Why this poison is interesting

The SSRIs exist because the tricyclics were lethal in overdose to a population defined by its risk of overdose. That was the design brief, and it was met: fluoxetine's UK label records that cases of overdose of fluoxetine alone usually have a mild course and that fatality attributed to overdose of fluoxetine alone has been extremely rare.⁴ A drug class whose own summary of product characteristics says that is a genuine pharmacological achievement.

So the interest is not in what an SSRI does alone. It is in what an SSRI does with something else. Serotonin toxicity is the library's clearest example of a poisoning that is combinatorial — produced not by a dose but by an arithmetic of mechanisms, in which a drug that increases release, a drug that blocks reuptake and a drug that prevents breakdown each contribute a term. Most single-agent SSRI overdoses do not produce it. Most severe serotonin toxicity involves two or more agents.

The third reason is diagnostic, and it is the reason clinicians get this wrong. The syndrome is defined at the neuromuscular junction, not in the mental state. Agitation and fever are the features that draw attention; clonus is the feature that carries the diagnosis, and it is the one nobody examines for unless they are looking.

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

The toxic principle

Serotonin toxicity is excess serotonin at the synapse, and the reason it is a useful concept rather than a vague one is that a small number of independent mechanisms feed the same pool. An SSRI blocks the reuptake transporter. A monoamine oxidase inhibitor prevents degradation. Amphetamines and MDMA force release. Some drugs — tramadol among them — do more than one. Lithium and triptans act on receptors downstream.

The receptor identity of the syndrome is usually attributed to 5-HT2A, which is the pharmacological basis for cyproheptadine and the reason the Hunter study used treatment with a 5-HT2A antagonist as one of its outcome measures.⁵ The receptor attribution is standard and is not seriously contested; what is much less secure is the leap from it to a specific benefit of a specific drug, which the appraisal section returns to.

The 'triad' is the wrong shape, and Hunter is the reason we know

Serotonin toxicity is traditionally taught as a triad of mental state change, autonomic instability and neuromuscular excitation, all three weighted equally. Dunkley and colleagues studied 2222 patients admitted after overdose of a serotonergic drug and derived decision rules by CART analysis from a learning dataset of 473 SSRI-alone overdoses. Of the many clinical features associated with serotonin toxicity, only clonus (inducible, spontaneous or ocular), agitation, diaphoresis, tremor and hyperreflexia were needed for accurate prediction — with hypertonicity and a maximum temperature above 38 °C added because they were universal in the life-threatening cases the learning dataset had lacked.⁵ The resulting Hunter criteria were more sensitive (84% vs 75%) and more specific (97% vs 96%) than Sternbach's.⁵ The triad survives in teaching; the data say the syndrome is mostly neuromuscular. Traditional teaching

The practical consequence of that finding is a change in examination technique rather than in knowledge. Clonus is elicited, not observed. Dorsiflex the ankle briskly and hold it. Look at the eyes for slow, continuous horizontal oscillation. Test tone. A patient can be agitated and febrile from a dozen causes; sustained inducible clonus in someone who has taken a serotonergic drug is a much narrower finding.

Toxicokinetics

The SSRIs are not one drug, and the kinetic differences between them are the reason citalopram behaves unlike sertraline in overdose. The table takes sertraline as the reference and calls out the divergences.

