Study Design:
Managing Pharmacodynamic Variability in Anesthesia:
Integrating the EEG, Effect Site Anesthetic Concentrations
and Measures of Anesthetic Depth
Steven L. Shafer, M.D.
Associate Professor of Anesthesia, Stanford University, Stanford, California
Staff Anesthesiologist, Palo Alto VA Health Care System
The holy grail of clinical pharmacology in anesthesia is to find a practical measure of “anesthetic depth.” Part of the difficulty of developing such a measure is our inability to define “anesthetic depth.” It is easy to define what constitutes a clinically acceptable anesthetic state: a patient who is still, unconscious, and hemodynamically stable. However, a patient can be rendered still using muscle relaxants without any assurance that the patient is adequately anesthetized. Stable hemodynamics can be achieved using vasoactive drugs, but again there is no assurance that the patient is unconscious. Because muscle relaxants and adrenergic blocking drugs are widely used, the patient who is still and hemodynamically stable may nevertheless be wide awake. Obviously we need to monitor of the mental state of the patient. Unfortunately, no such monitoring device exists.
[pic]
Figure 1: Probability of response as a function of steady-state plasma alfentanil concentration.
[pic]
Figure 2: Study relating thiopental concentration to clinical measures of drug effect for Verbal, electrical Tetanus, trapezius Squeeze, Laryngoscopy, and Intubation.
To understand what anesthetic depth means, we must first understand the relationship between drugs that induce the anesthetic state and measures of response. For the intravenous anesthetics, Ausems set a high standard with his careful examination of the relationship between plasma alfentanil concentration and the likelihood of response to noxious stimulation, shown in figure 1.[i] Ausems’ data show that when alfentanil is used with nitrous oxide 80% of patients will respond to laryngoscopy at an alfentanil concentration of 400 ng/ml, while fewer than 5% of patients will respond at a concentration of 700 ng/ml. If anesthetic depth is a measure of the CNS effect of anesthetic drugs, then Ausems’ data suggest we should define anesthetic depth as the probability of response to noxious stimulation.
Similar work has been done for thiopental. Hung and colleagues used a computer controlled infusion pump to obtain two steady state target concentrations of thiopental, as shown in figure 2.[ii] After allowing 5 minutes for plasma-effect site equilibration, a series of graded stimuli were applied to the patient: verbal electrically induced tetanus, trapezius squeeze, laryngoscopy, and intubation. Movement was the measure of response.
Figure 3 shows individual move/no move data as a function of thiopental concentration. We can calculate the probability of response vs concentration using logistic regression, as shown in figure 4. Figure 4 suggests an ordering of the stimuli based on the C50 of thiopental in suppressing response: V < T < S < L < T. Because of the biphasic EEG response (activation followed by depression), the EEG measure used in this study did not correlate with the clinical response.
[pic]Figure 3 Individual response/no response data vs. thiopental concentration.
Anesthesia is usually performed by combining an hypnotic drug with an opioid analgesic. When more than one drug is involved, the drug concentration may still give information about the likelihood of response to noxious stimulation, but the interaction between drugs must be considered. For example, Vuyk et al recently examined the interaction between propofol and alfentanil for intraoperative maintenance of anesthesia (an “analgesic” endpoint, since the stimulus is surgical pain) and emergence from anesthesia (an “hypnotic” endpoint).[iii]
Their results (figure 5) show how two drugs, taken together,
predict the likelihood of response to stimulation.
[pic]
Figure 4: Curves demonstrating the probability of response vs thiopental concentration.
Does the EEG respond to stimulation as, for example, blood pressure and heart rate might? Figure 6 shows the response of the canonical EEG variate (an EEG parameter optimized for opioids) to changing alfentanil concentrations and noxious stimulation in a patient receiving alfentanil by computer controlled infusion pump (CCIP). The concentration in the effect site is shown by the thin line in the bottom panel. The temporal relationship between the EEG and effect site alfentanil concentration is obvious. The crosshatched areas show the patient's hemodynamic response to four stimuli. The canonical EEG variate responded briskly to the stimulation, but was the response attenuated at
higher alfentanil concentrations. The EEG can thus be a sensitive measure of opioid drug effect that both correlates with effect site opioid concentration and simultaneously integrates stimulus-induced CNS activation with opioid-induced CNS depression. This integration of opioid effect and CNS response to stimulus suggests great potential of the EEG as a measure of anesthetic depth.
[pic]
Figure 5: Propofol/Alfentanil interaction on the concentrations required to suppress response to noxious intraoperative stimulation, and those associated with emergence from anesthesia.
