Drug resistance in idiopathic generalized epilepsies: Evidence and concepts

Abstract

Although approximately 10%–15% of patients with idiopathic generalized epilepsy (IGE)/genetic generalized epilepsy remain drug‐resistant, there is no consensus or established concept regarding the underlying mechanisms and prevalence. This review summarizes the recent data and the current hypotheses on mechanisms that may contribute to drug‐resistant IGE. A literature search was conducted in PubMed and Embase for studies on mechanisms of drug resistance published since 1980. The literature shows neither consensus on the definition nor a widely accepted model to explain drug resistance in IGE or one of its subsyndromes. Large‐scale genetic studies have failed to identify distinct genetic causes or affected genes involved in pharmacokinetics. We found clinical and experimental evidence in support of four hypotheses: (1) “network hypothesis”—the degree of drug resistance in IGE reflects the severity of cortical network alterations, (2) “minor focal lesion in a predisposed brain hypothesis”—minor cortical lesions are important for drug resistance, (3) “interneuron hypothesis”—impaired functioning of γ‐aminobutyric acidergic interneurons contributes to drug resistance, and (4) “changes in drug kinetics”—genetically impaired kinetics of antiseizure medication (ASM) reduce the effectiveness of available ASMs. In summary, the exact definition and cause of drug resistance in IGE is unknown. However, published evidence suggests four different mechanisms that may warrant further investigation.

Key Points.

• There is no consensus on the definition of drug resistance in IGE

• Experimental data do not consistently explain drug resistance in IGE

• Four different mechanisms may contribute to drug resistance

• Altered networks, minor cortical lesions, impaired interneurons, and/or drug kinetics are putative mechanisms involved

1. INTRODUCTION

According to the current International League Against Epilepsy (ILAE) classification of epilepsies, idiopathic generalized epilepsies (IGEs; also called genetic generalized epilepsies) comprise four different clinical presentations referred to as childhood absence epilepsy (CAE), juvenile absence epilepsy (JAE), juvenile myoclonic epilepsy (JME), and epilepsy with generalized tonic–clonic seizures alone (EGTCS) in adults. ¹ These four syndromes may lie on a clinical spectrum, which could explain the difficulties in identifying distinct diagnostic criteria for them. ² In contrast, the diagnosis of classical IGE cases is usually unproblematic and is based on a combination of generalized seizures, electroencephalogram (EEG) with generalized spike–wave discharges (GSWDs)/polyspike–wave discharges, normal cognitive function, and normal brain imaging. Atypical patients remain a diagnostic challenge. ¹ There is less consensus about which symptoms are incompatible with the diagnosis of IGE, and there is an ongoing discussion about the degree of intellectual disability that definitively excludes the diagnosis of IGE, whether focal onset seizures may be part of the spectrum, and whether resistance to antiseizure medication (ASM) is compatible with an IGE diagnosis. ³

The combination of rhythmic spike‐wave discharges and generalized seizures is not unique to IGE, and there is a clinical overlap between monogenetic epileptic encephalopathies due to Gamma‐aminobutyric acid receptor subunit alpha or Glut1. ⁴ , ⁵ In addition, focal lesions close to the frontal callosal fibers (e.g., cortical dysplasias) are associated with EEG patterns similar to those in IGE, and metabolic disorders (e.g., Unverricht–Lundborg disease) and distinct chromosomal alterations (e.g., 16q13.11 or 15q13.3) may mimic IGE. ⁶ , ⁷ Thus, a part of ASM‐resistant patients may represent “borderline patients,” with a substantial variability between IGE cohorts.

High heritability of IGE is widely accepted and proven by, for example, twin studies, but Mendelian inheritance is rare. ¹ Penetrance of epileptic seizures is variable, and the proportion of close relatives with asymptomatic EEG patterns associated with IGE is high and substantially exceeds the proportion of relatives with seizures. ⁸ Monogenetic causes of IGE are very rare and explain only a few percent of patient cases. ¹ Over the past decade, large‐scale genetic studies have provided a new model explaining the genetics of IGE. Genome‐wide association studies have revealed numerous single nucleotide polymorphisms significantly associated with IGE with a low effect size. According to this model, the cumulative and possibly synergistic effect of the common genetic variants explains a substantial proportion of the heritability. A recent meta‐analysis estimated that 46%–58% of heritability was explained by the polymorphisms identified so far. ⁹

