Overestimation of Bispectral Index in Sedated Intensive Care Unit Patients Revealed by Administration of Muscle Relaxant

This prospective study was conducted in accordance with the Helsinki declaration, and the protocol was approved by our local ethical committee (Comité Consultatif pour la Protection des Personnes se prêtant aux Recherches Biomédicales, Pitié-Salpêtrière, Paris). Care of the patients conformed to standard procedures currently used in our ICU, including standard equipment monitoring with continuous electrocardiography, invasive arterial blood pressure monitoring, pulse oximetry and capnography, and the additional use of noninvasive BIS monitoring. Patients were eligible for this study if they were mechanically ventilated and continuously sedated with midazolam and sufentanil to achieve a Sedation-Agitation Scale value equal to 1 (table 1). 16We prospectively included patients who required administration of muscle relaxant either for adaptation to mechanical ventilation or to facilitate invasive clinical procedures such as transesophageal echocardiography or bronchoscopy. Exclusion criteria were hemodynamic instability, administration of sedative drugs other than midazolam and sufentanil, change in midazolam or sufentanil administration regimen during the study period, and recent (<3 h) administration of muscle relaxant before the study period. Finally, because brain injury and hypothermia may decrease BIS values, we also excluded patients with neurologic injury, head trauma, or hypothermia (tympanic temperature < 36.0°C). 9,15,17,18 

The BIS was measured with an Aspect A-2000 monitor (Aspect Medical Systems, Newton, MA; software version 2.10). In every patient, the skin was carefully cleaned before positioning the BIS sensor (ZipPrep® three-electrode sensor; Aspect Medical Systems) on the forehead as specified by the manufacturer. Before starting recording, we verified that electrode impedance was below 5 kΩ. The BIS, electromyographic activity, total power of the signal (TOTPOW), and signal quality index (SQI) were continuously recorded and stored on a computer for off-line analysis. Because it has been shown that many devices such as forced-air warming systems can interfere with BIS monitoring, all of those artifacts were systematically noted and removed from computer file recordings during off-line analysis. 19,20Similarly, physical examination and care of the patient, such as tracheal suction, 7were carefully noted because they could markedly interfere with BIS value either through neurologic stimulation of the patient or through artifact production.

Elsewhere, because a new version of the BIS® monitor had been recently commercialized (BIS® XP, software version 3.12, four-electrode sensor; Aspect Medical Systems), which should better discriminate artifacts such as electromyographic activity, we decided to investigate some of the patients simultaneously with the old and the new BIS® monitors, as previously described. 21For these patients, the two BIS® sensors were simultaneously placed on opposite sides of the head.

The effect of neuromuscular blocking agents on muscle activity was monitored through stimulation of the ulnar nerve and acceleromyography at the adductor pollicis muscle. 22,23The stimulation surface electrodes were placed over the ulnar nerve on the volar side of the wrist. The acceleration transducer was placed at the distal part of the thumb, with its largest flat side against the thumb. The train-of-four (TOF) sequence (2 Hz) was recorded after supramaximal stimuli (200-μs duration) repeated every 15 s (TOF-Watch SX Monitor 1.2; Organon, Dublin, Ireland). The first response in the TOF sequence (T1) was used as the baseline value to which all subsequent T1 responses were compared. A preliminary study (data not shown) reported that TOF watch monitoring at the adductor pollicis did not interfere with BIS monitoring.

Before administration of myorelaxant, BIS, electromyographic activity, TOTPOW, and SQI were recorded for a 20-min period without clinical stimulation to the patient (baseline period) to determine the value of these variables in the absence of muscle relaxant. Immediately after this 20-min recording, control twitch height of the adductor pollicis was calibrated to set the T1 100% baseline level. Then, intravenous administration of myorelaxant (0.5 mg/kg atracurium) was performed. As soon as T1 decreased below 5% of baseline value, BIS, electromyographic activity, TOTPOW, and SQI were determined from a second 20-min recording period without clinical stimulation to the patient (myorelaxation period).

