Modern antiepileptic drug development has failed to deliver: Ways out of the current dilemma


Address correspondence to W. Löscher, Department of Pharmacology, Toxicology, and Pharmacy, University of Veterinary Medicine Hannover, Bünteweg 17, D-30559 Hannover, Germany. E-mail: and
D. Schmidt, Epilepsy Research Group, Goethestr.5, D-14163 Berlin, Germany. E-mail: dbschmidt@t-online.deBoth authors contributed equally to this review.


Despite the development of various new antiepileptic drugs (AEDs) since the early 1990s, the available evidence indicates that the efficacy and tolerability of drug treatment of epilepsy has not substantially improved. What are the reasons for this apparent failure of modern AED development to discover drugs with higher efficacy? One reason is certainly the fact that, with few exceptions, all AEDs have been discovered by the same conventional animal models, particularly the maximal electroshock seizure test (MES) in rodents, which served as a critical gatekeeper. These tests have led to useful new AEDs, but obviously did not help developing AEDs with higher efficacy in as yet AED-resistant patients. This concern is not new but, surprisingly, has largely been unappreciated for several decades. A second—admittedly speculative—reason is that progress in pharmacologic treatment of drug-resistant epilepsy will not be made unless and until we develop drugs that specifically target the underlying disease. Although better preclinical approaches will not be able to circumvent regulatory requirements, more efficacious drugs may allow us to abandon clinically questionable trials with intentionally less efficacious controls and noninferiority designs, and require evidence for comparative effectiveness. The failure of AED development has led to increasing disappointment among clinicians, basic scientists, and industry and may halt any further improvement in the treatment of epilepsy unless we find ways out of this dilemma. Therefore, we need new concepts and fresh thinking about how to radically change and improve AED discovery and development. In this respect, the authors of this critical review will discuss several new ideas that may hopefully lead to more efficacious drug treatment of epilepsy in the future.

Historically, antiepileptic drugs (AEDs) can be classified into three generations (Fig. 1). The first generation, entering the market from 1857 to 1958, includes potassium bromide, phenobarbital (PB), and a variety of drugs that were derived mainly by modification of the barbiturate structure, including phenytoin (PHT), primidone (PRM), trimethadione, and ethosuximide (ESM) (Krall et al., 1978a; Shorvon, 2009a). The second generation AEDs, including carbamazepine (CBZ), valproate (VPA), and the benzodiazepines, which were introduced between 1960 and 1975, differed chemically from the barbiturates (Shorvon, 2009b). The superior tolerability of CBZ and PHT for focal seizures over the more sedative barbiturates, PB and PRM, was shown in a double-blind benchmark trial comparing all four AEDs (Mattson et al., 1985). Despite their well-known dose-related central nervous system (CNS) side effects, PHT and PB have never been shown to be less efficacious for focal seizures than CBZ (Mattson et al., 1985) and both are still in widespread use in many parts of the world, mainly because of their low cost. However, CBZ, PHT, PB, and PRM (which is metabolized to PB) have two clinically important disadvantages. They are potent enzyme inducers, leading to clinically important adverse drug interactions, and they cause hypersensitivity reactions (Schmidt & Beyenburg, 2009). Astute clinical observations by French physicians established VPA as an efficacious drug for idiopathic generalized and focal epilepsy (Löscher, 1999b). However, VPA turned out to be somewhat less efficacious for complex partial seizures than CBZ (Mattson et al., 1992) and has three other clinically important disadvantages: As an enzyme-inhibitor it is involved in deleterious drug interactions, it causes hepatic failure in predisposed individuals, and it is the most teratogenic among the currently marketed AEDs (Schmidt & Beyenburg, 2009). Following the introduction of VPA in the 1960s, no new AEDs were entering the market for almost two decades, except some additional benzodiazepines (Fig. 1). Not surprisingly, patients and physicians had great expectations for therapeutic gains over older drugs when the new, third-generation AEDs entered the market, which arbitrarily include all drugs for epilepsy marketed after the introduction of CBZ or VPA in the 1960s (Fig. 1). Despite the shortcomings of second-generation AEDs outlined above, the widespread use and the unsurpassed clinical efficacy of CBZ and VPA made them benchmarks for comparison with third-generation AEDs. The era of the third-generation AEDs started in the 1980s with “rational” developments such as progabide, vigabatrin (VGB), and tiagabine (TGB), that is, drugs that were designed to selectively target a mechanism that was thought to be critical for the occurrence of epileptic seizures (Löscher & Schmidt, 1994).

Figure 1.

Introduction of AEDs to the market from 1853 to 2009 (adapted from data by Shorvon, 2009a,b). Licensing varied from country to country. We give here the year of first licensing or the first mention of clinical use in a country of Europe, the United States, or Japan (see Shorvon, 2009a,b, for further details). We have not included all derivatives of listed AEDs or AEDs used solely for treatment of status epilepticus.

The new, third-generation AEDs have undoubtedly expanded the therapeutic options, in particular for those in need of a change in medical regimen. However, the efficacy of many, but unfortunately not all, new AEDs for treatment of new-onset epilepsy is similar to that of older AEDs (Kwan & Brodie, 2000; Brodie et al., 2007; Marson et al., 2007; Glauser et al., 2010). For patients with previously drug-resistant seizures, adjunctive new AEDs have been shown to be more efficacious than adding placebo (Beyenburg et al., 2010). In addition, new AEDs have other benefits over some of the older drugs for epilepsy. Treatment with some of the new AEDs avoids adverse drug interactions and hypersensitivity reactions (Elger & Schmidt, 2008) and some new AEDs have clinically important utility for disorders other than epilepsy (Rogawski & Löscher, 2004b).

Despite these encouraging and welcome advantages that new AEDs have brought for the management of epilepsy, there is growing concern that the efficacy of drug treatment of epilepsy has not substantially improved with the introduction of new AEDs (Shorvon, 2009b). Although we have no population-based studies if the prognosis of epilepsy has changed since and because of the introduction of new AEDs, this dilemma is highlighted by the following disconcerting observations. Currently, 38% of adult patients with new-onset epilepsy have AED-resistant seizures (Mohanraj & Brodie, 2005). Forty years ago, Coatsworth (1971) reported in his summary of AED trials that 33–38% of those with focal seizures were unchanged or worse. This has possibly not changed since the 1880s when Gowers reported that 36% of his patients did not become seizure-free with potassium bromide, the first modern AED (Gowers, 1881). This shows incidentally that medical intractability is not a new phenomenon. To illustrate the widespread disillusion about the efficacy of new AEDs, Perucca (2010) noted in a recent commentary that it is unrealistic to expect that a new AED will be more efficacious than established agents at full dosages. This current dilemma of AED development has led to increasing disappointment among clinicians, basic scientists, and industry and may halt any further improvement in the treatment of epilepsy unless we find ways out of this dilemma. Therefore, we need new concepts and fresh thinking about how to radically change and improve AED discovery and development. This review critically examines if new AEDs have provided therapeutic gains in the treatment of epilepsy, it discusses several possible reasons for the apparent failure of modern AED development to discover drugs with higher efficacy, and proposes new ideas for AED discovery and development that may eventually lead to more effective drugs for epilepsy. In view of the summary of AED trials by Coatsworth in 1971, it may be worthwhile to examine if we have made therapeutic gains in the last 40 years.

Have We Made Therapeutic Gains in AED Efficacy in the Last 40 Years?

The introduction of each new AED into the market raises valid expectations in patients and physicians for more effective treatment of epilepsy. Although safety, tolerability, and lack of drug–drug interactions through enzyme induction and having additional treatment options are important, better efficacy is a highly desirable feature for any new AED. This section is limited to a discussion of therapeutic gains in AED efficacy of new versus old AEDs. In that regard, the two most important clinical questions are: Is monotherapy with a new AED superior compared to older drugs in terms of seizure remission in recent-onset epilepsy? And: Is adjunctive or substitution therapy with a new AED superior in terms of seizure remission compared to older drugs in previously refractory epilepsy?

New AEDs for recent-onset idiopathic generalized epilepsy

The efficacy of lamotrigine (LTG) and topiramate (TPM) versus VPA for treatment of idiopathic generalized epilepsy was studied in a subset of Arm B of the SANAD trial (Marson et al., 2007). SANAD was designed to assess whether LTG or TPM should become first-line treatment and thereby replace VPA as the existing first-line agent (Marson et al., 2007). The result was that VPA was more efficacious than LTG and similar in efficacy to TPM for all patients (Fig. 2) and also for the subgroup of patients with idiopathic generalized epilepsy (Marson et al., 2007). Although the study showed equal efficacy of TPM and VPA, it should be added that TPM was grossly inferior in terms of effectiveness. LTG was also shown to be less efficacious versus VPA for previously untreated juvenile myoclonic epilepsy (Mohanraj & Brodie, 2007). More recently, a multicenter double-blind randomized trial compared treatment with ESM, LTG, or VPA in 453 children with new-onset absence epilepsy of childhood (Glauser et al., 2010). After 16 weeks of therapy, the freedom-from-failure rates for ESM and VPA were similar (53% and 58%), but for both the rates were higher than for LTG (29%; p < 0.001 for both comparisons).

Figure 2.

Comparative efficacy of new AEDs as measured by time to 12-month remission. None of the new AEDs was superior in efficacy compared to CBZ or VPA in the SANAD trial (Marson et al., 2007). Data are hazard ratio (HR) and 95% CI. HR >1 indicates that 12-month remission occurs more rapidly on drug compared to baseline drug CBZ or VPA. HR of 1 was assigned to CBZ and VPA. GBP was found to be less efficacious compared to CBZ (log-rank test statistic = 11.26, df = 4, p < 0.0238; indicated by asterisk). LTG was less efficacious compared to VPA in this trial (log-rank test statistic = 6.384, df = 2, p = 0.041; indicated by circle). For details see Marson et al. (2007).

