The benzodiazepine withdrawal syndrome

The FDA emphasizes that ‘physical dependence can occur when benzodiazepines are taken steadily for several days to weeks, even as prescribed’ (FDA Drug Safety Communication, 2020). Withdrawal symptoms from benzodiazepines can develop after just several days or weeks of continuous use and even after intermittent use (FDA Drug Safety Communication, 2020; Horowitz & Taylor, Reference Horowitz and Taylor2024). Normal-dose physical dependence was observed from the 1960s leading to their restriction by the FDA in 1975 (Guina & Merrill, Reference Guina and Merrill2018; Lader, Reference Lader1991). In 1980 the Committee on the Review of Medicines in the UK focused on the issue of physical dependence and withdrawal and recommended that all benzodiazepine patients should have gradual withdrawal (Committee on the Review of Medicines, 1980). The 1980s and 1990s also saw regulatory restrictions for certain benzodiazepines (triazolam in particular) in Europe, Hong Kong, and the United States (Cloos et al., Reference Cloos, Bocquet, Rolland-Portal, Koch and Chouinard2015). With the later entry of the z-drugs, manufacturers also placed warnings of tolerance, physical dependence, and withdrawal symptoms for these agents (Parsons, Reference Parsons2012).

The benzodiazepine withdrawal syndrome includes numerous physical and emotional symptoms due to the myriad effects of benzodiazepines on bodily systems (Box 2). As the withdrawal syndrome includes anxiety, panic, and insomnia, it may not only be misdiagnosed as a relapse of the original condition for which the benzodiazepine was prescribed but also a putative new-onset psychiatric disorder, thereby perpetuating unnecessary long-term treatment (Markota, Rummans, Bostwick, & Lapid, Reference Markota, Rummans, Bostwick and Lapid2016; Moore, Pariente, & Begaud, Reference Moore, Pariente and Begaud2015). Up to 90% of long-term users experience withdrawal effects on dosage reduction or stopping (Schweizer, Rickels, Case, & Greenblatt, Reference Schweizer, Rickels, Case and Greenblatt1990).

Box 2.

The benzodiazepine withdrawal syndrome (Brunner et al., Reference Brunner, Chen, Klein, Maust, Mazer-Amirshahi, Mecca and Lindsay2025; Cosci & Chouinard, Reference Cosci and Chouinard2020; Ng et al., Reference Ng, Le Couteur and Hilmer2018; Sokya, Reference Sokya2017)

Box 2. Long description

At the center of the diagram is a box labeled Benzodiazepine withdrawal syndrome. Arrows point outward from this central box to seven surrounding boxes that categorize specific symptoms.

* Top-Left box: Severe rapid reduction. Symptoms include Catatonia can be fatal, Seizures can be fatal, Delirium tremens, Confusion, Hyperthermia, Mania, Organic brain syndrome, Psychosis e.g. paranoia, Suicidal ideation or suicide, and Violence and aggression.

* Top-Center box: General Somatic. Symptoms include Chest pain, Dry mouth, Gastrointestinal including nausea, diarrhea, vomiting, Headache, Perspiration, Palpitations, Fatigue and weakness, Muscle pain, Sweating, and Elevated blood pressure.

* Top-Right box: Psychological. Symptoms include Agitation and anxiety, Terror and panic attacks, Depersonalisation/derealisation, Depression can be severe, Hypochondriasis, Mood instability, Paranoia, Obsessive compulsive symptoms, Suicidality and self-harm, Irritability, agitation, aggression, and Anxiety.

* Middle-Right box: Sleep Disturbance. Symptoms include Insomnia, Hypersomnia, REM sleep rebound, Nightmares, and Excessive dreaming.

* Lower-Right box: Autonomic. Symptoms include Postural hypotension, Hypertension, and Tachycardia.

* Bottom-Right box: Cognitive. Symptoms include Confusion, Impaired concentration, and Impaired memory.

* Bottom-Center box: Sensory symptoms. Symptoms include Hearing disturbance, Dysesthesia, Neuropathic pain, Tingling, numbness, altered sensation, Visual disturbances, Tinnitus, Sensory hypersensitivity light, sounds, taste, smell, and Paraesthesia.

