Drug-resistant epilepsy (DRE) is not simply epilepsy that has not responded to medication – it is a clinically defined condition with a specific diagnostic threshold that determines when a patient should be evaluated for non-pharmacological therapies. Applying this definition consistently is the most consequential step in the management of patients with refractory seizures, because premature acceptance of drug resistance delays surgical candidacy evaluation by years in many patients.
The International League Against Epilepsy (ILAE) published the consensus definition of drug-resistant epilepsy in 2010: failure of adequate trials of two tolerated and appropriately chosen anti-seizure drugs (ASDs), whether as monotherapy or in combination, to achieve sustained seizure freedom. This definition has three operationally important components. First, the number of trials is two – not three, four, or five. Patients who have failed two adequate drug trials should be referred for a comprehensive epilepsy evaluation, not kept on successive pharmacological regimens indefinitely. Second, the trials must be adequate: a drug trial is adequate only if the ASD was appropriate for the patient's seizure type and epilepsy syndrome, was given at a dose sufficient to produce therapeutic blood levels or the maximum tolerated dose, and was maintained long enough to assess efficacy (typically at least three to six months). A trial that was discontinued due to intolerable adverse effects before an efficacy conclusion was possible does not count as evidence of pharmacoresistance. Third, the definition applies only when seizure freedom is the outcome measure; reduction in seizure frequency short of freedom does not satisfy the criterion for drug responsiveness.1
Drug-resistant epilepsy affects approximately 30% of people with epilepsy – a figure that has remained stable across decades despite the introduction of more than twenty new ASDs since 1990. The 70/30 rule describes this pattern: roughly 70% of patients with newly diagnosed epilepsy achieve seizure freedom with the first or second ASD tried, while the remaining 30% have a low probability of achieving seizure freedom with any subsequent ASD regardless of how many additional drugs are tried. Kwan and Brodie's landmark prospective study in 2000, which followed 470 newly diagnosed patients, established this probability distribution and demonstrated that once two adequate ASD trials have failed, the chance of becoming seizure-free with a third drug is approximately 11% and declines further with each subsequent trial.2
The clinical evaluation required before confirming a diagnosis of drug-resistant epilepsy must rule out several conditions that mimic pharmacoresistance without representing true refractory epilepsy. The most common of these is incorrect epilepsy diagnosis – particularly the misdiagnosis of psychogenic non-epileptic seizures (PNES) as epileptic seizures. Up to 25% of patients referred to tertiary epilepsy centers with a diagnosis of drug-resistant epilepsy are ultimately found to have PNES as the primary or contributing diagnosis.3 Other pseudoresistance causes include incorrect ASD selection for the patient's actual seizure type (carbamazepine worsening absence or myoclonic seizures, for example), subtherapeutic dosing, poor medication adherence, and interactions reducing ASD levels. Video-electroencephalography (VEEG) monitoring during a clinical event is the definitive diagnostic tool for distinguishing epileptic from non-epileptic seizures and is a standard component of the presurgical evaluation.7
For a patient to meet the ILAE definition of drug-resistant epilepsy: (1) two ASDs must have been tried – further pharmacological trials before referral are not evidence-based; (2) each trial must have been adequate – appropriate drug for seizure type, adequate dose, adequate duration; and (3) the outcome measure must be seizure freedom – not merely reduction. When all three conditions are met, the patient should be referred for comprehensive epilepsy center evaluation, including VEEG monitoring, high-resolution magnetic resonance imaging (MRI), and surgical candidacy assessment. Delay in referral is the most common and consequential management error in DRE.
Understanding why ASDs fail in drug-resistant epilepsy is not merely an academic exercise – the mechanism of resistance determines which strategies have the best chance of overcoming it. Two major hypotheses have accumulated the most experimental and clinical support: the transporter hypothesis and the target hypothesis. These are not mutually exclusive and may operate simultaneously in the same patient.
