Why this poison is interesting

Benzodiazepines are in Band A because of a negative result. They are among the most commonly taken drugs in overdose in the UK and among the least likely to kill anyone who takes them alone. That is not luck, and it is not because the doses are small. It is a direct consequence of how the molecule works at its receptor, and it is one of the few places in toxicology where a mechanism can be used to reason confidently about safety.

The second reason is the corollary. The ceiling is a property of the drug acting on its own, and it does not survive combination. An opioid depresses respiration by a separate mechanism at a separate site; a benzodiazepine's inability to depress respiration much by itself says nothing about what it will do to a patient who has also taken one. Most benzodiazepine deaths are combination deaths, and the mechanism explains why that is not a coincidence.

The third is that the antidote is a case study in a treatment that works and should usually not be given — and, unusually for this library, there is randomised evidence rather than case reports.

A poison is a drug whose kinetics have escaped its pharmacology.

The toxic principle

The poison is the drug itself, acting at a specific allosteric site on the GABA-A receptor — a ligand-gated chloride channel. The critical word is allosteric.

The comparison that makes this concrete is with the drugs that lack the ceiling:

Barbiturates: At higher concentrations they directly open the chloride channel, independently of GABA. There is no endogenous ligand limiting the effect, so the curve keeps rising — which is why barbiturate overdose is lethal in a way that benzodiazepine overdose is not, and why EXTRIP has a barbiturate recommendation and no benzodiazepine one.² Established
Ethanol: Modulates GABA-A among several other targets, but also acts at NMDA receptors and at concentrations measured in millimoles rather than nanomoles. Its respiratory depression is dose-related without a comparable plateau — see ethanol. Inferred
Opioids: Act at an entirely different receptor on a different neuronal population — see opioids. Their respiratory depression is not limited by GABA availability and is not opposed by anything a benzodiazepine ceiling provides. Established

Toxicokinetics

Benzodiazepines — a class of very different kinetics with one shared mechanism
Parameter	Therapeutic	In overdose	Why it changes
Absorption	Well absorbed orally. Onset of effect tracks lipid solubility rather than dose — diazepam enters the brain rapidly, lorazepam more slowly	Unchanged in kind. There is no anticholinergic gastric stasis as in tricyclics	The fast-onset agents produce the most abrupt clinical picture and are the most abused; the slow-onset agents produce a gentler curve. Neither difference changes the ceiling — it changes when the plateau is reached.
Protein binding	High for most of the class	Not a limiting factor; no clinically important saturation	High binding plus a moderate-to-large volume of distribution is the standard profile of a poison that cannot be dialysed — and here it does not matter, because nobody needs to.
Volume of distribution	Moderate to large; the lipophilic agents redistribute rapidly into fat	Unchanged	Redistribution into fat explains why a single dose of a lipophilic benzodiazepine wears off faster than its elimination half-life suggests, and why repeated dosing then accumulates. In overdose it means duration is governed by redistribution first and elimination later.
Half-life	Enormously variable across the class — hours for midazolam, days for diazepam once its active metabolites are counted	Prolonged in the elderly, in hepatic impairment, and with co-administered enzyme inhibitors	The relevant number is not the parent's half-life but the active metabolites'. Diazepam yields desmethyldiazepam, oxazepam and temazepam, all active; a diazepam overdose therefore produces sedation lasting far beyond diazepam's own clearance.
Elimination route	Hepatic oxidation (CYP3A4 and others) then glucuronidation; lorazepam, oxazepam and temazepam are conjugated directly without an oxidative step	Unchanged	The direct-conjugation agents are less affected by hepatic impairment and by enzyme inhibition — which is a prescribing fact rather than a poisoning one, but explains why the same overdose behaves differently in a patient with liver disease depending on which benzodiazepine it was.
Order of kinetics	First order	First order	Nothing saturates. Benzodiazepine poisoning is a receptor-occupancy problem with an upper bound, not a kinetic one.
Dialysability	—	Not clinically relevant. EXTRIP has published no benzodiazepine recommendation²	High protein binding and large volumes of distribution make removal ineffective, and the absence of a recommendation reflects that nobody has needed one: supportive care resolves the poisoning. The workgroup has published on barbiturates, which is a precise mechanistic contrast — the drug without the ceiling is the one that needed a dialysis guideline.

Metabolism and the metabolites

There is no bioactivation. Metabolism reduces effect — but slowly, and for several agents through metabolites that are themselves active benzodiazepines with longer half-lives than the parent.

