Prediction of Movement during Propofol/Nitrous Oxide Anesthesia: Performance of Concentration, Electroencephalographic, Pupillary, and Hemodynamic Indicators

With approval from the Committee on Human Research at the University of California, San Francisco, we studied ten healthy volunteers aged 25 +/-4 yr (mean+/-SD), weighing 66+/-15 kg, and with a lean body mass (LBM) of 50+/-11 kg. Lean body mass was determined from height (cm) and weight (kg) using a formula [15]:Equation 1.

Lactated Ringer's solution was infused at 120 ml/h. A computerized anesthesia machine (Modulus CD, Ohmeda, Madison, WI) was used to monitor heart rate, oxygen saturation, and end-tidal gas concentrations. These were recorded directly using automatic record-keeping software (IdaCare, Premier Anesthesia Systems, Atlanta, GA). Arterial blood pressure was measured directly in the radial artery of the non-dominant arm.

Anesthesia was induced with propofol (Zeneca Pharmaceuticals Group, Wilmington, DE) using an infusion pump (Ohmeda). The infusion rate was increased in three 10-min steps from 0 to 21 mg *symbol* kg LBM sup -1 *symbol* h sup -1. After loss of consciousness, 60% nitrous oxide in oxygen was administered, and airway support was provided as necessary. Before induction, volunteers were asked to grasp a 20-ml plastic syringe firmly between the thumb and forefinger of one hand. Dropping the syringe and loss of the eyelash reflex were used to indicate loss of consciousness.

Thirty minutes after the propofol infusion was started, the trachea of each volunteer was intubated without administration of muscle relaxants. If coughing or movement was observed, an intravenous bolus of 20–40 mg propofol was administered. Because substantial movement was observed in the first four subjects, we altered the protocol to include 40 mg intravenous propofol immediately before intubation in the remaining volunteers. The lungs then were mechanically ventilated to maintain an end-tidal carbon dioxide partial pressure of 35–40 mmHg.

After intubation, the propofol infusion rate was decreased to 15.4 mg *symbol* kg LBM sup -1 *symbol* h sup -1, and kept constant for 15 min. An electrical stimulus was applied at 10 min, and volunteers' responses were observed for 5 min. Electrical stimuli were administered via 25-G needles placed intradermally on the left thigh and consisted of a 65- to 70-mA, 100-Hz electric current lasting 10 s. A positive response was defined as gross movement of the right leg, arms, or head within the first minute after stimulation. To prevent desensitization at the needle insertion site, the electrodes were moved [nearly equal] 1 cm cephalad after each stimulation.

At the end of the 15-min test period, the infusion rate was changed and the protocol was repeated. The infusion rate was increased by 1.75 mg *symbol* kg LBM sup -1 *symbol* h sup -1 when movement occurred, and similarly decreased when it did not. These up-and-down sequences were continued until volunteers "crossed over" from movement to nonmovement (or vice versa) four times, or 7 h of anesthesia elapsed.

The propofol infusion rate then was increased to 21 mg *symbol* kg LBM sup -1 *symbol* h sup -1, nitrous oxide discontinued, the endotracheal tube removed, and oxygen administered. The propofol infusion rate was decreased to zero over a 50-min period in 10-min steps. The times at which the eyelash reflex returned and volunteers were able to recall their birthdays were recorded.

To avoid any confounding effect of hypothermia, distal esophageal temperature was maintained between 36 and 37 degrees C using a Bair Hugger forced-air warming cover positioned over the legs (model 300 cover, model 200 blower, Augustine Medical, Eden Prairie, MN). [16].

Pupillary reflexes were evaluated using a portable infrared pupillometer (Fairville Medical Optics, Larkings Green, Amersham, England), as described previously. [12]The instrument was programmed to search for a stable pupillary diameter, flash a 0.5-s light stimulus, and initiate a 2-s, 10-Hz scan at the start of the stimulus. All measurements were taken from the right eye and ambient light was excluded from the left eye.

The EEG was recorded continuously during each study. After preparation of the skin, gold cup electrodes were filled with conductive paste and attached to the skin with adhesive. The electrodes were positioned at FP1 and FP2, with the reference electrode at the nasion and the ground electrode at the mastoid. A microprocessor-based, four-channel monitor (B500, Aspect Medical Systems, Natick, MA) was used to collect raw EEG data, which were recorded for off-line analysis (Version 3.0) by Aspect Medical Systems.

