Valproate in Genetic Generalized Epilepsy: Contraception Versus Conception
Two women with genetic generalized epilepsy, both stabilized on valproate, at opposite ends of family planning. The drug and the diagnosis are the same; the risk calculation is not.
K.S., a 24-year-old woman, was diagnosed with juvenile myoclonic epilepsy at fifteen after a series of morning myoclonic jerks progressed to a generalized tonic-clonic seizure at school. She works as a structural engineer, a job that puts her on active construction sites several days a week, and has been seizure-free for five years on valproate 500mg twice daily — the only regimen, after two prior drug trials, that fully controlled both her myoclonic jerks and her generalized seizures. She has a copper IUD, placed three years ago specifically for its non-hormonal reliability, and has told her neurology team consistently, at every annual visit, that she does not want children.
Valproate has come with a real, secondary cost for her: irregular menses and a pattern of weight gain and hirsutism her gynecologist has attributed to valproate's known association with polycystic-ovary-like hormonal changes, distinct from and in addition to the drug's better-known teratogenicity risk. She has raised this with her neurologist before and decided, each time, that stable seizure control on a job that requires full alertness around heavy equipment outweighs it.
Her total daily dose matters to how that risk should be read. EURAP's registry data show valproate's malformation risk rising with dose rather than sitting at a fixed rate, with the sharpest increases above roughly 1,500mg daily; at 1,000mg she sits below that inflection, which lowers the number without making it comparable to the alternatives. Her contraception sits on the other side of the same ledger: a copper IUD carries a first-year failure rate under one percent, among the most reliable methods available and, unlike anything she has to remember daily, not dependent on adherence — which is the specific reason it was chosen. The question in front of the team today isn't new — it resurfaces at every annual visit — but her IUD is now due for routine replacement, which is what actually brought the conversation back to the table this time rather than leaving it for next year. That replacement is also the one moment in a five-year cycle when the contraceptive protection the whole plan rests on is briefly, genuinely in question.
At the annual visit, with the IUD due for replacement
Continue valproate. She has been explicit and consistent for five years that she doesn't want children, she has a copper IUD — among the most reliable contraceptive methods that exist, failure rates well under one percent a year — and valproate is the only drug that has ever fully controlled both her myoclonic and generalized seizures. Switching a stable, well-contracepted patient off her most effective drug because of a pregnancy risk she has taken active, durable steps to prevent isn't proportionate.
I'm not arguing her IUD is unreliable — it isn't. What I'd push back on is treating “reliable contraception” and “zero risk” as the same thing. IUDs fail, circumstances change, and valproate's teratogenicity isn't a small effect at the dose she's on — EURAP's registry data put major malformation risk meaningfully above baseline, and NEAD's cohort found a real, dose-dependent cognitive effect in children exposed in utero even when the pregnancy itself was otherwise unremarkable.
Five years of consistent intent is real and worth weighing, but intent isn't the variable EURAP and NEAD measured — exposure is, and exposure only stops the day the prescription does.
There's a version of this that doesn't force a choice between her stated plans and the registry data. This is exactly what the formal valproate pregnancy-prevention framework exists for — not a one-time conversation but an annual, documented risk acknowledgment, paired with a explicit standing plan for what happens if her contraception ever fails or her circumstances change, checked at the same visit her regimen gets reviewed anyway.
That keeps her on the drug that's actually controlling her seizures, keeps the risk conversation active rather than settled once and forgotten, and gives her an explicit, agreed-upon trigger — not just her own future self remembering — for revisiting the drug choice if anything about the contraceptive picture changes.
Agreed: valproate continues unchanged, the IUD is replaced as scheduled, and a documented annual risk-acknowledgment review is added to her existing visit structure rather than created as a separate appointment she might skip.
Not agreed, and named explicitly rather than smoothed over: the Maternal-Fetal Medicine Specialist's underlying position — that any woman of childbearing potential on valproate carries real risk regardless of stated intent — wasn't overruled, only judged not to outweigh five years of full seizure control and a genuinely reliable contraceptive method in her specific case.
