Benzodiazepine Site Map

Journal of Psychopharmacology 6(3) (1992) 357-363

A pilot study of the effects of flumazenil on symptoms persisting after benzodiazepine withdrawal

Malcolm H. Lader and Sally V. Morton

Institute of psychiatry, De Crespigny Park, Denmark Hill, London SE5 8AF, UK

The potential of the benzodiazepine antagonist flumazenil (Ro 15-1788) to lessen persisting benzodiazepine withdrawal symptoms was demonstrated in 11 patients who had been drug free for between 1 month and 5 years. Doses ranging from 0.2 to 2.0 mg divided into three intravenous injections over a few hours relieved long-standing symptoms to varying extents. These included clouded thinking, tiredness, muscular symptoms such as neck tension, cramps and shaking and the characteristic perceptual symptoms of benzodiazepine withdrawal, namely, pins and needles, burning skin, pain and subjective sensations of bodily distortion. Mood disorder, when present, also improved but the reduction in anxiety and depression may have reflected relief of physical symptoms. The onset of maximum response was sometimes delayed by as much as a day but was usually prompt. Side effects were reported to be either absent or typically described as lightheadedness or dizziness, lasted only a few minutes and were usually well tolerated. The benefits last between a few hours and several days despite flumazenil's otherwise short duration of action. However, symptoms did return to varying degrees in most cases, suggesting the need for repeated doses.

Key words: flumazenil; benzodiazepine withdrawal; benzodiazepine antagonist


It is now widely accepted that treatment with a benzodiazepine is often accompanied by dependence, as manifested by a physical withdrawal syndrome (American Psychiatric Association Task Force Report, 1990). Hence authoritative recommendations on limitations of their use to 2-4 weeks and tapering when stopping (Committee on Safety of Medicines, 1988). Rebound anxiety can occur after abrupt discontinuation of therapeutic doses taken for only a few weeks (Fontaine, Chouinard and Annable, 1984). More specific withdrawal symptoms, particularly perceptual disturbances, were identified over 10 years ago in long-term users stopped abruptly (Tyrer, Rutherford and Hugget, 1981) and in those withdrawn over a 2-week period (Petursson and Lader, 1981). In most instances the withdrawal is short-lived and mild especially if a long-acting benzodiazepine is is tapered off after long-term use (Busto et al., 1986; Schweizer et al., 1990). However, in some patients the symptoms are quite severe and relapse back onto medication is common (Rickels et al., 1990b).

In an unfortunate minority, various symptoms of withdrawal have been noted to continue for months or even years (Ashton, 1984, 1987; Golombok et al., 1987; Higgitt, Lader and Fonagy, 1985; Higgitt et al., 1990; Rickels et al., 1990a). There is some debate about this group of patients with the so called 'benzodiazepine post-withdrawal syndrome' Tyrer, 1991) as to whether it truly reflects a continuing disturbance arising from the dependence process. The alternative view is that this phase, which patient groups and the media have incorrectly ascribed directly to prior drug use, should be regarded as a new episode of illness or as a recrudescence of the origional neurotic disorder for which benzodiazepines were prescribed. However, against this is the observation that patients report their symptoms in prolonged withdrawal to be similiar to those experienced earlier in withdrawal but different to those of their previous illness. Also in support for an iatrogenic cause is the finding that nine patients suffering from prolonged withdrawal symptoms, 5-42 months after stopping benzodiazepines, did not have psychophysiologial reactions characteristic of either anxiety or conversion hysteria but, conversely, were hard to distinguish from a normal control group (Higgitt et al., 1990). Some, though not all, patients with prolonged withdrawal problems develop significant depression typically within the first 3 months of abstinence (Olajide and Lader, 1984). However, the prolonged withdrawal syndrome and return of anxiety probably occur commonly together.