* Middle-Left box: Neurological/ neuromuscular symptoms. Symptoms include Akathisia, restlessness, Ataxia, Blurred vision, Dilated pupils, Dizziness, Hypnagogic hallucinations, Mild to moderate aphasia, Muscular spasms, cramps, discomfort or fasciculations, Myoclonus, Tremor, Hot and cold spells, Photophobia, Restless legs syndrome, Co-ordination, balance problems, Stiffness, Increased urinary frequency, and Loss of appetite and weight loss.

Withdrawal effects have been classified in various ways, such as ‘new withdrawal symptoms’ (novel symptoms not experienced by the patient previously), ‘rebound symptoms’ (symptoms that the patient has experienced previously but with greater intensity), and ‘persistent post-withdrawal disorder’ (where withdrawal symptoms last for months or years after cessation) (Cosci & Chouinard, Reference Cosci and Chouinard2020). The FDA has emphasized that withdrawal effects can last from weeks to more than a year (Ashton, Reference Ashton1987; FDA Drug Safety Communication, 2020). Due to the persisting cognitive, affective, and somatic symptoms in these protracted withdrawal syndromes, experts have suggested the term ‘Benzodiazepine Induced Neurological Dysfunction’(BIND) as more appropriate, recognizing that such protracted symptoms often do not respond to re-instatement (Ritvo et al., Reference Ritvo, Foster, Huff, Finlayson, Silvernail and Martin2023; Shade et al., Reference Shade, Ritvo, Silvernail, Finlayson, Bressi, Foster and Martin2025), like similar post-cessation-induced neurological dysfunction seen from other classes of psychiatric drugs (Horowitz & Davies, Reference Horowitz and Davies2024). These protracted withdrawal syndromes have also been named ‘persistent post-withdrawal disorders’ by Chouinard and colleagues (Cosci & Chouinard, Reference Cosci and Chouinard2020).

Pharmacokinetic and pharmacodynamic variation and effect on withdrawal

Withdrawal arises from both pharmacokinetic and pharmacodynamic adaptations and so withdrawal risk varies between benzodiazepines according to half-life, potency, lipophilicity, receptor affinity, and the presence of active metabolites (Shapiro, Reference Shapiro2025; Teboul & Chouinard, Reference Teboul and Chouinard1990, Reference Teboul and Chouinard1991). Benzodiazepines with shorter elimination half-lives more commonly cause severe withdrawal problems (Chouinard, Reference Chouinard2004; Horowitz & Taylor, Reference Horowitz and Taylor2024). Aspects such as the alpha half-life (the rate of decline in plasma concentrations due to the process of drug redistribution from the central to the peripheral compartment), derived from single dosing studies, may be even more influential on withdrawal risk. This is particularly the case for alprazolam and lorazepam, with short alpha half-lives (Teboul & Chouinard, Reference Teboul and Chouinard1990, Reference Teboul and Chouinard1991). High-potency benzodiazepines (greater receptor occupancy per milligram and tighter binding), such as alprazolam, clonazepam, and lorazepam, cause more withdrawal than lower-potency benzodiazepines (Chouinard, Reference Chouinard2004). However, clonazepam’s favorable pharmacokinetic profile (longer alpha and beta half-lives) lends itself to less plasma fluctuations and inter-dose withdrawal anxiety despite its high potency (Teboul & Chouinard, Reference Teboul and Chouinard1991). Further risk factors for the withdrawal syndrome are shown in Box 3, including higher doses, longer duration of treatment, whether benzodiazepines have active metabolites with long half-lives, the presence of inter-dose withdrawal (Teboul & Chouinard, Reference Teboul and Chouinard1991), and a history of multiple drug switches or withdrawal attempts.

