Development and Clinical Application of Electroencephalographic Bispectrum Monitoring

UNTIL recently, anesthesiologists lacked the ability to monitor the effects of anesthetics on the brain in terms of "depth" or "adequacy" of anesthesia. Typically, surrogate measures of autonomic activity, such as changes in blood pressure and heart rate, have been used to assess the adequacy or inadequacy of anesthesia. Because it is believed that general anesthetics block consciousness by depressing the central nervous system, and electrical activity of the cerebral cortex can be measured using the electroencephalogram (EEG), it is expected that some component of the EEG should relate to adequacy of anesthesia. Such a relation was first suggested in 1937. 1With the advent of the microcomputer technology, it became possible to reduce the amount of data obtained from an EEG to various processed derivatives. 2Derivatives such as the power spectral edge, median frequency, and zero-crossing frequency, among others, have been described as potential measures of anesthetic effect on the central nervous system. 3–6In that these measures were found to depend on spec ific drug combinations and were not monotonically related to drug effect or clinical response, no gold standard for measuring the entire spectrum of anesthetic effect has been widely accepted.The first and only technology approved by the U.S. Food and Drug Administration (October 1996) for marketing as an EEG-based monitor of anesthetic effect is the bispectral analysis derivative known as the Bispectral Index Scale (BIS, Aspect Medical Systems, Natick, MA). The purpose of this review is to describe the clinical development of this technology and to assess our current understanding of its utility in clinical practice.Bispectral analysis is a statistical technique that allows study of phenomena with nonlinear character, such as surf beats and wave breaking. 7Bispectral analysis provides a description to a continuous pseudo–randomly varying signal (e.g. , EEG) that is an alternative to other conventional power spectral analysis techniques derived from fast Fourier transformation. The mathematics of bispectral analysis have been described elsewhere. 7–11The first studies of EEG bispectral analysis were published in 1971. 12Bispectral analysis is computationally intensive, and it was not until fast microprocessors were developed that online bispectral analysis of the EEG in the operating room became possible.Conventional analysis of the EEG using fast Fourier transformation produces information regarding the power, frequency, and the phase of the EEG signal. Typical displays, such as the compressed spectral array, graph power and frequency information and discard the phase information. 2Bispectral analysis represents a different description of the EEG in that interfrequency phase relations are measured, i.e. , the bispectrum quantifies relations among the underlying sinusoidal components of the EEG. 2Additional details regarding the computation of bispectral data can be found in Sigl and Chamoun 13and in a review by Rampil. 2The data contained in both the bispectral analysis and conventional frequency–power analyses of the EEG are used to create the proprietary parameter of the bispectral index, or BIS. 2,13BIS is a dimensionless number scaled from 100–0, with 100 representing an awake EEG and zero representing complete electrical silence (cortical suppression). During development, BIS went through several revisions (table 1) and the currently available versions (versions 3.3 and 3.4) are scaled as shown in figure 1. The BIS integrates various EEG descriptors into a single variable. The mixture of subparameters of EEG activity was derived empirically from a prospectively collected database of anesthetized volunteers with measures of clinically relevant sedative endpoints and hypnotic drug concentrations. 14The process by which BIS was derived is shown schematically in figure 2. The EEG was recorded onto a computer and was time-matched with clinical endpoints and, where available, drug concentrations. The raw EEG data were inspected, sections containing artifact were rejected, and spectral calculations were then performed to produce both bispectral and power spectral variables. Following statistical ranking, the variables correlating best with the clinical endpoint were chosen. These were then fitted to a multivariate statistical model using the maximum likelihood solution to a logistic regression analysis to produce a continuous series of BIS values. This index was then tested offline in a prospective manner on a new database, and studies evaluated its clinical utility. The parameters used in the current implementations of BIS have been detailed by Rampil. 2The BIS monitor represents the successful effort to model EEG versus behavioral responses. The BIS algorithm uses various derivatives from conventional EEG power spectral analysis as well as elements of bispectral analysis.In the absence of a gold standard for determining anesthetic depth, initial clinical studies evaluated the predictive power of BIS for clinical endpoints including patient movement to skin incision (similar to the determination of minimum alveolar concentration [MAC]) and autonomic responses to stimulation (hypertension and tachycardia [MACBAR]). Data from the first two clinical studies were combined to form the database from which BIS version 1.1 was derived. 