Flumazenil is an imidazobenzodiazepine that promptly reverses via competitive inhibition the hypnotic/sedative effects of benzodiazepines on gamma-aminobutyric acid receptors. Endogenous benzodiazepine ligands (endozepines) were isolated in urine, cerebrospinal fluid, and breast milk of women who had not received benzodiazepines. The bispectral index (BIS), an electroencephalographically derived parameter widely used for monitoring the effects of anesthetic/hypnotic drugs, was shown to correlate to various conditions that could influence electroencephalography. The authors examined the hypothesis that 0.5 mg of flumazenil administered to healthy unpremedicated patients during deep surgical remifentanil/propofol anesthesia would increase the BIS value and might expedite recovery from anesthesia.


Sixty healthy unpremedicated patients were randomly allocated to the flumazenil or control groups. After study drug administration, the authors compared BIS values and various recovery parameters in the flumazenil and control groups.


BIS baseline values in the flumazenil group (38.7 +/- 3.8) increased 15 min after flumazenil administration (53.2 +/- 4.7), with a significant difference over time (P < 0.0001) between the two groups. Mean recovery parameters time, comprising time to spontaneous breathing, eye opening/hand squeezing on verbal command, extubation, and date of birth recollection, was significantly shorter (P = 0.0002) in the flumazenil group (6.9 +/- 2.6 min) compared with the control group (9.8 +/- 2.9 min).


This study demonstrates that flumazenil given to healthy unpremedicated patients during propofol/remifentanil anesthesia significantly increased the BIS value and allowed earlier emergence from anesthesia. This may indicate that flumazenil could be used on a case-by-case basis to reverse endogenous or exogenous endozepines that might play a role during anesthesia.

FLUMAZENIL is an imidazobenzodiazepine that promptly reverses via  competitive inhibition the hypnotic/sedative effects of benzodiazepines on γ -aminobutyric acid receptors.1Flumazenil was shown to have a beneficial analeptic stimulant effect in hepatic encephalopathy patients; a recent meta-analysis2of several double-blind randomized placebo-controlled trials described improvement in electroencephalographic signs of hepatic encephalopathy after flumazenil administration in patients who had not received sedative/hypnotic medications.2Endogenous benzodiazepines ligands (endozepines)3were isolated in urine,4cerebrospinal fluid,5and breast milk6of women who had not received benzodiazepines. It has been suggested that there might be a beneficial analeptic stimulant effect as a result of flumazenil's displacement of these endogenous benzodiazepine ligands.

The bispectral index (BIS), an electroencephalographically derived parameter widely used for monitoring the effects of anesthetic/hypnotic drugs, was shown to correlate to various conditions that could influence electroencephalography.7Meanwhile, there is paucity of information regarding the effects of central nervous system (CNS) stimulants on BIS monitoring when they are added to general anesthesia. Because flumazenil is a readily available medication that might exert a stimulant effect expediting anesthesia recovery, we tested the hypothesis that flumazenil administration in unpremedicated patients during steady-state deep surgical anesthesia would lead to an increase in BIS value as a surrogate parameter suggesting a lighter plane of anesthesia. Our primary endpoint was the effect of flumazenil administration on deep surgical propofol/remifentanil total IV anesthesia of BIS around 40. The secondary endpoints were the differences between the two groups in various anesthesia recovery parameters.

Our report of a prospective clinical consecutive randomized study was prepared in conformity with the guidelines of the consolidated standards of reporting trials (CONSORT) statement.8After approval by the Peking University First Hospital (Beijing, People's Republic of China) ethics committee, all patients who agreed to participate in the study gave written informed consent. Exclusion criteria were body mass index less than 20 or greater than 26 kg/m2, treatment with cardiovascular or sedative/hypnotic drugs that might affect BIS monitoring,7and medical conditions that could affect the level of consciousness such as stroke, stupor, or dementia.

