the trials, though there was a higher percentage of males (i.e. > 60%) in three studies. 46,48,51 One study did not report the sex of the included patients. 47

All but one of the studies reported patients' weight. 47 The majority of studies reported weight in kilograms, ranging from a mean of 68–91 kg in the adult studies, and of 17–28 kg in the children studies. In addition to reporting weight in kilograms, one trial also reported body mass index (BMI), which was between 28 and 30 kg/m². 49 Another trial reported weight only in terms of BMI, with a mean of 30 kg/m², indicating an overweight/obese population. 44 The Cochrane BIS review34 included one study of obese patients.

Racial origin was reported in only one trial, in which the population was predominantly (> 80%) classified as white. 44 The countries in which the trials were conducted included the USA,45,49,52,53 USA/Canada,44 China,40,47,51 Germany,62 Egypt48 and India. 46 In the Cochrane BIS review34 the majority of studies were conducted in Europe or the USA. Seven of the trials were conducted in single centres,45,46,48,51–53,62 with one trial taking place in two centres,47 another taking place in three centres,44 one trial taking place in 13 centres40 and a trial not reporting the number of centres. 49

The type of surgery reported in the adult trials varied: open heart,44 major non-cardiac,47 major orthopaedic,49 orthopaedic,62 and elective moderate abdominal surgery. 48 The surgical procedures in the trials of children included tonsillectomy and/or adenoidectomy,45 urogenital/urological surgery46,51 and dental rehabilitation. 52,53

Only two of the trials reported patient risk factors for awareness. 40,44 To be included in the trial by Avidan and colleagues44 patients had to be at high risk for intraoperative awareness, demonstrating one or more of the following risk factors: planned open heart surgery; aortic stenosis; pulmonary hypertension; use of opiates; use of benzodiazepines; use of anticonvulsant drugs; daily alcohol consumption; ASA status 4; end-stage lung disease; history of intraoperative awareness; history of, or anticipated, difficult intubation; cardiac ejection fraction of < 40%; and marginal exercise tolerance. The trial by Zhang and colleagues40 included patients receiving TIVA, which they cited as a risk factor for intraoperative awareness. The Cochrane BIS review34 included four trials that were classified as including patients at high risk of intraoperative awareness. 27,78,79,82

The eligibility criteria employed by the trials generally excluded patients with significant comorbidities, or factors that may interfere with EEG readings, including epilepsy, cerebrovascular disease, dementia, treatment with opioids and antipsychotic medication, and illicit drug use. Two of the studies permitted inclusion of children with mild cerebral palsy without significant neurological deficit. 52,53 The trials included in the Cochrane BIS review34 also generally excluded patients with the above factors. Some of these trials also excluded patients considered obese, or patients with diabetes or impaired renal or hepatic function.

The ASA physical status classification of the patients in the trials was generally between I and II, indicating that they were generally healthy, with only mild disease. In three of the trials the ASA status was not reported45,47,48 (although in one of these trials the inclusion criteria specified patients had to be within I–III48). In one trial the proportion of patients with ASA status I–II was 50%, and the remainder of patients were classified as III (severe systemic disease). 49 There was one trial in which patients were predominantly classified as III–IV (IV being classified as a patient with severe systemic disease that is a constant threat to life). 44

Technologies

The trials varied in the level of detail given on the BIS module and monitors used. Two studies did not provide any information other than that a BIS module was used. 44,47 Most commonly reported was the BIS Monitor Model A-2000 as mentioned in four trials. 40,45,48,62 In one this was described as: ‘IP X 2’46 in another ‘version XP, software version 4.0’62 and in the third trial using Aspect Medical Systems ‘Software program Datex-Ohmeda S/5 Collect (v4.0)’ (Aspect Medical Systems Inc., Norwood, MA, USA). 48 One trial used BIS (version 3.3, Aspect Medical Systems) using an A-1050 EEG monitor,45 while another used BIS monitor (XP, algorithm 3.4; Aspect Medical Systems). 49 A further two trials reported using BIS (Aspect Medical Systems), but gave no further information on the software version or the monitor used. 52,53 Although most studies reported using Aspect Medical Systems BIS, one trial reported using the BIS monitor as manufactured by Phillips but using ‘Aspect Medical Systems' XP platform technology’. 51 Given the variability in reporting it is not clear how comparable the trials are in terms of the software and BIS algorithms used, which may have implications for the interpretation of the results of the trials.

All of the trials reported the target BIS values to be achieved during anaesthesia. In five trials the target was 40–60. 40,44,45,47,51 In one of these trials the target was increased to 60–70 during last 15 minutes of surgery. 45 In the remaining trials the target values were higher: 45–60;46 50 during maintenance (target value of 60 to facilitate rapid emergence from anaesthesia 15 minutes before expected end of surgery);62 50–60;48,49 55–65;53 and 60–70. 52

Although all of the trials compared BIS against standard clinical monitoring, the monitored parameters varied. Only one trial measured ETAC in order to detect possible intraoperative awareness. 44 Audible alarms sounded if the age-adjusted MAC fell outside of 0.7 to 1.3. The remaining nine trials used clinical signs to guide anaesthetic use. In general, a combination of signs were monitored in each trial, most commonly: blood pressure;46,48,49,52,62,94 heart rate;48,49,52,53,62 surgical stimulation;52,53 sweating;62 tear production;62 and movement. 62

Two trials did not explicitly define which signs were monitored other than that they were clinical signs and haemodynamic changes. 45,47 A further trial mentioned that the aim of standard clinical monitoring was to maintain haemodynamic stability while avoiding patient movement and achieving a rapid recovery. 51

Some of the trials reported that clinical signs were also monitored in the BIS arm, suggesting that adjustment of anaesthesia was based on signs of inadequate anaesthesia as well as BIS values. 48,52,53,62 For example, in one trial48 changes in anaesthesia were guided by the presence of clinical signs in relation to the BIS value. If the patient exhibited hypertension or tachycardia and the BIS was > 60 then sevoflurane was increased. If BIS was in the target range of 50–60, then fentanyl was given. If BIS was < 50 then sevoflurane was decreased and the patient checked for lack of analgesia. In the one trial that used ETAC as the comparator to BIS,44 it was stated that both forms of monitoring were used as part of structured protocols. It was not intended that these protocols would prescribe or restrict the use of anaesthetic agents. Practitioners were able to increase or decrease anaesthetic administration at their discretion if a patient's condition was haemodynamically unstable. The protocols were designed to increase vigilance and to provide warnings that patients might be experiencing awareness. Some trials did not explicitly report whether or not clinical signs were monitored in the BIS arm, and it is possible that in these studies anaesthesia was adjusted based on BIS monitoring in conjunction with changes in clinical signs.

All trials reported that a BIS monitor was used in the standard clinical monitoring arm, but that the values were hidden from the anaesthetist, for example by placing it out of their line of sight, or using a curtain or cover, and also switching off any audible alarms.

The majority of trials did not explicitly report where or when monitoring commenced and ceased. Where details were provided, monitoring started prior to anaesthesia induction45,46 and in the operating theatre. 46,51,62 Three studies reported cessation of monitoring: until patients achieved discharge criteria from the recovery room (Steward score of 6)46 and until discharge from the PACU. 52,53

The training and experience of the anaesthetist in using BIS was rarely mentioned in the trials. The trial by Avidan and colleagues44 reported that summaries of BIS and ETAC protocols were given to the practitioners to provide education and to increase adherence. Furthermore, signs were affixed to anaesthesia machines to remind practitioners to check BIS/ETAC and consider patient awareness. One of the trials mentioned that the anaesthetist was experienced, but provided no further information. 62

Anaesthetic agents and protocols

Five of the trials reported that an inhaled general anaesthetic was used for both induction and maintenance. 44,45,51–53 In all but one of these trials sevoflurane was the inhaled anaesthetic used. 53 Two of these trials also gave nitrous oxide in oxygen. 45,51 In the fifth trial patients either received isoflurane, sevoflurane or desflurane. 44 Three trials reported that both i.v. and inhalational general anaesthetic were used. 47–49 In two of these propofol was used for induction of anaesthesia and sevoflurane was given for maintenance. 48,49 The third trial implied that both propofol and an inhalational anaesthetic were given, but did not provide any further detail. 47 Three trials reported that propofol was given for both induction and maintenance of general anaesthesisa. 40,46,62 One of these also used nitrous oxide in oxygen during the maintenance period. 46

Only one trial stated that regional anaesthesia was used, although no information was provided on which agent was used. 62 One trial mentioned that regional anaesthesia was used for postoperative pain management. 49 In the remaining nine trials44–48,51–53,62 it was either reported that regional anaesthesia was not used or the use of regional anaesthesia was not stated.

Use of analgesia at various points during surgery was reported by seven of the trials, including fentanyl,49,51–53 fentanyl or morphine45,46 or remifentanil (during induction). 62 One trial reported that analgesia was used at the discretion of the anaesthetist. 40 In three trials the use of analgesia was not stated. 44,47,48 Premedication with midazolam was used in seven trials. 40,44–46,52,53,62 In two of these trials ketamine was also used as premedication. 52,53

Muscle relaxants were used in seven of the trials, including atracurium,46,48 cisatracurium,63 vecuronium bromide49 and rocuronium bromide. 52,53 One trial did not specify which agent was used. 40

Duration of anaesthesia was reported by five of the BIS trials46,48,49,51,62 and ranged from a mean of 40 minutes (paediatric urological surgery)51 to 126 minutes (major orthopaedic surgery in adults). 49 In the trials featuring adults, duration of anaesthesia was, in general, between 100 and 120 minutes. Duration of surgery was reported by seven of the BIS trials,40,45,46,48,51–53 and ranged from around 30 minutes (in children undergoing tonsillectomy and/or adenoidectomy)45 to 160 minutes (children undergoing dental surgery). 52 Not all trials reported both duration of anaesthesia and duration of surgery.

Outcomes

Table 7 illustrates the distribution of outcomes reported by the trials included in this systematic review. The table also shows the frequency of the outcomes in this review, the Cochrane BIS review34 and the grand total for both reviews.

The most commonly reported outcome was anaesthetic consumption (n = 30 trials), followed by recovery outcomes such as time to extubation (n = 26 trials); time to eye opening (either spontaneously or in response to command) (n = 22 trials); and time to discharge from the PACU. Intraoperative analgesic consumption was reported in 11 trials.

Adverse outcomes were less commonly reported, such as PONV (n = 3 trials); and emergence delirium (n = 1 trial59). One trial, by Leslie and colleagues,50 reported stroke, myocardial infarction, mortality for all surviving and available patients 30 days post operation. This is a long-term follow-up (median = 4.1 years) publication of the B-Aware trial [NB. A publication of the short-term results of this trial by Myles and colleagues 200479 (primary outcome: intraoperative awareness) was included in the Cochrane BIS review34].

Six of the 11 BIS trials40,44,46,49,59,62 included in this systematic review specified a primary outcome measure. In two trials the primary outcome measure was anaesthetic consumption,46,62 and in another trial the primary outcome measure was time to first movement response. 59 In the other three trials the primary outcome measure was intraoperative awareness. 40,44,49

In the trial by Avidan and colleagues44 – which recruited patients classified as at high risk of intraoperative awareness – the incidence of definite intraoperative awareness was the primary outcome measure. The incidence of definite or possible awareness was a secondary outcome. Awareness was assessed by a modified Brice questionnaire (references cited), and assessments were made 72 hours after surgery and 30 days after extubation. Patients who reported memories of the period between ‘going to sleep’ and ‘waking up’ were contacted by a different evaluator, who asked additional structured questions. Responses to the questionnaire from patients who had reported memories were reviewed by three independent experts, who determined whether the reported event involved definite awareness, possible awareness or no awareness. Where there was a difference in judgement over an awareness episode a fourth expert made the final determination. This study was designed specifically to evaluate the effects of BIS on intraoperative awareness, and to overcome methodological limitations of a previous single-centre trial by the same investigators (the B-Unaware trial27 – included in the Cochrane BIS review34) by including a study sample sufficiently large enough to detect a relatively rare outcome such as awareness.

