Why this poison is interesting

Every other poison in Band A has an antidote, a removal route, or a metabolic step you can block. Tricyclics have none. The drug is active as swallowed, the volume of distribution is so large that extracorporeal removal is arithmetically pointless, and the only pharmacological intervention is a manipulation of sodium and pH aimed at the ion channel rather than at the drug. What is left is a poison you have to out-wait while defending a heart.

Two things make it worth a mechanism page rather than an algorithm. The first is that the ECG — a free, immediate, repeatable test — outperforms the drug concentration so decisively that concentrations are not measured in UK practice at all, and the mechanism explains exactly why. The second is that acidaemia makes the poisoning worse through at least two independent routes, which turns a supportive intervention into a therapeutic one and makes intubation-and-ventilation a pharmacological decision.

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

The toxic principle

There is no toxic metabolite to identify. The parent drug is the poison, and several tricyclics generate metabolites that are themselves marketed antidepressants — amitriptyline yields nortriptyline, imipramine yields desipramine, dosulepin yields dothiepin-S-oxide.¹ The overdose is therefore a mixture of active compounds with overlapping pharmacology and long half-lives.

What makes tricyclics lethal is that they act at five targets simultaneously, and the resulting picture is not the sum of five toxidromes but a set of interacting failures.

Fast sodium channels — the lethal one: TCAs produce cardiac sodium channel blockade and can be classified as having type IA antiarrhythmic properties.¹ The block is use-dependent: it deepens with heart rate, which is precisely the situation an antimuscarinic tachycardia creates. Slowed phase 0 depolarisation widens the QRS, slows conduction and impairs contractility. Established
Alpha-1 adrenoceptors: Blocked, producing vasodilatation and hypotension — which lowers coronary perfusion of a myocardium that is already conducting and contracting badly. Established
Muscarinic receptors: Blocked, producing tachycardia, delirium, mydriasis, dry skin, urinary retention and gastric stasis that prolongs absorption.¹ The tachycardia is not benign: it deepens the use-dependent sodium block. Established
GABA-A receptors: Antagonised,¹ which is the usual explanation for seizures. A seizure then produces a lactic acidosis, and acidaemia worsens the sodium-channel block — the single most dangerous loop in this poisoning. Inferred
Histamine H1 receptors and potassium channels: Sedation from the first; QT prolongation from the second. Neither is usually what kills, but both contribute to the picture. Established

