A Crossover Comparison of the Sensitivity and the Specificity between BIS and AEP in Predicting Unconsciousness in General Anesthesia

Background. /ere is an increasing concern of awareness and recall during general anesthesia for both the patient and the anesthetist. /e bispectral index (BIS) is used to assess the level of sedation and depth of anesthesia and detect consciousness in different anesthetic drugs. Middle-latency auditory evoked potentials (AEPs) also quantify action of anesthetic drugs and detect the transition from consciousness to unconsciousness. We aim to compare the sensitivity and specificity between BIS and AEP in predicting unconsciousness in inhalational sevoflurane anesthesia and intravenous propofol anesthesia. Methods. Totally, 40 patients were randomly allocated into two groups: propofol or sevoflurane group. In the propofol group, anesthesia was induced with target-controlled infusion propofol. In the sevoflurane group, anesthesia was induced by increasing concentrations of sevoflurane. /ere were 3 end points during induction: sedation, unconsciousness, and anesthesia. Target and effect-site concentrations of propofol, end-tidal concentration of sevoflurane, and BIS and AEP were recorded at each stage. Results. We obtained good EC50 with both monitors, at which there is a 50% chance that the patient has reached the end point, but the index variation was affected by the anesthetic technique. Propofol had higher correlations with stage of anesthesia, BIS, and AEP than sevoflurane. BIS had higher correlations with depth of anesthesia than AEP, but we did not find an anesthetic depth monitor that had high sensitivity and specificity and is not affected by the anesthetic technique. Conclusions. /e prediction powers of BIS and AEP do not seem as good as some papers mentioned.


Introduction
Awareness and recall during general anesthesia, which are unintended accidental, represent failure of successful anesthesia and cause a serious complication of general anesthesia that is feared by patients and anesthetists alike [1][2][3]. It is difficult to describe and identify return to consciousness, so the reported incidence rates vary widely. Evidence suggests that the overall risk of awareness during anesthesia is between 0.1 and 0.5% [2,[4][5][6], and awareness has been considered as a potentially important factor for the occurrence of some diseases in patients, such as severe emotional distress and posttraumatic stress disorder [4,[6][7][8]. It also has important professional, personal, and financial consequences for the anesthetists [8][9][10][11]. e bispectral index (BIS), derived from electroencephalogram, is the most commonly used and accepted monitor for assessing the level of sedation and depth of anesthesia [10,[12][13][14][15]. BIS predicts movement in response to surgery and detects consciousness under different anesthetic drugs [15][16][17][18]. It is also a tool that may reduce the incidence of unexpected recall [10,12,13,18].
e AEP index (AEP) is a dimensionless number scaled from 100 (awake) to 0 and a mathematically derived variable measuring the amplitude and latency of the cortical midlatency auditory evoked potential that occurs in response to sound (a "click") [21,23,24]. Sevoflurane inhalational and propofol intravenous anesthesia are two widely used anesthetic techniques. However, there are no reports about the comparison of ability of predicting the awareness by BIS or AEP during these two anesthesia techniques. Herein, this study is designed to compare the sensitivity and the specificity between BIS and AEP in predicting unconsciousness with sevoflurane inhalational and propofol intravenous anesthesia.

Materials and Methods
e study was approved by the institutional review board. Unpremedicated patients who had given informed consent were recruited into the study. Demographic data and ASA classification were recorded. Routine monitoring plus monitoring for BIS and AEPi was established before the induction of anesthesia. Awake values for BIS and AEPi were recorded before the induction of anesthesia. Patients breathed oxygen through a standard anesthetic breathing circuit during induction. Patients were randomised into propofol or sevoflurane groups. ere were 3 end points during induction: (1) Sedation: patent was asleep and responded to gentle shaking or loud auditory stimulus (stage 4 of Ramsay scale). (2) Loss of consciousness: patient showed no response to verbal command and loss of eyelash reflex. (3) Anesthesia: patient gave no purposeful movement on tetanic stimulation to the ulnar nerve (50 Hz, 80 mA, 0.25 ms pulses) at the wrist using a constant current peripheral nerve stimulator.
e BIS and AEPi were recorded at each stage. In the propofol group, anesthesia was induced with target-controlled infusion (TCI) propofol. e TCI was initially set at 1 µg·l −1 and increased by 0.5 µg·l −1 every 2 minutes until anesthesia. Target and effect-site concentrations of propofol were recorded at each end point. In the sevoflurane group, anesthesia was induced by increasing concentrations of sevoflurane. End-tidal concentration of sevoflurane was recorded at each end point.