The trial by Zhang and colleagues40 also reported incidence of confirmed awareness, or possible awareness, using a Brice questionnaire. Assessments were made on the first and fourth day following surgery. An independent evaluating committee was used to verify cases of awareness. The patients in this trial were noted to be at increased risk of intraoperative awareness after receiving TIVA.

The trial by Kerssens and colleagues49 measured explicit awareness, via a patient interview, as well as implicit awareness, via a word recognition test. This is the only trial identified by the current systematic review that measured implicit awareness. The underlying hypothesis was that intraoperative memory could occur either because of insufficient anaesthetic or stress-induced learning mechanisms during unconsciousness (i.e. intraoperative memory could be dependent on and/or independent of depth of anaesthesia). Six hours after surgery, patients were interviewed using questions similar to the Brice interview, consisting of five questions, with additional questions asked as necessary. Following the interview a recognition memory test was performed. During anaesthesia, sequences of pre-determined neutral words were played to patients through headphones. The postoperative memory test involved playing pre-determined combinations of words that had been used during anaesthesia, and distractor words, to patients though headphones. Patients were instructed to listen to each test sequence and select the word played during surgery, or to guess if necessary. The main analysis of this study was the effect of study group assignment on recognition memory test performance, but given the low incidence of explicit recall (awareness) the study was not powered to detect differences in explicit recall. An arbitrary sample size of 100 patients was chosen to assess recognition memory.

Intraoperative awareness was also reported as a non-primary outcome by three other BIS trials included in this systematic review. 48,51,62 In these trials, awareness was one of a number of outcomes measured, and patients were not identified as being at particular risk. Awareness was assessed by a patient interview administered at various times up to 3 days post operation. In the trial by Ellerkmann and colleagues62 interviews took place on the first and third postoperative days, in the trial by Kamal and colleagues48 interviews took place on the first, second and third days postoperatively, and in the trial by Liao and colleagues51 the timing was not specified. Little detail of the interviews was given other than that ‘patients were questioned for recall of events, hearing vague sounds, feeling surgical instruments or dressing application, or dreaming’48 or patients were asked ‘whether they could recall any event or dreaming during the intraoperative period’51 or that a ‘standardised interview’ was used (reference cited). 62

The Cochrane BIS review conducted a meta-analysis of explicit intraoperative awareness, which included four RCTs. 27,78,79,82 The Cochrane review also included a further eight trials61,63,66,83,84,87–89 that reported explicit intraoperative awareness, but the review did not classify these as featuring patients at higher risk. They were not included in any meta-analysis and the impact on awareness was not commented on by the Cochrane review. The Cochrane BIS review did not report whether any of the included trials measured implicit awareness or assessed awareness during surgery using techniques such as the isolated forearm technique.

Intraoperative awareness

Table 8 gives the results of the six trials included in this systematic review which measured the impact of BIS monitoring on explicit intraoperative awareness, as assessed by patient interview.

No cases of awareness were reported at all in three trials,48,51,62 and a very low number of cases were reported in a fourth trial. 49 As stated earlier, these trials were not specifically designed to detect the effect of depth of anaesthesia monitoring on awareness, and therefore are unlikely to have sufficiently large enough sample sizes for relatively rare awareness events. In the trial by Avidan and colleagues,44 which included patients classified at higher risk for intraoperative awareness and was statistically powered for this outcome, there was a higher percentage of both definite awareness, and of definite or possible awareness cases, in the group who received BIS monitoring than the group who had standard clinical monitoring. However, these differences were not statistically significant. Avidan and colleagues44 also reported patient distress and sequelae resulting from intraoperative awareness, as a post hoc secondary outcome. Distress was measured using the Michigan Awareness Classification tool (reference supplied) and was characterised by reports of fear, anxiety, suffocation, sense of doom or sense of impending death. There was a higher percentage of distress reported in the BIS-monitored group, but no statistically significant difference between groups.

In contrast to Avidan and colleagues,44 Zhang and colleagues40 reported a significantly lower incidence of confirmed intraoperative awareness in patients monitored by BIS than in those who received standard clinical monitoring. Incidence of possible awareness was also lower for BIS-monitored patients, although not statistically significant. The incidence of confirmed or possible awareness was significantly lower for BIS-monitored patients.

Intraoperative awareness was the primary outcome measure in the Cochrane BIS review. 34 However, as stated earlier, the review reported awareness outcomes only for trials in its set which were conducted with patients considered to be at higher risk of awareness (n = 4). 27,78,79,82 The Cochrane review combined these four trials in a fixed-effect meta-analysis, and we have updated this meta-analysis to include the two trials from our study set that featured higher risk patients. 40,44Figure 2 reports the results of this meta-analysis.

FIGURE 2.

Meta-analysis of intraoperative awareness during BIS monitoring (patients classified at higher risk of awareness)

The meta-analysis included three subgroup analyses: trials that used inhaled GA only; trials that used a mixture of inhaled and i.v. anaesthesia; and trials that used TIVA. The original overall pooled Peto's odds ratio (OR) from the Cochrane review was 0.33 [95% confidence interval (CI) 0.13 to 0.84], indicating a statistically significant difference between groups favouring BIS. The addition of the trials by Avidan and colleagues44 and Zhang and colleagues40 increased the OR to 0.45 (95% CI 0.25 to 0.81). Caution is advised in the interpretation of this result as, overall, there was statistically significant heterogeneity (p = 0.009; I² = 79%). In the subgroup of trials that used only inhaled anaesthesia the Peto's OR was 1.79 (95% CI 0.63 to 5.11) in favour of standard clinical monitoring. This is in contrast with the other two subgroups, which favoured BIS monitoring.

Explicit intraoperative awareness was an outcome measured in a further eight trials included in the Cochrane BIS review. However, as stated earlier, the review did not report the results of these trials for this outcome. We examined these studies (data not formally extracted) and note that no patients in any of these eight trials reported experiencing intraoperative awareness. It is unlikely that these studies were adequately statistically powered to detect awareness.

The trial by Kersens and colleagues49 was the only study to report implicit awareness, that is awareness that the patient does not necessarily recall experiencing. The probability of postoperatively selecting a word presented during anaesthesia (target) was higher in the BIS monitoring group (mean 0.371 ± 0.132) than in the standard clinical monitoring group (mean 0.323 ± 0.132). The probability of postoperatively selecting a word not presented during anaesthesia (distractor) was lower in the BIS monitoring group (mean 0.315 ± 0.117) than in the standard clinical monitoring group (mean 0.338 ± 0.119). It was not reported whether or not differences between study groups were statistically significant. Intragroup and overall differences between postoperative target and distractor word recall suggest that BIS-monitored patients were more likely to select words presented during anaesthesia than words not presented during anaesthesia, but standard clinical monitoring patients performed no better than chance in word selection (within-group difference in probability of selecting target word or distraction word: BIS: p = 0.001; standard clinical monitoring: p≥ 0.05).

Anaesthetic consumption

Table 9 reports the impact of BIS monitoring on intraoperative general anaesthetic requirement.

Six of the 11 BIS trials included in this systematic review reported this outcome measure,45–47,49,51,62 two of which reported it to be the primary outcome. 46,62 Three of the trials reported volatile anaesthetic consumption, all of which were for sevoflurane. Two of these three trials were conducted in children. 45,51 The mean end-tidal sevoflurane concentration (%) during maintenance of GA in each of these three trials was statistically significantly lower in the BIS-monitored group than in the standard clinical monitoring group. The other three trials46,47,62 reported i.v. anaesthetic consumption, all of which used propofol. One of these trials was conducted with children. 46 In two of the three trials the maintenance dose was higher in BIS-monitored patients than standard clinical monitoring, but with no statistically significant differences between groups. 46,62 The third trial was reported in a conference abstract, and limited results are given, except that there was a 25.3% reduction in propofol consumption compared with standard clinical monitoring. 47

The Cochrane BIS review34 conducted random-effects meta-analyses for anaesthetic consumption, producing separate meta-analyses for volatile anaesthetic consumption and for propofol consumption. We have updated these meta-analyses with studies included in our systematic review. Figure 3 shows the results of the meta-analysis of volatile anaesthetic consumption (sevoflurane).

FIGURE 3.

Meta-analysis of volatile anaesthetic consumption (sevoflurane) during BIS monitoring, MAC equivalents. IV, inverse variance, SD, standard deviation.

As stated, two of the three studies measuring sevoflurane consumption in our systematic review were conducted in children. The Cochrane BIS review34 only included studies of adults, therefore we have only updated their meta-analysis with the one study of adults from our set (Kerssens and colleagues49). The original mean difference in MAC equivalents from the Cochrane review for sevoflurane consumption was −0.16 (−0.29 to −0.04), indicating a statistically significant difference in favour of BIS. Updating the meta-analysis with the trial by Kerssens and colleagues49 reduced the mean difference slightly to −0.15 (95% CI −0.25 to −0.06), but remained statistically significant. However, caution is advised because of a high degree of unexplained statistical heterogeneity (p < 0.00001; I² = 85%).

Figure 4 shows the results of the meta-analysis of propofol consumption.

FIGURE 4.

Meta-analysis of propofol consumption during BIS monitoring, mg/kg/minute. IV, inverse variance; SD, standard deviation.

As stated, one of the three studies measuring propofol consumption in our systematic review was conducted in children. 46 As the Cochrane BIS review34 only included studies of adults, therefore we have updated their meta-analysis with one of the two studies of adults from our set. 62 (NB. The other adult study47 was only reported in a conference abstract and the results were not reported in a format amenable to meta-analysis.) The original mean difference propofol consumption (mg/kg/minute) in the Cochrane review was −1.44 (−1.95 to −0.93), indicating a statistically significant difference in favour of BIS. Updating the meta-analysis with the trial by Ellerkmann and colleagues62 reduced the mean difference slightly to −1.30 (95% CI −1.83 to −0.76), but remained statistically significant. Again, caution is required because of highly significant unexplained statistical heterogeneity (p < 0.00001; I² = 80%).

Outcomes related to postanaesthesia care unit stay

Five of the 11 BIS trials45,46,48,52,53 in our systematic review reported this outcome, of which four were conducted with children. 45,46,52,53 In none of the trials was use of PACU a primary outcome. All of the studies appear to have reported the time to discharge from the PACU. However, it was not always clear exactly when the time to discharge began (e.g. from the end of skin closure, termination of anaesthetic or from admittance to the PACU). Bannister and colleagues45 reported time from end of surgery to PACU discharge, whereas Kamal and colleagues48 and both the trials by Messieha and colleagues52,53 stated measuring the end of general anaesthetic to PACU discharge (although in one of these trials48 data do not appear to be reported for that outcome). Bhardwaj and colleagues46 did not provide any detail on timing. Detail of discharge criteria varied between the trials. Bannister and colleagues45 and Kamal and colleagues48 both used the Aldrete scoring system (score of > 9), whereas Bhardwaj and colleagues46 used the Steward recovery scoring system (eligibility = score of 6). Messieha and colleagues52,53 did not report use of discharge criteria.

Table 10 shows the results of the trials relating to stay in the PACU.