M.A., a 29-year-old woman, was also diagnosed with juvenile myoclonic epilepsy as a teenager and has been seizure-free for four years on valproate 500mg twice daily after an earlier trial of lamotrigine failed to control her myoclonic jerks, a period she remembers as frightening — jerks strong enough some mornings to knock a coffee cup out of her hand, on top of two breakthrough convulsive seizures that summer. She married last year, teaches middle-school science, and she and her husband have agreed to start actively trying to conceive within the next two months — a decision she raised herself at today's visit, having read enough about valproate and pregnancy online to arrive already worried, before her neurologist said a word about it.
Her worry runs in both directions and she said so plainly: she knows valproate carries a real risk of birth defects and of the kind of subtle cognitive effect the NEAD study described, but she has also read enough to know that convulsive seizures during pregnancy carry their own risk to a fetus — reduced oxygen during a generalized tonic-clonic seizure, physical trauma from a fall, and in the most severe case, status epilepticus. She does not want to trade one risk for what turns out to be a worse one, and two months is not, by her own read of what she found online, a long runway to sort that out.
Two features of her history do most of the work in reading her case, and they pull against each other. NEAD's cognitive finding was dose-dependent and measured at school age rather than at birth, which means the exposure that matters is not a single moment but the whole of a first trimester that will be underway before a pregnancy test turns positive — an argument for deciding now rather than later. But her failed lamotrigine trial is not a generic note in a chart either. Lamotrigine is the drug most women with juvenile myoclonic epilepsy are switched to, and in her it did not hold the myoclonus; it is also known to worsen myoclonic jerks in some patients with this syndrome. That she has already failed the obvious alternative narrows the field before the conversation starts, and it means levetiracetam is being asked to do a job that one reasonable substitute has already proven unable to do for her.
Two months before she starts trying
Start the switch to levetiracetam now, before conception, not after a positive test. Valproate's teratogenicity is dose-related and highest with first-trimester exposure — by definition the weeks before most women even know they're pregnant — and NEAD's neurodevelopmental finding held even in children whose mothers had otherwise unremarkable pregnancies. The safest exposure is no exposure, and the only way to guarantee that is to be off the drug before conception, not during it.
You're right that pre-conception is the correct window if a switch happens at all — I'm not arguing for waiting until she's pregnant. But her own lamotrigine trial already told us levetiracetam or any alternative isn't guaranteed to hold her myoclonic seizures the way valproate does, and she's naming the real risk on the other side of this herself: a convulsive seizure during pregnancy is not a hypothetical harm either.
Framing valproate's teratogenicity data as settling this on its own skips past the actual comparison she needs — not valproate's risk in isolation, but valproate's risk against the real chance of losing seizure control on a drug that's already failed her once, at exactly the moment losing control matters most.
Both of those are real, and the two-month runway is tight but not impossible if we start today rather than debating it further. Cross-taper to levetiracetam now, with EEG and clinical follow-up at two and four weeks to catch breakthrough myoclonus early, before she starts actively trying — not after.
High-dose folic acid starts today regardless of which drug she ends up on; the neural-tube-defect risk reduction evidence is real, if incomplete for valproate specifically. And if levetiracetam genuinely fails to hold her myoclonic seizures over these two months, the honest fallback isn't declaring the switch a failure and reverting by default — it's the lowest effective valproate dose, split into smaller, more frequent doses rather than fewer large ones, since divided dosing blunts the peak fetal exposure a twice-daily regimen produces, discussed with her explicitly as a real, second-choice option rather than defaulted into.
Agreed: cross-taper to levetiracetam starts today, with EEG and clinic follow-up at two and four weeks; high-dose folic acid starts immediately regardless of outcome. She was told directly that if myoclonic control is lost, the fallback is a genuine, discussed decision — lowest effective divided-dose valproate — not an automatic reversion.
Not agreed: whether two months is really enough runway to confirm the switch is holding before she and her husband begin actively trying, or whether the couple should be advised to extend the timeline by a cycle or two to give the taper more room. She was left to make that call herself once she'd heard both positions, rather than the team deciding it for her.