Different Propofol–Remifentanil or Sevoflurane–Remifentanil Bispectral Index Levels for Electrocorticographic Spike Identification during Epilepsy Surgery

EPILEPSY is a chronic brain condition characterized by recurring seizures as a result of excessive epileptogenic activity. Medical therapy, the cornerstone of managing epilepsy, still fails a substantial portion of patients who are refractory to therapy, or intolerable to antiepileptic drugs.¹  Surgical epileptogenic foci excision, in medically intractable patients, is currently an effective option.²  Because some epileptogenic foci are in close proximity to motor, memory, speech, or sensory vital cortical areas,³  intraoperative electrocorticography from electrodes directly placed on the cerebral cortex,⁴  could help locate precisely the origin of epileptic activity, and delineate the resection margins.² 

The bispectral index (BIS), an electroencephalographic processed parameter is currently a widely used tool for quantifying the effects of anesthetic or hypnotic agents. The BIS proprietary algorithm using a single dimensionless number employs three descriptors that dominate sequentially, as electroencephalography changes its character.⁵  However, as revealed by several publications, BIS cannot be considered a true “depth of anesthesia” measure,^5,6  but rather a measure of the effects of certain anesthetic agents. BIS was successfully used for monitoring neurosurgical procedures.⁷ 

Because different anesthetic agents and regimes could either obtund or provoke epileptiform activity, many aspects regarding the influence of anesthetics on electrocorticography epileptiform activity are still not guided by scientific evidence but rather by local convention. Propofol^8,9  and sevoflurane,^2,10  commonly used epilepsy surgery anesthetics, have both raised controversies from conflicting reports of pro- and antiepileptogenic effects on electrocorticography. To date, we have seen no comparative studies answering the core issues of whether increasing anesthetics wipes out the clues as to which tissue the surgeon should excise, and whether the leads exhibiting spikes at sedation levels are the same as those exhibiting spikes at deep-anesthetic BIS levels. We hypothesized that decreasing BIS levels would obtund the spike frequency and amplitude of electrocorticographic epileptiform activity. Our null hypothesis was that different BIS levels would bear no effect on spike frequency or amplitude.

The methods we used in the current study were based in part upon some of our earlier studies, especially regarding the study registration and logistics,^7,11  intraoperative use of BIS monitoring, anesthetic management techniques, perioperative care,^6,11  and some aspects of the statistical analysis used.⁷ 

We registered our study at the European Community Clinical Trials Database EudraCT trial registration number: 2009-015745-23.^†† Our report of a consecutive randomized parallel-group clinical trial was prepared in conformity with the “consolidated standards of reporting trials (CONSORT)-statement” guidelines.¹²  Second Affiliated Hospital of Dalian Medical University ethics committee approval (number 2011-002-08) was on May 16, 2011, at a meeting chaired by Professor Dr. XianZhi Meng, M.D., Chairman of the ethics committee. Subjects or their next of kin, who gave written informed consent, were recruited in our study between May 23 and September 30, 2011. We excluded subjects suffering from hepatic or renal disease. Twenty-two subjects with medically intractable epilepsy, undergoing surgical resection of epileptogenic foci were recruited in the study. Before surgery, all subjects underwent video electroencephalography and contrast-enhanced magnetic resonance imaging.

BIS-Vista Quatro® sensors were placed on subjects’ forehead and connected to BIS-Vista® monitor (BIS®; Covidien, Dublin, Ireland). Electroencephalographic signals were processed in real time using BIS algorithm version 1.4. After verifying less than 5 kΩ electrodes impedance and more than 95% signal quality index, a blinded investigator digitally recorded BIS variables in the once every 5-s mode. We set the BIS smoothing window at 15 s. The BIS monitor displays, in decibel (dB) units, the 70–110 Hz frequency band frontal electromyography relative to a logarithmic transformation of 0.01 μV. We chose the level of 35 dB or lesser as an indicator of adequate frontal electromyographic suppression.¹³  Using a computer generated program, subjects were randomly allocated to the propofol or sevoflurane groups and were further randomized to four different BIS levels sequence order.

Subjects’ regular anticonvulsant medications were discontinued on the morning of surgery and they received no premedication. For induction, after entering subjects’ anthropometric data, we started remifentanil Orchestra® Base Primea, (Fresenius Kabi, Bad Homburg, Germany) and propofol Diprifusor® (AstraZeneca Pharmaceuticals, Macclesfield, United Kingdom) target-controlled infusion pumps, incorporating Minto et al.¹⁴  and Marsh et al.¹⁵  pharmacokinetic models, respectively. The trachea was intubated after rocuronium 600 μg/kg administration and 10% lidocaine laryngeal spray. We ventilated the lungs mechanically with 40% oxygen in air, which was adjusted to maintain 30–34 mmHg end-tidal carbon dioxide partial pressure. Subjects’ monitoring included heart rate, mean arterial pressure, and the Neuromuscular Transmission Module¹⁶  (GE Healthcare, Madison, WI) kinemyography.¹⁷  We used a forced-hot-air-blanket to maintain core temperature more than 36°C.

