Why this poison is interesting

Valproate is the hinge between two families this library keeps separate. Like salicylate and phenytoin it is a saturation poison — protein binding saturates in overdose, the free fraction rises, and the drug becomes dialysable precisely when it becomes dangerous. But unlike either of them, valproate has a toxic metabolite and depletes a cofactor, which is why it produces a hyperammonaemic encephalopathy and hepatotoxicity that pure sedatives do not — and why, alone in the saturation cluster, it has a genuine mechanism-directed antidote.³

Most of an overdose is sedating and benign. At up to five or six times the therapeutic maximum the label expects little more than nausea, vomiting and dizziness.² The sedation is the parent drug, and it is usually self-limiting.
The dangerous part is metabolic and delayed. Valproate depletes L-carnitine — the shuttle mitochondria need for fatty-acid oxidation — and inhibits the urea cycle, so ammonia rises, sometimes to encephalopathic levels, even when transaminases are normal.²³ Cerebral oedema and intracranial hypertension are recognised.²
Protein binding saturates. The free fraction is normally 6–15% but rises as the concentration climbs,² which is the salicylate mechanism again — the drug that overwhelms its binding is the drug the dialyser can reach.

A poison is a drug whose kinetics have escaped its pharmacology — and valproate's escape happens twice, once in the plasma and once in the mitochondrion.

The toxic principle

There are three toxic principles, not one, and they act on different timescales.

The parent drug depresses the central nervous system. Valproate enhances GABAergic transmission; in overdose that becomes sedation, coma, muscular hypotonia, hyporeflexia, miosis and respiratory depression.² This is the early, dominant and usually self-limiting picture. Established
A metabolite is hepatotoxic. Valproate is metabolised partly by cytochrome P450 to unsaturated species, of which 4-ene-valproate is implicated in the mitochondrial injury behind valproate hepatotoxicity — the same shift toward a minor, toxic metabolic route that defines paracetamol and ethylene glycol.³ Inferred
Carnitine depletion drives hyperammonaemia. Valproate sequesters and depletes L-carnitine, the cofactor that shuttles fatty acids into mitochondria; impaired β-oxidation and inhibition of the urea cycle raise ammonia, producing an encephalopathy that can occur at or below therapeutic concentrations and does not require liver failure.²³ Established

Toxicokinetics

Valproate — saturable binding, and a metabolism that turns toxic at the edges
Parameter	Therapeutic	In overdose	Why it changes
Absorption	Well absorbed orally; enteric-coated and prolonged-release preparations (Epilim gastro-resistant, Epilim Chrono) delay the peak²	Delayed and prolonged after a large ingestion, especially of modified-release forms — gastric lavage is described as useful for up to 10–12 hours²	As with theophylline and salicylate, the modified-release formulation can keep the concentration rising for many hours, so an early level is a floor. The label's willingness to lavage at 10–12 hours is itself a statement about how slow the absorption can be.
Protein binding	Normally high — free fraction only 6–15% of total²	Saturates; the free fraction rises as the concentration climbs²	This is the salicylate mechanism, and it has the same two consequences: the total concentration understates severity as the dose rises, and the drug becomes dialysable at exactly the concentrations where it is dangerous — which is why EXTRIP grades it dialysable at all despite its being highly bound at therapeutic levels.
Volume of distribution	Small — largely confined to plasma and extracellular fluid	Rises as unbound drug redistributes when binding saturates	A small volume of distribution, enlarging as the free fraction grows, is the second reason high-concentration valproate is a reasonable dialysis target — most of the free drug is in a compartment the machine can reach.
Metabolism	Chiefly hepatic — glucuronidation (~40%, via UGT1A6/1A9/2B7) and mitochondrial β-oxidation, with a minor P450 route²	The minor P450 route matters more in overdose, generating unsaturated hepatotoxic metabolites such as 4-ene-valproate³	This is the row that separates valproate from its cluster-mates. Salicylate and phenytoin have inert metabolites; valproate's minor oxidative pathway produces a toxic species, so a metabolic-branch problem sits underneath the saturation-kinetics one.
Half-life	Usually 8–20 h²	Prolonged in massive overdose as clearance pathways are overwhelmed	Shorter than phenytoin's, but long enough — and lengthened enough by a large load — that the encephalopathy and acidosis can persist for a day or more and can appear after the initial sedation has been noted and filed as stable.
Dialysability	—	Moderately dialysable, EXTRIP level of evidence B¹; the free, dialysable fraction rises with the concentration	The paradox is the same as salicylate's: a drug that is highly protein-bound and therefore 'undialysable' at therapeutic levels becomes efficiently removable once binding saturates in overdose. EXTRIP graded valproate's dialysability B — better evidence than phenytoin's C — and recommends dialysis for the sickest patients.