Diazepam — why the duration outlasts the drug

DiazepamRapid brain entry; short distribution half-life, long elimination half-life
CYP2C19 / CYP3A4 N-demethylationDesmethyldiazepam (nordazepam)Active, and longer-lived than diazepam itself
CYP3A4 hydroxylationTemazepamActive — and a marketed hypnotic in its own right
OxazepamActive. The common downstream metabolite, and also a marketed drug
Inactive glucuronide, renally excreted

What changes the answer

Co-ingestion, which is the dominant variable and the reason the ceiling stops mattering. Opioids most dangerously; also alcohol, gabapentinoids, sedating antihistamines and antipsychotics. Established
Age and frailty. The ceiling limits respiratory depression; it does not limit sedation, and sedation in a frail patient produces aspiration, falls, hypothermia and a long lie. The morbidity of benzodiazepine overdose in the elderly is not respiratory. Inferred
Tolerance. A dependent patient may be barely affected by a dose that would obtund a naïve one — and, conversely, is the patient in whom flumazenil is most likely to precipitate seizures.
Which benzodiazepine, and whether it is a benzodiazepine at all. Novel or 'designer' benzodiazepines — etizolam, flualprazolam, bromazolam and others — circulate in the UK illicit supply, often in tablets sold as diazepam. Their receptor pharmacology is broadly similar; their potency and duration frequently are not, and they are not detected by routine immunoassays. Inferred

Elimination and accumulation

There is nothing to enhance and nothing to trap. Benzodiazepines are not ionisable in a useful way, do not recirculate enterohepatically to any manageable degree, are too highly bound and too widely distributed to dialyse, and have no toxic metabolite whose formation could be blocked. The only accumulation that matters is the accumulation of active metabolites, and time is the only thing that clears them.

Target organs — and why those

Brain — cortex and limbic system

TargetGABA-A receptors containing the appropriate α subunits, modulated allosterically

Why hereGABA-A receptors are the principal inhibitory receptor of the central nervous system, and the benzodiazepine site exists on the subtypes concentrated in cortex, limbic structures and thalamus. Enhancing inhibition there produces anxiolysis, amnesia, sedation and — at the top of the flattened curve — coma. Because the effect is modulatory, the depth is bounded by how much GABA is being released, which is why the coma is usually rousable and usually uncomplicated. Established

At the bedsideSedation without significant hypoventilation is the characteristic isolated presentation. A benzodiazepine-only overdose that is profoundly hypoventilating should prompt a search for the second drug.

Brainstem — respiratory network

TargetThe same receptors, in the neurons generating respiratory rhythm

Why hereThe effect is present but limited, and limited for the same reason. Compare this directly with opioids, where µ-receptor agonism silences a specific respiratory population without any comparable endogenous constraint.³ Two sedatives, two mechanisms, two entirely different lethalities — and in combination the constraint on one does not compensate for the absence of a constraint on the other. Inferred

At the bedsideThe respiratory rate is the measurement in any sedative poisoning. In isolated benzodiazepine overdose it is usually reassuring, and that reassurance is the finding — not the sedation.

Airway — indirectly

TargetProtective reflexes rather than any receptor population of interest

Why hereThis is not organ selectivity; it is the consequence of sedation in a patient lying down. Loss of airway reflexes permits aspiration, and aspiration is a far commoner mechanism of harm in benzodiazepine poisoning than respiratory depression. It is also the mechanism by which the elderly are harmed disproportionately. Inferred

At the bedsidePositioning and airway assessment do more for a benzodiazepine-poisoned patient than any pharmacological intervention, which is an unglamorous conclusion that the mechanism supports directly.

Timeline of effects

Benzodiazepines — a course defined by what fails to happen

Time

What you seeWhat is happening

0–2 hOnset
What you seeDrowsiness, ataxia, dysarthria, anterograde amnesia. Progressive sedation to a rousable coma. Respiratory rate and oxygen saturation typically remain adequate.
What is happeningAllosteric enhancement of GABA-A chloride conductance across cortex and limbic system. The effect rises with concentration until endogenous GABA availability limits it, and then plateaus. Established
2–12 hThe plateau
What you seeDeep but stable sedation. The patient is rousable to voice or pain, protects their airway, and does not deteriorate further. The absence of progression is the diagnostic feature.
What is happeningThe dose–response ceiling. This is the phase in which taking more drug produces very little more effect — the mechanistic reason a patient can present having taken an alarming number of tablets and be, clinically, merely asleep.
Any timeThe combination case
What you seeFalling respiratory rate, rising CO₂, hypoxia. Looks like an opioid overdose because it usually largely is one. This is where benzodiazepine-related deaths occur.
What is happeningA second depressant acting at a different site with no comparable constraint. Naloxone reverses only the opioid component, so an incomplete response is expected and does not imply a wrong diagnosis — see opioids. Inferred
12 h – daysThe long tail
What you seeProlonged drowsiness, particularly with diazepam and in the elderly. Aspiration pneumonia, pressure injury and rhabdomyolysis in those found late.
What is happeningActive metabolites — desmethyldiazepam, oxazepam, temazepam — sustain receptor occupancy long after the parent drug has been cleared. Redistribution from fat contributes. Established