Electroencephalographic data were acquired in 4-s epochs, with each successive epoch overlapping the previous one by 3 s. After rejection of spurious data due to artifact, both the power spectrum and the bispectrum of each epoch were calculated. Individual power spectral parameters from the last 60 epochs were averaged together. High and low cutoff frequencies were 30 Hz and 0.5 Hz, respectively. The relative power spectral parameters calculated were the delta power (0.5–3.75 Hz), theta power (4.0–7.75 Hz), alpha power (8.0–13.5 Hz), beta power (13.75–30.0 Hz), 95% spectral edge, and median frequencies. The bispectra of 60 successive epochs were averaged and a series of intermediate bispectral parameters were computed from the averaged bispectrum. These intermediate parameters were combined into a multivariate Bispectral Index using coefficients developed on a separate EEG database that did not include the present subjects. Processed data from the two frontal channels were combined and saved every 15 s.

During each 15-min infusion, samples of arterial blood were obtained at 9 elapsed min in all volunteers. Samples also were obtained at 14 min in the first three volunteers, to assess the constancy of propofol blood concentrations during stimulation. They were placed in heparinized tubes and stored at 4 degrees C for up to 6 weeks (propofol blood concentrations decrease less than 0.2%/week at 4 degrees C). Samples were analyzed using a high-performance liquid chromatography assay, modified from the method of Plummer. [17]This assay is linear to at least 20 micro gram/ml and has a detection limit of 0.025 micro gram/ml and a coefficient of variation of 4% at 2 micro gram/ml in our laboratory.

A single measurement of arterial blood pressure and heart rate was recorded at each time. Electroencephalographic data sets collected 60 s before painful stimulation were used in the prediction of movement analysis. Electroencephalographic data collected during induction and awakening were not analyzed because EEG activation at low anesthesic concentrations may complicate interpretation of results. [18]Three sets of pupillary data were averaged, producing one averaged scan from which pupillary parameters were recorded for volunteers at each time during induction and awakening, and before stimulation. To determine the response of indicators to painful stimulation, pupillary and EEG data collected at -1.5, -0.5, 0, 0.33, 0.66, 1.5, 2.5, 3.5, and 4.5 min were analyzed.

To determine the value of each indicator for identifying loss and return of consciousness as defined by the dropping of the syringe, loss and return of the eyelash reflex, and ability to recall birth date, we used repeated-measures analysis of variance and Scheffe's F tests.

Electroencephalographic, pupillary, and hemodynamic data collected just before and for 5 min after each stimulus were analyzed using repeated-measures analysis of variance and Dunnett's tests. Paired t tests were used to compare the indicator data obtained when volunteers moved and did not move.

Effect-site concentrations were calculated by convolving the plasma concentrations over time against the function, k^e0e sup -k sub e0 *symbol* t, the disposition function of the effect site. The convolution interpolated adjacent plasma concentrations, assuming a linear rise when plasma concentrations were increasing, and a log-linear fall when plasma concentrations were decreasing. A value of 0.239 min sup -1 was taken as the k^e0.**** Theoretical background for these calculations is provided by the work of Verotta and Sheiner. [18]The relationships between movement in response to electrical stimulation and EEG, pupillary, hemodynamic, and propofol concentration data were defined using logistic regression (BMDP Statistical Systems, Los Angeles, CA and SPSS, Chicago, IL). Responses to the first three stimuli in each volunteer were not included in analysis, to exclude the effect of changing the protocol to include a 40-mg bolus of propofol before intubation. The remaining 130 stimuli were used for analysis. Logarithmic transformation of propofol blood and effect-site concentration, relative beta power, and median frequency was performed before analysis. Individual logistic curves were obtained for each volunteer, as well as an overall curve obtained by performing logistic regression on all 130 stimulus values. For the latter, the standard errors of the estimate, provided by BMDP for the variable values at the 50% and 95% nonresponse levels, were based on the assumption that the 130 stimuli were independent. These standard errors of the estimate were used to compute the 95% confidence limits. The significance value for the SPSS statistic "-2 log likelihood" was used as a measure of logistic curve fit. This significance value is obtained for the null hypothesis that the logistic curve explains the quantal data perfectly, that is, that the likelihood equals one. A larger significance value corresponds to a better logistic curve fit. See Smith et al. [19]for a further discussion of this measure.

