Epilepsy: Independent Evidence on Treatment, Pregnancy & Funding

Table of contents

• Evidence summary
• What epilepsy is
• Symptoms, seizure types and red flags
• Causes and risk factors
• Diagnosis: EEG and MRI limits, and a first seizure
• Treatments: what works and what does not
• Antiseizure medicines
• Pregnancy planning and the valproate programme
• SUDEP, safety and driving
• Surgery and neurostimulation
• Ketogenic diet and cannabidiol
• Supplements, lifestyle and complementary claims graded
• Risks and side effects
• Who to see
• Follow the money: makers, trials and patient groups
• Related research
• Frequently asked questions
• Sources and funding notes

Evidence summary

Independent evidence and credibility scorecard

Independence tiers: 1 = regulator or government body; 2 = academic or Cochrane team with no declared commercial funding for the work; 3 = independent authors analysing mostly manufacturer-run trials, or minor declared industry ties; 4 = manufacturer-funded or manufacturer-authored. Credibility runs from A (highest) to D. For this condition we treat professional societies and patient charities that take drug-company money as tier 3, and trial authors employed by a device or drug maker as tier 4. Our sources cluster in the UK and US, with trials from Canada and multinational registries, so the verdicts reflect Western health-system lenses; the WHO figures are the main exception.

What epilepsy is

Epilepsy is a long-term brain condition that causes seizures. It cannot currently be cured, but treatment often keeps it under control (NHS). The WHO defines it as two or more unprovoked seizures, and says a single seizure does not mean epilepsy (WHO). The NHS adds that diagnosis usually follows more than one seizure, or test results that show a high chance of another (NHS).

About 50 million people worldwide have epilepsy, and nearly 80% live in low- and middle-income countries (WHO). Risk of early death is up to three times that of the general population, and many of those deaths, such as falls, drowning, burns and prolonged seizures, are potentially preventable (WHO). Stigma and myths still discourage treatment (WHO).

Seizures that look like epilepsy but are not include functional (dissociative) seizures, which the DVLA handles separately (DVLA). Infant sleep jerks can also mimic seizures; see our guides on functional neurological disorder and benign sleep myoclonus of infancy.

Symptoms, seizure types and red flags

Seizures can cause stiffening or limpness, falls, jerking movements, loss of bladder control, a blank stare, or unusual sensations or behaviour, and usually last seconds to a few minutes (NHS). The NHS lists the main types as tonic-clonic, tonic, atonic, absence, focal and myoclonic (NHS).

Doctors use the International League Against Epilepsy (ILAE) 2017 system: focal onset (starting in one part of the brain, with awareness kept or impaired), generalised onset (both sides from the start, motor or absence), and unknown onset. Older terms such as "partial" and "complex partial" were dropped (Fisher 2017, ILAE). The type guides medicine choice, which is why a clear description from a witness matters.

Red flags that need an ambulance: a first-ever seizure, one lasting longer than usual or over 5 minutes, slow recovery, repeated seizures without recovery, or three or more in 24 hours (NHS). Contact NHS 111 or an urgent GP after a first seizure where hospital was not attended, or after a seizure following a long seizure-free spell (NHS).

Causes and risk factors

The cause is unknown in about half of cases. Known causes fall into structural, genetic, infectious, metabolic, immune and unknown groups, including birth injury, severe head injury, stroke, brain infections such as meningitis and brain tumours (WHO). Epilepsy starts most often in young children and people over 50 (NHS). In adults who have had a first seizure, a previous brain insult such as head injury, stroke or tumour raises the chance of recurrence (AAN/AES). See our guides on stroke and adult brain tumours.

Diagnosis: EEG and MRI limits, and a first seizure

A specialist reviews what happened before, during and after the seizure, and a witness account helps. Tests can include EEG (brain electrical activity), ECG (heart), MRI or CT, blood tests and genetic testing (NHS). The ECG matters because some faints and heart rhythm problems look like seizures.

The limits of the tests. According to the Epilepsy Society, a normal EEG does not exclude epilepsy and a normal MRI does not either; neither test alone confirms or rules out the diagnosis, which rests on history, witness accounts and test results together (Epilepsy Society, via search record). In the other direction, an EEG showing epilepsy-type activity, a significant brain-imaging abnormality, or a seizure during sleep all mark a higher chance of another seizure after a first one (AAN/AES).