Using a computer-generated program, 60 unpremedicated patients with American Society of Anesthesiologists classification I-II undergoing upper/lower extremities or general surgical procedures of around 3 h with expected blood loss of less than 1 l were randomly allocated to the flumazenil or control groups. A BIS Quatro sensor (Aspect Medical Systems, Newton, MA) was placed on the patients' forehead according to manufacturer's recommendations and connected to a BIS-XP monitor (version 3.4; Aspect Medical Systems). Data from the BIS and electromyography power displayed in decibel (dB) units were continuously collected and stored every 5 s on a laptop computer, and the BIS smoothing window was set at 30 s. Recordings were started after verifying a signal quality index of greater than 95% and electrodes impedance of less than 5 kΩ.

The expected surgical procedure duration was around 3 h; therefore, the short-acting opioid fentanyl (1.5 μg/kg) was used for induction to avoid hypotension that might result from using remifentanil for induction.9Propofol target-controlled infusion (TCI) using a Diprifusor infusion pump (AstraZeneca, Macclesfield, United Kingdom) incorporating a three-compartment Marsh pharmacokinetic model10was started after entering patients' anthropometric data. Propofol TCI was set to reach a plasma concentration of 3 μg/ml over a period of 2 min, during which patients were allowed to breathe spontaneously via  a facemask. Time to loss of consciousness as patients lost their eyelash reflex and no longer responded to verbal command, propofol TCI, and BIS values at loss of consciousness were recorded. Remifentanil (0.1–0.3 μg · kg−1· min−1) continuous infusion was used for maintenance. Neuromuscular block at the adductor pollicis muscle was evaluated using train-of-four (TOF)–Watch (Organon, Oss, The Netherlands). The acceleration piezo-transducer of the TOF-Watch was attached to the thumb, and the ulnar nerve was stimulated supramaximally at the wrist (pulse width, 200 μs, square wave) via  surface electrodes with TOF stimuli (2 Hz for 2 s) at 12-s intervals. T1(first twitch of the TOF) expressed as percentage of control response and the TOF ratio (T4:T1) were used for the evaluation of neuromuscular block. Rocuronium (600 μg/kg) was administered for tracheal intubation followed by 200 μg/kg rocuronium top-up doses maintaining T1at 25%. The lungs were mechanically ventilated with 40% oxygen in air and adjusted to maintain 30–40 mmHg end-tidal carbon dioxide. Patients were warmed using a forced-hot-air-blanket to maintain esophageal core temperature greater than 36°C. Patient fluid requirements were replaced with crystalloid solutions. Blood loss, estimated from swab/drape weighing and suction bottles, was replaced by 6% hydroxyethyl starch 130/0.4. Because exogenous catecholamines were shown to evoke changes in BIS readings,11,12patients who required exogenous catecholamine administration such as ephedrine or phenylephrine were excluded from the study. A stable BIS 40 for deep surgical anesthesia13was maintained with propofol TCI ± 0.2 μg/ml rate adjustments.

An assigned anesthesia nurse, the only one with access to the randomization code, prepared 0.5 mg of flumazenil or saline in identical syringes. All anesthesiologists were blinded to the group allocation. Forty-five minutes before the expected completion of the surgical procedure, remifentanil infusion was kept constant in the two groups at 0.15 μg · kg−1· min−1, corresponding to remifentanil 3.75 ηg/ml plasma TCI.13Patients were allowed to recover spontaneously from neuromuscular block up to 0.9 TOF ratio. Thirty minutes before the expected completion of the surgical procedure, a blinded research assistant administered the study drug. Mean arterial pressure and heart rate were recorded before and after study drug administration and during recovery from anesthesia. After discontinuation of anesthesia, a dedicated blinded anesthesiologist used a uniform, one verbal command every 20 s method to assess times to spontaneous respiration, eye opening/hand squeezing on verbal command, extubation, and date of birth recollection.

