Epilepsy Precision Medicine: Core Findings in 4 Subgroups
Where a molecular diagnosis changes the prescription, and where it changes nothing at all.
One-Minute Takeaway
- Evidence for epilepsy precision medicine reaches its highest tier — randomised, placebo-controlled trials — in only a handful of molecular subgroups; the rest rests on cohorts and functional studies.
- Rapid genome sequencing identified a diagnosis in 43% of 100 infants with new-onset epilepsy (95% CI 33–53), and that diagnosis informed treatment in 24 of the 43.
- Adjunctive everolimus reduced seizure frequency by a median 39.6% (95% CI 35.0–48.7) versus 14.9% on placebo (95% CI 0.1–21.7) across 366 patients with tuberous sclerosis complex.
- Fenfluramine produced a 62.3% greater reduction in monthly convulsive seizures than placebo (95% CI 47.7–72.8) in 119 patients with Dravet syndrome.
- Sodium channel blocker choice tracks variant function rather than gene name: benefit in gain-of-function, aggravation in loss-of-function.
Why Epilepsy Precision Medicine Matters
For paediatric neurologists the question is no longer whether a genetic cause can be found. It is whether finding it changes what you prescribe.
That question now has data behind it. One retrospective cross-sectional study followed 418 patients carrying a pathogenic or likely pathogenic variant on a multigene panel. No management change was reported for 178 of them (42.6%), meaning management shifted for the majority.
Median age at testing was 4 years, so this is largely a paediatric dataset. Among the 167 patients with follow-up information, 108 (64.7%) reported reduction or elimination of seizures. The design is observational, so these outcomes are associated with a genetic diagnosis rather than shown to be caused by one.
A minority moved the other way: 6 patients (3.6%) reported increased seizure frequency.
The Gene-STEPS cohort sharpens the paediatric case. Yield reached 74% in infants with neonatal-onset seizures against 36% with later infantile onset. Referral setting mattered similarly: 71% among infants referred from intensive care against 30% from outpatient clinics.
The authors nonetheless concluded that testing should not be restricted to infants in intensive care, since an outpatient yield near one in three still exceeds most diagnostic tests routinely ordered in this population.
This is the premise of epilepsy precision medicine: the diagnostic step is now reliable enough that the interesting problem has moved downstream, to what the molecular result licenses you to do.
A Practical Taxonomy of Epilepsy Precision Medicine
Gene lists are a poor organising principle at the bedside. Hundreds of genes are now associated with epilepsy, and the list is not a treatment algorithm.
What is actionable is mechanism. Four molecular categories account for nearly all of the treatment-changing findings a paediatric neurologist will encounter. Each carries a distinct kind of evidence and a distinct level of certainty, which is why epilepsy precision medicine is better read as four evidence bases than one.
The table below is organised around that distinction. Read it as four separate evidence bases rather than one, because the strength of the data varies more between categories than within them.
