1 Oulu, Finland 2 Hamilton, New Zealand
EditorIn a recent issue, Sleigh and Barnard1 discuss the properties of the spectral entropy index monitor, and why it fails to recognize unconsciousness in nitrous oxide anaesthesia. This finding is no surprise to any neurophysiologist who has experience of the neurophysiology of unconscious patients. A discussion on what this index really represents is therefore necessary.
Neurophysiologists have known for >50 yr the typical patterns of the EEG during anaesthesia, including the topographic features and reactivity, as well as differences in EEG patterns between different drugs. The same sequence of events is seen in several metabolic disorders and intoxications. After diffuse ischaemic brain damage the same pattern is seen in the order of recovery: suppression, burst suppression, slow activity, and then a normal EEG. This sequence is generated by pathophysiological events inside the brain, without the contribution of drugs or a metabolic disorder. But there are also many other causes of unconsciousness, such as an epileptic seizure, where the EEG does not follow this pattern. Ketamine- or nitrous oxide-induced unconsciousness are other examples. Furthermore, in spite of the common features, there are also considerable differences between different drugs and these are particularly evident in the EEG at the burst suppression level. And finally, as pointed out by Kissin,2 clinically similar anaesthesia can be produced by drug combinations with pharmacologically reversible effect, such as benzodiazepines and opiates, and drugs whose action cannot be reversed such as volatile anaesthetics: in both cases this same sequence of EEG events may be seen.
It is important to understand that we do not have a reasonable neurophysiological theory, which covers the whole variety of EEG patterns seen during anaesthesia. Yet, recent sleep research by Steriade and Amzica3 has considerably increased our understanding of slow sleep oscillations and their relation to sleep. We also have models of sleep spindles, which is another process. We have some idea about burst suppression, which is a third process, essentially independent of the other two. But most details of these patterns are still poorly understood. Developments in sleep research and epileptology, among other fields of neurophysiology, can be expected to increase our understanding of general anaesthesia in the near future.
The new EEG indices have brought no new understanding to what we knew 50 yr ago. They depend on successful merging of estimates of different processes such as the fast and slow oscillations, and burst suppression with its slower modulation and reactivity. Their interpretation is often based on physiologically erroneous assumptions, such as considering EEG suppression with one anaesthetic to be physiologically identical to that by another. The power spectrum is typically calculated from a non-stationary signal, which is a mathematical violation, as discussed by Miller and colleagues.4 The entropy literature sometimes refers to application of Shannons function to distribution of amplitude values, sometimes to the power spectrum, sometimes to approximate entropy, etc. All these are totally different phenomena, which may or may not correlate (i.e. they may change in the same direction or opposite directions when the EEG changes). Spectral entropy is erroneously claimed to estimate predictability and regularity of the signal, when it actually only measures how sinusoidal the signal is. A square wave signal is every bit as regular and predictable as a sine wave, but its spectral entropy is much higher than that of a sine wavesat least if the power spectrum refers to the power spectrum calculated with sine waves (i.e. FFT) although this is not the only way to calculate it.
There seems to be an irrational idea of entropy of the EEG, whichever of these totally different measures is referred to, to reflect somehow regularity or entropy at neural level or in larger systems in the brainand hence, consciousness. This reflects a total lack of understanding of how the EEG is generated and lack of experience of the clinical EEG. Spectral entropy was, in fact, first applied to alpha activityawake!1 Identical alpha activity, with identical spectral entropy, can be recorded from an irreversibly unconscious patient in alpha coma. Very regular delta activity (of very low spectral entropy) can be recorded during hyperventilation in awake young subjects. Many regular or irregular interictal epileptic discharges have no effect whatsoever on consciousness or other physiological functions: depth of anaesthesia-monitors may give any values between 100 and 0 in these instances. Their reading is meaningful only in the context of anaesthesia, and even then they do not measure consciousness. Consider the Wada test: one half of the brain can be anaesthetized to burst suppression, while the other half is awake.
What about entropy during ketamine anaesthesia? Beta activity increases and therefore the power spectrum has a clear peak at beta frequencies. By using an appropriately narrow bandpass we can construct a spectral entropy index for ketamine, only it relies on the beta and not the delta band, which is used for spectral entropy index when applied to most anaesthetics.
During burst suppression, the EEG is less sinusoidal than immediately before it, and this applies when bursts and suppressions occur either separately or together. This is only too evident if you look at the raw EEG, which is too non-stationary for a meaningful spectral estimate to be made of the spectral entropy of the delta activity.5 Spectral entropy of delta activity is changing in the wrong direction and has to be replaced by the burst suppression ratio, and the spectrum is best discarded.
Thus, there is no uniform EEG spectral entropy, which would change over the range of anaesthetic concentrations we meet in clinical practice. Neither is there any meaningful uniform theory of regularity or predictability, which is more sensitive in the clinical range. So there is really no point in referring to discussion of von Neumann and Shannons work.
