Why this poison is interesting

Carbamazepine is the library's best example of a single molecule that poisons through three mechanisms borrowed from three other pages. Its three-ring iminostilbene skeleton is close kin to the tricyclic antidepressants, so it blocks cardiac sodium channels and widens the QRS.² It has intrinsic antimuscarinic activity, so it produces the anticholinergic toxidrome — delirium, mydriasis, tachycardia, urinary retention and, critically, reduced gut motility. And that last effect turns the drug on itself: by slowing the gut it delays and prolongs its own absorption, so the poisoning relapses as a gut reservoir releases more drug on the second or third day.²

It blocks sodium channels like a tricyclic. The QRS widens, and the same bicarbonate logic that applies to the drugs on the sodium-channel blockade page applies here. Established
It is anticholinergic. Delirium, mydriasis, tachycardia and — the mechanistically important one — an ileus that becomes part of the kinetics.² Established
It has an active metabolite. Carbamazepine-10,11-epoxide is pharmacologically active and is itself measured in some assays; it outlasts the parent and contributes to the prolonged and fluctuating course.²
It causes hyponatraemia. An ADH-like effect produces water retention and hyponatraemia, occasionally severe.²

Most poisons obey their kinetics. Carbamazepine rewrites them — it slows the gut it is sitting in, and then poisons the patient twice.

The toxic principle

There are two toxic species and three toxic actions. The species are carbamazepine and its active 10,11-epoxide metabolite; the actions are sodium-channel blockade, antimuscarinic effect, and an ADH-like disturbance of water balance.

Sodium-channel blockade. Carbamazepine's anticonvulsant action is use-dependent blockade of voltage-gated sodium channels — the same mechanism as phenytoin and the tricyclics. In overdose it extends to the myocardium, producing QRS widening, conduction disturbance and, at the extreme, the syncope-and-cardiac-arrest picture the label describes.² Established
Antimuscarinic activity. Carbamazepine blocks muscarinic receptors, giving the anticholinergic picture: delirium, mydriasis, tachycardia, urinary retention, and reduced bowel motility and delayed gastric emptying.² The gut effect is the one that reshapes the kinetics. Established
An ADH-like effect. The label attributes water intoxication and hyponatraemia to an ADH-like action;² the hyponatraemia can be a chronic-therapy problem or an acute-overdose one, and can itself lower the seizure threshold. Inferred

Toxicokinetics

Carbamazepine — slow, self-delaying absorption and an active metabolite
Parameter	Therapeutic	In overdose	Why it changes
Absorption	Almost completely but relatively slowly absorbed; conventional tablets peak within ~12 h, bioavailability 85–100%²	Slower and more prolonged still in overdose — its own antimuscarinic ileus delays gastric emptying and gut transit, and relapse on the 2nd and 3rd day due to delayed absorption should be anticipated²	This is the row that defines the poisoning. Carbamazepine is slow to absorb even therapeutically, and in overdose it slows itself further by paralysing the gut it is in — creating a reservoir that releases drug for days. It is the clearest example in the library of a poison altering its own kinetics.
Protein binding	70–80% bound to serum proteins; free fraction reflected in CSF and saliva²	Unchanged; a substantial free fraction remains	Lower than phenytoin's ~90%, which is part of why EXTRIP graded carbamazepine dialysable despite the binding — there is more free drug available to the machine than the headline binding figure suggests.
Volume of distribution	0.8–1.9 L/kg²	Unchanged	A moderate volume of distribution; combined with a 20–30% free fraction, it makes extracorporeal removal worthwhile in the severe cases, though the delayed gut reservoir means dialysis alone does not end the exposure — the charcoal has to keep working on the gut.
Metabolism	Hepatic; the epoxide pathway (via CYP3A4) is dominant, producing the active 10,11-epoxide, then the inactive trans-diol²	The active epoxide accumulates and contributes to and prolongs the toxicity²	Auto-induction means chronic users clear the drug faster than naïve ones (single-dose half-life ~36 h falls to 16–24 h on repeated dosing²), so the same ingested amount behaves differently in a regular taker and a first-time overdose — and enzyme-inducing co-medication shortens it further.
Elimination	Renal excretion of metabolites (72% urine, 28% faeces); only ~2% unchanged in urine²	No useful renal route to enlarge	As with the rest of this cluster, urinary manipulation offers nothing. Removal means charcoal on the gut reservoir and, in severe cases, dialysis on the blood.
Dialysability	—	Moderately dialysable — EXTRIP suggests it in severe poisoning (2D) despite the protein binding, with MDAC continued throughout (1D)¹	The combination of a 20–30% free fraction, a moderate volume of distribution and a life-threatening severe presentation makes dialysis reasonable in the worst cases. That MDAC continues during dialysis reflects the two-compartment problem: the machine clears the blood while the charcoal works on the gut reservoir that would otherwise refill it.