New AEDs for recent-onset focal epilepsy

Arm A of the SANAD trial was designed as a pragmatic trial to assess whether any of the new AEDs LTG, gabapentin (GBP), TPM, or oxcarbazepine (OXC) should become first-line treatment and thereby replace the existing first-line agent CBZ (Marson et al., 2007). Based on efficacy criteria alone, and only those will be discussed here, none of the new AEDs were superior in efficacy to CBZ (Fig. 2). However, LTG and OXC were considered to be noninferior in efficacy, whereas CBZ was reported to be more efficacious compared to GBP (Marson et al., 2007). That GBP has been found to be less efficacious indicates that the SANAD trial has assay sensitivity to differentiate efficacious from less-efficacious treatment. Levetiracetam (LEV), which entered the market later, could not be studied in SANAD. However, a well-controlled noninferiority trial has shown that, at per-protocol analysis, 73.0% of patients randomized to LEV and 72.8% receiving controlled-release CBZ were seizure free at the last evaluated dose (adjusted absolute difference 0.2%, 95% CI −7.8% to 8.2%) for at least 6 months, indicating equivalent seizure remission for LEV versus slow release CBZ (Brodie et al., 2007). One recent study was inconclusive in establishing noninferiority of TPM (100 mg/day) versus oral PHT for new-onset focal seizures (Ramsay et al. 2010). This section does not include a discussion of monotherapy trials in recent-onset epilepsy in those older than 65 years, which is discussed elsewhere (Schmidt, 2011). Finally, it should be noted that the evidence base of comparative trials or large-scale clinical observations for comparing of newer versus older AEDs cannot be considered robust. An expert panel of the International League Against Epilepsy (ILAE) commission has voiced serious methodologic concerns (Glauser et al., 2006). Curiously, the current evidence base for comparing older versus newer AEDs in new-onset epilepsy after the drugs were on the market falls short of the benchmark trials comparing several older AEDs for new-onset epilepsy (Mattson et al., 1985, 1992).

In summary, none of the new AEDs were superior in efficacy to old AEDs such as CBZ and VPA in large, well-controlled trials of recent-onset epilepsy (Marson et al., 2007; Glauser et al., 2010). However, several new AEDs such as LEV, LTG, and OXC have been shown to be noninferior in seizure control by a predefined margin to CBZ. Moreover, GBP was shown to be less efficacious versus CBZ in mostly untreated focal epilepsy, and LTG was less efficacious compared to VPA (Marson et al., 2007). LTG was also shown to be less efficacious versus VPA and ETS for previously untreated childhood absence epilepsy (Glauser et al., 2010). This shows convincingly that current postmarketing trial designs for new-onset epilepsy—albeit not designs used for regulatory purposes—are able to detect less efficacious treatment, if it exists.

Although we could not find any evidence that any new AEDs are more efficacious than standard AEDs for new-onset epilepsies, VGB has received orphan drug status for new-onset West syndrome, although it does not seem to be more efficacious than hormonal treatment in the long-run (Darke et al., 2010). However, in earlier small trials VGB was shown to be superior in efficacy in new-onset West syndrome due to cerebral malformations or tuberous sclerosis, whereas adrenocorticotropic hormone (ACTH) proved more efficacious in cases with perinatal hypoxic/ischemic injury. The efficacy of the two drugs was similar in cryptogenic cases (Vigevano & Cilio, 1997). In another study of West syndrome, VGB was more efficacious than hydrocortisone (Chiron et al., 1997). VGB has caused concentric visual field defects requiring age-appropriate visual field testing and discontinuation in cases where spasm or seizure improvement is not achieved within 12 weeks of initiation (Willmore et al., 2009).

New AEDs for previously refractory epilepsy

The evidence base for comparing the efficacy of newer versus older drugs for refractory epilepsy is surprisingly weak. We found only one novel comparative trial, which included placebo as a third arm, showing noninferiority of PGB versus LTG for refractory focal seizures (Baulac et al., 2010). Although several clinical observations suggested that a change to a new medical regimen, which included third-generation AEDs, has led to seizure-freedom for at least 6–12 months in 14–28% of patients previously considered to have refractory focal seizures (Callaghan et al., 2007; Luciano & Shorvon, 2007), these reports provide no firm evidence that modern AEDs have substantially improved the treatment of epilepsy. The reason is that the publications included no compelling evidence that the patients had received adequate prior treatment with maximum doses of older AEDs. Maximum dose treatment has been shown to lead to seizure-freedom without a change of drug in up to 31% of patients with prior refractory epilepsy (Schmidt, 1983). Although the introduction of several new AEDs as orphan drugs such as stiripentol for Dravet syndrome (Chiron et al., 2000), rufinamide for refractory Lennox-Gastaut syndrome (Glauser et al., 2008), and brivaracetam for symptomatic myoclonus (U.S.A.) can, in general, be considered as a therapeutic gain, none of the orphan drugs has been compared in its efficacy to other drugs.

In clinical drug development, modern adjunctive drugs for refractory epilepsy are usually compared with adding placebo to the existing medication, but not with addition of a standard older drug (see below). Overall, we found no compelling evidence for therapeutic gain in efficacy with new AEDs versus older standard drugs in refractory epilepsy. This raises concern, given the seemingly unimproved proportion of patients with drug-resistant epilepsy, as discussed in the introduction. Finally, another serious concern with the current evidence base is that physicians and cost bearers are not provided evidence if a new AED is superior in any aspect at the time it enters the market and the pricing is determined.

Have We Made Therapeutic Gains in AED Tolerability or Safety in the Last 40 Years?

Although many believe that some modern AEDs are better tolerated than older ones, the authors of the ILAE Treatment Guidelines have suggested that statistically this has been very hard to show, except in a few studies (Glauser et al., 2006). There is concern that many of these studies were designed to support marketing strategies, and some of the methods used in these trials can skew the results in favor of the sponsor’s product (Aronson, 2006). For example, choice of inclusion and exclusion criteria, choice of comparator drug and formulation (slow release or not), dosing intervals, titration rates, and end points can influence outcome (Glauser et al., 2006) There is no doubt, however, that lower risk of hypersensitivity reactions and detrimental drug interactions have made some new AEDs such as LEV or GBP better tolerated and easier to use than some of the first-generation AEDs such as CBZ or PHT (Schmidt & Beyenburg, 2009). Among third-generation AEDs, LEV has advantages such as ease of use, lower risk of rash and of drug interaction, and good utility also for adjunctive use in juvenile myoclonic epilepsy that have made it one of the most often used third-generation AEDs. The use of LEV is likely to increase even more when it is widely available as generic medication, as in the United States.

Although a detailed discussion of life-threatening adverse effects of new versus old AEDs is beyond the scope of this review, the lower risk of hypersensitivity and idiosyncratic reactions for some third-generation AEDs such as GBP and LEV may offer advantages over second-generation AEDs such as CBZ, PHT, or PB (Schmidt, 2009).

However, serious idiosyncratic adverse effects have also been reported for several new AEDs such as VGB (concentric visual field defects), felbamate (FBM; aplastic anemia, hepatic failure), LTG, and OXC (Stevens–Johnson syndrome) (Schmidt, 2009). For some of the third-generation AEDs, clinical trials indicated depressive symptoms in more than 1% of treated patients (FBM, LEV, TGB, TPM, VGB, zonisamide), whereas frequencies of <1% have been noted for GBP, LTG, OXC, and PGB (Mula & Sander, 2007). Use of newer AEDs with a higher potential of causing depression was associated with a threefold increased risk of self-harm/suicidal behavior [odds ratio (OR) 3.08; 95% confidence interval (CI) 1.22–7.77] as compared with no use of AEDs during the last year (Andersohn et al., 2010). However, use of first- or second-generation AEDs or newer AEDs such as OXC, GBP, PGB, and LTG was not associated with an increased risk of self-harm/suicidal behavior (Andersohn et al., 2010). This seems to be in principal agreement with a report that epilepsy patients without depression are not at increased risk of suicidal ideation, behaviour, or both, when using second- or third-generation AEDs (Arana et al., 2010). However, Arana et al. (2010) did not specify outcome for those on third-generation AEDs that were implicated to be associated with suicidality by Andersohn et al. (2010). Therefore, the study by Arana et al. (2010) is not well suited to clarify the association of a subgroup of third-generation AEDs with suicidality, if it exists. It is worthwhile to note that all AEDs regardless of the data presented currently carry an FDA “class label” indicating that patients taking an AED are at a greater risk for suicide (US Food and Drug Administration, 2008).

Unfortunately, regulatory trials are usually too short and too small to capture rare adverse effects that emerge during long-term treatment. In addition, fixed-dose titration, which is used in many regulatory trials, is prone to overstate the side-effect profile seen with prudent flexible dosing in clinical practice or with single drug treatment. Furthermore, surprisingly, there are no comparative trials that have evaluated the general clinical benefit associated with a lower risk of hypersensitivity reactions and drug interactions seen with some of the new AEDs such as LEV or GBP versus older AEDs such as PHT and CBZ. However, a double-blind comparative trial of LEV versus CBZ-constant release (CR) mimicking clinical practice reported a similar proportion of patients in the LEV (79.6%) and CBZ-CR groups (80.8%) experienced at least one adverse event during the treatment period (Brodie et al., 2007). Overall, there was no substantial difference in the adverse effects reported between LEV and CBZ and the proportion of patients who discontinued therapy because of adverse events (Brodie et al., 2007). Although LTG had a lower rate of drug discontinuation because of side effects compared to CBZ in a large randomized but unmasked study of patients with mostly new-onset epilepsy (Marson et al., 2007), the same study reported no differences in the proportion of patients with at least one adverse effects during treatment with LTG, GBP, TPM, or OXC compared with CBZ or VPA (Fig. 3).

Figure 3.

Comparative tolerability of new AEDs. Given is the proportion of patients with mostly new-onset epilepsy reporting at least one clinically important adverse event. None of the new AEDs was superior in tolerability compared to CBZ (A) or VPA (B) in the SANAD trial (adapted from Marson et al., 2007).

The proportion of patients withdrawing medication due to adverse effects was similar for several new AEDs when compared to conventional or slow-release CBZ (Brodie et al., 2007; Marson et al., 2007). Teratogenic effects, which are a clinically important adverse outcome, could not be assessed in the SANAD study or the study by Brodie et al. (2007). Among old AEDs, VPA is known to be the most teratogenic agent, whereas CBZ seems to have a lower teratogenic potential than VPA (Tomson & Battino, 2009). Among the often-used new AEDs, LTG and possibly LEV seem to have a low teratogenic potential, which seems to be similar to that of CBZ and lower than that of VPA (Tomson & Battino, 2009). However, as a note of caution, we currently have no large-scale comparative teratogenicity data of new versus old AEDs.

