Prophylactic Levetiracetam Before Craniotomy: One Recommendation, Two Levels of Evidence
The recommendation against routine seizure prophylaxis in brain tumor patients is unusually strong. The recommendation about the days immediately around surgery itself is a separate, much weaker one — and this case turns entirely on which one actually applies.
W.H., a 61-year-old retired postal carrier, has spent most weekends this year in his garage bringing a rust-eaten 1967 pickup back to something roadworthy — a project his son gave up trying to talk him out of two winters ago. The headaches started small enough that he blamed the truck’s exhaust fumes, then grew steadily over three weeks into something that woke him at night and came with a word he’d search for and lose mid-sentence, which is what finally got him into his primary care doctor’s office rather than back under the hood. An MRI found a large, irregularly enhancing left temporal mass with surrounding edema, radiographically concerning for a high-grade glioma; he has never had a seizure, witnessed or otherwise, and his neurologic exam today is notable only for mild word-finding hesitation consistent with the lesion’s location. Resection is scheduled in two days. He is otherwise healthy apart from hypertension controlled on a single agent and a thirty-pack-year smoking history he quit fifteen years ago.
The question in front of the team — whether to start levetiracetam before he goes to the operating room — sounds like it should already be settled by the guideline that governs it, and in one sense it is: the 2021 SNO/EANO update to the AAN practice parameter (Walbert et al.) gives a Level A recommendation, its strongest grade, against prescribing antiepileptic drugs to reduce seizure risk in newly diagnosed brain tumor patients who haven’t had a seizure. But that recommendation was built from trials of ongoing, weeks-to-months prophylaxis in seizure-naive patients generally — not from the narrower, few-days window immediately around a craniotomy itself, when cortical manipulation carries its own separate and real seizure risk. For that specific window, the same guideline update grades the evidence Level C — insufficient to recommend for or against — a genuinely weaker and different statement than the Level A verdict everyone quotes first. Whether he needs an AED at all for the months after he’s recovered from surgery isn’t actually contested here — nobody in the room believes ongoing prophylaxis is indicated once he’s past the operative window. The disagreement in front of them is narrower than that, and turns entirely on which of the guideline’s two grades governs the next 48 hours.
The pre-op huddle, two days out
The Level A recommendation exists because the trials asking this question keep returning the same answer. Forsyth randomized 100 seizure-naive brain tumor patients to anticonvulsants or nothing and stopped accrual early — at three months, 87 percent of the treated arm and 90 percent of the untreated arm were seizure-free, and the investigators worked out they would have needed nine hundred patients to detect anything worth waiting for. The Cochrane review of post-craniotomy prophylaxis — ten randomized trials, eighteen hundred seizure-naive patients — found no consistent evidence that prophylaxis lowers seizure risk at all. Levetiracetam isn’t free of cost either — mood and behavioral side effects are real and dose-related. Starting it here treats a Level A ‘don’t’ as if it were a suggestion.
I’d read the guideline you’re citing more carefully before calling this settled. The same 2021 update — Walbert and colleagues — gives the perioperative window its own, separate grade — Level C, insufficient evidence to recommend for or against — specifically because none of the trials behind the Level A verdict were designed around the days immediately surrounding a craniotomy. Cortical manipulation during resection is a mechanistically distinct seizure trigger from the general tumor-related risk those trials measured. A short, defined course carries a low adverse-event burden — mostly transient somnolence and irritability, resolving once the drug stops. That’s a narrowly-scoped hedge for a genuinely under-studied window, not a contradiction of Level A.
The ‘don’t start unindicated AEDs’ argument you’re making is exactly right for the population those trials actually enrolled — patients well past their operative window, being followed for months. It doesn’t automatically transfer to a decision the trials never tested.
You’re both making a reasonable case about day one. I want to talk about day ninety. Chart-review data on exactly this population show that once a perioperative course starts, it tends not to stop. Lapointe’s seven-year series followed 205 seizure-naive glioma patients started on prophylaxis at surgery: 47 percent were still taking it at three months, median duration fifty-eight days. That’s not a hypothetical drift risk — it’s close to half of them, and it’s the documented norm, not the exception. If we start levetiracetam today, I want a specific discontinuation date — post-op day seven, say — written into the discharge plan itself, not left to whoever happens to see him at his next follow-up. Otherwise we’re not making the narrow, evidence-consistent decision you’re describing; we’re making the indefinite one the Level A recommendation was written to prevent.
Agreed: a seven-day perioperative course of levetiracetam, with the stop date written explicitly into the discharge summary rather than left implicit — a compromise the pharmacologist accepted once the Level A/Level C distinction was on the table, and the neurosurgeon accepted once a hard stop date was attached to what she’d proposed as an open-ended “short course.”
Not agreed: whether this should become the team’s default approach for every seizure-naive patient headed to craniotomy, or reserved for cases with a specific reason to expect higher intraoperative or perioperative seizure risk — extensive resection, eloquent or perilesional cortex, a lesion this close to the temporal speech areas. The neurosurgeon wants it as a standing default, arguing the Level C evidence doesn’t support selectively applying it only when someone’s worried; the pharmacologist wants case-by-case justification documented each time, worried that a quiet default is exactly how a seven-day plan becomes a three-month one again.