Although the underlying mechanisms of benzodiazepine dependence are still not entirely understood, research in animals points to the clinical potential for the benzodiazepine antagonist flumazenil (Ro 15-1788) to reverse benzodiazepine dependence and tolerance and prevent withdrawal (Whitwam, 1988). Administration of flumazenil reverses tolerance and dependence to benzodiazepines (Gonsalves and Gallager, 1985) but precipitates recognizable withdrawal symptoms . However, if given during chronic treatment flumazenil can, by similarly reversing tolerance, prevent subsequent withdrawal syndromes in primates (Gallager, Heninger and Heninger, 1986) and in rats (Baldwin, Hitchcott and File, 1990). One explanation for this is the antagonism and depletion of an endogenous benzodiazepine receptor ligand with inverse agonist and thus anxiogenic activity by the antagonist flumazenil (Baldwin, Hitchcott and File, 1990). However, levels of the proposed 'anxiety peptide' ligand associated with diazepam binding inhibitor (DBI) were not found to be increased by the administration of diazepam in rats (Ball et al., 1987). An alternative explanation is that chronic agonist use causes a persistent conformational change and thus a shift in benzodiazepine receptor efficacy in the direction of inverse agonist function (Little, Nutt and Taylor, 1987) and that flumazenil resets the receptor's sensitivity (Nutt and Costello, 1988). In binding to the benzodiazepine receptor flumazenil may alter the coupling of the elements of the GABA/benzodiazepine macromolecular complex modified by benzodiazepine binding, thus restoring the GABA recognition site to its pre-drug affinity (Gonsalves and Gallager, 1985).

Very little parallel work has been done in man. Healthy volunteers given lorazepam for 1 or 2 weeks developed various features of tolerance which were not affected by a small dose of flumazenil (0.2 mg) (Cittadini and Lader, 1991) but were reversed by a larger dose (3.0 mg) (Evans et al., 1990).

Because we considered that flumazenil might, by displacing the benzodiazepine, induce an abrupt withdrawal in chronic users with the risk of acute anxiety, psychosis or fits, we deemed it too hazardous to attempt to prevent subsequent withdrawal by prior administration of flumazenil. Instead we studied the effects of flumazenil in patients with symptoms persisting after they had become drug free.



The study was conducted at the maudsley Hospital over a 2 year period starting in 1989. The protocols were approved by the appropriate Ethics Committee and fully informed written consent was obtained from each patient. Since this was not a licensed use of flumazenil, the study was run on a 'named patient' basis.

Patients, aged between 18 and 65 and of either sex, were offered flumazenil if they had stopped taking long-term therapeutic doses of benzodiazepines at least 3 weeks, but preferably several months or years, before and were considered by both themselves and the investigator to be suffering primarily from continuing withdrawal symptoms such as anxiety and perceptual and muscular symptoms (Table 1) for which they were seeking treatment. Exclusion criteria included any psychotropic medication and serious physical disorder. Some of the demographic details and benzodiazepine history relating to these patients are available in Table 3.

Drug administration

All drug administrations took place on a research ward where the subjects came for the day or were in-patients, depending on clinical and geographic need. Emergency medical cover including i.v. diazepam was available but never required. A venflon or butterfly cannula was sited in an arm vein and kept patent with boli of heparin. Bolus doses of flumazenil were injected at a rate of 0.2 mg (i.e. 2 ml) per 15 s. A small test dose was always given first preceded by a single blind placebo bolus of saline matched in volume to identify immediate placebo responders.

For the first four patients the total dose was small (0.5 mg) and given in a series of boli over a few hours. Flumazenil (F) was mixed with saline so that each of three divided doses were the same volume and a further three were saline placebo (P) only. Thus the following was given single-blind: P, F 0.1 mg, P, F 0.2 mg, F 0.2 mg, P; total F 0.5 mg with 30 min between each dose.

Following encouraging results with one patient but only limited response in three others to the small dose, the protocol was expanded to administer a total dose of 2.0 mg to a further six patients. The series followed was: P, F 0.2 mg, F 0.8 mg, F 1.0 mg; total F 2.0 mg, as before all on the same day. Thus the first of the divided doses was again given single-blind after saline in matching volumes of 2 ml to compare drug effect with placebo response but the other two larger doses were given un-blind and undiluted. An 11th patient was treated with 2.0 mg to replace patient 5 who took only 0.2 mg in total.

In a further extension of the study, a second dose was given a few days or weeks later to three of those patients who were found to have gained relief from the first dose but whose symptoms had returned, and a year later to a fourth who had not originally responded. Due to the exploratory nature of the study the doses given were not identical but totalled 2.5 mg in two cases and 2.0 mg for the other two patients. A third dose of 2.0 mg was given to patient 10, making a total of 16 treatments on various occasions given to 11 patients.