Ability of propofol concentration, EEG, pupillary, and hemodynamic indicators to predict movement was evaluated using prediction probability (P^K), which compares the performance of indicators having different units of measurement. The mathematical basis of P^Kwas described by Smith et al. [19]Numerically, P^Kis the probability that an indicator predicts correctly which of a pair of randomly selected stimuli, one causing movement and the other not, will cause movement. An indicator that predicts perfectly whether a movement response will occur has a P^Kvalue of 1.0, whereas an indicator that performs no better than chance has a P^Kvalue of 0.5. Values for P^Kfor an indicator were computed for each volunteer, and an average of these P^Kvalues then was calculated. A value of P^Kalso was computed for all 130 stimuli combined, and the jackknife method was used to compute the standard error of the estimate, [19]based on the assumption that the 130 stimuli were independent.

The predictive performance of effect-site concentration was compared to the Bispectral Index, 95% spectral edge frequency, pupillary reflex amplitude and systolic arterial blood pressure. First, a Wilcoxon test was used to determine whether the individual-subject P^Kvalues for the alternative indicators differed from those for effect-site concentration. Individual-subject P^Kvalues for an indicator express the indicator's ability to predict a response if a subject's individual response curve is known. Second, a paired-data jackknife analysis [19]was used to determine whether combined-subject P^Kvalues for alternative indicators differed from the combined-subject P^Kvalue for effect-site concentration. Performance, as measured by the combined subject P^K, is of interest if the population response curve is known, but the individual-subject response curves are not (as is typically the case). For multiple comparisons, we used the Bonferroni method, as modified by Holm. [20].

Results are presented as mean+/-SD, except for the logistic curve 50% and 95% nonresponse levels, which are presented as estimate+/-standard error of the estimate. P < 0.05 was considered statistically significant.

Induction of anesthesia was smooth in all cases and no excitatory movements were observed. Recovery from anesthesia was similarly uneventful.

Pupillary, hemodynamic, and propofol concentration indicators during loss and return of consciousness changed in a sequence, consistent with gradually increasing anesthetic depth during induction and decreasing anesthetic depth during awakening. The volunteers were still awake when the syringe was dropped but not when the eyelash reflex was lost.

Loss and return of the eyelash reflex occurred at greater propofol effect-site concentrations than either dropping the syringe or recall of the birth date. Initial pupillary diameter was smaller when the eyelash reflex was lost than when it was regained: the other parameters did not differ significantly at the two times. The pupillary reflex had returned to normal by the time volunteers could recall their birth date, but they remained drowsy, and this sedation was reflected by residual blood propofol concentrations (Table 1).

Non-response levels were defined by the fraction of volunteers predicted not to move at a particular blood or effect-site concentration, or value of an EEG, pupillary or hemodynamic indicator (Table 2). The propofol effect-site concentration at the 50% nonresponse level (C^p50) was 1.8 micro gram/ml (95% confidence limits: 1.4–2.3 micro gram/ml;Figure 1). To prevent movement on 95% of occasions (C^p95) a predicted effect-site concentration of 9.7 micro gram/ml (4.8–19.4 micro gram/ml) would be required. Comparable values for propofol blood concentration were 1.7 micro gram/ml (1.2–2.3 micro gram/ml) and 12.3 micro gram/ml (5.1–29.6 micro gram/ml) respectively. The predicted systolic arterial blood pressure at the 95% non-response level was 74 mmHg (64–83 mmHg;Figure 2), a 50% decrease from its pre-induction value. Data for the Bispectral Index, 95% spectral edge frequency and pupillary reflex amplitude are presented in Figure 3, Figure 4, Figure 5.

Prediction probability (P^K) values were calculated to determine the relative performance of the EEG, pupillary, hemodynamic and concentration indicators (Table 2). The P^Kvalue (based on all 130 stimuli) for propofol effect-site concentration was 0.76 and for blood concentration, 0.74. In contrast, the Bispectral Index had a P^Kof 0.86, and 95% spectral edge frequency, 0.81. Median frequency and relative beta power also had P^Kvalues exceeding 0.80. When compared to the effect-site concentration, however, no indicator proved significantly different. All indicators, with the exception of heart rate, predicted movement after painful stimulation significantly better than chance alone. When P^Kwas calculated using the individual values in each volunteer, prediction improved. However, no significant differences in performance were demonstrated between effect-site concentration and bispectral index, 95% spectral edge frequency, pupillary reflex amplitude or systolic arterial blood pressure. A similar analysis performed using propofol blood concentration instead of effect-site concentration produced the same result, except that the Bispectral Index was significantly different from blood concentration (in the direction of improved prediction) in the paired-data jackknife comparison of combined-subject P^Kvalues.