| Study / trial | Design | n | Population | Target or comparison | Key result |
|---|---|---|---|---|---|
| Sheidley 2022 | Systematic review, random-effects meta-analysis | 154 studies | Cohorts ascertained for epilepsy | Diagnostic yield by test modality | Overall 17%; genome sequencing 48%, exome 24%, multigene panel 19%, microarray 9% |
| Gene-STEPS (D’Gama 2023) | Prospective multicentre cohort | 100 | Infants, new-onset epilepsy, 4 countries | Rapid trio genome sequencing | Diagnosis in 43% (95% CI 33–53); 34 distinct genes or regions; treatment informed in 24 of 43 |
| McKnight 2022 | Retrospective cross-sectional | 418 | Epilepsy with P/LP panel variant, median age 4 y | Management change after diagnosis | No change in 178 (42.6%); 108 of 167 with follow-up (64.7%) reported reduced or eliminated seizures |
| Wolff 2017 | Cohort with patch-clamp characterisation | 201 (71 new, 130 reviewed) | SCN2A-related disorders | Sodium channel blockers by age at onset | Seizure freedom with onset under 3 months; rarely effective and sometimes aggravating with later onset |
| Johannesen 2022 | International cohort, functional studies, modelling | 392 | SCN8A-related disorders | Sodium channel blockers vs other agents by variant function | Gain-of-function carriers (136 tested) responded significantly better to sodium channel blockers than to other antiseizure medications |
| Lagae 2019 | Phase 3 RCT, double-blind, placebo-controlled | 119 | Dravet syndrome | Fenfluramine 0.7 or 0.2 mg/kg/day vs placebo | 0.7 mg/kg: 62.3% greater reduction in mean monthly convulsive seizures (95% CI 47.7–72.8). 0.2 mg/kg: 32.4% (95% CI 6.2–52.3) |
| Laux 2026 | Phase 1–2a open-label, uncontrolled, plus extensions | 81 (75 into extension) | Dravet syndrome, ages 2–18 | Zorevunersen, an SCN1A-directed antisense oligonucleotide | Median change in convulsive-seizure frequency −58.82% to −90.91% across monthly intervals over 20 extension months. No placebo comparator |
| EXIST-3 (French 2016) | Phase 3 RCT, double-blind, placebo-controlled | 366 | TSC with treatment-resistant focal seizures; enrolled median age 10.1 y (range 2.2–56.3), 28% under 6 y | Adjunctive everolimus, high or low exposure, vs placebo | Median reduction 39.6% high exposure (95% CI 35.0–48.7), 29.3% low exposure (95% CI 18.8–41.9), 14.9% placebo (95% CI 0.1–21.7) |
| EPISTOP (Kotulska 2021) | Randomised at 6 sites, open-label allocation at 4 | 94 | Infants with TSC, no seizure history | Pre-emptive vs conventional vigabatrin | Time to first clinical seizure 364 d (95% CI 223–535) vs 124 d (95% CI 33–149) in the randomised arm; pooled 24-month OR 0.23 for drug-resistant epilepsy |
| PREVeNT (Bebin 2024) | Phase IIb RCT, double-blind, placebo-controlled | 56 randomised (84 enrolled) | Infants with TSC and epileptiform EEG | Early vigabatrin vs placebo | No difference in Bayley-III cognition, epilepsy incidence, drug-resistant epilepsy or time to first seizure at 24 months. Infantile spasms less frequent and later |
| Marigold (Knight 2022) | Phase 3 RCT, double-blind, placebo-controlled | 101 | CDKL5 deficiency disorder, ages 2–21 | Adjunctive ganaxolone vs placebo | Median change in 28-day major motor seizure frequency −30.7% (IQR −49.5 to −1.9) vs −6.9% (IQR −24.1 to 39.7); p=0.0036 |
| Zhang 2025 | Retrospective multicentre cohort | 90 | Glut1 deficiency syndrome (SLC2A1) | Ketogenic dietary therapy | Seizures controlled in 47 of 57 treated patients (82%); movement disorder resolved in 18 of 47 |
Category 1 — Sodium Channelopathies
This is the subgroup where epilepsy precision medicine most often changes the prescription, and where the change is sometimes a subtraction rather than an addition.
Variant function, not gene name, directs the drug
The organising insight is that SCN1A, SCN2A and SCN8A do not behave as three separate treatment questions. They behave as one question asked three times: does this variant increase or decrease channel activity?
Wolff and colleagues characterised 201 patients with SCN2A-related disorders and found the answer tracked age at onset. Sodium channel blockers were associated with seizure freedom when epilepsy began before 3 months, an age window populated by gain-of-function variants.
With onset after 3 months the same drugs were rarely effective and sometimes aggravated seizures. Truncating variants, which are loss-of-function by construction, appeared exclusively in the late-onset non-responder group.
Johannesen and colleagues found the same architecture in 392 individuals with SCN8A variants. Gain-of-function carriers responded significantly better to sodium channel blockers than to other antiseizure medications; loss-of-function carriers instead presented with generalised epilepsy and absence seizures.
