Electroencephalogram Bispectral Analysis Predicts the Depth of Midazolam-induced Sedation

Twenty-six (ASA physical status 1 and 2) men scheduled for elective surgery under regional anesthesia were studied according to a protocol approved by the institutional review board. Written informed consent was obtained from all the study patients. Patients with known neurologic disease, as well as significant cardiovascular, respiratory, and hepatic disease, were excluded from participating in this study.

Patients ranged in age from 22 to 70 yr (52+/-15 yr; mean+/-SD) and in weight from 65 to 120 kg (86+/-14 kg). All patients fasted for at least 6 h before surgery and received no preoperative medication. Routine monitors included electrocardiogram, noninvasive automatic blood pressure cuff, and a pulse oximeter. In addition, gold cup EEG electrodes (Grass Instruments, Quinly, MA) were placed on the scalp with conductive gel in the following configuration: FP sub 1 -C^zand FP²-C^z(International 10–20 System of electrode placement). A ground electrode was placed behind the right ear lobe, and the electrodes were secured with collodion. The impedance of each electrode was less than 5K Omega. To minimize artifacts, patients were instructed not to open their eyes, talk, or move during the EEG recording intervals before making the OAA/S assessments. The EEG waveforms were recorded at 128 Hz using the EEG monitor (model B500, Aspect Medical Systems, Natick, MA). The quantitative EEG parameters were calculated at 4-s epochs and recorded at 15-s intervals. During the surgical procedure, all patients received supplemental oxygen via nasal cannula at a flow rate of 3–4 l *symbol* min sup -1.

After placement of the monitoring devices, a regional anesthetic block (i.e., spinal, epidural, or axillary) was performed depending on the operative requirements. After achieving adequate surgical analgesia, 1 mg intravenous midazolam was administered. Thereafter, midazolam was titrated in 0.5–1-mg increments at 6–10-min intervals to achieve an OAA/S sedation level of 1. The total doses of midazolam ranged from 4.5 to 20 mg, with a mean value (+/-SD) of 9.1+/-4.3 mg. After attaining an OAA/S score of 1, no additional sedative medication was administered, and recovery from sedation was observed until the OAA/S score returned to 4. During the onset and recovery phases, sedation was assessed at 6–10-min intervals using the OAA/S scale. Assessments of sedation (OAA/S) were performed by one investigator (J.L.) to minimize interobserver variability.

The EEG parameters analyzed included the total power in the delta (< 4 Hz), theta (4–8 Hz), alpha (8–13 Hz), and beta (14–40 Hz) frequency bands; MF (i.e., the frequency at which half the power was at higher frequencies and half at lower frequencies); and 95% SEF (i.e., the frequency at which 95% of the total power was at lower frequencies); and the BI. A detailed description of the fundamentals of bispectral analysis has been reported in previously published studies. [17,18]All the patients remained well oxygenated (Sp^O2> 95%) and normotensive (MAP > 80% of the baseline values) throughout the study period.

In this study, the BI values reported are the average of real-time bispectral indices obtained from channels 1 and 2 (FP¹-C sub z and FP²-C^z). The values of EEG parameters at each OAA/S score were calculated by averaging all the values during the 45-s interval immediately before assessment of the sedation level. Mean values and SD were calculated for all EEG parameters at each OAA/S score. The mean EEG values between OAA/S scores 3 and 2 and between OAA/S scores 2 and 1 during the onset and recovery phases from midazolam-induced sedation were defined as the EEG⁵⁰values for response to verbal command (EEG⁵⁰-VC) and to shaking of the head (EEG⁵⁰-SH), respectively. Coefficients of variation for EEG⁵⁰-VC and EEG⁵⁰-SH were calculated by dividing the SD by the mean EEG value and multiplying by 100.

Kruskal-Wallis analysis of variance was used to determine significant changes in each EEG parameter during the onset of and recovery from sedation. The nonparametric Spearman's rank-correlation analysis was performed to evaluate the relationship between OAA/S scores and EEG-BI or 95% SEF. The EEG-BI-quantal response relationships were plotted as the EEG-BI and the percentage of patients responding to a verbal command or to shaking of their head. Differences with a P value < 0.05 were considered statistically significant.

With progressively increasing sedation from an OAA/S score of 5 to 1, the EEG-BI values decreased from 95.4+/-2.3 to 69.2+/- 13.9 (Figure 1, Table 1). Similarly, the 95% SEF values decreased from 20.1+/-4.5 to 15.7+/-3.8 as the level of sedation progressed from an OAA/S score of 4 to 1 (Figure 2, Table 1. During recovery, the EEG-BI increased from 69.2+/-13.9 to 90.8 +/-6 as the OAA/S score increased from 1 to 4 (Figure 1, Table 2). In contrast, the 95% SEF did not increase significantly as OAA/S level progressed from 1 to 4 (Figure 2, Table 2). The EEG-BI correlated closely with the OAA/S scores during both the onset (Spearman's Rho = 0.815; P < 0.0001) and recovery (Spearman's Rho = 0.596; P < 0.0001) phases (Figure 3). The 95% SEF was closely correlated only during the onset phase of sedation as OAA/S progressed from 4 to 1 (Spearman's Rho = 0.459; P < 0.0001)

With the decrease in EEG-BI, the cumulative percentage of patients responding to verbal commands decreased. The EEG-BI versus cumulative percentage response to shaking of the patient's head followed a similar pattern but was shifted to the left of the EEG-BI responses to verbal commands (Figure 4). These data suggest that a greater depth of sedation (i.e., lower EEG-BI values) is required to prevent a response to a physical stimulus (i.e., shaking the patient) than to a verbal stimulus. The coefficients of variation of BI for EEG⁵⁰-VC (10.2%) and EEG⁵⁰-SH (19.8%) were less than with the other EEG parameters studied (Table 3).