Statistical Analyses

The effect of flumazenil on BIS monitoring in unpremedicated patients was not previously examined; therefore, an a priori  sample size power analysis was not possible. On the basis of the first 20 pilot patients in whom mean BIS increase after flumazenil administration was 11 ± 9 compared with a BIS increase of 3 ± 5 in placebo patients, we performed an interim power analysis t  test (α= 0.05); this showed that a group size of 19 patients would be required to reveal a statistically significant difference between the two groups with 90% power. We then increased our sample size to 30 patients to match other previously published studies of doxapram and aminophylline stimulant analeptic effects.14,15We used a two-way analysis of variance (ANOVA) model (group × time) to compare parameter differences between the two groups. Dunnett two-sided multiple-comparison post hoc  test was used to compare BIS values at different time points. We used a simple model to compare recovery parameter differences between the two groups, namely the mean recovery parameters time, which comprised mean values of five recovery parameters (time to spontaneous breathing, eye opening/hand squeezing on verbal command, extubation, and date of birth recollection). Data were expressed as means ± SD, and P < 0.05 was considered statistically significant. Statistical analyses were performed using Number Crunching Statistical System 2007 (NCSS Inc., Kaysville, UT) and StatXact (Cytel Software Corporation, Cambridge, MA).

There was no significant difference between the two groups in patient characteristics, time to loss of consciousness, propofol/BIS values at loss of consciousness, duration of surgical procedures, surgical blood loss, and propofol TCI requirements for stable BIS 40 (table 1). Electromyography throughout the recordings was below 35 dB. Data were normally distributed. Two-way ANOVA (group × time) revealed a significant difference (P < 0.0001) between the two groups over time in BIS values (fig. 1). The degree-of-freedom was 1, 14, 14, and the F ratio was 76.8, 98.2, and 53.8 for the group-, time-, and group-time-effect, respectively. Dunnett two-sided multiple-comparison post hoc  test revealed a significant difference between the two groups starting 6 min after flumazenil administration. Mean recovery parameters time was significantly shorter (P = 0.0002) in the flumazenil group compared with the control group (table 2).

Although mean arterial pressure and heart rate before flumazenil administration (86 ± 10 mmHg, 64 ± 11) decreased slightly 15 min after flumazenil administration (82 ± 13 mmHg, 61 ± 8), there were no significant differences over time between the two groups. We encountered no serious cardiovascular events after flumazenil administration.

Our main finding that flumazenil administration in unpremedicated patients results in a significant increase in BIS value during propofol/remifentanil anesthesia could be attributed to flumazenil partially antagonizing the sedative effects of general anesthesia. Here we can speculate that, although propofol is not thought to bind to the benzodiazepine site on the γ-aminobutyric acid receptor, it seems possible that flumazenil could have interacted in some way with the γ-aminobutyric acid receptor to reverse the effect of propofol. Earlier report suggest that flumazenil lacks intrinsic activities; 0.3 mg of flumazenil administered after midazolam led to a prompt restoration of acoustical and somatosensory-evoked cortical responses in volunteers, but flumazenil administered without benzodiazepines did not change acoustical or somatosensory-evoked cortical responses.16Nevertheless, our observed BIS arousal effect could still be attributed in part to an intrinsic direct CNS stimulant effect, as BIS values attained significant difference between-groups shortly after flumazenil administration. Although flumazenil has hitherto been clinically described as an agent with few intrinsic properties with a primary effect lying in its ability to reverse benzodiazepines, we have shown that flumazenil might possess some intrinsic CNS stimulant activity that could result in an enhanced arousal from general anesthesia.

Another hypothesis proposes flumazenil inhibition of γ-aminobutyric acid-benzodiazepine receptor complex of endogenous benzodiazepine-like ligands (endozepines) in patients who had not received benzodiazepines, thus resulting in a significant increase in BIS values. The discovery of high-affinity benzodiazepine binding sites in human cerebral cellular tissues17marked the beginning of a search for putative endogenous receptor ligands according to a review article by Sand et al .18Naturally occurring endozepines that did not result from environmental contamination with pharmaceutical benzodiazepines have been reported in different mammalian tissues.19Endogenous benzodiazepines in potentially physiologic concentrations were demonstrated in human cerebrospinal fluid,5serum,20plasma,21urine,4ultrafiltrates4of patients who had not received benzodiazepines, and breast milk of healthy, newly delivered women who were not taking benzodiazepines.6,22These comprised a variety of 1,4-benzodiazepines such as diazepam, N -desmethyldiazepam, oxazepam, and lorazepam that corresponded to commercially available drugs.18In 1990, proof was given of genuine benzodiazepines in human brains that had been conserved in paraffin since the 1940s, i.e. , well before the era of industrial benzodiazepine synthesis.23 

