Bispectral Electroencephalographic Monitoring

In the 1990s, Aspect Medical Systems, a medical device company in Natick, Massachusetts, undertook an integrated research effort to develop the EEG as a measure of anesthetic depth. ^* The Aspect EEG monitor quantitates the anesthetic effects on the brain, specifically, the hypnotic component of anesthesia. The device presents

1251

1. The simultaneous use of multiple EEG signal-processing approaches captured incremental information that was not captured with traditional approaches based on a single signal-processing approach, such as fast Fourier transformation.
2. Multiple clinically relevant measures (movement, hemodynamics, drug concentrations, consciousness, recall) in patients and volunteers were gathered with concurrent EEG data.
3. Advanced multivariate statistical data analysis was used to correlate the components of the multiple EEG signal-processing approaches with the clinical data to create the univariate BIS parameter.
4. Prospective clinical evaluation of the BIS index was performed at multiple institutions under varying anesthetic and surgical conditions.
5. The BIS index was recognized as measuring the hypnotic components of the anesthetic and was relatively insensitive to the analgesic (e.g., opioid) components of an anesthetic.
6. Prospective clinical trials demonstrated that BIS monitoring could improve the outcome of an anesthetic regimen.
7. Simple hardware and sensors were developed and are commercially available to facilitate high-quality signal capture despite the noisy electrical environment of the operating room.

Bispectral Electroencephalographic Signal Processing

Bispectral EEG signal processing has been reviewed by Rampil^[151] and by Sigl and Chamoun.^[152] To date, most EEG signal processing has involved a form of spectral analysis, which examines the EEG signal during a small slice of time (epoch) as a function of frequency. The frequency analysis decomposes the EEG signal into a series of sine waves by Fourier analysis. Each sine wave is described with an amplitude, frequency, and phase angle. The sine wave amplitude is half the peak-to-peak voltage, the frequency is the number of cycles per second, and the phase angle describes the time offset of the sine wave relative to the start of the epoch. The output of the Fourier analysis is a histogram showing the frequency, amplitudes, and phase of the sine waves that taken together, create the EEG signal. Fourier analysis assumes that the signal pattern is constant during the epoch. Thus, it poorly handles signals that abruptly change during an epoch, such as burst suppression. Traditional spectral analysis ignores the phase data and assumes that the frequency constituents are uncorrelated (linear).

In contrast, BIS analysis assumes that frequencies may be correlated and uses the phase information to look for evidence of phase coupling among frequency bands, termed "bicoherence." The meaning of EEG phase relationships is not yet clear; however, the general notion is that an awake brain has multiple signal generators, working independently, and hence displays little synchronization. As the brain falls asleep, fewer independent signal generators are active, and thus the resulting EEG is more likely to reflect coupling (bicoherence) between signal generators. In addition, BIS analysis has several useful properties that promote noise suppression beyond what is possible with conventional spectral analysis.^[151] Thus, BIS analysis provides additional information about the EEG signal than is captured by traditional spectral analysis approaches that consider only frequency and amplitude.

Developing the Bispectral (BIS) Index

The BIS index is a complex, proprietary EEG parameter that has been under development since 1987 by Aspect Medical Systems. In 1996, the U.S. Food and Drug Administration approved the commercially available version as a monitor of anesthetic effect on the brain. Figure 31-19 displays the conceptual process whereby Aspect took EEG and clinical data from approximately 1500 subjects and 5000 hours of EEG signal gathered under a broad range of anesthetic regimens. The EEG was processed by first removing high- and low-frequency artifacts, electrocardiographic signals, pacemaker spikes, eye blinks, wandering baseline, and alternating current interference. The EEG data were then analyzed by three different approaches: Fourier, BIS, and time domain analysis ( Fig. 31-20 ). Time domain analysis was used to characterize burst suppression and isoelectricity. The EEG was smoothed across the power spectrum and bispectrum by using a moving average. The analysis then examined the degree of beta or high frequency (14 to 30 Hz), the amount of low-frequency synchronization, and the presence of fully or nearly suppressed periods (i.e., isoelectric signal).