Over the past 18 months year we have been analyzing a remarkable data base gathered by Aspect Medical Systems of Natick Massachusetts. Aspect Medical Systems is developing an EEG monitor to be used as an intraoperative measure of the anesthetic state. Their monitor uses a proprietary algorithm based on bispectral analysis to develop a measure of drug effect called the “BIS.” Bispectral analysis itself is based on Fourier analysis. It incorporates phase information to identify those frequencies that are harmonics of underlying frequencies. If there is information in the phase relationships between frequencies in the EEG, the bispectral index will capture information that is otherwise lost in Fourier derived analyses of the EEG. Aspect used canonical correlation to derive the univariate index “BIS” from the multivariate response vector produced by bispectral transformation.
[pic]
Figure 6: The relationship between heart rate, blood pressure, an EEG measure of opioid drug effect, and the predicted plasma and effect site alfentanil concentration.
Aspect has performed several large multi-center trials to understand how BIS relates to anesthetic depth. The first multicenter trial examined EEG and movement in response to incision (an analgesic endpoint) in 345 subjects (183 men, 162 women) ranging in age from 17-81 years, ASA class I (n=90), II (n=199), or III (n=56) anesthetized with a combination (depending on study center) of propofol, alfentanil, thiopental, midazolam, thiopental, isoflurane, and nitrous oxide. The second trial examined EEG and recall, sedation, and eyelash reflex (hypnotic endpoints) in 92 healthy volunteers receiving a combination (depending on study center) of propofol, alfentanil, midazolam, isoflurane, and nitrous oxide. The multicenter trials have been conducted by an experienced team of investigators, including Drs. Peter Sebel (Emory), Carl Rosow (Harvard/MGH), Lee Kearse (Harvard/MGH), Peter Glass (Duke), Marc Bloom (U. Pittsburgh), Ty Smith (UCSD), Ira Rampil (UCSF), Randy Cork (University of Arizona), and Mark Jopling (Ohio State University).
This data base, gathered at the cost of nearly 10 million dollars, may be the most comprehensive data base relating dose, concentration, clinical measures of anesthetic drug effect, and EEG measures during anesthesia that will ever be gathered. It is highly relevant to our question of managing pharmacodynamic variability in anesthesia. We turned to this data base to ask whether the EEG can be used as a tool for clinical management of pharmacodynamic variability in the operating room. Simply stated, can we develop models using the EEG that tell us more about the probability of response to stimulation than conventional measures such as heart rate and blood pressure?
Table 1:
|Available measures (predictors) of |Endpoints (general class) |
|anesthetic depth | |
|Blood Pressure |Recall (hypnotic) |
|Heart Rate |Sedation (hypnotic) |
|Change in blood pressure |Eyelash reflex (hypnotic) |
|Change in heart rate |Movement in response to incision (analgesic) |
|Bispectral Index | |
|Measured plasma drug concentration | |
|Predicted plasma drug concentration | |
|Predicted effect site drug concentration | |
The analysis that follows will show models that relate the measures potentially available (left side of table 1) to the endpoints (right side of table 1) that represent distinct definitions of the anesthetic state. Only models and conclusions that have remained statistically significant in cross-validations will be presented. All modeling was performed with NONMEM[iv] and graphed with S-plus (Mathsoft, Inc).
[pic]
Figure 7: The relationship between blood pressure, heart rate, and the probability of recall in the unstimulated patient.
First, the bad news. Blood pressure and heart rate are the primary end-points observed during anesthesia, particularly in the paralyzed patient. Figure 7 shows the probability of recall as a function of heart rate and blood pressure. In figure 7 and those that follow, 0 represents no response, and 1 represents a response. Figure 7 shows that heart rate and blood pressure contain no information about the ability to form memories that can be recalled in the unstimulated patient. This is not good news for anesthesiologists, because blood pressure and heart rate are the most commonly used measures of anesthetic adequacy. In other words, the paralyzed, unstimulated, patient could be fully conscious and yet the anesthesiologist would have no clue from the most
commonly used measures of anesthetic adequacy.
[pic]
Figure 8: The relationship between blood pressure, bispectral index, and the probability of recall
[pic]
Figure 9: The relationship between drug concentration, EEG, and the probability of recall.
Does the EEG help? Figure 8 shows the relationship between the bispectral index, blood pressure, and the probability of recall in the unstimulated subject. Clearly there is information in the EEG about the level of awareness of the subject. Specifically, if the bispectral index is less than 50, then there is nearly no possibility of awareness, while if the index is greater than 90, then the probability of awareness is high. This suggests that the EEG contains information about the level of consciousness, which will be generally referred to as “hypnosis,” with ability to form memories that can be recalled representing an “hypnotic endpoint.” In other words, the EEG can be used to help manage pharmacodynamic variability.