For patients who required myorelaxation for adaptation to mechanical ventilation, the study period was then finished. For the other patients who transitorily required myorelaxation to facilitate transesophageal echocardiography or bronchoscopy, the following procedure was subsequently performed: After completion of this investigation, as soon as T1 recovered 90% of baseline value, BIS, electromyographic activity, TOTPOW, and SQI were determined from a last, third 20-min recording period without clinical stimulation to the patient (recovery period). We also studied the shape of both BIS and electromyographic activity recovery curves and arbitrarily defined two pathways of recovery: (1) smooth recovery when the shapes were smooth and regular; and (2) stepped recovery when there were sudden increases in BIS (>20 units) and/or electromyographic (>5 dB) values occurring in less than 1 min.

For the baseline and myorelaxation periods, patients were conventionally classified in four levels of sedation, according to standard BIS range guidelines 11: light sedation (BIS ≥ 85), deep sedation (65 ≥ BIS < 85), general anesthesia (40 ≥ BIS < 65), and deep hypnotic state (BIS < 40). Elsewhere, as previously described, we defined a threshold of 42 dB, above which electromyographic activity is considered to artifactually increase BIS values. 9,13 

Data are expressed as mean ± SD. Means ± SDs of BIS, electromyographic activity, TOTPOW, and SQI were calculated from each 20-min recording period. Comparison of two means was performed using the unpaired and paired Student t  tests when appropriate. Comparison of several means was performed using repeated-measures analysis of variance and Newman-Keuls test. Correlation between two variables was performed using the least squares method. Multiple regression analysis was performed to research predictor baseline variables of the overestimation of BIS. All P  values were two-tailed, and a P  value of less than 0.05 was required to reject the null hypothesis. All statistical analysis were run on a computer using NCSS 6.0 software (Statistical Solutions Ltd., Cork, Ireland).

Forty-five patients, aged 48 ± 22 yr (35 men, 10 women; weight, 71 ± 9 kg), were investigated prospectively. All of them required sedation for more than 24 h in the ICU for multiple trauma (n = 18) or postoperative care after major vascular (n = 13), orthopedic (n = 6), abdominal (n = 4), or maxillofacial (n = 4) surgery. Fifteen of these patients required long-period myorelaxation for adaptation to mechanical ventilation, whereas 30 patients required transitory administration of myorelaxant for facilitation of transesophageal echocardiography or bronchoscopy (transitory group).

No change in midazolam (6.5 ± 4.0 mg/h) or sufentanil (25 ± 13 μg/h) administration regimen occurred in any of the 45 patients during the study periods. For the 45 patients, hemodynamic variables remained unchanged between the baseline and myorelaxation study periods (table 2). For the 30 patients of the transitory group, hemodynamic variables remained unchanged among the baseline, myorelaxation, and recovery study periods (table 3).

After administration of myorelaxant, BIS (67 ± 19 vs.  43 ± 10, P < 0.001) and electromyographic activity (37 ± 9 vs.  27 ± 3 dB, P < 0.001) significantly decreased, whereas SQI significantly increased (90 ± 8 vs.  94 ± 7%, P < 0.001) and TOTPOW was unchanged (53.1 ± 3.2 vs.  52.7 ± 2.9 μV², NS ). Figure 1shows two representative examples of the effect of administration of neuromuscular blocker on BIS and electromyographic activity, respectively, in one patient who required long-period myorelaxation for adaptation to mechanical ventilation and in another patient who required transitory myorelaxation for facilitation of bronchoscopy.