Figure 31-19 Key steps used during development of the bispectral algorithm. Statistical modeling was used to identify the best features of the electroencephalogram for recognition of clinical end points. The circular path notes that the process was iterative. (Redrawn from Kelley SD: Monitoring Level of Consciousness during Anesthesia and Sedation. Natick, MA, Aspect Medical Systems, 2003.)

1252

Figure 31-20 Schematic diagram of signal processing paths integral to generating a single bispectral (BIS) index value. After digitization and artifact rejection, original electroencephalographic (EEG) epochs undergo three primary paths of analysis—power spectral analysis, BIS analysis, and time-based analysis for suppression/near suppression—to look for key EEG features. The BIS algorithm, based on statistical modeling, combines the contributions of each of the identified features to generate the scaled BIS index. (Redrawn from Kelley SD: Monitoring Level of Consciousness during Anesthesia and Sedation. Natick, MA, Aspect Medical Systems, 2003.)

Clinical Development and Validation of the Bispectral Index

The initial studies with early versions of the BIS algorithm in the early 1990s attempted to correlate the BIS index to predict movement in response to skin incision with the use of isoflurane/oxygen,^[152] propofol/nitrous oxide^[153] or propofol/alfentanil anesthesia.^[154] In the first major clinical evaluation of the BIS technology, a large, multicenter study by Sebel and colleagues^[155] randomized 300 patients into two groups. Both groups received an anesthetic regimen designed so that approximately half of all patients would move at skin incision. In the control group, the BIS index was passively recorded and not used for titration. In the BIS-titrated group, drug concentrations were increased to produce a BIS value less than 60. In the control group, the BIS index was a mean of 66 ± 19(standard deviation [SD]) with a 43% movement rate. In the BIS-adjusted group, the BIS index was lower, a mean of 51 ± 19 (SD) with a significantly lower movement rate of 13%. In this multicenter trial, some centers used primarily opioids with nitrous oxide, whereas others used propofol or isoflurane as the primary anesthetic. The study also found that reliability of the BIS index was strongly influenced by anesthetic technique. When hypnotic drugs such as propofol or isoflurane are used as the primary anesthetic, changes in the BIS index correlate with the probability of a movement response to skin incision. When opioid analgesics are used at higher doses, the correlation to patient movement became much less significant. This information, when linked to the understanding that purposeful movement to skin incision reflects the spinal action of anesthetics instead of their cortical effects, provided fundamental understanding of how development of the BIS algorithm had to be focused on the hypnotic components of anesthesia, specifically, the clinical measures of consciousness and memory. A subsequent series of studies described in the following paragraphs provided the data and clinical understanding to create the current algorithm to calculate the BIS index.

1253

Glass and coworkers^[156] designed a study to examine the relationship among BIS index, measured drug concentrations, and increasing levels of sedation when the anesthetic drugs propofol, midazolam, isoflurane, and alfentanil were given to volunteers in a controlled manner. Data from this study were used to further optimize the BIS algorithm. Seventy-two volunteers received a single drug at four or five defined, increasing target plasma concentrations that ultimately created an unconscious subject. The BIS index, arterial drug or end-tidal concentrations, and subjective sedation and memory scores were measured. The BIS index was significantly correlated to measured drug concentrations, as well as clinical measures of sedation. None of the alfentanil patients lost consciousness and had minimal change in the BIS index, a finding confirming that the BIS index is not sensitive to low concentrations of opioid analgesics. Fifty percent and 95% of the volunteers were unconscious at BIS values of 67 and 50, respectively.

Liu and colleagues^[157] demonstrated in 26 surgical subjects that the BIS index accurately tracked the degree of clinical sedation with intermediate (4 mg) to large (20 mg) doses of midazolam given during regional anesthesia. Response to a loud voice (observer assessment of alertness/sedation [OAA/S] score of 3) corresponded to a BIS index of 87 ± 6 (SD) and a 40% probability of recall. A deeper end point, lack of response to mild prodding (OAA/S score of 2) corresponded to a BIS value of 81 ± 8 (SD) and resulted in complete lack of recall. When midazolam created an unresponsive subject, the mean BIS index was 69.2 ± 13.9 (SD). Similar results were found by these investigators for propofol when given under regional anesthesia.^[158] Katoh and coworkers^[159] had similar results with low, sedating doses of sevoflurane and also with higher, anesthetic concentrations in 69 surgical patients. Both the BIS index and end-tidal sevoflurane concentrations were correlated to clinical sedation scores