Toxicokinetics

Tricyclics — figures from the EXTRIP review's pharmacokinetic table¹
Parameter	Therapeutic	In overdose	Why it changes
Absorption	Rapidly absorbed from the gastrointestinal tract. Bioavailability is modest and variable — amitriptyline 31–61%, imipramine 29–77%, dosulepin about 30%¹	Antimuscarinic effects decrease gastrointestinal motility and can prolong the time to peak drug concentration¹	The drug slows its own absorption. That is why a patient can look well at two hours and be in trouble at five, and why the window for gastrointestinal decontamination is not the usual one. It is the same phenomenon as opioid or anticholinergic co-ingestion in paracetamol poisoning — except here the drug is doing it to itself.
Protein binding	High — 82–96% for amitriptyline, 90–98% clomipramine, 93–97% trimipramine; bound mainly to alpha-1 acid glycoprotein and lipoproteins¹	Binding is pH-sensitive: acidaemia reduces it and raises the free fraction; alkalosis increases it¹	This is why pH is a therapeutic variable rather than a supportive one. It is also why the acidosis from a seizure, or from hypoventilation, is not merely a marker of severity — it liberates more drug. The loop runs the other way too: alkalinisation is proposed to work partly because increased protein binding reduces free drug availability.¹
Volume of distribution	Enormous, and the defining property. Amitriptyline 5–20 L/kg; nortriptyline 21–57; imipramine 15–30; dosulepin 11–78¹	Unchanged — it was already vast	A Vd of 20 L/kg in a 60 kg adult is a notional 1,200 litres. Almost none of the drug is in plasma at any moment, which is the arithmetic that makes every extracorporeal technique futile and every plasma concentration uninformative.
Half-life	Long: amitriptyline 31–46 h, nortriptyline 18–93 h, protriptyline 54–198 h, clomipramine 22–84 h¹	Longer still, with active metabolites contributing¹	Long half-lives with a huge Vd mean recovery is by redistribution and slow hepatic clearance, not by elimination of a plasma pool. The clinical corollary is that improvement is not evidence the drug has gone, and neither is the passage of a few hours.
Elimination route	First-pass hepatic metabolism with high endogenous clearance; numerous metabolites, many pharmacologically active, most excreted in urine¹	Unchanged in kind. Nothing saturates in a way that changes the mechanism	There is no enzyme worth blocking and no cofactor worth replacing, because there is no bioactivation. That absence is the reason this is the one Band A poison with no antidote directed at the drug.
Order of kinetics	First order	First order	Tricyclics are not a saturation poison. The dose–response steepness comes from the use-dependence of the channel block and from the pH feedback, not from any change in clearance.
Dialysability	—	Not dialysable. EXTRIP: "We recommend NOT to perform ECTR in patients with TCA poisoning" (1D).¹ The workgroup screened 1,312 studies, included 77, and abstracted data on 108 patients including 12 fatalities — all case reports, case series and one poor-quality observational study¹	The review makes the reason explicit and arithmetical: the limiting factor is the massive volume of distribution of TCAs and not extraction by the filter or adsorbent column.¹ Their worked example: a 60 kg patient who ingests 2,400 mg of amitriptyline (Vd 20 L/kg) has a plasma concentration of 2,000 ng/mL; four hours of charcoal haemoperfusion at 350 mL/min blood flow, even assuming a perfect extraction ratio of 100%, removes a trivial fraction of the body burden. A perfect filter cannot clean a compartment the drug is not in.

Metabolism and the metabolites

This section exists in every monograph on this site for consistency, and in tricyclic poisoning it is nearly empty — which is worth saying rather than hiding. There is no bioactivation. Hepatic metabolism generates numerous metabolites, many with pharmacological activity, most excreted in urine.¹ Metabolism here reduces toxicity slowly; it does not create it.

Tricyclics — a receptor map, since there is no metabolic one

Tricyclic antidepressantLipophilic weak base. Active as swallowed¹
Distribution — not metabolismMyocardial and brain concentrations 40–200 times plasma¹
Drug in cardiac and cerebral tissueThe compartment that matters, and the one no filter can reach
Fast Na⁺ channel block (type IA)Wide QRS, negative inotropy, arrhythmiaUse-dependent — deepens with tachycardia and with acidaemia¹
α₁-adrenoceptor blockVasodilatation and hypotensionLowers coronary perfusion of an already failing myocardium
Muscarinic blockTachycardia, delirium, gastric stasisProlongs absorption¹; the tachycardia worsens the channel block
GABA-A antagonism¹SeizuresAnd a seizure produces the acidosis that worsens everything else

What changes the answer

Anything that lowers pH. Seizure, hypoventilation, hypoperfusion, and a ventilator set to a normal minute volume in a patient who was compensating. Established
Anything that raises heart rate, because the sodium channel block is use-dependent — including the drug's own antimuscarinic tachycardia. Inferred
Co-ingested drugs that also block sodium channels — flecainide, quinine, chloroquine, cocaine, some antihistamines and carbamazepine. The effects are additive at the same channel, and the ECG will not distinguish them.
Which tricyclic. The class is not uniform: half-lives run from 8 hours to nearly 200, and volumes of distribution from 5 to 78 L/kg.¹ Dosulepin has a particularly poor reputation in UK overdose, and lofepramine a comparatively benign one.

