Why Monitor Depth of Anesthesia?

The problem of awareness during general anesthesia has become an increasingly prominent issue for both the anesthesiology community and the lay public. The Joint Commission issued a Sentinel Alert regarding awareness in 2004, and the American Society of Anesthesiologists (ASA) subsequently established a Task Force on Intraoperative Awareness that published a practice advisory regarding the use of brain function monitoring. This attention reflects the clinical significance of awareness because patients distressed by the recall of intraoperative events may develop severe psychological sequelae, including PTSD.

To address this problem, various devices have been developed to help assess the depth of anesthesia. It is worth noting that estimates of the incidence of awareness have ranged from 0.0068% in a study by Pollard et al to 1.0% in a study by Errando et al, making the identification of patients at risk for awareness and the reduction of this rare event a challenging task. There is a possibility that depth of anesthesia (DoA) monitors may also help anesthesia providers titrate medications, and help prevent oversedation, which may result in decreased morbidity and mortality. Regardless of the motivation for monitoring brain function, the anesthesia provider should consider the evidence regarding the clinical utility of the device with respect to the outcome of interest.

Currently Available Depth of Anesthesia Monitors

This chapter will discuss the following devices:

• ▪
  

  Bispectral index monitors (Aspect Medical Systems)

  
  
• ▪
  

  S/5 Entropy module (GE Healthcare)

  
  
• ▪
  

  Narcotrend monitor (MonitorTechnik)

  
  
• ▪
  

  Cerebral state monitor (Danmeter)

  
  
• ▪
  

  SEDLine/patient state analyzer (Hospira)

  
  
• ▪
  

  SNAP II monitor (Stryker)

  
  
• ▪
  

  AEP Monitor/2 (Danmeter)

We will focus on these monitors because they represent the main devices that are currently used and researched and were the devices mentioned in the ASA Practice Advisory for Intraoperative Awareness and Brain Function Monitoring. Bowdle, Bruhn, and Jameson and Sloan have written reviews outlining the basic principles behind EEG monitoring and discuss the devices as they existed in 2006.

The Ideal Monitor

The ideal DoA monitor would (1) detect whether the patient is too lightly anesthetized and is at risk of awareness; (2) detect whether the patient is unnecessarily deeply anesthetized, and thus at risk for prolonged recovery; (3) work similarly across different patients; and (4) work similarly regardless of anesthetic modality and medications. Bruhn discusses the principles of validity and reliability of DoA monitors. The first two criteria relate to the concept of validity, which is the “accuracy” of the monitor in correctly determining the anesthetic depth. The latter two criteria relate to the concept of reliability, or “consistency” of the monitor in reporting the same result for a given depth of anesthesia.

The clinical utility of the device depends on how well it meets these criteria. By being accurate and consistent, a device should theoretically be able to demonstrate improvements in certain clinical end points, such as decreased awareness events or recovery times.

Assessment and Comparison of Awareness Monitors

This chapter will demonstrate the underlying computational approach of key monitors and examine the available evidence regarding the clinical utility of the devices. The intention is not to compare technologies with each other to determine the “best” equipment to use. Algorithms and technologies are continually updated, and there is much heterogeneity in studies in terms of patient populations and clinical end points. Generalizations made across various studies are thus often misleading, if not invalid. Sources purporting to provide product comparisons should be examined judiciously for both accuracy and bias. For example, an online “distributive evidence database” ( brainmonitor.doctorevidence.com ) offers to objectively synthesize and compare the outcome-oriented evidence for DoA monitors, but is funded by Aspect Medical Systems.

To discuss the merits of each DoA monitor, it is useful to have an easily communicated metric to describe the accuracy and reliability of the device. Sensitivity/specificity analyses or the comparison of likelihood ratios are two possible approaches. An alternate approach, adopted by the ASA Practice Advisory for Intraoperative Awareness and Brain Function Monitoring along with the majority of publications on DoA monitors, is to compare prediction probabilities.

Smith et al explains the concept of prediction probability (Pk) as applied to DoA monitors. Conceptually, Pk represents a measure of how well an index value (such as the BIS index) provided by the device can differentiate between different clinical states (e.g., response vs. no response to command, or levels of a graded sedation scale). If an index value were perfectly associated with the clinical state (i.e., a higher index value always represents a higher clinical level of arousal), the Pk value would be 1. If the index value were no better than chance in predicting a higher or lower clinical state, the Pk value would be 0.5.

The Pk is independent of scale, and does not make assumptions about the underlying distribution. It is thus commonly used for evaluating and comparing different monitors. The Pk essentially represents “criterion validity” (which is the agreement of the monitor with another instrument). However, the range of Pk values produced by various studies performed across different patients and anesthetic regimens are sometimes used as a measure of the reliability of the monitoring device.