15,16BIS was compared with other commonly used power spectral derivatives to predict movement following skin incision in patients receiving thiopental–isoflurane anesthetic. 17EEG variables 2.0 min before incision were used as individual controls. A statistically significant difference between BIS levels, but not in spectral edge or median frequency, in subjects who moved at skin incision (BIS 65 ± 15, mean ± SD) was noted compared with those who did not move (BIS 40 ± 16). The accuracy (overall accuracy of prediction)‡was 83%, but the ability to correctly identify nonmovers (specificity) was only 63%. Power spectral derivatives did not predict movement in response to skin incision; this was confirmed in a recent study during thiopental–isoflurane anesthesia. 18The BIS version 1.1 was also evaluated for its ability to predict hemodynamic responses (more than 20% increase in blood pressure or heart rate) to laryngoscopy during a thiopental–nitrous oxide–opioid anesthetic technique. 16A statistically significant difference was found between patients who mounted a hemodynamic response (BIS 67 ± 10) compared with those who did not (BIS 45 ± 14). In this study, power spectral edge and median frequency did not distinguish those subjects who responded from those who did not. However, other researchers have found power spectral edge to be a useful predictor of hemodynamic response to laryngoscopy. 5To evaluate the predictive ability of BIS for movement using different anesthetic techniques, a prospective comparison was conducted using computer-controlled infusions of propofol (target plasma concentration, 4 μg/ml) plus alfentanil (125 ng/ml) compared with isoflurane (end-tidal concentration, 0.5%) plus alfentanil (125 ng/ml) anesthetics, techniques expected to achieve a 50% movement response to skin incision. 19In the period before skin incision, BIS was statistically significantly different for those who moved at incision compared with those who did not for each anesthetic technique, whereas other EEG derivatives were not significantly different. However, there was no difference between the patients in the isoflurane–alfentanil group that did not move (BIS 63 ± 10) and those in the propofol–alfentanil group who did move (BIS 63 ± 9). These studies demonstrate that BIS version 1.1 could predict movement response to incision but depended on the anesthetic agents used.Based on these results, a multicenter study of 300 patients from seven study sites using seven different anesthetic techniques was undertaken. 20Anesthetic technique was specific to each site and did not vary within each site, although there was significant overlap among drugs used at the various sites. One half of the patients at each site were randomized to receive anesthetic doses in which 50% of patients were expected to move in response to skin incision. The other half was randomized to a treatment group in which the anesthetic drug dose was adjusted to produce a BIS value of less than 60. The percentage of patients who moved in the group where a 50% movement rate was expected was 43% (BIS 66 ± 19 before incision). In the BIS-guided group (in whom anesthetic doses were larger), the movement response rate was significantly lower (13%), as was the BIS (51 ± 19). Overall, as BIS decreased, the probability of a movement response also decreased. At some sites where opioid doses were relatively large, there was no apparent relation between BIS and the probability of movement. Retrospective pharmacodynamic modeling using STANPUMP (Steven Shafer, VA Medical Center, Palo Alto, CA) was performed to estimate the effect-site concentrations of the intravenously administered anesthetics and opioids during balanced anesthesia. Using logistic regression analysis, an interaction model for the effects of the inhalational and intravenous anesthetics and opioids was derived. As the concentration of isoflurane and propofol increased, a decreasing BIS was associated with a decreasing probability of movement. In contrast, increasing opioid dose was associated with a decreased probability of movement without significant changes in BIS. Thus, when large doses of opioids are used, there is a poor association between the probability of remaining immobile after incision and BIS.Concurrently, several studies furthered our understanding of the anatomic pathways underlying the movement response to surgery. In rats, Rampil et al. 21demonstrated that MAC did not change following removal of the forebrain structures via craniotomy. They also demonstrated in the same model that spinal cord transection at C1–C2 level did not alter MAC. 22Antognini et al. 23separated the systemic and cranial circulations in the goat using bypass circuits to selectively anesthetize either the head or the body (including spinal cord). When the whole animal was anesthetized, MAC of isoflurane was 1.2%. When the cranial circulation alone was anesthetized, MAC was 2.9%. The conclusion from these three studies was that the movement response–reflex to skin incision is mediated primarily at spinal cord level. 