In all trials, time to discharge from the PACU was statistically significantly greater in the standard clinical monitoring group than in the BIS monitoring group, with mean differences in the range of 6.7–30 minutes. One trial did not report data for this outcome, mentioning that time to discharge was comparable between groups. There was also a statistically significant difference in the one trial that measured time to arrival at the PACU, with reduction of 4.7 minutes for BIS monitoring. 48 The two trials that reported duration of stay in the PACU both reported statistically significant differences in favour of BIS. 52,53

Eligibility for discharge from the PACU unit was one of the secondary outcomes from the Cochrane BIS review. 34 The review meta-analysed the outcome ‘PACU’ stay, including data from 12 trials. Examination of characteristics of the trials included in this meta-analysis, as summarised in the Cochrane review, show that some of the trials reported time to arrival in the PACU, time to discharge from the PACU and length of stay in the PACU. These all appear to have been included in the same meta-analysis, and there is no discussion about how timings may differ according to these different outcomes. Given this lack of clarity, and the fact that the Cochrane review included only trials of adults, we decided not to update this meta-analysis with data from trials identified in the current review. The pooled random-effects mean difference reported in the Cochrane review was −7.63 minutes (95% CI −12.50 to −2.76 minutes) in favour of BIS. However, caution is advised for the reasons given above, as well as a high degree of statistical heterogeneity (p < 0.00001; I² = 82%). The results of the meta-analysis are similar to the results of the trials included in the current review (i.e. showing a benefit for BIS monitoring).

Time to recovery from anaesthesia

The trials included in the current systematic review reported a variety of outcomes relating to recovery from anaesthesia, including time to tracheal extubation, time to eye opening and movement responses.

Table 11 reports the time to tracheal extubation following surgery.

Five of the 11 BIS trials included in the current systematic review measured time to extubation, of which four were conducted with children. 45,46,52,53 None of these studies considered this to be a statistically powered primary outcome measure. Timing was reported to have begun from end of surgery in three studies,45,52,53 and from termination of anaesthetic in two studies. 46,48 Extubation times were shorter for BIS-monitored patients than for those receiving standard clinical monitoring by as much as 5 minutes or as little as 0.5 minutes. Differences between groups were reported to be statistically significant in two trials,45,53 but not in two other trials. 48,52 One trial did not report numerical data, stating that times were comparable between groups. 46

A sixth study, conducted with children, reported time to laryngeal mask airway removal following surgery as an outcome. 51 The mean time [standard deviation (SD)] in minutes was 1.8 (1.6) in the BIS-monitored group, and 2.1 (2.4) in the standard clinical monitoring group (p = 0.93), indicating no statistically significant differences between groups.

Time to extubation was one of the secondary outcomes from the Cochrane BIS review. 34 The review meta-analysed data from 21 trials. Given that four of the five trials45,46,52,53 in the current systematic review were conducted in children and the Cochrane review was restricted to trials of adults, we have not updated their meta-analysis. The overall random-effects mean difference in time to extubation was −2.87 minutes (95% CI −3.74 to −1.99 minutes), indicating a statistically significant difference in favour of BIS. Caution is advised as there was a high degree of statistical heterogeneity (p < 0.00001; I² = 79%).

Table 12 reports time to eye opening following surgery.

Three trials included in the current systematic review reported time to eye opening, two of which were conducted with children. 46,51 Timing was reported to have begun immediately after the last surgical stitch in two studies48,51 and from the end of surgery in one trial. 46 Times were shorter in BIS-monitored patients, although by modest duration (up to 1 minute) and there were no statistically significant differences between groups. One trial provided only narrative results, reporting comparable times between groups.

Time to eye opening was one of the secondary outcomes from the Cochrane BIS review. 34 The review meta-analysed data from 19 trials. Given that two of the three trials in the current systematic review were conducted in children and the Cochrane review was restricted to trials of adults, we have not updated their meta-analysis. The overall random-effects mean difference in time to extubation was −2.14 minutes (95% CI −2.99 to −1.29 minutes), indicating a statistically significant difference in favour of BIS. Caution is advised as there was a high degree of statistical heterogeneity (p < 0.00001; I² = 83%). The results of the meta-analysis are more conclusive than those of the relatively smaller number of trials included in the current review.

Table 13 reports the results of three trials that reported other recovery outcomes.

All three of the trials45,46,51 reporting other recovery outcomes were conducted with children. Bannister and colleagues45 reported mean time to first movement, with a statistically significant reduction for BIS-monitored patients of 2.8 minutes. Similarly, Liao and colleagues51 reported a statistically significant reduction in time to first spontaneous movement of 2.5 minutes. This trial51 also reported a shorter time to phonation (making a vocal sound) of 4.5 minutes, but this was not statistically significant. Bhardwaj and colleagues46 reported time to response to commands, commenting that this was comparable in the two groups but not reporting any numerical data.

Postoperative nausea and vomiting

Postoperative nausea and vomiting was reported by only one of the trials included in the current systematic review, the trial by Liao and colleagues. 51 There was no difference between patients in the BIS and standard clinical monitoring groups in terms of nausea [n = 5 (10%); n = 6 (11%), respectively, p = 0.95] or vomiting [n = 2 (4%); n = 3 (6%), respectively, p = 0.88]. Postoperative nausea and vomiting was not reported by the Cochrane BIS review. 34

Emergence delirium

Liao and colleagues51 also reported the incidence of emergence delirium, as measured by the Paediatric Anaesthetic Emergence Delirium (PAED) instrument (noted to be valid and reliable by the authors, reference cited). Assessment took place by a trained observer in the PACU every 5 minutes after awakening for 30 minutes. The highest score during this period was used in the final PAED score. (NB. A description of the instrument and what the scores mean is not given.) There was no statistically significant difference between BIS and standard clinical practice monitored patients [median (interquartile range) score 18 (14–16); 15 (13–15), respectively, p = 0.94].

Postoperative cognitive dysfunction

The only trial to report postoperative cognitive dysfunction was that of Chan and colleagues, who studied an elderly patient population. 47 Cognitive dysfunction was assessed by a battery of eight neuropsychology tests before and at 1 and 3 weeks after surgery (no information on the tests reported). POCD was confirmed when two or more test parameters or the combined z-value > 1.96 (no further information given). There was no statistically significant difference between BIS and standard clinical monitoring in rates of dysfunction at 1 week post surgery [146 (32.5%); 177 (39.1%), respectively, p = 0.07]. However, the difference between groups become significant at 3 months post surgery [36 (8.1%); 54 (12%), respectively, p = 0.03; OR 1.6 (95% CI 1.0 to 2.4)]. Caution is advised as this trial47 was reported in a conference abstract therefore detail of its characteristics are lacking, prohibiting a thorough appraisal of its methodological quality. As the abstract was published in 2010 a full publication potentially may be available in the near future.

Mortality, myocardial infarction and stroke

One trial, by Leslie and colleagues,50 reported stroke, myocardial infarction and mortality for all surviving and available patients 30 days post operation (Table 14). This is a long-term follow-up (median = 4.1 years) publication of the B-Aware trial79 in patients classified at higher risk of intraoperative awareness because of factors such as type of surgery (e.g. high-risk cardiac surgery), health status (e.g. cardiovascular impairment) and lifestyle (e.g. heavy alcohol intake). [NB. A publication of the short-term results of this trial by Myles and colleagues79 (primary outcome: intraoperative awareness) was included in the Cochrane BIS review. 34 Results of this trial are presented earlier in this report.]

There was no statistically significant difference between BIS-monitored patients and patients who received routine care in mortality, myocardial infarction or stroke.

Summary of Bispectral Index assessment

• Six trials included in this systematic review measured the impact of BIS monitoring on explicit intraoperative awareness. Four of these trials reported few or no cases of awareness; however, they were not statistically powered to detect this outcome. The other two trials were powered to detect awareness and we added them to the meta-analysis from the Cochrane BIS review (restricted to patients considered to be at higher risk of awareness). The pooled Peto's OR was 0.45 (95% CI 0.25 to 0.81), in favour of BIS. However, there was statistically significant heterogeneity and a non-significant difference in the subgroup of trials in which only inhaled GA was used.

• Three trials included in this systematic review reported changes in sevoflurane consumption, all of which were statistically significantly lower with BIS monitoring. We updated the Cochrane meta-analysis with one of these trials, producing a pooled mean difference of −0.15 (95% CI −0.25 to −0.06) MAC equivalents in favour of BIS (with unexplained statistically significant heterogeneity).

• Three trials included in this systematic review reported changes in propofol consumption. In two of these the maintenance dose was higher in BIS-monitored patients than standard clinical monitoring, but not statistically significant. In the third trial propofol consumption was lower for BIS. We updated the Cochrane meta-analysis with one of these trials, producing a pooled mean difference of −1.33 mg/kg/minute (95% CI −1.82 to −0.84 mg/kg/minute), in favour of BIS (with unexplained statistically significant heterogeneity).

• Five trials included in this systematic review reported time to discharge from the PACU, all of which appeared to be secondary outcomes. In all trials time to discharge was statistically significantly shorter in BIS-monitored patients, with mean differences in the range of 6.7–30 minutes. The Cochrane BIS review did a meta-analysis of the outcome ‘PACU stay’ (including time to arrival in the PACU, time to discharge from the PACU, and length of stay in the PACU). The pooled mean difference was −7.63 minutes (95% CI −12.50 to −2.76 minutes) in favour of BIS (with unexplained statistically significant heterogeneity).

• Five trials included in this systematic review measured time to tracheal extubation, as a secondary outcome. Extubation times were shorter for BIS-monitored patients compared with standard clinical monitoring by as much as 5 minutes, and as little as 0.5 minutes, but not always statistically significant. The pooled mean difference in the Cochrane review for this outcome was −2.87 minutes (95% CI −3.74 to −1.99 minutes) in favour of BIS (with unexplained statistically significant heterogeneity).

• Three trials included in the current systematic review reported time to eye opening as a secondary outcome. Times were shorter in BIS-monitored patients, although by modest duration (up to 1 minute), and there were no statistically significant differences between groups. The pooled mean difference in the Cochrane review for this outcome was −2.14 minutes (95% CI −2.99 to −1.29 minutes), indicating a statistically significant difference in favour of BIS (with unexplained statistically significant heterogeneity).

• Postoperative nausea and vomiting was reported by only one trial. Incidence of nausea and vomiting was low (around 10% or less) and there was no statistically significant difference between groups.

• Only one trial reported the incidence of postoperative cognitive dysfunction. There was no statistically significant difference between groups in rates of dysfunction at 1 week post surgery. By 3 months post surgery, incidence had fallen to around 8–12%, with a significant difference in favour of BIS. This study was reported only as a conference abstract and it is not clear whether or not this outcome was adequately statistically powered.

• Longer-term postoperative outcomes of stroke, myocardial infarction and mortality were reported by only one trial (median of 4.1 years post operation), as secondary outcomes. Mortality was lower in BIS-monitored patients, although not statistically significant. Incidence of stroke and myocardial infarction was similar between groups.

• In summary, BIS monitoring was associated with overall lower rates of explicit intraoperative awareness (limited to patients classified at higher risk of awareness, and non-significant effects in the subgroup of patients receiving only inhaled anaesthesia), lower general anaesthetic consumption and shorter recovery times (e.g. PACU discharge, time to extubation, time to eye opening). Generally, there was little difference between BIS and standard clinical monitoring in complications arising from excessive anaesthetic dose (e.g. nausea, vomiting and cognitive dysfunction). Caution is advised in the interpretation of the results as not all outcomes appeared to be adequately statistically powered, and there was significant heterogeneity. There was much variation between the trials in terms of patient characteristics and surgical procedures.

