National Institute for Clinical Excellence Guidance on Measuring Depth of Anaesthesia: Limitations of EEG-Based Technology
Total Page:16
File Type:pdf, Size:1020Kb
Volume 110, Number 3, March 2013 British Journal of Anaesthesia 110 (3): 325–8 (2013) doi:10.1093/bja/aet006 EDITORIAL I Downloaded from https://academic.oup.com/bja/article/110/3/325/250090 by guest on 30 September 2021 National Institute for Clinical Excellence guidance on measuring depth of anaesthesia: limitations of EEG-based technology J. J. Pandit1* and T. M. Cook2 1 Nuffield Department of Anaesthetics, Oxford University Hospitals, Oxford, UK 2 Royal United Hospital, Bath, UK * E-mail: [email protected] Accidental awareness during anaesthesia, especially with The available evidence on the impact of the technologies on pain and subsequent recall of the event, is a terrifying pros- reducing the likelihood of intraoperative awareness is limited. Overall, [EEG-base monitors are] not associated with a statistic- pect. A high proportion of those who experience it are ally significant reduction in intra-operative awareness in reported to go on to develop symptoms similar to post- patients classified as at higher risk ... traumatic stress disorder.1 The reported incidence of recall after general anaesthesia, as ascertained by later question- What is the anaesthetist reader (or patient) to make of these ing using the Brice questionnaire, is high, 1–2 per 1000 contemporaneous and conflicting conclusions? cases (although the majority of these do not involve painful The bodies providing these two sources of advice had very experiences, are very brief recollections, or both).2 –8 There- different remits, which may explain the opposing conclu- fore, a monitor to help the anaesthetists identify those sions. The Technology Assessment Report was a review (in- patients who are awake during surgery would be extremely cluding a meta-analysis which incorporated results of a 11 useful. recent Cochrane review). The NICE guidance, on the other The publication of Diagnostic Technologies Guidance from hand, is a consensus opinion which takes into account the the National Institute for Clinical Excellence (NICE) on the findings of the Technology Report but also includes expert role of EEG-based monitoring in reducing the risk of aware- and individual views, perhaps importantly including that ness during general anaesthesia may lead patients, politi- from companies which manufacture the devices under con- cians, and planners (and some anaesthetists) to conclude sideration. The broad aim of the NICE Diagnostics guidance that we now have an answer.9 The guidance concludes that: programme is explicitly to promote rapid adoption of clinic- ally innovative and cost-effective diagnostic technologies, The use of EEG-based depth of anaesthesia monitors is recom- based on their potential (not necessarily proven ability)to mended as an option during any type of general anaesthesia in improve care (see: http://www.nice.org.uk/media/A0B/97/ patients considered at higher risk of adverse outcomes. DAPManualFINAL.pdf). However, that is not quite the full story. The guidance was itself On the ‘probability of unconsciousness’ 10 based on a Technology Assessment Report (commissioned In several places, the NICE guidance appropriately stresses by National Institute of Heath Research Health Technology As- that a monitor’s output reflects the probability of (un)con- sessment programme) which concluded (in contrast to NICE’s sciousness in a given patient. It is repeatedly stated that a final guidance): BIS (and also E-entropy) output reading of 40 (or 40–60) & The Author [2013]. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: [email protected] BJA Editorial I indicates a ‘low’ probability of awareness. However, it is not indicated otherwise (i.e. the false negatives, FN) and the specified what ‘low’ means. Also not explained is how the number of cases of appropriate anaesthesia in which the probability of awareness varies as the monitor reading monitor indicated otherwise (the false positives, FP). These increases: 40 may be a low probability, but how much are then combined to yield quantities such as the sensitivity, higher is, say, 50 or 65? In other words, what is the shape specificity, and positive (and negative) predictive values of of the relationship between monitor output and probability, the diagnostic instrument. In the B-Aware trial, results if plotted graphically? were not presented in this way (and in the group of patients We already have answers to these questions, if they are whose anaesthesia was not BIS-guided, the monitor was posed in relation to expired volatile agent (end-tidal) concen- turned off, so no data were available). Nonetheless, the trations. This is the basis of the concept of the MAC data can be re-arranged to show the relevant discriminatory (minimum alveolar concentration), which