SSRIs — a class whose members differ where it matters
Parameter	Therapeutic	In overdose	Why it changes
Absorption	Sertraline peaks at 4.5 to 8.4 h; food does not significantly change bioavailability¹. Citalopram Tmax about 3.8 h, oral bioavailability about 80%²	Peak effects correspondingly late; the label-derived windows are the basis for a long observation period	A peak measured in hours, not minutes, is why an SSRI overdose looks well early. A patient who is asymptomatic at two hours has not yet reached the peak of a sertraline ingestion.
Protein binding	Sertraline approximately 98%¹; citalopram and escitalopram below 80%²³	Unchanged in any way that alters management	Sertraline's binding is the direct reason its label says dialysis is unlikely to help¹. Citalopram's much lower binding does not rescue it, because its volume of distribution is enormous.
Volume of distribution	Citalopram about 12.3 L/kg²; escitalopram 12 to 26 L/kg³	Unchanged	A drug distributed through 12 to 26 litres per kilogram is overwhelmingly not in the blood. This is the digoxin argument applied to an antidepressant — see digoxin — and it defeats any extracorporeal approach on its own.
Metabolism	Sertraline: extensive first-pass, via CYP3A4, CYP2C19 and CYP2B6¹. Escitalopram: demethylation primarily by CYP2C19, with CYP3A4 and CYP2D6 contributing³	Pathways unchanged; interaction potential is the clinical issue	Multiple pathways make the SSRIs relatively robust to any single interaction — unlike tramadol, whose activation depends on one polymorphic enzyme. Sertraline levels are about 50% higher in CYP2C19 poor metabolisers, and the label says the clinical meaning is not clear.¹
Active metabolites	Sertraline → N-desmethylsertraline, half-life 62 to 104 h¹. Citalopram → demethyl- and didemethylcitalopram, all active metabolites are themselves SSRIs, weaker than the parent²	Contribute to a long tail rather than to acute severity	A metabolite that is a weaker version of the parent is not a bioactivation. Contrast paracetamol, where the metabolite is the poison. Here the metabolites merely extend the pharmacology.
Half-life	Sertraline approximately 26 h (range 22–36)¹; citalopram about 1.5 days²; escitalopram about 30 h³	Prolonged in hepatic impairment — sertraline's AUC rises three-fold in liver damage¹	Long half-lives mean a serotonergic drug is still present days after it was stopped. This is the kinetic basis of the washout interval before starting an MAOI, and the reason fluoxetine's is the longest of all.
Elimination	Sertraline: extensively metabolised, less than 0.2% excreted unchanged in urine¹. Citalopram: 85% hepatic, 15% renal, about 12% of the daily dose as unchanged drug in urine²	Unchanged	Hepatic clearance of a highly distributed, highly bound drug. There is no renal handle to pull — no ion trapping, no urinary alkalinisation, nothing analogous to salicylate.
Dialysability	—	No. Sertraline's label: due to the large volume of distribution of sertraline, forced diuresis, dialysis, haemoperfusion and exchange transfusion are unlikely to be of benefit¹. Fluoxetine's label says the same⁴. EXTRIP has published no recommendation covering the SSRIs⁶	Two independent UK labels reach the same conclusion, and sertraline's states the premise — a large volume of distribution, alongside its 98% protein binding.¹ Fluoxetine's states only the conclusion.⁴ Citalopram's and escitalopram's put a number on the distribution volume: 12.3 and 12 to 26 L/kg.²³ This is one of the few dialysability rows in the library where an EXTRIP absence costs nothing, because the physical chemistry answers the question.

Metabolism and the metabolites

There is no toxic metabolite anywhere in this class, and the pathway below is drawn to show the interaction, because that is where the danger is. The relevant biochemistry is not what happens to the SSRI but what happens to serotonin while the SSRI is present.

Serotonin at the synapse — and the four places drugs interfere

TryptophanDietary precursor
Serotonin (5-HT), stored in vesicles
Released into the synapseSynaptic serotoninThe pool that matters. Amphetamines and MDMA force release into it — see amphetamines and MDMA
Monoamine oxidase A5-hydroxyindoleacetic acidDegradation. MAO inhibitors remove this limb entirely — the mechanism with no ceiling
Reuptake into the presynaptic neuroneThe therapeutic target. Blockade is saturable: once every transporter is occupied, more drug adds nothing
Serotonin toxicityNeuromuscular excitation first, then autonomic instability, then mental state change⁵

Citalopram's metabolism deserves separate mention because it is the only member of the class whose own label reads like a cardiac drug's. Citalopram is metabolised to demethylcitalopram, didemethylcitalopram, citalopram-N-oxide and an inactive deaminated propionic acid derivative, and the label is explicit that all the active metabolites are also SSRIs, although weaker than the parent compound, with unchanged citalopram the predominant compound in plasma.² Nothing in that metabolic scheme explains the QT effect, which belongs to the parent molecule and is discussed under target organs.

Elimination and accumulation

Elimination is hepatic, slow, and almost entirely irrelevant to the acute management of an overdose — which is itself worth stating, because it means the whole question of enhanced elimination can be dismissed on arithmetic rather than deferred to a guideline.

Accumulation matters in a different sense: between drugs rather than within one. Fluoxetine and its metabolite are the longest-lived agents in the class, and the reason a washout interval is required before starting a monoamine oxidase inhibitor is that a drug stopped a fortnight ago may still be occupying serotonin transporters. Fluoxetine's label makes the mirror-image point about the tricyclics — that an extended time for close medical observation may be needed in a patient who has taken a tricyclic overdose if they are also taking, or have recently taken, fluoxetine.⁴ A drug's half-life becomes a safety issue when the danger is combinatorial.