Study populations

Two of the seven E-entropy trials were conducted with children, with median age 4–6 years (range 3–12 years). 54,56 The remaining five trials were in adults, with the mean age of patients ranging from 33 years55 to 69 years. 58 The trials varied in their sex composition. One trial was entirely on adult women,55 whereas another trial was almost entirely on young boys (the trial included 12% girls in one study arm only). 56 One trial included more elderly men than women (men–women ratio approximately 4:1),58 whereas another trial included more middle-aged women than men (male–female ratio approximately 1:3). The remaining three E-Entropy trials included a more even balance of males and females. 54,61,62 In all seven trials the mean body weight of patients appeared to be within the normal range, with mean weights ranging from 16 kg to 22 kg in the child studies and from 65 kg to 82 kg in the adult studies. One trial was conducted at six centres in three countries (Finland, Sweden and Norway). 57 The remaining trials appeared to be single-centre studies (not explicitly stated in two trials) that were each carried out in one country: Germany,55,62 France,61 India,56 South Korea54 and Taiwan. 58 None of the E-Entropy trials reported the ethnicity of their participants.

Four of the E-Entropy trials were in patients undergoing a mix of abdominal, urological, gynaecological and/or orthopaedic surgical procedures,56,61,62 which also included breast and thyroid surgery in one trial. 57 One trial specifically involved children undergoing tonsillectomy or adenoidectomy. 54 Another trial was carried out specifically in women undergoing laparoscopic gynaecological procedures. 55 The remaining trial focused on total knee replacement surgery. 58 Only one of the E-Entropy trials was clearly limited to day surgery patients. 56 None of the seven trials identified any specific risk factors for intraoperative awareness among their populations and none reported whether or not patients had any comorbidities that affect EEG monitoring. However, all the E-Entropy trials stated that they excluded patients with any history of cerebrovascular and/or neurological disorders. The ASA grade of patients was I–II in four of the trials,54–56,58 and I-III in the remaining three trials. 57,61,62 The proportion of grade III patients varied by study groups within these three trials, ranging 1–3%,57 11–15%61 and 3–26%. 62

Technologies

Four of the seven E-Entropy trials reported that they used the E-Entropy module manufactured by GE Healthcare,55,57,61,62 and six of the trials reported that they used the S/5TM monitor (Datex-Ohmeda). 54–58,61 Very little other information about the modules and monitors was provided: only one trial mentioned the version of the S/5 monitor used (Avance),56 and none of the studies stated the version of the E-Entropy algorithm software used.

The target E-Entropy values during anaesthesia maintenance were mostly in the range 40–65. Four trials specified target ranges for state entropy, which were either 40–6054,55 or 45–65. 56,57 A further trial specified a specific state entropy target of 50. 62 The remaining two trials specified target ranges for both state entropy and response entropy, which were 35–4558 and 40–60. 61 Four of the trials that specified target values for state entropy permitted an increase in the state entropy value during the last 15 minutes of surgery. During this period, the target values were specified as 60,62 65–70,56 ‘ideally 65, but not > 70’57 and ‘> 60 acceptable’. 55 In addition to the target values of state and response entropy, three trials also specified target values of the difference between response entropy and state entropy: these were < 10 in two trials55,57 and 5–10 in the remaining trial. 58

Two of the seven E-Entropy trials reported that E-Entropy monitoring for anaesthesia delivery was done in conjunction with monitoring haemodynamic changes. One of these trials specified that heart rate and blood pressure were to be kept within ± 20% of their baseline (preoperative visit) values. 57 The second trial stated that E-Entropy was used to guide anaesthesia unless (unspecified) haemodynamic changes of 30% persisted for > 5 minutes.

In addition to titrating anaesthesia to maintain the specified target entropy values, two trials specified corrective action if target values were exceeded. One trial specified intermittent provision of a sufentanil bolus if the response entropy–state entropy difference exceeded 10 for > 2 minutes. 61 The other trial specified administration of a propofol bolus if the state entropy value increased suddenly above 65.

In all seven of the E-Entropy trials, the E-Entropy monitoring was initiated in the operating theatre. Two trials stated57 or implied58 that E-Entropy monitoring was started before anaesthesia induction, and two trials stated that E-Entropy monitoring began after anaesthesia induction. 55,56 The remaining three trials did not report whether E-Entropy monitoring commenced before or after anaesthesia induction.

Comparators

Standard clinical monitoring was based on blood pressure and heart rate in three trials. 54,57,61 As well as blood pressure and heart rate, a further two trials also monitored sweating, lacrimation or movement,62 or coughing, chewing, grimacing or purposeful movement. 55 The remaining trials monitored heart rate, mean arterial pressure and lacrimation, and either movement in response to surgical stimulation,56 or sweating, flushing or wrinkling of frontal facial muscles, together with monitoring the end-tidal anaesthetic concentration. 58 Quantitative thresholds for the clinical parameters that were used to guide anaesthesia titration were specified in five of the seven E-Entropy trials. 54–58

In addition to titrating anaesthesia according to the clinical parameters, in one trial58 the ETAC was adjusted to maintain mean arterial pressure and heart rate fluctuations to within ± 30% of the baseline values. In another trial, i.v. fentanyl was given if clinical parameters were not stabilised after increasing the anaesthetic concentration to 1.3 MAC. 56

Anaesthetic agents and protocols

Three of the seven trials used i.v. propofol for anaesthesia induction. 55,61,62 One trial used i.v. propofol with alfentanil analgesic for induction. 57 A further trial employed propofol if patients had an i.v. line, but otherwise used inhaled sevoflurane for induction. 56 The remaining two trials both used inhaled sevoflurane for induction in all their patients. 54,58

For maintenance of anaesthesia, three trials used inhaled sevoflurane,54,58,61 and one trial used inhaled isoflurane. 56 The remaining trials used i.v. delivery of propofol,62 propofol and remifentanil,55 or propofol and alfentanil analgesic. 57

Overall, two trials used the same inhaled agent (sevoflurane) for both induction and maintenance;54,58 three trials used i.v. agents (all included propofol) for both induction and maintenance;55,57,62 and two trials used an i.v. anaesthetic for induction followed by an inhaled anaesthetic for maintenance. 56,61

Regional anaesthesia was only clearly reported in one of the E-Entropy trials, in which a caudal block was placed with bupivacaine. 56 Two trials stated that regional anaesthesia was not used. 58,61 One trial referred to regional anaesthesia in the publication abstract but did not provide details. 62 The remaining three trials did not refer to regional anaesthesia.

One of the E-Entropy trials stated that analgesics were not used during induction or maintenance of anaesthesia, although ketorolac was used after anaesthetic cessation. 54 One trial used i.v. sufentanil during induction and maintenance, with morphine during the last 15 minutes of surgery, followed by paracetamol, nefopam or non-steroidal anti-inflammatory drugs postoperatively. 61 Two trials used fentanyl during anaesthesia maintenance. Of these, one also used fentanyl and lidocaine during induction,58 whereas the other used fentanyl postoperatively, according to the patient's pain score. 56 One trial used piritramide during the last 15 minutes of surgery only. 55 The remaining two trials did not refer to analgesia either during induction, maintenance or post surgery. 57,62

Premedication was reported in five of the E-Entropy trials. The agents used were oral hydroxyzine,61 oral midazolam alone,62 oral midazolam with a benzodiazepine,55 i.v. midazolam54 and oral diazepam (in five of six study centres). 57 The remaining two trials did not specify whether or not premedication was used.

All of the E-Entropy studies except one56 used muscle relaxants. The muscle relaxants were atracurium,58,61,62 rocuronium54,55 or were not specified a priori but were chosen at the anaesthetist's discretion when needed. 57

In five trials anaesthesia was administered in the operating theatre. 56–58,61,62 The two remaining trials did not report where anaesthetics were administered.

The mean duration of anaesthesia was reported in six studies and ranged from 64.3 minutes for tonsillectomy or adenoidectomy procedures in children54 to 190.8 minutes for general surgical procedures in adults. 61 The remaining study reported median duration of anaesthesia which was 68–72 minutes (range 32–180 minutes) for lower abdominal or urological surgical procedures in children. 56

Duration of the surgery itself was reported less precisely than the duration of anaesthesia. Surgical duration was described as a minimum of 1 hour,55,61 approximately 1.5 hours,58 a mean of 41.4–48.1 minutes,54 or a median of 29–30 minutes (range 15–95 minutes)56 or was not reported. 57,62

The training and experience of the anaesthetists in E-Entropy module use was reported in four of the seven E-Entropy trials. 55,57,61,62 One trial stated that anaesthetists were allowed to accustom themselves to the use of E-Entropy monitoring for 3 weeks, and all participants had substantial previous experience with EEG-based depth of anaesthesia monitors. 57 In the remaining three trials the descriptions provided for training or experience were only superficial: ‘more than 3 months of routine use’61 ‘experienced anaesthesiologist’;62 and ‘anaesthesia was supervised by an experienced staff anaesthetist’. 55

Outcomes

Anaesthetic consumption was the primary outcome in four of the seven E-Entropy studies (Table 15). 54,58,61,62 The method of assessing anaesthetic consumption was by weighing the vaporiser,61 measuring the end-tidal concentration,54 using data from the S/5 anaesthetic delivery system58 or was not reported. 62 In the remaining three trials the primary outcomes were time to eye opening55,56 and time to response to a verbal command,57 after cessation of anaesthesia.

The most frequently reported outcomes overall for which quantitative results were reported were: anaesthetic consumption (a primary outcome in four trials54,58,61,62 and a secondary outcome in three trials55–57); entropy values (a secondary outcome in all seven trials); time to eye opening (a primary outcome in two trials55,56 and a secondary outcome in four trials54,57,61,62); intraoperative awareness (a secondary outcome in all except one trial56); haemodynamic profiles (a secondary outcome in all except one trial62); time to extubation (a secondary outcome in three trials54,57,61); and postoperative pain (a secondary outcome in two trials55,56). Other outcomes that were reported quantitatively in one trial each were postoperative pain, analgesia consumption, PONV, time to recovery based on Aldrete or Steward scores, time spent with adverse haemodynamic profiles, probability of emergence and (in a study with children) parental satisfaction. Some of the trials provided only a narrative report of outcomes. These outcomes were not extracted from the primary trials as no estimates of effect or variance could be determined. For example, two trials57,58 stated narratively that pain scores, analgesic use and incidence of PONV did not differ between E-Entropy and clinical practice groups but no quantitative results were reported for these outcomes and so these are not included in Table 15.

Three of the six trials that measured intraoperative awareness employed versions of standard patient questionnaires published by Brice and colleagues24 (two studies57,61) or Nordström and colleagues96 (one study62). The three remaining trials stated only that intraoperative recall was assessed by independent nurses;54 patients were questioned about memory and awareness;55 or the level of awareness was assessed. 58 Four trials reported the timing of the intraoperative awareness assessments, which were 24 hours after surgery,55 on the first postoperative day,62 in the PACU and on the first day post surgery,57 or on the first and third days post surgery. 61 The remaining two trials did not specify the timing of the awareness outcome assessments. No further details of the methods for assessing intraoperative awareness were reported.

Length of follow-up was relatively short in all the trials, being 1 day post surgery (for intraoperative awareness) in three trials,54,55,57 3 days post surgery (for intraoperative awareness) in three trials,58,61,62 and only 2 hours post surgery (for pain assessment) in the remaining trial. 56 The duration of follow-up would not have been adequate for detecting delayed onset of awareness recall, which may occur more than 1 week post surgery.

Intraoperative awareness

Only one case of intraoperative awareness was reported in the six trials that measured this outcome (Table 16). This was experienced by an adult woman in the standard clinical practice group of the trial by Gruenewald and colleagues. 55 It should be noted that the sample sizes of these studies may have been too small to detect rare events such as intraoperative awareness.

Anaesthetic consumption

Four trials that assessed volatile anaesthetic consumption either as the primary outcome for sevoflurane54,58,61 or a secondary outcome for isoflurane,56 all demonstrated statistically significant reductions in the E-Entropy-guided anaesthesia group compared with the standard clinical monitoring group (Table 17). In the trial by Aime and colleagues,61 the rates of sevoflurane consumption, but not the total amount consumed, were significantly lower in the E-Entropy group. In this trial61 the difference in sevoflurane consumption rates between groups was more pronounced when the consumption rate was normalised to patients' body weight.