Brunklaus and colleagues extended this reasoning across the sodium channel gene family, showing that biophysical characterisation of variants in one SCN gene can inform predicted function in another. This matters because functional testing is resource-intensive and performed for only a minority of variants.
Clinical Pearl
Age at seizure onset is the cheapest available proxy for variant function in the sodium channelopathies. Onset before 3 months favours gain-of-function and a sodium channel blocker trial; onset after 3 months favours loss-of-function, where the same drug class may worsen seizures. The proxy is imperfect, and where the variant is novel or the phenotype atypical, functional prediction rather than gene name should drive the decision.
Dravet syndrome: the exception that proved the rule
Dravet syndrome, driven predominantly by SCN1A haploinsufficiency, was the first widely adopted example of a genetic diagnosis mandating drug avoidance. It is also where the field’s most active pipeline sits.
Fenfluramine produced the largest controlled effect in this category: a 62.3% greater reduction in mean monthly convulsive seizure frequency than placebo (95% CI 47.7–72.8) among 119 patients. Echocardiography during the trial showed valve function within normal physiological range in all patients.
Zorevunersen moves closer to the actual mechanism, using an antisense oligonucleotide to raise NaV1.1 production from the non-mutated SCN1A allele. Across 81 patients in two phase 1–2a studies and their extensions, median convulsive-seizure frequency fell between 58.82% and 90.91% from baseline across monthly intervals.
Those studies were open-label and uncontrolled, so the figures carry no comparator and cannot be read as a treatment effect. Two participants died of sudden unexpected death in epilepsy and one of malnutrition. The phase 3 EMPEROR study, which excludes gain-of-function SCN1A variants, has not reported.
Category 2 — mTOR-Pathway Epilepsies
Tuberous sclerosis complex supplied epilepsy precision medicine with its first mechanism-directed antiseizure drug. It also supplies the field’s most instructive contradiction.
Everolimus: a target-directed effect of moderate size
EXIST-3 randomised 366 patients aged 2 to 65 with treatment-resistant focal seizures to adjunctive everolimus at two trough exposures or placebo. Median seizure-frequency reduction was 39.6% at high exposure (95% CI 35.0–48.7) against 14.9% on placebo (95% CI 0.1–21.7).
The gradient between exposures is the part worth noting: 29.3% at low exposure (95% CI 18.8–41.9) sits between placebo and high exposure, which is what a genuine target effect should look like.
The cost is real. Grade 3 or 4 adverse events occurred in 24% of the high-exposure group against 11% on placebo. Serious adverse events were reported in 14% against 3%.
Pre-emptive vigabatrin: two trials, two answers
TSC is one of the few epilepsies that can be diagnosed before the first seizure, which makes prevention testable. Two trials tested it and disagreed.
EPISTOP followed 94 infants with TSC and no seizure history using monthly video EEG, starting vigabatrin either at the first epileptiform EEG or after the first seizure. Time to first clinical seizure was 364 days (95% CI 223–535) with pre-emptive treatment against 124 days (95% CI 33–149).
Pooled 24-month analysis reported an odds ratio of 0.23 for drug-resistant epilepsy (p=0.022). That pooling combined the randomised arm with an open-label arm from four additional sites, which weakens the inference.
PREVeNT then randomised 56 infants under double-blind, placebo-controlled conditions. It found no difference in Bayley-III cognitive composite at 24 months, and none in epilepsy incidence, drug-resistant epilepsy, or time to first seizure.
The two trials agree on one point: infantile spasms were less frequent and occurred later with pre-emptive vigabatrin. Beyond that, the honest reading is that pre-emptive treatment in TSC remains contested rather than settled.
Category 3 — Metabolically Treatable Epilepsies
This subgroup carries the weakest trial evidence in epilepsy precision medicine and, paradoxically, the highest stakes for a missed diagnosis. The intervention is substrate, not receptor, and the therapeutic window closes with time.
Glut1 deficiency and ketogenic dietary therapy
Glut1 deficiency syndrome, caused by SLC2A1 variants, produces epilepsy that resists conventional drugs because the problem is fuel delivery across the blood–brain barrier rather than excitability.
