Anesth Analg 2005;100:141-148
© 2005 International Anesthesia Research Society
doi: 10.1213/01.ANE.0000138057.61904.FD
TECHNOLOGY, COMPUTING, AND SIMULATION
Steven J. Barker Section Editor
The Association Between Propofol-Induced Loss of Consciousness and the SNAPTM Index
Cynthia A. Wong, MD,
Robert J. Fragen, MD,
Paul C. Fitzgerald, RN, MS, and
Robert J. McCarthy, PharmD
Department of Anesthesiology, Northwestern University Feinberg School of Medicine, Chicago, Illinois
Address correspondence to Cynthia A. Wong, MD, Department of Anesthesiology, Northwestern University Feinberg School of Medicine, 251 E. Huron St., F 5-704, Chicago, IL 60611. Address e-mail to c-wong2{at}northwestern.edu Reprints will not be available from the authors.
 |
Abstract
|
|---|
The SNAPTM is a processed electroencephalogram monitor that uses an algorithm based on low- and high-frequency spectral components to derive a SNAPTM index. In this study we sought to determine the relationship of the SNAPTM index with loss of consciousness in subjects receiving a bolus of propofol. Unpremedicated subjects were randomized to receive 1 of 11 doses of IV propofol (0, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, or 2.4 mg/kg; n = 20 per group). The SNAPTM index was recorded when the subject became unconscious (end-point) or at 160 s after the injection. Sixty-five percent of subjects achieved the end-point (defined as the time at which the subject dropped a weighted syringe). The 50% effective dose for propofol was 0.97 mg/kg (95% confidence interval [CI], 0.86–1.07 mg/kg). The median awake SNAPTM index was 92 (range 78–99) and did not differ between subjects who reached the end-point and those who did not. The end-point SNAPTM index decreased from baseline in the subjects who dropped the syringe to a median of 76 (range, 57–94) at doses 
1.0 mg/kg but was not different among doses. The index was not different from baseline at 160 s in subjects who did not reach the end-point. Binary logistic regression models predicted a SNAPTM index 95% effective dose for loss of consciousness of 71 (95% CI, 63–74) and 19 (95% CI, 16–22) for changes in SNAPTM index from baseline. The areas under the receiver operator characteristic curves for these models were 0.837 and 0.864. The SNAPTM index correlated with propofol-induced loss of consciousness. It appears to be a useful indicator of loss of consciousness and should be further investigated as a monitor of anesthesia depth.
IMPLICATIONS: The SNAPTM is a processed electroencephalogram monitor that uses an algorithm based on low- and high-frequency spectral components to derive the SNAPTM index. The SNAPTM index correlated with propofol-induced loss of consciousness.
|
Introduction
|
|---|
Several monitoring modalities have been investigated to assess depth of anesthesia or level of hypnosis during anesthesia [e.g., bispectral electroencephalogram (EEG) analysis (1) and auditory and somatosensory evoked potential]. The SNAPTM monitor (Viasys Healthcare, Inc., Conshohocken, PA; Everest Biomedical Instruments, Inc., St. Louis, MO) derives a dimensionless number from 0 to 100 (with 100 representing the awake state) from a single-channel processed EEG signal. A description of the SNAPTM monitor is presented in Appendix 1.
The SNAPTM index is derived from an algorithm based on low-frequency (0.1–40 Hz) and high-frequency (80–420 Hz) EEG components. Similar to other monitors of depth of anesthesia, the low-frequency component of the SNAPTM index is determined by spectral analysis of EEG signals in the 0.1- to 40-Hz range (
, ß, and 
frequencies). In contrast to other monitors, spectral frequencies from 40 to 80 Hz are ignored. This range is predominated by muscle activity and other environmental noise signals, but it also contains 
frequency components that have been shown to be modified by anesthetics (2,3). The SNAPTM algorithm, however, incorporates high-frequency (80–420 Hz) EEG components that contain 150- to 200-Hz oscillations that represent direct electrotonic coupling of neurons in the mammalian cortex and hippocampus (4). Axonal networks appear to underlie the in vivo ripple (the approximately 200-Hz field potential oscillations) and to drive activity (5,6). Anesthetics may directly or indirectly influence these fast potentials, and, therefore, analysis of this component of the EEG may contribute to the ability to differentiate between the awake and anesthetized state.