Three trials that assessed consumption of i.v. anaesthetics55,57,62 showed mixed results (Table 17). Propofol consumption in the E-Entropy group was statistically significantly lower than in the standard clinical practice group in two trials that assessed anaesthetic consumption as secondary outcomes,55,57 but not in a trial that assessed anaesthetic consumption as the primary outcome. 62 Remifentanil consumption was significantly higher in the E-Entropy group in one trial that assessed this as a secondary outcome,55 but did not differ between groups in the trial that assessed this as the primary outcome. 62 Alfentanil consumption, assessed as a secondary outcome in one trial, did not differ significantly between the study groups. 57

The trials that assessed anaesthetic consumption measured outcomes in different ways, expressed their outcomes in different units (total consumption or rates) and, as noted above, differed in the patient populations that they included. These differences would preclude the meaningful pooling of the anaesthetic consumption outcomes that were reported (Table 17).

Time to recovery from anaesthesia

Results are summarised in Table 18 for the trials that reported time to eye opening;54–57,61,62 extubation;54,57,61 spontaneous breathing;57 recovery of orientation;54,57 response to commands;57 recovery defined by Aldrete score;54 and recovery defined by modified Steward score. 56

Time to eye opening was significantly shorter, by approximately 2–4 minutes, in the E-Entropy group than in the standard clinical practice group in two of six trials. 56,57 One of these assessed this as a primary outcome in children56 and the other assessed it as a secondary outcome in adults. 57 In the remaining four trials54,55,61,62 (one of which specified this as a primary outcome55) the time to eye opening did not differ between the study groups (Table 18).

Time to extubation (a secondary outcome) was shorter, by approximately 3–4 minutes, in the E-Entropy group than in the standard clinical monitoring group in all three trials that assessed this outcome. 54,57,61 The differences were stated as statistically significant in two of the trials54,57 but statistical significance was not reported in the remaining trial61 (see Table 18).

The times to spontaneous breathing (a secondary outcome);57 recovery of orientation (a secondary outcome);54,57 response to commands (a primary outcome);57 and recovery defined by an Aldrete score of at least 9 (a secondary outcome)54 were each significantly shorter in the E-Entropy group than the standard clinical practice group in the two trials54,57 that reported these outcomes (see Table 18). However, the time to recovery as defined by reaching a Steward score of 6 (a secondary outcome) did not differ between the study groups in one trial that assessed this outcome. 56

Outcomes related to postanaesthesia care unit stay

The time from discharge from the operating room to the PACU was shorter by approximately 3–4 minutes in the E-Entropy group than the standard clinical practice group in the two trials that monitored these outcomes56,57 (Table 19). The differences in both trials were statistically significant, although only marginally so in one of the trials. 56

The time to discharge from the PACU was shorter in the E-Entropy group than the standard clinical monitoring in the only trial that assessed this outcome,57 although the difference was not statistically significant. The time from which discharge from the PACU was measured was not reported, however, which makes interpretation of this outcome unclear57 (see Table 19).

Postoperative pain

Two trials reported postoperative pain, using different rating scales (Table 20). Pain was assessed as a score on a 0–10 scale55 or using the Children's Hospital of Eastern Ontario Pain Score (CHEOPS). 56 Pain scores were significantly lower in the E-Entropy group than standard clinical practice for the adult population. 55 In the paediatric population, the CHEOPS scores were significantly lower in the E-Entropy group at 60, 90 and 120 minutes after arrival in the PACU but not at 30 minutes after arrival. 56

Analgesic consumption

Only one E-Entropy trial assessed analgesic consumption. 61 Consumption of sufentanil was slightly lower in the E-Entropy group than the standard clinical monitoring group during both induction and maintenance of anaesthesia, but the differences were not statistically significant (Table 21).

Postoperative nausea and vomiting

One trial that assessed PONV after arrival in the recovery room55 reported similar frequencies in the E-Entropy and standard clinical monitoring that did not differ significantly (Table 22).

In addition to the outcomes reported above, the E-Entropy trials reported that the following outcomes did not differ between E-Entropy and standard clinical practice groups (data not extracted): patient satisfaction scores;55 parent satisfaction scores for children at 24 hours post surgery;55 time spent by patients with adverse haemodynamic profiles;61 and treatment for haemodynamic events. 61

Summary of E-Entropy assessment

• Six trials monitored intraoperative awareness in adults and children receiving different volatile and i.v. anaesthetics. Only one case of awareness occurred, in the standard clinical practice group of one trial. However, sample sizes were relatively small in these trials.

• Four trials monitored consumption of volatile anaesthetic (three monitored sevoflurane as a primary outcome, one monitored isoflurane as a secondary outcome). Consumption was significantly lower in the E-Entropy monitoring than standard clinical practice groups of all trials, with the proviso that in one of these trials the difference in sevoflurane consumption was statistically significant for rates of consumption but not for total anaesthetic dose.

• Three trials that monitored consumption of i.v. anaesthetic yielded mixed results. Trials that monitored consumption of propofol, remifentanil and alfentanil as primary outcomes found no statistically significant differences between the study groups. However, significantly lower consumption of propofol and remifentanil in the E-Entropy group was reported in trials that assessed these as secondary outcomes.

• Time to eye opening was significantly shorter in the E-Entropy group than the standard clinical practice group in two of six trials, one of which assessed this as a primary outcome, but did not differ in the remaining four trials.

• Time to extubation (a secondary outcome) was shorter in the E-Entropy group than the standard practice group in all three trials that assessed this outcome. The differences were stated as statistically significant in two of the trials but statistical significance was not reported in the remaining trial.

• The times to spontaneous breathing (a secondary outcome), recovery of orientation (a secondary outcome), response to commands (a primary outcome) and recovery defined by an Aldrete score of at least 9 (a secondary outcome) were each significantly shorter in the E-Entropy group than in the standard clinical practice group. Except for time to orientation (two trials), these outcomes were reported by only one trial each. The time to recovery as defined by reaching a Steward score of 6 (a secondary outcome) did not differ between the study groups in one trial that assessed this outcome.

• The limited evidence available (from two trials which assessed secondary outcomes only) suggests that E-Entropy monitoring favours shorter time to discharge to and from the PACU, but it is unclear whether or not the time gains are clinically important.

• No firm conclusions can be drawn about effects of E-Entropy monitoring on postoperative pain because the only two trials that assessed this used different rating scales, and the effect of E-Entropy monitoring on pain scores was temporally variable in one of the trials. Analgesic consumption and frequency of PONV were assessed in one trial each and did not differ between the E-Entropy and standard clinical practice groups. Postoperative pain, nausea and vomiting, and analgesic consumption were assessed only as secondary outcomes in these trials.

• In summary, compared with standard clinical monitoring, E-Entropy monitoring favoured: lower consumption of volatile anaesthetics and some, but not all, i.v. anaesthetics; and shorter times to recovery and discharge to and from the PACU, assessed by various measures. E-Entropy monitoring had no consistent impact on other outcomes that were monitored, including intraoperative awareness, but the small sample sizes in the trials may not have provided adequate statistical power to detect meaningful differences in rare events. Pooled effect estimates would not be estimable reliably for these outcomes, because of the uniqueness of the individual studies (which included different populations in terms of age, sex and ethnicity, undergoing different surgical procedures) and differences between the trials in the way that outcomes were assessed and reported. Also, the majority of the outcomes were secondary and may not have been adequately powered statistically to detect clinically relevant differences between the E-Entropy and standard clinical practice groups.

Study populations

In all trials of Narcotrend, the study population was adults (mean age 40–50 years) and 33–50% of participants were males for the three studies reporting sex. Mean weight ranged from 60 kg to about 84 kg. All studies appeared to be single-centred studies with three conducted in Germany and one in China. 59 Ethnicity of participants was not reported in any study.

The type of surgery was minor orthopaedic surgery,63,64 microwave coagulation for liver cancer,59 and all kinds of elective surgery, including surgery for ‘malignoma’ and peripheral vascular surgery. 60 No trial reported risk factors for awareness. Comorbidities were reported in two trials:59,60 hypertension was reported in both of these and one trial60 also reported cardiac arrhythmia, diabetes type II, asthma and miscellaneous comorbidities. Three trials included the number of participants with ASA grade I, II or III, with most grade II and fewest grade III; the fourth trial59 reported only that participants were ASA grade II or III.

Technologies

The Narcotrend monitor with software version 2.0 AF was used in three trials,60,63,64 whereas in the fourth trial59 no details of the software version are reported. Two trials report using the MonitorTechnik (Germany) with Blue Sensor (Denmark). 59,60

The Narcotrend target value during maintenance anaesthesia was D₀ and then adjusted to C₁ 15 minutes before the expected end of surgery in two studies,63,64 and D₂–E₀ during maintenance adjusted to D₀–D₁ 10 minutes before the end of surgery in one study. 59 In the fourth study60 the Narcotrend target value was D₂–E₀ with no further details given. The two studies59,60 using Narcotrend target values of D₂–E₀ therefore used deeper levels of anaesthesia and hypnosis than the other two studies. Monitoring started in the operating theatre in two studies,63,64 in the computed tomography department where surgery took place in one study59 and was not reported in the fourth study. 60

Only one trial59 explicitly stated that observational indices of electrocardiography (ECG), heart rate and mean arterial blood pressure were continuously monitored alongside Narcotrend scores. The other three studies did not explicitly state whether or not standard clinical monitoring took place in addition to Narcotrend. However, as signs of inadequate anaesthesia were based on vital signs and clinical parameters it can be assumed that it did. For example, signs of inadequate anaesthesia were hypertension, tachycardia or patient movement, eye opening, swallowing, grimacing, lacrimation and sweating. 63,64 Vital clinical parameters of heart rate, pulse oximetry readings, rectal temperature and end-expiratory carbon dioxide were continuously measured in the fourth study. 60

Comparators

Standard clinical continuous monitoring included heart rate, systemic arterial blood pressure, respiratory rate, oxygen saturation and end-tidal concentrations of carbon dioxide63,64 plus end-tidal desflurane64 and heart rate, pulse oximetry readings, rectal temperature and end-expiratory carbon dioxide. 60 In one study59 heart rate, blood pressure and body movement were used for monitoring.

Anaesthetic agents and protocols

Three studies used total i.v. anaesthesia: two60,63 used propofol–remifentanil for induction and maintenance anaesthesia; one used propofol–fentanyl induction and propofol anaesthesia maintenance. 59 The fourth study used desflurane–remifentanil anaesthesia. 64 Regional anaesthesia was not reported in any of the studies. Premedication was used in three studies in the form of midazolam60,64 and diazepam. 63 Analgesia included metamizol with sodium chloride,63,64 fentanyl59 and novaminsulfone, piritramide or morphine. 60 Muscle relaxants used included atracurium,64 cisatracurium63 and rocuronium. 60

Mean duration of anaesthesia ranged from 113 to 125 minutes,64 from 108 to 127 minutes,63 from 88 to 91 minutes59 and from 105 to 111 minutes60 in the four trials, with no significant differences between groups within each study. Duration of surgery was not reported in any study. Three studies60,63,64 reported that all patients were anaesthetised by the same experienced anaesthesiologist, one of which mentions specific experience in Narcotrend. 63 No details are given for the length of experience/training of the anaesthetist in the fourth study. 59

Outcomes

The primary outcome (statistically powered) specified in three trials was time to eye opening63,64 and time to extubation60 (Table 23). Time to tracheal extubation was also an outcome in two other studies. 63,64 All four studies59,60,63,64 report anaesthetic consumption and intraoperative awareness. Other reported outcomes include time to arousal time59 (defined as the time between cessation of drugs and the patient being able to open their eyes on command) and time to recovery of orientation (defined as the time between a patient opening their eyes on command and the restoration of orientation). 59 Two studies63,64 report time to discharge to the PACU and two report PONV. 59,60

Intraoperative awareness

No patients in any of the trials of Narcotrend reported intraoperative awareness as explicit memory during anaesthesia, although two patients (8%) receiving Narcotrend anaesthetic monitoring recalled dreaming during anaesthesia. 60

Anaesthetic consumption

Three studies report consumption of propofol; two59,63 found a statistically significant reduction in the group receiving Narcotrend monitoring compared with standard clinical monitoring, whereas the third60 found no difference in consumption between groups (Table 24).