Despite this undisputable progress, the exact pathophysiology of IGE remains unknown. A thalamic involvement in the generation of rhythmic spike–wave discharges (the most prominent and defining feature of IGE) has been discussed since the 1940s, and this hypothesis was finally proven in humans using combined functional magnetic resonance imaging (MRI) and EEG recordings. Gotman et al. ¹⁰ showed that spike–wave discharges were preceded by thalamic overactivation followed by decreased cortical activation. This initial study was complemented by a series of studies mainly performed in JME patients that confirmed the involvement of various parts of the brain. However, irrespective of the methodology used, MRI, magnetic resonance spectroscopy, voxel‐based volumetric studies, and diffusion tensor imaging indicate an involvement of at least (1) the anterior thalamus and (2) the supplementary motor areas and associated white matter tracts. ¹¹

This led to the understanding of IGEs as network or system disorders, ¹² implying additional symptoms beyond seizures. The recognized phenotypic spectrum of IGE (also referred to as endophenotype) has broadened and distinct neuropsychological and neurophysiological phenotypes have been described in recent years. ¹³ , ¹⁴

The processes that link the identified polymorphisms to the detectable changes in cerebral networks are still essentially unknown. A recent study based on two large independent datasets provided the first evidence of a shared genetic basis of IGE and background EEG oscillations reflecting corticothalamic functioning. ¹⁵ A similar study on distinct EEG features such as spike–wave discharges is still pending.

Although there is a certain consensus on the genetic mechanisms causing IGE and the anatomical location of the cerebral networks involved, the exact mechanisms triggering seizures and eliciting the different phenotypes seen in patients are unknown. An unsolved question is whether the different IGEs comprise a disease spectrum with variable phenotypes or whether they are distinct entities with distinct etiology but sharing symptoms of thalamocortical dysfunction. ¹ , ²

The lack of knowledge and consensus on the etiology of the different syndromes classified as IGE relates directly to the lack of knowledge and consensus on the etiology of drug‐resistant IGE. Despite the availability of a wide range of ASMs, drug‐resistant IGE is a substantial clinical and socioeconomic challenge, with increased risk of sudden unexplained death in epilepsy. ¹⁶ , ¹⁷ Comparability of studies is further impaired by variable definitions of drug resistance, the higher efficacy of valproic acid (VPA) compared to other adequately chosen ASMs, and the challenge of pseudoresistance. ¹⁸

A substantial number of studies have provided data and hypotheses about the possible mechanisms involved in drug resistance. In this paper, therefore, we report on a comprehensive review of the literature with the aim of summarizing the available evidence on drug‐resistant IGE and describing common themes and concepts.

2. MATERIALS AND METHODS

We searched for articles describing studies investigating drug‐resistant IGE and possible causalities in PubMed and Embase using the following search terms: idiopathic generalized epilepsy OR genetic generalized epilepsy OR juvenile absence epilepsy OR juvenile myoclonic epilepsy OR absence epilepsy OR absence epilepsy, childhood OR absence epilepsies, childhood OR absence epilepsies, juvenile OR absence epilepsies, myoclonic OR epilepsy with generalized tonic–clonic seizures alone AND drug resistance OR drug refractory OR treatment resistant OR medically refractory OR medically unresponsive OR uncontrolled OR refractory OR resistant OR treatment refractory OR unresponsive OR medically resistant. Reference lists were checked for additional eligible articles.

The last search was performed on April 1, 2022, and search results for the definition of drug resistance are given a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta‐Analyses)‐compliant flowchart (Figure S1). ¹⁹ A less strict selection of papers was made for the narrative parts of the review; included papers are given in Table S1. The search included retrospective or prospective studies on possible mechanisms of drug resistance in IGE, patient studies suggestive of causality in drug‐resistant and drug‐responsive IGE, and reviews looking at possible mechanisms of treatment resistance. Studies were included irrespective of the definition of drug resistance used. Each author's (variable) definition of drug resistance was used and is referred to as “drug resistance” in this paper.

Acceptable epilepsy subsyndromes were CAE, JAE, JME, EGTCS, or IGE (alone or in combination). Studies including patients with possible Lennox–Gastaut syndrome (LGS) or epileptic encephalopathies were excluded. Only articles in English with an available abstract and published since 1980 were included.

We excluded animal studies, case reports, comments, and conference abstracts as well as studies with exclusively pediatric patients and studies without a clear definition of treatment resistance or solely studying unresponsive or uncontrolled but not drug‐resistant IGE. All search results were reviewed by title and abstract, and full‐length texts were reviewed in eligible articles.