Pk values are best used to compare devices in the same study. Comparisons between different studies may be confounded by differences in patient populations and anesthetic regimens. Pk values are also dependent on the clinical scale and assessment method used for comparison. For example, the Pk value for the prediction of loss of response to verbal command may be different from the Pk value for the prediction of loss of response to painful stimulus. The order of events is also important due to potential physiological hysteresis (i.e., loss of response during induction is not necessarily the reverse process of regaining response during emergence).

For purposes of evaluating a device’s ability to detect and prevent impending awareness, it may be best to use Pk values that assess some measure of response to stimulus upon emergence. It is also very important to note that Pk depends on the coarseness of the scale. A Pk value calculated using an “awake” versus “not awake” dichotomization would tend to have a higher value compared with a Pk calculated using multiple ordinal categories such as a 1 to 10 scale. This underscores the need to compare Pk values very cautiously.

In addition, Pks can be misleading if they are used as an indicator of the “believability” of a particular DoA monitor in predicting awareness in a given patient. Pks represent how well the monitor can predict a particular state across multiple patients; it does not offer direct information about the significance of a particular index value in a given patient. Index values must always be used as one piece of information that helps modify the individual pretest probability of awareness as determined by the clinician.

Basic Principles of EEG Analysis

In 1937, Gibbs et al published a study on the relationship between electroencephalography (EEG) patterns and drug-induced changes in neuronal activity. The authors presciently recognized the potential application of EEG technology as a monitoring device, commenting:

“a practical application … might be the use of the electro-encephalogram as a measure of the depth of anesthesia during surgical operations. The anesthetist and surgeon could have before them on tape or screen a continuous record of the electrical activity of both heart and brain.”

The regular adoption of EEG as a monitoring technology can be facilitated by the development of computerized EEG processing systems to analyze complex waveform patterns. Most current awareness monitors similarly collect EEG information with frontal electrodes, but they differ significantly in how these EEG signals are processed to yield an index value representing the depth of anesthesia. Processing techniques unique to certain devices will be discussed in the device-specific section of this chapter. Rampil provides an excellent technical primer on the physiology of EEG physiology and monitoring, while Sigl and Chamoun provide a very clear explanation of the basic mathematics behind EEG signal analysis.

EEG Activity and Anesthetic Depth

Synaptic activity of cells in the cortex results in voltage changes that can be detected by electrodes placed on the scalp. There are various waveform patterns within certain frequency ranges, which correspond to various neurophysiological processes. These patterns are grouped into frequency bands called, in order of increasing frequency, δ-, -, α-, β-, and γ-waves. Each band exhibits certain changes under the influence of anesthetic agents. A review by Jameson et al provides a detailed description of these neurophysiological and EEG effects.

Delta waves (<3 Hz) have the slowest frequencies of the various frequency bands and are seen in deep sleep; they may result from extreme thalamic depression. Θ-waves (4 to 7 Hz) may similarly arise from the inhibition of thalamic pacemaker cells. Theta waves are readily seen in young children and become less prominent with age, although they are normally seen during sleep at any age. Δ- and -waves are often referred to as “slow waves.” α-Waves (8 to 12 Hz) are prominently seen in awake patients and are prominent over the vertex in a relaxed or sedated state. They are thought to reflect the cyclical activity of thalamic pacemaker cells when anesthetics decrease the inhibition of the thalamus by the nucleus reticularis. The pyramidal area of alpha wave prominence tends to move to more frontal regions with increased depth of anesthesia in a process called “anteriorization” or “frontal predominance.”

Beta waves (12 to 24 Hz) are present in the prefrontal regions, and presumably reflect thalamocortical pathways with higher frequencies than α-waves due to desynchronization of the thalamus by sensory stimuli. Beta waves are normally present during alert states and can increase during initial central nervous system depression due to disinhibitory effects. This “β-activation” can be seen in Figure 12–1 . β activity is most prominent in the frontal regions and tends to move posteriorly with increased depth of anesthesia.

EEGs showing the progression of characteristic changes associated with increased anesthetic depth.

(Image from Tonner PH, Bein B. Classic electroencephalographic parameters: median frequency, spectral edge frequency etc. Best Pract Res Clin Anaesthesiol 2006;20:147-159.)