Metabolism and the metabolites

This is where valproate parts company with the rest of its cluster. Salicylate, theophylline and phenytoin have metabolic maps with no toxic branch. Valproate has two toxic outcomes wired into its metabolism — a hepatotoxic metabolite, and a cofactor it consumes.

Valproate — a metabolism with a toxic branch and a stolen cofactor

Valproate (ingested)Sedating in its own right — enhances GABAergic transmission
Glucuronidation (~40%, UGT)Valproate glucuronideThe major route; non-toxic, renally excreted²
Mitochondrial β-oxidation — requires carnitine3-oxo- and other β-oxidation productsThe route that depletes carnitine. As carnitine is consumed, β-oxidation fails and the urea cycle is inhibited — ammonia rises³
Minor P450 oxidation (major in overdose)4-ene-valproate and other unsaturated metabolitesHepatotoxic. Implicated in the mitochondrial liver injury of valproate — a minor route that becomes significant when the majors saturate³

What changes the answer

Pre-existing carnitine deficiency worsens everything — young children, the malnourished, those on long-term valproate or with mitochondrial disease start with less of the cofactor to lose, and are more prone to hyperammonaemic encephalopathy. Inferred
A normal liver does not exclude the metabolic lesion. The hyperammonaemia is a urea-cycle and carnitine problem, so it can precede or occur without hepatotoxicity.³ Established
Modified-release preparations delay the whole sequence, so both the sedation and the metabolic tail can arrive later than the ingestion time suggests.²

Elimination and accumulation

Valproate is cleared by the liver, and in overdose that clearance is doubly stressed — the major glucuronidation route is loaded and the minor oxidative route is producing more toxic metabolite. There is no useful renal exit to enlarge, so removal in severe cases means the dialyser, and correction of the metabolic lesion means carnitine.

The EXTRIP position

79 articles, 82 overdose patients, toxicokinetic grading in 55; very low quality of evidence, but dialysability graded level B — better-evidenced than phenytoin.¹

Recommended in severe VPA poisoning (1D), and specifically for [VPA] >1300 mg/L (9000 µmol/L), cerebral oedema, or shock (all 1D).¹
Suggested for [VPA] >900 mg/L (6250 µmol/L), coma or respiratory depression requiring ventilation, acute hyperammonaemia, or pH ≤7.10 (all 2D).¹
Stop when clinical improvement is apparent (1D) or [VPA] is 50–100 mg/L (350–700 µmol/L) (2D).¹
Intermittent haemodialysis preferred (1D); haemoperfusion (1D) or CRRT (2D) are alternatives.¹

Target organs — and why those

Brain — sedation

TargetGABAergic transmission, enhanced by the parent drug

Why hereValproate's therapeutic action is potentiation of GABA; in overdose this becomes CNS depression, hypotonia, hyporeflexia, miosis and respiratory depression.² This is the early and usually dominant picture, and it is the parent drug's doing. Established

At the bedsideComa with small pupils and preserved-then-failing respiration. Naloxone has reversed valproate-induced coma in scattered case reports,² a curiosity rather than a mainstay. This part of the poisoning is generally self-limiting with airway support.

Brain — hyperammonaemic encephalopathy and cerebral oedema

TargetAstrocytic ammonia handling and cerebral water balance

Why hereRising ammonia — from carnitine depletion and urea-cycle inhibition — is neurotoxic, and severe valproate poisoning is associated with cerebral oedema and intracranial hypertension.²³ This is a metabolic brain injury layered on top of the sedative one, and it can lag it. Established

At the bedsideWorsening consciousness, or a failure to improve as the sedation should be wearing off, with a rising ammonia — the signal to think about carnitine and, if severe, dialysis. Cerebral oedema is a strong (1D) EXTRIP dialysis indication.¹

Liver

TargetHepatic mitochondria, injured by 4-ene-valproate and impaired fatty-acid oxidation

Why hereThe unsaturated oxidative metabolites and carnitine depletion together impair mitochondrial β-oxidation, producing hepatotoxicity that ranges from transient transaminase rise to, rarely, fulminant failure.³ Acute massive overdose can add to a chronic-use risk. Inferred

At the bedsideCheck liver function and lactate, but do not use a normal liver to exclude the hyperammonaemic encephalopathy — the two do not always travel together.³ Carnitine is advocated for the hepatotoxicity as well as the ammonia.