Metabolism and the metabolites

Unlike its cluster-mates, carbamazepine has an active metabolite that matters — the 10,11-epoxide — but the more distinctive metabolic feature is not the epoxide itself. It is that the drug's antimuscarinic action makes its own absorption, not just its metabolism, part of the toxic picture.

Carbamazepine — an active epoxide, and a gut that will not empty

Carbamazepine (ingested)Blocks sodium channels and muscarinic receptors. The antimuscarinic effect slows the gut it is sitting in
A gut reservoir of unabsorbed drugThe self-inflicted latent phase — delayed, prolonged absorption; relapse on day 2–3²
Carbamazepine-10,11-epoxidePharmacologically active; outlasts and supplements the parent, and is not measured by every assay²
10,11-trans-diol and its glucuronideInactive; renally excreted²

What changes the answer

Chronic users clear it faster. Auto-induction cuts the half-life from ~36 h to 16–24 h,² so a naïve first-time overdose can behave more sluggishly than the same amount in a long-term user — the opposite of the theophylline chronic-toxicity story. Established
Enzyme-inducing co-medication shortens the parent half-life but not the danger — the epoxide and the gut reservoir still drive the course. Inferred
Hyponatraemia lowers the seizure threshold, so the ADH-like effect and the sodium-channel effect can combine to make seizures more likely in a way neither would alone.² Inferred

Elimination and accumulation

Carbamazepine leaves by hepatic metabolism, with almost nothing excreted unchanged.² So the two levers that work are charcoal — on the gut reservoir and the recirculating drug — and dialysis, in the severe cases, on the blood. The delayed, relapsing absorption means neither works as a single event: the gut keeps releasing drug, so the charcoal keeps working.

The EXTRIP position

74 articles, 173 patients, 6 fatalities; very low quality of evidence, and the workgroup concluded carbamazepine is moderately dialysable and suggested extracorporeal removal in severe poisoning despite the high protein binding.¹

Suggested in severe carbamazepine poisoning (2D).¹
Recommended for multiple seizures refractory to treatment (1D) or life-threatening dysrhythmias (1D) — the two strong indications, both end-organ.¹
Suggested for prolonged coma or respiratory depression requiring ventilation (2D), or if significant toxicity persists — especially with rising or persistently elevated concentrations — despite MDAC and supportive care (2D).¹
Stop when clinical improvement is apparent (1D) or the concentration is below 10 mg/L (42 µmol/L) (2D).¹
Intermittent haemodialysis preferred (1D); haemoperfusion (1D) or CRRT (3D) alternatives. Continue MDAC throughout (1D).¹

Target organs — and why those

Brain

TargetCerebellar and vestibular pathways, muscarinic receptors, and consciousness

Why hereSodium-channel blockade gives the cerebellar signs (nystagmus, ataxia, dysarthria) shared with phenytoin; antimuscarinic activity adds delirium, agitation and hallucination; and at high concentration consciousness is depressed to coma. Hyponatraemia and the epoxide compound the CNS picture.² Established

At the bedsideFluctuating consciousness is characteristic — patients wax and wane, partly because of the delayed, pulsatile absorption. A cyclical or relapsing coma in an anticonvulsant overdose is carbamazepine until proven otherwise.²

Heart

TargetMyocardial sodium channels and the conducting system

Why hereThe tricyclic-like sodium-channel blockade widens the QRS and disturbs conduction; the label records syncope with cardiac arrest at the severe end.² Anticholinergic tachycardia sits on top. This is the mechanism that links carbamazepine to the sodium-channel blockade page. Established

At the bedsideA widening QRS is treated on the tricyclic model — sodium bicarbonate — and is one of EXTRIP's strong (1D) dialysis indications when the dysrhythmia is life-threatening.¹

Gut — and therefore the kinetics

TargetMuscarinic receptors on gut smooth muscle

Why hereAntimuscarinic blockade reduces gastric emptying and bowel motility, so unabsorbed drug pools in a slow gut and is released over days.² The target organ is also, uniquely, the reason the poisoning relapses. Established

At the bedsideReduced bowel sounds and delayed absorption are a reason to continue charcoal and to keep observing beyond the first day — the relapse on day two or three is the gut releasing its reservoir.²

Sodium and water balance

TargetThe ADH axis

Why hereAn ADH-like effect causes water retention and hyponatraemia, occasionally symptomatic.² It is a recognised chronic-therapy adverse effect and can be part of the acute picture. Inferred

At the bedsideCheck the sodium; a low sodium can itself provoke seizures and confounds the neurological assessment. Correction is a matter of rate and cause, guided by TOXBASE.