The responses of representative indicators to painful stimulation are shown in Figure 6. At each measurement time, most indicator values were significantly greater in volunteers who moved than in those who did not. The pupillary reflex amplitude and systolic arterial blood pressure changed statistically significantly after stimulation. Pupillometry demonstrated the most dramatic response. There was swift dilatation of the pupil, which was similar in magnitude in both groups, but sustained longer in those who moved than in those who did not. The increase in systolic arterial blood pressure after stimulation was statistically significant, but not clinically important. The Bispectral Index and 95% spectral edge frequency did not change significantly after painful stimulation. Heart rate did not differ significantly between those who moved and those who did not; nor did heart rate increase after stimulation.

We assessed differences between propofol blood concentrations at 9 and 14 elapsed min of the 15-min constant-rate infusions in the first three volunteers (48 sample pairs). No propofol blood concentration at 14 min differed from that at 9 min by more than 34%. When the infusion rate was increased at time zero, propofol blood concentration changes ranged from +34 to -23%(1+/-14%) between 9 and 14 min. When infusion rate was decreased, propofol blood concentration changes ranged from +19 to -20%(2+/-10%). Modeled propofol effect-site concentration changes ranged +15 to -16%(3+/-10%) and +11 to -21%(1 +/-7%), respectively (Figure 7).

This study demonstrates that indices of pharmacodynamic effect, such as the Bispectral Index and 95% spectral edge frequency of the EEG, pupillary reflex amplitude, and systolic arterial blood pressure, predict movement as well as propofol effect-site concentration in healthy volunteers administered propofol/nitrous oxide anesthesia.

Systolic arterial blood pressure performed surprisingly well as a predictor of movement in these controlled circumstances. However, the effects of disease, surgery, fluid administration, and drugs affecting the circulation are likely to reduce the predictive value of arterial blood pressure in the clinical setting. [6]Administration of opioids may similarly reduce the performance of the pupillary light reflex.

The generally steeper slopes of the individual logistic curves compared to the curve generated by analyzing all 130 stimuli together indicates that knowing the responsiveness of a person facilitates prevention of movement and provision of adequate anesthesia in that person, [21]even though such individual "calibration" rarely is available clinically. The benefit of knowing individual responsiveness is quantified by the amount to which average individual-subject P^Kvalues exceed the overall combined-subject P^Kvalue.

The propofol blood concentration at which 50% of volunteers were predicted to move after supramaximal electrical stimulation (C^p50) was 1.70 micro gram/ml (95% confidence limits: 1.24–2.32 micro gram/ml). This finding was consistent with values previously reported during propofol/nitrous oxide anesthesia: 1.66 micro gram/ml during propofol/nitrous oxide/morphine anesthesia [22]and 2.5 micro gram/ml during propofol/nitrous oxide/lorazepam anesthesia. [23]In contrast, the C^p50 of propofol alone reportedly is 15.2 micro gram/ml. [24].

No indicator we studied perfectly predicted movement after supramaximal electrical stimulation. There are several possible explanations for this lack of predictive value that relate to propofol's pharmacokinetics and pharmacodynamics. To eliminate the confounding effects of pharmacokinetic variability in the response of subjects to a stimulus, constant blood concentrations and blood-to-effect-site equilibration are required. [7]Constant-rate infusions, such as those used in this study, result in exponentially declining rates of rise or fall in drug concentration, [25]and potentially may introduce error. We took advantage of the near steady-state concentrations toward the end of each step, and were able to maintain reasonably constant propofol blood concentrations between 9 and 14 min (as evidenced by data from the first three volunteers). As electrical stimulation was applied at 10 min, we expect that propofol blood concentration changed even less between 9 and 10 min, than between 9 and 14 min. Smith et al. [24]used only preassessment and postassessment samples which were within+/-35% of each other for their determination of propofol C^p50; all our sample pairs fell within these boundaries. The homogeneous nature of our study group, and the use of lean body mass to calculate dosage, [26]probably contributed to the observed relatively constant levels.