No consistent correlations were found for the delta or theta power bands or MF values and OAA/S score during the onset of or recovery from sedation. The total power in the alpha band increased as the OAA/S score changed from 5 to 2 (OAA/S score: alpha power = 5:9.8+/-8; 4:14.3+/-11; 3:24.2+/-11.7; 2:28.3+/-11.8). In contrast, beta power decreased as the OAA/S score decreased from 4 to 1 (OAA/S score: beta power = 4:26.1+/-12.5; 3:25.4+/-12.6; 2:18.1+/-7.6; 1:13.2+/-9.9). During the recovery phase, the alpha or beta power did not change significantly as OAA/S score increased from 1 to 4.

This study demonstrates that the EEG-BI correlates with midazolam-induced sedation as assessed by the OAA/S rating scale. With increases in sedation, there were predictable decreases in the EEG-BI. Similarly, 95% SEF decreased as OAA/S score decreased from 4 to 1. During the recovery phase, decreasing sedation was associated with increasing EEG-BI values. In addition, the small coefficients of variation of EEG⁵⁰-VC and EEG⁵⁰-SH for EEG-BI suggest that the EEG-BI may provide a better correlation with sedation than the other EEG parameters studied, including the 95% SEF. Therefore, of the indexes we evaluated, the EEG-BI appears to be the most accurate for quantifying midazolam-induced sedation during regional anesthesia.

In this study, alpha power increased as the OAA/S score decreased from 5 to 2, whereas beta power decreased as the OAA/S score decreased from 4 to 1. It has been reported that benzodiazepine-induced sedation increases high-frequency beta power. [19,20]However, use of beta power as an effect parameter requires constant verbal stimulation and a responsive subject. Greenblatt et al. found that the increased EEG beta power band was lost when the subject slept, decreasing the usefulness of beta power band for studying the effects of larger doses of midazolam. [21]In a study involving dogs, Fleischer et al. demonstrated that the infusion of midazolam produced dose-related EEG changes with high-frequency alpha and beta power changing to low-frequency delta and theta power. [9]In our clinical study, we did not observe an increase in the delta or theta power after administration of 9.1+/-4.3 mg intravenous midazolam (mean+/-SD). This apparent disparity in results could be due to species differences and to differing definitions of the power bands, variable settings for the frequency filters, and subtle low-frequency motion artifacts induced by eye movements. The EEG power bands, therefore, would not appear to be good indicators for assessing benzodiazepine-induced sedation.

The 95% SEF has been shown to predict hemodynamic responses to laryngoscopy and intubation. [8]In the current study, a good correlation existed between 95% SEF and clinical responses at deeper sedation during onset phase (OAA/S scores from 4 to 1). These results confirm the preliminary findings of Billard et al. during intravenous anesthesia. [20]Buhrer et al. demonstrated that the initial transient increase in 95% SEF correlated with the midazolam-induced power shift from alpha to beta bands. [22]However, with deeper sedation, the power increase in the delta band moved the 95% SEF toward lower frequencies. [22]The coefficients of variation for both EEG⁵⁰-VC and EEG⁵⁰-SH were less with the 95% SEF compared to the power bands or MF. These data suggest that the 95% SEF is superior to the use of the EEG power bands and MF in assessing midazolam-induced sedation. However, 95% SEF is a less accurate indicator of the degree of midazolam-induced sedation than the EEG-BI during both the onset and recovery phases.

When using sedative medication as part of a monitored anesthesia care technique, the anesthesiologist attempts to titrate the drug to optimize patient comfort while maintaining cardiorespiratory stability and intact protective reflexes. Based on the EEG-BI response curves, we recommend a EEG-BI value greater than 80 to achieve and maintain an OAA/S score of 3 or greater. Therefore, anesthesiologists may be able to improve the accuracy of titration of midazolam and other sedative-hypnotic drugs (e.g., propofol) during local and regional anesthesia by using the EEG bispectral monitor as an adjunct to their routine clinical assessment. The major limitations in using this EEG device to monitor sedation relate to the time required to place the monitor leads and the expense of the monitoring device. Future studies need to examine the relationship between OAA/S scores and EEG parameters with other centrally active sedative and analgesic drugs in the operating room and intensive care settings.

In conclusion, the EEG-BI appears to accurately predict the OAA/S score in patients receiving midazolam for sedation during regional anesthesia. Of the EEG variables evaluated in this study, the EEG-BI appeared to be the most accurate for assessing the depth of sedation.

The authors thank the postanesthesia care unit nurses and the residents in the Department of Anesthesiology and Pain Management, for their cooperation during this study. They also thank Nassib Chamoun, Paul Manberg, Ph.D., and Pat Embree, C.R.N.A., Aspect Medical Systems, Inc.