Guided by preoperative investigations, a subdural 8 × 1, 8 × 2, or 8 × 4 platinum iridium (Pt-Ir) alloy grid electrodes array, 10-mm intercontact distance (Beijing HuaKeHengSheng Healthcare Co., Beijing, People’s Republic of China) was placed, after craniotomy and dural incision, on the potential epileptogenic zone and sutured to the dura. We used Nicolet® 32 Model-I electrocorticographic monitor (Nicolet Biomedical Inc., Middleton, WI) using Cz as monopolar reference electrode, with 70 Hz frequency, filtered at 0.01–1,500 Hz bandwidth and 0.03 s time constant 32-channel amplifier. We synchronized the BIS-Vista® and Nicolet® monitors’ clocks to the exact “h:min:s” time. Soon after the grid electrodes were placed, using sevoflurane minimum alveolar concentration, or propofol target-controlled infusion adjustments, subjects were randomly allocated to four 15-min BIS85 (BIS 71–85), BIS70 (BIS 56–70), BIS55 (BIS 41–55), and BIS40 (BIS ≤40) sequence order. Unlike the primary hypnotic-based inhalational or intravenous anesthetics, opioids in analgesic concentrations produce minimal electrophysiological alterations on the cerebral cortex.¹⁸  Maintaining remifentanil at levels which allow subjects to comfortably tolerate the endotracheal tube for the 15-min analogo-sedation¹⁹  segment of our study would not necessarily change BIS values.²⁰  Because our BIS 56–70 and BIS 71–85 15-min segments as such could indicate analgo-sedation and not general anesthesia, our study subjects were fully informed and consented that they could be exposed, for the study purposes, to analgo-sedation for the 15-min segment of our study.

Electrocorticographic recordings, imported into “American Standard Code for Information Interchange” format, were visually analyzed in the off-line mode by our electroencephalographer (Dr. Su), who was blinded to the anesthetics used and the four 15-min BIS sequence order assignments. Our electroencephalographer quantified electrocorticography spike/min frequency, microvolt voltage amplitude, and whether the active firing “spiking leads” at the sedation BIS levels are the same spiking leads at deep-anesthetic BIS levels. To distinguish true epileptiform electrocorticography spike signals from any pharmacologically induced signals of γ-amino-butyric acid-ergic anesthetics, we used the criteria first defined by Chatrian et al.²¹  and Gloor et al.,²²  which are still widely considered as the internationally recognized standards. Spikes are defined as “paroxysmal, spontaneous, isolated, high-voltage, electrical discharge, characterized by an acute triangular form with a duration of less than 70 ms, which could be clearly distinguished from background electroencephalography.”^21,22  Such criteria are essential in preventing background noise from being erroneously considered as epileptiform spike discharges. We classified spiking leads’ spatial distribution as “multilobar” or “lobar”, when they occur within one cerebral lobe.

We quantified the epileptiform activity, for 15-min visual counting selection, using the three most widely used criteria namely, spike amplitude, spike frequency, and number of spiking leads according to the electrocorticography analysis standards of Cleveland Clinic Foundation.²³ 

One of the major electroencephalographic “graphoelements” are the K-Complexes, first described by Loomis et al.²⁴  Today, K-Complexes are still defined, using Rechtschaffen and Kales criteria,²⁵  as “electroencephalographic waveforms with a well delineated negative sharp wave, which is immediately followed by a positive component. The total duration of the complex should exceed 0.5 s.”²⁵  Conventionally, most researchers now agree that K-Complexes are delineated as high-voltage, large amplitude 75–200 µV waveform, bi (tri) phasic slow waves, characterized by an initial short surface–positive transient wave, a slower larger surface–negative phase that peaks at 300–750 ms, immediately followed by a final positive phase that peaks near 900–1,300 ms.²⁶  It is believed that the K-Complexes constitute a physiologic correlate to arousal. This notion is mostly based on the association of the “spontaneous K-Complex” with signs of anesthesia lightening. The “response K-Complex” could also occur in response to surgical stimuli.²⁷ 

In order to differentiate epileptiform spike discharges from other drug-induced, nonepileptiform waves such as K-Complexes, we considered the wave width as a reliable characteristic to easily distinguish true epileptiform spike discharges of less than 70 ms from the K-Complexes, which are more than 0.5 s.