Three studies reported remifentanil consumption and all found no statistically significant difference between Narcotrend and standard clinical monitoring. 60,63,64

Desflurane consumption per patient was not different between the Narcotrend monitoring group and standard anaesthetic practice, although desflurane consumption per patient per minute was statistically significantly lower in the Narcotrend group. 64

Time to arrival at postanaesthetia care unit

Two studies reported time to arrival at PACU and found statistically significantly shorter times in the Narcotrend monitoring group compared with the standard care monitoring group63,64 (Table 25).

Time to eye opening

Time to eye opening was the primary outcome in two trials and results between the studies differ (Table 26). One trial64 reported no statistically significant difference between Narcotrend monitoring and standard clinical monitoring, whereas the other trial63 reported a statistically significant reduction in time to eye opening of 5.9 minutes in the Narcotrend group compared with standard care.

Time to extubation

Time to tracheal extubation was the primary outcome in one study60 and no difference was found between monitoring of anaesthesia by Narcotrend and standard clinical monitoring (Table 27). In contrast, two other studies that reported time to extubation found statistically significant reductions in time to extubation of between 1.4 to 6 minutes with Narcotrend monitoring compared with standard clinical monitoring. 63,64

Other measures of time to emergence from anaesthesia

Time to arousal (defined as the time between cessation of drugs and the patient being able to open their eyes on command) was statistically significantly shorter in the group receiving Narcotrend monitoring than the group receiving standard clinical monitoring. 59 Duration of orientation recovery was also shorter with Narcotrend monitoring (Table 28). 59

Postoperative nausea and vomiting

One study found that no nausea or vomiting was reported after surgery in either group. 59 Another study60 reported that nausea scores were statistically significantly higher in the group receiving anaesthesia monitoring by standard clinical practice than by Narcotrend at 10 minutes after extubation (mean ± SD, 24.06 ± 34.04 vs 6.88 ± 15.2, respectively, p = 0.005); however, there were no significant differences at other time points.

Analgesic consumption

Two studies59,60 reported consumption of pain-relieving drugs and found no statistically significant differences between Narcotrend and standard care monitoring groups.

Summary of Narcotrend assessment

• Four trials59,60,63,64 monitored intraoperative awareness in adults receiving different volatile and i.v. anaesthetics; no patients reported explicit memory during anaesthesia although two patients receiving Narcotrend monitoring recalled dreaming during anaesthesia.

• Three studies59,60,63 that measured consumption of propofol reported different results; significantly lower consumption was found in the Narcotrend group in two studies, whereas no difference was reported between groups in the third study.

• Three studies60,63,64 found no significant difference between groups in remifentanil or desflurane consumption.

• Two studies63,64 reported time to arrival at PACU and found statistically significantly shorter times in the Narcotrend group compared with standard care.

• Time to eye opening was the primary outcome in two studies63,64 which yielded conflicting results; one reported a significantly lower time in the Narcotrend group than with standard care and the other reported no difference between groups.

• Time to extubation was the primary outcome in one study60 which found no difference between groups; two other studies63,64 that reported this measure as a secondary outcome found significantly shorter time to extubation with Narcotrend monitoring than with standard care.

• Time to arousal and duration of orientation recovery were reported to be shorter with Narcotrend monitoring compared with standard care in the one study59 reporting these outcomes.

• Results suggest that there are no differences between groups in PONV after surgery or analgesic consumption from the two studies that report these outcomes.

• In summary, Narcotrend monitoring compared with standard practice during minor orthopaedic surgery resulted in shorter recovery times (eye opening, arrival at PACU and time to extubation) and reduced propofol consumption. It was also associated with lower doses of propofol and shorter recovery during TIVA with propofol and fentanyl in liver cancer microwave coagulation. Narcotrend-assisted propofol–remifentanil anaesthesia did not reduce propofol or remifentanil consumption or time to extubation compared with standard clinical assessment in patients undergoing a range of elective surgery. The majority of the outcomes reported in the studies of Narcotrend were secondary and may not have been adequately powered statistically to detect clinically relevant differences. Also, the trial results are applicable to the specific patient groups included in the studies for the type of anaesthesia used and are not generalisable beyond this.

Overview

A decision-analytic model was developed to assess the cost-effectiveness of depth of anaesthesia monitoring, compared with standard clinical monitoring, in accordance with the scope of the appraisal issued by NICE. Separate analyses are presented for each of the three included technologies (the included technologies are not compared with each other).

The model was structured to include outcomes identified in the scope issued by NICE for this appraisal, where suitable data on the relative effectiveness of included technologies was identified in our systematic review of patient outcomes (see Results of systematic review of patient outcomes). The model evaluates costs (UK pounds using a 2011 price base) from the perspective of the NHS and Personal Social Services. Outcomes in the model are expressed as QALY. Both costs and outcomes are discounted using a 3.5% annual discount rate, in line with current guidance. 38,39

Modelling approach and model structure

The model developed for this assessment was a simple decision tree, which accounted for patients' risk of experiencing short-term anaesthetic-related complications (such as PONV) and more serious complications that may be associated with risk of morbidity or mortality. These were included, in addition to a risk of experiencing intraoperative awareness, see Figure 6.

FIGURE 6.

Decision tree evaluating cost-effectiveness of depth of anaesthesia monitoring compared with standard clinical monitoring.

Each of the short-term anaesthetic-related complications could be associated with additional treatment costs (such as antiemetic medication for patients experiencing PONV, whereas for patients experiencing POCD there may be in-hospital costs of managing the condition, additional days of hospital stay and, for longer-term cases, additional costs of managing the condition following discharge). No direct cost-consequences for intraoperative awareness are included in the model. However, it is assumed that a proportion of patients who experience awareness will suffer psychological symptoms arising from the awareness episode and that a proportion of those will develop PTSD and may seek treatment.

We assumed that monitoring of basic clinical signs, including blood pressure and heart rate (mandatory worldwide), would be common components to standard clinical monitoring and to depth of anaesthesia monitoring using EEG devices (as discussed in Description of technologies under assessment) and, therefore, these have not been costed in the model. The key cost components identified for the standard clinical monitoring branch of the model are the costs of anaesthesia, costs of anaesthesia-related complications and costs of managing long-term sequelae of intraoperative awareness, with baseline levels (unit costs, estimated baseline consumption of anaesthetics and estimated baseline incidence of anaesthesia-related complications/intraoperative awareness) defined at the root node of the tree. The effects of EEG-based depth of anaesthesia monitoring (using the included technologies) compared with standard clinical monitoring, which have been identified and assessed in the systematic review of patient outcomes, are applied to the baseline estimates, at the depth of anaesthesia monitoring node. These are applied as proportionate changes or OR/relative risks.

No quality-of-life (QoL) impact (utility loss) is included in the model for short-term anaesthesia-related complications (such as PONV) as these are expected to be of limited duration. Similarly, the model does not include an estimate of the QoL impact (utility loss) for an intraoperative awareness episode. The most significant quality-of-life QoL impact of any intraoperative awareness experience is assumed to be captured by estimating the incidence of psychological symptoms arising as a result of the awareness episode (including cases of PTSD).

As indicated, data population of the model required the estimation of baseline risks for a number of parameters in addition to the effectiveness estimates drawn from the systematic review of patient outcomes. The following section identifies the model parameters and the data sources used in the model.

Cost of depth of anaesthesia monitoring

The costs of depth of anaesthesia monitoring consist of the capital costs associated with acquisition of the module and recurring costs associated with sensors which are attached to the patient. Table 31 summarises the costs supplied by manufacturers for each of the modules included in the assessment.

Equivalent annual costs for each module (assuming a 5-year useful life for the equipment and a discount rate of 3.5%) are presented in Table 32.

The annual throughput of patients for each module is assumed to be 1000 patients per year (equivalent to four patients per day for a working year of 250 days) if used for patients at average risk of intraoperative awareness, based on discussion with clinical experts. We assumed that throughput would be halved if depth of anaesthesia monitoring was limited only to patients at high risk of intraoperative awareness (equivalent to two patients per day for a working year of 250 days) – the impact of assumptions regarding patient throughput on the unit costs for DoA modules is tested in scenario analyses.

Additional costs

The manufacturers' submissions to NICE indicate minimal additional power consumption associated with the modules. Therefore no additional costs were added to account for this.

The need for additional training for staff to operate the monitor appears to vary by model, according to the industry submissions. Narcotrend models require a day for the delivery of a lecture and training in the operating theatre or intensive care unit (ICU). The technical part of the training (handling the Narcotrend device, electrode placing) requires < 1 hour. The manufacturer of the E-Entropy model states that a 30-minute introductory training session is required in placement of the sensors, whereas no additional training is required for the use of a BIS monitor. This is not currently accounted for in the model.

The Narcotrend device included in this assessment is a stand-alone monitor [although the manufacturer's submission states that it can also send data in real time to other anaesthesia monitors (makes and coverage not specified)], whereas BIS and E-Entropy are modules designed to operate with other anaesthesia monitors. BIS is compatible with a range of monitoring platforms.

(Commercial-in-confidence information has been removed.) E-Entropy is compatible with GE Healthcare's most recent monitor range (CARESCAPE Monitors B850 and B650), but not older software levels (in GE Healthcare monitors) or with monitors produced by other manufacturers. The manufacturer's submission estimates that 45% of all UK operating theatres would be compatible with E-Entropy – for the remaining 55% significant investment in new monitoring equipment may be required for compatibility. Costs based on Table 31 would not be representative for facilities requiring such investment in new monitoring equipment.

The manufacturers did not supply any information on maintenance costs or costs of maintenance contracts for any of the depth of anaesthesia modules. As a result, the base case excludes any costs for recurrent maintenance. The potential impact of maintenance costs are examined in scenario analyses using assumptions regarding maintenance costs (annual maintenance costs estimated at 10% and 20% of the module acquisition cost).

Summary of unit costs for depth of anaesthesia modules

Unit costs for DoA modules include acquisition costs for the module (annualised, assuming a 5-year effective life, and converted to an average cost per patient using assumptions on patient throughput) and recurring costs arising from the single-use sensors attached to the patient.

Unit costs included in the base case do not include estimates of the cost of formal training or familiarisation with equipment or maintenance costs.

Anaesthetic dose

Postoperative nausea and vomiting

Our systematic review of patient outcomes identified limited evidence of the impact of DoA monitoring on the risk of PONV. A baseline risk of PONV (30%)102–104 for standard clinical monitoring and DoA monitoring has been included in the model. The sensitivity of the results to the potential impact of depth of anaesthesia monitoring on the risk of PONV is explored in a scenario analysis using data from a meta-analysis on the effectiveness of BIS on a range of outcomes including PONV by Liu. 105 We assumed that all treatments (such as prophylaxis against PONV) were the same for each monitoring group, and that all patients experiencing PONV were treated using 4 mg ondansetron by intramuscular or slow i.v. injection (unit cost = £5.39; BNF33).

Postoperative cognitive dysfunction

Intraoperative awareness

Sequelae of intraoperative awareness

Incidence of psychological sequelae

A targeted search for studies reporting symptoms of patients who had reported awareness during surgery was undertaken in order to understand the health-related consequences of intraoperative awareness.