We hypothesized that the SNAPTM index is associated with the change in level of consciousness induced by a propofol bolus. The purpose of this study was to validate the SNAPTM index as a measure of loss of consciousness induced by a propofol bolus in unpremedicated adults.
|
Methods
|
|---|
This study was approved by the Northwestern University IRB. ASA physical status I or II patients scheduled for elective surgical procedures under general anesthesia were eligible to participate. Exclusion criteria included chronic therapy with sedative/hypnotic, antidepressant, or centrally acting antihypertensive drugs; pregnancy at the time of study; history of alcohol or drug abuse; allergy to propofol; body mass index (BMI) <18 kg/m2 or BMI >30 kg/m2; or age <18 or >55 yr.
After we obtained written, informed consent, an IV catheter was placed in a large forearm vein. No premedication was administered; the subject was brought to the operating room, and standard monitors were placed. A SNAPTM electrode montage was placed on the subject’s forehead (Appendix 1), low impedance (<10 k
) was verified, and the SNAPTM monitor signal averaging time was set to zero. The subjects were then given a weighted 20-mL syringe to hold in the hand opposite the arm with the IV and were instructed not to drop it. This was followed by a 3- to 5-min period to allow for patient relaxation and stabilization of the EEG signal. A baseline SNAPTM index was recorded as the average value during a 1-min period immediately preceding study drug injection.
Group assignments for all subjects were made by using a computer-generated random number sequence. After consent was obtained, subjects were randomized to 1 of 11 groups to receive propofol by IV bolus (0, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, or 2.4 mg/kg; n = 20 per group). The propofol syringe was prepared by an anesthesiologist investigator. The syringe was diluted with normal saline to a final volume of 20 mL and covered to blind both the study subject and the observer-investigator to the propofol dose. The propofol was injected over 15 s in a rapidly running IV infusion at a stopcock placed between the IV catheter and IV tubing.
The SNAPTM index was recorded continuously until a nadir was reached or for 160 s, whichever was longer. The observer-investigator recorded the SNAPTM index when the subject dropped the syringe (end-point) or at 160 s after the injection if the subject did not drop the syringe (end-point SNAPTM value). Loss of consciousness was defined as occurring when the subject dropped the syringe (7). To verify that syringe drop was actually associated with loss of consciousness, the degree of sedation beginning at the time of injection, every 20 s thereafter, and at the end-point (just after syringe dropped or 160 s) was assessed by using the Observer Assessment of Alertness/Sedation scale (8). The study ended after 160 s or after the SNAPTM index nadir was reached if this occurred after 160 s. The anesthesiologist then proceeded to induce general anesthesia if the subject was still awake.
The sample size determined for this study was calculated by using the following assumptions: 1) approximately 50% of the patients would reach the end-point (drop the syringe), 2) the variation in the probability of the event would change by 10% with a 1 SD change in the SNAPTM index at this end-point, and 3) power was set at 90% and to 0.05. These assumptions resulted in 223 patients with a binary logistic regression model. Calculations were made with PASS 2002 (NCSS Statistical Software, Kaysville, UT). The randomization was performed in 2 blocks of 110.
The 50% effective dose for propofol was determined by using probit analysis of the log10 of the propofol doses. Baseline SNAPTM index values were compared between subjects who reached the end-point versus those who did not by using the Mann-Whitney U-test. Baseline SNAPTM index values and time from injection to end-point were compared at each propofol dose by using the Kruskal-Wallis H-test. The difference in the SNAPTM index at end-point (or 160 s) from the baseline overall, as well as at each propofol dose, was analyzed with Wilcoxon’s signed rank test. Tests were adjusted for multiple comparisons by using the Bonferroni method. P < 0.05 was required to reject the null hypothesis.