The probability of recall is partly determined by the concentration of the anesthetic drugs. Figure 9 shows the probability of recall as a function of the “propofol equivalents” and the bispectral index. The “propofol equivalents” represents the relative contributions of all drugs, as determined by an interaction model. To model the interaction, every drug used in the trial enters the model, both by itself, and (to account for interactions), as a cross product with the concurrent concentration of every other drug. All drug concentrations, and their cross products, have a coefficient that is determined by the regression except for propofol, whose coefficient is set to 1 (hence, the final results are in units of equivalent propofol concentration). For the volunteer trials, where the drugs were propofol, isoflurane, midazolam, and alfentanil, the regression model for concentration was:
[pic] Eq. 1
Only coefficients significantly different from 0 were retained in the final model. None of the interaction terms came out as significant (more on this later), and the final model was:
[pic] Eq. 2
Equation 2 suggests that isoflurane, in vol %, is approximately 3 fold more potent than propofol, in (g/ml. If the MAC awake/asleep for isoflurane is 0.7, then the concentration of propofol associated with loss of consciousness should be 2.0, both of which are in agreement with prior literature. As an aside, it is interesting to compare the performance of measured concentrations as opposed to predicted effect site concentrations. The data base contained measured plasma propofol, midazolam, and alfentanil concentrations, and end-tidal isoflurane concentrations. We also had access to predicted end tidal, plasma and effect site concentrations, based on application of published PK/PD models for propofol, midazolam, and alfentanil, and predicted brain concentrations for isoflurane, based on application of the isoflurane model published by Lerou and colleagues.[v] Equation 1 was solved for both measured concentrations, and concentrations predicted by PK/PD models. The performance of the predicted concentration was at least as good as, and possibly better than, the performance of the measured concentrations. This is a surprising result. The implication is that predicted concentrations, which can be obtained in real time from a computer in the operating room, are at least as useful as the actual concentrations that would be obtained were real-time measurements of concentration available.
Returning to figure 9: while there is information in the EEG about the probability of recall, there is only information about this end point at drug concentrations less than 1.5 (g/ml propofol equivalents. Above this concentration, the probability of recall is 0, regardless of the EEG. This analysis suggests that of the measures of anesthetic depth potentially available to the anesthesiologist, predicted effect-site drug concentration, in terms of propofol equivalents, is the single best predictor of the probability of recall. However, there is useful additional information in the bispectral index when the predicted propofol equivalents are between 0.5 and 1.5 (g/ml.
[pic]
Figure 10: The probability of deep sedation as a function of bispectral index and propofol equivalents.
Figures 10 and 11 show the relationship between the EEG and two other “hypnotic” end-points: deep sedation and eyelash reflex, respectively. Deep sedation is defined as lack of response to loud verbal command. The shapes of these relationships were identical, with a slight offset between them, suggesting that eyelash reflex is a “lighter” stimulus than response to loud verbal command. The only difference between these responses we could find with NONMEM was a difference in C50 for the propofol equivalents and BIS values. Otherwise, the interaction model was identical for the two responses:
[pic] Eq. 3
Figures 9, 10, and 11 show that there is information in both the predicted drug concentrations and in the bispectral index for hypnotic endpoints. Statistically, predicted drug concentration and bispectral index contribute approximately equally to the model, and the prediction of response is better with both measures than with either alone. The conclusion from figures 9, 10, and 11 is that the best predictors of the hypnotic state in the unstimulated patient are the predicted drug concentration at the effect site and the EEG. It is interesting that both of these are readily available to anesthesiologists using software (e.g., . med., STANPUMP and STELPUMP) and hardware such as the Aspect EEG monitor.
[pic]
Figure 11: The probability of no eyelash reflex as a function of bispectral index and propofol equivalents.
The classic definition of anesthetic depth is movement in response to noxious stimulation.. The clinical relevance of movement to anesthesia has changed over the past 30 years because 1) muscle relaxants are widely used to suppress movement directly, 2) Rampil and colleagues have demonstrated that movement in response to noxious stimulation is a spinal reflex,[vi] and thus has little relevance as a measure of cortical activity (i.e., awareness), and 3) the extensive use of opioids in anesthesia may render patients sufficiently analgesic that they don’t move in response to noxious stimulation, but are nonetheless conscious. Observations 2 and 3 suggest that movement is an “analgesic” endpoint, in that it is a response to pain that is suppressed mostly by the analgesic properties of anesthetic drugs, requiring low concentrations for drugs that have primary analgesic efficacy (e.g., opioids, ketamine) and high concentrations for drugs that do not have primary analgesic efficacy (e.g., propofol, isoflurane).