The transporter hypothesis proposes that overexpression of drug efflux transporters at the blood-brain barrier reduces ASD penetration into epileptogenic brain tissue to subtherapeutic concentrations, even when plasma levels are adequate. The most studied transporter in this context is P-glycoprotein (P-gp), encoded by the ABCB1 gene and expressed on the luminal surface of brain capillary endothelial cells. P-glycoprotein actively pumps many ASDs – including phenytoin, carbamazepine, phenobarbital, and lamotrigine – out of the brain endothelium and back into the systemic circulation. In drug-resistant epilepsy, both animal models and human resection specimens have demonstrated upregulation of P-gp expression in and around the seizure focus, creating a pharmacokinetic sanctuary that is spatially localized to the area that most needs drug exposure. The clinical implication is that standard plasma-level monitoring does not capture the local brain tissue drug concentration, which may be substantially lower in the epileptogenic zone than in surrounding brain regions.4
The target hypothesis proposes that structural or functional changes in the molecular targets of ASDs reduce drug binding affinity and efficacy at those targets. The best-documented example is voltage-gated sodium channel modification in drug-resistant epilepsy. In patients with pharmacoresistant focal epilepsy, resected tissue has shown alterations in the expression and splicing of sodium channel subunits – particularly Nav1.1 (SCN1A), Nav1.2 (SCN2A), and Nav1.6 (SCN8A) – and in the inactivation kinetics of sodium channels in epileptogenic neurons. These changes reduce the sensitivity of the channel to sodium channel-blocking ASDs such as carbamazepine, phenytoin, and lamotrigine, whose mechanism depends on preferential binding to the inactivated state of the channel. When channels recover from inactivation more rapidly than normal, the use-dependent block that these drugs rely on becomes less effective. Similar target-level changes have been described for gamma-aminobutyric acid type A (GABA-A) receptors, where subunit composition shifts in epileptogenic tissue reduce sensitivity to benzodiazepines and barbiturates.5
Pharmacogenomics contributes a third dimension to the understanding of drug-resistant epilepsy. Genetic variants in ABCB1 (encoding P-gp) have been associated with differences in ASD response in several cohort studies, though clinical predictive value remains limited. Variants in cytochrome P450 (CYP) enzymes – particularly CYP2C9 and CYP2C19 – alter the metabolism of phenytoin and several other ASDs, leading to either subtherapeutic levels (rapid metabolizers) or toxicity (poor metabolizers) at standard doses. The SCN1A variant rs3812718 has been associated with differential response to carbamazepine and phenytoin by influencing sodium channel splicing. These pharmacogenomic associations support individualized ASD selection but have not yet translated into routine clinical genotyping algorithms in most centers.6
An important practical point from the mechanistic understanding of drug resistance is that adding multiple ASDs with the same mechanism of action provides little additional benefit in truly pharmacoresistant patients. If sodium channel blocking drugs have failed due to target-level changes in those channels, adding a second or third sodium channel blocker is unlikely to succeed. Rational polypharmacy in drug-resistant epilepsy should favor agents with complementary or distinct mechanisms – for example, combining a sodium channel blocker with a synaptic vesicle protein 2A (SV2A) ligand such as levetiracetam or brivaracetam, or with an alpha-2-delta (α2δ) subunit ligand such as gabapentin or pregabalin, rather than combining multiple sodium channel-targeting drugs.5
If drug resistance is driven primarily by P-glycoprotein overexpression (transporter hypothesis), strategies to circumvent it include using ASDs that are poor P-gp substrates (levetiracetam, valproate), or using P-gp inhibitors – though no clinically viable P-gp inhibitor has been validated for this indication. If resistance is driven by target-level changes (target hypothesis), the rational response is to switch to drugs that act at different targets. In practice, both mechanisms likely contribute in most patients with true pharmacoresistance, and the primary clinical response should be surgical evaluation rather than continued pharmacological trials.
Resective epilepsy surgery is the most effective treatment for drug-resistant focal epilepsy when a discrete, resectable epileptogenic zone can be localized. In properly selected patients, temporal lobe surgery achieves seizure freedom in 60–80% of cases – an outcome that no ASD regimen can approach in a truly pharmacoresistant patient. Yet the median time from diagnosis of drug-resistant epilepsy to surgical evaluation remains more than ten years in most healthcare systems, a delay that carries substantial cost in injury, cognitive decline, employment, and mortality.