Target organs — and why those

Spinal cord and neuromuscular system

TargetSerotonergic modulation of motor neurone excitability

Why hereThe organ that defines the syndrome. Dunkley's analysis of 2222 serotonergic overdoses found that clonus, agitation, diaphoresis, tremor and hyperreflexia were sufficient for accurate prediction of serotonin toxicity⁵ — four of those five are neuromuscular. The descending raphe–spinal projections that modulate motor neurone excitability are the anatomy usually invoked, and they also account for the rostrocaudal gradient, though that specific explanation is a reconstruction rather than a demonstration. Inferred

At the bedsideInducible, spontaneous or ocular clonus; hyperreflexia; tremor; hypertonicity — greater in the legs than the arms. Clonus must be elicited deliberately: brisk ankle dorsiflexion, held. This is the examination that makes the diagnosis.

Heart — citalopram and escitalopram specifically

TargetCardiac repolarisation, and at high concentration the sodium channel

Why hereThe one organ where the class is not a class. Citalopram's overdose section lists QT prolongation, torsade de pointes, QRS prolongation, bundle branch block, bradycardia, cardiac arrest and atrial and ventricular arrhythmia²; sertraline's lists QTc prolongation and torsade and recommends ECG monitoring in all sertraline overdoses¹; fluoxetine's lists arrhythmias and very rare torsade⁴. Escitalopram's label lists QT prolongation and arrhythmia among overdose features but describes most cases as mild or asymptomatic, with 400–800 mg taken without severe symptoms.³ The QRS widening reported for citalopram is what places it, alone in this class, alongside the sodium-channel blockade page. Established

At the bedsideECG in every SSRI overdose, on the labels' own instruction¹². Citalopram is the member for which the ECG most often changes management, and the one whose label reports QRS widening and bundle branch block.

Brain — thermoregulation and autonomic control

TargetHypothalamic and brainstem serotonergic pathways

Why hereHyperthermia in serotonin toxicity is produced by muscle — sustained hypertonicity and clonus generating heat faster than it can be lost — with central thermoregulatory failure contributing. This distinction is not academic: it is why the temperature tracks the neuromuscular findings and why cooling a rigid patient without addressing the rigidity is inadequate. Dunkley added hypertonicity and a maximum temperature above 38 °C to the criteria precisely because both were universal in the life-threatening cases.⁵ Inferred

At the bedsideHyperthermia, diaphoresis, tachycardia, labile blood pressure, mydriasis, diarrhoea. Temperature above 38 °C with hypertonicity is the combination that marks severity, not agitation alone.

Kidney and the sodium balance

TargetAntidiuretic hormone regulation

Why hereSSRIs are a well-recognised cause of hyponatraemia through inappropriate ADH secretion, particularly in older patients, and escitalopram's overdose section lists hyponatraemia and hypokalaemia explicitly among the reported features.³ The mechanism is usually described as SIADH; the serotonergic step linking transporter blockade to vasopressin release is not established in humans, which is why this card is inferred rather than established. Inferred

At the bedsideConfusion or seizure in an older patient recently started on an SSRI, with a low sodium — a therapeutic-dose problem more often than an overdose one, and a reason to check the sodium in any SSRI-related presentation.

Skeletal muscle

TargetSustained contraction in severe serotonin toxicity

Why hereRhabdomyolysis is the downstream consequence of unrelieved hypertonicity and clonus, not a direct drug effect on muscle. Citalopram's label records that in severe poisoning there may be hyperpyrexia and elevation of serum creatine kinase, and that rhabdomyolysis is rare.² The label's word is rare, and this page keeps it — the complication is real and its frequency is low. Established

At the bedsideCreatine kinase and renal function in any patient with hypertonicity or hyperthermia. It is the mechanism by which a neurological syndrome becomes a renal one.