In the ASPECT patient trials, movement was the primary endpoint. Ethical concerns precluded maintaining surgical patients at sufficiently light planes of anesthesia to assess recall and other “hypnotic” endpoints. These trials involved (at separate institutions) the use of fentanyl, alfentanil, and sufentanil. In our analyses these opioids were normalized to fentanyl concentrations by dividing the concentration by the potency as determined from prior EEG studies.
Figures 12, 13, and 14 relate the probability of movement to heart rate, blood pressure (figure 12), bispectral index and opioid concentration (figure 13), or bispectral index and propofol concentration (figure 14). Figure 12 demonstrates that neither blood pressure or heart rate predicted of movement. Figures 13 and 14 show that in this study the bispectral index offered approximately as much information as opioid concentration of propofol concentration, but there was no threshold for either BIS or concentration that separated patients that moved from patients that didn’t move.
[pic]
Figure 12: The probability of not moving as a function of blood pressure and heart rate.
[pic]
Figure 13: The probability of not moving as a function of bispectral index and opioid concentration.
[pic]
Figure 14: The probability of not moving as a function of bispectral index and propofol concentration.
[pic]
Figure 15: The correlation between opioid and propofol in the Aspect trials relating EEG to movement.
Figures 13 and 14 suggest that BIS may be a useful predictor of movement. However, BIS contains little new information if predicted opioid or propofol concentrations are included in the model. This is similar to the results reported recently by Leslie et al,[vii] although they achieved better prediction with both BIS and propofol concentrations then we observed here.
Figures 13 and 14 suggest that movement in this trial was equally well predicted from opioid or propofol concentration. At first blush this appears bizarre, given data, such as shown in figure 5, demonstrating profound synergy between propofol and alfentanil. Figure 15 explains why an interaction was not seen. In this clinical trial, clinicians titrated drugs they way they usually do. As a result, opioid and propofol concentrations correspond closely because light anesthesia was treated by adding more propofol and more opioid, and excessive anesthesia was treated by lowering propofol and opioid levels. Thus, the propofol concentration was an excellent predictor of opioid concentration, precluding analysis of their interactions.
Conclusion:
Because of pharmacodynamic variability, it is possible that the still, hemodynamically stable patient may be conscious during anesthesia and surgery. No monitor exists that can accurately measure the mental state of the paralyzed patient. If we define anesthetic depth in terms of the probability of reaching specific hypnotic endpoints, such as recall or sedation, then that best predictor of the probability of recall is a the effect site concentration of anesthetic drugs, predicted using published pharmacokinetic/pharmacodynamic models. The best predictor of sedation is the EEG, using the Bispectral Index. Combining these two information streams results in better prediction of the anesthetic depth than either measure alone. Clinically, this is good news because both of these measures are readily available with existing technology. Additionally, this study confirms many prior reports that blood pressure and heart rate are, in general, poor predictors of anesthetic adequacy.
References
-----------------------
[i] Ausems ME, Hug CC, Stanski DR, Burm AGL: Plasma concentrations of alfentanil required to supplement nitrous oxide anesthesia for general surgery. Anesthesiology 65:362-373, 1986
[ii] Hung OR, Varvel JR, Shafer SL, Stanski DR. Thiopental pharmacodynamics: II. Quantitation of clinical and EEG depth of anesthesia. Anesthesiology 77:237-244, 1992.
[iii]. Vuyk J et al.. The pharmacodynamic interaction of propofol and alfentanil during lower abdominal surgery in women. Anesthesiology 83:8-22, 1995.
[iv]. Beal SL, Sheiner LB: NONMEM User’s Guide. San Francisco, University of California San Francisco, 1979
[v] Lerou JG, Dirksen R, Beneken Kolmer HH, Booij LH. A system model for closed-circuit inhalation anesthesia. I. Computer study. Anesthesiology 75:345-355, 1991
[vi] Rampil IJ, Anesthetic potency is not altered after hypothermic spinal cord transection in rats. Anesthesiology 80:606-610, 1994
[vii]. Leslie K, Sessler DI, Smith WD, Larson MD, Ozaki M, Blanchard D, Crankshaw DP. Prediction of movement during propofol/nitrous oxide anesthesia. Performance of concentration, electroencephalographic, pupillary, and hemodynamic factors. Anesthesiology 84:52-63, 1996
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.
Related searches
- 3d design something
- 2019 toyota highlander design change
- product design website
- design something online
- design your own monogram free
- design your own tea towel
- design your dream house virtual
- 3d virtual home design free
- design a virtual house free
- design things online for free
- 2020 toyota highlander design change
- design something for free