Anesthesia Induction.
Routine monitoring plus monitoring for BIS and AEP was established before the induction of anesthesia. Awake values for BIS and AEP were recorded before the induction of anesthesia. Patients breathed oxygen through a standard anesthetic breathing circuit during induction. Patients were randomised into propofol or sevoflurane groups. ere were 3 end points during induction: (1) sedation: patient was asleep and responded to gentle shaking or loud auditory stimulus (stage 4 of Ramsay scale); (2) unconsciousness: patient showed no response to verbal command and loss of eyelash reflex; (3) anesthesia: patient gave no purposeful movement on tetanic stimulation to the ulnar nerve (50 Hz, 80 mA, 0.25 ms pulses) at the wrist using a constant current peripheral nerve stimulator. e BIS and AEP were recorded at each stage. In the propofol group, anesthesia was induced with target-controlled infusion (TCI) of propofol. e TCI was initially set at 1 µg·l −1 and increased by 0.5 µg·l −1 every 2 minutes until anesthesia. Target and effect-site concentrations of propofol were recorded at each end point. In the sevoflurane group, anesthesia was induced by increasing concentrations of sevoflurane. End-tidal concentration of sevoflurane was recorded at each end point.

Statistical
Analysis. GraphPad Prism version 5 (GraphPad Software, Inc) was used for data analysis. Demographic data were analyzed by the chi-square test and t-test. Haemodynamic data were analyzed by repeated measures analysis of variance and post hoc pair-wise comparison for difference stages of anesthesia. Spearman correlation analysis, logistic regression analysis, receiver operating characteristic (ROC) analysis, sensitivity and specificity, and prediction probability (P K ) were used for analyzing the depth of anesthesia, drug concentration, BIS, and AEP. P < 0.05 was considered to have statistically significant difference.

Patient Characteristics.
Forty-two patients were assessable for intraoperative BIS and AEP data, including 22 patients with sevoflurane anesthesia and 20 patients with propofol anesthesia. Two patients of the sevoflurane group were censored because of the unreasonable high BIS and AEP in the anesthesia stage. One patient of the sevoflurane group swapped the effect-site concentrations on sedation and unconsciousness because of the unreasonable high concentrations (4 and 4.6 mcg/ml) in the sedation stage. No significant difference in gender, height, weight, smoking history, alcohol intake, pain in sedation stage, or American society of Anesthesiologists status was found between two groups (Table 1).

Haemodynamic Data.
Systolic blood pressure (SBP), heart rate (HR), and respiratory data (RR) in two groups were analyzed in four stages: base, sedation, unconsciousness, and anesthesia (Tables 2 and 3 and Figure 1).For SBP, time effect was significantly different at the 0.05 level of significance. Time and group interaction effect was significantly different at the 0.01 level of significance. On the propofol group, the SBP on all the other stages was significantly different from the baseline (P � 0.0003, <0.0001, and <0.0001 in sedation, unconsciousness, and anesthesia stages, respectively). e SBP on the sedation stage was also significantly different from the SBP on the unconsciousness and anesthesia stages. On the sevoflurane group, the SBP on all the other stages was significantly different from the baseline (P � 0.0354, 0.0053, and 0.0031 in sedation, unconsciousness, and anesthesia stages, respectively). e SBP on the sedation stage was significantly different from the SBP on unconsciousness and anesthesia stages.
On HR, both time effect and group interaction effects were significantly different at the 0.05 level of significance. On the propofol group, the heart rates on all the other stages were significantly different from the baseline heart rate (P � 0.0266, 0.0034, and 0.0188 in sedation, unconsciousness, and anesthesia stages, respectively). e heart rate on the sedation stage was also significantly different from the heart rate on the unconsciousness stage (P � 0.0133). On the sevoflurane group, there were not significantly different in all different stages.