The current review suggests that, in agreement with the literature (Wilby et al., 2005), taken together, the available evidence on the tolerability and safety does not support a general advantage of all new AEDs over all old AEDs. However, it is clinically important that some of the new AEDs such as LEV, GBP, or LTG seem to be safer compared to some other new AEDs, such as VGB and FBM, and to VPA, among the old AEDs. In addition, some of the new AEDs such as LEV, GBP, or LTG seem to be better tolerated than conventional CBZ, particularly in the elderly. Compared to slow-release CBZ, however, a head-to-head study in elderly patients found only a nonsignificant trend for better tolerability of LTG (Saetre et al., 2007).

Preclinical Strategies to Develop AEDs: The Last 40 Years

The discovery and development of a new AED relies heavily on the preclinical use of animal models to establish efficacy and safety prior to first trials in humans (White et al., 2006). This approach has been very successful and crucially contributed to the development of numerous clinically effective AEDs. Indeed, animal models with a similarly high predictive value do not exist for other CNS disorders, such as bipolar disorders or migraine (Perucca et al., 2007). At least three strategies have been used for generating new compounds for testing of anticonvulsant activity in animal models: (1) random screening of newly synthesized compounds of diverse structural categories with as yet unknown mechanisms; (2) structural variation of known AEDs; and (3) rational drug design by developing drugs that selectively target a mechanism that is thought to be critical for the occurrence of epileptic seizures (Löscher & Schmidt, 1994; Löscher, 1998; Löscher & Schmidt, 2002; Rogawski & Löscher, 2004a; Bialer & White, 2010). All three strategies have generated clinical useful AEDs, although many scientists believe that rational drug development resulting in highly target-selective compounds has advantages over the more traditional other strategies. Why have these preclinical strategies obviously failed to discover new AEDS with better efficacy compared to the old AEDs? After the anticonvulsant effect of the first AEDs had been discovered mostly by serendipity following observations with hypnotic drugs in patients with epilepsy (potassium bromide by Sir Charles Locock in 1857 and PB by Alfred Hauptmann in 1912), the systemic search for AEDs was initiated by Merritt and Putnam’s work at the Boston City Hospital in the 1930s, which represents a pharmacology milestone (Anderson, 2009). Although Merritt and Putnam were not the first to use electroshock for seizure induction in animals, they devised a simple and reliable method, the so-called electroshock threshold method in cats, to assess drugs for anticonvulsant effects and then used it to systematically screen hundreds of compounds for efficacy. They showed that the anticonvulsant and sedative effects of drugs could be separated, inspiring not only their own research but also that of many other investigators for more selective drugs. They validated their laboratory observations by demonstrating that their animal model accurately predicted anticonvulsant efficacy in humans, and PHT, their first discovery (Putnam & Merritt, 1937), was also the first anticonvulsant drug to be tested in animals before it was given to humans. The electroshock test was later modified for mice and rats, and the maximal electroshock seizure (MES) test created by Toman, Swinyard, and Goodman (1946) is still the most commonly used first screen in the search for new AEDs, being quite effective in identifying drugs that block generalized tonic–clonic seizures in humans (Bialer & White, 2010).

The second “gold standard” for the early detection of anticonvulsant activity is the s.c. pentylenetetrazole (PTZ) test, which is thought to be useful in identifying drugs that block generalized nonconvulsive (absence, myoclonic) seizures (White et al., 2006). The γ-aminobutyric acid (GABA)A receptor antagonist PTZ (also known as pentetrazol and metrazol) was introduced as a convulsant in 1926 and has been broadly used experimentally to study seizure phenomena and, until the late 1930s, for convulsive therapy in patients with major depression (Löscher, 2009). In 1944, Everett and Richards used the PTZ seizure model in mice to demonstrate the anticonvulsant effect of trimethadione, which was subsequently demonstrated to block absence seizures in humans and introduced for this indication in 1946. Everett and Richards (1944) also showed that PHT was ineffective in the PTZ model, which is in line with its lack of efficacy against absence seizures in patients (Table 1). Therefore, two simple animal models, the MES and PTZ tests, could be used to differentiate AEDs with different clinical effects, which subsequently formed the basis for Swinyard (1949) and Swinyard et al. (1952) to propose the MES and s.c. PTZ tests in mice and rats as standard procedures for predicting clinical anticonvulsant activity of investigational drugs. Some years after the discovery of trimethadione, the PTZ test was crucial to identify the succinimides (Chen et al., 1951), including phensuximide, methsuximide, and ESM, which rapidly replaced oxazolidinediones such as trimethadione because of their superior tolerability. Interestingly, based on experiments with a range of compounds in the MES and PTZ tests, Swinyard et al. (1952) noted that, unless the candidate AEDs under investigation are chemically related to compounds known to possess clinical anticonvulsant usefulness, the assay results obtained by these methods may have only limited value for predicting clinical efficacy or specificity for various seizure types.

Table 1.   Anticonvulsant spectrum of AEDs in models and man
DrugAnticonvulsant effect in rodent modelsClinical efficacy (seizure suppression)
MES (mice/rats)s.c. PTZ (mice/rats)Amygdala-kindling (rats, focal seizures)Partial seizuresGeneralized seizures
  1. NE, not effective.

  2. Adapted from Rogawski and Löscher (2004a) and Bialer et al. (2009).

Predominant Na+ (and Ca2+) channel activity
Predominant Ca2+ channel activity
GABA systems
Novel targets

The fact that the same two tests that were developed more than 60 years ago are still considered “gold standards” for detection of new AEDs (Bialer & White, 2010) is most likely the main reason that preclinical strategies failed to discover new AEDS with better efficacy and tolerability compared to the old AEDs. For many decades, drugs that were not effective in these tests were excluded from further development. This strategy was also recommended by the Anticonvulsant Drug Development (ADD) Program of the National Institute of Neurological Disorders and Stroke (NINDS) of the National Institutes of Health (NIH; Bethesda, MD), which was initiated in 1975 to stimulate the discovery and development of new chemical entities for the symptomatic treatment of human epilepsy (Krall et al., 1978a,b; White et al., 2006). In addition to evaluating drugs for anticonvulsant activity in the MES and s.c. PTZ tests, this program includes tests for “neurotoxicity” such as the rotarod test to calculate the “protective (or therapeutic) index” between anticonvulsant and neurotoxic doses of test compounds. Since its initiation, the ADD program has evolved and various additional models have been implemented (White et al., 1995; White, 1997; White et al, 1998; White, 2003; White et al., 2006; Smith et al., 2007; Bialer & White, 2010), more recently also models of AED-resistant seizures, to avoid that interesting compounds are missed when only using the MES and s.c. PTZ tests (Fig. 4). However, although the early models (MES and s.c. PTZ) are now only a part of the program’s screening process (Fig. 4), they are still used extensively as critical gatekeepers for the identification of promising drug candidates (Stables et al., 2002; Bialer & White, 2010).

Figure 4.

Schematic diagram illustrating the initial screen of the NINDS-sponsored University of Utah Anticonvulsant Drug Development (ADD) Program. An investigational compound is initially screened for efficacy in the MES and s.c. PTZ tests. The activity of those compounds with demonstrated efficacy and minimal behavioral toxicity is subsequently quantitated (ED50 and TD50) at the time of peak anticonvulsant effect. Compounds found inactive in the MES and s.c. PTZ tests are evaluated in the LEV-sensitive 6-Hz seizure test in mice. For those compounds that are found to be active in the 6-Hz test, their activity is quantitated at their respective time of peak effect. All compounds found to be active in one or more of these three identification screens are then differentiated on the basis of their activity in additional seizure models, including the hippocampal kindled rat model of TLE. Adapted from White et al. (2006).

More than 27,000 compounds from academic and pharmaceutical institutions were evaluated by the NINDS-sponsored ADD program, leading to the discovery of various new chemical entities with anticonvulsant activity (White et al., 2006). However, it has been argued that the use of “old” models will only identify “me-too” drugs and is unlikely to identify drugs that will have an effect on the refractory epilepsies (Löscher, 1998; Stables et al., 2002; White et al., 2006; Bialer & White, 2010). Furthermore, interesting compounds with novel mechanisms of action may be missed by the MES and PTZ tests. An interesting example in this regard is the third-generation AED LEV. In the late 1980s, Alma Gower, a pharmacologist working at UCB Pharma in Belgium, observed in routine screening that LEV, the (S)-enantiomer of the ethyl analog of the nootropic piracetam, showed potent anticonvulsant activity in audiogenic seizure-prone mice (Gower et al., 1992). Gower et al. (1992) also reported that LEV protected against MES and PTZ seizures in mice, but subsequent studies in our laboratory failed to confirm these findings when the MES and s.c. PTZ tests were performed with standard parameters (Löscher & Hönack, 1993). This failure of the MES and s.c. PTZ tests to identify the anticonvulsant activity of LEV was later confirmed by several other groups, including reports from UCB and the NINDS-sponsored ADD program (Klitgaard et al., 1998; White et al., 2006). However, LEV was found to be quite effective in chronic models of partial and primary generalized seizures, including the kindling model of temporal lobe epilepsy (TLE) (Löscher & Hönack, 1993; Klitgaard et al., 1998). Whereas the MES and PTZ models induce seizures in healthy, neurologically intact rodents, kindling is a chronic model in which the repeated application of electrical stimuli via a depth electrode in the limbic system (amygdala or hippocampus) induces permanently enhanced seizure susceptibility and other enduring brain alterations that are similar to those occurring in human TLE (Sato et al., 1990). The kindling model is the only chronic model that is currently used by most AED discovery programs, including the NINDS-sponsored ADD program (Bialer & White, 2010). It is the only model that adequately predicted the clinical utility of novel AEDs against partial seizures in patients with epilepsy (Table 1). However, AEDs, such as LEV, which are particularly effective in the kindling model, do not exhibit higher clinical efficacy in patients with pharmacoresistant partial seizures, as discussed earlier, so that kindling does obviously not predict anticonvulsant efficacy in pharmacoresistant partial epilepsy. AED testing in large groups of amygdala-kindled rats showed, however, that about 20% of kindled rats do not respond to standard AEDs such as PHT, so that such AED-resistant subgroups of kindled rats (but not all kindled rats per se) may serve as a model of refractory TLE (Löscher, 2006).

