Epilepsy: diet works, antioxidants don't, and a disclosure we have to make

Every post in this series so far has ended with some version of "we sell nothing for this." Epilepsy is different. We sell two things that appear in the epilepsy literature, and neither of them appears well. One is ineffective. The other is an interaction risk.

So this page does two jobs: it explains the condition, and it tells you plainly what that evidence means for our own products. We are not doctors, and nothing here is treatment advice.

This is the second post in our series on the neurological conditions beyond dementia, after Parkinson's disease.

What epilepsy actually is

Epilepsy is a tendency to recurrent unprovoked seizures — bursts of abnormally synchronised electrical activity in the brain. It is one condition by name and many by cause: genetic, structural (after a stroke, injury, tumour or malformation), metabolic, immune, infectious, or unknown.

The most useful thing a non-specialist can know is that most seizures are not convulsions.

  • Focal aware seizures — the person stays conscious and experiences something: a smell, a rising sensation in the stomach, a feeling of intense familiarity, a twitch in one limb.
  • Focal impaired-awareness seizures — altered consciousness with automatic behaviour: lip-smacking, fumbling with clothes, wandering, unresponsiveness for a minute or two with no memory of it afterwards. These are routinely mistaken for daydreaming, confusion, intoxication or a psychiatric episode.
  • Absence seizures — brief blank staring, often in children, frequently recorded as inattention at school.
  • Bilateral tonic-clonic seizures — the convulsion everyone pictures, and the one that carries most of the risk discussed below.

Antiseizure medications control seizures in the majority of people who take them. Around 30% do not respond to initial pharmacological treatment and are classed as having drug-resistant epilepsy — a figure that recurs consistently across the literature, and which comes up again below.

The risk most people with epilepsy are never told about

SUDEP — sudden unexpected death in epilepsy — is the commonest directly epilepsy-related cause of death. A 2026 review sets out the scale:1

  • Across all people with epilepsy, SUDEP affects roughly 1 in 1,000 per year.
  • Depending on severity, the annual rate can exceed 10 per 1,000 patient-years.
  • Because people live with epilepsy for decades, the cumulative average lifetime risk is 5–20%.

The mechanism, as currently understood, runs like this: a bilateral tonic-clonic seizure, often nocturnal, is followed by pathologically impaired arousal, then apnoea, then bradycardia, progressing to fatal asystole. Established risk factors include frequent bilateral tonic-clonic seizures — especially nocturnal ones — living alone, long epilepsy duration, and drug-resistant epilepsy.

A 2026 update on the field adds newly identified risk factors from prospective studies: peri-ictal apnoea, disrupted sleep homeostasis, extratemporal epilepsies and elevated BMI.2 It also reports something that matters more than any of them: multiple surveys highlight substantial gaps in SUDEP communication, with many people with epilepsy and their carers remaining insufficiently informed. And alongside that, accumulating evidence that optimising seizure control is likely to reduce SUDEP risk.

In the severe childhood epilepsies the numbers are concrete. A Bayesian meta-analysis of 37 randomised trials and 15 extension studies covering 3,757 patients with developmental and epileptic encephalopathies estimated all-cause mortality at 8.76 per 1,000 person-years (95% credible interval 5.50–13.89) and SUDEP at 4.32 per 1,000 person-years (2.66–6.64). In Dravet syndrome specifically, SUDEP ran at 7.59 per 1,000 person-years.3

We are including this because the research itself says people are not being told, and because the practical implications are things a person can act on: taking medication consistently, not sleeping alone where that is avoidable, and treating nocturnal convulsions as a specific thing to raise with a neurologist rather than a general worry.

Where that leaves sleep — and us

Note what two of those risk factors are: nocturnal seizures, and disrupted sleep homeostasis. Sleep is not a peripheral wellness concern in epilepsy. It sits inside the mechanism of the condition's main cause of death.

That is precisely why it is not a sales opportunity, and the next two sections explain why.