To account for effect-site equilibration in a series of exponentially changing rates of increase or decrease in blood concentration, [27]we calculated predicted effect-site concentrations, using a k^e0= 0.239 min sup -1 (t¹/2e0 = 2.9+/-2.2 min). This value was calculated by Schuttler et al. using the EEG median frequency.**** The assumptions inherent in this analysis that may introduce error are:(1) the plasma concentration versus time curves in our volunteers roughly captured the shape of the curves in the model, which assumed a linear rather than log-linear rise;(2) the effect-site concentration was related to the plasma concentration by a first-order process;(3) the EEG median frequency reflected propofol's action at its effect site; and (4) the intersubject and intrasubject variability in k sub e0 was low. Intersubject and intrasubject variability in k^e0is likely with propofol, given the large standard deviation in t¹/2e0, and is known to exist for other intravenous anesthetics. [27]In addition, variability in the interaction of the anesthetic molecule with its site of action, the time between interaction and effect, and the magnitude of that effect, [25]all may reduce k^e0prediction accuracy. Indicators of pharmacodynamic effect, such as EEG, pupillary, and hemodynamic parameters, predict response from actual central nervous system activity and therefore remove the effect of intersubject or intrasubject variability. We were unable, however, to demonstrate that the performance of such indicators was statistically superior to calculated effect-site concentration.

Imperfect prediction by any one indicator in our study may reflect involvement of more parts of the central nervous system than were measured. Electroencephalographic measures, which reflect cortical activity, also failed to predict movement perfectly during thiopental anesthesia in humans [28]and isoflurane anesthesia in rats. [29]As the origin of movement in response to painful stimulation during anesthesia apparently is below the cortex, [30]measures of brain stem or spinal cord function might perform better. However, in this study, hemodynamic and pupillary indicators were not superior to EEG. A multivariate analysis may have achieved better prediction. [31].

Another reason for the imperfect prediction of the EEG relates to a possible biphasic effect of propofol on the EEG. Initially, activation occurs, with an increase in power in the beta range (propofol blood concentration: 1–2 micro gram/ml)[32]; depression, and finally burst suppression follow. [33]There is a problem therefore interpreting the meaning of a particular value of an EEG indicator, [34]because one indicator value may result from two different propofol concentrations. For this reason, EEG values at end points of induction and awakening were not reported, because propofol concentrations at these times were in the potentially biphasic range. During the stimulation phase of this study, however, propofol concentrations generally were greater than 1.5 micro gram/ml (i.e., in the linear part of the concentration/effect relationship). In addition, we used nitrous oxide as well as propofol, which likely results in a greater depth of anesthesia: deeper than the biphasic range. Despite potential problems, EEG indicators proved predictive of movement.

The frontal leads are appropriate for monitoring EEG changes during propofol anesthesia.***** The reference electrode is not usually positioned at the nasion because this site is susceptible to interference from facial movement. However, both the frontal leads and the nasion are convenient locations for electrode placement, especially during cranial surgery. The EEG performed well as a predictor of movement, despite this montage.

A constant, supramaximal stimulus is essential to the determination of anesthetic potency. [14]Previous studies demonstrated that tetanic stimulation of the magnitude we applied is supramaximal. [35]Recently, Laster et al. [36]reported that electrical stimulation gave results comparable to tail clamp in rats given volatile anesthesia: only a 40-V stimulus produced a minimum alveolar concentration value significantly greater than tail clamp. We used the 15-V stimulus pattern of Laster et al., stimulus intensity was the same on each occasion, and the stimulating needles were moved after administration of each tetanic stimulation to avoid desensitization of nerve endings. We conclude, as did Laster et al., that, while electrical stimulation potentially is not supramaximal, it may be used as a substitute for conventional forms of stimulation in volunteers.

Using multiple identical stimuli in each volunteer, while economical, may result in intravolunteer correlation between stimuli, as described earlier. We potentially introduced error into our results by assuming, as have others, [28]that each measurement in each volunteer was independent. Logistic regression and the P^Kstatistic make the assumption of independent data; no comparable statistic has been developed that specifically permits nonindependent data. The assumption of independent data may have led to optimistic bias (underestimation) of confidence intervals for the 50% and 95% nonresponse levels and the standard errors of P^K(all 130 stimuli). The removal of optimistic bias could only strengthen our result that there were no significant differences in performance between effect-site concentration and Bispectral Index, 95% spectral edge frequency, pupillary reflex amplitude, or systolic arterial blood pressure.