3. RESULTS

3.1. Definition of drug resistance

The literature search yielded 1021 results, of which 55 studies were included in the final analyses (Table 1, Figure S1). An additional six studies were identified from cross‐references. Drug resistance was defined according to the ILAE ²⁰ in 20 of the 61 studies. The remainder used variable observation periods for treatment response and different definitions of drug resistance based on observation periods of different length and different numbers and combinations of antiepileptic drugs (AEDs); in some cases, lack of response to VPA was considered essential for drug resistance (Table 1). In summary, we found no consensus on the optimal definition of drug resistance in IGE.

TABLE 1.

Definitions of drug resistance used in studies found, divided into before and after the publication of the ILAE definition (reviews not included)

Definition of drug resistance used in the articles	Studies published before ILAE definition (number of drug‐resistant patients [%])	Studies published after ILAE definition (number of drug‐resistant patients [%])
ILAE		Baheti [1] (116 [100%]) Pietrafusa [2] (21 [34.3%]) Cerulli Irelli [3] (29 [14.6%]) Choi [4] (121 [20.5%]) Gesche [5] (50 [12.3%]) Cação [6] (121 [50.4%]) Gesche [7] (33 [27.5%]) Gomez‐Ibañez [8] (122 [46.3%]) Shakeshaft [9] (165 [21.6%]) Wolking [10] (154 [17.2%]) Conrad [11] (31 [30.1%]) Jensen [12] (18 [11.1%]) Pegg [13] (23 [69.7%]) McKavanagh [14] (23 [69.7%]) Li [15] (17 [35.4%]) Wang [16] (12 [38.7%]) Gesche [17] (72 [52.6%]) Badawy [18] (49 [71.0%]) Badawy [19] (14 [23.0%]) Cerulli Irelli [20] (69 [34.2%])
ILAE, valproic acid required as one of the failed ASMs		Sun [21] (25 [29.4%]) Pegg [22] (18 [48.6%])
ILAE, required time period of seizure freedom specified differently than by the ILAE		Karakis [23] (25 [49.0%])
Number of ASMs tried for a defined time period		Kamitaki [24] (118 [50.9%]) Pegg [25] (18 [48.6%])
Failure of valproate	Szaflarski [26] (45 [16.9%]) Baykan [27] (8 [16.7%]) Gelisse [28] (24 [17.1%]) Fernando‐Dongas [29] (10 [30.3%])	Paiva [30] (20 [20.4%]) Jayalakshmi [31] (31 [16.0%]) Kay [32] (13 [26.0%]) Szaflarski [33] (22 [36.7%])
Failure of a given number of ASMs/defined number of seizures in a defined time period	Kwan [34] (11 [15.3%]) Jayalakshmi [35] (43 [16.2%])	Qu [36] (204 [42.2%]) Qu [37] (204 [42.2%]) Qu [38] 204 [42.1%]) Balan [39] (201 [100%]) Balan [40] (201 [100%]) Wu [41] (204 [42.2%]) Guo [42] (204 [42.2%]) Özek [43] (10 [100%])
Lack of seizure control for defined time period	Dasheiff [44] (12 [100%]) Mohanraj [45] (27 [26.2%])	Gürer [46] (83 [38.6%]) Guaranha [47] (25 [38.5%]) Lobato [48] (18 [45.0%]) Kay [49] (23 [31.9%]) Bolden [50] (14 [46.7%])
Failure of specific drug regimen or named drugs	Wolf [51] (41 [17.9%]) Létourneau [52] (11 [39.3%])	Wolking [53] (775 IGE patients)
Lack of seizure control despite “adequate treatment”	Szaflarski [54] (9 [100%])	Arntsen [55] (19 [47.5%]) Szaflarski [56] (23 [100%])

Note: The complete citation of each article is given in the supplements. Please note that the numbers in brackets after author names refer to the numbers in the Supporting Information list of references.

Abbreviations: ASM, antiseizure medication; ILAE, International League Against Epilepsy.