We measured spike amplitude using the automated capture “Nicolet monitoring software” of our electrocorticography monitoring window (fig. 1). Basically the spike amplitude is expressed as “volatility”, where the amplitude represents the “nadir-zenith vertical length of a wave”, or the mean of the two nadir-zenith perpendicular lines for biphasic waves (fig. 2).

We used data from the previously mentioned study³  for our a priori sample size power analysis, in which the preoperative mean spike frequency of 5 ± 4 declined to 3 ± 2 spikes/min with general anesthesia. For statistically significant difference with more than 90% power, our a priori one-sample paired t test (α = 0.05) power analysis revealed that we require a group size of 10 subjects. We then increased our sample size to 11 subjects to accommodate for dropouts. The “spike amplitude” and the “number of spiking leads” secondary parameters were not a priori-powered because there were no studies that examined the effect of anesthesia on “spike amplitude” or the “number of spiking leads.”

We used two-way ANOVA (group × BIS level) StatXact (Cytel Software Corporation, Cambridge, MA) and Number Crunching Statistical System 2007 (NCSS Inc., Kaysville, UT) statistical analyses to compare electrocorticographic variables between propofol and sevoflurane groups. Because the same number of spiking leads in each subject could be attained at different BIS levels, but with spiking leads at different locations, we further verified whether the “spiking leads” are at the same location at different BIS levels. Data were expressed as means ± SD. We considered P value less than 0.05 as statistically significant.

We further post hoc normalized our raw data to an arbitrary level of BIS85, as we a priori hypothesized a decline in spike parameters from that particular level, thus, giving the percentage change.

Subjects’ characteristics and history of epilepsy and pathology are presented in tables 1 and 2. Electromyographic values were consistently below 35 dB, with no significant difference in electromyography values at the four BIS levels or between the two groups. Mean ± SD estimated plasma and predicted effect site propofol concentrations were 2.5 ± 0.3 and 1.0 ± 0.2 µg/ml at BIS85, 3.2 ± 0.3 and 1.6 ± 0.1 µg/ml at BIS70, 4.5 ± 0.4 and 2.4 ± 0.2 µg/ml at BIS55, and 5.6 ± 0.6 and 3.1 ± 0.4 µg/ml at BIS40, respectively.

Seizure type was generalized secondary tonic–clonic seizure, with complex partial seizure in 9 subjects, generalized secondary tonic–clonic seizure in 10 subjects, and complex partial seizure in 3 subjects. Surgical procedures were lesionectomy, with corticoectomy in 5 subjects, lesionectomy in 4 subjects, anterior medial temporal resection in 12 subjects, and selective amygdalo-hippocampectomy in 1 subject. Our study patients had a favorable postoperative clinical course.

Mild increase in electrocorticography background noise, occasionally observed in our study, did not affect our recognition of spike activity. Two-way ANOVA revealed no difference between groups in spike frequency (P = 0.720), spike amplitude (P = 0.647), or number of spiking leads (P = 0.653). Our blind review revealed that decreasing BIS levels in the propofol and sevoflurane groups increased spike frequency (fig. 3; P < 0.001 and P < 0.001) and spike amplitude (fig. 4; P = 0.006 and P = 0.009). Spiking leads at BIS85 and BIS70 sedation were at the same location as the spiking leads at BIS55 and BIS40 deep-anesthetic levels. In the propofol and sevoflurane groups, compared with BIS85 and BIS70 the one or two “newly appearing spiking leads” at deep BIS55 and BIS40 did not result in a significant difference (P = 0.057 and P = 0.109) in the number of spiking leads (fig. 5). In the propofol and sevoflurane groups, the number of lobes involved was limited to the temporal lobe in 10 subjects, and temporal-occipital multilobar in one subject in each group.

In the propofol and sevoflurane groups, compared with BIS85, spike frequency in the propofol and sevoflurane groups increased 100 and 170% at BIS70, 116 and 180% at BIS55, and 132 and 230% at BIS40. Whereas, compared with BIS85, spike amplitude in the propofol and sevoflurane groups increased 108 and 113% at BIS70, 121 and 125% at BIS55, and 128 and 170% at BIS40.

In electrocorticography representative raw traces (figs. 6, 7, 8, and 9) we highlighted some spike discharges distinguishable from background, and we demonstrated an increase in spike frequency and amplitude with deepening BIS levels. The ovals do not point out to all spikes for the chosen sweep, rather a representative trace, where some leads, such as CS-01 that did not have spikes at high BIS levels, begin to exhibit spiking behavior at deeper BIS levels.

