Valproate and perampanel emerge as the strongest signals in hyperammonemic encephalopathy
A pharmacovigilance analysis of the FDA Adverse Event Reporting System points to hyperammonemic encephalopathy as a clinically important adverse event across multiple antiepileptic drugs, with sodium valproate carrying the strongest association. Investigators reviewed FAERS reports from the first quarter of 2013 through the third quarter of 2024 and identified 1,456 HE-related events, 93.06% of them linked to antiepileptic therapy. The signal was not confined to a single legacy drug. Perampanel also showed a strong independent association, while several other agents appeared to become more hazardous when valproate entered the regimen.
The magnitude of the valproate signal was striking. VPA reached a reporting odds ratio of 122.14, with a 95% confidence interval of 110.16 to 135.41, dwarfing the other drugs in the analysis. Perampanel followed with an ROR of 52.62, and that signal appeared independent of valproate coadministration. Eight additional AEDs showed positive associations, although many of those signals were influenced by concurrent valproate exposure, especially topiramate and levetiracetam. In practical terms, the analysis suggests that clinicians should not view HE as a problem reserved for one older antiseizure medication alone; the adverse event can surface in more complex regimens where valproate acts as the biochemical amplifier.
Combination therapy sharpens the risk signal around ammonia toxicity
The clearest interaction involved valproate plus topiramate. This pairing was associated with a significantly higher risk of HE, with an odds ratio of 3.38 and a 95% confidence interval of 2.25 to 5.06. Antipsychotic co-therapy also carried a measurable signal: olanzapine plus valproate reached an OR of 1.65, and quetiapine plus valproate reached an OR of 1.95. Those combinations matter because they mirror real-world prescribing, where seizure control, mood stabilization, agitation, and behavioral symptoms often drive layered prescribing across neurology and psychiatry.
One sentence captures the bedside consequence: ammonia monitoring becomes more urgent when valproate is paired with other centrally acting drugs. The biological rationale is plausible even within a spontaneous-reporting framework. Valproate can impair ammonia handling and destabilize mitochondrial metabolism, creating a vulnerable metabolic state in which additional agents may tip the patient into encephalopathy. Topiramate has long been suspected in this setting, and the present signal is consistent with that concern rather than dismissing it as a rare anecdote.
Age and indication also influenced the signal. Minors had a lower risk than adults, with an odds ratio of 0.61 and a 95% confidence interval of 0.50 to 0.76. Psychiatric indications were also associated with a lower risk than other indications, with an OR of 0.74 and a 95% confidence interval of 0.62 to 0.89. Those patterns may reflect prescribing differences, underlying disease mix, dose selection, reporting behavior, or combinations used in epilepsy compared with psychiatric care. A spontaneous-reporting database cannot separate each of those possibilities cleanly, but the pattern still points to a high-risk clinical neighborhood around valproate-heavy regimens.
Bedside monitoring needs to follow the regimen, not just the diagnosis
The study’s central message is straightforward: HE risk tracked not only with the drug itself, but with how that drug was deployed. Valproate remained the dominant signal, yet the analysis also identified perampanel and several combination patterns that should make clinicians pause before assuming encephalopathy is unlikely simply because a modern AED is in use. Polypharmacy can obscure the cause of confusion, lethargy, vomiting, ataxia, or altered mental status until ammonia is finally measured. In that setting, delayed recognition can prolong hospitalization and complicate antiseizure management.
The authors also emphasized that VPA+TPM increased risk without worsening outcomes, which is a useful clinical nuance. It suggests the combination may increase the chance of the adverse event without necessarily guaranteeing a worse final course once recognized and managed, although spontaneous reports cannot prove that directly. Even so, the finding supports a lower threshold for checking ammonia in patients receiving valproate with topiramate, or with antipsychotics that add further metabolic and sedating burden.
The analysis relied on FAERS, so the signals reflect reporting patterns rather than incidence or causal proof. Spontaneous reports are vulnerable to underreporting, duplicate reports, confounding by indication, dose uncertainty, and incomplete clinical detail. Those limitations are especially relevant in a syndrome like HE, where mental status change may be multifactorial and often attributed first to infection, metabolic derangement, intoxication, or the neurologic disease itself. Still, when a signal repeats across a large reporting window and aligns with established pharmacology, it becomes hard to ignore.
For neurologists, psychiatrists, and hospitalists, the practical question is no longer whether valproate can drive HE. It can. The more urgent question is which combinations, in which patients, should trigger earlier ammonia checks, closer neurologic observation, and faster reconsideration of the drug regimen before encephalopathy deepens.