Linear tapering in clinical practice guidelines

In two scoping reviews of benzodiazepine tapering studies (Brandt et al., Reference Brandt, Bressi, Lê, Neal, Cadogan, Witt-Doerring and Wright2024; Pollmann et al., Reference Pollmann, Murphy, Bergman and Gardner2015), most original investigations and guidelines employed reductions of 10%–25% every 1–2 weeks in linear fashion (i.e. 75% of the original daily dose, then 50% of the original dose, 25% of the original dose), and then discontinuation (Pollmann et al., Reference Pollmann, Murphy, Bergman and Gardner2015). The final dose before complete cessation was generally 25% or 50% of the original dose, when reported, in these studies (Pollmann et al., Reference Pollmann, Murphy, Bergman and Gardner2015), with some implying a final dose of 5%–10% of the original dose (Brandt et al., Reference Brandt, Bressi, Lê, Neal, Cadogan, Witt-Doerring and Wright2024). Similarly, an influential clinical practice guideline recommends reducing the dose by ‘25% per week’ for rapid reduction and ‘10% every 2–4 weeks’ as a slower method (Kaiser Permanente, Reference Permanente2019). Notably, most guidelines recommend that percentage reductions all refer to the original dose of medication taken by the patient and therefore represent linear tapers (Pottie et al., Reference Pottie, Thompson, Davies, Grenier, Sadowski, Welch and Al2019).

Proportionate/hyperbolic tapering in clinical practice guidelines

In contrast, other guidelines (Ashton, Reference Ashton2002; Horowitz & Taylor, Reference Horowitz and Taylor2024; Nice, 2018) recommend tapering according to a proportionate pattern. For example, the NICE Clinical Knowledge Summary on benzodiazepines and z-drug withdrawal suggests that benzodiazepines are tapered at approximately 5%–10% of the most recent daily dose every 1–2 weeks, meaning that the size of the reductions becomes smaller and smaller as the total dose gets lower (Nice, 2018). When patients are taking 40 mg, ‘5%–10%’ corresponds to 2–4 mg every 1–2 weeks, but when down to 10 mg, this corresponds approximately to 1 mg every 1–2 weeks (Box 1) (Nice, 2018). Generally, 0.5–1 mg of diazepam is the final dose suggested by such tapering regimens before complete cessation (approximately 1%–2% of the original dose) (Ashton, Reference Ashton2005; Horowitz & Taylor, Reference Horowitz and Taylor2024). Similarly, the NICE guidelines on safe withdrawal of prescribed drugs of dependence recommend ‘a slow, stepwise rate of reduction proportionate to the existing dose, so that decrements become smaller as the dose is lowered’ (NICE, 2022).

In 2025, a joint clinical practice guideline on benzodiazepine tapering developed by the American Society of Addiction Medicine (ASAM), funded by the FDA, partnered with nine other medical societies including the American Psychiatric Association, recommended linear dose reductions as the preferred strategy (5%–10% of the daily dose every 2–4 weeks) (Brunner et al., Reference Brunner, Chen, Klein, Maust, Mazer-Amirshahi, Mecca and Lindsay2025). However, the ASAM guideline also cites the Maudsley Deprescribing Guidelines: Antidepressants, Benzodiazepines, Gabapentinoids and Z-drugs to recommend hyperbolic tapering for patients who experience withdrawal symptoms in order to mitigate these effects (Brunner et al., Reference Brunner, Chen, Klein, Maust, Mazer-Amirshahi, Mecca and Lindsay2025; Horowitz & Taylor, Reference Horowitz and Taylor2024).

Existing evidence for linear versus hyperbolic tapering

There has been limited research evaluating these different methods of tapering. However, clinical studies have reported that linear tapering is unsuccessful for a significant proportion of patients. For instance, one trial found 90% of those who linearly tapered (25% of reduction per week from the original dose) experienced withdrawal symptoms, with 32% of those on long half-life benzodiazepines and 42% of those on short half-life benzodiazepines unable to cease their medication because of withdrawal symptoms (Schweizer et al., Reference Schweizer, Rickels, Case and Greenblatt1990). Only 36% of patients were able to successfully stop their benzodiazepine in a similar trial with a 4-week taper at 15-month follow-up (Voshaar et al., Reference Voshaar, Gorgels, Mol, van Balkom, van de Lisdonk, Breteler and Zitman2003).