Timeline of effects

Carbamazepine's timeline is not a curve but a series of waves, because the drug is released from a slowed gut in instalments and regenerates an active metabolite as it goes.

Carbamazepine — a relapsing course from a self-delayed absorption

Time

What you seeWhat is happening

0–6 hOnset
What you seeNystagmus, ataxia, dysarthria and drowsiness, with anticholinergic features — mydriasis, dry mouth, tachycardia, agitation or hallucination. Vomiting is common.
What is happeningSodium-channel blockade produces the cerebellar signs and antimuscarinic activity the anticholinergic ones. Absorption is slow, so the concentration is still climbing — and the antimuscarinic ileus is already beginning to delay it further.²
6–24 hPeak, and the first trough
What you seeFluctuating and often deepening consciousness, a widening QRS in significant overdose, urinary retention and reduced bowel sounds. The patient may appear to improve, then not.
What is happeningPeak parent concentrations (conventional tablets peak ~12 h even therapeutically²), maximal sodium-channel and anticholinergic effects, and the active epoxide accumulating. The gut reservoir is forming behind a paralysed bowel.
Day 2–3Relapse from the reservoir
What you seeA patient who was improving deteriorates again — deeper coma, recurrent seizures, or a QRS that re-widens. The label instructs you to expect exactly this.
What is happeningThe self-inflicted latent phase: unabsorbed drug pooled in a slow gut is released in a second wave, and the continuously generated epoxide sustains the toxicity.² This is why MDAC continues and observation is prolonged — the poisoning is not over when the first level falls.
ResolutionSlow clearance
What you seeRecovery over days as the gut reservoir empties and hepatic metabolism clears parent and epoxide. Chronic users clear faster than the naïve.
What is happeningClearance is hepatic and, in chronic users, auto-induced; the half-life falls from ~36 h to 16–24 h with regular dosing.² The last of the exposure is the epoxide and the tail of the gut reservoir. Established

What the mechanism predicts at the bedside

Why the QRS is read like a tricyclic's

Because it is caused by the same lesion. Carbamazepine's sodium-channel blockade widens the QRS exactly as a tricyclic does,² so a broad QRS in a carbamazepine overdose is a marker of cardiac sodium-channel toxicity and is treated on the same model — sodium bicarbonate — as on the sodium-channel blockade page. A life-threatening dysrhythmia is also a strong EXTRIP dialysis indication.¹ Established

Why the absorption is the problem, and why charcoal is the answer

The antimuscarinic ileus makes carbamazepine slow to absorb and prone to relapse from a gut reservoir.² Multi-dose charcoal is therefore disease-modifying rather than merely decontaminating — it binds the reservoir before it is absorbed and interrupts recirculation, which is why EXTRIP continues it even during dialysis.¹ The mechanism turns a supportive measure into a specific one. Established

Why a single level cannot close the case

A parent-carbamazepine concentration can fall and then rise again as the gut releases more drug, and the active epoxide — not always measured — sustains toxicity independently.² Serial levels and continued observation through day two or three are the safe course; a single downward trend is the classic false reassurance here. Inferred

Why the sodium matters

The ADH-like hyponatraemia can be symptomatic and can lower the seizure threshold, compounding the sodium-channel-mediated seizure risk.² Checking and, where needed, cautiously correcting the sodium is part of managing the neurological picture, not a separate issue.

Why chronic and acute behave differently

Auto-induction means long-term users clear carbamazepine roughly twice as fast as first-time takers,² so the same ingested amount produces a more prolonged course in a naïve overdose — the reverse of theophylline's chronic-toxicity pattern, and a reminder to ask about regular use when judging how long to watch.

The antidote, from the poison's side

There is no single antidote, but there are two mechanism-directed treatments, each aimed at one of carbamazepine's borrowed toxidromes — and a removal strategy that has to respect the two-compartment problem.