After a first seizure. The chance of another is 21% to 45% in the first two years. Starting medicine right away is likely to lower the two-year recurrence risk but is unlikely to change long-term remission, and medicine side effects occur in 7% to 31% (Krumholz 2015). The guideline gives no yes-or-no answer; it asks patient and doctor to weigh risks and preferences (AAN/AES). NICE NG217, is the UK guideline for diagnosis and treatment (NICE); its recommendations (last updated 5 August 2026) say to consider lamotrigine or levetiracetam as first-line monotherapy for focal seizures, and to offer a choice of lamotrigine, levetiracetam or sodium valproate for generalised tonic-clonic seizures, following the MHRA valproate safeguards.

Treatments: what works and what does not

Antiseizure medicines

Choosing a medicine depends on seizure type, age, sex, other conditions and pregnancy potential. SIGN 143 states that for focal seizures lamotrigine is the drug of choice, and for genetic generalised epilepsy valproate is the most effective, with lamotrigine or levetiracetam options for women of childbearing age (SIGN 143 quick reference guide, via search record). SANAD II backs both points: lamotrigine had fewer treatment failures than levetiracetam (HR 0.60, 95% CI 0.46 to 0.77) or zonisamide (HR 0.46, 0.36 to 0.60), and the authors did not support routine first-line levetiracetam or zonisamide for focal epilepsy (Marson 2021). For generalised epilepsy, levetiracetam was "neither clinically effective nor cost-effective" compared with valproate (Marson 2021).

The NICE guideline NG217 covers drug choice for adults, young people and children. NICE NG217 was last updated on 5 August 2026 (NICE). For focal seizures it considers lamotrigine or levetiracetam first; for generalised tonic-clonic seizures it offers a choice of lamotrigine, levetiracetam or sodium valproate, with the MHRA valproate safeguards applied. That is a broader choice than the SANAD II cost-effectiveness finding alone.

The NHS says finding the right medicine can take a few tries, dose changes should only be made with a specialist, and after two years without seizures a gradual stop may be considered (NHS). Benzodiazepines such as clonazepam are used in some epilepsies and as emergency rescue; see our clonazepam evidence review, plus lorazepam and diazepam.

Pregnancy planning and the valproate programme

Healthy pregnancies are possible with epilepsy, but anyone planning one, or who finds they are pregnant, should contact their specialist team promptly without stopping medicine on their own (NHS).

Valproate

The MHRA says valproate is contraindicated in girls and women of childbearing potential unless the conditions of its Pregnancy Prevention Programme are met, and should be used only if other treatments are ineffective or not tolerated. Roughly 30% to 40% of exposed children have serious developmental disorders and about 10% have congenital malformations; no safe dose in pregnancy is established; and anyone on the programme should be reviewed at least annually (MHRA, February 2020). From 31 January 2024 valproate must not be started in anyone under 55 unless two specialists independently document that no other effective, tolerated treatment exists (SIGN page summarising MHRA).

For men, the MHRA issued precautionary advice in September 2024 after a Nordic study: cumulative risk of neurodevelopmental disorders in children was 4.0% to 5.6% with fathers on valproate versus 2.3% to 3.2% with lamotrigine or levetiracetam (pooled adjusted HR 1.50, 95% CI 1.09 to 2.07). The MHRA says causation is unconfirmed, the study had no untreated group, and the country-level results showed no association on their own. It advises effective contraception during treatment and for 3 months after, and tells men not to stop valproate without advice (MHRA, September 2024).

Other medicines

The MHRA's pregnancy review called lamotrigine and levetiracetam the safer options, citing more than 12,000 and 1,800 pregnancies, with no increase in major malformations at usual doses. Carbamazepine raises malformation risk, dose-dependently, in about 9,000 exposed pregnancies. Women taking antiseizure medicines who plan a pregnancy should be offered folic acid 5 mg daily, started beforehand (MHRA, January 2021). The EURAP registry found the same ranking: valproate 10.3%, carbamazepine 5.5%, lamotrigine 2.9%, levetiracetam 2.8% (Tomson 2018).

Topiramate: since June 2024 it is contraindicated in women of childbearing potential unless a prevention programme is met, and in pregnancy for epilepsy unless no other suitable treatment exists. Effective contraception is advised during treatment and for at least four weeks after (MHRA, June 2024). Seizures themselves also matter in pregnancy, which is why abrupt stopping is discouraged (MHRA).

SUDEP, safety and driving

SUDEP

The AAN/AES guideline put adult SUDEP at about 1.2 per 1,000 patient-years (95% CI 0.64 to 2.32, low confidence) and childhood SUDEP at 0.22 per 1,000 (95% CI 0.16 to 0.31, high confidence). Tonic-clonic seizures are the major risk factor, and risk rises as their frequency rises. Being seizure-free, especially from tonic-clonic seizures, is strongly linked to lower risk (the exact multiplier and any bedroom-supervision finding were not verified by us, so we omit them) (Harden 2017). The NHS describes SUDEP as rare and says effective treatment lowers the risk (NHS). The WHO fact sheet does not mention SUDEP (WHO).