Eight studies were identified5,7,8,10,11,19,114,115 (Table 41). These suggested that the patients who had experienced intraoperative awareness fall into three groups: those who do not experience any sequelae, those who experience ‘late psychological symptoms’ and those who go on to suffer from PTSD. Late psychological symptoms (LPS) comprise anxiety, chronic fear, nightmares, flashbacks, indifference, loneliness and a lack of confidence in future life. Anxiety, nightmares and flashbacks appeared to be the predominant symptoms in the study by Samuelsson and colleagues115 in patients with an LPS duration of < 2 months; those experiencing symptoms for a longer duration reported nightmares and flashbacks alone. A diagnosis of PTSD is made if all six criteria of the clinician-administered PTSD scale (CAPS) are positive. These include symptoms of re-experiencing trauma, avoidance, hyper-arousal, significant distress and the duration of symptoms lasting longer than 1 month. 8

TABLE 41 - Studies reporting incidence of LPS and PTSD in patients who experienced awareness – summary of characteristics, methods and results

DSM-IV, Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition; NR, not reported.

Total claims for adverse outcomes between 1961 and 1995 in closed claims project was 4183.

Of the 18 patients experiencing awareness identified by Sandin and colleagues,9 two could not be contacted, six declined to participate and one had died.

Samuelsson identified eight (17%) patients as having a total symptom score (summed across seven symptoms) of > 2 (no rationale for this threshold).

Of the 13 patients experiencing awareness in the B-Aware trial,79 six had died.

Leslie and colleagues8 pooled their estimate of PTSD with the estimates of Lennmarken and colleagues11 and Samuelsson and colleagues115 for severe psychological sequelae (n = 8, 17%) to derive an incidence of 26% (95% CI 15% to 37%).

Study	Date	Method of recruitment	Identification/classification of LPS and PTSD	Aware	LPS		PTSD
				n	n	%	n	%
Evans116	1987	Advertisement in four British newspapers		27
Moerman et al.5	1993	Referral from university hospital anaesthesiology department	Response to (open-ended) interview question – ‘have there been any consequences?’ (of the identified awareness episode). Patients reported sleep disturbance, dreams and nightmares, flashbacks and anxiety during the day	26	18	69	NR
Schwender et al.114	1998	Advertisements in four German papers and on internet (n = 21) or referral from three hospital anaesthesia departments (n = 24)	Response to questionnaire items on after effects (including anxiety and nightmares). No definition for PTSD reported (simply states ‘whether … PTSD syndrome developed’)	45	22	49	3	7
Domino et al.19	1999	Retrospective analysis of American Society of Anaesthetists Closed Claims Project (malpractice claims) – data from 1961 to 1995a	No definitions – reports states ‘% (n) sustained temporary emotional distress, whereas in % (n) post-traumatic stress disorder developed’	61	51	84	6	10
Osterman et al.10	2001	Advertisement in newspapers, fliers in hospitals, self-referral following print and TV news stories or referral by anaesthetist	PTSD defined using CAPS	16	NR		9	56
Lennmarken et al.11	2002	18 patients identified as experiencing awareness during GA in two hospitals (reported by Sandin and colleagues9) were followed up for interview regarding psychological symptomsb	PTSD defined using diagnostic criteria A1-F in the DSM-IV, American Psychiatry Association117	9	7	78	4	44
Samuelsson et al.115	2007	Consecutive patients who had undergone GA were interviewed regarding awareness during previous GA	LPS were any one of: anxiety, chronic fear, nightmares, flashbacks, indifference, loneliness and lack of confidence in future life (each rated on a scale from zero to two). PTSD appears to be defined on basis of existing clinical diagnosis (not specifically identified or classified in study)	46	15c	33	1	2
Ghoneim et al.7	2009	Data extracted from published case reports on ‘awareness’ and ‘anaesthesia’ – from PubMed between 1950 and August 2005	No definition of LPS	271	NR	22	NR
Leslie et al.8	2010	13 patients identified as experiencing awareness in the B-Aware trial (reported by Myles and colleagues79) were followed up for interview regarding psychological symptomsd	PTSD defined using CAPS	7	NR		5	71

Just two of the studies had a prospective design. 8,115 The study by Samuelsson and colleagues115 reported 46 awareness cases in a cohort of 2681 interviewed after surgery. This is therefore the strongest evidence for development of PTSD and LPS that was identified in the targeted search. Leslie and colleagues,8 although reporting a small cohort, were the only authors among those identified to report time to onset and duration of symptoms. However, some cases of PTSD reported were ongoing, and it is unclear how this may impact on the duration results. The two prospective studies were used to inform the baseline data inputs, for the states of LPS and PTSD, into the model, as presented in Table 42 below. The six remaining studies were small, with limited usefulness for understanding the prevalence of psychological symptoms associated with awareness, because of retrospective design, participant recruitment methods or low recruitment levels. 5,7,10,11,19,114

TABLE 42 - Baseline values for probability of LPS and PTSD in patients experiencing awareness

Sequelae	Value (95% CI)	Method	Source
LPS
Probability, given awareness	0.326 (0.195 to 0.480)	15/46 patients with awareness	Samuelsson et al.115
PTSD
Probability, given awareness	0.177 (0.113 to 0.230)	Pooled proportion of subjects with LPS having PTSD or severe symptoms, from (two) studies reporting this proportion, applied to probability of LPS	Samuelsson et al.115 and Lennmarken et al.11
		Pooled estimate based on 0.57111 (4/7) and 0.0533115 (8/15) = 0.542 (95% CI 0.345 to 0.733)
		Probability PTSD = (15/46) × 0.542

Duration of post-traumatic stress disorder

Leslie and colleagues8 reported a median duration of 4.7 years (range 4.4 to 5.6 years) for patients experiencing symptoms of PTSD. No further information on the distribution is provided so it unclear how well the median approximates to the mean duration of symptoms, as these cases of PTSD reported were ongoing, and it is unclear how this may impact on the duration results. Targeted searches did not identify any other studies reporting duration of PTSD symptoms associated with intraoperative awareness. One study was identified which reported duration of PTSD (median duration and survival curves) in a non-institutionalised, civilian population aged 15–54 years, conducted in the USA. 118 These data were from the National Comorbidity Survey (a survey designed to study the distribution, correlates and consequences of psychiatric disorder in the USA) and included 5877 respondents from 48 states. Response rates to part 2 of the survey, which included components related to PTSD, were between 98.1% (for those screening positive for any lifetime diagnosis in part 1 of the survey) and 99% (for a random subsample of those not screening positive in part 1 of the survey). The median duration of symptoms for respondents who had ever sought professional treatment (n = 266) was 36 months and for those who had not sought professional treatment (n = 193) was 64 months. We estimated the mean duration of PTSD symptoms for the population who had not sought professional treatment, by fitting a regression (assuming a Weibull distribution for the survival function; see Appendix 10 for details) to the reported survival curves. The mean duration of PTSD symptoms derived in this analysis was 152 months (12.7 years) (Figure 7).

FIGURE 7.

Survival curve based on duration of symptoms for respondents who did not seek treatment for PTSD, reported by Kessler and colleagues,118 and fitted Weibull model.

Quality of life impact of psychological sequelae

A review of the HRQoL of patients with PTSD was undertaken in order to explore the differences in scores between PTSD patients and those who had also experienced trauma, but had not gone on to develop PTSD. These scores were used to inform those in the model for patients experiencing awareness and developing psychological symptoms.

Methods

A systematic search was undertaken in order to identify studies reporting utility values associated with PTSD. The details of the search strategy are documented in Appendix 11. A total of 334 studies were initially identified by the search. The abstracts were screened by two independent reviewers and 21 full papers were retrieved (Figure 8). These were assessed against the inclusion criteria detailed in Table 43.

FIGURE 8.

Flow chart of identification of QoL studies for inclusion in the review.

TABLE 43 - Inclusion criteria for QoL review

EQ-5D, European Quality of Life-5 Dimensions; SF-6D, Short Form questionnaire-6 Dimensions; TTO, time trade-off.

Criteria	Include	Exclude
Participants	Adults with PTSD	Studies related to or concerning specific morbidities, with the exception of psychiatric (or related) illness
Design	Studies that report a utility value, based on generic preference based measures for QoL, such as EQ-5D, SF-6D, or other standard valuation technique such as standard gamble or TTO
Interventions	Any
Other	Articles published in English	Articles in languages other than English
		Conference abstracts

Summary

• Two papers met the inclusion criteria for this review of utility scores in PTSD. Six other papers reporting SF-36 scores for people with PTSD were also retained.

• Neither of the studies meeting the inclusion criteria (reporting a utility value based on a generic, preference-based measure) was based on the EQ-5D.

• One study reported a utility score for patients with PTSD based on TTO. 119 However, no score for patients without PTSD was reported, and therefore no difference in these can be derived.

• The second study reported scores for patients both with and without PTSD, but these were based on the SF-36 and converted to a utility score. 120

• Therefore the evidence base for HRQoL in patients with PTSD is limited.

• Six further studies provide SF-36 scores, which have been transformed into utility values. These can provide context and values for sensitivity analysis.

Post-traumatic stress disorder costs

The costs of treating PTSD have been estimated based on assumptions contained in the national cost impact report133 associated with NICE Clinical Guideline no. 26133 on the management of PTSD in adults and children in primary and secondary care. 134 The costing report acknowledged that there has been little systematic collection of information about PTSD, on services provided to people with PTSD or on uptake of these services. This limited the feasibility of developing a comprehensive bottom-up costing model and resulted in the costing being based on a series of assumptions – developed and validated through discussion with members of the Guideline Development Group (GDG) and key clinical practitioners in the NHS. These assumptions, in terms of uptake and services available, are summarised in Figure 9 and are discussed below.

FIGURE 9.

Care pathways and costing assumptions developed for NICE.

Data from the adult psychiatric morbidity survey,135 which reported that 24% respondents assessed as having a neurotic disorder were receiving treatment of some kind at the time of interview, were used as the basis for estimating the current proportion of people with PTSD who seek treatment. On the basis of additional data from the same survey, 62.5% of these were assumed to be receiving pharmacological therapy alone, 16.7% were receiving counselling or therapy alone and 20.8% were receiving both. It was assumed that, following implementation of the guideline, the proportion receiving treatment would increase by 10%, to 34%. Moreover, the guideline proposed a substantially different care pathway with significantly fewer patients expected to receive medication (with a recommendation that drug treatments not be offered routinely as first line, but to provide trauma-focused psychological treatment to more patients with PTSD symptoms). We estimated an average cost for management of PTSD using the assumptions regarding take-up of treatment options (70% of patients accept psychological treatment, and 30% initially accept pharmacological treatment) and severity (30% patients have mild symptoms and are initially managed through watchful waiting, and 20% have severe symptoms and are offered trauma-focused psychological treatment within the first month after the traumatic event). Table 47 summarises the unit costs, assumptions regarding the proportion of patients receiving each treatment and the overall cost estimated for PTSD.

TABLE 47 - Unit cost and treatment uptake assumptions used to calculate costs of managing PTSD

Patients are assumed to remain on pharmacological therapy for the duration of their PTSD symptoms (12.7 years in the base case). Costs are discounted at 3.5%.

Patients requiring ongoing psychological therapy are assumed to continue treatment for the duration of their PTSD symptoms (12.7 years in the base case). Costs are discounted at 3.5%.