Descriptive Statistics of Drug Concentrations, BIS, and
AEP at Different Stages of Induction. redicted blood and effect-site propofol concentrations and inspired and endtidal sevoflurane concentrations during different stages of induction are shown in Table 4. Both BIS and AEP showed a trend of diminishing level of consciousness with both anesthetic techniques.

Correlation Analysis.
On correlation analysis of BIS and AEP vs. propofol and sevoflurane concentration, only 6 correlation coefficients in the propofol group were significant at the 0.05 level of significance. r � −0.50, −0.49, and −0.45 when BIS is in the unconsciousness stage vs. predicted blood concentration of propofol in sedation, anesthesia stages, and effect-site concentration of propofol in the anesthesia stage, respectively. r � 0.56 when AEP is in the anesthesia stage vs. predicted blood concentration and effect-site concentration of propofol in the unconsciousness stages, and r � −0.53 when AEP of baseline vs. effect-site concentration of propofol is in the sedation stage. ey were around 0.5, just fair correlated. All the others are not significantly correlated (Tables 5 and 6). On correlation analysis     (Tables 7 and 8).
Both monitors showed a trend of diminishing level of consciousness with both anesthetic techniques. e index showed good correlation with stage of induction (Table 9), except with AEP when used with sevoflurane which gave a low correlation coefficient of −0.61 and a 95% CI crossing −0.5. Results showed that propofol had higher correlations between stage of anesthesia and BIS and AEP than sevoflurane. BIS had higher correlations with depth of anesthesia than AEP (given the prior probabilities for the sedation/ unconsciousness/anesthesia to be 0.5).
Effective concentrations EC5, EC50, and EC95 referred to drug concentration at which 5%, 50%, and 95% of the patients, respectively, reached the predefined end point. EC5, EC50, and EC95 of predicted blood and effect-site propofol and inspired and end-tidal sevoflurane as well as BIS and AEP values at sedation, unconsciousness, and anesthesia and their sensitivity, specificity, and P K values are shown in Tables 10-12, respectively. Effect-site propofol concentration had a smaller 95% CI of EC50 than that of blood at all stages from sedation to anesthesia. is difference is not noticed between inspired and end-tidal sevoflurane. BIS gave similar 95% CI of EC50 with propofol and sevoflurane at sedation and unconsciousness, but a range of smaller values with sevoflurane at anesthesia. For AEP, propofol always showed a range of smaller values from sedation to anesthesia. P K is the probability that the indicator values of the data points predict correctly which of the data points are the lighter (or deeper). A value of P K � 0.5 means that the indicator correctly predicts the anesthetic depths only 50% of the time, i.e., no better than a 50:50 chance. A value of P K � 1 means that the indicator predicts the anesthetic depths correctly 100% of the time. Table 13 shows the P K for depth of anesthesia measured by the two monitors, BIS and AEP with different anesthetic techniques, propofol and sevoflurane. BIS had good P K values with both propofol and sevoflurane with both above 0.8, and the P K with propofol was better than that with sevoflurane. On the contrary, AEP did not have good P K values, especially with sevoflurane. e P K values with BIS are significantly better than those with sevoflurane. e prediction powers of BIS and AEP do not seem as good as some papers mentioned (P K > 0.9).