3.2. Clinical factors associated with drug resistance

Five reviews (including one meta‐analysis and 21 studies) examined clinical factors associated with treatment resistance. Twelve studies examined JME patients only, whereas eight studies examined all subsyndromes of IGE. Clinical characteristics associated with refractory JME were recently studied in two narrative reviews and in one well‐designed meta‐analysis; two reviews included other IGEs as well. ²¹ , ²² , ²³ , ²⁴ , ²⁵ In line with previous reports and reviews, ²¹ , ²⁴ the meta‐analysis confirmed the following risk factors for drug‐resistant epilepsy: absence seizures, young age at onset, presence of all three seizure types (absences, myoclonic seizures, and generalized tonic–clonic seizures), psychiatric comorbidity, praxis‐induced seizures, and CAE developing to JME. ¹⁸ , ²¹ , ²⁶

More recent studies added catamenial seizures, status epilepticus, and resistance to VPA to the list of clinical characteristics associated with drug resistance and confirmed the association between previous febrile seizures and treatment outcome. ²⁶ , ²⁷ , ²⁸ Some authors argued that atypical seizures or atypical seizure components and the number of generalized tonic–clonic seizures are relevant to the development of drug resistance. ²¹ , ²⁶

The meta‐analysis did not confirm female sex and family history of epilepsy as postulated risk factors for drug resistance as suggested by a recent large cohort study. ²⁸ , ²⁹ However, a very recent study on JME patients showed a sex‐dependent role of seizure precipitants for drug resistance. ³⁰ Furthermore, systemic disorders such as thyroid disease may modulate the seizure threshold and thereby trigger breakthrough seizures. ²¹ Despite the heterogeneity of the studies, risk factors for drug resistance did not differ between cohorts that included all IGE subsyndromes and cohorts that included JME patients only (Table 2).

TABLE 2.

Clinical risk factors for treatment resistance in cohorts with JME patients only and cohorts including all IGEs

Study	All IGE/other IGEs								JME only
	Monhanraj et al. [45]	Gesche et al. [7]	Cerulli Irelli et al. [3]	Gomez‐Ibañez et al. [8]	Wolf et al. [51]	Kamitaki et al. [24]	Szaflarski et al. [26]	Choi et al. [4]	Dasheiff et al. [44]	Fernando‐Dongas et al. [29]	Baykan et al. [27]	Pietrafusa et al. [2]	Guaranha et al. [47]	Paiva et al. [30]	Baheti et al. [1]	Gelisse et al. [28]	Jayalakshmi et al. [31]	Gürer et al. [46]	Cação et al. [6]	Shakeshaft et al. [9]
N	103	137	199	227	229	232	267	655	12	33	48	61	65	112	116	155	201	215	240	765
Risk factors studied
Absences		x			x	x						x						x		x
Age at onset		x		x	x								x						x
All three seizure types			x			x		x					x			x
Psychiatric comorbidity		x		x				x			x	x	x	x		x	x	x	x
Praxis induction													x
History of childhood absence epilepsy					x
Catamenial seizures						x		x												x
Febrile seizures	x		x
History of status epilepticus				x	x													x
Atypical features					x					x									x
Intellectual difficulties					x					x
Number of generalized tonic–clonic seizures					x							x
Electroencephalographic features			x	x	x	x	x		x	x		x	x				x
Resistance to valproic acid		x	x				x											x
No clinical features															x
JME diagnosis			x
Abuse		x
Somatic disorders											x							x
Intellectual deficiency										x

Note: The complete citation of each article is given in the supplements. Please note that the numbers in the brackets refer to the numbers in the Supporting Information list of references.

Abbreviations: IGE, idiopathic generalized epilepsy; JME, juvenile myoclonic epilepsy.

3.3. Coexistence of focal and generalized epilepsy

There is a clinical overlap in the seizure semiology of primary generalized tonic–clonic seizures and focal to bilateral tonic–clonic seizures. ²¹ Absences and myoclonic seizures can be difficult to distinguish clinically from focal seizures with or without impaired consciousness, ³¹ and focal seizure symptoms are reported by a substantial proportion of patients with established IGE. ¹² However, unambiguous motor symptoms (e.g., focal stereotyped unilateral clonic or dystonic involvement of limb) indicative for focal seizures remained rare. Conversely, asymmetric motor features like head deviation or myoclonus regularly occur in IGE. ³ , ³² Two retrospective case series reported that <1% of all IGE patients develop focal epilepsy later in life, and two patients underwent successful epilepsy surgery. ³³ , ³⁴ Of the two case series describing drug resistance, Usui et al. reported drug resistance in 15 of 26 patients, and Gelisse et al. found no association based on 10 JME patients. ³⁵ , ³⁶ The low number of cases and the likely selection bias of specialized epilepsy centers challenge generalizability, and the available data neither confirm nor refute an association between drug resistance and coexistence of focal and generalized epilepsy.