In a retrospective cohort of 90 Chinese patients, seizures were controlled in 47 of 57 treated patients (82%) on ketogenic dietary therapy. Movement disorder resolved in 18 of 47 and improved in a further 26.
No randomised trial exists, and none is likely: the mechanistic rationale and the observed response rate make equipoise difficult to defend. Evidence here is consistent across cohorts but observational throughout.
Pyridoxine-dependent epilepsy
Pyridoxine-dependent epilepsy from ALDH7A1 deficiency presents with neonatal seizures resistant to standard drugs but responsive to pharmacological pyridoxine. Seizure control, however, has not translated into normal development for most patients.
The 2021 International PDE Consortium consensus guideline recommends lysine reduction — dietary lysine restriction and arginine supplementation — as adjunct therapy. It reports improved biochemical parameters and cognitive development in many but not all patients.
The guideline grades its own statements using GRADE, and its recommendations include statements resting on expert consensus at the lowest evidence levels. That candour is itself useful, and it is the appropriate register for this subgroup.
Clinical Pearl
The metabolically treatable epilepsies are the strongest argument for testing early rather than testing eventually. A trial-grade effect size is not the relevant standard when the alternative is years of ineffective antiseizure medication and avoidable developmental cost.
Category 4 — Synaptic and Neurosteroid-Responsive Epilepsies
The fourth category covers developmental and epileptic encephalopathies where the molecular lesion sits in synaptic or signalling machinery rather than in an ion channel or a metabolic pathway.
CDKL5 deficiency disorder and ganaxolone
The Marigold trial randomised 101 patients aged 2 to 21 with a pathogenic or likely pathogenic CDKL5 variant to adjunctive ganaxolone or placebo. Median change in 28-day major motor seizure frequency was −30.7% (IQR −49.5 to −1.9) against −6.9% (IQR −24.1 to 39.7) on placebo, p=0.0036.
The effect is real and modest. The placebo interquartile range extends to +39.7%, which is a reminder of how variable seizure counts are in this population and how much a trial of this size can and cannot resolve.
Ganaxolone is a positive allosteric modulator of GABAA receptors. Its selection for CDKL5 rests on neurosteroid biology rather than on correcting the CDKL5 defect, which places it closer to syndrome-targeted than to variant-targeted therapy.
Where this category is still mostly promise
For most genes in this group — STXBP1, SYNGAP1, KCNQ2 among them — a molecular diagnosis currently delivers prognosis, recurrence-risk counselling and trial eligibility rather than a different prescription.
That is a legitimate return on testing, and the Gene-STEPS data quantify it. A genetic diagnosis informed prognosis in 37 of 43 cases and recurrence-risk counselling in every case. Treatment was informed in 24 of 43.
Comparative Analysis
Reading the four subgroups side by side, three patterns in epilepsy precision medicine emerge that no single trial shows.
First, effect sizes are moderate and remarkably consistent across mechanisms. Everolimus, fenfluramine and ganaxolone were tested in different diseases against different targets, yet all landed in the range of a 30% to 60% placebo-adjusted reduction in seizure frequency. None produced seizure freedom at a population level.
Second, most licensed “precision” therapy is syndrome-targeted rather than variant-targeted. Fenfluramine treats Dravet syndrome; it does not correct SCN1A haploinsufficiency. Ganaxolone treats CDKL5-associated seizures through GABAA modulation, not through CDKL5 itself.
Everolimus and zorevunersen are the two genuine mechanism-directed agents in this table. Only one of them has controlled data.
Third, the largest clinical gains may come from subtraction. Withdrawing a sodium channel blocker from a child with loss-of-function SCN1A or late-onset SCN2A disease costs nothing and requires no new drug. No randomised trial has quantified that benefit, and the cohort evidence is consistent enough that one is unlikely to be run.