is an ED50; the con- capacity of the monitor (Supplementary Table S1). Examined centration at which 50% of patients move in response to in this manner, the monitor’s utility is far from encouraging, a standard stimulus. The ED50 (or EDx; points when x%, with the sensitivity and specificity of 50% and the positive Downloaded from https://academic.oup.com/bja/article/110/3/325/250090 by guest on 30 September 2021 95%, 99% of patients do not move, etc.) has a constant predictive value of just 0.2%. value in arbitrary units, regardless of the agent used and In contrast to the B-Aware trial, the B-Unaware trial (with the concept can be used for a variety of endpoints. For a similar study design) did not find a reduction in awareness example in MAC, the endpoint is response to a standard inci- with BIS monitoring.2 This time, BIS records were available sion, while in MAC-awake, the endpoint is response to verbal for both groups. Again, re-arranging the data to explore diag- stimulus. In this way, MAC (or any EDx) ‘normalizes’ the popu- nostic efficacy of the monitor does not suggest great utility lation response across agents, and indicates the probability (Supplementary Table S2), with the sensitivity 25%, specifi- of no-response at any given dose, regardless of agent.12 – 15 city 45%, and positive predictive value of just 0.1%. A third Therefore, at 1 MAC (the ED50), there is always the same trial (BAG-Recall) had limited information of this type (it (50%) probability of no-response with halothane, isoflurane, also found no difference between BIS-guided anaesthesia or any other volatile anaesthetic. If a DOA monitor accurately and alternative groups), but the supplementary data reflected this fact, then its output should be the same at 1 appear to suggest the BIS remained ,60 in 12 of 27 patients MAC of any agent. Were the monitor to read, say, 35 at 1 who experienced awareness.3 MAC sevoflurane but 60 at 1 MAC halothane, it would not NICE did not analyse the data in this way. In using the trial be a very useful monitor. Yet this is exactly what is reported data in this manner, we have assumed (as did the trial inves- with EEG-based monitoring.16 Thus, a BIS value of just below tigators) that a BIS value of .60 is always indicative of 60 (i.e. within the recommended target values for both BIS awareness during anaesthesia, regardless of how brief this and E-entropy monitors) might in fact indicate an unaccept- period is. It is in fact not known what exactly the BIS criteria able risk of consciousness with some agents.17 NICE recom- are that suggest awareness (nor does NICE guidance help on mends that DOA monitoring should be considered when total this point). Is it the highest BIS in a certain time interval, or i.v. anaesthesia with propofol is used, but the data of Glass the average value, or a certain percentage increase over and colleagues18 clearly demonstrate similarly disparate time, etc.? In other words, should an anaesthetist be con- probabilities at equi-BIS values: at, say, a BIS value of 70, cerned during an anaesthetic if the BIS value is 40 for the probability of unconsciousness is 50% for isoflurane many hours but jumps to 75 for just 10 s? Or much more con- but just 15% for propofol. cerned in a situation where the BIS value is steady at 61 for 30 min? Or are these situations equivalent? We also do not The interpretation of existing clinical know if in these studies the anaesthetist (or anaesthetic agent) acted sufficiently promptly when the monitor indi- trial data cated that the risk of awareness was high.19 Both the NICE guidance and Technology Assessment Report To complicate matters, BIS appears to be ‘blind’ to several use data from large clinical trials to inform their conclusions. anaesthetic agents. Nitrous oxide is very well established to The B-Aware trial reported two of 1225 cases of awareness add to the effect of other volatile agents. Yet, the BIS value when anaesthesia was guided using a BIS monitor (to is unchanged (indeed, in one study, removing nitrous oxide achieve a BIS reading of 40–60) and 11 of 1238 without a caused BIS paradoxically to decline).20 21 Ketamine is BIS monitor (P¼0.022).5 The attractive conclusions are that popular with some anaesthetists, especially for emergency using the monitor has significantly reduced the occurrence surgery,22 but does not influence (or can even paradoxically of awareness and, furthermore, that maintaining the BIS in increase) BIS values.23 24 this range (i.e. ,60) will prevent consciousness. However, a monitor is not an intervention like a drug, Problems in applying the NICE guidance whose efficacy can be measured as the simple incidence of beneficial outcome. Rather a monitor is a diagnostic tool, into practice whose efficacy is usefully measured as the proportion of Even those enthusiastic about EEG-based monitoring may correct diagnoses it makes.