The presurgical evaluation is a multidisciplinary process aimed at answering a single question: is there a discrete epileptogenic zone whose removal or disconnection will render the patient seizure-free without causing an unacceptable neurological deficit? The core components of this workup are prolonged video-electroencephalography (VEEG) monitoring to capture habitual seizures and localize their electrographic onset, high-resolution 3-Tesla magnetic resonance imaging (MRI) using epilepsy-specific protocols to identify a structural lesion, and neuropsychological testing to establish baseline cognitive function and predict functional risk from proposed resection. In patients where these studies are concordant – meaning the semiology, electroencephalography (EEG) onset, and MRI lesion all point to the same brain region – surgery can often be planned without additional invasive testing. When studies are discordant or the epileptogenic zone abuts eloquent cortex, additional tests are required: fluorodeoxyglucose positron emission tomography (FDG-PET), ictal single-photon emission computed tomography (SPECT), magnetoencephalography (MEG), and ultimately intracranial electrode implantation (stereo-EEG or subdural grids) for direct electrocortical recording and stimulation mapping.7
Temporal lobe epilepsy (TLE) with mesial temporal sclerosis (MTS) is the most common surgically treated syndrome and the one with the strongest evidence base. Randomized controlled trial data, including the landmark Wiebe trial in 2001, demonstrated that anterior temporal lobectomy produced seizure freedom in 58% of patients at one year versus 8% with continued medical therapy, with 38% versus 3% achieving freedom from seizures that impaired awareness. Long-term follow-up studies show that approximately 50–60% of TLE patients maintain seizure freedom at five to ten years after surgery.8 Neocortical resections for non-TLE focal epilepsy have lower success rates (30–50% seizure freedom), reflecting the greater difficulty in precisely delineating the epileptogenic zone when no discrete lesion is present on MRI. In pediatric patients with hemispheric epilepsy syndromes, hemispherotomy or functional hemispherectomy can achieve seizure freedom in 70–80% of cases, though the procedure is applicable only in patients with preexisting contralateral hemiplegia.7
Neuromodulation therapies are device-based treatments for patients who are not candidates for resective surgery – either because the epileptogenic zone cannot be localized, because it overlaps eloquent cortex, or because the patient declines resection. Vagus nerve stimulation (VNS) delivers intermittent electrical stimulation to the left vagus nerve via an implanted pulse generator. VNS reduces seizure frequency by more than 50% in approximately 50% of patients at two years, with responder rates improving over time and continuing to increase with years of therapy. VNS does not eliminate seizures in most patients but reduces their frequency and severity and improves postictal recovery time. Responsive neurostimulation (RNS; NeuroPace system) uses a closed-loop device that continuously monitors electrocortical activity via chronically implanted electrodes and delivers a brief burst of stimulation when a seizure is detected. In pivotal trials and long-term follow-up, RNS produced a median seizure reduction of approximately 53% at two years, increasing to approximately 75% at nine years – a progressive improvement not seen with open-loop devices. Deep brain stimulation (DBS) targeting the anterior nucleus of the thalamus (ANT-DBS) received FDA approval in 2018 for drug-resistant focal epilepsy based on the SANTE trial, which demonstrated a 40% median seizure reduction at three months, increasing to 69% median reduction at five years.9
Prospective studies show that in patients with drug-resistant epilepsy who eventually undergo surgery, the average delay from pharmacoresistance to surgical referral exceeds ten years. During this period, patients experience ongoing seizures, fall injuries, cognitive decline from repeated ictal and postictal states, loss of employment and driving privileges, and excess mortality from sudden unexpected death in epilepsy (SUDEP). Surgical evaluation is not a last resort – it is a medical necessity that should be triggered by the ILAE definition of DRE (failure of two adequate ASD trials), not by exhaustion of the pharmacological armamentarium.
The ketogenic diet (KD) is a high-fat, adequate-protein, very-low-carbohydrate dietary intervention that induces a state of nutritional ketosis producing sustained elevation of ketone body levels in blood and brain. Its antiseizure efficacy has been established in multiple randomized and prospective controlled trials, and it is recommended as a treatment option in drug-resistant epilepsy across all age groups, with strongest evidence in children and in patients with specific metabolic epilepsy syndromes where it is the treatment of choice rather than an adjunct.