Binary logistic regression analysis were performed to construct predictive models of the probability of reaching the end-point based on both the absolute SNAPTM index at end-point and the difference in the end-point SNAPTM index from the baseline index. The logistic model was internally validated with a 10% cross-validation method. The predictive power of the models was determined with receiver operating characteristic (ROC) curves by plotting the sensitivity (subjects who reached the end-point correctly predicted by the model divided by the total number predicted to reach the end-point) against 1 – specificity (subjects who did not reach the end-point correctly predicted by the model divided by the total number predicted not to reach the end-point) and calculating the area under the curve (AUC). Brier scores were determined for the training and test samples (9,10). The Brier score measures the difference between a forecast probability of an event and its occurrence, expressed as 0 or 1, depending on whether the event has occurred or not. Scores range from 0 to 1, with lower scores signifying greater concordance with the model. The bias of the internal validation method was calculated as the average difference between the test sample and the training sample. Accuracy was assessed by using the root mean squared error of the internal validation samples.
The relationship between the SNAPTM nadir and the propofol dose was examined by fitting the baseline SNAPTM index and SNAPTM index nadir (by using the Marquardt algorithm and nonlinear regression) to a modified Hill equation. Statistical analysis was performed with NCSS 2001 (NCSS Statistical Software, Kaysville, UT).
|
Results
|
|---|
We evaluated 220 patients scheduled for operative procedures under general anesthesia. The mean age was 38 ± 9 yr. There were no differences in sex distribution (male/female, 101:119) or BMI (mean, 24 ± 3 kg/m2) among propofol dosage groups.
The propofol dose-response relationship for all subjects is shown in Figure 1. One-hundred-forty-two subjects (65%) reached the end-point (dropped the syringe). The 50% effective dose of propofol was 0.97 mg/kg (95% confidence interval [CI], 0.86 to 1.07 mg/kg).
View larger version (16K):
[in this window]
[in a new window]
|
Figure 1. Probability of achieving end-point (syringe drop) versus propofol dose. Open triangles are observed frequencies. The solid line is predicted by the model, and dashed lines are the 95% confidence intervals for the model.
|
|
The median baseline SNAPTM index was not different between subjects who dropped the syringe and those who did not (Fig. 2). The end-point SNAPTM index was different from baseline in subjects who dropped the syringe (median, 77), but not in subjects who did not drop the syringe (median, 92). Compared with the awake baseline, the end-point SNAPTM index was lower in subjects who dropped the syringe. The Observer Assessment of Alertness/Sedation score was 5 at baseline in all subjects and was 2 (n = 5) or 1 (n = 137) immediately after the syringe was dropped in all subjects who dropped the syringe.
The effect of the propofol dose on the SNAPTM index is shown in Figure 3. For subjects who dropped the syringe, the end-point SNAPTM index was different from baseline for all propofol doses 1 mg/kg (Fig. 3). There were no differences in the median end-point SNAPTM indices at the syringe-drop end-point among the propofol doses. The time from injection to syringe drop was shorter for propofol doses of 2.2 and 2.4 mg/kg compared with 0.6 mg/kg (Fig. 3). For subjects who did not drop the syringe, the baseline SNAPTM index did not differ from the end-point SNAPTM index for any propofol dose.
The logistic regression model predicted that loss of consciousness (syringe drop) in 95% of subjects would occur at a mean absolute SNAPTM index of 71 (95% CI, 63–74) and at a mean SNAPTM index difference from baseline of 19 (95% CI, 16–22) (Fig. 4). The predicative performance (AUC) of the model and the average prediction error (Brier score) in the cross-validation analysis are shown in Table 1. Cross-validation demonstrated good agreement between test and training samples for both predictive performance and average predictive error. Figure 4 shows the ROC curve for the apparent models.