24This anatomic separation of EEG generator sites from the somatic motor control sites in the spinal cord may explain the inability of BIS, which is derived from cortical EEG, to predict reflex movement. Therefore, clinical endpoints used during the development of the BIS version 1.1 were reevaluated.These data indicate that the hypnotic component of anesthesia (i.e. , "sleep") differs from the analgesic component 25(fig. 3) and suggest that a satisfactory anesthetic state can be obtained by a balance of hypnotic drugs (e.g. , volatile or intravenous anesthetics) and analgesic drugs (e.g. , opioids), resulting in unconsciousness and areflexia. Generally, a balance between hypnosis and analgesia is sought. If the dose of hypnotic agent is large, then relatively smaller amounts of analgesic are needed. If analgesic doses are relatively large, then hypnotic medications are decreased to avoid hemodynamic instability. Sedation was selected as the most appropriate clinical endpoint of hypnosis, and BIS was reformulated (version 2.0 and greater) from the existing database 26(table 1). A study in 72 volunteers established the relation between BIS, plasma drug concentrations, and level of sedation. 14These data were also used to develop BIS version 3.0 offline. Steady-state equilibration of plasma drug concentration and effect-site or brain concentration were achieved using computer-controlled, pharmacokinetically driven infusion devices targeted to hold plasma drug concentrations constant for a minimum of 15 min. In these volunteers, the relation between BIS, sedation, and memory function were evaluated using propofol, midazolam, isoflurane (end-tidal concentration held constant) or alfentanil, administered individually. Concentrations of each individual drug were increased in a stepwise fashion after equilibration at each level in a sequence of three to four steps to beyond the level that would normally cause unconsciousness. Subsequently, doses were decreased in a stepwise manner and increased again, and then patients were allowed to recover, so any EEG evidence of acute tolerance could be evaluated. The BIS version 3.0 score (r = 0.883) correlated significantly better with the Observer's Assessment of Awareness/Sedation (OAA/S) than did the measured propofol concentration (r =−0.778, P < 0.05). 27The correlations between BIS and OAA/S for isoflurane and midazolam were 0.85 and 0.75, respectively; these values were not statistically different from the correlation obtained between measured drug concentrations and OAA/S. BIS values representing unconsciousness (OAA/S = 2) in 50% and 95% of volunteers were 67 and 50, respectively. BIS version 3.0 also had a very high prediction probability (PK) 28§(0.88–0.98) for correctly identifying loss of consciousness. Alfentanil (50 or 100 ng/ml), alone or in combination with propofol, 29did not influence this version of BIS. Gajraj et al. 30studied 12 patients with spinal anesthetics (but no surgical stimulation) during repeated transitions from consciousness to unconsciousness following propofol infusions. At a BIS of 55, all patients were unconscious. No data presently exist on the effect of surgical stimulation on the thresholds (BIS) for awareness and memory under general anesthesia.The BIS version 3.0 was also found to predict responsiveness to verbal command during sedation or hypnosis better than either targeted or measured serum propofol concentration (with or without nitrous oxide). 31Katoh et al. 32demonstrated the value of this BIS version as a tool for predicting depth of sedation and hypnosis in patients anesthetized with sevoflurane. The PK29for BIS and sevoflurane concentration (0.966) was consistent over the entire sedative range. Both BIS and sevoflurane concentration had a linear relation with OAA/S. Loss of response to mild prodding, defined as a transition from OAA/S score of 2 to 1, occurred at a mean ED50BIS of 66 (95% confidence interval [CI], 64–68; ED95= 58). No EEG parameter, including BIS, was a significant predictor of movement in response to skin incision in this study. Other studies confirmed the relation between BIS and level of sedation after midazolam, 33intraoperative recall after propofol sedation, 34and suppression of learning after propofol. 35Taken together, these data suggest that BIS accurately reflects the degree of sedation with volatile and intravenous hypnotic agents, including midazolam. However, reformulation of the BIS decreased the ability to predict movement responses or hemodynamic changes to painful surgical stimulation. 36The ED50for unconsciousness (BIS 67) in volunteers 14was confirmed in paralyzed patients anesthetized with thiopental or propofol. 37In this study, patients received a single dose of propofol or thiopental and were paralyzed with vecuronium (0.1 mg/kg). The forearm was isolated from the neuromuscular blocking agent by a tourniquet inflated above systolic blood pressure, and return of consciousness was defined as the patient squeezing the investigator's hand twice in response to command. In this study, no patient recovered consciousness with a BIS less than 58, and a BIS of 65 signified a less than 5% probability of return of consciousness within 50 s. BIS did not specifically identify when a particular patient would return to consciousness. This was confirmed by other investigators. 