Timeline of effects

SSRI overdose — late peak, and a syndrome that can appear at therapeutic doses

Time

What you seeWhat is happening

0–2 hDeceptively well
What you seeNausea and vomiting; often nothing else. The patient looks fine.
What is happeningAbsorption under way but incomplete. Sertraline's peak is at 4.5 to 8.4 h¹ and citalopram's Tmax about 3.8 h² — nothing has reached its maximum yet.
1 hThe decontamination window closes
What you see—
What is happeningCitalopram's label puts charcoal at more than 5 mg/kg within 1 hour, and records 50% reduction in absorption when charcoal was given half an hour after ingestion.² After that the argument weakens quickly.
4–8 hPeak concentration
What you seeSomnolence, tachycardia, tremor, agitation, dizziness, vomiting.¹ Coma is reported but less frequent.¹
What is happeningPeak plasma concentrations.¹ For most single-agent SSRI overdoses this is as bad as it becomes.
6–12 hThe citalopram window
What you seeSeizures and QT prolongation, in a patient who may have looked well.²
What is happeningCitalopram's overdose section lists convulsion first among reported features and includes QT prolongation, torsade de pointes and QRS prolongation.² This is the reason a citalopram overdose is observed longer than a sertraline one, and why the ECG is repeated rather than done once.
Hours, any timeSerotonin toxicity — if a second agent is involved
What you seeClonus, hyperreflexia, tremor, agitation, diaphoresis, hypertonicity, temperature above 38 °C.⁵ Legs worse than arms.
What is happeningNot a phase of SSRI overdose so much as a separate event superimposed on it. Onset is usually within hours of the combination, not of the overdose — which is why it can appear in someone who has taken no overdose at all, having simply been prescribed linezolid.
24 h+Resolution, or a long tail
What you seeRecovery in most single-agent cases. Serotonin toxicity resolves as the agents clear.
What is happeningHalf-lives of 26 h for sertraline¹, 1.5 days for citalopram² and 30 h for escitalopram³, with active metabolites longer still — N-desmethylsertraline at 62 to 104 h.¹ Clinical recovery precedes pharmacological clearance by days, which matters only for what is started next.

What the mechanism predicts at the bedside

Examine for clonus, deliberately. It is the single highest-yield finding in the Hunter rules⁵ and it does not appear unless you look for it.
Look at the legs before the arms. The rostrocaudal gradient is characteristic and is easily missed if the examination stops at the biceps reflex.
Ask what else the patient takes, including antibiotics. Linezolid is a reversible MAOI and is the interaction most often missed; methylene blue is the second.
Which SSRI it is changes the plan. Citalopram's label lists convulsion, QT prolongation, torsade, QRS prolongation and bundle branch block²; escitalopram's describes most overdoses as mild³. These are not interchangeable drugs in overdose.
Get an ECG in every SSRI overdose — sertraline's own label recommends ECG monitoring in all ingestions of sertraline overdoses¹ — and repeat it.
A well patient at two hours is uninformative. The peak is at four to eight hours¹, and citalopram's seizures fall later still.
Check the sodium, particularly in an older patient. Escitalopram's label lists hyponatraemia among overdose features³ and the therapeutic-dose problem is commoner than the overdose one.
Check creatine kinase if there is hypertonicity or fever. Rhabdomyolysis is the route from a neurological syndrome to a renal one, and citalopram's label names it as rare.²
Do not ask for dialysis. Two UK labels state independently that it will not help¹⁴, and the volume of distribution explains why.²³
A partial response to benzodiazepines does not exclude serotonin toxicity — sedation reduces the agitation and the muscle activity, which is useful, without touching the serotonin.

The antidote, from the poison's side

All four UK labels consulted for this page say there is no specific antidote.¹²³⁴ What exists instead is a set of interventions aimed at consequences: benzodiazepines for the agitation and the muscle activity, active cooling for the temperature, sodium bicarbonate for a widened QRS, and cyproheptadine for the receptor.

Benzodiazepines: The first-line intervention in practice, and mechanistically sensible: they reduce agitation, reduce muscle tone and therefore reduce heat production. Citalopram's label specifically directs that convulsions be controlled with intravenous diazepam if frequent or prolonged.² They do nothing to the serotonin — they treat the consequences that kill.
Active cooling and paralysis: For hyperthermia in severe serotonin toxicity. The heat is generated by muscle, so the intervention that most reliably stops it is the one that stops the muscle. This is the mechanistic reason paralysis features in severe cases where antipyretics do nothing.
Cyproheptadine: A 5-HT2A antagonist, and the closest thing to a mechanism-directed treatment. Dunkley's study used treatment with a 5-HT2A antagonist as one of its three outcome measures⁵, which tells you the drug was in use in that cohort — it does not tell you the drug worked, and the study was not designed to answer that. Oral only, in a patient who may be too unwell to swallow. Traditional teaching
Sodium bicarbonate: For QRS widening. Citalopram's label lists QRS prolongation and bundle branch block among overdose features², which is the sodium-channel picture described on the sodium-channel blockade page. The bicarbonate treats the channel, not the SSRI.
Activated charcoal: The one intervention with a number attached: 50% reduction in citalopram absorption at half an hour, considered above 5 mg/kg within 1 hour.² Sertraline's and fluoxetine's labels both say charcoal may be as or more effective than lavage.¹⁴
Extracorporeal removal: No. Forced diuresis, dialysis, haemoperfusion and exchange transfusion are unlikely to be of benefit¹⁴; EXTRIP has published no recommendation covering the SSRIs.⁶
Stopping the second serotonergic drug: Not usually listed as a treatment, and arguably the most important one. The syndrome is combinatorial; removing one term reduces the sum.