The classic ketogenic diet provides calories in a ratio of 4 parts fat to 1 part combined protein and carbohydrate by weight – a ratio that generates sufficient ketone body production to sustain nutritional ketosis in most patients. The primary ketone bodies produced are beta-hydroxybutyrate (BHB) and acetoacetate (AcAc), with BHB predominating in blood. When circulating glucose is restricted and ketone bodies serve as the primary fuel for the brain, multiple antiseizure mechanisms are engaged: direct inhibition of voltage-gated sodium channels by polyunsaturated fatty acids (PUFAs) released during fat metabolism; enhancement of the adenosine triphosphate-sensitive potassium (K-ATP) channel open state, hyperpolarizing neurons and raising the seizure threshold; increased production of gamma-aminobutyric acid (GABA) from glutamate via increased transamination; inhibition of mechanistic target of rapamycin (mTOR) pathway activity; and mitochondrial biogenesis with improved energetic efficiency in neurons. No single mechanism fully explains the diet's efficacy, and the antiseizure effect appears to be multifactorial.10
Clinical evidence for the ketogenic diet comes from multiple prospective and randomized controlled trials. The most rigorous evidence is from a multicenter randomized controlled trial in children with drug-resistant epilepsy published by Neal et al. in 2008, which demonstrated that 38% of children assigned to the ketogenic diet achieved more than 50% seizure reduction at three months compared with 6% in the control group, with 7% achieving more than 90% reduction. Longer-term observational data show that approximately 50–55% of patients who initiate and maintain the diet for one year achieve more than 50% reduction in seizure frequency, and approximately 15% achieve seizure freedom. These outcomes are comparable to what can be expected from a third or fourth ASD trial in a truly drug-resistant patient, making the ketogenic diet a high-yield alternative or adjunct in that setting.11
The ketogenic diet is the treatment of choice – not merely an adjunct – in two specific metabolic conditions: glucose transporter type 1 (GLUT1) deficiency syndrome and pyruvate dehydrogenase (PDH) deficiency. In GLUT1 deficiency, the glucose transporter responsible for moving glucose across the blood-brain barrier is impaired, producing cerebral glucose deficiency. Ketone bodies cross the blood-brain barrier via monocarboxylate transporters independently of GLUT1, making the ketogenic diet the only intervention that provides adequate brain fuel in these patients. Failure to diagnose GLUT1 deficiency and institute ketogenic diet therapy causes progressive cognitive decline and pharmacoresistant epilepsy that responds specifically to dietary therapy. GLUT1 deficiency should be suspected in any child with early-onset epilepsy refractory to multiple ASDs, particularly if seizures are triggered by fasting or exercise.10
Several dietary variants of the classic ketogenic diet offer lower fat ratios and greater palatability with evidence of similar efficacy in select populations. The modified Atkins diet (MAD) uses a less restrictive fat-to-protein-plus-carbohydrate ratio (approximately 1:1 to 2:1), limits carbohydrates to 10–20 grams per day without strict fat requirements, and does not require calorie counting or weighing of food portions. The low glycemic index treatment (LGIT) uses a moderate fat ratio (approximately 60% of calories) and restricts carbohydrates to those with glycemic index below 50. Both MAD and LGIT have responder rates approaching those of the classic KD in pediatric and adult observational series. Adverse effects requiring monitoring across all variants include kidney stones (in 5–8%), dyslipidemia, growth impairment in children, selenium and carnitine deficiency, and the cardiomyopathy risk associated with selenium deficiency that requires supplementation.11
In GLUT1 deficiency syndrome, the ketogenic diet is not an adjunct – it is the primary treatment and should be initiated as soon as the diagnosis is confirmed, regardless of seizure burden. In pyruvate dehydrogenase deficiency, the ketogenic diet provides an alternate metabolic substrate and is similarly first-line. In Dravet syndrome and certain other developmental and epileptic encephalopathies, early initiation of the ketogenic diet produces better outcomes than waiting until multiple ASDs have been tried. When these conditions are suspected, metabolic and genetic evaluation should precede or proceed in parallel with ASD initiation, not wait until pharmacoresistance is established.