Gamma waves, also known as β ₂ -bands, are of a higher frequency range (25 to 50 Hz) and may play a role in sensory processing and perception. Recent evidence suggests that organized γ activity may be essential for consciousness and interrupted during anesthesia. The EEG of an alert individual mostly consists of higher frequency α- and β-waves. With the addition of anesthetic agents, the overall amplitude of the EEG waveform initially increases as the EEG is synchronized in the 8 to 10 Hz range, and subsequently decreases with increased anesthetic depth. Similarly, the predominant frequencies initially increase during the early stages of anesthesia, after which the EEG shifts to lower frequency – and δ-waves.

With sufficiently deep anesthetic states (often minimal alveolar concentration [MAC] values of 1.5 or higher ), the EEG demonstrates a bilateral pattern of slow and mixed waves. A pattern of high amplitude activity (bursts) upon a flat baseline (suppression) is called “burst suppression.” A burst suppression ratio, comparing the percent duration of suppression, is often calculated. Burst suppression can also be seen with brain-injured conditions, such as postischemic states. Further deepening of the anesthetic state results in no discernable voltage changes (isoelectricity) ( Figure 12–1 ).

EEG Processing Techniques

The EEG can be analyzed in either the time domain or the frequency domain. Many of the improvements in DoA monitoring have resulted from the ability to process the EEG in the frequency domain. In the time domain, voltage changes in the EEG are plotted against time. Burst suppression is identified in the time domain. There are other time-domain techniques, such as the calculation of zero crossing frequency (ZXF), which is the number of times the EEG crosses the baseline, or a periodic analysis, which calculates the time between two consecutive minima of the waveform. Although these metrics are useful parameters in limited circumstances, complex signals cannot be fully analyzed by solely using time domain-based methods.

The advancement of microprocessing over the past few decades enabled fast calculations of Fourier transformations, which enable analysis of the frequency domain, wherein the frequencies of component signals present in the EEG are compared with the degree to which these frequencies are present. The general premise of EEG-based DoA monitors is that changes in EEG patterns can be detected, through time and frequency domain calculations, to provide an estimation of the level of consciousness.

Fourier Transformation and the Power Spectrum

Fourier analysis (also called spectral analysis) is based on the concept that complex waveforms can be approximated as a sum of sinusoids of various frequencies, amplitudes (power), and phase relationships, as shown in Figure 12–2 . A discrete Fourier transformation (DFT) converts a sample of periodic brain waves from the “time domain” to the “frequency domain” and “phase domain.” A fast Fourier transform (FFT) is an algorithm that very efficiently performs DFTs. Applied to an EEG, an FFT takes as an input the raw EEG waveform measured over time, and outputs the frequencies and phase of the component waves of the raw waveform. (A loose analogy for the FFT would be a prism separating out white light into individual colors of various wavelengths.)

The five upper waveforms (with frequencies of 1, 2, 3, 4, and 5 Hz and amplitudes of 1, 1.5, 1, 2, and 1.5, respectively) combine to yield the waveform on the bottom . A Fourier transformation allows the combined waveform to be deconstructed into the five constituent sinusoids.

(Image from Sigl JC, Chamoun NG. An introduction to bispectral analysis for the electroencephalogram. J Clin Monit 1994;10:392-404.)

The information about the frequency domain is often presented as a graph of power (the magnitude of the frequency contribution) versus frequency. The resulting “power spectrum” thus provides a measure of the extent to which various frequencies (corresponding to α-, β-, γ-waves, etc.) are present in the EEG.

Median Frequency and Spectral Edge Frequency

Early frequency-based analysis included simple parameters such as the peak frequency, the median frequency, or the spectral edge frequency. The peak frequency is often the frequency with the greatest power in a given wave region. The median frequency is the frequency that divides the power in half, and the spectral edge frequency is calculated as the frequency below which the majority (often, 90% or 95%) of the power lies ( Figure 12–3 ).

A schematic of the power spectrum. The horizontal axis represents the frequency of the component wave, and the vertical axis represents the power, or the degree to which the frequency is represented. Here, 90% of the power exists below the spectral edge frequency.

(From Tonner PH, Bein B. Classic electroencephalographic parameters: median frequency, spectral edge frequency etc. Best Pract Res Clin Anaesthesiol 2006;20:147-159.)

Nonlinearity and Phase Coupling

The cerebral cortex can be thought of as a system in which sinusoidal inputs (from neuronal triggers such as thalamic pacemakers) are entered into a “system” which outputs a certain EEG signal. Examination of the power spectrum alone provides information about the power and frequency, but not the phase relationships of the inputs. In a “linear” system, the output represents simple addition of each input, and the phase (time offset of the sinusoid) of the input does not matter. The neuronal system is complex, however, and should be thought of as a non linear system. As consequence of non linearity, the phase of the output signal depends on the phases of the input signals. The phases are then considered to be “phase coupled.”