Treatment	Unit cost (£)	Proportion (%)	Cost (£)
Watchful waiting	20.20	16.50	3.33
Pharmacological therapy	240.32	30.00	915.62
Combined pharmacological and psychological therapy	240.32	5.00	152.60
Psychological therapy (severe acute cases < 1 month)	272.40	14.00	38.14
Psychological therapy (> 1 month after traumatic event)	437.39	38.53	168.50
Psychological therapy (severe acute cases > 1 month)	437.39	4.62	20.21
Additional, ongoing psychological therapy	181.60	9.97	229.86
Total			1528.26

The NICE guideline133 does not include any estimates for inpatient care for people with PTSD. Targeted searches did not identify any UK studies of health service use, in particular use of secondary services and inpatient care for people with PTSD. One US study identified by the searches reported health-care utilisation, derived from electronic medical records, for civilian primary care patients, including a proportion who had current PTSD. 136 This study reported an incidence rate ratio of 2.22 (adjusted for age, sex, income, substance dependence, depression and comorbidity) for hospitalisation in subjects with PTSD compared with those without PTSD. Unadjusted mean number of hospitalisations among the PTSD group was 0.43 compared with 0.18 in those without PTSD. No further details are reported on the reason for hospitalisation or length of stay. In the absence of data specific to the UK for people with PTSD, we have assumed, based on the mean values reported in this study, an excess hospitalisation probability of 0.25 per year among people with PTSD. We derived a crude estimate of the average cost of hospitalisation (£2590), based on 2010–11 NHS Reference Costs,137 by summing the total costs reported for elective and non-elective inpatient HRG data and dividing by the total activity under these headings. On this basis we estimated an additional £7576 for hospitalisations among people with PTSD over the average duration of symptoms of 12.7 years.

The total cost associated with PTSD was £9104 (undiscounted) or £6128 (discounted at 3.5%).

Summary of model inputs

The following tables contain a summary of the input parameters in the model, the base-case value and a brief overview of how the data were derived including a source, where relevant. Table 48 provides a summary of the cost per patient of each depth of anaesthesia technology, including an estimated cost per patient of the depth-monitoring device as well as the cost of consumables (single-use sensors attached to the patient). Table 49 provides a summary of the baseline cost of anaesthetic drug calculated for standard clinical monitoring in each comparison and the proportionate reduction in consumption associated with depth of anaesthesia monitoring. We have assumed that the reduction in consumption of anaesthetic will be realised only for the general surgical population and not in the population at high risk of awareness, as the raised risk of awareness may be an indication that this group of patients are already at a risk of being underdosed.

TABLE 48 - Model input parameters–cost per patient of DoA modules

Parameter	Value (£)	Source
BIS
Cost per patient of depth monitoring device	0.96	Equivalent annual cost for depth monitor (acquisition cost £4350) assuming an effective life of 5 years and using a discount rate of 3.5%. Patient throughput assumed at 1000 per year
Cost per patient of depth monitor sensors	14.50	Manufacturer's price of £362.50 for a box of 25 sensors (for Vista monitor)
E-Entropy
Cost per patient of depth monitoring device	1.19	Equivalent annual cost for depth monitor (acquisition cost £5352) assuming an effective life of 5 years and using a discount rate of 3.5%. Patient throughput assumed at 1000 per year
Cost per patient of depth monitor sensors	8.68	Manufacturer's price of £217 for a box of 25 sensors
Narcotrend
Cost per patient of depth monitoring device	2.28	Equivalent annual cost for depth monitor (acquisition cost £10,285, mid-point of range quoted by manufacturer) assuming an effective life of 5 years and using a discount rate of 3.5%. Patient throughput assumed at 1000 per year
Cost per patient of depth monitor sensors	0.56	Average across manufacturer's price of £0.14 per sensor, using three for one-channel recording and five for two-channel recording

TABLE 49 - Model input parameters – anaesthetic drug consumption

Parameter	Value (95% CI)	Source
BIS
Baseline inhaled anaesthetic cost	£11.04	Cost for 1.6 MAC hours (95 minutes) of sevoflurane (concentration of 1.8% and fresh gas flow rate of 4 l/minute). Unit cost of £0.59 per ml, based on price of £148 per 250 ml
Reduction in consumption of inhaled anaesthetic using depth monitor (proportionate reduction compared with standard clinical care)	−0.202 (−0.330 to −0.074)	Mean difference of −0.15 from a (weighted) mean consumption of 0.765 MAC equivalents
Baseline i.v. anaesthetic cost	£20.92	Cost for 2 hours of propofol [at 6.77 mg/kg/hour (from control arms of RCT in meta-analysis) and patient average weight of 77 kg]. Unit cost of £0.0202 per mg
Reduction in consumption of i.v. anaesthetic using depth monitor	−0.193 (−0.272 to −0.113)	Mean difference of −0.130 from a (weighted) mean consumption of 6.73 mg/kg/hour
E-Entropy
Baseline inhaled anaesthetic cost	£15.93	Cost for 2.3 MAC hours (137 minutes) of sevoflurane (concentration of 1.8% and fresh gas flow rate of 4 l/minute). Unit cost of £0.59 per ml, based on price of £148 per 250 ml
Reduction in consumption of inhaled anaesthetic using depth monitor	−0.286 (−0.492 to 0.079)	Mean difference of −0.04 from patient normalised consumption of 0.14 g/kg/hour (in standard care arm, Aime et al.61)
Baseline i.v. anaesthetic cost	Propofol = £18.85	Ellerkmann et al.62
	Remifentanil = £4.26
	Propofol = £14.35	Gruenewald et al.55
	Remifentanil = £14.94
Reduction in consumption of i.v. anaesthetic using depth monitor	0.050 (−0.075 to 0.174)	Propofol mean difference of 5 from baseline rate of 101 mg/kg/hour (Ellerkmann et al.62)
	−0.111 (−0.232 to 0.010)	Remifentanil mean difference of −0.01 from baseline rate of 0.09 mg/kg/hour (Ellerkmann et al.62)
	−0.147 (−0.237 to −0.058)	Propofol mean difference of −14 from baseline rate of 95 mg/kg/hour (Gruenewald et al.55)
	0.179 (0.085 to 0.274)	Remifentanil mean difference of 0.07 from baseline rate of 0.39 mg/kg/hour (Gruenewald et al.55)
Narcotrend
Baseline inhaled anaesthetic cost	£24.09	Cost for 2.1 MAC hours (125 minutes) of desflurane (concentration of 6.6% and fresh gas flow rate of 4 l/minute). Unit cost of £0.30 per ml, based on price of £76 per 250 ml
Reduction in consumption of inhaled anaesthetic using depth monitor	−0.156	Mean difference of −69 mg/minute from 443.6 mg/minute (in standard care arm, Kreuer et al.64)
Baseline i.v. anaesthetic cost	Propofol = £19.39	Cost for 108 minutes of propofol [at 6.81 mg/kg/hour (from control arms of RCT) and patient average weight of 80 kg]. Unit cost of £0.0202 per mg
	Remifentanil = £10.79	Cost for 108 minutes of remifentanil [at 0.120 mg/kg/hour (from control arms of RCT) and patient average weight of 80 kg]. Unit cost of £5.12 per mg
Reduction in consumption of i.v. anaesthetic using depth monitor	−0.292 (−0.429 to −0.155)	Propofol mean difference of −1.99 from baseline rate of 6.8 mg/kg/hour60,63
	−0.054 (−0.158 to 0.050)	Remifentanil mean difference of −0.01 from baseline rate of 0.25 mg/kg/hour60,63

Table 50 provides a summary of model inputs related to awareness including the baseline risks for patients considered at high risk of awareness and a general surgical population, the risk reduction associated with depth of anaesthesia monitoring and a list of assumptions underlying the estimation of the cost and outcomes associated with the psychological sequelae of intraoperative awareness.

TABLE 50 - Model input parameters – intraoperative awareness

Parameter	Value (95% CI)	Source
Intraoperative awareness
Baseline awareness in surgical population at high risk of awareness	0.45% (0.06% to 1.19%)	Pooled estimate from control arms of RCT in high-risk patients
Reduction in awareness using depth monitor		Meta-analysis of RCT in high-risk patients, undertaken as part of this review (see Assessment of outcomes: Bispectral Index)
High-risk patients undergoing TIVA (Peto's OR)	0.24 (0.10 to 0.60)
High-risk patients undergoing anaesthetic induction with i.v. and maintenance with inhaled anaesthetic (Peto's OR)	0.45 (0.25 to 0.81)
Baseline awareness in general surgical population	0.16% (0.10% to 0.23%)	Pooled estimate from studies reporting incidence of awareness, not specified to be high risk
Reduction in awareness using depth monitor
General surgical population undergoing TIVA (Peto's OR)	0.24 (0.10 to 0.60)	Meta-analysis of RCT in high-risk patients, undertaken as part of this review (see Assessment of outcomes: Bispectral Index).
General surgical population undergoing anaesthetic induction with i.v. and maintenance with inhaled anaesthetic (Peto's OR)	0.45 (0.25 to 0.81)	Effect assumed to be the same as for high-risk patients
Psychological sequelae of intraoperative awareness
Probability of LPS, given awareness	0.326 (0.195 to 0.480)	Samuelsson et al.115
Duration of LPS	6 months	Assumption
Unit cost of LPS	0	Assumption
Utility reduction due to LPS	Same as PTSD	Assumption
Probability of PTSD, given awareness	0.177 (0.113 to 0.230)	Samuelsson et al.115 and Lennmarken et al.11
Duration of PTSD	12.7 years	Kessler et al.118
	7.32 years, discounted at 3.5% (8.2 to 21.6 years, 5.6 to 9.6 discounted at 3.5%)
Unit cost of PTSD	£9104	NICE [consists of £915.62 (60%) pharmacological therapy, £456.71 (30%) psychological therapy and £152.60 (10%) combined pharmacological and psychological therapy]. Excess risk of hospitalisation 25% annually.136 Average cost of inpatient stay. NHS Reference Costs 2010–2011137
Utility reduction due to PTSD	0.12	Various

Table 51 provides a summary of model inputs relating to anaesthetic complications (PONV and POCD in the model), including the baseline risks and the risk reduction associated with depth of anaesthesia monitoring for POCD.

TABLE 51 - Model input parameters – postoperative complication (PONV and POCD)

Parameter	Value	Source
PONV
Baseline PONV	30%	Cohen MM, Duncan PG, DeBoer DP, Tweed WA. The postoperative interview: assessing risk factors for nausea and vomiting. Anaesth Analg 1994;78:7–16
Reduction in PONV using depth monitor	Not included in base case	Included as a scenario analysis
Unit cost of PONV	£5.39	£5.39 (4 mg of ondansetron)
Utility reduction due to PONV	0
POCD
Baseline POCD	Average duration of 29.65 days	ISPOCD study reported POCD in 25.8% (95% CI 23.1% to 28.5%) of patients at 1 week and in 9.9% (95% CI 8.1% to 12.0%) of patients at 3 months after surgery: compared with 3.4% and 2.8%, respectively, in UK controls. At median follow-up of 532 days, 10.4% patients had cognitive dysfunction compared with 10.6% controls (47 non-hospitalised volunteers of similar age). Assume excess of 22.4% at 7 days, reducing to excess of 7.10% at 3 months and excess of 0% at 1.5 years (532/365.25 years) – area under curve = 29.65 days
Reduction in POCD using depth monitor	Average duration of 21.10 days	Chan and colleagues'47 abstract reported 32.5% (BIS) vs 39.1% (standard clinical monitoring) at 7 days and 8.1% (BIS) vs 12% (standard clinical monitoring) at 3 months
		OR estimated as 0.75 (at 7 days) and 0.646 (at 3 months) – applied to excess proportions above. Assume average duration of 21.10 days
Unit cost of POCD	0
Utility reduction due to POCD	0.05	Jonsson et al.:110 difference in utility between an MMSE score > 25 (0.69) and an MMSE score between 21 and 25 (0.64). Normal to mild cognitive dysfunction

Bispectral Index compared with standard clinical monitoring