In the last 20 years, a number of animal models of pharmacoresistant epilepsy have been developed (Löscher, 2006), but none of these models is routinely used by AED discovery programs. The only exception is the 6-Hertz (6-Hz) test in mice, in which low-frequency (6-Hz), long-duration (3-s) electrical stimulation via corneal electrodes is used to induce partial seizures that are resistant to PHT and many other AEDs, when using high stimulus intensities (Barton et al., 2001). As such, the 6-Hz seizure model may represent a potential therapy-resistant model wherein seizures can be evoked in normal mice (White et al., 2006). As shown in Fig. 4, this model is now included in the NINDS-sponsored ADD program that is performed at the University of Utah, because, in contrast to the MES and s.c. PTZ tests, this model would have discovered the anticonvulsant activity of LEV. However, it remains to be seen whether any new AED discovered by the strategy illustrated in Fig. 4 exhibits clear advantages in as yet pharmacoresistant patients with epilepsy.

In conclusion, the MES and s.c. PTZ tests developed >60 years ago have been extremely useful for identifying various new AEDs, but obviously they did not help develop AEDs with higher efficacy in as yet AED-resistant patients. This concern is not new (e.g., Löscher, 1986, 1998, 2002a; Löscher & Schmidt, 1988, 2002; Stables et al., 2002) but, surprisingly, has not been fully appreciated for several decades. The fact that preclinical models used for identification and development of novel AEDs have been originally validated by “old” AEDs may explain why none of the new AEDs possesses significant advantages in efficacy and tolerability compared with the old drugs (Löscher, 1998). Drugs identified by the MES test often resemble PHT, which acts by modulating voltage-dependent sodium channels, whereas drugs identified by the PTZ test have often a benzodiazepine-like mode of action, potentiating the inhibitory effect of GABA (Meldrum, 1997), although both tests also picked up various drugs with other mechanisms (Bialer & White, 2010). Because of logistical problems when testing large numbers of compounds, more laborious models of epilepsy such as kindling, are used at only later stages of drug development, although, in contrast to acute seizure models such as the MES test, drug testing in the kindling model predicted the negative outcome from clinical testing of drugs blocking the N-methyl-d-aspartate (NMDA) subtype of glutamate receptors (Löscher & Hönack, 1991). The latter finding is one important example for failure of a “rational” strategy of AED development, but several other rational concepts also failed (Löscher & Schmidt, 1994). However, one important limitation when evaluating the success of rational strategies is that we do actually not know how many of the numerous investigational AEDs that have been tested by the NINDS-sponsored ADD program and other programs were actually derived from a mechanistic drug discovery program. It may very well be that many mechanisms have been evaluated in not only traditional but also less traditional models, but their efficacy was not sufficient to support advancement to the next level of the drug development process. Therefore, it would be extremely helpful if pharmaceutical industry would allow access to unpublished preclinical data.

Clinical Strategies to Develop AEDs: The Last 40 Years

The principal aim of clinical development of new AEDs in the last 40 years has been to fulfil the minimum requirements of regulatory agencies to achieve market authorization. Regulatory agencies in the European Union and in the United States are currently mandated to require evidence from well-controlled trials with adequate assay sensitivity showing that the experimental AED is safe and effective [European Medicines Agency (EMEA), 2010; U.S. Food and Drug Administration (FDA), Center for Drug Evaluation and Research (CEDR), 2010]. Assay sensitivity is the ability of the trial to distinguish an effective treatment from a less effective or ineffective treatment (EMEA, 2001).

Clinical development of AEDs for refractory epilepsy

An experimental AED is currently considered to be effective for refractory epilepsy in the United States and in the European Union, if it is more effective than placebo during adjunctive treatment. Although placebo-controlled trials have successfully identified efficacious and nonefficacious experimental AEDs, they have a number of disadvantages (Table 2).

Table 2.   Advantages and disadvantages of conventional placebo controlled add-on trials for refractory epilepsy
Superiority design detects AEDs that are and those that are not more efficacious than placebo (assay sensitivity)May overestimate efficacy because seizure-free rates are for 12–24 weeks only. Usually measures 50% seizure reduction which does not lead to substantial improvement of quality of life (Birbeck et al., 2002)
Accepted worldwide “as gold standard” by regulatory agenciesDoes not provide a clinically meaningful seizure-free outcome that can directly guide the choice among several drugs. Placebo is not an approved drug for treatment of epilepsy
Optional third arm for comparison with standard drug at full dose (Baulac et al., 2010; see text)Unexplained variability of placebo response from study to study. Incomplete understanding what drives the placebo response
Allows double-blind designDo not allow optimal treatment conditions in terms of patient selection
 Denies placebo patients treatment with an active comparator drug

The utility of placebo controls has received critical attention, mainly because response to placebo has shown unexplained variability from trial to trial. In a recent meta-analysis the proportion of patients reporting 50% or more seizure reduction versus baseline was 15%, with a range of 0–39% for those on placebo as add-on treatment (Beyenburg et al., 2010). Overall in 30 studies, 2.1% (with a wide range of 0–17%) of controls receiving adjunctive placebo were seizure-free (Beyenburg et al., 2010). In addition, unusually high and unexplained placebo response rates may possibly have played a role in the failure of some recent experimental AEDs, such as carisbamate and brivaracetam, to show efficacy superior to placebo in individual add-on trials and in a comparative trial (Baulac et al., 2010). Furthermore, placebo exposure has raised ethical concerns for withholding active treatment from patients with frequent seizures during the trial (Chadwick & Privitera, 1999).

For monotherapy use, in later stages of the development, evidence from well-controlled trials is required to show that the experimental AED is more efficacious for treatment of refractory epilepsy than pseudoplacebo in withdrawal-to-monotherapy designs (U.S.A.). This approach has allowed the efficacy of several new AEDs to be demonstrated, including FBM, TPM, LTG, and OXC, but not GBP and TGB (Beydoun & Kutluay, 2003). This design, however, has also met with increasing criticism from the epilepsy community, mainly because randomizing patients with frequent seizures to a deliberately suboptimal treatment (pseudoplacebo) is ethically questionable (Perucca & Tomson, 1999). Pseudoplacebo is usually a lower dose of the test AED or of a standard comparator AED that is expected to be less efficacious than the standard AED given at a fully effective dose in clinical use (EMEA, 2001).

Clinical development of AEDs for new-onset epilepsy

In the European Union, the EMEA recommends that monotherapy studies be conducted in patients with newly diagnosed epilepsy, employing “randomised, double-blind positive controlled trials aiming to demonstrate at least a similar benefit/risk balance of the test product as compared to an acknowledged standard product at its optimal use” (CPMP, 2000). As a consequence, monotherapy approval in Europe generally requires evidence of noninferiority (i.e., in clinical terms of not being worse for seizure control) by a predefined margin to an established comparator. Following that procedure, a monotherapy license was granted for LEV use in new-onset focal epilepsy (Brodie et al., 2007). In the United States, a noninferiority approach to demonstrate efficacy of AEDs is not accepted by the FDA (Perucca, 2010). The FDA argues that no sufficient evidence exists to determine that established AEDs can be consistently differentiated from less effective or ineffective treatments under any specific study design, and, therefore, the finding of equivalence or noninferiority does not allow to exclude that, in the particular population and under the specific conditions in which the trial was conducted, both treatments could have been similarly ineffective (Leber, 1989). Based on these compelling premises, the FDA requests that evidence of efficacy be obtained by demonstrating superiority over a comparator (Perucca, 2010). However, the use of suboptimal treatments as controls to achieve assay sensitivity has been increasingly criticized on ethical grounds, as noted above. A meta-analysis has now demonstrated that patients with refractory epilepsy randomized to suboptimal treatments in all previous trials had similar outcomes, thereby allowing the buildup of a dataset of historical controls against which response to investigational AEDs can be compared in future trials (French et al., 2010; see Katz, 2006 for further discussion). Although the avoidance of suboptimal treatments in future trials is a welcome development, the conversion-to-monotherapy design does not allow for comparative effectiveness against best standard drug.

Why are new AEDs not more efficacious in previously refractory epilepsy? A clinical perspective

For our discussion of possible reasons, apart from preclinical development, that new AEDs are not more efficacious for previously refractory epilepsy, several possible explanations can be considered, including prior life-time exposure to more than five AEDs (Schiller & Najjar, 2008). In an add-on trial of lacosamide, for example, a total of 82.1% of patients had been treated previously with four or more lifetime AEDs, including 48.3% who had been treated with seven or more AEDs in their lifetime (Chung et al., 2010). Only 16.4% of patients had 1–3 lifetime AEDs (Chung et al., 2010). Although the trial design was able to detect a difference between placebo and test drug, it would be useful to explore in a post hoc analysis the relationship, if any, between trial outcome and exposure to lifetime AEDs to explore the clinical impact of prior AEDs for seizure-free outcome in clinical trials of refractory epilepsy, if any.

Although a detailed discussion about the possible mechanism(s) of why AEDs work less well after many earlier AEDs failed to control seizures is also beyond the scope of this review (see Schmidt & Löscher, 2005), several explanations may be briefly considered. (1) The previous AEDs have already exhaustively and unsuccessfully tried to affect the relevant mechanisms for seizure initiation, propagation, and termination, and other seizure-related mechanisms exist that are not affected by current AEDs. (2) Cross-tolerance to the present AED has developed during prior AED exposure (for review see Löscher & Schmidt, 2006). (3) Drug resistance to several AEDs is caused by functional changes associated with the underlying epilepsy, and epileptogenic processes do not seem to be favorably influenced by mechanisms of current AEDs (our current favorite explanation). Whatever the reason, a population in which a large number of prior AEDs has failed to control seizures does not provide useful efficacy data for patients with drug-resistant epilepsy exposed to five or fewer lifetime AEDs who can be expected to become seizure-free (up to 16%) after change in medical regimen (Schiller & Najjar, 2008). Population-based studies have shown that the number of previous AEDs is a significant prognostic factor (MacDonald et al., 2000). Studies in patients with five or fewer prior lifetime AEDs are more likely to show differences in efficacy, if they exist, and would support the use of such AEDs for early use in refractory epilepsy. In addition, this would limit the repeated inclusion of the same patient in subsequent trials. The downside is, of course, that the marketing license might be limited to such populations. On the other hand, having a license for early add-on use in refractory seizures may be an advantage to gain market share.