The categories also differ in evidence architecture. The mTOR and neurosteroid subgroups rest on randomised trials with modest effects. The channelopathy subgroup rests on large cohorts with functional corroboration — observational, but mechanistically coherent.
The metabolic subgroup rests on cohorts and consensus alone.
That asymmetry is not a defect in the field so much as a consequence of prevalence. Randomising a disorder that affects a few hundred children worldwide is often not feasible, and demanding trial-grade evidence uniformly would leave the treatable metabolic epilepsies untreated.
Quality and Consistency of the Evidence
Strength varies by outcome rather than by topic. Graded by outcome, the picture is as follows.
- Diagnostic yield of sequencing in childhood epilepsy — high. A meta-analysis of 154 studies and a prospective four-country cohort converge: yield is highest for genome sequencing and highest again in neonatal onset. Point estimates differ between reviews because inclusion criteria differ, not because the reviews conflict.
- Everolimus for TSC-associated refractory seizures — moderate. One large phase 3 RCT with a coherent exposure gradient, non-overlapping confidence intervals against placebo, and a known adverse-event burden. Not replicated by a second independent phase 3 trial.
- Fenfluramine for Dravet syndrome — moderate. Randomised, double-blind and placebo-controlled with a large effect, but a single-syndrome population of 119 and industry sponsorship. A second randomised trial in stiripentol-treated patients and a third trial support the direction.
- Ganaxolone for CDKL5 deficiency disorder — moderate. Randomised and placebo-controlled at n=101, with a modest median effect and wide interquartile ranges. First controlled evidence in this disorder.
- Genotype-directed sodium channel blocker use — moderate. Two large international cohorts (n=201 and n=392) with independent functional corroboration and a consistent direction. Observational, retrospective and susceptible to reporting bias.
- Pre-emptive vigabatrin in TSC — low, and contested. EPISTOP and PREVeNT disagree on seizure and cognitive outcomes and agree only on infantile spasms. EPISTOP’s pooled analysis mixes randomised and open-label allocation; PREVeNT randomised 56 infants.
- Ketogenic therapy in Glut1 deficiency — low by design. Consistent, large cohort responses with strong mechanistic rationale, but no randomised evidence and none anticipated.
- Zorevunersen for Dravet syndrome — low, pending phase 3. Open-label, uncontrolled, n=81, with substantial reported reductions but no comparator. Phase 3 EMPEROR has not reported.
What the Evidence Does Not Show
Named gaps in the current evidence
Each of these is a specific unanswered question, not a general call for further study.
No trial has tested a diagnosis-to-treatment pathway as the intervention. The studies above measured what clinicians did after a result, not what happens to outcomes when sequencing is randomised against usual care. The central claim of epilepsy precision medicine — that testing early improves outcomes — remains untested by that design.
No controlled evidence quantifies the benefit of withdrawing contraindicated drugs. The sodium channelopathy literature supports drug avoidance on cohort and functional grounds. The size of that benefit, and how much of it is recoverable if withdrawal happens late, is unmeasured.
Developmental outcome has not moved. PREVeNT found no cognitive benefit from pre-emptive vigabatrin at 24 months. Pyridoxine controls seizures in ALDH7A1 deficiency without normalising development in most patients.
Seizure reduction and developmental trajectory have not been shown to be the same target.
Head-to-head comparisons are absent. Fenfluramine, cannabidiol and stiripentol have each been compared with placebo in Dravet syndrome, never with one another. Sequencing decisions in this syndrome rest on indirect comparison.
Populations are narrow. The Glut1 cohort was Chinese, and Gene-STEPS recruited from four high-income countries. Generalisability to settings without rapid sequencing access is unestablished.
EXIST-3 spanned 99 centres in 25 countries, but it was not a paediatric trial by design. Median age was 10.1 years (range 2.2–56.3), with 104 patients (28%) under 6 years and 67 (18%) aged 18 or over, so the paediatric estimate rests on a subgroup rather than the headline result.
Long-term safety of the newest agents is unknown. Zorevunersen requires repeated intrathecal dosing. Raised CSF protein was reported in 45% of extension participants and post-lumbar puncture syndrome in 25% of the phase 1–2a cohort.
