IN this issue of Anesthesiology, Struys et al . report that bispectral analysis (BIS®, Aspect Medical System Inc., Newton, MA, USA) and a derivative of the middle-latency auditory evoked response (MLAER) were similar in their ability to track levels of sedation and loss of consciousness during infusions of propofol, and that both had poor predictive power with respect to movement in response to noxious stimulus. 1The MLAER was analyzed using a soon-to-be commercially available device (A-Line®, Alaris Medical Systems, Inc., San Diego, CA). The appearance of articles like this one and the increasing availability of monitors that attempt to use cerebral electrophysiologic signals to track depth of anesthesia (BIS®, A-Line®, PSA 4000®[Physiometrix, N. Billerica, MA]) beg numerous questions. Among them are: What are we trying to accomplish with these monitors and can we accomplish those objectives with the monitors that are available today? As a lesser but related issue, if these monitors are eventually going to be effective (if they are not so now), is success more likely to occur with the recording of spontaneous electrophysiologic activity (BIS®, PSA 4000®) or with the recording of evoked signals (A-Line®)?
Before dealing with those questions, clinicians may benefit from considering some of the technical issues pertaining to these monitors. In particular, the principles underlying the processing performed by the A-Line®monitor may be unfamiliar to many. The A-Line®generates an index derived from analysis of the configuration of the MLAER. Auditory evoked responses have been studied since the early 1980s by groups from London, England (Thornton et al . 2–5), and Munich, Germany (Schwender et al. 6–9). However, a commercial device has now become feasible because the increasing availability of compact computing power and refinements in signal processing techniques have essentially made on-line monitoring possible. The MLAER consists of waves, which occur between 20 and 100 ms after auditory stimuli. The MLAER in a normal awake subject consists of a typical configuration with three vertex positive peaks. Increasing concentrations of a volatile agent “stretch” this waveform into a two-wave pattern with reduced amplitude and increased latencies. In at least one study, prolongation of the negative wave (Nb), which occurs normally at about 40 ms, to or beyond a latency of 47 ms, quite reliably distinguished between subjects who could and could not recall events during anesthetic administration. 2,3Among the original difficulties was that signal averaging of at least one thousand responses was necessary to extract this signal, making it impractical as an online monitor. The “autoregressive” (ARX) modeling technique used by the A-Line®monitor uses advanced filtering techniques to extract the waveform of interest more rapidly. The essence of the technique is that filtering parameters are modified on an ongoing basis. After acquisition of the baseline signal, filtering parameters are adjusted to create what functions as a “keyhole” that will admit only the signal of interest. That allows rapid verification of the continuing presence of that signal. As the signal varies, it cannot pass through the keyhole and it is possible therefore to make rapid notification of change.
The BIS®device is probably more familiar. One of the issues that has been of concern to some clinicians is that the precise workings of the innards of some of these devices (BIS®, PSA®) is proprietary and opaque to the user. While the general approach to the derivation of the BIS index has been described in this journal, 10the specifics of that index are, in fact, proprietary. It is known that the BIS®is calculated from several variables derived from the electroencephalogram as independent predictors of the probability of consciousness and that these predictors are combined, with various weightings, in a prediction rule to render a measure of hypnosis on a linearized 0–100 BIS®scale. The exact details of the parameters extracted and their weighting in determining the final score are not known to the user. Furthermore, the BIS algorithm is continuously being refined. While improving the reliability of the monitor has obvious merit, it has the potential disadvantage of making the validity of comparisons of results obtained by investigators using different versions of the monitor uncertain. This also leaves the clinician uncertain as to whether conclusions drawn from the study of earlier versions of the monitor remain clinically valid.
To return to the questions posed in the first paragraph, there are several possible objectives in the use of monitors of depth of anesthesia. Preventing awareness is the one that has made the biggest “splash” with the public and may perhaps loom largest with some clinicians. However, preventing unwanted hemodynamic responses, avoiding motor responses to noxious stimulus, preventing autonomic and adrenergic responses to stress, minimizing expenditures on anesthetic agents and expediting both awakening and postanesthesia care unit (PACU) discharge are all objectives that have been considered. All may be worthwhile. However, while it may be possible to argue that the depth of anesthesia monitors mentioned above (and perhaps others) can be potentially useful adjuncts to achieving various of these end-points, it is difficult to build a case from the published literature that they can do so definitively. The exception may be the use of BIS®monitoring in expediting awakening and discharge and reducing the cost of anesthetic agents. However, because we do not wish the economics of medicine to be the focus of this editorial, we will not plunge deeply into an examination of the merits of these two applications of monitors of depth of anesthesia. Suffice it to say that it is our opinion that the economic and outcome benefits of these applications, using any device, are not clearly established. The most recent related publication that we have seen demonstrated that BIS®monitoring resulted in a 3.6 min reduction in time to responsiveness (P < 0.05), a 12-min reduction in time to PACU discharge (NS) and a 2.1 ml reduction in the consumption of isoflurane (P < 0.05) in patients undergoing 2 h of anesthesia for hip or knee replacement procedures 11That isoflurane reduction results in cost savings of US .21¢ at both the University of California, San Diego Medical Center and the Utrecht University Medical Center (where a 100-ml bottle of isoflurane costs approximately US $10.00).