A second, also largely unexplored reason that new AEDs have not been shown to be more efficacious in apparently refractory epilepsy may be the high baseline seizure frequency of those entering clinical trials for refractory focal epilepsy. Baseline seizure frequency has typically been on average eight seizures during the 8-week baseline period (Sharief et al., 1996). Although such a high seizure frequency has the advantage of allowing baseline duration of as short as 4 weeks, population-based studies have shown that a high seizure frequency of, for example, one or more seizures per week, is a predictor for poor drug response, at least for childhood-onset epilepsy (Sillanpää & Schmidt, 2009). Post hoc analysis is needed to clarify the relationship between baseline seizure frequency and seizure-free outcome in patients undergoing add-on trials for refractory epilepsy, if any.

A third, also largely unexplored reason that new AEDs have not been shown to be more efficacious than older AEDs may be that we have overlooked that certain combinations of AEDs may be more efficacious than other combinations, as suggested by isobolographic analysis (Czuczwar et al., 2009). Although a detailed discussion of rational polytherapy is beyond the scope of this review, combining two AEDs with widely different mechanisms of action may have advantages. A post hoc analysis of two phase III trials combining the third-generation AED lacosamide, which has a sodium channel mechanism, with AEDs that have a non-sodium channel mechanism is an example (Sake et al., 2010). Evaluating treatment of suitable, well-tolerated AED combinations versus single drug treatment may be a promising new idea to enhance the efficacy of third-generation AEDs. In addition, combining two AEDs with different mechanisms may offer advantages in tolerability (Sake et al., 2010). However, the interpretation of such trials is not as straightforward as it may seem. Confounding factors that need to be considered include the number of life-time AEDs and the baseline seizure frequency, as discussed earlier.

Is trial design to blame for the dilemma of modern AEDs?

Although the use of placebo and pseudoplacebo in trials of refractory epilepsy and noninferiority trials in new-onset epilepsy have rightly received critical attention and met with ethical concerns (Chadwick & Privitera, 1999; Garattini & Bertele’, 2007, Perucca, 2008), it would be unfair, at least in our view, to lay the blame for the current dilemma of new AEDs on the doorsteps of the regulatory agencies and their rules for trial design, mainly for two reasons. (1) Regulatory trials have been mandated to provide assay sensitivity and be it at the expense of introducing intentionally weaker efficacy controls, even if they are recognized to have no or little clinical merit as treatment for epilepsy. (2) Current guidelines do not preclude a drug company from presenting evidence for comparative efficacy of their test AED versus best AED treatment as control. Although any improvement of trial design is welcome, it is not altogether unreasonable to consider that drug companies themselves could have viewed their new test AEDs as not being unequivocally superior to best treatment and opted for a more risk-averse efficacy trial design. If we had better preclinical drug development resulting in more efficacious drugs, we would not need inactive or less active control to show superior efficacy of the new AED.

Proposal for Future Preclinical Strategies to Develop More Efficacious AEDs

Previous AED discovery programs, which were based primarily on the ability of an investigational drug to limit an evoked seizure in a normal mouse or rat, have been highly effective in identifying new AEDs and predicting anticonvulsant activity of such drugs in patients, but have failed to discover drugs with markedly higher efficacy compared to older AEDs. What are the possible reasons for this failure and what can we learn from them when thinking about conceptional shifts in AED development?

First, the question has always been whether the MES and s.c. PTZ seizure models would identify all potential AEDs or whether these old models would fail to identify compounds that had great potential efficacy but worked through mechanisms not tested by these models (Stables et al., 2002). One important example in this respect is LEV that would not have been developed when solely relying on the MES and s.c. PTZ models. In this respect it is also important to note that the MES test is a model of generalized tonic–clonic seizures and does not reliably predict anticonvulsant effects against partial seizures, but may result both in false-negative (e.g., LEV, VGB, TGB) and false-positive (e.g., NMDA antagonists) conclusions in this respect (Table 1). Furthermore, it has been suggested that the MES test preselects certain adverse effects of AEDs, apparently as a result of the model identifying compounds with specific molecular targets (Meldrum, 2002). Meldrum (2002) has proposed several immediate and more remote strategies for overcoming this problem. For a discussion on how to develop better tolerable AEDs, the interested reader is referred to several previous reviews (Löscher & Schmidt, 2002; Meldrum, 2002; White et al., 2008). However, animals have only modest value for predicting human tolerability (Perucca et al., 2007), although this can be improved by using kindled or epileptic rodents for CNS adverse-effect testing (Löscher & Schmidt, 2002; Meldrum, 2002). Similar to the limitations of the MES test, the s.c. PTZ test does not always reliably predict activity against absence seizures but may produce false-positive (e.g., PB, VGB, TGB) and false-negative (e.g., LTG) data (Table 1), so that additional models of nonconvulsive seizures, such as rat substrains with spontaneously occurring spike-wave discharges in the electroencephalogram (EEG) (e.g., the Strasbourg rat), should be included in the characterization of novel AEDs (Löscher, 1999a; White et al., 2006).

Second, the fact that AEDs discovered by the MES and PTZ models do not work in about 30% of epilepsy patients indicates that these early discovery models lack sufficient predictability for pharmacoresistant seizures. However, except for the 6-Hz mouse model in the ADD program of the NINDS (Fig. 4), none of the available models for AED-resistant seizures is routinely included in initial testing of investigational drugs in current academic or pharmaceutical AED discovery programs. It is not surprising that the MES and s.c. PTZ models failed to identify drugs that have an effect on the refractory epilepsies, because these models were initially introduced because of their ability to identify conventional AEDs for new-onset seizures. However, it has to be kept in mind that selection of models used in AED discovery programs such as the ADD program of the NINDS was based in part on cost, model efficiency, and the apparent prediction for clinical success by models available at the time. The models also were chosen to allow screening of large numbers of compounds, which is not possible when using more complex and time-consuming models such as kindling or epileptic animals with spontaneous recurrent seizures. Therefore, when thinking about new preclinical strategies, one has to consider the necessity of testing large numbers of compounds during early drug development. Currently, the 6-Hz model of AED-resistant seizures is the only model of pharmacoresistance that allows mass-screening of novel compounds. However, validation of the 6-Hz model is an unresolved issue. We do not know yet whether novel compounds that act in the 6-Hz model have higher antiepileptic efficacy in as yet AED-resistant patients. Overall, choosing an optimal model for the search of more effective compounds is a complex endeavor in the quest to balance out the requirements of mass-screening versus the validity of models to discover drugs for previously drug-resistant epilepsies. Furthermore, one important additional problem is that there may be hundreds of compounds that are active in the 6 Hz or other pharmacoresistant models that will never reach the point of clinical trials because of a lack of support by the pharmaceutical industry. There is a clear need for a better collaboration between the supplier of novel compounds, the regulatory bodies, the clinical trial consortia, and pharmaceutical industry, so that the truly novel compounds will at least be given their day in court.

To our knowledge, we were the first to propose more than 20 years ago to include in AED testing also models in which conventional AEDs fail to suppress seizures (Löscher, 1986) and demonstrated that AED-resistant subgroups of rats can be selected from kindling and post–status epilepticus (SE) models of TLE (cf., Löscher, 2006), but these models of pharmacoresistant TLE have the disadvantage that they are not suited for initial drug screening. Furthermore, the fact that simple screening models such as the MES test are not only predictive of anticonvulsant activity in epilepsy patients but also identified drugs that were subsequently found to exert beneficial effects in other neurologic or psychiatric diseases, including neuropathic pain, migraine, and bipolar disorders, has been a strong impetus for the pharmaceutical industry to invest in AED development (Rogawski & Löscher, 2004b). Efficacy in one or more of these disorders could triple the market potential of any new AED (Bialer & White, 2010). It seems plausible that shared molecular actions underlie the efficacy of AEDs in epilepsy and nonepileptic conditions, suggesting that epilepsy, pain syndromes, and affective disorders and the like have common pathophysiologic mechanisms (Rogawski & Löscher, 2004b). Whether AEDs that are ineffective in traditional seizure models such as the MES test would exert therapeutic effects in nonepileptic conditions is not known, but it is interesting to note that LEV, which is inactive in the MES test, failed to exhibit efficacy in a neuropathic pain trial (Finnerup et al., 2009). Whether LEV has potential benefit in other neurologic or psychiatric disorders remains to be established (Farooq et al., 2009).

Third, it has recently been argued that the seizure types used as end points in the MES, kindling, and other models included in current AED screening programs may result primarily in the development of new, but redundant, drugs that primarily target convulsive (e.g., tonic–clonic) seizures (d’Ambrosio & Miller, 2010). This is a result of current definitions of experimental seizures that often focus on specific types of motor seizures with a defined minimum duration, but tend to ignore short nonconvulsive seizures, which often resemble human complex-partial seizures more than those seizure types used as end points for drug testing. Therefore, during screening of potential AEDs, new agents that may control human complex-partial seizures more effectively than existing AEDs might be missed (d’Ambrosio & Miller, 2010).

Fourth, the typical approach of AED testing in animal models focuses primarily on drug potency and not efficacy. Therefore, different investigational drugs are compared in terms of their anticonvulsant ED50s, that is, the dose-suppressing seizures in 50% of the animals, which is calculated from dose–response curves, testing one group of animals per dose. The lower the ED50, the more potent is the drug, and high potency is often an important argument for selecting drugs for further development. However, it is the antiepileptic efficacy that finally determines the clinical usefulness of a new AED. In animal models, efficacy is more difficult to determine than potency, but one approach is determining ED50s in the 6-Hz model at increasing current intensities (22, 32, 44 mA) as proposed by Barton et al. (2001). At the highest stimulus intensity (44 mA), most first- and second-generation AEDs are ineffective at suppressing the seizures, but a few third-generation AEDs (e.g., retigabine) are effective (Bialer et al., 2009), thus allowing the differentiation of AEDs in terms of anticonvulsant efficacy in this mouse model. Another approach has been illustrated by testing a large series of AEDs in PHT-nonresponding and PHT-responding kindled rats (Löscher, 2006). Interestingly, in these experiments LEV was the only AED that was highly effective in PHT-resistant rats, whereas all other AEDs were significantly less efficacious or not efficacious at all in PHT nonresponders, demonstrating that the PHT resistance in this model extends to various other old and new AEDs (Löscher, 2006), thus simulating multidrug resistance in patients with TLE. One might expect that, based on these preclinical findings, LEV could possibly be more effective than other AEDs in the treatment of drug-resistant epilepsy. Although LEV is often used in previously drug-resistant seizures, such evidence does not exist. Current regulatory trial designs do not evaluate comparative effectiveness of treatments for drug-resistant epilepsy, as discussed below.