Elimination and accumulation

Tricyclic poisoning is the library's clearest example of a poison that accumulates somewhere you cannot reach. There is no ion trapping to exploit, no enterohepatic recirculation worth interrupting, no saturated pathway to unblock, and no removal technique that works. What determines recovery is redistribution out of tissue and first-pass hepatic clearance, both of which are slow.

The one manipulable variable is pH, and it acts on distribution rather than on elimination. Alkalinisation does not remove tricyclic from the body; it changes how much of what is there is free and how tightly the free fraction engages the channel. That is a genuinely unusual therapeutic target and worth naming precisely.

The EXTRIP position, and why it is a strong recommendation on weak evidence

EXTRIP's TCA statement is one of its shortest and firmest: "We recommend NOT to perform ECTR in patients with TCA poisoning" (1D) — a strong recommendation resting on very low quality evidence.¹ That combination is unusual, and it is defensible here for a reason that does not apply to most of their other poisons: the recommendation does not depend on the clinical data at all. It follows from the volume of distribution, which is measured, large and not in dispute. The clinical literature could hardly have shown benefit even if benefit existed.

Target organs — and why those

Heart — the fast sodium channel

TargetVoltage-gated Na⁺ channels in Purkinje fibres and ventricular myocardium

Why hereTwo reasons compound. Concentration: myocardial drug is reported at 40–200 times plasma.¹ Physiology: the block is use-dependent, so it is deepest in the tissue depolarising most often — and the drug's own antimuscarinic tachycardia raises that rate. The result is slowed phase 0 depolarisation, which widens the QRS, slows intraventricular conduction and reduces contractility at the same time. Terminal right-axis deviation appears because the right bundle and right ventricle are affected disproportionately. Established

At the bedsideQRS duration is the measurement, and prominent R wave in aVR with a rightward terminal axis is the same phenomenon seen from a different lead. A Brugada-like pattern may appear.¹

Brain — seizures

TargetGABA-A receptors, antagonised¹

Why hereThe same 40–200-fold tissue concentration applies to brain as to myocardium.¹ Reduced GABA-ergic inhibition lowers seizure threshold, and the antimuscarinic and antihistaminergic actions add delirium and sedation on top. What makes the seizures dangerous is not the seizure itself but its metabolic consequence: a lactic acidosis that immediately worsens the cardiac lesion. Inferred

At the bedsideSeizures occurred in 16 of 79 patients (20%) in the Liebelt cohort and are strongly associated with QRS prolongation.⁴ A seizing tricyclic patient is a cardiac emergency.

Vasculature

Targetα₁-adrenoceptors, blocked

Why hereThis is not organ-selective so much as unavoidable: α₁ blockade produces vasodilatation wherever arteriolar tone depends on it. Its importance is entirely in its interaction — hypotension in a heart with impaired conduction and impaired contractility reduces coronary perfusion, which worsens both. Established

At the bedsideHypotension that is fluid-unresponsive is one of the standard triggers for bicarbonate,¹ and it is also the point at which a pure vasodilator picture and a pure cardiogenic picture become impossible to separate at the bedside.

Timeline of effects

Tricyclic poisoning has no latent phase, and unusually for this library that is the dangerous property rather than a reassuring one. Absorption is prolonged by the drug's own antimuscarinic effect,¹ so the picture builds; but once it builds, it can move to arrest within minutes.