A period was left between each of the divided doses for observation of the onset and duration of immediate side effects and benefits. At first 30 min was allowed but it became apparent that side effects were invariably immediate and subsided within 5 min. Benefits were also immediate with no obvious further progress after ~ 15 min, or were delayed for many hours. The minimum interval was therefore halved to 15 min.


For entry into the study patients were required to have no clinical significant abnormality on standard blood testing of haematological, renal, liver and thyroid function and a negative urine drug screen for benzodiazepines and opioids. The Hamilton Anxiety Rating Scale (HAM-A) (Hamilton, 1959) and self-rated Hospital Anxiety and Depression Scale (HAD) (Zigmond and Snaith, 1983) were completed to give baseline comparisons of anxiety and depression. The Eyssenck Personality Questionnaire (EPQ) giving scores for psychoticism (P), extraversion (E), neuroticism (N) and for the 'Lie Scale' (L) (Eysenck and Eyseck, 1975) and Locus of Control of Behaviour (LCB) (Craig, Franklin and Andrews, 1984) were also completed to assess personality characteristics.

A 10-item list of persistent benzodiazepine withdrawal symptoms comprising seven muscular and perceptual symptoms, one overall rating and two mood variables was patient-related for severity on a 9-point scale (Table 1).The somatic symptoms were chosen as those most commonly presented by our patients as being peculiar to their illness since benzodiazepine withdrawal. Emphasis was also placed on selection of those items which corresponded to the muscular and perceptual hypersensitivity symptoms reported from other studies as occurring during withdrawal (Ashton, 1984; Pertursson and Lader, 1981; Tyrer, Owen and Dawling, 1983). Validated benzodiazepine withdrawal rating scales have since been published (Busto, Sykora and Sellers, 1989; Tyrer, Murphy and Riley, 1990) but were not available at the start of the study and were not designed for use over the time scale of this type of study. The items on the list that were present at baseline were rated regularly throughout the period of dosing and for follow-up until no further changes were noted. When rating the 'overall' item patients were allowed to refer to all their physical and mental symptoms thereby including some probably unrelated to withdrawal but important to them.

A record was made of the nature, severity, precise timing of onset from the start of the injection and duration of any immediate side effects reported.


For analysis of the 10 rated symptoms the scores for the combination of the seven somatic symptoms presented by each patient plus the 'overall' item were averaged to give a 'somatic mean' score. The effectiveness of flumazenil for each of the 16 treatments given to 11 patients was then assessed by calculating the percentage reduction in this somatic mean from baseline, but after any placebo response, to the point of maximum relief. This measure of response was compared with clinical background and flumazenil dose to identify predictive factors.

The average effect of flumazenil on the somatic mean and on each of the 10 items on the list was similarly calculated from baseline to peak effect using the scores for all treatments in which the symptom was present.

The 10-item list was thus not used as a rating scale and is not proposed as a measure of withdrawal severity.



Both somatic symptoms and mood responded to flumazenil as assessed by the self-rated 10-symptom list (Table 1). On all but one of the 16 treatment occasions, tense head or neck was presented as a symptom as was tiredness for 12 treatments. The other five symptoms rated were presented by between six and nine patients. The symptoms varied in the extent of their severity and response to flumazenil (Fig. 1) with the mean score for pins and needles only reducing by 27% whereas shaking improved 82% (Table 1). The average somatic mean relief was 48% but with a wide variation from 5 to 100%. Self-rated mood improvements averaged 57% for anxiety ('self anx' in Table 3) and 70% for depression ('self dep'). Patients had tended to report variability in the severity of their symptoms in the past but the flumazenil, if effective, relieved these to a greater extent than ever previously experienced. The most noticeable response was the early onset of muscular relaxation, particularly of neck tension.

Flumazenil Table 1

In addition to the decrease in ratings of mood and somatic symptoms listed in Table 1, individual patients also reported improvement in depersonalisation, calmness, alertness, unclear eyes and insomnia and in subjective distortion, numbness and pain in the face and throat.


Flumazenil Figure 1

Timing of response

Benefits from flumazenil generally started to be felt within a few minutes of drug administration and built up to a peak effect over a few hours (Table 2). Maximum relief usually lasted for the rest of the day with further partial relief for a few days. For most patients the benefit gradually faded but a few patients gave accounts of some symptoms returning more suddenly or forcefully the next day.