Timeline of effects

Valproate's timeline has two layers running at different speeds: a sedative layer that arrives early and recovers, and a metabolic layer that can arrive late and worsen. The danger is in the gap between them.

Valproate — sedation first, the metabolic tail behind it

Time

What you seeWhat is happening

0–4 hEarly
What you seeNausea, vomiting, dizziness and increasing drowsiness. After a moderate overdose this may be the whole illness. Miosis and hypotonia in larger ingestions.
What is happeningThe parent drug's GABAergic sedation, rising with absorption. With a modified-release preparation the concentration is still climbing, and gastric lavage may still be worthwhile up to 10–12 hours.²
4–24 hThe metabolic tail builds
What you seeIn a significant overdose, the patient who should be waking may instead become more confused. A metabolic acidosis, a rising ammonia and — in massive ingestion — hypotension develop. The valproate level may already be plateauing or falling.
What is happeningCarnitine is being depleted and the urea cycle inhibited, so ammonia climbs; the minor oxidative pathway is generating hepatotoxic metabolites.³ This is the genuine latent phase — a metabolic injury accumulating behind a sedative picture that looked as though it had peaked. Encephalopathy here is a reason to check the ammonia and consider carnitine and dialysis, not to deepen sedation.
24 h onwardSevere / decompensation
What you seeHyperammonaemic encephalopathy, cerebral oedema, coma, shock and, rarely, hepatic injury. This is where the deaths occur, and it is where EXTRIP's strong dialysis indications sit — >1300 mg/L, cerebral oedema, shock.¹
What is happeningMaximal metabolic derangement: ammonia unbuffered, mitochondrial fatty-acid oxidation failing, protein binding saturated so the free, active and dialysable fraction is high. Dialysis removes both valproate and ammonia; carnitine addresses the biochemistry. Established

What the mechanism predicts at the bedside

Why the ammonia is the number to follow

Because the dangerous part of valproate poisoning is metabolic, not sedative, and the ammonia tracks it while the valproate level tracks the sedation.²³ A rising ammonia — especially with worsening consciousness and a normal-ish liver — is the signal that the latent phase has arrived, and it is an EXTRIP dialysis indication in its own right.¹ Established

Why L-carnitine is a real antidote, not a tonic

Valproate depletes carnitine; carnitine is the cofactor mitochondrial β-oxidation cannot run without.³ Replacing it restores the pathway and helps the urea cycle clear ammonia — a defined lesion with a defined answer. The 2025 UK review found evidence supporting carnitine for valproate-induced hyperammonaemia and hepatotoxicity in both chronic use and acute overdose, while noting that the optimal dose and route remain uncertain.³ The indication is sound; the regimen is a TOXBASE and NPIS matter. Established

Why the total valproate level can mislead

Protein binding saturates in overdose, so the free, active fraction rises out of proportion to the total,² exactly as with salicylate. A total level that looks only moderately raised can accompany a high free concentration and a sick patient — and it is the free drug, not the total, that the dialyser removes efficiently. Inferred

Why dialysis treats two problems at once

In the severely poisoned patient, haemodialysis removes valproate — increasingly efficiently as binding saturates — and clears ammonia directly.¹ So the one intervention addresses both the poison and its most dangerous consequence, which is why EXTRIP's strong indications are the metabolic ones (cerebral oedema, shock, a very high level) rather than the sedative ones.

Why a normal liver does not reassure

The hyperammonaemic encephalopathy is a urea-cycle and carnitine problem and can occur with normal transaminases and synthetic function.³ Using a clean liver panel to exclude serious valproate toxicity is a predictable error the mechanism warns against.

The antidote, from the poison's side

Valproate is the one poison in the saturation cluster with an antidote worth the name, and it is a cofactor-replacement one — the same shape of treatment as acetylcysteine for paracetamol, aimed at a specific depleted molecule rather than at the drug.

L-carnitine refills the cofactor valproate depletes, restoring mitochondrial fatty-acid oxidation and helping the urea cycle clear ammonia — a mechanism-directed antidote for the hyperammonaemia and hepatotoxicity, not for the sedation.³ Established
It does nothing for the drowsiness, which is the parent drug's GABAergic effect and resolves with time and airway support. Carnitine is for the metabolic tail, and giving it makes sense only once you are thinking about ammonia and liver, not coma alone.
Dialysis is the removal step for the sickest, and it treats valproate and ammonia together;¹ carnitine and dialysis are complementary, not alternatives — the biochemistry and the body burden addressed at once.
Naloxone and other reported reversals are curiosities, not mainstays.² The two interventions that follow from the mechanism are carnitine for the ammonia and dialysis for severe or high-concentration poisoning.