Fifth, in the MES and s.c. PTZ models, but also the 6-Hz model now included in the ADD program sponsored by the NINDS (Fig. 4), efficacy is defined by the ability of an investigational drug to limit an evoked seizure in a normal animal (usually mouse or rat). In contrast, clinical seizures associated with human epilepsy evolve spontaneously from an altered CNS substrate. Therefore, one may argue that the pharmacologic sensitivity of evoked and spontaneous seizures differs, so that models with acutely evoked seizures produce false-positive data. However, anticonvulsant potencies of clinically used AEDs are comparable in models of evoked generalized seizures and genetic models of generalized epilepsy, including DBA/2 mice, genetically epilepsy prone rats (GEPRs), and epileptic gerbils (Löscher, 1984, 1999a; Löscher & Meldrum, 1984). In contrast, the focal-onset, secondarily generalized seizures in the kindling model and post-SE models of TLE are typically more difficult to suppress than acutely evoked generalized convulsive seizures in the MES and PTZ models in normal rats, which is in line with clinical experience that partial seizures are more difficult to treat than generalized tonic–clonic seizures (Löscher et al., 1986; Löscher, 1997, 1999a, 2002b). On the other hand, the pharmacology of evoked seizures in the kindling model and spontaneous seizures in post-SE models is similar (Löscher, 2002b). Therefore, the chronic brain alterations associated with kindling and post-SE models change the pharmacology more than the fact of whether a seizure is evoked or spontaneous (Löscher, 2002b).

Sixth, despite the fact that 70% of epilepsy begins in childhood, drug screening is done in adult animals. Although one may argue that age probably does not have a major impact on preclinical screening, it is a concern of pediatric neurologists. One wonders if there are molecules that have different efficacy and tolerability in the immature brain and that are missed with screening in adult animals. Various factors contribute to enhanced seizure susceptibility in the developing brain (Rakhade & Jensen, 2009). An understanding of these factors could yield potential therapeutic targets for new antiepileptic and antiepileptogenic agents.

Seventh, one lesson to be learned from the previous failure of several rational strategies aimed toward more and more selective drugs is that an absolute selectivity for one target may not be desirable for a multifactorial disease such as epilepsy (Löscher & Schmidt, 1994, 2002). Therefore, most clinically efficacious AEDs act by a combination of several mechanisms (e.g., blockade of voltage-dependent sodium or calcium channels, potentiation of GABAergic inhibition, limitation of glutamatergic excitation), so that a “rational” combination of mechanisms in a single drug, for example, by pharmacophore-based drug design, may be a more successful strategy for creating novel broadly acting AEDs than development of highly selective compounds. One interesting example in this regard is ADCI [(±)-5-aminocarbonyl-10,11-dihydro-5H-dibenzo[a,d]cyclohepten-5,10-imine], a tricyclic compound structurally related to the N-methyl-d-aspartate (NMDA) antagonist dizocilpine (MK-801) and the AED CBZ, sharing the mechanisms of these two compounds, that is, blocking both voltage-activated sodium channels and NMDA-receptor–gated channels (Rogawski et al., 1991; Sun & Lin, 2000). The combination of these two mechanisms of action made ADCI (also known as SGB-017) an effective AED in several different animal models of epilepsy, and Neurogen and Wyeth-Ayerst started to develop this drug as a potential treatment for epilepsy, seizures, and stroke (Varming, 1998). However, naturally, previous concepts of designer drugs had to rely on the limited knowledge of the pathophysiology of ictogenesis and epileptogenesis, which may explain the limited success of such concepts. Therefore, the way in which rational approaches are applied to drug discovery is often too simplistic, ignoring the complex alterations of brain functions induced by epilepsy. Furthermore, targeting endogenous processes such as GABAergic or glutamatergic transmission bears the risk of severe adverse effects, which may be enhanced by the brain alterations associated with epilepsy as illustrated by the unexpected adverse effects induced by the competitive NMDA antagonist D-CPP-ene in patients with complex partial seizures (Löscher & Schmidt, 1994).

An alternative strategy to rational drug design is “reverse target development,” in which only anticonvulsant drugs not acting by known mechanisms of AEDs are further characterized, possibly an interesting strategy for the future (Löscher & Schmidt, 2002). For several of the AEDs that have been introduced in recent decades, for example, LEV, retigabine, or GBP, the mechanism of action has been unknown at the time of their development and was identified only later or is still not fully elucidated (Rogawski & Löscher, 2004a; Rogawski & Bazil, 2008). However, even new mechanisms of seizure suppression, such as modulation of the synaptic vesicle protein SV2a by LEV or KCNQ2 potassium channels by retigabine, obviously do not significantly enhance antiepileptic efficacy in previously AED-resistant epilepsy patients. Whether the same is true for future compounds acting by other mechanisms is not known, and, in theory, it might be that there is a strikingly more effective compound among the numerous AEDs that are currently in the clinical or preclinical pipeline (Bialer et al., 2009; Bialer & White, 2010). However, this chance is probably low unless we understand more about the basic mechanisms of pharmacoresistance, thus allowing more rational development of compounds.

Eight, current strategies of AED development search for drugs that symptomatically suppress seizures by diverse mechanisms. As pointed out above, we do not think that anticonvulsant efficacy can be markedly enhanced by any of the new mechanisms of seizure suppression of those numerous investigational drugs that are currently in the AED pipeline (Perucca et al., 2007; Bialer et al., 2009; Bialer & White, 2010). Instead, we strongly believe that progress in the efficacy of AEDs, particular with regard to pharmacologic treatment of drug-resistant epilepsy, will not be made unless and until we develop drugs that specifically target the underlying disease. Indeed, already in 2001, a workshop organized by the NINDS to explore the current problems, needs, and potential usefulness of existing methods of discovery of new therapies to treat epilepsy patients concluded that the epilepsy research community should undergo a conceptual shift to move away from using models that identify therapies for the symptomatic treatment of epilepsy to those that may be useful for identifying therapies that are more effective in the refractory population and that may ultimately lead to an effective cure in susceptible individuals (Stables et al., 2002). To realize the goal of a cure, the molecular mechanisms of the next generation of therapies must necessarily evolve to include targets that contribute to epileptogenesis and pharmacoresistance in relevant epilepsy models.

Which targets may be relevant in this respect? Potential mechanisms of AED resistance in epilepsy have been studied intensively over recent years (Sisodiya, 2003; Löscher & Potschka, 2005; Schmidt & Löscher, 2005; Remy & Beck, 2006; Rogawski & Johnson, 2008; Schmidt & Löscher, 2009). Several major hypotheses have evolved, including (1) the transporter hypothesis, suggesting that overexpressed efflux transporters at the blood–brain barrier (BBB) restrict brain penetration of AEDs, resulting in too low AED levels at their targets sites; (2) the target hypothesis, suggesting that acquired alterations to the structure and/or functionality of target ion channels and neurotransmitter receptors result in loss of AED efficacy; (3) the gene variant hypothesis, suggesting an inherent resistance governed by genetic variants of proteins involved in the pharmacokinetics or pharmacodynamics of AED action; (4) the network hypothesis, suggesting that structural brain alterations and/or network changes (e.g., hippocampal sclerosis) result in resistance; and (5) the intrinsic severity hypothesis, proposing that there is a continuum in severity of the disease, which determines its relative response to medication. The observation that a high frequency of seizures prior to onset of treatment is a prognostic signal of increased severity and future drug refractoriness of epilepsy suggests that common neurobiologic factors may underlie both disease severity and pharmacoresistance (Schmidt & Löscher, 2009). Furthermore, it is important to consider that patients who are resistant to one major AED are often also resistant to various other AEDs, including drugs that act by different mechanisms, which argues against specific target alterations as a major cause of AED resistance. Rather, unspecific mechanisms, such as seizure-induced overexpression of multidrug transporters, most importantly P-glycoprotein (Pgp), at the BBB may be involved in AED resistance (Löscher & Potschka, 2005). Therefore, one strategy to overcome AED resistance would be to combine AEDs with Pgp inhibitors or to develop AEDs that are not substrates of Pgp or other efflux transporters expressed at the BBB. The European Union is currently supporting a “European research initiative to develop imaging probes for early in-vivo diagnosis and evaluation of response to therapeutic substances” (EURIPIDES; cf., that will explore the causes of drug resistance in patients with major neurologic diseases, including epilepsy. A key aim of this initiative is the development of new positron emission tomography (PET) ligands, including 11C-labeled AEDs, to study alterations in functionality and expression of Pgp in epilepsy that could ultimately permit identification of patients who would benefit from coadministration of Pgp inhibitors.

However, it would be naive to believe that one single mechanism, such as Pgp overexpression at the BBB, is responsible for AED resistance in the majority of patients, but rather resistance is due to various factors, which may coexist in the same patient (Schmidt & Löscher, 2009). Experimentally it has been shown that pharmacoresistance is associated with high seizure frequency, AED target alterations, network alterations, and Pgp overexpression in the same epileptic rats, although not all AED-resistant rats display high seizure frequency (Löscher, 2006; Schmidt & Löscher, 2009). Inhibition of Pgp selectively increases AED concentrations in epileptogenic brain regions that share overexpression of Pgp and reduced AED target sensitivity, which may explain the surprisingly high efficacy of the Pgp inhibitor tariquidar to counteract pharmacoresistance in epileptic rats (Brandt et al., 2006). In other words, Pgp inhibition may counteract the consequences of both Pgp overexpression and loss of target sensitivity in epileptogenic brain regions by increasing AED concentrations at target sites. It remains to be established whether this interesting hypothesis can be translated to the clinical arena, but the data illustrate that epilepsy models that allow comparison of AED responders and nonresponders yield valuable insights into mechanisms of AED resistance. A preliminary analysis of PET studies with radiolabeled Pgp ligands in patients with epilepsy performed so far by the EURIPIDES consortium suggests that ≥30% of AED-resistant patients exhibit overexpression of Pgp in epileptogenic brain regions (Löscher, 2010). Whether such patients would benefit from coadministration of Pgp inhibitors will be studied in clinical trials with selective inhibitors such as tariquidar.