Sodium bicarbonate for the QRS — the sodium-channel toxicity is treated on the tricyclic model, and a widening QRS is the trigger, exactly as on the sodium-channel blockade page.² Established
Multi-dose activated charcoal for the absorption — disease-modifying here because it binds the gut reservoir the antimuscarinic ileus creates and interrupts recirculation, and EXTRIP continues it during dialysis.¹ This is the treatment the mechanism most specifically demands.
Dialysis for the severe cases, and specifically for refractory seizures or life-threatening dysrhythmia,¹ with charcoal continued alongside because the gut reservoir would otherwise refill the blood.
Time and observation for the relapse — the single most useful thing after the acute measures is not to be caught out by the day-two deterioration the label tells you to expect.²

Critical appraisal

The self-delayed, relapsing absorption is the page's central and best-evidenced claim Established — it is stated in the label as something to anticipate, not merely a possibility,² and it follows mechanically from the drug's own antimuscarinic effect on the gut. It is the single most useful thing to know about carbamazepine overdose.
The sodium-channel account of the QRS is secure Established and clinically load-bearing: it dictates bicarbonate treatment and links the page to the tricyclics and the sodium-channel-blockade entry. The one caveat is that severe conduction disturbance is a high-concentration phenomenon and is less prominent than in a comparable tricyclic overdose.
The active-epoxide contribution is real but hard to quantify at the bedside. Inferred The epoxide is pharmacologically active and outlasts the parent,² but because it is not universally measured, its precise share of any individual patient's toxicity is usually unknown — which is itself a reason to treat the patient and the trajectory rather than a single parent-drug number.
EXTRIP's recommendations rest on 173 patients and very low quality evidence,¹ and the decision to suggest dialysis despite high protein binding reflects a moderate free fraction and the severity of the sickest cases rather than strong removal data. The concentration thresholds are consensus cut-points, and the strong indications are the end-organ ones — refractory seizures and life-threatening dysrhythmia.
The ADH-like hyponatraemia mechanism is inferred and its acute contribution is variable. Inferred It is well recognised in chronic therapy; in acute overdose its role is real but usually secondary to the CNS and cardiac effects, and it matters mainly as a modifiable contributor to the seizure threshold.
The chronic-versus-acute clearance difference is the mirror image of theophylline's. Here it is the acute overdose in a naïve patient that runs longer, because auto-induction has not shortened their half-life.² The practical lesson is the same in both drugs — ask about regular use — but the direction is reversed, which is a useful thing to hold explicitly rather than to assume.

References

1
Ghannoum M, Yates C, Galvao TF, Sowinski KM, Vo TH, Coogan A, Gosselin S, Lavergne V, Nolin TD, Hoffman RS; EXTRIP Workgroup. Extracorporeal treatment for carbamazepine poisoning: systematic review and recommendations from the EXTRIP workgroup. Clin Toxicol (Phila) 2014;52(10):993–1004. PubMed 25355482 · Recommendation set also at extrip-workgroup.org/carbamazepine. Source of the moderately-dialysable conclusion, the 2D suggestion in severe poisoning, the strong (1D) indications for refractory seizures and life-threatening dysrhythmia, the 2D indications for prolonged coma and for persistent toxicity despite MDAC, the <10 mg/L cessation threshold, the continue-MDAC-during-dialysis recommendation, and the 74-article / 173-patient / 6-death evidence base. Verified 12 Sep 2026 from the abstract and the workgroup's published recommendation page.
2
Tegretol 200 mg Tablets — Summary of Product Characteristics, Novartis. emc product 7845. §4.9 and §5.2 fetched and read in full. Source of the overdose features (CNS depression, nystagmus, ataxia, dysarthria, mydriasis, convulsions; QRS widening and syncope with cardiac arrest; respiratory depression and pulmonary oedema; delayed gastric emptying and reduced bowel motility; hyponatraemia and water intoxication from an ADH-like effect), the explicit instruction to anticipate relapse and aggravation on the 2nd and 3rd day due to delayed absorption, the recommendation for gastric evacuation, charcoal and haemodialysis, the ~12 h peak and 85–100% bioavailability, the 70–80% protein binding, the 0.8–1.9 L/kg volume of distribution, the CYP3A4 epoxide pathway and active 10,11-epoxide with its ~6 h half-life, the auto-induction shortening the parent half-life from ~36 h to 16–24 h, and the therapeutic range of 4–12 mg/L. Verified 12 Sep 2026.
3
TOXBASE — carbamazepine. National Poisons Information Service. toxbase.org (NHS login required. NPIS: 0344 892 0111. The authoritative UK source for charcoal regimens, bicarbonate thresholds, sodium correction and dialysis criteria. Login-gated, so not quoted here.)

Carbamazepine

At a glance