Diet works here. That is unusual, and it is worth taking seriously

Ketogenic diet therapy is one of the few non-drug interventions anywhere in neurology with randomised evidence behind it. A 2026 systematic review of randomised controlled trials screened 1,193 articles and included 17 RCTs — 11 in children, 6 in adolescents and adults, with follow-up from 6 to 24 months:4

  • In children, 37% may achieve a 50% or greater reduction in seizure frequency on any form of ketogenic diet therapy — moderate-certainty evidence.
  • About 6 more children per 100 may achieve a 90% or greater reduction, though at low certainty.
  • In adolescents and adults, the diet may produce a ≥50% reduction in 16 more individuals per 100 compared with usual care — again moderate certainty. Its effect on ≥90% reduction is uncertain, with too few events and too much imprecision.
  • Adherence may be slightly lower on the diet than on usual treatment, in both children and adults.
  • Quality-of-life, cognitive and behavioural outcomes: scarce, heterogeneous, very low certainty. No conclusions available.

A 2025 umbrella review of 18 meta-analyses, appraised with AMSTAR 2, found the same direction with a different diet coming out on top.5 Dietary therapies — particularly the Low Glycemic Index Treatment — significantly increased the chance of a ≥50% reduction (relative risk 1.95, 95% CI 1.58–2.33) and a ≥90% reduction (RR 3.33, 1.51–5.14).

But not seizure freedom. The pooled estimate for becoming seizure-free on dietary therapy was RR 0.68, with a published interval of 0.00–1.36 — which is to say no effect, reported with an interval wide enough to be uninformative. Fewer seizures, not no seizures.

And the price of the diet that works

This is the part that gets left out of every enthusiastic article about keto and the brain. The same umbrella review found that dietary therapies and antioxidants together significantly increased the overall incidence of adverse events (RR 1.56, 1.19–1.92). Broken out for the ketogenic diet specifically:5

  • Kidney stones: RR 5.24 (3.73–6.74)
  • Weight loss: RR 4.80 (3.43–6.17)
  • Dyslipidaemia: RR 3.56 (2.00–5.11)
  • Constipation: RR 3.02 (1.55–4.48)

A separate 2026 review of micronutrient risk on ketogenic therapy adds that the diet's restrictive nature drives falling adherence over time, with gastrointestinal problems, potential for malnutrition and a high risk of micronutrient deficiencies — concluding that a registered dietitian monitoring growth and adjusting the plan is central to doing it safely.6

So the honest summary of the one dietary intervention with real evidence in a neurological disease: it works, modestly; it does not produce seizure freedom; it roughly quintuples kidney stone risk; and it requires a dietitian. That is medical nutrition therapy prescribed and monitored by clinicians. It is the opposite of something you start yourself because an article said keto is good for the brain.

The disclosure

Now the part that applies to our shelf.

Antioxidants do not work for epilepsy. The same umbrella review tested vitamin D, vitamin E and melatonin as adjunctive antioxidant therapy. In the subgroup analysis of ≥50% seizure reduction, antioxidants did not demonstrate significant efficacy, and the authors' stated conclusion is that current evidence does not support the use of antioxidants as effective adjunctive therapy for epilepsy.5

Melatonin specifically was ineffective at reducing seizure frequency (reported as RR 0.30, 95% CI 0.00–0.63 — an interval running to zero, which is itself a sign the pooled estimate is imprecise, and which we are printing as published rather than tidying).5

We sell melatonin. It is in Sleep Support at 10 mg and Sleep Formula at 2 mg. We have written about what the melatonin dose-response evidence actually shows for sleep onset, which is a real if modest literature. None of it extends to seizures, and the pooled epilepsy evidence says it does not reduce them. If you have epilepsy, melatonin is not a seizure intervention and should not be treated as one.

St John's Wort is an interaction risk, and it is in Sleep Support. A 2026 systematic review of cytochrome P450 herb-drug interactions, grading evidence from in-vitro work through human pharmacokinetic studies and documented case reports, assessed Hypericum perforatum among its herbal candidates and found clinically relevant induction or inhibition of CYP activity — particularly CYP3A4 — with the potential to alter drug exposure, reduce therapeutic efficacy or increase toxicity.7

Antiseizure medications are among the most interaction-dense drug classes in medicine; several are themselves enzyme inducers, and several have narrow margins where a drop in blood level means a breakthrough seizure. Reduced therapeutic efficacy of an antiseizure medication is not an inconvenience. Given what the section above says about seizure control and SUDEP risk, it is a safety matter.

