Established Status Epilepticus: What the ESETT Trial Settled

The short answer

The Established Status Epilepticus Treatment Trial (ESETT), published in the New England Journal of Medicine in 2019, compared three second-line seizure medications head to head and found no winner. Levetiracetam, fosphenytoin, and valproate each stopped seizures and restored responsiveness within an hour in roughly half of patients whose convulsions had not responded to a benzodiazepine, with similar rates of serious adverse events. That is a null comparative result, and it is one of the more useful things a trial can deliver. When three reasonable options perform the same, the choice can shift to what actually varies between hospitals and patients: availability, cost, drug interactions, and contraindications.

What problem ESETT was built to solve

Status epilepticus is a seizure that does not stop on its own. Convulsive status epilepticus, the form ESETT studied, is a medical emergency because prolonged seizure activity injures neurons and becomes harder to abort the longer it runs. The first-line treatment is well established: a benzodiazepine such as lorazepam or midazolam. But benzodiazepines fail to control a substantial share of patients, and at that point the seizure is called established or benzodiazepine-refractory status epilepticus.

For decades, what to give next was genuinely unsettled. Fosphenytoin (a phenytoin prodrug) was the traditional choice, but valproate and levetiracetam were both widely used, largely on the strength of habit, familiarity, and small studies rather than a direct comparison. Guidelines listed all three without ranking them, because the evidence to rank them did not exist. ESETT was designed to fill that gap.

How the trial was designed

ESETT, funded by the National Institutes of Health through its neurological emergency and pediatric research networks, enrolled patients across dozens of US emergency departments. Eligible patients had convulsive status epilepticus that persisted after an adequate dose of a benzodiazepine. They were randomly assigned, in blinded fashion, to receive levetiracetam, fosphenytoin, or valproate.

Two design features deserve attention. First, it was blinded: neither the treating clinicians nor the outcome assessors knew which drug a given patient received, which guards against the expectation effects that plague open-label comparisons. Second, it used a response-adaptive Bayesian design. Rather than fixing group sizes in advance, the trial updated the randomization as data accumulated, steering more patients toward whichever arm was performing better, and it allowed pre-planned interim looks at the data. If one drug had pulled ahead, this design was built to find it efficiently. If none had, the design was also built to recognize that and stop.

The primary outcome was concrete and clinically meaningful: absence of clinically apparent seizures with improving responsiveness at 60 minutes after the infusion started, without needing any additional anticonvulsant. That combination matters, because a drug that quiets the visible convulsion but leaves the patient obtunded, or that only works when rescued by a second agent, is not the same as a drug that resolves the emergency.

What the trial found

The three drugs were nearly indistinguishable. In the primary analysis, the seizure stopped and responsiveness improved in about 47 percent of patients given levetiracetam, 45 percent given fosphenytoin, and 46 percent given valproate. The trial's Bayesian machinery estimated the posterior probability that each drug was the single most effective option at roughly 0.41, 0.24, and 0.35, respectively. No arm cleared the threshold for superiority.

Because the accumulating data showed the drugs converging rather than separating, the trial met its pre-specified futility criterion and stopped early, after enrolling around 400 patients out of a planned maximum near 720. Stopping for futility is not a failure. It means the design worked exactly as intended: it recognized that continuing was unlikely to identify a best drug and stopped exposing patients to a comparison that had already answered its question.

Safety followed the same pattern. Rates of endotracheal intubation, hypotension, seizure recurrence over the following hours, and death were broadly similar across the three arms, with no drug standing out as clearly safer. A companion analysis published in The Lancet in 2020 broke the results down by age group, from children to older adults, and again found no meaningful difference in efficacy or safety among the three drugs within any age band. The equivalence held across the lifespan.

Why a null result is worth having

It is tempting to read "no difference" as "the trial found nothing." The opposite is closer to the truth. Before ESETT, the belief that one of these drugs was better rested on tradition rather than evidence, and that belief carried real consequences, since institutions stocked, protocolized, and defaulted to particular agents partly on faith. ESETT replaced that uncertainty with a well-supported conclusion: on average, these three drugs are interchangeable for this emergency.

That conclusion has practical force. If efficacy and safety are a wash, the decision can rest on the factors that genuinely differ. Valproate is generally avoided in pregnancy and in certain metabolic conditions. Some agents carry more drug-interaction burden. Availability at the bedside, speed of preparation, and cost vary between hospitals. A null comparative result does not tell a clinician nothing; it tells them the choice is theirs to make on other grounds, which is a form of freedom rather than a dead end.

A few limits are worth keeping in mind when reading any single trial this way. ESETT reports averages, and an average of no difference does not prove that no patient subgroup responds differently to one drug, only that none was detected. The trial studied specific drugs at specific doses for a specific emergency, so its findings do not automatically extend to other agents, other doses, or seizures that are not convulsive. And equivalence at 60 minutes is not a claim about every downstream outcome. Even so, for the question it was built to answer, ESETT answered clearly.

This article is educational and is not medical advice; decisions about seizure treatment belong with the clinicians caring for a given patient.