3.4. EEG features associated with drug resistance

Numerous publications studied associations of EEG features and seizure outcome. However, only nine EEG studies specifically examined treatment resistance in IGE patients, and three of these included only JME patients. The current literature on focal EEG features of IGE is reviewed elsewhere. ³ Associations between EEG features and drug resistance were reported in 10 clinical cohorts. Five different EEG features were reported as being prognostic for drug response. Evidence for an association of the epileptic discharge burden and drug resistance was low. ³⁷

3.4.1. Focal EEG asymmetries

Asymmetries in generalized paroxysms and in focal EEG abnormalities (such as focal slowing and focal discharges) are common in IGE and can lead to diagnostic difficulties, especially when focal features are reported as part of the seizure semiology. ²¹ Their relationship with drug resistance has been debated, and there is no consensus on whether asymmetries or focality on EEG are associated with drug resistance, with some studies finding no association and others finding significantly increased odds of drug resistance in patients with EEG asymmetries. This has led to theories about drug‐resistant patients having different localization of seizure generators than drug responders. ³⁸ Szaflarski et al. ³⁹ , ⁴⁰ supplemented EEG studies with functional MRI (fMRI) and found increasing frontal blood oxygen level‐dependent correlates of GSWD generators, indicating that drug‐resistant patients have cortical and not subcortical seizure generators.

3.4.2. Generalized polyspike trains (GPTs)

Three studies confirmed the prognostic value of generalized polyspike trains (GPTs)—a feature usually associated with LGS, which was first described by Sun et al. in a cohort of patients referred for long‐term EEG monitoring, and confirmed by Cerulli Irelli et al. and Kamitaki et al. ⁴¹ , ⁴² , ⁴³ Conrad et al. complemented these studies by showing that GPTs were twice as often found in sleep than in wake EEGs, stressing the need for prolonged EEG recordings (>24 h) to capture GPTs, which is in keeping with findings by Jensen et al., showing that GPTs are rare on routine EEGs. ⁴⁴

3.4.3. Prolonged epileptiform runs

Arntsen et al. studied the association between prolonged epileptiform runs of ≥3 s (whether provoked or unprovoked) and the clinical course of Norwegian JME patients and found an association with persisting seizures, VPA resistance, and the occurrence of absence seizures. ⁴⁵ The association with drug resistance was confirmed by Jensen et al. ⁴⁴

3.4.4. Generalized paroxysmal fast activity

Generalized paroxysmal fast activity (GPFA) is traditionally linked to LGS but is seen in 5.6%–18.9% of patients with otherwise typical IGE with absence seizures. ⁴⁶ , ⁴⁷ In several large independent cohorts, GPFA has been linked to persistent seizures and drug resistance in both children and adults. ⁴² , ⁴⁷

3.4.5. Photoparoxysmal response

Photoparoxysmal response (PPR) may be present in various conditions including epilepsy with eyelid myoclonia (Jeavons syndrome) and JME. ⁴⁸ It shows an autosomal‐dominant inheritance and is seen in 3%–4% of the general population. ⁴⁹ PPR shows a female preponderance and maximal penetrance in the young. ⁴⁹ In JME, PPR may be associated with favorable treatment response. ³⁰ The association between PPR and favorable treatment response is in contrast to the association of other seizure triggers such as catamenial epilepsy or praxis ²⁸ , ³⁰ and is likely explained by involvement of different cortical networks as shown by fMRI/EEG. ³⁰ How these different networks interfere with response to treatment is not yet known.