We report electrocorticography evaluation at various BIS-Vista levels using two different anesthetic regimes. Our study results suggest that our hypothesis of a reduction in epileptiform activity with decreasing BIS levels and our null hypothesis were both rejected as we observed a stark contrast. Decreasing BIS levels actually enhanced electrocorticographic epileptiform activity. Despite the fact that epilepsy preoperative diagnostic modalities are making great strides, intraoperative electrocorticography, which might be influenced by anesthetic agents, still remains an essential tool for defining and delineating epileptogenic foci. We addressed the controversy that general anesthetics could hamper reliable detection of electrocorticography epileptogenic spikes because deeper BIS levels actually increased spike frequency and amplitude of sharp epileptiform waves. This suggests that in medically intractable epilepsy patients, deep BIS anesthetic activation of epileptiform activity, may prove to be an important diagnostic tool conveniently supporting neurosurgeons’ assessment for an accurate tailored resection of the epileptogenic foci.

It is believed that K-Complexes constitute a physiologic correlate of arousal. This notion is mostly based on the association of “spontaneous K-Complexes” with lightening of anesthesia. The “response K-Complexes” could also occur in response to surgical stimuli.²⁷  Si et al.²⁸  compared changes in the frequency and amplitude of K-Complexes in refractory frontal lobe epileptic and control patients. K-Complexes’ frequency and amplitude were higher in epileptic patients than in controls.²⁸  To differentiate true epileptiform spike discharges from other drug-induced nonepileptiform waves, such as K-Complexes, we considered wave width as a reliable index that can easily distinguish true epileptiform spikes of less than 70 ms^21,22  from other γ-amino-butyric acid-ergic anesthetics’ pharmacologically induced nonepileptiform waves, such as spontaneous or response K-Complexes of greater than 0.5 s width.

Miyauchi et al.²⁹  studied electroencephalographic background noise activity in epileptic and control patients. In epileptic patients, electroencephalographic background noise showed a marked increase compared with controls.²⁹  This brings in question the usefulness of our study in such surgical procedures when a grid of electrodes is directly placed on to the cerebral cortex providing a raw signal, from which we could have computed the power spectrum of the electrocorticography channels. This simple analysis of spectral changes in electrocorticographic signals for both propofol and sevoflurane would have provided a much better idea of the background brain state in which spiking is observed. However, our a priori registered study design did not allow us to utilize the “spectral signal: noise ratio analysis” as a study parameter. That was beyond the scope of our a priori registered study design.

Despite the common notion that increasing anesthetics are thought to induce progressive loss of brain responsiveness, Kroeger and Amzica³⁰  presented the first in vivo electroencephalographic patterns of deep anesthesia-induced dramatically increased brain excitability. Using intracellular recordings of neurons and glia, extracellular calcium, and electroencephalographic recordings, they demonstrated that deep anesthesia increased brain excitability via complex mechanisms, such as excitatory N-methyl-d-aspartate modulation, gap junction transmission, and extracellular calcium concentration.³⁰  Voss et al.³¹  further clarified that “there still seems to be no unifying neural mechanism of anesthetics-related epileptogenesis.” Epileptiform activities are not caused simply by “too much excitation”, but rather by excitation of neurons mass already primed to such excitation. Deep anesthesia-induced γ-amino-butyric acid-ergic inhibition can sensitize the cerebral cortex to lower the excitation threshold. This occurs through network “inhibitory lag” prolongation, “inhibiting-the-inhibitors” subpopulations of interneurons, or changing the potential of chloride reversal “excitatory γ-amino-butyric acid” at the synaptic level.³¹ 

At BIS 56–70 and BIS 71–85 15-min segments of our study, maintaining remifentanil at analgo-sedation¹⁹  levels allowing patients to tolerate the endotracheal tube did not necessarily change BIS values.²⁰  Shafer et al.¹⁸  demonstrated that dosages of almost five times the analgesic concentrations would be required for the appearance of electroencephalographic depression.¹⁸  Combined with propofol, remifentanil of almost eight times the analgesic dosage for endotracheal intubation, did not modify the BIS value.²⁰  This is due to the fact that noncortical structures, which are undetectable by electroencephalography such as locus coeruleus-noradrenergic system, are involved in the mechanism of opioids drug effect.³²  This would result in the BIS, although able to estimate the sedative hypnotic component of anesthetics such as propofol or sevoflurane, would not reflect the direct opioid effect of remifentanil.²⁰ 