Figure 31-21 A, Clinical correlations of the bispectral (BIS) index. Maintaining the BIS index from 45 to 60 during general anesthesia appears to ensure unconsciousness with a hypnotic/opioid anesthetic technique while providing for rapid emergence (Adapted from Johansen JW, Sebel PS: Development and clinical application of electroencephalographic bispectrum monitoring. Anesthesiology 93:1336, 2000.) B, Electroencephalographic (EEG) changes observed with increasing depth of anesthesia. (Redrawn From Kelley SD: Monitoring Level of Consciousness during Anesthesia and Sedation. Natick, MA, Aspect Medical Systems, 2003.)

Kearse and associates^[161] undertook a more detailed examination of memory and response to command during propofol/nitrous oxide sedation with BIS monitoring. These investigators found a strong association between the BIS index and response to command, whether the command consisted simply of a uniform voice asking a volunteer to move a hand or foot or incorporated graded and then varied stimuli. The relationship between the BIS index and responsiveness scores remained consistent over time and with increases or decreases in propofol concentrations. No subject was responsive when the BIS index was less than 57.

Johansen and Sebel^[162] have reviewed the historical clinical development and current applications of BIS monitoring in clinical anesthesia.

Interpreting the Bispectral Index

Figure 31-21 displays the numerical value of the BIS index, from 0 to 100 relative to clinical end points, and the underlying EEG signal. BIS values of 0 represent an isoelectric EEG, whereas values of 100 represent an awake CNS. After administration of a hypnotic drug, the BIS index decreases from an awake value of 100 as the patient's

1254

A limited number of studies examined the relationship of the BIS index to direct measures of brain function. Alkire^[163] studied the correlation between the cerebral metabolic rate, as measured by positron emission tomography, and the BIS index during propofol and isoflurane anesthesia. Volunteers served as their own controls. After baseline measurements, volunteers received increasing amounts of propofol or isoflurane until they were unconscious (BIS values of 30 to 50). This study demonstrated a linear relationship between decreasing BIS index and a reduction in cerebral metabolic rate, thus indicating a physiologic link of the quantitated EEG and cerebral metabolism under the influence of hypnotic drugs.

Ludbrook and colleagues^[164] examined the relationship between propofol brain concentrations, as estimated by simultaneous arterial/jugular bulb venous blood sampling during propofol induction of anesthesia, and cerebral blood flow and BIS measurements. Although significant intersubject variability was seen, calculated propofol brain concentrations and BIS values were closely correlated.

The relationship between noxious stimuli, opioid response, and the BIS index has been determined. Early studies had demonstrated that low concentrations of

Figure 31-22 The reduction in verbal responsiveness to command (red line) and memory (black line) with decreased bispectral (BIS) index values. BIS values less than 60 are associated with a very low probability of awareness. (Redrawn from Kelley SD: Monitoring Level of Consciousness during Anesthesia and Sedation. Natick, MA, Aspect Medical Systems, 2003.)

Guignard and associates^[166] extended this understanding by examining the response of the BIS index to the interaction of noxious stimuli (laryngoscopy/tracheal intubation) during propofol administration with and without remifentanil in surgical patients. Computer-controlled infusion of propofol was used to achieve a constant target effect-site concentration of 4 µg/mL. Hemodynamics, movement response, and the BIS index were measured during tracheal intubation. With only propofol at a target effect-site concentration of 4 µg/mL, the BIS index decreased to a mean of approximately 45. Intubation resulted in a significant increase in BIS values to a mean of 70. When computer-controlled infusion of remifentanil was added to achieve target effect-site concentrations of 2 to 16 ng/mL, the interaction of remifentanil on BIS values was obvious. Before intubation, these remifentanil concentrations did not change the BIS values during propofol hypnosis. Progressively increasing concentrations of remifentanil blunted the increase in BIS index after intubation. Remifentanil target effect-site concentrations of 8 and 16 ng/mL resulted in a minimal increase in the BIS index, hemodynamic stability, and no purposeful movement. This study demonstrated that significant noxious stimuli given with only a hypnotic present (e.g., propofol, inhaled anesthetics) will result in significant hemodynamic response with an increase in BIS index values. The addition of adequate opioid analgesia results in hemodynamic control and less or no change in the BIS value.