The synthesis of naturally occurring benzodiazepines by vegetal cells had been originally suggested in 1987,24and it has been supported by recent compelling evidence from sterile cultures showing that wheat and potatoes could produce pharmacologically active benzodiazepine molecules during germination.25,26Moreover, several benzodiazepines that are not currently available for therapeutic use, such as deschlorodiazepam and isodiazepam, have been isolated from plants in support of a biosynthetic pathway, resulting in the de novo  formation of benzodiazepines.25Natural benzodiazepines, such as delorazepam and temazepam, found in soil, plant, and animal tissues, are virtually indistinguishable from benzodiazepines of industrial origin in terms of chemical structure and pharmacological activity.18To date, the confirmation by mass spectrometric analyses25,26of the occurrence of at least nine different pharmacologically active natural 1,4-benzodiazepines in a number of plants and nutritive plant products commonly used for human consumption, such as potato tuber, wheat, rice, soy beans, cherries, maize, mushrooms, lentils, and grapes,25,27,28strongly suggests that these agents may be already incorporated in the food chain.18Although the levels of these “natural” endozepines might be quite low, reversing endogenous (or exogenous from food) benzodiazepine-like substances could be another possible mechanism.

We used BIS monitoring as a surrogate parameter suggesting a lighter plane of anesthesia after flumazenil administration. Likewise, previous studies have used BIS monitoring to quantify the analeptic effect of CNS stimulants. They demonstrated a significant increase of BIS values after doxapram/aminophylline administration during steady-state inhalational anesthesia.14,15This still raises the question of the real clinical relevance of our observations; analeptics or CNS stimulants are not commonly used in current anesthetic practice in the light of the predictable pharmacokinetic profile of modern anesthetics with short offset times. We believe that our study could still provide clinically relevant information regarding arousal by a readily available drug that could be used on a case-by-case basis in patients whose recovery from anesthesia is unexpectedly prolonged.

In our study, 45 min before the expected completion of surgical procedure, we kept remifentanil infusion constant in both groups at 0.15 μg · kg−1· min−1, corresponding to remifentanil 3.75 ηg/ml plasma TCI; this probably kept the two study groups as equal as is clinically possible. Remifentanil allows rapid recovery with a very short context-sensitive half time of 3–4 min independent of the duration of infusion or the total dose given.29Whereas opioids in the usual clinical doses were shown not to affect BIS monitoring; unlike intravenous or inhalational anesthetics, opioids in analgesic concentrations produce minimal or no electrophysiological alterations on the cerebral cortex.30,31Shafer and Varvel clearly demonstrated that opioid doses of almost five times the analgesic concentrations would be required for the appearance of a noticeable electroencephalographic depression30due to the fact that noncortical structures undetectable by the electroencephalograph, such as locus coeruleus-noradrenergic system, are involved in the mechanism of opioid drug effect.31As a result, the opioids we used in our study were unlikely to confound our results.

Electromyographic activities are artifact signals that occur within the frequency “range of interest” of the bispectrum and could simulate the BetaRatio, one of the BIS component descriptors that would be misinterpreted by the BIS algorithm as electroencephalographic activity.32In our study, electromyographic values were below what could be considered the cutoff value of 35 dB,33clearly indicating that high electromyographic activity did not confound the results of our study.

Our findings of a slight, nonsignificant decline in mean arterial pressure and heart rate after flumazenil administration concur with two previous double-blind, randomized placebo-controlled studies in which 0.2–0.5 mg of flumazenil administration reduced mean arterial pressure by 3–5 mmHg and heart rate by 4–7 beats/min in healthy volunteers who did not receive benzodiazepines.34,35We encountered no serious cardiovascular events after flumazenil administration. Although ventricular arrhythmias have been reported with flumazenil administration in two benzodiazepine-overdose comatose patients,36,37we are unaware of any serious cardiovascular adverse events being reported with flumazenil administration in patients who had not received benzodiazepines.

Animal studies have shown that flumazenil administration could potentially increase cerebral blood flow and intracranial pressure during the reperfusion phase after incomplete global ischemia.38This could represent a serious limitation for the use of flumazenil in patients with compromised intracranial physiology. However, the design of our study that included only patients without neurologic conditions did not allow us to explore such an effect.