As the degree of phase-coupling of various EEG frequency components is thought to reflect complex neurophysiological changes associated with various levels of consciousness, DoA monitors can use this phase coupling information to help predict the depth of anesthesia.

BIS Technology (Aspect Medical Systems)

Bispectral (BIS) technology is based upon an empirically determined proprietary algorithm developed by Aspect Medical Systems; it was approved by the U.S. Food and Drug Administration in 1996. It converts the recordings from a frontal EEG to a single number, the BIS index, which ranges from 0 (isoelectric) to 100 (awake) and represents the level of consciousness by the patient. The BIS system is the most widely used DoA monitor in the United States. According to information provided on the manufacturer’s website as of June, 2009, the BIS platform is installed in 75% of U.S. operating rooms, and 19% of procedures requiring general anesthesia or deep sedation use the BIS monitor. As such, most of the research pertaining to DoA monitors has been done on the BIS. In 2006, a PubMed search for the phrase “bispectral index” yielded 852 citations ; at the time of writing, the same search parameters revealed 1543 citations.

Following the earlier A-2000 and BIS VIEW monitoring platforms, the latest implementation of the BIS technology is the BIS VISTA system, which provides an option for bilateral monitoring. The BIS technology is available as a stand-alone system or can be incorporated into “BIS modules” for use with monitoring systems of external manufacturers ( Figure 12–4 ).

The BIS VISTA monitor.

(Image from Aspect Medical Systems. (website): http://www.aspectmedical.com/assets/Images/photos/right_column/bis-vista.jpg .) Accessed November 20, 2008.

Underlying Principle: Bispectral Analysis

Bispectral analysis (also called two-dimensional spectral analysis) refers to the statistical process of quantifying quadratic nonlinearities of the system. The bispectrum, or bispectral value, represents the degree of phase coupling of two fundamental frequencies and a modulation frequency represented by the sum of the two frequencies ( Figure 12–5 ). A related parameter is the bicoherence, which is obtained by normalizing the bispectrum to a 0% to 100% range, to reflect the effect of relative amplitudes of the component signals. Part of the BIS algorithm calculates the logarithm of the ratio of the sum of bispectrum peaks in the range from 0.5 to 47 Hz to the sum of the peaks from 40 to 47 Hz. This value, known as the “SyncFastSlow,” is one of the many parameters used to calculate the BIS index. A full explanation of the calculation of the bispectral value is beyond the scope of this chapter, but can be found in the primer by Sigl and Chamoun.

The linear combination of two frequencies (f1 = 10 Hz, f2 = 3 Hz) results in a new waveform that contains the sum of the original waves. Bispectral analysis of these waveform results in a peak bispectral value, B (f1, f2), which reflects the phase-locked constituent frequencies.

(Image from Rampil IJ. A primer for EEG signal processing in anesthesia. Anesthesiology 1998;89:980-1002.)

The overall premise behind bispectral analysis is that changes in the clinical state are represented by changes in phase coupling. The neurophysiological significance of a large bispectral value being generated by two particular frequencies is not clear, but may represent the degree of shared EEG pacemaker elements.

What It Outputs

To be useful for the clinician, the information resulting from the bispectral analysis is converted into an easily interpretable and clinically meaningful value.

For the BIS monitor, this value is the “BIS index,” which is calculated using a proprietary algorithm. The formula was empirically developed by correlating clinically assessed depths of anesthesia with features of the EEG including information from the frequency domain (i.e., power spectrum, bispectrum, and bicoherence) and the time domain (burst-suppression analysis).

The BIS VISTA display shows the BIS index value and trend and the raw EEG waveform and EMG indicator ( Figure 12–6 ).

The BIS VISTA display shows the BIS index value and trend and the raw EEG waveform and EMG indicator.

(Image from Aspect Medical Systems. BIS VISTA Monitoring System Operating Manual, website: http://www.aspectmedical.com/ Files/file/Doc/BISVISTAmanual.pdf .) Accessed July 21, 2010.

The BIS bilateral system provides the option of a density spectral array (DSA) to be displayed ( Figure 12–7 ). This displays the trend of the power spectrum from each hemisphere simultaneously.

The density spectral array (DSA) can be viewed when the BIS bilateral system is used. The frequency scale with a range of 0-30 Hz is on the horizontal axis , with the vertical axis representing the time progression. The area to the midline with respect to the white spectral edge line contains 95% of the power.

(Image from Aspect Medical Systems. BIS VISTA Monitoring System Bilateral Monitoring Addendum, website: http://asrs.aspectms.com/Manuals/Vista/070-1024%20VISTA%20Op%20Man%20Bilat%20Add.pdf .) Accessed July 21, 2011.