3.5. Network changes

Based on the idea of IGE being a network disease, several MRI studies have examined structural and network changes in IGE patients. ¹¹ IGE patients and their healthy siblings have altered networks and coactivation of networks that are not found in controls, indicating a genetic etiology. ⁵⁰ Several groups have studied the functional connectivity of separate brain regions and their coactivation patterns. Pegg et al. studied interictal EEG functional network topology ⁵¹ , ⁵² complemented with fMRI. ⁵³ IGE patients had very regular network topology that may make the networks more vulnerable to synchronicity and hence seizures. ⁵¹ , ⁵³ However, the authors could not identify network topologies associated with seizure control. Using a different definition of drug resistance (VPA resistance), Kay et al. found a correlation between resting state functional connectivity and GSWD frequency and a negative correlation between disease duration and default mode network connectivity, indicating that the brain changes associated with lack of seizure control also lead to attention deficits and executive dysfunction. ⁵⁴ , ⁵⁵ Another study focusing on EGTCS patients revealed distinct functional connectivity changes in different brain areas without compensatory enhancement in drug‐resistant patients as opposed to drug‐sensitive patients. Again, functional connectivity was negatively correlated with disease duration. ⁵⁶ Vollmar et al. ⁵⁷ complemented these data by showing that the extent of motor cortex coactivation correlated with disease activity and ASM treatment, indicating that more extensive network changes may be associated with poor seizure control.

Lobato et al. ⁵⁸ examined structural changes using diffusion tensor imaging and found diffuse changes in white matter bundles and tracts in IGE patients compared to controls, but none of these changes was associated with seizure status in 40 IGE patients. In line with these findings, Szaflarski et al. ⁵⁹ found no differences in the white matter of VPA‐resistant and nonresistant patients. In contrast, McKavanagh et al. ⁶⁰ described bihemispheric structural and network alterations that differed between drug‐resistant and nonresistant patients. Li et al. ⁶¹ also found differences between drug‐resistant and nonresistant patients, where atrophy of the basolateral region of the left amygdala was associated with drug resistance. In summary, no reproducible structural changes have yet been identified in drug‐resistant patients.

3.6. Transcranial magnetic stimulation studies

Cortical excitability differs between IGE patients and healthy population controls. Patients with IGE show increased excitability, and there is evidence that cortical excitability varies during the day, with hormonal changes, and with ASM. ⁶² Transcranial magnetic stimulation (TMS) has been suggested as a useful treatment option in drug‐resistant IGEs. ⁶³ Classical TMS studies indicate impaired early intracortical inhibition in JME patients when compared to controls. ⁶² Impaired γ‐aminobutyric acid type A (GABA_A) receptor signaling as a main mechanism is supported by in vivo studies of patients with GABRG2(R43Q) ⁶² mutations and by pharmacological studies. ⁶⁴ Badawy et al. ⁶⁵ found in TMS studies that patients with JME had higher cortical excitability than controls and patients with JAE or EGTCS. Normalization of early intracortical inhibition was associated with ASM response, whereas impairment of early intracortical inhibition remained unchanged in patients with drug‐resistant IGE. Furthermore, changes in cortical excitability were associated with treatment response.

In a 3‐year longitudinal TMS study, Badawy et al. ⁶⁶ found that cortical excitability increased over time in treatment‐resistant patients. Among drug‐naïve patients, cortical excitability did not differ between seizure‐free patients and those who developed treatment resistance.

Chowdhury et al. ¹³ described a completely new TMS endophenotype associated with IGE that was characterized by polyphasia of motor evoked potentials in the hand. This phenotype was reproduced but was not associated with treatment response, and its significance remains unclear. ⁶⁷ , ⁶⁸

3.7. Neuropsychological features

The network changes in the IGEs produce distinct neuropsychological profiles that include executive dysfunction and distinct personality traits involving high scores on harm avoidance, novelty‐seeking behavior, and impulsiveness that may represent an endophenotype. ¹² , ⁶⁹ However, only few studies have focused on treatment resistance, and most are impaired by an ASM effect on frontal lobe function. ⁷⁰ A small American study comparing patients with drug‐resistant temporal lobe epilepsy and patients with IGE observed lower intelligence scores in IGE patients. ⁷¹ Walsh et al. and Thomas et al. studied drug‐resistant JME patients and found impairments in intelligence, executive function, and memory that were related to age at epilepsy onset, current number of ASMs, abnormal personality traits, and psychiatric disorders. ⁷² , ⁷³ Valente et al. ⁷⁴ reported that poorer seizure control on VPA was associated with worse cognitive performance and higher expression of impulsive traits. Combining neuropsychological tests with TMS, Bolden et al. ⁷⁵ found that significantly poorer performance in attentional tasks in treatment‐resistant IGE patients was significantly associated with higher cortical excitability. In summary, distinct cognitive deficits are probably part of the endophenotype of IGE. It remains unclear whether the described associations between neuropsychological features and drug resistance reflect a common endophenotype or are secondary to ASM or the consequences of having drug‐resistant epilepsy. ⁷⁶