In conclusion, flumazenil administration during propofol/remifentanil anesthesia in healthy unpremedicated patients significantly increased the BIS value and allowed earlier emergence from anesthesia. This may indicate that flumazenil, on a case-by-case basis, could be useful for reversing endogenous or exogenous endozepines that might play a role during anesthesia.

The authors thank Ms. Eugenia Lamont, Medical Editor, Section for Surgical Research, Department of Surgery, University Hospital/Medical University of Graz, Graz, Austria, for her great help with reading and streamlining the manuscript.

Amrein R, Hetzel W, Hartmann D, Lorscheid T: Clinical pharmacology of flumazenil. Eur J Anaesthesiol Suppl 1988; 2:65–80
Goulenok C, Bernard B, Cadranel JF, Thabut D, Di Martino V, Opolon P, Poynard T: Flumazenil versus  placebo in hepatic encephalopathy in patients with cirrhosis: A meta-analysis. Aliment Pharmacol Ther 2002; 16:361–72
Rothstein JD, Garland W, Puia G, Guidotti A, Weber RJ, Costa E: Purification and characterization of naturally occurring benzodiazepine receptor ligands in rat and human brain. J Neurochem 1992; 58:2102–15
Beaumont K, Cheung AK, Geller ML, Fanestil DD: Inhibitors of peripheral-type benzodiazepine receptors present in human urine and plasma ultrafiltrates. Life Sci 1983; 33:1375–84
Deckert J, Kuhn W, Przuntek H: Endogenous benzodiazepine ligands in human cerebrospinal fluid. Peptides 1984; 5:641–4
Dencker SJ, Johansson G, Milsom I: Quantification of naturally occurring benzodiazepine-like substances in human breast milk. Psychopharmacology (Berl) 1992; 107:69–72
Dahaba AA: Different conditions that could result in the bispectral index indicating incorrect hypnotic state. Anesth Analg 2005; 101:765–73
Moher D, Schulz KF, Altman DGfor the CONSORT group: The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomised trials. Lancet 2001; 357:1191–4
for the CONSORT group
Noseir RK, Ficke DJ, Kundu A, Arain SR, Ebert TJ: Sympathetic and vascular consequences from remifentanil in humans. Anesth Analg 2003; 96:1645–50
Marsh B, White M, Morton N, Kenny GNC: Pharmacokinetic model driven infusion of propofol in children. Br J Anaesth 1991; 67:41–8
Andrzejowski J, Sleigh JW, Johnson AT, Sikiotis L: The effect of intravenous epinephrine on the bispectral index and sedation. Anaesthesia 2000; 55:761–3
Ishiyama T, Oguchi T, Iijima T, Matsukawa T, Kashimoto S, Kumazawa T: Ephedrine but not phenylephrine, increases bispectral index values during combined general and epidural anesthesia. Anesth Analg 2003; 97:780–4
Hoymork SC, Raeder J, Grimsmo B, Steen PA: Bispectral index predicted and measured drug levels of target-controlled infusions of remifentanil and propofol during laparoscopic cholecystectomy and emergence. Acta Anaesthesiol Scand 2000; 44:1138–44
Wu CC, Lin CS, Wu GJ, Lin YH, Lee YW, Chen JY, Mok MS: Doxapram and aminophylline on bispectral index under sevoflurane anaesthesia: A comparative study. Eur J Anaesthesiol 2006; 23:937–41
Wu CC, Mok MS, Chen JY, Wu GJ, Wen YR, Lin CS: Doxapram shortens recovery following sevoflurane anesthesia. Can J Anaesth 2006; 53:456–60
Kochs E, Wüst P, Blanc I, Schulte am Esch J: Acoustic and somatosensory evoked cortical potentials in sedation with midazolam and drug antagonism with flumazenil. Anasth Intensivther Notfallmed 1989; 24:49–56
Moehler H, Okada T: Benzodiazepine receptor: Demonstration in the central nervous system. Science 1977; 198:849–51