Discussion
In this study, we investigated the usefulness and consistency of two anesthetic depth monitors, BIS and AEP with different anesthetic techniques, propofol intravenous anesthesia and sevoflurane inhalational anesthesia. BIS and AEP are two popular anesthetic depth monitors. It is important for them to perform consistently with different anesthetic techniques.
EC50 is analogous to the concept of minimum alveolar concentration for volatile anesthetics and is defined as the concentration of an i.v. anesthetic agent at which 50% of the     Scientific Programming patients will not move or respond to skin incision. is clinically useful concept allows prediction of propofol concentration in the blood and at the effect site [25,26].
We defined the anesthesia stage as when the patient showed no gross purposeful movement to tetanic stimulation of the ulnar nerve, which was easy to perform and had the advantage over skin incision as a repeatable stimulus. A study showed no significant difference between the effective concentration of propofol which prevented half of the patients to move (EC 50 ) at tetanic stimulation and that at skin incision in somatic response, but significant differences in haemodynamic response [25,26]. Tetanic stimulation was useful in this study as a reproducible and repeatable stimulus at different propofol and sevoflurane concentrations. Similar to the results from Milne's group, the range of effect-site concentrations to include 90% of patients (EC5-EC95) was smaller than the predicted blood concentration range and hence a more useful figure to guide propofol administration [27]. Similarly, in the sevoflurane group, the range of endtidal concentrations was smaller than the inspired, but to a lesser extent. Both monitors had distinctly different EC50s with small 95% CI. BIS had similar EC50s with both propofol and sevoflurane, but AEP showed different values between the two anesthetic techniques.
In this study, 90% of the patients were sedated at a BIS value between 90 and 71 with propofol or between 100 and 60 with sevoflurane.
is indicates that BIS is therefore better at predicting sedation with propofol. AEP showed very wide range of values in order to induce 90% of the patients at sedation with both propofol (AEP value range 100-11) and sevoflurane (AEP value range 96-27), and therefore, AEP did not seem to be useful in guiding sedation. At unconsciousness, BIS showed a smaller range with propofol (BIS value range 77-37) than with sevoflurane (BIS value range 93-23), which might again indicate that BIS performs better with propofol. AEP showed a wide range with both propofol (AEP value range 61-4) and sevoflurane (AEP value range 85-12) at unconsciousness. At anesthesia, BIS again had a smaller range with propofol (BIS value range 61-31) than with sevoflurane (BIS value range 67-11). AEP showed a narrow range with propofol (AEP value range 28-6) but a wide one with sevoflurane (AEP value range 75-0). BIS appeared to be a good indicator of depth of anesthesia with propofol, which was reflected by the high P K value of 0.91. Anesthetic seemed to have an effect on performance of the monitors, particularly with AEP monitor. BIS overall performed well with both anesthetic techniques, i.v. propofol and inhalational sevoflurane, but with a higher P K with propofol. AEP showed poorer performance than BIS in our study. With a P K of 0.56 with sevoflurane, AEP became doubtful as an anesthetic depth monitor which means the prediction powers of BIS and AEP do not seem as good as some papers mentioned [21,22,28,29]. Considering the difference results between this study and previous ones, different protocols of studies might be the reason [22,[28][29][30]. We use detected more drug concentrations at more time points with more accurate statistical methods, but we still think we need more studies to verify the results. And this result might remind the clinicians that both BIS and AEP are not as reliable as they thought.
In summary, we obtained good EC50 with both monitors, but the index variation was affected by the anesthetic technique. e performance of the anesthetic depth monitors was better when propofol was used. Very wide variation was found in the combination of AEP and sevoflurane [22,25,[31][32][33]. It seems the monitors are at best at giving the EC50, at which there is a 50% chance that the patient has reached the end point, and we have not yet found an anesthetic depth monitor that has high sensitivity and specificity and not affected by the anesthetic technique.

Data Availability
All the underlying data supporting the results of this study can be found in IRB of Second Affiliated Hospital of Dalian Medical University.

Disclosure
Haitao Yang and Guan Wang are the co-first authors.

Conflicts of Interest
e authors declare that there are no conflicts of interest.

Authors' Contributions
Haitao Yang and Guan Wang contributed equally to this work.