3.8. Genetic studies

Different genes targeting either receptors or channels possibly involved in seizure propagation or ASM response have been studied. Investigations of genes related to sodium and potassium channels in a large Chinese cohort of IGE patients have not yielded any association with treatment resistance, ⁷⁷ , ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸¹ and studies in India have shown no significant associations between drug resistance and ASM transporters and receptors. ⁸² , ⁸³ Finally, genome‐wide association studies on IGE patients of European descent coupled with clinical data on treatment response to the most widely used ASMs (valproic acid, levetiracetam, lamotrigine) found a minimal enrichment of missense and truncating variants of genes involved in pharmacokinetics of VPA and levetiracetam in drug‐resistant patients, indicating a potential genetic cause of drug resistance. The clinical significance of the findings remains unknown, however. ⁸⁴ , ⁸⁵ Evidence in support of the drug transporter hypothesis or the target theory remains sparse. ⁸⁶

4. DISCUSSION

The aim of this paper was to present a concise overview of current research on drug‐resistant IGE. Although there is neither consensus on the clinically most useful diagnostic criteria for treatment resistance nor conclusive evidence on its causes, the available data appear to support four hypotheses on treatment resistance in IGE. The condensation of the available data to these four testable hypotheses may help to trigger and inspire future research, but we wished to avoid extensive reflections over individual concepts, given the paucity of available data. Furthermore, it is currently unknown whether these postulated mechanisms for drug resistance are complementary, mutually exclusive, or synergistic.

4.1. Network hypothesis

According to the network hypothesis, drug‐resistant patients have more severe network alterations than drug responders. Although the exact mechanisms of thalamic nuclei involvement are not fully elucidated, the thalamocortical networks are likely crucial, given the thalamus' significant role in the generation and propagation of spikes and waves. ⁸⁷ , ⁸⁸ The network hypothesis is supported by evidence from fMRI, EEG, and (with limitations) neuropsychological studies that suggests altered functional connectivity in IGE patients compared to healthy controls, indicating that IGEs are network diseases. ¹¹ Different resting state networks have been studied, and drug‐resistant patients seem to be more affected by alterations than drug responders. ⁵⁷ , ⁶⁰ , ⁶¹ This makes their brain more susceptible to seizures and explains the praxis induction and cognitive and psychiatric problems seen in IGE patients. ⁵⁴ , ⁵⁵ , ⁵⁶ Network alterations in drug‐resistant patients, apart from giving possible explanations for drug resistance, may lead the way for new treatment options, like deep brain stimulation, previously not used in IGE patients. ⁸⁹ However, it remains unclear how and which networks are altered by ASMs, seizure frequency, and disease duration. Given that various research groups have reproduced a negative association between network changes and disease duration, it remains unclear whether treatment resistance is a result of or the cause of network changes. ⁹⁰ Studies proving a direct association of the extent of, e.g., corticothalamic connections and treatment response would support this hypothesis, whereas studies showing that clinical or imaging correlates are not linked to treatment response may falsify it.

4.2. Minor focal lesion in a predisposed brain hypothesis

This hypothesis proposes that the crucial factor for drug resistance is a focal lesion in a brain predisposed to IGE. This concept is supported by various but conflicting reports on focal asymmetries and spikes in the EEGs of severely affected patients with focal and generalized epilepsies. ²¹ , ³⁹ If correct, it has important implications for the management of drug‐resistant IGE patients, for example, by enabling the selection of patients with generalized epilepsy who would be amenable for epilepsy surgery. Validation or rejection of this hypothesis would also contribute to the discussion of IGE as a spectrum of diseases and its potential relationships to other generalized epilepsies and epileptic encephalopathies. ¹ , ² Machine learning algorithms applied to large‐scale imaging databases (e.g., ENIGMA database) or EEG samples may provide crucial data in support of this hypothesis.

4.3. Interneuron hypothesis

According to the interneuron hypothesis, persistent impaired GABAergic signaling of interneurons despite pharmacological treatment contributes to drug resistance in IGE. This concept is essentially based on the lack of normalization of intracortical inhibition in TMS studies in drug‐resistant patients compared to responders. ⁶² The model is, however, further supported by genetic data that indicate a crucial role of GABA receptor mutations in the development of IGE and epileptic encephalopathies. In view of motor evoked potential studies suggesting that cortical excitability varies throughout the day and a study indicating hormone dependence, this model may explain several clinical phenomena associated with drug resistance (e.g., its association with catamenial seizures). ⁶² If correct, the model would allow rapid identification and monitoring of drug‐resistant patients by TMS. However, it does not fully explain whether the relationship between seizures and cortical excitability is causal or consequential and why seizures tend to subside with age. ⁹⁰ Advanced imaging modalities, animal experiments, and a better understanding of the clinical phenotype of distinct GABA receptor mutations may further contribute to supporting or refuting this hypothesis.