Tricyclics — the untreated course, and what is happening underneath it

Time

What you seeWhat is happening

0–2 hAntimuscarinic
What you seeDry mouth, mydriasis, warm dry skin, urinary retention, sinus tachycardia. Drowsiness. The patient often looks like an anticholinergic overdose and not much more.
What is happeningMuscarinic and H1 blockade dominate because they need lower concentrations than sodium-channel block. Gastric emptying is already slowing, which prolongs absorption¹ — so this apparently mild phase is one in which the dose is still arriving.
1–6 hCardiac and CNS — the risk window
What you seeQRS widening. Prominent terminal R wave in aVR. Hypotension. Agitation or reduced consciousness. Seizures. Broad-complex arrhythmias. Deaths overwhelmingly occur in this window.
What is happeningTissue concentrations, 40–200 times plasma,¹ reach the threshold for use-dependent sodium-channel block. The antimuscarinic tachycardia deepens it. GABA-A antagonism lowers the seizure threshold,¹ and a seizure adds a lactic acidosis that reduces protein binding and deepens the block further.
6–24 hPlateau, then slow improvement
What you seeIf the first six hours are survived without significant QRS widening, the risk falls substantially. Conscious level improves before the ECG fully normalises. Delirium may persist longer than anything cardiac.
What is happeningRedistribution from tissue continues and hepatic clearance grinds on, but half-lives of 31–46 hours for amitriptyline and up to 93 hours for nortriptyline¹ mean the body burden is barely reduced. Improvement reflects falling tissue concentration at the channel, not elimination.
Beyond 24 h
What you seePersistent delirium and tachycardia in some. Late deterioration is uncommon in an alert patient with a normal ECG, but a patient who has required bicarbonate is not finished when their pH is corrected.
What is happeningActive metabolites with their own long half-lives continue to circulate.¹ Any new acidosis — a hospital-acquired pneumonia, a period of hypoventilation, an unfortunate ventilator setting — can re-liberate free drug from protein binding. Inferred

What the mechanism predicts at the bedside

Why the QRS predicts and the drug concentration does not

Boehnert and Lovejoy studied 49 patients prospectively, split by maximal limb-lead QRS duration.³ In the 13 patients with a QRS under 0.10 s there were no seizures and no ventricular arrhythmias. In the 36 with a QRS of 0.10 s or longer there was a 34% incidence of seizures and a 14% incidence of ventricular arrhythmias. Seizures occurred at any QRS ≥0.10 s; ventricular arrhythmias were seen only at ≥0.16 s. And the key negative finding: "serum drug levels failed to predict the risk of seizures or ventricular arrhythmias accurately."³ Established

The mechanism explains the whole result. The QRS measures the consequence of channel blockade in cardiac tissue, which is where the drug is; the serum concentration measures the compartment holding a small and variable fraction of the body burden.¹ This is one of the cleanest demonstrations in toxicology that the right test is the one sampling the right compartment.

What lead aVR adds, and what it does not

Liebelt and colleagues compared aVR measurements with QRS duration in 79 patients, of whom 16 (20%) had seizures and 5 (6%) ventricular arrhythmias.⁴ The terminal R wave in aVR was larger in those who had events (4.4 versus 1.8 mm, p<0.001), as was the R/S ratio (1.4 versus 0.5, p<0.001). Sensitivity of RaVR ≥3 mm was 81% and of R/SaVR ≥0.7 was 75%, against 82% for QRS >100 ms. Positive predictive values were 43%, 46% and 35% respectively. On multiple logistic regression, RaVR ≥3 mm was the only ECG variable that significantly predicted seizures and arrhythmias (OR 6.9, 95% CI 1.2–40, p=0.03).⁴ Established

Why acidaemia is a treatment failure, not just a marker

Because it acts on the lesion through two routes at once: it reduces protein binding, raising the free fraction of a highly bound drug,¹ and it worsens the sodium-channel block directly. Alkalinisation reverses both. This is why a period of hypoventilation during intubation, or a ventilator set to a textbook minute volume in a patient who had been compensating, is a pharmacological event and not merely a physiological one. Hyperventilation alone — systemic alkalosis achieved without any sodium — also improves hypotension and cardiac conduction disturbances.¹ Established