Everyday safety

The NHS suggests medical ID, a trigger diary, a seizure alarm, showers rather than baths, and care when swimming or cooking, and warns against missing doses, changing treatment without advice, and heavy alcohol use (NHS).

Driving: UK (DVLA)

After any seizure, driving must stop and the DVLA must be informed (NHS). For a car or motorcycle (Group 1): a first unprovoked seizure means stopping for 6 months, or 12 months if an underlying risk factor is present; established epilepsy means 12 months from the last seizure; seizures only in sleep have their own concession; and a provoked seizure usually means 6 months. During medicine withdrawal do not drive, and not for 6 months after the last dose. For lorry and bus (Group 2): 10 years seizure-free without medicine for epilepsy, or 5 years after an isolated seizure (DVLA). Also see our guide on drowsy driving.

Driving: United States (DMV)

US rules are set by each state. Epilepsy Foundation New England says every state restricts driving after seizures, the motor-vehicle agency (not the doctor) decides in most states, and the patient has a duty to report. In its four states the required seizure-free period ranges from as little as 3 months to 2 years (Epilepsy Foundation New England). A US consensus statement proposes 3 months, with state intervals running from about three months to a year (Sanchez and Krumholz, NeurologyLive 2018). We could not open the Epilepsy Foundation's national state-by-state database; check your state agency.

Surgery and neurostimulation

Surgery

In the Canadian trial of 80 people with temporal lobe epilepsy, 58% of the surgery group were free of awareness-impairing seizures at one year against 8% on medicines, with adverse effects in 4 (10%) surgical patients (Wiebe 2001). A US trial of earlier surgery enrolled 38 people (planned 200) and stopped for slow recruitment: 11 of 15 surgical patients and none of 23 on medicines were seizure-free in year 2; memory decline occurred in 4 surgical participants (Engel 2012). Surgery needs full specialist work-up, and the NHS lists it as a rare option (NHS).

Vagus nerve stimulation

The Cochrane review found a 50% or greater reduction more likely with high-level than low-level stimulation (RR 1.73, 95% CI 1.13 to 2.64; moderate certainty), with hoarseness more common at high levels (RR 2.17, 99% CI 1.49 to 3.17). It called VNS effective and well tolerated but said evidence is moderate to low certainty (Panebianco 2022).

Brain-implanted stimulators

The responsive-stimulation trial randomised 191 adults: seizures fell 37.9% with active stimulation and 17.3% with sham at 12 weeks (P=0.012) (Morrell 2011). The thalamic trial of 110 adults found a 29% greater reduction than control in the final blinded month, 56% median reduction at 2 years, five deaths unrelated to the device, and more reports of depression and memory problems in the stimulated group (Fisher 2010).

Ketogenic diet and cannabidiol

Ketogenic diet

The Cochrane review included 13 trials (932 people; 711 children, 221 adults). All were at high risk of bias because they could not be blinded, and certainty was low to very low. In children, seizure freedom was more likely (RR 3.16, 95% CI 1.20 to 8.35; 385 children) and so was a 50% or greater reduction (RR 5.80, 95% CI 3.48 to 9.65). Vomiting, constipation and diarrhoea were common, also in usual-care groups. In adults nobody became seizure-free and the 50% reduction result was not significant. The authors consider it reasonable for people who are not suitable for surgery (Martin-McGill 2020; record). Diets are started under dietitian and doctor supervision, not as a do-it-yourself plan.

Cannabidiol (Epidiolex)

Epidiolex is a prescription cannabidiol solution. The US label covers seizures in Lennox-Gastaut syndrome, Dravet syndrome and tuberous sclerosis complex from age 1. Liver enzyme rises above 3 times the upper limit occurred in 13% on the drug versus 1% on placebo, with valproate raising the risk most. Liver tests are required before starting and at 1, 3 and 6 months. It raises levels of the active clobazam metabolite, and somnolence occurred in 32% versus 11% on placebo in LGS/Dravet. Label doses: start 2.5 mg/kg twice daily, maintenance up to 10 mg/kg twice daily for LGS/Dravet (Epidiolex label).