What the mechanism predicts at the bedside

Why an isolated benzodiazepine overdose is usually an observation problem

Because the mechanism has an upper bound and the drugs have no organ toxicity. There is nothing being destroyed, nothing accumulating behind a bottleneck, and no metabolite to intercept. Supportive care and time are not a fallback here — they are the treatment that matches the mechanism.

Why a benzodiazepine overdose that is hypoventilating deserves a search

Because the mechanism predicts that it should not be. Significant respiratory depression from a benzodiazepine alone is uncharacteristic, so its presence is positive evidence of something else — an opioid, an alcohol, a gabapentinoid, an airway that is not being protected, or a patient with very little respiratory reserve to begin with. The ceiling is more useful as a diagnostic prior than as a reassurance. Inferred

Why flumazenil is a real antidote that is usually the wrong choice

Flumazenil is a competitive antagonist at the benzodiazepine site and it works. The question is not whether it reverses sedation — it does — but whether reversing sedation in this poisoning is worth its harms. Unusually for toxicology, that question has been addressed with randomised data.

What the randomised evidence shows

A systematic review with meta-analysis of 13 randomised trials, 994 patients randomised (990 evaluable), comparing flumazenil with placebo in known or suspected benzodiazepine overdose:¹

Adverse events: 138/498 with flumazenil versus 47/492 with placebo — risk ratio 2.85 (95% CI 2.11–3.84), p < 0.00001.¹
Serious adverse events: 12/498 versus 2/492 — risk ratio 3.81 (95% CI 1.28–11.39), p = 0.02.¹
The commonest adverse events with flumazenil were agitation and gastrointestinal symptoms; the commonest serious adverse events were supraventricular arrhythmia and convulsions.¹
No patients died during the blinded phase of the trials.¹

The authors' conclusion: use of flumazenil in patients presenting to an emergency department with known or suspected benzodiazepine intoxication "is associated with a significantly increased risk of (S)AEs compared with placebo. Flumazenil should not be used routinely, and the harms and benefits should be considered carefully in every patient."¹ Established

The mechanism explains both halves of that result. Reversing an allosteric enhancer in a brain that is adapted to benzodiazepines removes an inhibitory tone it has come to depend on — hence seizures, and hence the particular danger in chronic users. And in a patient who has co-ingested a proconvulsant such as a tricyclic antidepressant, benzodiazepine effect may be the only thing holding the seizure threshold up. Inferred

Why the risk–benefit is so poor here specifically

Compare with opioids. Naloxone reverses a poisoning that is killing the patient, so accepting withdrawal is an easy trade. Flumazenil reverses a poisoning that is usually not killing the patient, so the same class of harm is being accepted for a much smaller benefit. A competitive antagonist is only as good as the danger it removes, and the ceiling effect means there is usually not much danger to remove. Inferred

Why a negative drug screen does not exclude a benzodiazepine

Routine urine immunoassays are built around the metabolites of the classical benzodiazepines and detect several of the novel ones poorly or not at all. Since etizolam, flualprazolam, bromazolam and similar compounds circulate in tablets sold as diazepam, a negative screen in a sedated patient with a plausible history means very little. This is a laboratory limitation rather than a pharmacological one, and it is the reason the clinical picture outranks the test. Inferred

The antidote, from the poison's side

Read from the poison's side, flumazenil is mechanistically elegant and clinically awkward, and both follow from the same fact: it competes at the benzodiazepine site, and that site is modulatory rather than essential.