Why sodium bicarbonate rather than sodium or bicarbonate alone

Because two mechanisms are wanted and it delivers both. The EXTRIP review states it carefully: sodium bicarbonate "may ameliorate hypotension due to volume and sodium loading, and improves myocardial conduction disturbances presumably by creating a sodium load and also by inducing alkalosis", with the alkalosis effect "potentially due to increased protein binding thereby reducing free drug availability and altering the charge of the TCA–receptor complex".¹ Note the hedging in the original — presumably, potentially. That is honest, and it is preserved here. Inferred

Hypertonic sodium alone has also demonstrated benefit in a few animal studies and isolated cases,¹ which suggests the sodium load carries real weight independently of pH. The workgroup's conclusion is that bicarbonate "combines the effect of sodium loading and alkalosis and remains the therapy of choice", particularly for seizures, fluid-unresponsive hypotension, or the typical ECG findings.¹

Why nobody sends a tricyclic concentration

Because it would not change anything. It does not predict seizures or arrhythmias,³ it samples the wrong compartment,¹ it takes longer to return than the patient takes to deteriorate, and no threshold triggers a treatment that the ECG does not trigger sooner. This is a case where the absence of a test is a mechanistic conclusion rather than a resource constraint.

The antidote, from the poison's side

The EXTRIP review is blunt: "there is no specific antidote for TCA poisoning", but "there is a fundamental role for sodium bicarbonate".¹ Read from the poison's side, that distinction matters. Bicarbonate is not an antagonist, a chelator or a scavenger — it does not interact with the drug at all. It changes the conditions under which the drug meets its target.

The sodium load raises the transmembrane sodium gradient, so the sodium current through unblocked channels is larger. It is a competition against the block rather than a removal of it. Inferred
The alkalosis increases protein binding, reducing free drug, and alters the charge of the drug–receptor complex.¹ Note that this is exactly the reverse of the manoeuvre used in salicylate poisoning, where alkalinisation ionises a weak acid to trap it. Same chemistry, opposite drug, opposite intent. Inferred
Hyperventilation achieves the alkalosis without the sodium¹ — useful to know when a patient is already intubated, and a reminder that pH and sodium are separable levers.
Lipid emulsion is a rescue therapy for refractory cardiovascular collapse, on the hypothesis that a lipid compartment sequesters a lipophilic drug. Given the 40–200-fold tissue distribution already present,¹ the mechanistic plausibility is lower here than in local anaesthetic toxicity, and the evidence is case-level. Unlicensed for this use. Inferred
Extracorporeal life support substitutes for the circulation while redistribution occurs. It is not drug removal and should not be confused with the ECTR that EXTRIP recommends against.¹

Critical appraisal

The QRS thresholds rest on one small prospective cohort in which nobody died. Inferred Forty-nine patients, 1985, all survived.³ The 100 ms and 160 ms figures are mechanistically coherent and universally taught, but they are descriptive findings from a single series, not validated decision thresholds. Using them as binary rules attributes a precision the study cannot support.
The aVR literature is more equivocal than its reputation. Inferred Liebelt's own comparison put RaVR ≥3 mm at 81% sensitivity against 82% for QRS >100 ms — essentially identical — with positive predictive values of 43% versus 35%, and an odds ratio whose confidence interval spans 1.2 to 40.⁴ The finding that aVR was the only variable surviving regression is real and worth knowing; the widespread claim that aVR outperforms the QRS overstates it.
The mechanism of sodium bicarbonate is not settled, and the source says so. Inferred The EXTRIP review uses presumably and potentially about the sodium-load and alkalosis contributions, and notes that hypertonic sodium alone has shown benefit in animal studies and isolated cases.¹ The relative weight of sodium loading, of increased protein binding, and of altered charge at the receptor has not been separated in humans. The therapy works; the account of why is a well-supported hypothesis.
The seizure–acidosis–arrhythmia loop is inferred rather than demonstrated. Inferred Each link is documented separately — GABA-A antagonism,¹ pH-dependent binding,¹ pH-dependent channel block — but the loop as a causal sequence in human overdose is a reconstruction. It is a good one, it predicts the observed tempo, and it should still be labelled as what it is.
EXTRIP's strong recommendation against dialysis rests on very low quality evidence, and this is the case where that combination is most defensible.¹ The recommendation follows from a measured volume of distribution rather than from the 108 patients of case-level data, and their worked example is reproducible by anyone. Contrast this with their paracetamol or ethylene glycol thresholds, where the numbers really are consensus positions.
The class is treated as homogeneous and is not. Half-lives span 8 to 198 hours and volumes of distribution 5 to 78 L/kg across the drugs in EXTRIP's own table.¹ "Tricyclic overdose" is a useful clinical category and a poor pharmacological one, and generalisations about time to safety are correspondingly weak.
Tricyclic prescribing has fallen, and the poisoning has not become less lethal. Amitriptyline is now prescribed more often for neuropathic pain and migraine prophylaxis than for depression, which changes who has access to it without changing what a large ingestion does. A page written around the drug class name risks under-representing where the tablets actually come from.