Ropcke and colleagues^[167] demonstrated similar EEG activation and increase in BIS values when desflurane only (no opioids) was given to surgical patients during intra-abdominal surgery. They demonstrated that in unstimulated patients before the initiation of surgery, a desflurane end-tidal concentration of 2.2% was needed to achieve a mean BIS value of 50. The desflurane end-tidal concentration needed to achieve a BIS value of 50 increased to 6.8% during intra-abdominal noxious surgical stimulation. The authors did not examine how opioid analgesia would change the response observed with desflurane. These two studies indicate that with inadequate analgesia (opioid) during significant noxious surgical stimuli, one can expect the BIS index to increase. Low concentrations of opioid in the presence of a hypnotic drug with no noxious stimuli present will not change the BIS value. Higher doses/concentrations of opioids can result in EEG slowing and a decrease in the BIS index, either alone or with hypnotic anesthetic drugs present.

Two studies suggest that nitrous oxide given in concentrations of up to 70% have minimal effect on the BIS index. Rampil and coworkers^[168] demonstrated in volunteers that although nitrous oxide concentrations up to 50% did alter the spectral content of the EEG, there were minimal changes in the BIS index. The subjects did not exhibit

1255

Katoh and coworkers^[170] examined the effects of increasing age on the relationship of sevoflurane sedation relative to the BIS index. They studied 96 surgical patients 18 to 85 years of age. The patients were sedated at sevoflurane end-tidal concentrations ranging from 0.4% to 0.9%, and verbal responsiveness to loud name calling (OAA/S score of 3) or mild prodding (OAA/S score of 2) was determined. The BIS index was linearly related to end-tidal sevoflurane concentrations. Sevoflurane end-tidal concentrations of 0.4% to 0.5% produced BIS values of 80 to 95, whereas sevoflurane end-tidal concentrations of 0.8% to 0.9% produced BIS values of 60 to 80. The authors found that MAC_awake was age related, with the elderly having a significantly lower value than younger patients did. The relationship of the probability of no response to verbal command versus the BIS index was not age related. These data suggest that the BIS index can be used to adjust for the aged-related change in sevoflurane potency. Similar studies will be needed to examine the age independence of BIS values at deeper clinical end points of hypnotic drug effect.

Several other factors that will be encountered in clinical anesthesia care can interact with the BIS index. The moderate changes in body temperature that occur during cardiopulmonary bypass can change the EEG signal and hence the BIS index. Mathew and colleagues^[171] showed that in patients undergoing cardiopulmonary bypass with constant effect-site concentrations of fentanyl and midazolam, hypothermia decreased the BIS index by 1.12 units per degree Celsius decline in temperature. Body temperature ranged from approximately 32°C to 37°C in this clinical study. In healthy volunteers, mild core hyperthermia (increase of 2°C) did not alter the BIS index.^[172]

Infusion of esmolol can also alter the BIS index. Johansen^[173] has shown in surgical patients that during a stable propofol/alfentanil anesthetic (propofol target effect-site concentration of 5.5 µg/mL, alfentanil effect-site concentration of 50 or 150 ng/mL), a 30-minute infusion of esmolol (1-mg/kg bolus, 250 µg/kg/min) decreased the BIS index from 37 ± 6 (SD) to 22 ± 6 (SD). Discontinuation of esmolol reversed these changes. Menigaux and associates^[174] demonstrated that when esmolol was given during induction with propofol only, the BIS index was not different from that of the control group (propofol, no esmolol) before intubation. BIS values were approximately 45 before intubation. With intubation, the BIS value increased in the control group to approximately 60, whereas the BIS value in the esmolol group was unchanged. These two studies indicate that there is a complex interaction of β-adrenoreceptor blockade during anesthesia with relevant noxious stimuli and measurement of the hypnotic component with the BIS index. β-Adrenoreceptor blockade both alters the EEG arousal response to noxious stimuli when inadequate analgesia is present and can decrease the BIS index during adequate clinical anesthesia.