In Dravet syndrome (120 patients), convulsive seizures fell from 12.4 to 5.9 a month on cannabidiol versus 14.9 to 14.1 on placebo; the share with at least a 50% reduction was 43% versus 27% (OR 2.00, P=0.08), and 5% versus 0% became seizure-free (Devinsky 2017). NICE appraisal TA614 covers cannabidiol with clobazam for Dravet seizures (NICE); NICE published it on 18 December 2019 and notes a commercial access agreement with the maker. A 2014 Cochrane review of earlier cannabinoid studies could draw no reliable conclusions (Gloss 2014). Over-the-counter CBD is not Epidiolex and is not an epilepsy treatment; see our CBD evidence review for quality, liver and legal issues.

Supplements, lifestyle and complementary claims graded

Trigger and lifestyle advice is common sense rather than trial-proven: the NHS recommends a trigger diary and avoiding heavy drinking (NHS).

Risks and side effects

Other documented effects: sedation and appetite loss with cannabidiol (Devinsky 2018); hoarseness with vagus nerve stimulation (Cochrane); depression and memory complaints with thalamic stimulation (Fisher 2010); digestive upset and, in one adult trial, higher cholesterol with ketogenic diets (Cochrane); memory decline after temporal resection in a small trial (Engel 2012). Medicine side effects in general ranged 7% to 31% in the first-seizure evidence (Krumholz 2015).

Interactions. Cannabidiol raises the active clobazam metabolite and liver-enzyme risk with valproate (label). Herbal products may interfere with medicines (Epilepsy Action Australia). Tell a pharmacist before adding anything.

Who to see

A first seizure needs urgent assessment (NHS). A specialist (neurologist or epilepsy specialist nurse) should choose and review the medicine; an experienced specialist must judge whether valproate is necessary for anyone who could become pregnant (MHRA). Drug-resistant epilepsy should be reviewed for surgery, devices or diet at a specialist centre (NHS). A dietitian is essential for ketogenic diets, and a pharmacist for interactions.

Follow the money: makers, trials and patient groups

Who makes and sells the treatments

• Cannabidiol: Epidiolex is made by GW Pharmaceuticals, bought by Jazz Pharmaceuticals (headquartered in Dublin, Ireland) in a deal completed on 5 May 2021 at $220 per GW American depositary share ($200 cash plus $20 in Jazz shares). Epidiolex exceeded $500 million in net sales in 2020 (company announcement via BioSpace). The label names Jazz as manufacturer (DailyMed).
• Devices: NeuroPace developed and funded the responsive stimulator trial (conference report); the first author works there (Morrell 2011). The thalamic trial's grants came from the US NINDS (Fisher 2010 record); ClinicalTrials.gov lists Medtronic as lead sponsor of that trial (NCT00101933).
• Medicines: UCB is the American Epilepsy Society's top-tier corporate supporter ($500,000+ a year); Jazz is in the next tier ($250,000 to $499,000), with SK Life Science, LivaNova, Takeda and others at $100,000 to $249,999 (AES, 2023 recognition). NeuroPace, Medtronic and many monitoring companies also appear.

Who funded the evidence

• Public: SANAD II (NIHR HTA, 09/144/09, funder had no role in design) (Marson 2021); the ketogenic diet and melatonin Cochrane reviews (NIHR) (record; record); the ERSET surgery trial (NINDS) (Engel 2012).
• Mixed: the EURAP pregnancy registry lists Bial, Eisai, GSK, Janssen-Cilag, Novartis, Pfizer, Sanofi-Aventis, UCB, the Netherlands Epilepsy Foundation and Stockholm County Council in its abstract, with 18 agencies in its grant list (Tomson 2018 record). SANAD II authors report grants or fees from UCB, Eisai, GW, GSK and others outside the submitted work, and one author sat on the NICE panel (disclosures). The SUDEP guideline authors list fees from UCB, Eisai, Bial, GSK, Upsher-Smith and device makers (Harden 2017).
• Manufacturer: both pivotal Epidiolex trials (2018; 2017) and the responsive stimulation trial.

Patient groups and professional societies

• American Epilepsy Society: accepts support under accreditation and commercial-support policies and discloses it by tier; the page does not explain how sponsors are kept from influencing guidance (AES).
• Epilepsy Foundation (US): ProPublica's Form 990 data show fiscal 2025 revenue of about $16.2 million and expenses of about $19.0 million, mostly contributions; the breakdown by donor is not shown (ProPublica). Its New England branch says it partners with the national body (Epilepsy Foundation New England).
• Epilepsy Action (UK): its published policy says it accepts pharmaceutical and other commercial funding with partnership agreements, editorial control retained, and reports gifts of £10,000 or more in its annual report (policy text seen in search results; we could not open the page).