Everolimus tolerability beyond the 18-week core phase of EXIST-3 is not characterised by the trial evidence cited here, and the adverse-event figures above describe that core phase only.
Practical Implications
The following describe what the evidence on epilepsy precision medicine supports, not what any individual child requires.
Testing threshold. The 2023 National Society of Genetic Counselors evidence-based guideline strongly recommends genetic testing for anyone with unexplained epilepsy, without age limitation. It places exome or genome sequencing, or a panel of more than 25 genes, as first-tier testing followed by chromosomal microarray, and conditionally favours sequencing over a panel.
Timing. Yield in Gene-STEPS was highest in neonatal onset (74%) and in intensive care referrals (71%). It nonetheless reached 30% in outpatients, which supports testing early rather than reserving it for the sickest infants.
Before the result returns. Age at seizure onset offers an interim signal in suspected sodium channelopathy. Onset before 3 months is compatible with gain-of-function and a sodium channel blocker trial; onset later is compatible with loss-of-function, where the class may aggravate seizures.
When the result arrives. Ask which of the four categories it falls into before asking which drug it implies. That question resolves the strength of the evidence you are about to act on, and it resolves whether the action is to add a drug or to stop one.
What to tell families. A genetic diagnosis informed prognosis in 37 of 43 diagnosed infants and recurrence-risk counselling in all of them, against treatment in 24 of 43. Prognosis and counselling are the more reliable yield, and framing them as the expected return sets accurate expectations.
Trial access. For SCN1A, CDKL5, STXBP1 and SYNGAP1, a confirmed molecular diagnosis is now an entry requirement for studies of agents targeting the underlying defect. Untested children are excluded from that pipeline by default.
Evidence Grade and Bottom Line
BModerate. Randomised placebo-controlled evidence exists in three molecular subgroups, with consistent, moderate effect sizes and well-established diagnostic yield. Genotype-directed prescribing rests on large, mechanistically corroborated cohorts rather than trials, and prevention and developmental outcomes remain contested or negative.
- Sequencing reliably finds a cause in childhood epilepsy, and the yield rises sharply with earlier onset and greater severity.
- A molecular result changes management for the majority of children in whom one is found, though the observational design of that evidence limits causal claims.
- Where randomised evidence exists — everolimus, fenfluramine, ganaxolone — effects cluster around a 30% to 60% placebo-adjusted seizure reduction, not seizure freedom.
- Most of what is called precision therapy today is syndrome-targeted. Everolimus, and zorevunersen if phase 3 confirms it, are the mechanism-directed exceptions.
- The clearest and cheapest gain is drug avoidance guided by variant function, and it is supported by cohorts rather than trials.
- Prevention in TSC and developmental outcome across subgroups remain the field’s two open problems.