30,38A limitation of all the "return of consciousness" studies described in this review is that they were conducted in the absence of noxious stimulation. It should also be noted that the definition of "return to consciousness" varies widely across the referenced studies and does not consistently include evaluation of complex command performance (e.g. , "move your left hand" or "squeeze my hand twice"). Ethical concerns make it impossible to intentionally provoke return of consciousness during the noxious stimulation of surgery. Thus, there are no data to provide confidence in transferring consciousness thresholds determined from volunteer studies into the practice of clinical anesthesia.Electroencephalographic power spectral parameters display complex relations with hypnotic drug dose that are unique to each class of agents. 39As mentioned previously, BIS and intravenous or volatile hypnotic dose have been shown to correspond in a statistically significant, linear, monotonic fashion during clinical trials, with BIS decreasing as hypnotic dose increased. 29,32,35,40When modeling effect-site concentrations of sevoflurane or isoflurane in surgical patients before intubation, BIS had a high predictive power (median coefficients of determination, 0.92 and 0.93, respectively) and displayed some hysteresis (effect site equilibration half time (t1/2ke0), 3.5 ± 2.0 min and 3.2 ± 0.7 min, respectively) with end-tidal anesthetic measurements. 40Quantitative analysis of hysteresis provides information on the speed of onset–uptake and offset–elimination of anesthetic action, whereas monitored or estimated plasma or effect-site drug concentrations does not. The only previous investigation of the dynamic relation between BIS (version 1.1) and end-tidal volatile anesthetic concentration cannot be directly compared because of subsequent reformulation of the BIS. 39To date, the most direct evidence linking BIS to brain cellular activity was provided by Alkire, who investigated the correlations between cerebral metabolic rate, sedation, and BIS. 41With each patient serving as his or her own awake baseline control, regional cerebral metabolic activity was imaged using positron emission tomography under three different conditions: propofol sedation, unconscious propofol, or isoflurane anesthesia. Alkire found that the magnitude of the anesthetic-induced changes in the EEG, evident during sedation and light anesthesia, paralleled the reduction in global cerebral metabolism. Reduction of whole-brain metabolic activity was dose-dependent and decreased in a linear fashion.The BIS has recently been used as a surrogate measure of anesthetic effect on the brain and employed as the control variable for closed-loop feedback for general anesthesia. et al. computer-controlled propofol infusions adjusted to an BIS of BIS value at which patients consciousness. This feedback model was to BIS by effect-site propofol concentration to within of values varying of are data on the relation between BIS and memory under sedation and anesthesia. et al. that BIS well with OAA/S during sedation with both propofol and midazolam during under regional anesthesia. OAA/S score of or response to a to a BIS value of ± and a probability of OAA/S score of 2 or response to mild to a BIS value of ± and a complete of In volunteers administered a propofol a of on nonlinear regression analysis, learning was by 50% at a BIS value of ± 1. These were by et al. volunteers during propofol anesthesia. was at BIS values than response to command with (95% and (95% for recall and (95% and (95% for consciousness. It should be noted that these studies were conducted in the absence of surgical et al. and memory in acute patients across a of BIS values during surgery. was tested by of No patient had However, there was a relation between BIS and the ability of patients to complete with during i.e. , at BIS levels, patients were to accurately complete than would be expected by information occurred at BIS between and This study demonstrated that memory was related to the depth of using BIS has been associated with a reduction in anesthetic agent in reduction in anesthetic dose could to an increase in the of The of awareness during general anesthesia has been to be between and date, there have been uses of BIS with an of awareness to Aspect Medical as of Aspect Medical Systems, Natick, BIS was 65 or in in which BIS were were because of either a of BIS or or of Therefore, although the of awareness may be of BIS to anesthetic does not to increase the likelihood of response to sedation and hypnosis is to predict because of a complex of including of medications and significant individual and pharmacodynamic of anesthetic effect using BIS should of drug to each both potential and of hypnotic The of hypnotic is defined by the absence of awareness and It should also be associated with the minimum dose of hypnotic of hypnotic should speed and et al. a multicenter in patients using a standard anesthetic