Epidural anesthesia can also decrease the amount of hypnotic anesthetic needed for sedation. Hodgson and Liu^[175] have shown that the end-tidal concentration of sevoflurane needed to achieve a BIS index of 50 was 0.59% in patients receiving a lidocaine epidural for surgical procedures whereas a sevoflurane end-tidal concentration of 0.92% was needed to achieve a BIS index of 50 in patients receiving general anesthesia or general anesthesia with an intravenous infusion of lidocaine to match the plasma concentrations achieved during the epidural anesthetic. This study demonstrates that the deafferentation induced by regional anesthesia results in supraspinal effects that change hypnotic anesthetic requirements.

The dissociative anesthetic ketamine can create excitatory effects on the EEG. Ketamine doses that create unresponsiveness (0.25 to 0.5 mg/kg) did not change the BIS index.^[176] When ketamine is used with other hypnotics that decrease the BIS index, such as propofol, an additive interaction occurs on the pharmacodynamic end points that is not reflected in the BIS index.^[177] Thus, the BIS index may not have clinical utility when used with ketamine alone or with other anesthetic drugs.

Bispectral Index and Clinical Utility/Outcome

Several studies have examined the clinical and economic outcome benefits of BIS monitoring in routine surgery. Gan and colleagues^[178] studied 302 subjects receiving a propofol/alfentanil/nitrous oxide anesthetic regimen in four different institutions. Half the subjects were randomized to a "standard practice" whereby the propofol and alfentanil were titrated to provide a stable anesthetic with the fastest possible recovery. The BIS index was recorded in this group but was not displayed to the practitioner. The remaining subjects received "standard practice plus BIS monitoring." In this group, propofol infusions were titrated to achieve a target BIS value between 45 and 60 during surgery and 60 to 75 during the final 15 minutes of the anesthetic regimen. Patients in the BIS group required significantly lower normalized propofol infusion rates (134 versus 116 µg/kg/min), were extubated significantly sooner (11.2 versus 7.3 minutes), had a higher percentage of patients oriented on arrival in the recovery room (43% versus 23%), and had more rapid discharge from the recovery room. This study demonstrated that titrating propofol with BIS monitoring during balanced anesthesia decreased propofol use and significantly improved recovery with no difference in intraoperative conditions.

Song and colleagues^[179] presented similar findings with inhaled anesthetics. Sixty surgical patients were randomized to receive desflurane or sevoflurane with 65% nitrous oxide and low doses of fentanyl. Half the patients were randomized to standard practice in which the inhaled

1256

Three clinical trials have been undertaken to examine the value of BIS monitoring relative to intraoperative awareness. The B-Aware trial is an Australian prospective, randomized, double-blind multicenter trial involving patients at high risk of awareness (cesarean section, high-risk cardiac surgery, trauma surgery, or rigid bronchoscopy).^[180] Patients were assessed for awareness at 2 to 6 hours, 24 to 36 hours, and 30 days. An independent, blinded committee assessed each report of awareness. BIS-guided anesthesia reduced the risk of awareness by 82% (n = 1227, 2 patients) in comparison to the control group (n = 1238, 11 patients).

The AIM study was a prospective multicentered study of 19,576 patients undergoing general anesthesia at seven geographically dispersed institutions in the United States.^[181] All patients were interviewed postoperatively and 1 week after surgery with a structured interview technique. A total of 25 cases (0.13%) of awareness were identified, and they occurred at a fairly consistent rate in the seven different institutions. High-risk procedures such as cardiac surgery had a higher incidence (0.95%) than lower-risk orthopedic procedures did (0.1%). These findings are similar to previous reports of the incidence of awareness from Australia and Sweden. Awareness is a ubiquitous phenomenon with an incidence between 1 and 2 cases per 1000 anesthetics.