It reverses the drug, not the sedation. Flumazenil displaces the benzodiazepine from its allosteric site and has no effect on GABA-A function otherwise, on alcohol, on opioids or on any other sedative. A patient who does not wake has not disproved the benzodiazepine — they have demonstrated a co-ingestion. Established
Its half-life is short, so re-sedation follows the same arithmetic as renarcotisation after naloxone — the antagonist leaves before the agonist does. Established
Seizure risk concentrates in exactly the patients most likely to be offered it: chronic benzodiazepine users, and those who have co-ingested a proconvulsant. The randomised data put serious adverse events — most often supraventricular arrhythmia and convulsions — at a risk ratio of 3.81 against placebo.¹ Established
Its most defensible use is diagnostic or iatrogenic — reversing procedural sedation, or clarifying a picture in a patient with no history of benzodiazepine use and no proconvulsant co-ingestion. Even there, the meta-analysis's conclusion stands: not routinely, and considered carefully in every patient.¹

Critical appraisal

"Benzodiazepines are safe in overdose" is true enough to teach and dangerous as a sentence. Inferred What the mechanism supports is that isolated oral benzodiazepine overdose in an otherwise well adult rarely causes fatal respiratory depression. It does not support safety in the elderly, in the presence of any other depressant, in patients with respiratory disease, or where the airway is unprotected. Most benzodiazepine-associated deaths involve other drugs, and the ceiling is not the reason they are safe — it is the reason the other drug is the one that killed them.
The ceiling effect is textbook pharmacology, well supported and rarely quantified. Established for the allosteric mechanism and the requirement for endogenous GABA; the position of the plateau in a human overdose has not been characterised, and there is no concentration above which further drug is known to be inert. It is a shape, not a threshold.
The flumazenil meta-analysis is the strongest evidence on any page in Band A, and it has limits.¹ Thirteen trials and 994 patients is a genuinely randomised base — better than anything EXTRIP has for any poison. But no patients died in the blinded phase, so the trials cannot speak to mortality benefit or harm, and the serious-adverse-event finding rests on 12 events against 2, with a confidence interval from 1.28 to 11.39. The direction is convincing; the magnitude is not precise.
The seizure mechanism is inferred rather than demonstrated. Inferred That flumazenil increases serious adverse events, of which convulsions were among the commonest, is established from the trial data — though no seizure-specific effect estimate is reported.¹ That it does so by unmasking withdrawal in an adapted brain, or by removing anticonvulsant tone in a patient with a proconvulsant co-ingestion, is a reconstruction — coherent, widely accepted, and not directly shown.
Novel benzodiazepines are a moving target and this page describes them only in outline. Inferred Potency, duration and receptor subtype selectivity vary across compounds appearing in the UK supply, and published human pharmacology lags the market by years. The one durable statement is analytical rather than pharmacological: routine immunoassays cannot be relied on to detect them.
Treating a class as one poison is a bigger compromise here than it looks. Half-lives span hours to days, active metabolites are present for some agents and absent for others, and onset varies several-fold with lipid solubility. The shared receptor mechanism is real; the shared clinical course is an approximation that fails most obviously when predicting how long someone will be asleep.

References

1
Penninga EI, Graudal N, Ladekarl MB, Jürgens G. Adverse events associated with flumazenil treatment for the management of suspected benzodiazepine intoxication — a systematic review with meta-analyses of randomised trials. Basic Clin Pharmacol Toxicol 2016;118(1):37–44. PubMed 26096314 Thirteen randomised trials, 994 patients randomised and 990 evaluable. Source of every figure quoted above: adverse events 138/498 versus 47/492 (RR 2.85, 95% CI 2.11–3.84), serious adverse events 12/498 versus 2/492 (RR 3.81, 95% CI 1.28–11.39), the commonest adverse and serious adverse events, the absence of deaths in the blinded phase, and the quoted conclusion. Verified 1 Sep 2026 from the abstract.
2
EXTRIP workgroup — published recommendations index. extrip-workgroup.org/recommendations Cited for a contrast: the workgroup's 22 published poison recommendations include barbiturates and do not include benzodiazepines. Verified 1 Sep 2026.
3
Liu S, Kim D-I, Oh TG, et al. Neural basis of opioid-induced respiratory depression and its rescue. Proc Natl Acad Sci U S A 2021;118(23):e2022134118. PMC8201770 Cited here only for the contrast with µ-opioid-receptor-mediated respiratory depression, which has no comparable endogenous-ligand constraint. Verified 1 Sep 2026.
4
TOXBASE — benzodiazepines; flumazenil. National Poisons Information Service. toxbase.org (NHS login required. NPIS: 0344 892 0111. The authoritative UK source for observation periods, the restricted indications for flumazenil, and disposition — and for the novel benzodiazepines currently in circulation. Login-gated, so not quoted here.)

Benzodiazepines

At a glance