References

1
Yates C, Galvao T, Sowinski KM, Mardini K, Botnaru T, Gosselin S, Hoffman RS, Nolin TD, Lavergne V, Ghannoum M; EXTRIP workgroup. Extracorporeal treatment for tricyclic antidepressant poisoning: recommendations from the EXTRIP Workgroup. Semin Dial 2014;27(4):381–9. PMC4282541 Open access. Source of the Table 1 pharmacokinetic figures (bioavailability, protein binding, half-life and volume of distribution for each drug), the 40–200-fold myocardial and brain tissue concentration, the alpha-1 acid glycoprotein binding, the type IA antiarrhythmic classification and GABA-A antagonism, the antimuscarinic gastric-stasis effect on absorption, the quoted sodium bicarbonate mechanism with its hedging, the hyperventilation and hypertonic-sodium observations, the not-dialysable recommendation with its 1,312/77/108-patient evidence base, and the worked haemoperfusion calculation. Verified 1 Sep 2026 from the full text.
2
Body R, Bartram T, Azam F, Mackway-Jones K. Guidelines in Emergency Medicine Network (GEMNet): guideline for the management of tricyclic antidepressant overdose. Emerg Med J 2011;28(4):347–68. PubMed 21436332 Cited as the UK emergency-medicine guideline for this poisoning. Not used as the source of any figure on this page. Citation verified 1 Sep 2026.
3
Boehnert MT, Lovejoy FH Jr. Value of the QRS duration versus the serum drug level in predicting seizures and ventricular arrhythmias after an acute overdose of tricyclic antidepressants. N Engl J Med 1985;313(8):474–9. PubMed 4022081 Prospective study of 49 patients. Source of the QRS <0.10 s versus ≥0.10 s comparison, the 34% seizure and 14% arrhythmia incidences, the ≥0.16 s arrhythmia threshold, and the finding that serum drug levels failed to predict risk. All 49 patients survived; the thresholds are descriptive, not validated. Verified 1 Sep 2026.
4
Liebelt EL, Francis PD, Woolf AD. ECG lead aVR versus QRS interval in predicting seizures and arrhythmias in acute tricyclic antidepressant toxicity. Ann Emerg Med 1995;26(2):195–201. PubMed 7618783 Prospective cohort of 79 patients, 16 seizures and 5 ventricular arrhythmias. Source of every aVR figure quoted above including the sensitivities, positive predictive values and the odds ratio of 6.9 (95% CI 1.2–40). Verified 1 Sep 2026.
5
TOXBASE — tricyclic antidepressants; amitriptyline; dosulepin. National Poisons Information Service. toxbase.org (NHS login required. NPIS: 0344 892 0111. The authoritative UK source for bicarbonate thresholds and dosing, arrhythmia management and observation periods. Login-gated, so not quoted here.)

Tricyclic antidepressants

At a glance