The SAFE-II study is Swedish awareness follow-up evaluation of 5100 patients who underwent BIS monitoring during surgery.^[182] The use of BIS monitoring was associated with a significantly reduced incidence of awareness (78% reduction) when compared with historical controls from the same hospitals and investigators.^[49] These three studies demonstrate that awareness occurs in current clinical practice and that routine BIS monitoring can significantly reduce this risk to patients.

The relationship of anesthetic management to patient morbidity and mortality has generally focused on the immediate perioperative time period, that is, the hours to days surrounding surgery and anesthesia. Information on the effects of anesthetic management on long-term postoperative mortality rates is limited. Weldon and coauthors^[183] reported that deep maintenance anesthetic levels are associated with higher 1-year postoperative death rates in patients 40 years and older undergoing major, noncardiac surgery. These investigators monitored the BIS index in 907 patients undergoing general anesthesia for major elective surgery (excluding cardiac and intracranial procedures) lasting at least 2 hours. The anesthetic technique was not controlled, and the anesthesiologist was blind to the BIS values. The percentage of time that the BIS index was less than 40, between 40 and 60, or greater than 60 was measured during the maintenance phase of the anesthetic. The authors examined the incidence of death at 1 year in the group. Mortality was significantly higher if the BIS index was less than 40 in patients older than 40 years. Increasing age and lower BIS values were both independently associated with higher mortality rates. Lennmarken and colleagues^[184] confirmed these finding with a retrospective analysis of a previously published study examining the incidence of awareness using BIS monitoring. The authors found in the cohort of 5077 consecutive subjects monitored with the BIS index that a low BIS value increased the risk of 1-year mortality. Monk and coworkers examined 1-year postoperative mortality in a large Medicare hospital database linking hospitals that used BIS monitoring.^[185] This retrospective analysis also suggested that hospitals that routinely use BIS monitoring have decreased 1-year postoperative mortality.

Clinical Use of the Bispectral Index

The previously described development of the BIS index and its application to patient care monitoring represent the first scientifically validated and commercially supported monitor of anesthetic drug effect on the CNS. The BIS index primarily measures the effects of hypnotics on the EEG. The synergistic interaction of opioids with volatile or intravenous hypnotics on clinical end points (hemodynamics, movement responses) is more profound than reflected by the EEG. As a result, the BIS index is most accurate when used with anesthetics consisting of a low or moderate dose of opioid analgesic and a hypnotic drug (volatile inhaled anesthetic, intravenous anesthetic) titrated to the BIS response. Low opioid doses enable the BIS index to accurately reflect the pharmacodynamics of the hypnotic drugs on the CNS. High-dose opioids yield significant synergy with the hypnotic drugs, as demonstrated by Glass and colleagues,^{[114] [115]} such that minimal amounts of hypnotic drug are required to obtain adequate anesthesia. Because the reduced hypnotic dose in high-dose opioid techniques results in less profound hypnotic EEG drug effect, the BIS response is less reliable with a high-dose opioid technique.

Clinical use of BIS monitoring involves separating the hypnotic and analgesic components of an anesthetic regimen. The concept entails titration of the hypnotic drug (e.g., isoflurane, desflurane, sevoflurane, propofol, midazolam) to lower the BIS value to 40 to 60 (see Fig. 31-21 ). This range appears to be the therapeutic window associated with a high probability of unconsciousness. A small to moderate amount of analgesic drug (opioid) is also given with the hypnotic drug. The anesthesiologist then evaluates the clinical and BIS responses to the surgical procedure over time. During times of intense surgical stimulation, if the BIS index increases and the patient exhibits movement and hemodynamic responses, the anesthesiologist should respond by increasing the hypnotic component to lower the BIS value to the 40 to 60 range. If the BIS value is in the 40 to 60 range and movement and hemodynamic responses continue, incremental opioid should be added to increase the analgesic component of the anesthetic until movement and hemodynamics are controlled. As the end of the anesthetic regimen approaches, the

1257

BIS monitoring provides an important new dimension to the ability to adjust the components of an anesthetic in a logical manner. Figure 31-23 presents a tabular summary of the algorithm that will allow maximal utility of the BIS index during anesthesia.