4.4. Changes in drug kinetics

The idea of a purely genetic cause of treatment resistance with altered ASM kinetics goes hand in hand with the concept of IGEs being genetic. It is the most obvious explanation for drug resistance, but extensive genetic studies have so far failed to identify a distinct genetic cause. Studies have identified mutations possibly associated with changes in ASM kinetics, but these mutations explained drug resistance in only a minority of patients. ⁸⁴ , ⁸⁵ Hopefully, further progress in our understanding of the genetic and epigenetic causes of IGE will result in a better understanding of drug resistance.

4.5. Drug resistance is not compatible with the diagnosis of IGE

Importantly, none of the experimental studies included in this review questioned whether drug resistance is compatible with an IGE diagnosis. Diagnosing “borderline” patients with asymmetrical clinical or EEG features is challenging, and these “atypical” patients represent a substantial portion of patients with drug resistance. In keeping with this, drug‐resistant patients have characteristic clinical features like absence seizures, younger age at onset, catamenial seizures, and psychiatric comorbidity. The concept that “good drug response” characterizes at least JME patients was included in the 1989 ILAE classification but not in the updated version in 2022, in which the existence of drug‐resistant seizures is acknowledged. ¹ The main argument to diagnose drug‐resistant IGE patients as IGE is the lack of a more plausible diagnosis. ¹ , ⁹¹ Currently, clinical, imaging, and genetic studies do not provide consistent data supporting the idea of drug‐resistant IGE being substantially different from bona fide IGE.

Although the attention and number of studies on drug resistance have increased since the publication of the ILAE definition in 2010, ²⁰ the current guidelines refer to epilepsies in general and do not consider different epilepsy types. Hence, the ILAE definition is not widely used in the field (Table 1), and there is still no consensus on the best definition of drug resistance in IGEs. ¹⁸ Furthermore, most studies have focused on JME, potentially due to the higher prevalence, resulting in sparse information on the other subsyndromes, such as adult patients with drug‐resistant absence epilepsy. The lack of consensus on subsyndromes and whether they represent distinct entities further challenges the comparability of studies and the generalizability of findings. ¹ , ²

Prerequisite for finding the cause(s) of drug resistance in IGE is an established consensus on the definition of drug resistance in IGE. This will allow similar identification of patients, increase reproducibility of crucial findings, and thus provide solid ground for future research. A pragmatic approach may be the use of valproate resistance as proxy for drug resistance in IGE. ¹⁸

One important step was the recent publication of diagnostic criteria for the IGEs. A next step may be widening research to encompass all subsyndromes and inclusion of patients not followed at tertiary centers. This will allow comparing drug‐resistant patients with the entire spectrum of patients with rapid response to low ASM doses. Collaborative multicenter studies are necessary to achieve high patient numbers for studies on drug resistance in IGE ⁹² and will allow gathering a sufficient number of patients required for, for example, machine learning approaches and identification of yet unknown patterns of resistance within IGE cohorts.

In conclusion, the exact cause of drug resistance in IGE is unknown. However, published evidence suggests four different mechanisms that may warrant further investigation.

AUTHOR CONTRIBUTIONS

Joanna Gesche: Analysis of literature, design of the study, data analysis, and writing and approval of the manuscript. Christoph P. Beier: Analysis of literature, design of the study, data analysis, and writing and approval of the manuscript.

FUNDING INFORMATION

The study was supported by a postdoctoral scholarship from the Region of Southern Denmark to J.G., by the free research grant, Odense University Hospital, and by a grant from the Jascha Foundation (2022‐0278).

CONFLICT OF INTEREST

C.P.B. has received honoraria from UCB, Eisai, and Arvelle. J.G. does not report possible conflicts of interest.

Supporting information

FIGURE S1

TABLE S1

APPENDIX S1

Gesche J, Beier CP. Drug resistance in idiopathic generalized epilepsies: Evidence and concepts. Epilepsia. 2022;63:3007–3019. 10.1111/epi.17410