Pupillary indicators revealed a more dramatic response to painful electrical stimulation than hemodynamic indicators, confirming our earlier observations. [37]In addition, the response of the EEG was attenuated compared to the pupillary light reflex. Time averaging of the EEG data may be responsible for the apparent lack of EEG response.

Simple indicators of loss of consciousness, such as dropping a syringe and loss of the eyelash reflex, often are used in studies of anesthetic induction. [27]We demonstrated in this study that loss and return of the eyelash reflex correspond to a deeper level of anesthesia than syringe-dropping or recall of the birth-date. The pupillary light reflex was most useful in predicting that a particular level of anesthesia had been attained. Propofol blood concentrations in this study at the time the eyelash reflex was lost (2.9+/-0.7 micro gram/ml) are similar to those reported previously (2.8+/-0.6 micro gram/ml)[38]and also are similar to concentrations when loss of responsiveness occurs during slow infusions. [24]Despite the slow changes in propofol blood concentration during induction and recovery, blood-to-effect site hysteresis still was observed. Elimination of hysteresis in blood concentration between loss and return of the eyelash reflex demonstrates the utility of estimating effect-site concentration.

The same indicators of anesthetic effect used during induction often are used during recovery. Return of the eyelash reflex in our volunteers corresponded to return of a responsive state. Volunteers were oriented to time at greater propofol concentrations than reported previously, [39,40]although administration of benzodiazepines and opioids in previous studies may account for this difference. Our data support the view that simple tests of cognitive function do not necessarily predict complete recovery from anesthesia.

No previous studies have evaluated the ability of so many potential indicators to predict movement during propofol/nitrous oxide anesthesia. In addition to determining propofol concentrations, we report EEG, pupillary, and hemodynamic data. The prediction probability (P^K) statistic allows us to directly compare these indicators, despite their different units of measurement. For the indicators we studied, Smith et al. [19]have shown that, after nonlinear scaling to improve curve fit, the relative values of a logistic curve goodness-of-fit measure correlated closely with the relative values of P^K. An advantage of P^Kis that it does not require a search for nonlinear scaling. We anticipate that this technique will have numerous applications in anesthesia research.

In summary, we conclude that indices of pharmacodynamic effect, such as the EEG, pupillary light reflex, and systolic arterial blood pressure, predict movement as well as effect-site concentration during propofol/nitrous oxide anesthesia. Indices of pharmacodynamic effect are easier to measure and results are immediately available. These data suggest that monitoring such indicators during anesthesia may facilitate prevention of movement in nonparalyzed patients and prevention of awareness in paralyzed patients.

The authors thank Steven L. Shafer, M.D., for advice on k^e0calculation; Andrew Bjorksten, Ph.D., for analysis of the propofol samples; Charles Hackman, F.A.N.Z.C.A. (assisted by an Australian and New Zealand College of Anaesthetists Research Project Grant), for programming the computer-controlled infusion; Joseph McGuire, B.S., for technical assistance; and Winifred von Ehrenburg, M.A., for editorial advice. They also thank Ohmeda, Inc., for loan of a Modulus CD integrated anesthesia machine and infusion pump; Mallinckrodt Anesthesia Products, Inc., for donation of thermocouples; Hermes Systems, Inc., for loan of an IdaCare automatic record-keeping system; and Augustine Medical, for donation of the forced-air warmer and disposable covers.

*Dutton RC, Smith WD, Smith NT: EEG prediction of arousal during anesthesia, with combinations of isoflurane, fentanyl and N²O (abstract). ANESTHESIOLOGY 1991;75:A448.

**Sebel P, Bowles S, Saini V, Chamoun N: Accuracy of EEG in predicting movement at incision during isoflurane anesthesia (abstract). ANESTHESIOLOGY 1991;75:A446.

***Vernon J, Bowles S, Sebel P, Chamoun M: EEG bispectrum predicts movement at incision during isoflurane or propofol anesthesia (abstract). ANESTHESIOLOGY 1992;77:A502.

****Schuttler J, Schwilden H, Stoeckel H: Pharmacokinetic-dynamic modeling of Diprivan (abstract). ANESTHESIOLOGY 1986; 65:A549.

*****Mahla M, Grundy B, Schmidit R, Richards R: Propofol does not cause epileptiform electroencephalographic activity. ANESTHESIOLOGY 1991; 75:A184.