A survey of specialized deprescribing clinics around the world report using hyperbolic tapering as their main approach to safely stopping psychiatric drugs, including benzodiazepines (Cooper et al., Reference Cooper, Ashman, Lomani, Moncrieff, Guy, Davies and Horowitz2023). Consistent with this, a recent systematic review of tapering strategies for psychiatric drugs, including benzodiazepines, concluded that ‘hyperbolic tapering…appears to be the most promising strategy for psychiatric drug discontinuation’ (Eserian et al., Reference Eserian, Blanco, Mercuri, Matos and Galduróz2023). Hyperbolic tapering has been adopted by the NICE guidelines for several other classes of drugs including opioids and antidepressants, and suggested for higher risk patients in antidepressant tapering guidelines in Australia and Canada, based on the biological plausibility of this approach and emerging empirical evidence(Horowitz & Wilcock, Reference Horowitz and Wilcock2022; Horowitz et al., Reference Horowitz, Murray and Taylor2021; Horowitz & Taylor, Reference Horowitz and Taylor2021; Jauca, Reference Jaucan.d.; NICE, 2022; Therapeutic Guidelines, 2025). However, currently there are no randomized controlled trials comparing linear tapering with hyperbolic tapering.

Pharmacodynamics of benzodiazepines and z-drugs

Benzodiazepines and z-drugs are positive allosteric modulators of the GABA-A receptor, producing their clinical effects by enhancing the effect of the neurotransmitter GABA at the GABA-A receptor (Cheng, Wallace, Ponteri, & Tuli, Reference Cheng, Wallace, Ponteri and Tuli2018; Gunja, Reference Gunja2013). GABA is a universal inhibitor of neural activity and decreases the release of many neurotransmitters with excitatory properties (acetylcholine, noradrenaline, dopamine, serotonin, and glutamate), thought to produce the subjective experience of anxiolysis and sedation (Cheng et al., Reference Cheng, Wallace, Ponteri and Tuli2018; Turton & Lingford-Hughes, Reference Turton and Lingford-Hughes2016).

The relationship between dose of benzodiazepines (and z-drugs) and their effect on GABA-A receptors is hyperbolic (Holford, Reference Holford2018). This relationship arises as a consequence of the law of mass action, whereby additional drug molecules produce increasingly small additional effects as receptors become increasingly saturated (Holford, Reference Holford2018). The hyperbolic nature of this relationship is often obscured by the common practice of plotting dose–response curves with dose on a logarithmic axis, yielding sigmoid curves with plots that appear linear at intermediate doses (Holford, Reference Holford2018).

PET scanning, using the radioligand [¹¹C]flumazenil in non-human primates (Figure 2a), allows visualization of this hyperbolic relationship (Atkins & Nimmo, Reference Atkins and Nimmo1975; Brouillet et al., Reference Brouillet, Chavoix, Bottlaender, Khalili-Varasteh, Hantraye, Fournier and Mazière1991; Holford, Reference Holford2018). This relationship is notable for the relatively shallow slope of the relationship between dose and GABA-A occupancy at high doses of diazepam, and the steep gradient at lower doses.

Figure 2.

The differential effects of tapering benzodiazepines according to a linear or hyperbolic pattern of reduction. (a) The relationship between dose of diazepam and action at GABA-A receptors. Note the hyperbolic relationship between dose and estimated GABA-A occupancy. This graph was adapted from Brouillet et al. (Reference Brouillet, Chavoix, Bottlaender, Khalili-Varasteh, Hantraye, Fournier and Mazière1991), a study conducted in baboons, with dosages converted to human equivalents, as outlined in Nair and Jacob (Reference Nair and Jacob2016). (b) Linear dose reductions of diazepam cause hyperbolically increasing reductions in effect at the GABA-A receptor. This may be associated with increasingly severe withdrawal effects. (c) Hyperbolically decreasing dose reductions produce linear reductions in effect at GABA-A receptors. This may be associated with more ‘evenly spread’ withdrawal effects.

Figure 2. Long description

Three line graphs, labeled A, B, and C, share identical axes. The x-axis represents Diazepam dose in milligrams from 0 to 100. The y-axis represents GABA-A occupancy as a percentage from 0 to 100. All three graphs feature a hyperbolic curve that rises steeply from the origin and flattens as it approaches 80 percent occupancy.