Documented integrity events

The regulatory history is itself a money-and-safety story: MHRA restricted valproate in girls and women from 2018, strengthened rules in 2020 after finding gaps in implementation, widened them to all new patients under 55 in January 2024, and added precautionary advice for men in September 2024 (MHRA 2020; MHRA 2024). We found no fraud or enforcement action in our sources and did not search for litigation.

Gaps

We could not trace the Epilepsy Foundation's donors from the record we saw, the national Epilepsy Society's finances, or the NeuroPace trial's full disclosure statements. Several author declarations were not available in the records.

Related research

For wider, evidence-graded context, see these Pure City Research guides:

• Clonazepam: independent evidence
• Lorazepam (Ativan): independent evidence
• Diazepam (Valium): independent evidence
• CBD: independent evidence
• Functional neurological disorder
• Benign sleep myoclonus of infancy
• Stroke
• Adult primary brain and CNS tumours
• Melatonin
• Magnesium
• Vitamin D
• Drowsy driving

Frequently asked questions

Sources and funding notes

What we could not verify: the full NICE technology appraisal text (we read only the TA614 overview page); the Epilepsy Society's own diagnosis page, the Epilepsy Action Australia complementary-therapies page and the Epilepsy Action industry-funding policy page (seen in search results only); the DVLA guidance beyond the headings summarised; the Epilepsy Foundation's national driving-law database; the full funding statement of the Wiebe surgery trial; and competing-interest statements for several trials. We did not search for lawsuits or enforcement actions against antiseizure drug makers. The MHRA topiramate page is dated 20 June 2024 by its own text, while some secondary sources give 11 June.

1. WHO. Epilepsy fact sheet — global figures; funded by member states and donors.
2. NHS. Epilepsy — patient information; UK government.
3. DVLA. Neurological disorders: assessing fitness to drive — UK licensing rules; UK government.
4. NICE. NG217 Epilepsies in children, young people and adults — guideline; read in full (PDF, updated 5 August 2026); UK government.
5. SIGN 143. Diagnosis and management of epilepsy in adults and quick reference guide — Scottish guideline; funder not stated.
6. AAN/AES. First unprovoked seizure in adults and Krumholz 2015 — guideline; supported by AAN.
7. Harden et al. 2017. SUDEP incidence and risk factors — guideline; AAN funded, authors report industry fees.
8. Fisher et al. 2017. ILAE operational classification of seizure types — position paper; grants incl. Eisai, Medtronic, Pfizer and public funders.
9. Epilepsy Society. Diagnosis — UK charity; seen in search results.
10. Marson et al. 2021. SANAD II focal epilepsy and generalised epilepsy — randomised trials; NIHR HTA.
11. Tomson et al. 2018. EURAP comparative malformation risk — registry; many drug-company and public funders.
12. MHRA. Valproate pregnancy prevention programme (2020); valproate use in men (2024); antiepileptic drugs in pregnancy (2021); topiramate (2024) — regulator; UK government.
13. Wiebe et al. 2001. Surgery for temporal-lobe epilepsy — randomised trial; non-US government support.
14. Engel et al. 2012. Early surgical therapy — randomised trial; NINDS.
15. Panebianco et al. 2022. Cochrane: vagus nerve stimulation — systematic review; no funding listed.
16. Morrell 2011. Responsive cortical stimulation and conference report — trial; NeuroPace.
17. Fisher et al. 2010. Anterior thalamic stimulation — trial; NINDS grants listed.
18. Martin-McGill et al. 2020. Cochrane: ketogenic diets (full text; AFP summary) — systematic review; NIHR.
19. Devinsky et al. 2018. Cannabidiol in Lennox-Gastaut and 2017 Dravet trial — trials; GW Pharmaceuticals.
20. DailyMed. Epidiolex label — label; Jazz Pharmaceuticals.
21. BioSpace. Jazz completes acquisition of GW — company announcement; Jazz.
22. NICE TA614. Cannabidiol with clobazam for Dravet syndrome — appraisal; not opened.
23. Brigo 2016. Melatonin; Panebianco 2017. Yoga; Cheuk 2014. Acupuncture; Gloss 2014. Cannabinoids — Cochrane reviews; melatonin review NIHR-funded, others no funding listed.
24. Epilepsy Action Australia. Complementary therapies — charity; seen in search results.
25. Epilepsy Foundation New England. Driving; Sanchez and Krumholz 2018 — US driving context; charity branch; trade publication, funder not stated.
26. American Epilepsy Society. Corporate support; ProPublica. Epilepsy Foundation; Epilepsy Action industry policy (page text seen in search results; its link returned 404 when we opened it) — funding disclosures.

Educational research reviewed 6 October 2026. This article is not medical advice; talk to a qualified clinician or pharmacist before starting, stopping or combining treatments.