Nevertheless, the effect of administration of neuromuscular blocker on BIS and electromyographic activity was highly variable among the 45 patients studied (fig. 2). In 13 patients, there was no change in BIS range assessment, according to the four standard BIS range guidelines; 19 patients showed a decrease of one step in BIS range assessment, 11 patients showed a decrease of two steps, and 2 patients showed a decrease of three steps (fig. 2, A ). These findings imply that BIS range assessment was overestimated in 32 of the 45 studied patients (71%). On the other hand, before administration of myorelaxant, only 10 of the 45 patients (22%) showed electromyographic activity greater than the threshold of 42 dB, above which electromyographic activity is considered to artifactually increase BIS values (fig. 2, B ). Figure 3presents trend recordings of two patients who respectively showed a nonsignificant change and a major decrease in both BIS and electromyographic activity following myorelaxant administration.

Figure 4shows the distribution of patients during the baseline and myorelaxation periods according to the four standard BIS range assessments. The number of patients under light or deep sedation versus  general anesthesia or deep hypnotic state was markedly overestimated before administration of myorelaxant (53 vs.  2%, P < 0.001).

We observed a significant correlation between BIS and electromyographic activity during the baseline period (R = 0.78, P < 0.001;fig. 5, A ). We also observed a significant correlation between ΔBIS (= baseline BIS value − myorelaxation BIS value) and both baseline BIS (R = 0.85, P < 0.001;fig. 5, B ) and baseline electromyographic activity (R = 0.74, P < 0.001;fig. 5, C ). Multiple regression analysis performed with ΔBIS as a dependent variable, and baseline variables as independent variables showed that ΔBIS was significantly correlated only to both baseline BIS and electromyographic values (table 4).

For the 30 patients in the transitory group, during the recovery period, the variables BIS, electromyographic activity, TOTPOW, and SQI recovered values, which were not significantly different from baseline values (table 3). Recovery of BIS and electromyographic activity occurred generally earlier than recovery of acceleromyography at the adductor pollicis (fig. 3, B ). According to our definition of recovery curve shape, recovery of BIS and electromyographic activity was smooth in 13 patients and parallel with acceleromyography. On the contrary, in the 17 remaining patients, we observed a stepped recovery of BIS and electromyographic curves, whereas recovery of acceleromyography at the adductor pollicis occurred progressively and smoothly (fig. 6).

Finally, among the 45 patients of this study, 16 had been studied using the old and new versions of the BIS® monitor simultaneously. We observed that the overestimation of BIS value because of high electromyographic activity was still present with the new BIS® XP monitor (fig. 7). Indeed, administration of neuromuscular blockade significantly decreased BIS values with both the old (−25 ± 15%, P < 0.001) and the new (−21 ± 14%, P < 0.001) monitors (table 5). In addition, there was no significant difference in these BIS decreases after myorelaxant administration between the two monitors.

In this study, we have shown that in sedated ICU patients, (1) BIS markedly decreased after administration of myorelaxant; (2) the decrease in BIS value following neuromuscular blockade was significantly correlated with BIS and electromyographic values before administration of myorelaxant; (3) using standard BIS range guidelines, the number of patients under light or deep sedation versus  general anesthesia or deep hypnotic state was markedly overestimated before administration of myorelaxant; and (4) overestimation of the BIS value because of high electromyographic activity was still present with the new BIS® XP monitor.

Sedation is an important issue in ICU patients, and assessing the level of unconsciousness of patients is mandatory to determine the amount of sedative drugs to be administered. 24Whereas adequate sedation reduces the stress response and improves the tolerance of routine ICU procedures, oversedation may be deleterious by increasing the duration of mechanical ventilation and the incidence of pneumonia. 10,25,26Several clinical scores such as the Ramsay scale and the Sedation-Agitation Scale (table 1) are subjective but reliable methods to assess the level of sedation in sedated ICU patients and have been used widely both clinically and in many comparative sedation trials. 16,24,27In the current study, we used the Sedation-Agitation Scale instead of the Ramsay scale. Indeed, although the Ramsay scale has existed for much longer, it is actually considered to be less reliable than the Sedation-Agitation Scale to assess sedation in ICU patients. 13,16Elsewhere, we have chosen to investigate deeply sedated patients with a Sedation-Agitation Scale of 1. Indeed, it should be kept in mind that this is below the sensitive range of this score and may represent a large potential source of patient variability. Nevertheless, one of the potential uses of BIS monitoring in the ICU should be to detect oversedation. Therefore, we have voluntary chosen to investigate BIS monitoring in patients with a Sedation-Agitation Scale score of 1, i.e. , those who were exposed to a risk of oversedation much more than others.