Efforts should be taken to do basic research using EEG, evoked potentials, fMRI etc., and experimental work at different neurophysiological levels to learn to understand how different anaesthetic agents work. The new indices have been useful in taking the anaesthetists interest from spinal cord reflexes to cerebral function in the daily assessment of drug effects. The entropy or rather delta spectral entropy index seems to behave fairly monotonously until burst suppression. The BIS index has non-linear components even at lower anaesthetic concentrations, which probably cause the oscillations seen in some versions of it.
However, the indices have no or only very limited scientific value. The basic reason for this is that they include so many oversimplifications, violations of mathematical rules, non-linear components and misunderstandings. Worst of all, they do not provide the means to analyse what really happens in the EEG, or the individual spectral or time domain parameters the index is calculated from. Most publications do not even try to analyse these, which is a violation of scientific common sense! It is only when these boxes provide us with the original EEG and the individual spectra, burst suppression ratios etc., that are used for the calculations, that we can use them for research of permanent value. Who will care less in 10 yr that a BIS index model in 2001 behaves differently during halothane and isoflurane anaesthesia?
We predict that many of these discussions about bispectra or entropies will look very amusing for their lack of understanding of the EEG and brain function, and for their belief in some mysterious truth about brain function, in the not too distant future. Until then, use these indices for clinical work but do not think that they are tools of scientific value.
V. Jäntti
S. Alahuhta
Oulu, Finland
EditorWe would like to thank Dr Jäntti and Prof Alahuhta for their detailed response to our editorial. They have highlighted the gulf that exists between pure science and clinical practice with regards to cerebral function monitoring. The tone of their commentary seems critical or even dismissive of our editorial; yet the facts they present do not contradict those in the editorial. With respect to the lack of effect of nitrous oxide on the spectral entropy, they state in their opening paragraph: this is not really a surprise to any neurophysiologist who has experience of the neurophysiology of unconscious patients. In other words, why bother publishing this? Yet they answer this question later by stating that the new indices (BIS, spectral entropy, etc.) have been useful in taking the anaesthetists interest from spinal cord reflexes (i.e. MAC) to cerebral function in the daily assessment of drug effects. Unfortunately, the neurophysiology of consciousness, sleep, and anaesthesia, and the neuropathophysiology of epilepsy, brain trauma and other diseases of the central nervous system, are not emphasized very strongly in current anaesthesia training. As a profession we have largely ignored the monitoring of drug effects on cerebral function in the clinical setting because of its perceived impracticality. Modern cerebral function monitors have changed that. They take the EEG out of the neurophysiologists laboratory and into the operating theatre. These monitors will become a routine part of anaesthesia monitoring and training, only if clinicians find them useful. A major component of this process is coming to understand the limitations of the available monitorsand that was the focus of our editorial. Once in widespread clinical use, we can expect a further evolution of these monitors. Cerebral function monitoring will evolve, in the same way that carbon dioxide monitoring evolved from a simple number to guide the adequacy of ventilation, to having a continuous waveform which provides more subtle information about the function of the breathing circuit and the lungs. We do not have to wait until a universal feature of the EEG that always reflects unresponsiveness is discovered. To the clinician it does not matter that the Datex-Ohmeda cerebral function monitor is poorly named, and is more a measure of spectral uncertainty than entropy. (The marketing team that successfully sold a spectral uncertainty monitor would become legend in their profession!) It does matter to anaesthetists that patients having general anaesthetics are unconscious, and many anaesthetists will have used cerebral function monitors to try to ensure unconsciousness. Many of these same clinicians will be unaware how grossly different the cortical electrical behaviour is during unconsciousness depending on the agents used. Spectral entropy reasonably tracks the hypnotic effect of volatile anaesthetic agents and GABA-ergic i.v. induction agents. The likelihood is that it will, like the BIS monitor, be shown to have some utility. A widespread decrease in cortical activityas measured by spectral entropy or BISseems to be sufficient but not necessary to cause unconsciousness. From a research perspective, it would add scientific validity to know the proprietary algorithms used in cerebral function monitors, but a strong clinical interest in the effects of anaesthesia on the EEG is probably more vital to progress in this line of research. Pain has been popularized as the fifth vital sign. On our anaesthetic monitors we routinely have waveforms for the ECG, the pulse oximeter, the capnograph, and the arterial blood pressure. Perhaps the frontal EEG should be popularized as the fifth vital waveform.
J. Barnard
J. W. Sleigh
Hamilton, New Zealand
References
1 Sleigh JW, Barnard JPM. Editorial I. Entropy is blind to nitrous oxide. Can we see why? Br J Anaesth 2004; 92: 15961
2 Kissin I. General anesthetic action: an obsolete notion? Anesth Analg 1993; 76: 21518[ISI][Medline]
3 Amzica F, Steriade M. Electrophysiological correlates of sleep delta waves. Electroencephalogr Clin Neurophysiol 1998; 107: 6983[CrossRef][ISI][Medline]
4 Miller A, Sleigh JW, Barnard J, Steyn-Ross DA. Does bispectral analysis of the electroencephalogram add anything but complexity? Br J Anaesth 2004; 92: 813
5 Huotari AM, Koskinen M, Suominen K, et al. Evoked EEG patterns during burst suppression with propofol. Br J Anaesth 2004; 92: 1824