Critical appraisal

The carnitine-depletion account of the hyperammonaemia is well established and clinically load-bearing.³ It explains why the ammonia rises without liver failure, why carnitine is a rational antidote, and why the vulnerable groups are the carnitine-poor. The 2025 UK review supports carnitine for the hyperammonaemia and hepatotoxicity while being candid that the optimal dose and route are unknown³ — an honest 'the indication is sound, the regimen is not settled'.
The 4-ene-valproate hepatotoxicity mechanism is inferred rather than proven in the acute human overdose. Inferred It is the standard explanation for valproate liver injury and is coherent with the mitochondrial picture, but the quantitative contribution of the toxic metabolite versus simple carnitine depletion to any individual case is not separable at the bedside.
The saturable protein binding is on firm ground Established and is quoted from the label's own 6–15% free fraction.² It is the same mechanism as salicylate's and it has the same two consequences — a misleading total level and a drug that becomes dialysable when it becomes dangerous.
EXTRIP's valproate recommendations rest on 82 patients and very low quality evidence,¹ though the dialysability grade (B) is better than phenytoin's. The concentration thresholds (1300, 900 mg/L) are consensus cut-points from case-level data, not validated population thresholds, and should be read alongside the clinical indications rather than instead of them.
The latent phase is real but variable. The delayed hyperammonaemic encephalopathy is well described,³ but its timing depends on the preparation, the dose, the baseline carnitine status and co-ingestants, so 'the encephalopathy can lag' is a warning to keep watching, not a fixed interval to wait out.
Naloxone reversal of valproate coma is a genuine but minor and inconsistent finding.² It is recorded here for completeness and explicitly not promoted to a treatment — the mechanism does not obviously predict it, and it should not distract from carnitine and dialysis.

References

1
Ghannoum M, Laliberté M, Nolin TD, MacTier R, Lavergne V, Hoffman RS, Gosselin S; EXTRIP Workgroup. Extracorporeal treatment for valproic acid poisoning: systematic review and recommendations from the EXTRIP workgroup. Clin Toxicol (Phila) 2015;53(5):454–65. PubMed 25950372 · Recommendation set also at extrip-workgroup.org/valproic-acid. Source of the moderately-dialysable (level B) grading, the >1300 mg/L / cerebral oedema / shock strong (1D) indications, the >900 mg/L / coma / acute hyperammonaemia / pH ≤7.10 suggested (2D) indications, the 50–100 mg/L cessation range, and the 79-article / 82-patient evidence base. Verified 12 Sep 2026 from the abstract and the workgroup's published recommendation page.
2
Epilim 200 mg Gastro-resistant tablets — Summary of Product Characteristics, Sanofi. emc product 519. §4.9 and §5.2 fetched and read in full. Source of the up-to-5–6× (minor symptoms) and 10–20× (massive: CNS depression, coma, hypotonia, hyporeflexia, miosis, metabolic acidosis, hypotension, shock) overdose descriptions, the intracranial-hypertension/cerebral-oedema note, the hypernatraemia-from-sodium-content warning, the L-carnitine-for-hyperammonaemia statement, the reported naloxone use, the 6–15% free fraction, the ~40% glucuronidation metabolism, the 8–20 h half-life, and the up-to-10–12-hour lavage window. Verified 12 Sep 2026.
3
Gziut T, Thanacoody R. L-carnitine for valproic acid-induced toxicity. Br J Clin Pharmacol 2025;91(3):636–47. PubMed 39261302 UK review (NPIS Newcastle). Source of the account of valproate metabolism producing both therapeutic and toxic metabolites, carnitine as the essential cofactor for mitochondrial fatty-acid oxidation, the evidence supporting carnitine for valproate-induced hyperammonaemia and hepatotoxicity after both chronic use and acute overdose, and the statement that the optimal dose and route remain unknown. Verified 12 Sep 2026 from the abstract.
4
TOXBASE — sodium valproate; valproic acid. National Poisons Information Service. toxbase.org (NHS login required. NPIS: 0344 892 0111. The authoritative UK source for carnitine indications and dosing, ammonia thresholds and dialysis criteria. Login-gated, so not quoted here.)

Valproate

At a glance