We already carry an interaction warning on Sleep Support's product page and we have written the ingredient up in full in St John's Wort: real efficacy, real interaction risk and in our products and mental health medication interactions. Stating it here as plainly as we can:

If you have epilepsy, do not take Sleep Support without clearing it with the doctor who prescribes your antiseizure medication. That is not a formality and it is not a legal disclaimer. The full ingredient list is in inside Sleep Support so you can take it to the appointment.

The one supplement with a real rationale here, which we do not sell

Several of the older antiseizure medications induce cytochrome P450 enzymes, which accelerates the breakdown of vitamin D. An observational study of 226 patients on long-term carbamazepine measured the effect: serum vitamin D fell from a mean of 45 nmol/L before treatment to 36.6 nmol/L after (95% CI for the difference −14.47 to −2.37, P = 0.007), with calcium and sodium also significantly reduced.8

That is a genuine, mechanistically coherent reason for someone on an enzyme-inducing antiseizure medication to have their vitamin D checked and, if low, corrected. It is also a monitoring and prescribing decision for a clinician who knows which drug you are on, not something to self-administer off the back of a blog post.

We do not sell a vitamin D product. We are pointing at it anyway, because it is the honest answer to "is there any supplement that matters in epilepsy?" — and the answer is one we have no commercial interest in.

When the drugs do not work

For the roughly 30% with drug-resistant epilepsy, the options are surgery, devices, or dietary therapy. The device evidence is worth seeing precisely because of the gap between two numbers that are often reported as if they were one.

A systematic review of 88 studies and 3,068 children with drug-resistant epilepsy treated with invasive neuromodulation found vagus nerve stimulation produced a pooled responder rate — meaning a 50% or greater seizure reduction — of 54% (95% CI 51–58%). Pooled seizure freedom was 6% (4–8%).9

A separate meta-analysis of 19 studies and 149 children on deep brain stimulation, at a minimum 12 months' follow-up, found a responder rate of 63.3% (53.2–72.3%, I² = 3%) and seizure freedom of 13.8% (8.2–22.3%). Age at implantation did not predict response, and differences between stimulation targets were not significant.10

"More than half respond" and "one in sixteen becomes seizure-free" describe the same intervention. Both are true, and which one a family hears changes what they expect.

Where we come in

We sell nothing that treats or prevents epilepsy, and two things we do sell need the cautions above. That is the least flattering position we have taken on this blog, and printing it is cheaper than the alternative, which is a reader with epilepsy taking our sleep product without mentioning it to their neurologist.

The broader pattern is the same one we found in Parkinson's disease, where mitochondrial antioxidants engaged their targets without changing the disease. Epilepsy tests the antioxidant thesis directly, in a condition whose biology genuinely involves oxidative stress, and the pooled answer is no. Our standing argument about why mechanism is not evidence is in why we don't sell brain blends.

The short version

  • Epilepsy is recurrent unprovoked seizures from many different causes. Most seizures are not convulsions — focal and absence seizures are routinely mistaken for inattention, confusion or a psychiatric episode.
  • Around 30% of people do not respond to initial drug treatment.
  • SUDEP affects roughly 1 in 1,000 people with epilepsy per year, over 10 per 1,000 patient-years in severe epilepsy, with a cumulative lifetime risk of 5–20%. Research says most people are not told.
  • Risk factors include nocturnal bilateral tonic-clonic seizures, living alone, drug resistance, and disrupted sleep homeostasis. Better seizure control probably lowers the risk.
  • Ketogenic diet therapy works: ≥50% seizure reduction in 37% of children, and in 16 more adults per 100 than usual care, both at moderate certainty. Low Glycemic Index Treatment performed best in the umbrella review (RR 1.95 for ≥50%).
  • Dietary therapy does not deliver seizure freedom, and the ketogenic diet carries RR 5.24 for kidney stones, 4.80 for weight loss, 3.56 for dyslipidaemia, 3.02 for constipation, plus high micronutrient-deficiency risk. It needs a dietitian.
  • Antioxidants showed no significant efficacy. Melatonin was ineffective at reducing seizure frequency. We sell melatonin.
  • St John's Wort induces CYP3A4 and can reduce the efficacy of other drugs. It is in our Sleep Support. If you have epilepsy, clear that product with your prescriber first.
  • The supplement with a real rationale is vitamin D, because enzyme-inducing drugs deplete it — 226 patients on carbamazepine dropped from 45 to 36.6 nmol/L. We do not sell it.
  • Vagus nerve stimulation: 54% respond, 6% become seizure-free. Deep brain stimulation: 63.3% respond, 13.8% seizure-free.