Another strategy would be prevention of AED resistance. At least in theory, this could be achieved by treating patients after an epileptogenic brain insult with a drug that, if not preventing epilepsy by an antiepileptogenic effect (which would be the ultimate goal), improves the prognosis of epilepsy by preventing pharmacoresistance. For instance, if neurodegeneration is involved in AED resistance, as suggested by the network hypothesis (Schmidt & Löscher, 2009), a neuroprotective drug may exert favorable effects. An alternative strategy would be developing a drug that prevents progression of epilepsy in patients after first diagnosis of seizures, thereby reducing worsening cases. Numerous studies have tested AEDs and various investigational drugs for antiepileptogenic or disease-modifying effects in rat models of symptomatic epilepsy, and a few of these drugs have also been tested in patients after traumatic brain injury (Temkin, 2009; Löscher & Brandt, 2010). As yet, all clinical and most experimental studies have been negative, but a few experimental reports indicated that (1) brain damage can be minimized by prophylactic treatment with a number of drugs after brain insults, (2) the frequency or severity of spontaneous seizures can be reduced by such treatment, and (3) the behavioral and cognitive consequences of brain insults can, at least in part, be prevented by pharmacologic treatments after brain insults such as SE (Löscher & Brandt, 2010). Epilepsy prevention trials thus far have been limited to a few first- or second-generation AEDs (Temkin, 2009), but, based on promising experimental data, the third-generation AEDs TPM and LEV are currently evaluated for antiepileptogenic or disease-modifying efficacy in patients after traumatic brain injury (Dichter, 2009; Löscher & Brandt, 2010). We presently assess in our laboratory whether neuroprotective treatment after SE prevents or minimizes AED resistance of spontaneous seizures. Overall, the emerging knowledge on the pathologic basis of epilepsy will soon affect the design of new therapeutics (Bialer & White, 2010).

Based on the above considerations, Fig. 5 illustrates a proposal for future preclinical strategies to develop more efficacious AEDs. Frankly, in our opinion, it makes no sense to continue with the development of novel AEDs that are effective in screening models and kindling, but do not have any clear preclinical advantage other than high potency and protective index compared to already used AEDs. New AEDs should be effective in animal models of pharmacoresistant seizures, including models with spontaneous seizures, and, ideally exert not only a seizure-suppressing antiepileptic (anticonvulsant, antiseizure) effect but, in addition, also an antiepileptogenic or disease-modifying effect. A discussion of how this can be achieved goes beyond the scope of the present review, but the interested reader is referred to several recent reviews on antiepileptogenesis in this respect (Löscher & Brandt, 2010; Kharatishvili & Pitkänen, 2010; Pitkänen, 2010). It has been argued that the mechanisms underlying antiepileptic (antiictogenic) and antiepileptogenic or disease-modifying effects may differ (e.g., Klitgaard & Pitkänen, 2003), but several newer AEDs, including TPM and LEV, have been shown to exert both antiepileptic and disease-modifying effects in animal models, which formed the impetus to test whether such drugs modify the development of posttraumatic epilepsy in patients (Dichter, 2009; Löscher & Brandt, 2010). Future antiepileptogenesis trials with investigational drugs in animal models should not only include kindling and post-SE models of TLE but also models in which epilepsy develops after traumatic brain injury. Such models exist, but, as yet, have only rarely been used for testing antiepileptogenesis (Kharatishvili & Pitkänen, 2010). The goal of combining antiepileptic and antiepileptogenic activities in one drug should not exclude alternative strategies for antiepileptogenesis. For instance, paradoxically, several drugs with proconvulsant activity in normal rats, including the cannabinoid receptor 1 antagonist rimonabant and the α2-receptor antagonist atipamezole, have recently been found to exert disease-modifying effects after brain insults, which may indicate that the molecular reorganization that develops after brain insults changes the pharmacology of such drugs (Löscher & Brandt, 2010; Pitkänen, 2010). More studies will be required to determine what properties of these agents—their proconvulsant nature or their mechanisms of action—are the ones that mediate their antiepileptogenic properties.

Figure 5.

Schematic diagram illustrating a modified decision tree for anticonvulsant drug testing. Note that a drug that is solely effective in the initial MES screen is not tested further unless it also exerts efficacy in at least one model of drug-resistant seizures. Compounds that are not active in the MES test are also evaluated in the 6-Hz test and/or other models of AED-resistant epilepsy (see text). It is currently not possible to decide whether the 6-Hz test alone can be used as the go/no go test, because its predictability in regards to pharmacoresistant epilepsy has not been validated. If a compound exerts efficacy in one or more models of pharmacoresistant seizures, advanced studies then include trials of seizure suppression in rats with spontaneous recurrent seizures plus antiepileptogenesis trials, for example, testing whether an investigational drug prevents or modifies the development of spontaneous seizures, neurodegeneration, or behavioral and cognitive alterations after a brain insult such as SE or traumatic brain injury. It should be noted that other, more innovative strategies for AED development, such as drugs targeting specifically epileptogenesis rather than ictogenesis, and drugs working as adjuvants of existing AEDs (for example, Pgp inhibitors) are not addressed in the figure, but are discussed in the text.

With respect to AED-resistant seizures, the decision tree illustrated in Fig. 5 should not solely rely on the 6-Hz model of AED-resistant focal seizures in mice, but should include more laborious models of pharmacoresistant seizures, such as the PHT-resistant kindled rat, the LTG-resistant kindled rat, AED-resistant rats with spontaneous seizures selected from post-SE models of TLE, or the methylazoxymethanol acetate (MAM) in utero model of nodular heterotopia (Löscher, 2006). Of particular interest is the finding that some of these models tend to better mimic the human condition, for example, the presence of AED responders and nonresponders, than any acute seizure model. By testing new drugs in animals that are found to be nonresponsive to existing therapies, we may be able to better predict efficacy in a particular patient population and thereby enrich our chances of identifying a truly novel therapy (Smith et al., 2007). The overall goal should be to get the mechanistically novel and more efficacious drugs into different models of resistance and tested against spontaneous seizures as quickly as practicable. This approach, albeit costly and labor intensive, should be considered when attempting to differentiate a new drug from those that are either on the market or currently undergoing clinical trials. In this respect, it is also important to note that certain adverse (“neurotoxic”) effects of a new compound in kindled or epileptic animals may be more pronounced than adverse effects in normal nonepileptic animals (Löscher & Hönack, 1991; Hönack & Löscher, 1995), so that ADD programs should include adverse effect testing in kindled or epileptic rats to exclude a false-positive estimate of their therapeutic index (Klitgaard et al., 2002).

Beyond the numerous mechanistic possibilities to interfere with processes underlying the initiation and propagation of seizures, we also need fresh provocative ideas on how to develop new AEDs that reliably suppress or cure seizures in the as-yet-resistant patient population. One example is the recent use of proconvulsant drugs for antiepileptogenesis (see above). Another novel idea is to better understand endogenous mechanisms involved in seizure termination and postictal refractoriness, which may lead to novel targets for AED development (Löscher & Köhling, 2010). In addition, it may also be helpful to look for analogies for new innovative therapies in other diseases, such as the use of β-adrenergic receptor blockers (or β-blockers) for treatment of congestive heart failure [for which they were once contraindicated (Doughty et al., 1994)]. Another area of interest is the exploration of potential antiepileptogenic agents such as antiinflammatory or neuromodulatory drugs for control of as-yet drug-resistant seizures (Löscher & Brandt, 2010). The underlying hypothesis is that patients with continuously drug-resistant seizures that have been unresponsive to five or more lifetime AEDs have functional or structural brain changes that may be amenable to antiepileptogenic agents. It is also conceivable that the mechanisms of a drug that works in AED-resistant patients differ radically from those of a drug that exerts anticonvulsant effects in AED-responsive patients. An analogy from another disease is proton pump inhibitor (PPI) refractory gastroesophageal reflux disease (GORD), in which mechanisms causing GORD differ from those in PPI-responsive patients, so that other strategies of treatment have to be used (Liu & Saltzman, 2009).

Proposal for Future Clinical Strategies for Testing New AEDs

What are the main problems of current clinical development strategies that may, at least in part, be related to the fact that current third-generation AEDs are not more effective than older AEDs or that our current regulatory trial design does not allow us to detect differences in efficacy between AEDs, if they exist? And, more importantly, how can we address these problems?

The most pressing problems of current trial design from a clinical perspective include:

  • 1. Lack of comparative effectiveness research. Comparative effectiveness research (CER) is the study of the relative effects of treatments to determine which will be most likely to improve overall health for a specific condition (French & England, 2010). As discussed earlier, we have no clinical trial design to directly determine prior to approval if a new add-on drug is superior in efficacy to a standard add-on drug. The use of intentionally less efficacious placebo controls has received critical attention (see above). Physicians are wary to recommend use of new add-on drugs because of their limited and uncertain impact on improvement of seizures unless they have evidence from trials that the new add-on drug has better efficacy than older drugs. The lack of CER may explain, at least in part, why only about 30% of all patients with epilepsy in Europe receive new AEDs 20 years after marketing and 70% still use first- or second-generation AEDs (Cramer et al., 2010). This concern is not new but, surprisingly, has been largely ignored for more than a decade, as has the next problem (Gram & Schmidt, 1993).

  • 2. Lack of comparative monotherapy trial design for new-onset epilepsy and for chronic epilepsy. The use of intentionally less efficacious low-dose regimens of standard drugs as controls has also received critical attention (Perucca, 2008). We have no clinical trial design to determine directly prior to approval if a new drug is superior in efficacy as monotherapy for new-onset epilepsy or chronic epilepsy versus standard drug treatment.