Article Information and References
- Sheidley BR, Malinowski J, Bergner AL, Bier L, Gloss DS, Mu W, et al. Genetic testing for the epilepsies: A systematic review. Epilepsia. 2022;63(2):375–387. DOI: https://doi.org/10.1111/epi.17141
- Smith L, Malinowski J, Ceulemans S, Peck K, Walton N, Sheidley BR, et al. Genetic testing and counseling for the unexplained epilepsies: An evidence-based practice guideline of the National Society of Genetic Counselors. J Genet Couns. 2023;32(2):266–280. DOI: https://doi.org/10.1002/jgc4.1646
- Coughlin CR 2nd, Tseng LA, Abdenur JE, Ashmore C, Boemer F, Bok LA, et al. Consensus guidelines for the diagnosis and management of pyridoxine-dependent epilepsy due to α-aminoadipic semialdehyde dehydrogenase deficiency. J Inherit Metab Dis. 2021;44(1):178–192. DOI: https://doi.org/10.1002/jimd.12332
- D’Gama AM, Mulhern S, Sheidley BR, Boodhoo F, Buts S, Chandler NJ, et al. Evaluation of the feasibility, diagnostic yield, and clinical utility of rapid genome sequencing in infantile epilepsy (Gene-STEPS): an international, multicentre, pilot cohort study. Lancet Neurol. 2023;22(9):812–825. DOI: https://doi.org/10.1016/S1474-4422(23)00246-6
- McKnight D, Morales A, Hatchell KE, Bristow SL, Bonkowsky JL, Perry MS, et al. Genetic testing to inform epilepsy treatment management from an international study of clinical practice. JAMA Neurol. 2022;79(12):1267–1276. DOI: https://doi.org/10.1001/jamaneurol.2022.3651
- Laux L, Sullivan J, Perry MS, Brunklaus A, Desurkar A, Schreiber JM, et al. Zorevunersen in children and adolescents with Dravet syndrome. N Engl J Med. 2026;394(10):969–982. DOI: https://doi.org/10.1056/NEJMoa2506295
- Lagae L, Sullivan J, Knupp K, Laux L, Polster T, Nikanorova M, et al. Fenfluramine hydrochloride for the treatment of seizures in Dravet syndrome: a randomised, double-blind, placebo-controlled trial. Lancet. 2019;394(10216):2243–2254. DOI: https://doi.org/10.1016/S0140-6736(19)32500-0
- French JA, Lawson JA, Yapici Z, Ikeda H, Polster T, Nabbout R, et al. Adjunctive everolimus therapy for treatment-resistant focal-onset seizures associated with tuberous sclerosis (EXIST-3): a phase 3, randomised, double-blind, placebo-controlled study. Lancet. 2016;388(10056):2153–2163. DOI: https://doi.org/10.1016/S0140-6736(16)31419-2
- Knight EMP, Amin S, Bahi-Buisson N, Benke TA, Cross JH, Demarest ST, et al. Safety and efficacy of ganaxolone in patients with CDKL5 deficiency disorder: results from the double-blind phase of a randomised, placebo-controlled, phase 3 trial. Lancet Neurol. 2022;21(5):417–427. DOI: https://doi.org/10.1016/S1474-4422(22)00077-1
- Kotulska K, Kwiatkowski DJ, Curatolo P, Weschke B, Riney K, Jansen F, et al. Prevention of epilepsy in infants with tuberous sclerosis complex in the EPISTOP trial. Ann Neurol. 2021;89(2):304–314. DOI: https://doi.org/10.1002/ana.25956
- Bebin EM, Peters JM, Porter BE, McPherson TO, O’Kelley S, Sahin M, et al. Early treatment with vigabatrin does not decrease focal seizures or improve cognition in tuberous sclerosis complex: the PREVeNT trial. Ann Neurol. 2024;95(1):15–26. DOI: https://doi.org/10.1002/ana.26778
- Johannesen KM, Liu Y, Koko M, Gjerulfsen CE, Sonnenberg L, Schubert J, et al. Genotype-phenotype correlations in SCN8A-related disorders reveal prognostic and therapeutic implications. Brain. 2022;145(9):2991–3009. DOI: https://doi.org/10.1093/brain/awab321
- Brunklaus A, Feng T, Brünger T, Perez-Palma E, Heyne H, Matthews E, et al. Gene variant effects across sodium channelopathies predict function and guide precision therapy. Brain. 2022;145(12):4275–4286. DOI: https://doi.org/10.1093/brain/awac006
- Wolff M, Johannesen KM, Hedrich UBS, Masnada S, Rubboli G, Gardella E, et al. Genetic and phenotypic heterogeneity suggest therapeutic implications in SCN2A-related disorders. Brain. 2017;140(5):1316–1336. DOI: https://doi.org/10.1093/brain/awx054
- Zhang MJ, Zhang SM, Zhang QP, Wen YX, Wang JP, Jiang YW, et al. Clinical and genetic characteristics of glucose transporter 1 deficiency syndrome in a large cohort of Chinese patients. World J Pediatr. 2025;21(3):274–283. DOI: https://doi.org/10.1007/s12519-025-00884-9