This article is for informational purposes only and has not been evaluated by the FDA. It is not intended to diagnose, treat, cure, or prevent any disease. We are not doctors. Epilepsy requires specialist diagnosis and management. Nothing here is a reason to start, stop or change any prescribed medication — abruptly stopping an antiseizure medication can cause status epilepticus, which is a medical emergency. Do not start a ketogenic or other therapeutic diet for epilepsy without medical and dietetic supervision. If you have epilepsy and take or are considering any supplement, including ours, tell the clinician who prescribes your antiseizure medication.

References

  1. Mann C, Schubert-Bast S, Rosenow F, Strzelczyk A. "Current perspectives in sudden unexpected death in epilepsy (SUDEP): epidemiology, research approaches and pathways to prevention." Neurological Research and Practice, 2026, 8(1):18. doi:10.1186/s42466-026-00480-w
  2. Ryvlin P, Rheims S. "SUDEP update 2026: recent advances in experimental and clinical research." Current Opinion in Neurology, 2026, 39(2):123–130. doi:10.1097/wco.0000000000001463
  3. Moro P, Borioni MS, Mazzeo A, Cocchi E, Di Bonaventura C, Cerulli Irelli E. "SUDEP and mortality in developmental and epileptic encephalopathies: A meta-analysis of randomized clinical trials and extension studies." Epilepsia, 2026, 67(9):4631–4648. doi:10.1002/epi.70348
  4. Vaccarezza MM, Sanguine VL, Balaciano G. "Ketogenic diet therapies for the treatment of drug-resistant epilepsy in children and adults: A systematic review." PLOS One, 2026, 21(7):e0333334. doi:10.1371/journal.pone.0333334
  5. Feng Y, Cai S, Wang M, Guo Y, Zhou Z, Wang R, Yang S. "Efficacy and safety of antioxidants and dietary therapies for epilepsy: an umbrella meta-analysis." Frontiers in Nutrition, 2025, 12:1723370. doi:10.3389/fnut.2025.1723370
  6. Singh B, Botten P, Richardson KP, Weaver C, Purohit S. "The Ketogenic Diet and Potential Micronutrient Risks in Drug-Resistant Epilepsy Management: A Literature Review." Nutrients, 2026, 18(7):1081. doi:10.3390/nu18071081
  7. Hsieh YH, Huang ST. "Interactions between cytochrome P450 enzymes and traditional Chinese herbs: A comprehensive review." Journal of the Chinese Medical Association, 2026, 89(8):578–592. doi:10.1097/jcma.0000000000001404
  8. Hakami AY, Althagafi AA, Alkhayat AS, Alturkistani BA, Mawar HE, Ibrahim IH, Abualhamail H, Ahmed ME. "Long-term effects of carbamazepine on bone-related biochemical markers in patients with epilepsy." Acta Epileptologica, 2026, 8(1):22. doi:10.1186/s42494-026-00262-6
  9. Zaidan SK, Ghaith HS, Abdelkader K, Lopez Ramos CG, Shafie B, Zook ZA, Cahill J, Gendreau J, Nugent JG, Greisman JD, et al. "Invasive Neuromodulation of Drug-Resistant Epilepsy in Pediatric Populations: A Comprehensive Systematic Review and Meta-Analysis." Neuromodulation, 2026. doi:10.1016/j.neurom.2026.06.477
  10. Mhanna A, Ciliberto MA, Ganganna ST, Dlouhy BJ, Owens JW, Boes AD. "Deep brain stimulation outcomes in pediatric drug-resistant epilepsy: A systematic review and meta-analysis." Seizure, 2026, 142:281–289. doi:10.1016/j.seizure.2026.09.009