Repeated doses given between 4 days and 5 weeks later resulted in some additive and increasingly persistent effect though they had less immediate effect in terms of score reduction (Table 3). This may be related to either reduced baseline score or increased expectation of relief each time. Several patients who experienced temporary relief each time were reluctant to have further doses because they could not face the possible return of symptoms afterwards.

Flumazenil Table 2


Since only one of the first four patients had shown much response the total dose was raised from 0.5 mg to 2.0 mg. Doses of 1.5 or 2.0 mg tended to be associated with at least a 40% improvement but no clear dose-effect relationship was apparent (Table 3).

Factors associated with response

The demographic details, benzodiazepine (BZ) history, and clinical measurements before dosing from all the mood, somatic symptom and personality rating scales were tabulated against ranked individual maximum relief from flumazenil (Table 3). These factors varied widely within the patient group but none were significant predictors of effectiveness. However, Fig. 2 shows that greater response (p<0.07) tended to be associated with a lower neuroticism (N) score on the EPQ. This was further demonstrated by patient 4 whose muscular symptoms responded much better to a second dose of flumazenil 1 year after her first, having in the meantime been successfully treated for the affective component of her withdrawal syndrome, resulting in reduction of her previously high neuroticism score (Table 3).


Flumazenil Table 3

Flumazenil Figure 2

Side effects

Adverse effects of varying severity were observed immediately after injection of the flumazenil. In many cases these were experienced within a minute of the start of administration and before the whole bolus had been given. Most such symptoms lasted between 1 and 5 min, usually 3 min. Patients often had difficulty describing these adverse effects but the typical impression given was of a sudden rush to the head of a mildly unpleasant sensation most often described as lightheadedness or dizziness. Several patients said it was reminiscent of the effects of i.v. and oral dyes used in radiological procedures. Only patient 3 said that the temporarily increased burning in her ears and neck and jittery legs were like a return of her acute withdrawal symptoms last experienced a year before. Patient 5 was so alarmed by his dizziness, tight chest and dimmed vision that he refused further doses. It appeared clinically that the sudden side effect of dizziness had triggered a panic attack accounting for all his other symptoms. This phenomenon was observed to a small degree with a few other patients. However, most patients felt that the immediate adverse effects were outweighed by the benefits and said they would recommend others to try the treatment.

The single-blind placebo control of the first dose was useful in establishing that on the whole the patients were not placebo responders. Immediate side effects were sometimes reported on placebo but were invariably more typical of anxiety reactions and distinct from the symptoms reported on flumazenil.


Our impression from this, our first experience in patients, is that the benzodiazepine antagonist flumazenil does have potential to treat persisting benzodiazepine withdrawal symptoms. To varying degrees, sometimes dramatically, it reduced both physical and mental symptoms which had persisted for months or even years after withdrawal. Musculo-perceptual symptoms such as muscle tension, spasms, shakes and burning halved in intensity. In the range tested of 0.2-2.0 mg, response was not obviously related to dose.

Only the immediate response to the first divided dose was placebo controlled and then only single-blind and our observations also suffer from missing data and lack of standardization due to the ever developing exploratory nature of this pilot study. However, there are some consistent findings which do not support the theory of an antagonist correcting a persisting receptor dysfunction. Although flumazenil has weak agonist properties at high dose (Higgitt, Lader and Fonagy, 1987) and its post-anaesthetic reversal of sedation is immediate and lasts only a few hours depending on doses (Dunton et al., 1988). Yet maximum response in our patients usually developed more gradually and invariably lasted very many times longer, all suggesting that flumazenil was not acting purely as an anxiolytic.

The role of patient expectation is difficult to assess. They were all hoping for a 'cure' but were told that this was only a theoretical possibility and that the effects of flumazenil in this situation were unknown. Hence no suggestion of a time course was given when obtaining consent. However, it was generally the less, and not more, neurotic patients who responded, suggesting that it was not a placebo effect. It is not clear whether the mood improvements were a direct drug effect or reaction to the relief of symptoms.