* Graph A shows the baseline hyperbolic relationship with data points at approximately 5, 10, 30, 50, 75, and 100 milligrams.

* Graph B illustrates linear dose reductions. Red dashed lines mark equal horizontal intervals on the x-axis at 50, 37.5, 25, and 12.5 milligrams. These correspond to increasingly large vertical drops on the y-axis, showing that equal dose cuts lead to accelerating losses in receptor occupancy.

* Graph C illustrates hyperbolic dose reductions. Red dashed lines mark equal vertical intervals on the y-axis at 80, 60, 40, and 20 percent occupancy. These correspond to progressively smaller horizontal dose reductions on the x-axis, showing that smaller dose cuts at lower levels are required to maintain a linear reduction in receptor effect.

Hyperbolic pattern in benzodiazepine effects

The hyperbolic dose–receptor occupancy relationship is mirrored in multiple neurochemical and behavioral effects, indicating clinical relevance. GABA-gated currents increase hyperbolically with increasing diazepam concentrations in cellular models (Berezhnoy et al., Reference Berezhnoy, Gravielle, Downing, Kostakis, Basile, Skolnick and Farb2008; Wongsamitkul et al., Reference Wongsamitkul, Maldifassi, Simeone, Baur, Ernst and Sigel2017), including recombinant human GABA-A receptors (Atack, Reference Atack2009) (Figure 3a). Seizure threshold shows a hyperbolic relationship to dose in zebrafish (Gupta, Khobragade, & Shingatgeri, Reference Gupta, Khobragade and Shingatgeri2014), rats (Kapur & Macdonald, Reference Kapur and Macdonald1997), and non-human primates (Bottlaender et al., Reference Bottlaender, Brouillet, Varastet, Le Breton, Schmid, Fuseau and Mazière1994) (Figure 3b). Anxiolytic effects increase hyperbolically in pigeons (Kleven & Koek, Reference Kleven and Koek1999), rats (Dubinsky et al., Reference Dubinsky, Vaidya, Rosenthal, Hochman, Crooke, Deluca and Shank2002), and non-human primates (Rowlett et al., Reference Rowlett, Platt, Lelas, Atack and Dawson2005) (Figure 3c). Z-drugs display hyperbolic effects on locomotion, seizure threshold, muscle relaxation, and ataxia in mice (Sanger & Depoortere, Reference Sanger and Depoortere1998).

Figure 3.

The relationship between dose of benzodiazepine or z-drug and many biological and clinical effects is hyperbolic. (a) Diazepam produces a hyperbolic relationship between benzodiazepine-site concentration and GABA current in human recombinant GABA-A receptors. Adapted from Atack (Reference Atack2009). (b) Triazolam produces a hyperbolic relationship between dose and increase in seizure threshold in non-human primates. Adapted from Bottlaender et al. (Reference Bottlaender, Brouillet, Varastet, Le Breton, Schmid, Fuseau and Mazière1994). (c) Lorazepam demonstrates a hyperbolic relationship between dose and degree of anxiolysis. Anxiolysis was measured by mean percentage of open arm entries on the elevated plus-maze in mice. Adapted from Dubinsky et al. (Reference Dubinsky, Vaidya, Rosenthal, Hochman, Crooke, Deluca and Shank2002). (d) Zolpidem produces a hyperbolic relationship between dose and total sleep time in human subjects. Adapted from Merlotti et al. (Reference Merlotti, Roehrs, Koshorek, Zorick, Lamphere and Roth1989).

Figure 3. Long description

Four panels labeled A through D.

Panel A. Top-left. The X axis is Diazepam concentration in nM from 0 to 400. The Y axis is GABA current percentage from negative 50 to 150. A curve rises sharply from the origin and plateaus near 125 percent at 300 n M.

Panel B. Top-right. The X axis is Triazolam dose in ug/kg from 0 to 300. The Y axis is increase in seizure threshold percentage from negative 200 to 600. The curve rises steeply from 0 and levels off at approximately 500 percent between 200 and 250 ug/kg.