  • 3. Lack of trial designs to detect antiepileptogenic drugs that affect the underlying pathology or structure of epilepsy.

Novel placebo-controlled add-on trial design

If we have better preclinical approaches resulting in more efficacious drugs, adding a third arm allowing for comparative AED control in patients with drug-resistant epilepsy is an option. A pilot comparative trial, which included a third placebo-arm, has shown noninferiority of PGB versus LTG for refractory focal seizures (Baulac et al., 2010; Fig. 6A). Although a detailed review of comparative trial designs is beyond the scope of this review, we briefly summarize the advantages and disadvantages of comparative trial designs (Table 3).

Figure 6.

Considerations for future comparative trial designs for new-onset and refractory epilepsy. It is worthwhile to note that although both comparative trials indicated in the figure had a noninferiority outcome, Glauser et al. (2010) reported a superiority outcome in new-onset epilepsy (see text and Table 3 for a brief discussion of the pros and cons of comparative trial designs).

Table 3.   Advantages and disadvantages of randomized active-control comparative trials for epilepsy allowing for dose optimization in each treatment arm
Allows for superiority design in double-blind design to detect AEDs that are more efficacious or better tolerated than a standard drug at full dose (Glauser et al., 2010). Does provide a clinically meaningful seizure-free outcome that can directly guide the choice among several drugsNo-difference outcome is inconclusive. Both drugs can be equally ineffective (Leber, 1989). Noninferiority outcome showing that the test drug by a predefined margin is no worse versus standard drug is of limited clinical value (accepted by EMEA but, justifiably, not acceptable to the FDA)
Avoids placebo or pseudoplacebo controls which are not indicated for the treatment of epilepsyMay require large sample sizes for superiority outcome if the expected treatment difference between test drug and standard drug is small, as it is today for most new drugs, and may thus pose substantial logistical challenges
Suitable for new-onset epilepsy and first add-on treatment (Lindberger et al., 2000)Patients with prior exposure to the control/standard drug are not suitable
May provide more realistic efficacy outcome because seizure-free rates are usually for 12 months or more 
Does allow optimal treatment conditions during the trial in terms of patient selection and adjustment of dose 

If the experimental AED is superior in efficacy to the full-dose treatment arm and to placebo after 12 weeks of maintenance treatment, this constitutes compelling evidence for a more efficacious new AED. If the experimental drug is superior in efficacy to only the placebo arm but not superior or only noninferior to the AED arm, we have another traditional add-on AED with well-defined limits of efficacy, an advantage compared to current trial design, which does not allow any characterization beyond stating that the new add-on AED is more efficacious than placebo. Finally, if the efficacy is not superior to placebo, the investigational drug failed the efficacy test. Transfer-to-monotherapy designs in add-on responders would be an option. Although the placebo-controlled plus actively controlled design addresses the issue of whether a new AED is superior versus an established agent and versus placebo, there are difficulties and limitations with such design. Large sample sizes would be required, and it may be difficult to find refractory patients who were never exposed to standard AEDs such as LEV or CBZ. Moreover, studying only one fixed dose in each of the two active treatment arms would make the comparison of efficacy/effectiveness dose-specific, and of less relevant to clinical practice. Adding more doses, or allowing dosing flexibility in a placebo-controlled trial could be logistically prohibitive. The scenario of substituting the placebo-controlled add-on trial with an active-control add-on trial aiming at superiority over an established agent might even place a hurdle for the drug development process.

Comparative monotherapy trial design for new-onset epilepsy

If we had more efficacious new drugs for evaluation of monotherapy in newly treated patients we could compare the experimental AED to best current AED treatment at full clinical dose in a superiority design employing a double-blind randomized retention design over 36 months. The power of the study should be able to detect a clinically relevant difference, if it exists. Experimental agents that show a noninferiority outcome to standard therapy are labeled as such (Brodie et al., 2007; Fig. 6B). The substantial issues of noninferiority outcome have been discussed above. Pricing of the new noninferior AED could be similar to that of the old drug comparator unless an added benefit is shown otherwise. It is worthwhile to note that current guidelines in both the United States and in the European Union already allow for comparative effectiveness designs for monotherapy in newly diagnosed patients.

Antiepileptogenic drug trial design

Although there is an unmet need for agents that affect the pathophysiology of the underlying epilepsy, we are just at the doorstep of developing such agents in preclinical experiments (Löscher & Brandt, 2010). By definition, an antiepileptogenic compound should prevent epilepsy in patients at risk (Temkin, 2009), whereas a disease-modifying compound should have the ability to improve the long-term clinical course of the epilepsy.

Even if we had an agent that is antiepileptogenic or disease-modifying in animal models, clinical trials of experimental antiepileptogenic or disease-modifying agents are fraught with a number of difficulties (Dichter, 2009; Temkin, 2009; Löscher & Brandt, 2010). In trials of epilepsy prevention after traumatic brain injury, the choice of a suitable patient population, the logistics of targeting the drug within the therapeutic window after injury, and the need for long-term exposure are crucial. Posttraumatic epilepsy may be associated with alcohol or recreational drug abuse, which renders trial management more difficult. Nevertheless, proposals have been made for trial design in posttraumatic epilepsy (Dichter, 2009). Epilepsy after initial SE could be another population for testing the ability of an antiepileptogenic agent to prevent epilepsy. However, the disadvantages of poststatus studies are numerous and include the mortality and morbidity of the underlying etiology of SE (and the SE itself) and the relatively small risk of developing poststatus epilepsy in humans (which incidentally is in marked contrast to more frequent poststatus seizures in rodents), and render it more difficult to detect the intrinsic tolerability and safety of the test compound. A perhaps more suitable population could be patients with nontraumatic spontaneous hemorrhage, particularly small lobar hemorrhages, which causes seizures in up to 50% (Faught et al., 1989). Finally, hospitals taking care of stroke patients have expertise for providing immediate care after injury, which could include an antiepileptogenic agent and placebo-control. Innovative trial designs to assess the antiepileptogenic activity of AEDs should include serial magnetic resonance imaging (MRI) to assess changes in the brain structure during epileptogenesis (Löscher & Brandt, 2010). A further difficulty is the need for long-term follow-up of 1 year or more to fully assess the effects of treatment that may possibly be as short as 1 month, although the duration of exposure may possibly vary from compound to compound and also depend on how quick treatment can be started after injury.


The main conclusions of the present review are:

  • 1. Some AEDs of the third generation have advantages (similar efficacy, lower risk of hypersensitivity reactions, lower risk of drug interactions, and apparently fewer AED-associated diseases including teratogenicity) over CBZ and VPA, and are often used for these reasons. In particular LEV seems to have advantages over other new AEDs and is probably the most often used third-generation AED. However, there is no compelling evidence that third-generation AEDs, including LEV, have made clinically relevant advances in the efficacy and the day-to-day tolerability of current drug treatment for epilepsy. The lack of improved efficacy and tolerability with the third-generation of AEDs is astounding and dictates new strategies.

  • 2. The reasons for the apparent lack of progress of AEDs are manifold and include problems in the preclinical and, to a lesser degree, in the clinical development of drugs for epilepsy.

  • 3. There will be no easy solution if preclinical drug testing does not deliver more efficacious AEDs, and if clinical development imposes ethical and logistical challenges by trying to maintain assay sensitivity.

  • 4. Therefore, we cannot continue with the current ways of preclinical and clinical drug testing and development but need new innovative strategies.

  • 5. In the preclinical arena, it makes no sense to continue with the development of novel AEDs that are effective in screening models (particularly the MES and s.c. PTZ tests) and kindling, but do not have any clear preclinical advantage other than high potency and protective index compared to already used AEDs.

  • 6. Instead, we need to include models for AED-resistant seizures and should search for drugs that do not only symptomatically suppress ictogenesis but interfere with the processes underlying epilepsy.

  • 7. For this goal, we need imaginative scientists with technical expertise in animal models and radically new ideas in AED discovery and development.

  • 8. The pharmaceutical industry should be aware of the complexity of the issue, which goes far beyond the use of simple seizure models for assessing anticonvulsant activity.

Although we probably need better drugs more than we need better trials, the current strategies for regulatory trial design are not well suited to answer clinically valid questions.

  • 1. Physicians, patients, and cost bearers want to know if the new AED that is entering the market has better efficacy or another valid clinical benefit compared to an older drug.

  • 2. We need to address two major problems in clinical testing of new AEDs. One is that we need to look for alternatives to continue testing new AEDs, often repeatedly, in patients with continuously drug-resistant seizures who, based on their treatment history, cannot become seizure free in any significant proportion.

  • 3. More importantly, we still use intentionally less-efficacious controls to maintain assay sensitivity, which is important. However, at present, that effectively prevents the assessment of comparative efficacy of the test AED versus standard treatment prior to approval.

If we had more effective drugs, we could abandon clinically questionable regulatory trials with less-efficacious controls including placebo or low-dose treatment and noninferiority designs, and enter the era of comparative effectiveness trials. It would be naive, however, to ignore the overwhelming logistical obstacles to performing comparative testing if we do not have available more efficacious experimental AEDs. The current failure to provide clinicians and patients and cost-bearers with AEDs that are proven to be better than earlier drugs at the time of approval is at the core of the dilemma with new AEDs.

Finally, the current dilemma with new AEDs should not result in assigning blame, for example, clinicians blaming experimental scientists for creating or using poor models, while we are all forgetting that some of the issues that we are discussing today have long been recognized but we have neglected to act on them. Instead, the current dilemma offers a chance to radically overhaul the current strategies and concepts of preclinical and clinical AED development. In this sense, the current review can only provide thought-provoking impulses, at best, but not final solutions for our dilemma with new AEDs.


We thank Dr. Michael A. Rogawski und Dr. Gregory L. Holmes for critical reading of an early version of the manuscript.


Dr. D. Schmidt has declared consulting or speaking fees over the last 12 months from the following companies: UCB, Lundbeck, Desitin, Grünenthal, Novartis, and GSK. Dr. W. Löscher received consulting or speaking fees over the last 12 months from the following companies: UCB, Pfizer, Lundbeck, Grünenthal, and Schering-Plough. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.