As we were experimenting with a non-licensed indication, we exercised the utmost caution in the first patients using only 0.5 mg. However, later patients easily tolerated the higher dose of 2.0 mg. As in normal subjects (Amrein et al., 1987), side effects were typically lightheadedness or dizziness. The side effects were mild for most patients though seemed to lead to panic in some. In one study (Nutt et al., 1990) panic attacks were induced by flumazenil 2.0 mg in 8/10 panic disorder patients but in none of 10 controls and were considered to be an anxiogenic effect rather than a catastrophic reaction to an unpleasant side effect. The burning or heat reported by three patients may well correspond to the flushing side effect reported by human subjects given 5-6 mg flumazenil (Amrein et al., 1987) rather than being a return of benzodiazepine withdrawal as one of our patients asserted. We were alerted to the possibility of flumazenil exacerbating withdrawal but do not consider that this was observed.

Our experiences suggest that a course of doses might be required for more sustained remission rather than a single treatment as has been successful in animal models. However, the animal studies typically utilized only a few weeks of administration of fairly high doses of benzodiazepines and this may not be an appropriate model for human low-dose long-term usage. Whether an antagonist could be administered routinely to benzodiazepine-dependent patients before discontinuation to prevent withdrawal, as animal studies predict (Gallager, Heninger and Heninger, 1986; Baldwin, Hitchcott and File, 1990), remains to be tested since the risk of severe withdrawal must be judged high. This was demonstrated by the precipitation of panic in two patients with a history of panic disorder and of high anxiety in a third chronic user of diazepam given flumazenil to test their level of dependence (Bernik, Gorenstein and Gentil, 1991). However, the finding that flumazenil could be given to benzodiazepine-dependent epileptic patients without precipitating fits but rather re-instating the efficacy of their benzodiazepine as an anticonvulsant by reducing tolerance, is encouraging (Savic, Widen and Stone-Elander, 1991). Their further finding that a dose of 1.5 mg produces optimum receptor occupancy when given with clonazepam also corresponds with our finding that a dose of 1.5-2 mg of flumazenil was clinically effective. To investigate further the potential of flumazenil we are handicapped by the unavailability of an oral preparation, although one was under study some years ago (Higgitt, Lader and Fonagy, 1986). It suffered the disadvantage of low bio-availability but we hope that oral flumazenil or other benzodiazepine antagonists such as B-carbolines will be developed. From these preliminary observations, placebo-controlled testing of flumazenil both pre-withdrawal and in the prolonged withdrawal syndrome would appear worthwhile.

Address for correspondence

M. H. Lader

Institute of Psychiatry

Denmark Hill

London SE5 8AF



American Psychiatric Association Task Force Report (1990) Benzodiazepine dependence, toxicity and abuse. Washington DC

Amrein R, Leishman B, Bentzinger C, Roncari G (1987) Flumazenil in benzodiazepine antagonism, actions and clinical use in intoxications and anaesthesiology. Med Toxicol 2: 411-429

Ashton H (1984) Benzodiazepine withdrawal: an unfinished story. Br Med J 288: 1135-1140

Ashton H (1987) Benzodiazepine withdrawal: outcome in 50 patients. Br Med J 82 665-671

Baldwin H A, Hitchcott P K, File S (1990) The use of flumazenil in prevention of diazepam dependence in the rat. Hum Psychopharmacol 5: 57-61

Ball J A, Burnet P W J, Ghatei M A, Mulderry P K, Bloom S R (1987) Flumazenil-precipitated withdrawal symptoms in chronic users of therapeutic doses of diazepam. J Psychopharmacol 5: 215-219

Busto U, Sellers E M, Naranjo C A, Cappell H, Sanchez-Craig M, Sykora K (1986) Withdrawal reaction after long-term therapeutic use of benzodiazepines. N Engl J Med 315: 854-859

Busto U E, Sykora K, Sellers E M (1989) A clinical scale to assess benzodiazepine withdrawal. J Clin Psychopharmacol 9: 412-416

Cittadini A, Lader M (1991) Lack of effect of a small dose of flumazenil in reversing short-term tolerance to benzodiazepines in normal subjects. J Psychopharmacol 5: 220-227

Craig A R, Franklin J A, Andrews G (1984) A scale to measure locus of control of behaviour. Br J Med Psychol 57: 173-180

Committee on Safety of Medicines (1988) Benzodiazepines, dependence and withdrawal symptoms. Curr Problems 21: 1-2