Critical appraisal

The 'triad' badge is earned by a citation for the doubt, not by an absence. Dunkley and colleagues studied 2222 serotonergic overdose admissions and found that only clonus, agitation, diaphoresis, tremor and hyperreflexia were needed for accurate prediction, with hypertonicity and temperature above 38 °C added for the life-threatening cases — producing rules more sensitive (84% vs 75%) and more specific (97% vs 96%) than Sternbach's.⁵ A criteria set that outperforms the triad on its own data contests the triad's equal weighting of its three limbs.
Cyproheptadine's badge is the more contestable of the two, and is stated narrowly. The receptor rationale (5-HT2A antagonism) is sound and uncontested. What is downgraded is the confident teaching that cyproheptadine is the antidote for serotonin toxicity. The evidence offered for it in practice is observational, and the study most often cited alongside it — Dunkley's — used 5-HT2A antagonist treatment as an outcome measure defining the cohort, not as an intervention under test.⁵ A drug being given is not evidence that it worked, and this is precisely the shape of claim this library flags.
The Hunter criteria were derived from overdose patients, mostly of SSRIs.⁵ Their performance in the commoner real-world scenario — therapeutic doses of two interacting drugs — is an extrapolation from the population in which they were validated, and a reasonable one, but an extrapolation.
The MAOI ceiling argument is inference from pharmacology, not from a comparative study. It is the standard account, it is internally consistent, and it has not been demonstrated by measuring synaptic serotonin in poisoned humans.
The rostrocaudal gradient's anatomical explanation is a reconstruction. The clinical observation is robust; attributing it specifically to raphe–spinal projection length is not demonstrated and is marked as inference.
Escitalopram's label reports doses of 400 to 800 mg taken without severe symptoms.³ That is quoted as a risk-assessment observation from a regulator-approved document, in the same spirit as the salicylate concentration row on its own page — it is not a threshold, and it says nothing about a patient who has also taken something else.
No fatal dose appears on this page, and citalopram's label statement that the fatal dose is not known² is not reproduced as a figure. Comparative lethality between class members is permanently out of scope here, and the differences described are differences in mechanism — QT, QRS, seizure — which are actionable.
EXTRIP's silence on the SSRIs is uninformative rather than awkward.⁶ Unlike clonidine, where the absence leaves a genuinely open question, here two labels answer it from physical chemistry.¹⁴

References

1
Lustral 100 mg film coated tablets (sertraline) — Summary of Product Characteristics. electronic medicines compendium, product 2835. Sections 4.9 (Overdose) and 5.2 (Pharmacokinetic properties). medicines.org.uk/emc/product/2835
2
Cipramil 20 mg film-coated tablets (citalopram) — Summary of Product Characteristics. electronic medicines compendium, product 992. Sections 4.9 (Overdose) and 5.2 (Pharmacokinetic properties). medicines.org.uk/emc/product/992
3
Cipralex 20 mg film-coated tablets (escitalopram) — Summary of Product Characteristics. electronic medicines compendium, product 7716. Sections 4.9 (Overdose) and 5.2 (Pharmacokinetic properties). medicines.org.uk/emc/product/7716
4
Fluoxetine 20 mg Capsules — Summary of Product Characteristics. electronic medicines compendium, product 13515. Section 4.9 (Overdose). medicines.org.uk/emc/product/13515
5
Dunkley EJ, Isbister GK, Sibbritt D, et al. The Hunter Serotonin Toxicity Criteria: simple and accurate diagnostic decision rules for serotonin toxicity. QJM 2003 Sep;96(9):635–42. PMID 12925718. (2222 serotonergic overdose admissions 1987–2002; learning dataset of 473 SSRI-alone overdoses; clonus, agitation, diaphoresis, tremor and hyperreflexia sufficient for accurate prediction, with hypertonicity and maximum temperature above 38 °C added; sensitivity 84% vs 75% and specificity 97% vs 96% against Sternbach's criteria.)
6
EXTRIP Workgroup — published recommendations index. Extracorporeal Treatments in Poisoning Workgroup. Cited for the absence of any recommendation covering the SSRIs. extrip-workgroup.org/recommendations
7
Zydol XL 400 mg prolonged-release Tablets (tramadol) — Summary of Product Characteristics. electronic medicines compendium, product 88. Section 4.9 (Overdose). Cited here for the statement that naloxone will not antagonise tramadol's inhibitory effects on monoamine reuptake or its serotonin-releasing effects. medicines.org.uk/emc/product/88

SSRIs and serotonin toxicity

At a glance