For several years, BIS monitoring has been used in sedated ICU patients as a tool for an objective assessment of sedation or hypnotic drug effect trials. 24Indeed, good correlations were reported between clinical sedation scores and BIS values in sedated ICU patients. 3–6,8,9However, among these studies, some of them pointed out that BIS was correctly correlated with sedation scores in only some sedated ICU patients. 4,5,8,9On the other hand, following the cases reported by Bruhn et al.  12in anesthetized patients, several authors have recently noted that BIS reliability could be severely impaired because of excessive electromyographic activity, which could contaminate BIS monitoring and therefore lead to overestimated and irrelevant BIS values in sedated ICU patients. 5,8,9,13Nevertheless, to our knowledge, the magnitude of BIS overestimation in sedated ICU patients without neuromuscular blockade has not previously been established. Our study clearly demonstrated that BIS overestimation in sedated ICU patients without neuromuscular blockade was significant and, in addition, showed that this overestimation was correlated both with BIS and electromyographic values before administration of myorelaxant (fig. 5).

In the current study, in sedated ICU patients, we used the BIS range guidelines, which have been initially developed for anesthetized patients. 11This point should be kept in mind because the relevance of these anesthetic guidelines for sedated ICU patients had not yet been proven. Nevertheless, to the best of our knowledge, no BIS range guidelines that describe appropriate and inappropriate levels of sedation when assessed by such a monitor have been widely accepted for the ICU.

After administration of myorelaxant, 32 of the 45 studied patients had a decrease of at least one step in BIS range assessment (fig. 2, A ), which means that BIS range assessment was overestimated in approximately 71% of our patients before administration of myorelaxant. This finding is important because only 10 of the 45 studied patients (approximately 22%) showed a baseline electromyographic value greater than the threshold of 42 dB, above which electromyographic activity is considered to artifactually increase BIS values 9,13(fig. 2, B ).

The number of patients under light or deep sedation versus  general anesthesia or deep hypnotic state was markedly overestimated before administration of myorelaxant (53 vs.  2%, P < 0.001;fig. 4). Therefore, clinicians who wish to determine the amount of sedation in ICU patients only from BIS monitoring would expose them to an unnecessary oversedation with potential morbidity, mortality, and increased cost. 10,25,26Elsewhere, because we found that on one hand, baseline BIS value was significantly correlated to baseline electromyographic activity, and on the other hand, that the magnitude of decrease in BIS after myorelaxation was significantly correlated both with BIS and with electromyographic values before administration of neuromuscular blocker, the inference is that the “true” BIS value could be roughly approximated from both BIS and electromyographic values before neuromuscular blockade. It should be remembered that before administration of neuromuscular blockers, overestimation of BIS was present, although SQI showed high values (90 ± 8%), indicating that BIS monitoring should have been reliable. Even with SQI near 100%, we still observed a significant decrease in BIS value after administration of myorelaxant drug (fig. 1). Finally, even if myorelaxation significantly increased SQI, we did not find any significant correlation between the decrease in BIS value after neuromuscular blockade and the SQI before myorelaxation.