Panel C. Bottom-left. The X axis is Lorazepam dose in mg from 0 to 15. The Y axis is Anxiolysis from 0 to 40. The curve shows a hyperbolic increase, starting at approximately 5 and reaching nearly 30 at a dose of 10 m g.

Panel D. Bottom-right. The X axis is Zolpidem dose in mg from 0 to 25. The Y axis is Total sleep time in minutes from 445 to 470. The curve begins at 447 minutes at zero dose and rises hyperbolically toward 465 minutes at a 20 mg dose.

In humans, hyperbolic relationships occur between benzodiazepine dose and saccadic slowing/sedation (Hommer et al., Reference Hommer, Matsuo, Wolkowitz, Chrousos, Greenblatt, Weingartner and Paul1986), zolpidem’s effects on sleep (Merlotti et al., Reference Merlotti, Roehrs, Koshorek, Zorick, Lamphere and Roth1989) (Figure 3d), diazepam’s impact on cortisol and growth hormone (Hommer et al., Reference Hommer, Matsuo, Wolkowitz, Chrousos, Greenblatt, Weingartner and Paul1986) and delirium risk in critically ill children (Mody et al., Reference Mody, Kaur, Mauer, Gerber, Greenwald, Silver and Traube2018).

Withdrawal symptoms also appear to be hyperbolically dose-dependent: larger reductions cause greater severity, particularly at lower doses (Ashton, Reference Ashton, Haddad, Dursun and Deakin2004). For example, it has been observed that 1 mg reductions are generally tolerable from 20 mg diazepam but not from 5 mg, where smaller decrements are preferred (Ashton, Reference Ashton, Haddad, Dursun and Deakin2004), suggesting a hyperbolic relationship between dose reduction and withdrawal severity.

Neurobiology underpinning tapering

Adaptations to a substance cause tolerance during exposure and predict withdrawal when the substance is removed and these adaptations act unopposed (Turton & Lingford-Hughes, Reference Turton and Lingford-Hughes2016). The physiological mechanisms of benzodiazepine tolerance and adaptation remain uncertain, with mixed findings on reduced number or sensitivity of GABA-A receptors (Cheng et al., Reference Cheng, Wallace, Ponteri and Tuli2018). Preclinical models suggest uncoupling – loss of allosteric modulation at the benzodiazepine site – as a likely mechanism (Cheng et al., Reference Cheng, Wallace, Ponteri and Tuli2018). Chronic benzodiazepine use also down-regulates adenosine receptors and up-regulates glutamatergic NMDA and AMPA receptors, likely compensating for enhanced GABA activity (Cheng et al., Reference Cheng, Wallace, Ponteri and Tuli2018).

Relative underactivity of inhibitory GABA activity, and its downstream effects, during benzodiazepine dose reduction likely cause withdrawal symptoms (Authier et al., Reference Authier, Balayssac, Sautereau, Zangarelli, Courty, Somogyi and Eschalier2009), supported by findings that administration of flumazenil – a short-acting selective GABA-A antagonist – induces rapid-onset withdrawal (Mintzer, Stoller, & Griffiths, Reference Mintzer, Stoller and Griffiths1999). Reduced GABA transmission allows increased excitatory activity, as neurotransmitters like glutamate act with less opposition (Ashton, Reference Ashton2005), perhaps explaining symptoms such as panic, anxiety, tachycardia, diaphoresis, and typical of excitatory states. Withdrawal likely resolves once receptor changes and downstream processes caused by adaptation to drug exposure sufficiently reverse (Ashton, Reference Ashton2005). Although not well understood, the rate of this reversal varies between individuals and with exposure duration, perhaps explaining withdrawal durations from weeks to years (Ashton, Reference Ashton1995; Lader & Morton, Reference Lader and Morton1992).

The rationale for tapering is that gradually reducing GABA-A receptor modulation by benzodiazepines minimizes disturbance to homeostatic balance achieved at steady-state dosing and thus withdrawal symptoms. Slow tapering allows neuroadaptations to resolve in step with reduced receptor activation (Ashton, Reference Ashton2005; Horowitz & Taylor, Reference Horowitz and Taylor2021).