View larger version (13K):
[in this window]
[in a new window]
|
Figure 4. Top, Logistic regression model predicting the probability of loss of consciousness for the difference in the SNAPTM index from baseline. Bottom, Receiver operating characteristic curve for the difference in the SNAPTM index from baseline.
|
|
The SNAPTM index nadir for each propofol dose and the relationship predicted from the modified Hill equation are shown in Figure 5. The propofol dose associated with half-maximal effect was 1.27 mg/kg (95% CI, 0.49–2.04 mg/kg), with a maximum (lowest) SNAPTM index of 59 (95% CI, 28–91) and a Hill coefficient of 2.06 (95% CI, 0.061–3.49).
View larger version (19K):
[in this window]
[in a new window]
|
Figure 5. Box plot of SNAPTM index nadir for each propofol dose and the relationship predicted from the modified Hill equation:
where E0 equals the baseline SNAP value, Emax equals the maximum EEG effect produced by the propofol, IC50 equals the propofol dose associated with the half-maximal effect, and equals the Hill coefficient or steepness of curve. Variable estimates ± SE for the model were as follows: Emax = 59.2 ± 16.2, IC50 = 1.27 ± 0.39, and = 2.06 ± 0.73. The r2 of the model was 0.357.
|
|
|
Discussion
|
|---|
Monitoring the level of consciousness during sedation or the depth of anesthesia during general anesthesia is desirable. However, measuring depth of anesthesia is not a straightforward process; there are several ways of defining anesthesia and adequate depth of anesthesia (11,12). The SNAPTM is a monitor approved by the Food and Drug Administration to monitor the effects of anesthesia on the brain. It uses an algorithm to derive an index from the low-frequency (0.1–40 Hz) and high-frequency (80–420 Hz) EEG components (Appendix 1). A previous study demonstrated that the SNAPTM index varied indirectly with the perceived clinical depth of anesthesia and also with the depth of anesthesia as assessed by the bispectral index (BIS) monitor (Aspect Medical Systems, Natick, MA) (13). The monitor was withdrawn from the market by Viasys Healthcare in 2003 for strategic business reasons but has recently been acquired by Everest Biomedical Instruments.
The results of this study demonstrate that the SNAPTM index is associated with loss of consciousness induced by a propofol bolus. There was reasonable separation of SNAPTM values at loss of consciousness compared with awake values. The predictive performance of the model, as described by the area under the ROC, was similar to that determined for other depth-of-anesthesia monitors. For example, by using a ramped infusion protocol to achieve a predetermined target concentration that was maintained for 10 minutes to ensure equilibrium with the effect site, Glass et al. (12) determined that the predictive probability score for loss of consciousness with the BIS index (Version 3.0) was 0.976 ± 0.006 (mean ± SE). Chen et al. (14) determined that the AUC for prediction of loss of consciousness with a balanced anesthesia technique (with data collected during both induction and emergence) was 0.95 ± 0.04 for the PSA 4000 (Physiometrix Inc., Billerica, MA) and 0.79 ± 0.04 for the BIS monitor.
The predictive performance of the SNAPTM index in this study is especially favorable in light of the fact that the SNAPTM index was obtained by using a bolus dose of propofol to induce loss of consciousness, as compared with ramped infusions or gradual emergence from general anesthesia. Our study design was also more likely to result in large intersubject variability in effect-site concentrations of propofol compared with targeted infusions, but it is more analogous to clinical scenarios. Despite the limitations of the study design, the performance of the model in the cross-validation analysis was favorable. For both concordances of the model, as well as the average prediction error, there was low average bias. The baseline normalized data had a lower root mean squared error in the error of the bias, suggesting that this may be a more reliable variable to monitor.
The SNAPTM index at loss of consciousness was independent of the propofol dose, as was the time to reach the end-point. Subjects who remained awake did not demonstrate significant changes in the SNAPTM index, regardless of propofol dose. However, the relationship between propofol dose and the SNAPTM index nadir could be reasonably described by the modified Hill equation, which suggests that the SNAPTM index describes the level of hypnosis obtained with different doses of propofol and may be useful as a pharmacodynamic measure of hypnotic drug effect. The Hill coefficient (2.06) determined by nonlinear regression suggests that for an increase in propofol dose of 1 mg/kg, a 20-point decrease from the SNAPTM baseline will be observed in the steep portion of the dose-response curve.