The genomic revolution in epilepsy has transformed the diagnostic landscape and is beginning to reshape treatment. In the past decade, next-generation sequencing has identified single-gene causes for hundreds of epilepsy syndromes previously classified as "cryptogenic," revealing specific pathogenic mechanisms amenable to mechanism-targeted intervention. The transition from symptomatic seizure suppression to disease-modifying or curative therapy is now underway for a small but growing number of genetic epilepsies.
Antisense oligonucleotides (ASOs) are short synthetic single-stranded nucleic acid sequences that bind to complementary messenger ribonucleic acid (mRNA) transcripts and alter their processing, stability, or translation. ASOs have emerged as a platform technology for epilepsies caused by gain-of-function mutations in ion channel genes, where the therapeutic goal is to reduce expression of the mutant allele. The most advanced clinical application is in SCN8A epilepsy (gain-of-function mutations in the Nav1.6-encoding SCN8A gene), which produces a severe early-onset epileptic encephalopathy that is uniformly resistant to standard ASDs. An ASO targeting SCN8A mRNA to reduce Nav1.6 protein expression reduced seizure frequency by more than 90% in animal models and has entered early-phase human clinical trials. A separate ASO approach targeting SCN1A – to upregulate Nav1.1 expression in Dravet syndrome – is also under clinical investigation, exploiting the observation that in most Dravet patients one allele is functionally normal and could be upregulated to compensate for the haploinsufficient allele.12
Adeno-associated virus (AAV)-based gene therapy offers the possibility of a permanent curative intervention for selected genetic epilepsies. AAV vectors can deliver functional copies of deficient genes, inhibitory RNA constructs, or gene-editing components directly into the brain parenchyma or cerebrospinal fluid (CSF). In preclinical models of Dravet syndrome, AAV-delivered SCN1A transgenes or inhibitory RNA constructs targeting the compensatory hyperpolarization-activated cyclic nucleotide-gated channel 1 (HCN1) have produced sustained reduction in seizures. The safety profile of AAV gene therapy in the central nervous system has been established in pediatric spinal muscular atrophy (SMA) trials, and CNS gene therapy trials for genetic epilepsies are anticipated or ongoing. Key challenges include ensuring adequate brain distribution of AAV vectors, long-term transgene expression stability, and avoiding immune responses against vector or transgene proteins.13
Mechanistic target of rapamycin (mTOR) inhibition represents the first approved disease-modifying therapy for a genetic epilepsy. Tuberous sclerosis complex (TSC) is caused by loss-of-function mutations in TSC1 or TSC2, the protein products of which normally inhibit mTOR complex 1 (mTORC1). Loss of TSC1/TSC2 function results in constitutive mTORC1 hyperactivation in neurons and glial cells, driving cortical tuber formation, aberrant synaptic connectivity, and epileptogenesis. Everolimus, an mTORC1 inhibitor, was approved by the FDA in 2018 for the adjunctive treatment of seizures associated with TSC in patients aged 2 years and older, based on the EXIST-3 trial. The EXIST-3 trial demonstrated median seizure reduction of 29.3% (low-exposure everolimus) and 39.6% (high-exposure everolimus) versus 14.9% for placebo, with responder rates (50% reduction) of 28.2% and 40.0% versus 15.1%. Everolimus also reduces the size of TSC-associated subependymal giant cell astrocytomas (SEGAs) and angiomyolipomas, making it a multi-target disease-modifying agent in TSC rather than a purely antiseizure drug.14
The broader landscape of genotype-directed epilepsy therapy is expanding rapidly. Quinidine, a sodium channel blocker used in cardiac arrhythmia, has shown preliminary efficacy in KCNT1 (potassium channel gene) gain-of-function epilepsy of infancy with migrating focal seizures. Memantine, an N-methyl-D-aspartate (NMDA) receptor antagonist, has been explored in GRIN2A and GRIN2B gain-of-function epilepsies driven by overactive glutamate receptors. Fenfluramine, originally approved for Dravet syndrome via serotonin modulation, has also shown activity in KCNQ2 neonatal epilepsy. Cannabidiol (CBD) acts via multiple mechanisms and has demonstrated efficacy beyond its original Dravet syndrome indication in several genetic epilepsy syndromes with mTOR pathway dysregulation. These examples illustrate the principle that identifying the precise genetic cause of a patient's epilepsy enables hypothesis-driven pharmacological targeting – a departure from the empirical trial-and-error approach that has defined epilepsy pharmacotherapy for the past century.15
Comprehensive genetic evaluation – including chromosomal microarray and epilepsy gene panel sequencing or whole-exome sequencing – is now recommended for all patients with drug-resistant epilepsy of unknown etiology, as well as for infants and children with any epileptic encephalopathy. A genetic diagnosis changes clinical management in a substantial proportion of patients: it may identify a precision therapy target (TSC/mTOR, SCN8A/ASO, GLUT1/ketogenic diet), reveal a contraindication to a specific drug class (sodium channel blockers in Dravet/SCN1A), or establish a prognosis and recurrence risk that guides family planning. The diagnostic yield of comprehensive genomic panels in drug-resistant epilepsy exceeds 30% in pediatric cohorts and 15–20% in adult cohorts.