Pharmacological rationale for hyperbolic tapering

Benzodiazepine dose demonstrates a hyperbolic relationship with both GABA-A occupancy and clinical effects. This suggests a linear relationship between GABA-A occupancy and clinical effects, supported by seizure-threshold studies in primates (Ashton, Reference Ashton2002; Framer, Reference Framer2021; Nice, 2018). A similar linear relationship between withdrawal effects and GABA-A occupancy is plausible. To minimize withdrawal, especially at lower doses, GABA-A activity should be reduced linearly, which requires hyperbolically decreasing dose reductions. The benzodiazepine dose–response curve illustrates the impact of linear tapering (Figure 2a): reducing diazepam by fixed 12.5 mg intervals from 50 mg yields progressively larger drops in GABA-A occupancy – 6.6 percentage points (p.p.) (50 to 37.5 mg), 10.0 p.p. (37.5 to 25 mg), 16.7 p.p. (25 to 12.5 mg), and 33.6 p.p. (12.5 to 0 mg) (Figure 2b, Table 1). At low doses, small reductions have disproportionately large effects; for example, 5 to 0 mg changes occupancy by 16.8 p.p., exceeding the change from 100 to 50 mg (13.3 p.p.), consistent with the clinical observation of increasing withdrawal severity near zero (Ashton, Reference Ashton, Haddad, Dursun and Deakin2004).

A regimen that reduces GABA-A occupancy linearly – e.g. 50 mg by 4 equal-sized steps of receptor occupancy – would entail 50 mg (66.9% occupancy), 24.9 mg (50.2%), 12.4 mg (33.5%), 5.0 mg (16.7%), and 0 mg (Figure 2c). A more gradual regimen reducing occupancy by approximately 1 percentage point at a time is shown in Table 2.

A hyperbolic reduction regimen can be closely approximated by an exponential one, which can be easier to calculate. For example, reducing the dose by 10% of the most recent dose at each step equates to approximately 2 percentage-point reductions in GABA-A occupancy, while 5% reductions equate to approximately 1 percentage point reductions. This seems to support the validity of the roughly exponential regimens (5%–10% reductions of the most recent dose) derived from clinical experience (Ashton, Reference Ashton2002; Framer, Reference Framer2021; Nice, 2018).

This analysis suggests that final benzodiazepine doses before cessation should be far below therapeutic levels, so the last reduction to zero is no larger (in receptor occupancy terms) than previously tolerated steps. This aligns with regimens recommending final diazepam doses of 0.2–0.5 mg (1%–2% GABA-A occupancy), with some patients requiring even smaller doses (Ashton, Reference Ashton2005; Gupta, Cahill, & Miller, Reference Gupta, Cahill and Miller2018).

Limitations of the review

Several limitations should be noted when interpreting the evidence base for hyperbolic tapering of benzodiazepines.

The PET imaging data used to model the dose-occupancy curve come from non-human primates rather than humans; while receptor pharmacology is broadly comparable, species differences in subunit composition, distribution, and pharmacokinetics may exist (Atack et al., Reference Atack, Wong, Fryer, Ryan, Sanabria, Zhou and Hargreaves2010; Friedman, Redmond, & Greenblatt, Reference Friedman, Redmond and Greenblatt1991). These data also represent single doses and population averages, which obscure inter-individual variability in receptor sensitivity, binding affinity, and pharmacodynamic response. No empirical studies have directly tested linear versus hyperbolic tapering; this remains a biologically plausible inference from receptor theory, supported by clinical experience, including patient-led experience. Modeling assumes steady-state pharmacokinetics and linear elimination, which may be disrupted by illness, polypharmacy, or metabolic variability.

Nonetheless, because pharmacodynamic relationships follow the law of mass action, the general shape of the dose-occupancy curve – dictated by ligand–receptor binding kinetics – is unlikely to differ substantially between primates and humans or across individuals, even if exact inflection points vary. Thus, while absolute parameters may shift, the hyperbolic nature of the relationship is unlikely to be an artifact, providing a sound rationale for hyperbolic tapering. In practice, individual differences are best managed empirically, using the intensity of withdrawal symptoms as the main guide to taper speed rather than a ‘one-size-fits-all’ approach.

Implementation in clinical practice