A limitation of all depth-of-anesthesia monitors is that the derived index of hypnosis is a dimensionless number from 0 to 100, and the scales are not necessarily interchangeable. For example, our model predicted that 95% of subjects would lose consciousness at a SNAPTM index of 71. This number is higher than that calculated by Glass et al. (12) for the BIS monitor (BIS95 = 51). In a previous study during clinical anesthesia, we also observed an offset between the SNAPTM and BIS indices during the maintenance of and emergence from general anesthesia (13). The BIS value was lower than the SNAPTM index.
An additional limitation of this study was the use of a signal averaging time of zero, which may have given greater weighting to momentary changes in EEG signals than with the default SNAPTM setting of 10 seconds. We elected to use the zero averaging time because of the short time between injection and loss of consciousness. This may have introduced additional error in the observed and modeled SNAPTM index at end-point.
In conclusion, this study demonstrates that the SNAPTM index is associated with loss of consciousness after a propofol bolus. Further studies are necessary to determine whether other hypnotic drugs also demonstrate this association and how combinations of anesthetics affect the index.
|
Appendix 1: The SNAPTM EEG Algorithm
|
|---|
The SNAPTM index (SI) is a combination of two independent variables created by processing the spectral energy from a single-channel EEG. A high-frequency (HF) variable is computed from HF (80–420 Hz) components of the EEG and a low-frequency (LF) variable from the LF (0.1–40 Hz) components of the EEG. The SI, a dimensionless number from 100 to 0, is derived from the following equation:
The SNAPTM monitor uses a fast Fourier transformation to separate the EEG signal into its spectral components. The algorithm processes the last 10 s of EEG data to calculate the HF and LF components of the algorithm. The HF component of the SI is most active during the awake or lightly anesthetized state. Increasing depth of anesthesia is associated with predominance of the LF component. Internally, the HF component is scaled to result in values from 0.0 to 1.0, and the LF component is scaled to result in values from 0.0 to 100. The HF component of the SI decreases as spectral energy increases (more awake). The LF component is computed from the EEG power spectrum by using the following equation:
|
Appropriately weighted frequency components of the EEG are placed into the numerator and denominator, resulting in an LF value that increases as the depth of anesthesia increases and a lower SI. Because the LF value is a quotient, it is possible to obtain the same value with two EEG signals with different root mean squared amplitudes, depending on the frequency content.
The SNAPTM sampling system consists of analog hardware and digital signal processing (Fig. 6). The system acquires a single-channel EEG signal by using a disposable pregelled self-adhesive frontal array electrode that is applied to either the right or left temple. Electrode positions I and III function as recording sites, and position II serves as the ground. Electrode impedances are checked automatically upon initiation of monitoring with the option to continuously check impedances at user-defined intervals. If impedance is high (>10 k), a bad-impedance icon is displayed.
View larger version (17K):
[in this window]
[in a new window]
|
Figure 6. SNAPTM sampling system. EEG = electroencephalogram; HF = high frequency; LF = low frequency; FIR = finite impulse response; ADC = analog to digital converter; FFT = fast Fourier transformation.
|
|
The single-channel EEG is amplified with a gain of 2000, followed by a low-pass analog antialias filter with a cutoff frequency of 1000 Hz. The signal is then digitalized by using a 16-bit analog-to-digital converter sampling at 5120 Hz. The digitalized EEG is passed to a digital signal processor that uses downsampling techniques to create effective sampling rates of 1024 and 204.8 Hz. The data are stored and processed by using the procedure described above, and the SI is displayed on the monitor.
Artifact rejection and low-signal detection are performed in the time domain by using amplitude and time techniques. When a low signal is detected, the SNAPTM displays "Low-Signal," and the real-time EEG is displayed. Cautery and EMG artifact are recognized because the signals are much larger than normal EEG signals. The SI value flashes on the monitor, and a gap appears in the trend line. EEG data continue to be processed in the background, and the SI is displayed again when the artifact situation is resolved.
|
Acknowledgments
|
|---|
Supported in part by Viasys Healthcare, Inc. (Conshohocken, PA).