Sand P, Kavvadias D, Feineis D, Riederer P, Schreier P, Kleinschnitz M, Czygan FC, Abou-Mandourm A, Bringmann G, Beckmann H: Naturally occurring benzodiazepines: Current status of research and clinical implications. Eur Arch Psychiatry Clin Neurosci 2000; 250:194–202
Medina JH, Pena C, Piva M, Paladini AC, De Robertis E: Presence of benzodiazepine-like molecules in mammalian brain and milk. Biochem Biophys Res Commun 1988; 165:547–53
Duthel JM, Constant H, Vallon JJ, Rochet T, Miachon S: Quantitation by gas chromatography with selected-ion monitoring mass spectrometry of “natural” diazepam, N-desmethyldiazepam and oxazepam in normal human serum. J Chromatogr 1992; 579:85–91
Wildmann J, Niemann J, Matthaei H: Endogenous benzodiazepine receptor agonist in human and mammalian plasma. J Neural Transm 1986; 66:151–60
Pena C, Medina JH, Piva M, Diaz LE, Danilowicz C, Paladini AC: Naturally occurring benzodiazepines in human milk. Biochem Biophys Res Commun 1991; 175:1042–50
Klotz U: “Natural” benzodiazepines in man (letter). Lancet 1990; 335:922
Wildmann J, Moehler H, Vetter W, Ranalder U, Schmidt K, Maurer R: Diazepam and Ndesmethyldiazepam are found in rat brain and adrenal and may be of plant origin. J Neural Transm 1987; 70:383–98
Wildmann J: Increase of natural benzodiazepines in wheat and potato during germination. Biochem Biophys Res Commun 1988; 157:1436–43
Kavvadias D, Abou-Mandour AA, Czygan FC, Beckmann H, Sand P, Riederer P, Schreier P: Identification of benzodiazepines in Artemisia dracunculus and Solanum tuberosum rationalizing their endogenous formation in plant tissue. Biochem Biophys Res Commun 2000; 269:290–5
Unseld E, Krishna DR, Fischer C, Klotz U: Detection of desmethyldiazepam and diazepam in brain of different species and plants. Biochem Pharmacol 1989; 38:2473–8
Klotz U: Occurrence of “natural” benzodiazepines. Life Sci 1991; 48:209–15
Kapila A, Glass PSA, Jacobs JR, Muir KT, Hermann DJ, Shiraishi M, Howell S, Smith RL: Measured context-sensitive half-times of remifentanil and alfentanil. Anesthesiology 1995; 83:968–75
Shafer SL, Varvel JR: Pharmacokinetics, pharmacodynamics, and rational opioid selection. Anesthesiology 1991; 74:53–63
Pan Y, Li D, Chen S, Pan H: Activation of μ-opioid receptors excites a population of locus coeruleus-spinal neurons through presynaptic disinhibition. Brain Res 2004; 997:67–78
Sleigh JW, Steyn-Ross DA, Steyn-Ross ML, Williams ML, Smith P: Comparison of changes in electroencephalographic measures during induction of general anaesthesia: Influence of the gamma frequency band and electromyogram signal. Br J Anaesth 2001; 86:50–8
Mathews DM, Kumaran KR, Neuman GG: Bispectral index-derived facial electromyography-guided fentanyl titration in the opiate-exposed patient. Anesth Analg 2003; 96:1062–4
Stroehle A, Kellner M, Holsboer F, Wiedemann K: Behavioral, neuroendocrine, and cardiovascular response to flumazenil: No evidence for an altered benzodiazepine receptor sensitivity in panic disorder. Biol Psychiatry 1999; 45:321–6
Neave N, Reid C, Scholey AB, Thompson JM, Moss M, Ayre G, Wesnes K, Girdler NM: Dose-dependent effects of flumazenil on cognition, mood, and cardio-respiratory physiology in healthy volunteers. Br Dent J 2000; 189:668–74
Marchant B, Wray R, Leach A, Nama M: Flumazenil causing convulsions and ventricular tachycardia. BMJ 1989; 299:860
Short TG, Maling T, Galletly DC: Ventricular arrhythmia precipitated by flumazenil. BMJ 1988; 296:1070–1
Kochs E, Roewer N, Peter A, Schulte am Esch J: Effect of flumazenil on global cerebral blood flow and on intracranial pressure in the reperfusion phase following incomplete global cerebral ischemia. Anasth Intensivther Notfallmed 1988; 23:159–62