Kwan P, Arzimanoglou A, Berg AT, et al. Definition of drug resistant epilepsy: consensus proposal by the ad hoc Task Force of the ILAE Commission on Therapeutic Strategies. Epilepsia. 2010;51(6):1069–1077.
doi:10.1111/j.1528-1167.2009.02397.xKwan P, Brodie MJ. Early identification of refractory epilepsy. N Engl J Med. 2000;342(5):314–319.
doi:10.1056/NEJM200002033420503Scheffer IE, Berkovic S, Capovilla G, et al. ILAE classification of the epilepsies: position paper of the ILAE Commission for Classification and Terminology. Epilepsia. 2017;58(4):512–521.
doi:10.1111/epi.13709Loscher W, Potschka H. Drug resistance in brain diseases and the role of drug efflux transporters. Nat Rev Neurosci. 2005;6(8):591–602.
doi:10.1038/nrn1728Remy S, Beck H. Molecular and cellular mechanisms of pharmacoresistance in epilepsy. Brain. 2006;129(1):18–35.
doi:10.1093/brain/awh682Zimprich F, Sunder-Plassmann R, Stogmann E, et al. Association of an ABCB1 gene haplotype with pharmacoresistance in temporal lobe epilepsy. Neurology. 2004;63(6):1087–1089.
doi:10.1212/01.WNL.0000141021.42763.F6Rosenow F, Luders H. Presurgical evaluation of epilepsy. Brain. 2001;124(9):1683–1700.
doi:10.1093/brain/124.9.1683Wiebe S, Blume WT, Girvin JP, Eliasziw M. A randomized, controlled trial of surgery for temporal-lobe epilepsy. N Engl J Med. 2001;345(5):311–318.
doi:10.1056/NEJM200108023450501Salanova V, Witt T, Worth R, et al. Long-term efficacy and safety of thalamic stimulation for drug-resistant partial epilepsy. Neurology. 2015;84(10):1017–1025.
doi:10.1212/WNL.0000000000001334Rho JM. How does the ketogenic diet induce anti-seizure effects? Neurosci Lett. 2017;637:4–10.
doi:10.1016/j.neulet.2015.07.034Neal EG, Chaffe H, Schwartz RH, et al. The ketogenic diet for the treatment of childhood epilepsy: a randomised controlled trial. Lancet Neurol. 2008;7(6):500–506.
doi:10.1016/S1474-4422(08)70092-9García Quilón P, Volpedo G, Cappato S, et al. Antisense oligonucleotides as a precision therapy for developmental and epileptic encephalopathies. CNS Neurosci Ther. 2024;30(11):e70050.
doi:10.1111/cns.70050Taher N, McKinney C, Patel Z, et al. Gene therapy approaches for epilepsy. Front Mol Neurosci. 2022;15:806516.
doi:10.3389/fnmol.2022.806516French JA, Lawson JA, Yapici Z, 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:10.1016/S0140-6736(16)31419-2Oyrer J, Maljevic S, Scheffer IE, Berkovic SF, Petrou S, Reid CA. Ion channels in genetic epilepsy: from genes and mechanisms to disease-targeted therapies. Pharmacol Rev. 2018;70(1):142–173.
doi:10.1124/pr.117.014456