Cannabidiol Off-Label for Refractory Focal Epilepsy, on Top of Phenytoin
A single patient with refractory focal epilepsy outside cannabidiol's labeled syndromes. The advocacy pressure to try it is real — so is a documented interaction with the drug he's already taking.
T.K., a 16-year-old boy, has had drug-resistant focal epilepsy since age eight, unrelated to any of the three syndromes cannabidiol is actually approved for, and has failed four prior antiepileptic drugs before landing on his current regimen of phenytoin, which controls roughly half his baseline seizure frequency. He plays adaptive wheelchair basketball on a competitive regional team, an activity his parents credit with keeping him engaged through years of imperfect seizure control, and his mother has spent the past several months in an online support community where cannabidiol comes up constantly — not for his syndrome specifically, but as a broadly discussed option for refractory epilepsy generally.
His parents requested cannabidiol directly at today's visit, citing its FDA approval as evidence of a fundamentally safe, well-studied drug, and asking why it shouldn't be tried given how much seizure burden remains despite phenytoin. Cannabidiol's approval really is real and its safety data in its approved populations are genuinely substantial — the GWPCARE program, Devinsky et al. (2017) in Dravet syndrome and Thiele et al. (2018) in Lennox-Gastaut, is what earned it. But the indication runs to three named syndromes, none of them his, and the specific obstacle relevant to T.K. is not safety in the abstract; it's phenytoin. Phenytoin is a strong inducer of CYP3A4, one of the enzymes partly responsible for cannabidiol's own metabolism, and adding cannabidiol on top of an established phenytoin regimen risks a real, documented reduction in cannabidiol's own effective exposure — the opposite problem from what most cannabidiol drug-interaction counseling actually focuses on, which is cannabidiol raising other drugs' levels, not having its own levels lowered by one already on board.
His most recent phenytoin level was therapeutic, and his seizure control, while incomplete, has been stable on the current dose for over a year — the phenytoin isn't the problem being solved; it's the fixed variable everything else has to work around.
With the family's request on the table
I understand why cannabidiol keeps coming up — it's a genuinely FDA-approved drug with real trial evidence, and his seizure control is incomplete. But that evidence is narrow. The GWPCARE program — Devinsky et al. (2017) in Dravet syndrome, Devinsky et al. (2018) and Thiele et al. (2018) in Lennox-Gastaut — is what earned Epidiolex its approval, and its indication runs to exactly three syndromes: Lennox-Gastaut, Dravet, and tuberous sclerosis complex. None of them is his. And there's a specific pharmacologic obstacle here that has nothing to do with whether cannabidiol is generally safe: the label itself warns that strong CYP3A4 and CYP2C19 inducers may reduce cannabidiol exposure, and phenytoin is a strong CYP3A4 inducer. His levels could sit below the trial range before we'd ever get a fair read on whether it helps.
That's a real interaction and I don't think anyone here disputes it. But 'his exposure would be lower than in the GWPCARE trials' isn't the same as 'it definitely won't work' — off-label cannabidiol use for refractory epilepsy outside the three approved syndromes has real, if less rigorously studied, clinical reports of benefit, and his family has been asking in good faith for months, not showing up today on impulse.
Treating the interaction as a reason to close the conversation skips past the more honest framing — it's a reason to dose and monitor around it deliberately, not a reason to refuse the drug outright to a family who's done real homework.
There's a way to actually test the drug rather than either refusing it or starting it blind to the interaction. Start cannabidiol at a higher end of the standard titration range specifically because we know phenytoin will pull his own exposure down, and check a cannabidiol level partway through titration — not routinely done for most patients, but genuinely informative here given the known interaction, so we know whether he's actually reaching a meaningful dose rather than guessing from seizure counts alone.
That gives the family a real trial rather than a symbolic one, and gives us an honest way to tell 'the drug didn't help him' apart from 'the drug never really got a fair chance at his phenytoin dose' — two very different findings that would otherwise look identical from the outside.
Agreed: cannabidiol started at the higher end of the standard titration range, with a mid-titration cannabidiol level check specifically to confirm meaningful exposure is being reached despite the phenytoin interaction, and a structured seizure diary to assess real effect against his baseline.
Not agreed: how long an inadequate trial should be allowed to run before declaring it a failure and stopping — the Epileptologist wanted a defined, relatively short window given cannabidiol's real cost and administrative burden; the Primary Care Physician wanted more room given the family's investment and off-label evidence's more gradual, less clean-cut nature. Left open, to be revisited once the level check comes back.