The BIS was initially developed as an objective measurement of the depth of anesthesia in the operating room, where patients are highly sedated and often paralyzed. 1In the ICU, patients are generally less profoundly sedated than in the operating room and, most of the time, do not receive neuromuscular blockers. 23Also, the data used to develop the BIS algorithm where obtained from a large database of anesthetized patients and may not necessary always apply to sedated ICU patients. 28In addition, the sources of interference during BIS monitoring are various and common in ICU. 13,29Among those artifacts, high electromyographic activity is probably the most common limitation to BIS monitoring in ICU patients. 5,8,9,13Nonetheless, the contamination of BIS by electromyographic activity is inherent in the calculation of BIS by the monitor: BIS® software uses electroencephalographic signals up to 47 Hz, whereas electroencephalographic and electromyographic signals are conventionally considered to exist in the 0.5- to 30-Hz and 30- to 300-Hz bands, respectively. 11This overlap of electroencephalographic and electromyographic signals will undoubtedly remain the main pitfall of BIS monitoring in the ICU or in the operating room in unparalyzed patients.

Recovery of electromyographic activity and thus of BIS occurred earlier than recovery of acceleromyography at the adductor pollicis (fig. 3, B ). Because the BIS® sensor was applied on the forehead, this could be explained by the known delayed recovery after myorelaxation of adductor pollicis as compared to facial muscles such as the orbicularis oculi. 30On the other hand, in 17 patients, recovery of BIS and electromyographic activity was unexpectedly abrupt and irregular, with several steps, whereas recovery of acceleromyography at the adductor pollicis occurred progressively and smoothly. This is probably linked to the ICU environment, where various stimuli are unavoidable. 29,31Indeed, Johansen et al.  32recently showed that administration of stimulus to sedated ICU patients could induce increases in both electromyographic activity and BIS and that these increases may also depend on the initial level of hypnosis. Moreover, noise alone, which is generally high in the ICU, could be seen as a noxious stimulus, causing sympathetic activation and therefore increasing BIS in lightly sedated patients. 33,34 

Most of the previous studies investigating the accuracy of BIS monitoring in ICU patients were performed without myorelaxation (which was even an exclusion criterion for some of them) and consequently involved exposure to the highly contaminating effect of electromyographic activity, which should limit the interpretation of their conclusions. 3–6,8,9In the extreme, BIS values up to 97, related to very high electromyographic activity, were recently reported in brain-dead patients, which demonstrated that BIS could be highly contaminated by electromyographic activity. 15Because BIS monitoring has been recently proposed as a monitor of neurologic function in unsedated brain-injured patients, physicians should keep in mind the potential major interference from high electromyographic activity. 18,35–43 

Finally, there are some limits in the interpretation of the results of our study. First, we only investigated patients who were deeply sedated with sufentanil and midazolam, and therefore, we could not extrapolate the magnitude of overestimation of BIS to other sedation protocols. Nevertheless, this was done purposely to investigate a homogeneously sedated group of patients under a common sedative drug protocol administration, in contrast to most previous studies of BIS in the ICU. 4–6,9,13Second, this study was performed with the monitor Aspect 2000 running software version 2.10 with a three-electrode sensor. BIS monitoring is a recent technology, and improvements in hardware, software, and sensor are continuously available (BIS® XP, software version 3.12, four-electrode sensor). From the point of view of AspectMS®, this new equipment should be better at detecting interference from electromyographic activity and other artifact conditions often found in the ICU. Indeed, Coluzzi et al.  21recently reported an improvement in sedation monitoring of ICU patients with the new BIS® XP monitor as compared to the old standard BIS® monitor. Nevertheless, our results from the 16 patients investigated simultaneously with the two monitors showed that neuromuscular blockade still induced a significant decrease in BIS values as assessed by the new XP monitor, which was similar to that observed with the old monitor (table 5).

In summary, BIS monitoring in the ICU may be dramatically overestimated because of high muscular activity. The magnitude of the BIS decrease following neuromuscular blockade is significantly correlated to both BIS and electromyographic values before myorelaxation. The corollary of this overestimation could be a BIS-induced oversedation, which is unnecessary and may be deleterious for ICU patients.

The authors thank David J. Baker, M.D., F.R.C.A. (Département d'Anesthésie-Réanimation, Hôpital Necker, Paris, France), for reviewing the manuscript.