technique. were randomized either to a standard practice group or to standard practice with BIS infusions were adjusted by clinical in the standard practice group and by to BIS values of during to in the in the standard practice group in BIS values in the compared with 50 in the The propofol infusion rate for of anesthesia was decreased in the group compared with the standard practice group the propofol dose used was lower in the BIS the of anesthesia was also significantly in this to was min (95% in the control group and decreased to min (95% with BIS In the 43% of patients were on in the compared with in the standard practice The of did not between This study demonstrated that hypnotic during anesthetic can speed and from anesthesia propofol et al. were to receive either or sevoflurane anesthesia, and the was either of BIS value or used BIS a value to volatile anesthetic BIS values in the 40 during anesthetic whereas those in the 60. anesthetic decreased significantly by compared with controls. to decreased from ± min ± SD) to ± min for and from ± 3.5 min to ± min for sevoflurane. BIS time to verbal responsiveness decreased from ± min to ± min for and ± min to ± 2.0 min for sevoflurane. However, time to of time to and time to were not by BIS to other anesthetic agents, is a anesthetic with effects on the EEG. doses of to produce did not reduce BIS. was used in with propofol sedation, there was an interaction to achieve hypnotic did not change BIS values. it that BIS cannot be used to monitor hypnosis during of nitrous at of to 50% does not alter BIS, does it cause unconsciousness. nitrous responsiveness to command is but BIS does not sedative concentrations of nitrous not to BIS, which is consistent with its as a hypnotic The of nitrous to plasma concentrations of propofol in volunteers decreased the probability of response to a of at any BIS level. no studies have investigated the effect of the of nitrous to a general anesthetic during surgical are currently regarding opioid and interaction of opioids a spectrum of with on BIS. No studies have evaluated the utility of BIS in anesthetics on large doses of is data to evaluate the of BIS in patients with In who was found to have a determined EEG, BIS values were baseline = the BIS did not in or activity may be in with EEG signal and the BIS EEG are to exist in the to and exist in the to although BIS uses EEG to This separation is not and can in the conventional EEG range. This activity is as the BIS. BIS values can also with high by or activity can be displayed on the there is no to the BIS Therefore, BIS values that are high on clinical should be with the amount of were used to develop and the BIS. The influence of and of the brain on BIS, as well as its correlation to drug effects and anesthetic is in exist to and awareness in the and should not be without between level of sedation, and anesthetic with BIS in have not been However, et al. an linear relation between BIS and end-tidal sevoflurane concentration in and BIS decreased by 50% in than 2 of at an end-tidal sevoflurane concentration of (95% compared with (95% in consistent with the known increase in MAC in this is to BIS provides an measure of hypnotic drug sedation such as the been used to measure the level of in patients and in the development of the BIS. As described previously, a number of have the high correlation between BIS, hypnotic drug concentration, and OAA/S for sedation. that BIS may be for sedation during monitored anesthesia and et al. the loss of consciousness as and the of recall as BIS correlated significantly than any other EEG variable with both loss of consciousness and return to consciousness after midazolam propofol sedation. studies suggest that BIS values of an loss in information and recall during and midazolam are both used for sedation in defined clinical endpoints and significant The influence of on hypnotic response is in these It would patients and speed from sedation to accurately monitor and hypnosis in the It is not known patients should receive hypnotic infusions or doses should be to of or can BIS although of using BIS may be direct measure of hypnotic would for and tolerance during hypnotic infusions. However, the of awareness and recall in the A number of be for continuous of patients in the (e.g. , The is an for EEG, and it is useful information can be derived in this the BIS in the is a of because data from the suggest that is was developed using clinical endpoints of sedation and monotonically to both the hypnotic component of anesthesia and to anesthetic drug It has been tested and in randomized clinical BIS both the potential for awareness and of hypnotic but does not predict movement or hemodynamic response to can it predict the consciousness exist to the of BIS. It is not useful during anesthesia or in patients with in technology have an this not function beyond the activity from over the and can and the BIS. be to activity and to be of the in monitor and A version of the BIS, to make the index less to is developed Aspect Medical Natick, move new have to be in a manner that both and utility in clinical understanding of the clinical of this new technology is in its and its to the practice of has to be