Increasing sevoflurane concentrations from 0.5 to 1.5 minimum alveolar concentration in epileptic patients, increased electrocorticography spikes’ frequency and number of spiking leads.¹⁰  Conversely, in epileptic patients sevoflurane’s 1.5 minimum alveolar concentrations, when combined with fentanyl, had significantly fewer electrocorticography spikes than 0.5 minimum alveolar concentration.²  Similarly, several conflicting studies demonstrated propofol’s controversial electrocorticography pro- and antiepileptogenic properties. Subanesthetic propofol as sole sedative in epileptic patients seemed to possess no electrocorticography epileptogenic activity,⁹  another controversy when electrocorticographic epileptiform spike activity were easily detectable, and not obtunded during craniotomy under propofol infusion.⁸  The effects of propofol on epileptic discharge activity may be related to its potential mechanism of intrinsic subcortical glycine antagonism,³³  as suggested by animal studies.³⁴ 

In our blind electrocorticography evaluation, we considered several parameters namely, spike frequency, spike amplitude, and the spiking leads. Obviously, knowing the location of the spiking leads is “crucial” for determining the extent of resection and should be formally precisely quantified. This gives more credence to the “location” of the spiking leads criteria than spikes’ frequency or amplitude parameters, which can just help in “better recognition and identification” of already existing epileptogenic foci. Interestingly, in our study, just spike frequency and amplitude electrocorticography parameters exhibited a significant temporal enhancement, with decreasing BIS levels but not the spiking leads. In the propofol and the sevoflurane groups, the leads exhibiting spikes at BIS85 and BIS70 sedation were at the same location as the spiking leads at BIS55 and BIS40 deep-anesthetic levels. Compared with BIS85 and BIS70, the one or two “newly appearing” deep-anesthetic BIS55 and BIS40 spiking leads did not result in a significant difference in the number of spiking leads. We believe that the one or two new spiking leads do not represent “new pharmacologically induced spikes in eloquent brain matter outside epileptiform zones” potentially provoking otherwise normal cortex into epileptic-like activity, rather, epileptogenic leads in existing epileptogenic zones that only become apparent with deep-anesthetic BIS levels.

In an epileptic patient under sevoflurane anesthesia, Chinzei et al.³⁵  herein described transient repeated episodes of BIS fluctuations associated with electroencephalography epileptogenic activity.³⁵  Similarly, we recently described repeated episodes of BIS transient fluctuations reflecting repeated high electromyography convulsions coinciding with electroencephalography epileptic seizures in a patient with “acute encephalitis with refractory, repetitive partial seizures.”³⁶  Because BIS values are derived from processed electroencephalographic data, abnormal electroencephalography from neurological conditions of excessive epileptogenic activity could significantly affect BIS values.^35,36  Furthermore, electromyographic activities fall within the bispectrum’s “range of interest”, as electromyography_{30-300 Hz} overlaps the BIS descriptor BetaRatio electroencephalography_{30-47 Hz},³⁷  which the BIS algorithm would misinterpret as lightening of anesthesia.³⁸  Because BIS cannot be considered as a true “depth of anesthesia” measure,⁵  high frequency “polyspike” epileptogenic activities will alter BIS values, including those we encountered in our current study, even when there is no potential for electromyography contamination as our study patients’ electromyography values were consistently below the 35 dB cutoff value.¹³  Here our study has its limitation because the BIS algorithms’ proprietary nature renders it impossible to scientifically evaluate or precisely quantify to what extent the electromyography and electroencephalography epileptogenic components influenced the BIS value. We consider that the observed BIS values in our study patients might not be accurately commensurate with propofol and sevoflurane anesthesia levels.

Another limitation in our study was seeking “optimal levels” that would imply a performance “decrement.” Our current design revealed that spiking increased with deepening BIS levels. However our study design did not allow us to investigate what would happen if one was to further bring BIS down to burst suppression, typically occurring below our investigated range that could have revealed more spiking. It seems that our study is incomplete, a question which our as such a priori registered study design did not allow us to expand.

We believe that our study did address an unresolved issue. We have demonstrated that decreasing BIS levels enhanced the morphology of epileptogenic spikes’ frequency and amplitude with the same location and number of spiking leads. In subjects with medically intractable epilepsy, deep BIS levels may prove to be an important diagnostic tool that could conveniently support neurosurgeons’ evaluation for an accurate tailored epileptogenic foci resection.

The authors thank Jun Zhu, B.S. (Research Assistant, Covidien, Hang Zhou, Zhe Jiang, People’s Republic of China), and Li Ding, B.S. (Research Assistant, AstraZeneca, Xi’an, Shaanxi, People’s Republic of China), for their technical support for the study data collection.