Dunton A W, Schwam E, Pitman V, McGrath J, Hendler J, Siegel J (1988) Flumazenil: US clinical pharmacology studies. Eur J Anaesthiol Suppl 2: 81-95

Evans S M, Guarino J G, Roach J D, Furman W, Liebson I A, Griffiths R R (1990) The effects of flumazenil (Ro 15-1788) following acute and repeated exposures to lorazepam in healthy volunteers. Abstract Poster, CINP Sept. 10-14, Kyoto, Japan

Eysenck H J, Eysenck S B G (1975) Manual of the Eysenck personality questionaire. Hodder and Stoughton, Sevenoaks, Kent

Fontaine R, Chouinard G, Annable L (1984) Rebound anxiety in anxious patients after abrupt withdrawal of benzodiazepine treatment. Am J Psychiatr 141: 848-852

Gallager D W, Heninger K, Heninger G (1986) Periodic benzodiazepine antagonist administration prevents benzodiazepine withdrawal symptoms in primates. Eur J Pharmacol 132: 31-38

Golombok S, Higgitt A, Fonagy P, Dodds S, Saper J, Lader M (1987) A follow-up study of patients treated for benzodiazepine dependence. Br J Med Psychol 60: 141-149

Gonsalves S F, Gallager D W (1985) Spontaneous and Ro 15-1788-induced reversal of subsensitivity to GABA following chronic benzodiazepines. Eur J Pharmacol 110: 163-170

Hamilton M (1959) The assessment of anxiety states by rating. Br J Med Psychol 32: 50-55

Higgitt A C, Lader M H, Fonagy P (1986) The effects of the benzodiazepine antagonist Ro 15-1788 on psychophysiological performance and subjective measures in normal subjects. Psychopharmacology 89: 395-403

Higgitt A, Fonagy P, Toone B, Shine P (1990) The prolonged benzodiazepine withdrawal syndrome: anxiety or hysteria? Acta Psychiatr Scand 82: 165-168

Little H J, Nutt D J, Taylor S C (1987) Kindling and withdrawal changes at the benzodiazepine receptor. J Psychopharmacol 1: 35-46

Nutt D J, Costello M J (1988) Rapid induction of lorazepam dependence and reversal with flumazenil. Life Sci 43: 1045-1053

Nutt D J, Glue P, Lawson C, Wilson S (1990) flumazenil provocation of panic attacks. Arch Gen Psychiatr 47: 917-925

Olajide D, Lader M (1984) Depression following withdrawal from long-term benzodiazepine use: a report of four cases. Psychol Med 14: 937-940

Petursson H, Lader M H (1981) Withdrawal from long-term benzodiazepine treatment. Br Med J 283: 643-645

Rikels K, Case W G, Schweizer E, Garcia-Espana F, Fridman R (1990a) Benzodiazepine dependence: management of discontinuation. Psychopharmacol Bull 26: 63-68

Rikels K, Schweizer E, Case W G, Greenblatt D J (1990b) Long-term therapeutic use of benzodiazepines. I. Effects of abrupt discontinuation. Arch Gen Psychiatr 47: 899-907

Savic I, Widen L, Stone-Elander S (1991) Feasibility of reversing benzodiazepine tolerance with flumazenil. Lancet 337: 133-137

Schweizer E, Rickels K, Case G, Greenblatt D J (1990) Long-term therapeutic use of benzodiazepines. II. Effects of gradual taper. Arch Gen Psychiatr 47: 908-915

Tyrer P (1991) The benzodiazepine post-withdrawal syndrome. Stress Med 7: 1-2

Tyrer P, Murphy S, Riley P (1990) The benzodiazepine withdrawal symptom questionnaire. J Affect Disorders 19: 53-61

Tyrer P, Owen R, Dawling S (1983) Gradual withdrawal of diazepam after long-term therapy. Lancet i: 1402-1406

Tyrer P, Rutherford D, Huggett T (1981) Benzodiazepine withdrawal symptoms and propranolol. Lancet i: 520-522

Whitwam J G (1988) Flumazenil: a benzodiazepine antagonist. Br Med J 297: 999-1000

Zigmond A S, Snaith R P (1983) Hospital anxiety depression scale. Acta Psychiatr Scand 67: 361-370

Forum Disclaimer SiteMap Home Contact us The Ashton Manual