With respect to the other potentially valid reasons for employing these monitors, clinicians must be certain that their use for one purpose does not actually defeat others. For instance, if clinicians undertake to use one of these depth of anesthesia monitoring modalities to reduce anesthetic agent costs or expedite PACU discharge by maintaining patients below but close to some threshold perceived to correlate with loss of conscious perception, is it possible that the autonomic responses to stress (the consequences of which, granted, are ill defined) will actually be increased? Or, if the chosen threshold is not highly reliable in terms of assuring the absence of conscious perception (awareness), is it possible that clinicians, in the course of achieving cost savings will actually and unwittingly increase the incidence of awareness? The data of Sandin et al . indicate that, in a large population of patients undergoing elective or urgent surgery with neuromuscular blockade, awareness occurs with an incidence of approximately 1 of every 556 patients. 12In the study of Wong et al . mentioned above, the patients in the BIS® monitoring group were given anesthetic agents in sufficient amounts to maintain BIS®level between 50 and 60. 11Let us, for argument sake, say that the average BIS level achieved in those patients was 55. If patients with a BIS level of 55 have the potential to formulate memory just 0.5% of the time (1/200), then there is the potential that the clinicians might actually cause the occurrence of awareness at a rate higher than that, which occurs spontaneously in a nonmonitored population (1/556). A more detailed consideration of this concern has already appeared in this journal. 13This line of reasoning is not intended to discourage clinicians from exploring the application of these monitors, but rather to encourage applying them with an understanding of the sensitivity and specificity of those monitors. The ideal monitor for the detection of an event that occurs at low frequency (such as awareness) is one for which the range of values seen in patients who do and do not have the end-point of interest essentially never overlap (high sensitivity and specificity). The data of Struys et al. , indicate that this is probably not the case for either the BIS®or the parameter extracted from the MLAER by the A-Line®monitor 1. For confirmation, the reader should consult figure 3A,B of the Struys et al . article. 1Those authors assessed the level of sedation with the Observer's Assessment of Alertness/Sedation (OAAS) scale. Fig 3A and B present box plots for the BIS®and AAI (A-Line ARX Index) values achieved for patients at various OAAS levels of sedation. A comparison of the values from the two monitors at sedation level 3 (“responds only after the name is called loudly and/or repeatedly”), and 0 (“no response after painful trapezius squeeze”), reveals that values that correspond to unresponsiveness to noxious stimulus in some patients correspond to responsiveness to voice in others. To relate this once again to the paper of Wong et al ., 11in the group studied by Struys et al ., there were apparently many patients with BIS®scores between 50 and 60 who were responsive to voice command or to minimal prodding. 1Again we say that while these monitors may be useful adjuncts to various clinical objectives, including rapid awakening and cost savings, these monitors do not yet have the high level of discriminative power to be definitive methods for identifying depth of anesthesia end-points, and data from them must be considered with careful simultaneous attention to all of the other traditional signs that we have used to assess depth of anesthesia.
Other potential clinical objectives were mentioned earlier in this editorial. With respect to the use of these monitors to anticipate and prevent movement to noxious stimulus, the results of Struys et al . reveal the poor predictive power of both the A-Line®and BIS®monitors. Predicting and preventing autonomic response to stress was also mentioned. To our knowledge, there have been no investigations that have attempted to correlate these electrophysiologic monitors with autonomic responses.
The second question was whether spontaneous signals or evoked responses were more likely to be the basis of effective monitors of depth of anesthesia. Ultimately the answer to that question will be an empiric one derived from studies like that performed by Struys et al . 1Intuition should count for little in medical science; Aristotle's intuitions retarded progress in medicine for a thousand years. Nonetheless, it seems likely that extracting definitive depth of anesthesia information from the spontaneous activity of many millions of neurons representing many disparate subpopulations will be very difficult. Not all anesthetic agents interact with the same populations of receptors or have the same effects on axonal conduction and, accordingly, the constellation of the effects of our various anesthetic agents and their myriad combinations on the total neuron pool is likely to be very varied. That is not to say that, with a sufficient number of observations, indices that describe the electrophysiologic behavior of that total neuron pool cannot be extracted. That is precisely the approach that was used in the empiric derivation of the BIS index, which arguably tracks level of sedation more effectively and consistently than any of the electroencephalogram derivatives that preceded it. Yet it still has not achieved, even in the context of pharmacologic monotherapy (propofol), the sensitivity and specificity (again see fig. 3A, in the article by Struys et al .) that would be ideal for a monitor of depth of anesthesia. Superficially, it would appear that tracking the response of a much smaller subpopulation of neurons (i.e., the neural pathway corresponding to an individual evoked response) should be easier. Certainly, the effects of anesthetic agents on evoked responses (the auditory evoked response in the case of the A-Line®monitor) are likely to be much easier to describe empirically. But once again the problem that our anesthetic agents do not all interact with the same populations of receptors or have the same effects on axonal conduction will intrude. While the auditory evoked response may provide a good correlation in individual patients with depth of anesthesia for one class of agents (and it appears to do so with volatile agents 2,3and propofol), it may offer a much less apparent correlation for others (e.g ., narcotics, 8benzodiazepines 9) that have less effect on the auditory pathway. Whatever the potential for the use of MLAERs as a measure of the depth of anesthesia, they suffer one very substantial disadvantage. The empiric observations are that the MLAER becomes attenuated to the point of near unrecordability relatively early after loss of consciousness. Therefore, its most obvious application may be in the prevention of awareness while its potential applicability in prevention of movement, prevention of cardiovascular responses and prevention of autonomic responses to noxious stimulus or to stress seem likely to be more limited. It may ultimately prove, if we deem the grail of depth of anesthesia monitoring worth pursuing, that the optimal monitor of depth will be one that integrates parameters extracted from both spontaneous and evoked cerebral electrophysiologic signals.