The authors thank Tom C. Krejcie, MD, Professor of Anesthesiology, and Michael J. Avram, PhD, Associate Professor of Anesthesiology, Northwestern University, Chicago, IL, for their help in study design and manuscript review.
|
References
|
|---|
- Rampil IJ. A primer for EEG signal processing in anesthesia. Anesthesiology 1998; 89: 980–1002.[ISI][Medline]
- Faulkner HJ, Traub RD, Whittington MA. Disruption of synchronous gamma oscillations in the rat hippocampal slice: a common mechanism of anaesthetic drug action. Br J Pharmacol 1998; 125: 483–92.[ISI][Medline]
- Faulkner HJ, Traub RD, Whittington MA. Anaesthetic/amnesic agents disrupt beta frequency oscillations associated with potentiation of excitatory synaptic potentials in the rat hippocampal slice. Br J Pharmacol 1999; 128: 1813–25.[ISI][Medline]
- Draguhn A, Traub RD, Schmitz D, Jefferys JG. Electrical coupling underlies high-frequency oscillations in the hippocampus in vitro. Nature 1998; 394: 189–92.[Medline]
- Traub RD, Draguhn A, Whittington MA, et al. Axonal gap junctions between principal neurons: a novel source of network oscillations, and perhaps epileptogenesis. Rev Neurosci 2002; 13: 1–30.[ISI][Medline]
- Traub RD, Kopell N, Bibbig A, et al. Gap junctions between interneuron dendrites can enhance synchrony of gamma oscillations in distributed networks. J Neurosci 2001; 21: 9478–86.[Abstract/Free Full Text]
- Crankshaw DP, Allt-Graham J. ED50 values for thiopentone, methohexital propanidid and Alfathesin: a clinical experiment. Anaesth Intensive Care 1978; 6: 36–43.[ISI][Medline]
- Chernik DA, Gillings D, Laine H, et al. Validity and reliability of the Observer’s Assessment of Alertness/Sedation Scale: study with intravenous midazolam. J Clin Psychopharmacol 1990; 10: 244–51.[ISI][Medline]
- Steyerberg EW, Harrell FE Jr, Borsboom GJ, et al. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol 2001; 54: 774–81.[ISI][Medline]
- Smith WD, Dutton RC, Smith NT. Measuring the performance of anesthetic depth indicators. Anesthesiology 1996; 84: 38–51.[ISI][Medline]
- Stanski DR. Monitoring depth of anesthesia. In: Miller RD, ed. Anesthesia. 5th ed. Philadelphia: Churchill Livingstone, 2000: 1087–116.
- Glass PS, Bloom M, Kearse L, et al. Bispectral analysis measures sedation and memory effects of propofol, midazolam, isoflurane, and alfentanil in healthy volunteers. Anesthesiology 1997; 86: 836–47.[ISI][Medline]
- Wong CA, Fitzgerald PC, McCarthy RJ. Comparison of depth of anesthesia indices (SNAPTM vs. bispectral) during balanced general anesthesia in patients undergoing outpatient gynecologic surgery [abstract A553]. 2002 ASA meeting abstracts. Available at: http://asa-abstracts.com.
- Chen X, Tang J, White PF, et al. A comparison of patient state index and bispectral index values during the perioperative period. Anesth Analg 2002; 95: 1669–74.[Abstract/Free Full Text]
Accepted for publication June 17, 2004.
This article has been cited by other articles:
|
|
|
|
 
C. A. Wong, R. J. Fragen, P. Fitzgerald, and R. J. McCarthy
A comparison of the SNAP IITM and BIS XPTM indices during sevoflurane and nitrous oxide anaesthesia at 1 and 1.5 MAC and at awakening
Br. J. Anaesth.,
August 1, 2006;
97(2):
181 - 186.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
Different from baseline; P < 0.05. 
Different from subjects who did not drop the syringe; P < 0.05.
