Identification of new treatments for epilepsy: issues in preclinical methodology

Summary

Preclinical research has facilitated the discovery of valuable drugs for the symptomatic treatment of epilepsy. Yet, despite these therapies, seizures are not adequately controlled in a third of all affected individuals, and comorbidities still impose a major burden on quality of life. The introduction of multiple new therapies into clinical use over the past two decades has done little to change this. There is an urgent demand to address the unmet clinical needs for: (a) new symptomatic anti-seizure treatments for drug-resistant seizures with improved efficacy/tolerability profiles, (b) disease modifying treatments that prevent or ameliorate the epileptogenic state, and (c) treatments for the common comorbidities that contribute to disability in people with epilepsy. New therapies also need to address the special needs of certain subpopulations, i.e. age- or gender-specific treatments. Preclinical development in these treatment areas is complex due to heterogeneity in presentation and etiology, and may need to be formulated with a specific seizure, epilepsy syndrome or comorbidity in mind. The aim of this report is to provide a framework that will help define future guidelines that improve and standardize the design, reporting, and validation of data across preclinical anti-epilepsy therapy development studies targeting drug-resistant seizures, epileptogenesis and comorbidities.

Introduction

Epilepsy affects 50 million people worldwide (World Health Organization 2006), with an estimated 2–3 million living in the United States (Epilepsy Foundation of America, Hirtz et al. 2007), 6 million in Europe (World Health Organization 2010), and at least 40 million in the developing world (World Health Organization et al. 2005). Epilepsy poses a significant burden on the quality of life of affected individuals and their families. Since the introduction of bromide as an anti-seizure drug in 1857, there has been an impressive expansion of therapies that are clinically effective in decreasing the frequency and severity of seizures in people with epilepsy. This class of symptomatic treatments is widely referred to as “antiepileptic drugs” (AEDs). In this article, we will avoid this term and use instead “anti-seizure drug”, to prevent confusion with disease-modifying therapies that have a sustained modulatory effect on the underlying epileptic state (i.e. the predisposition to generate spontaneous recurrent seizures) and/or with treatments that ameliorate associated comorbidities. The newer anti-seizure drugs have been identified through systematic screening in batteries of an increasing number of in vivo and in vitro seizure and epilepsy models (Loscher and Schmidt 2011). Undoubtedly, the clinical availability of a broad range of anti-seizure treatments has significantly improved the management of the disorder. At present, two thirds of all individuals with epilepsy will achieve seizure freedom with available medications. This translates into better quality of life and reduces the risk of seizure-associated injuries and death. However, a third of people with epilepsy will not have adequate seizure control with the current medications. For these patients the situation has improved very little in the last few decades. In addition, current screening methods have failed to elucidate which drugs are more or less likely to produce clinically significant adverse effects that may impair the quality of life or limit dosing to levels insufficient to completely control seizures. Other important concerns are the risks related to drug-drug interactions and the potential for teratogenicity, which may limit the use of effective anti-seizure medications in women of child bearing potential.

There is an urgent need for more effective and better tolerated treatments to control drug-resistant seizures, as well as for innovative therapies to prevent, stop or reverse the development of epilepsy and epilepsy-related comorbidities (White 2003, Smith et al. 2007, Jacobs et al. 2009, Loscher and Schmidt 2011). Such treatments may include not only individual pharmacological compounds or combination therapies but also devices and other novel therapeutic interventions. Here we will use the general term anti-epilepsy treatment (AET) to include all these types of treatments. Where appropriate, more specific terms will be used to indicate the treatment indication, adopting the following definitions modified from those recommended by Pitkanen (Pitkanen 2010).

1. Symptomatic treatment

• Anti-seizure treatment: A treatment that stops or reduces the frequency or severity of seizures, irrespective and potentially independently of the underlying epileptic state or disease progression. Although this term may not be easily translatable to languages other than English, it serves to differentiate it from treatments for epileptogenesis.

• Anti-comorbidity treatment: A treatment that alleviates or reverses the symptoms related to various comorbidities of epilepsy, such as neurocognitive deficits, neuropsychiatric conditions and cardiovascular events.

2. Disease-modifying treatments

These treatments alter the development or progression of epilepsy, comorbidities, and also the associated pathology. They include:

• Anti-epileptogenesis treatments: When an anti-epileptogenesis treatment is given prior to epilepsy onset, it prevents or delays the development of epilepsy. If seizures occur, they may be fewer in frequency, shorter, or of milder severity. When such a treatment is given after the diagnosis of epilepsy, it can alleviate seizure severity or prevent or reduce the progression of epilepsy, or change the seizures from drug-resistant to drug-sensitive. Cure is achieved when there is a complete and permanent reversal of epilepsy such that no seizures occur after treatment withdrawal.

• Comorbidity-modification: A treatment that alleviates or reverses the symptomatic development or progression and also the associated pathology related to various comorbidities of epilepsy.

The validation of new AETs that address currently unmet clinical needs is a multistep process, requiring rigorous testing through all stages of development, using experimental models of seizures or epilepsy (Table 1). While these models have provided new candidate molecular targets for drug-resistant seizures, epileptogenesis and certain comorbidities (Loscher and Potschka 2005, Loscher 2007, Pitkanen 2010, Pitkanen and Lukasiuk 2011), there is concern that these findings may not translate into clinically meaningful interventions. This situation poses a significant burden and barrier to investment, as the costs associated with development of a new anti-seizure treatment currently approximates $1 billion.

Table 1. Representative rodent models of seizures, epileptogenesis, and epilepsies in preclinical studies.

(Smyth et al. 2002, White 2003, Pitkanen et al. 2006, Velisek et al. 2007, Lee et al. 2008, Cortez et al. 2009, Hunt et al. 2009, Marsh et al. 2009, Price et al. 2009, Dube et al. 2010, Dyhrfjeld-Johnsen et al. 2010, Kharatishvili and Pitkanen 2010, Scantlebury et al. 2010, Eastman et al. 2011, Faingold et al. 2011, Raffo et al. 2011, Rakhade et al. 2011). This table is largely based on the reported use of the models in the literature, not on the demonstrated suitability of these models for the indicated purposes of AED testing.

Model of seizures or epilepsy	Seizure type	NIH AED screening program	Used in testing effects in drug-resistant seizures or epilepsy	Used in testing anti-epilepto-genesis effects	Used in testing effects on comorbidities	Used in early life AET studies
Maximal electroshock seizures	Tonic clonic	Yes	No	No	No	No
Pentylenetetrazole model	Clonic	Yes	No	No	No	No
Low dose pentylenetetrazole model	Generalized absence	No	No	No	No	No
Bicuculline, picrotoxin models	Clonic	Yes	No	No	No	Yes
Flurothyl model	Clonic	No	No	No	No	Yes
Spike-wave discharge models:	Generalized absence
- Genetic Absence Epilepsy Rats of Strasbourg,		No		Yes	Yes	No
- WAG/Rij, γ-butyrolactone		No		Yes	Yes	No
Monogenic mouse models of absence epilepsy	Generalized absence	No
Tetanus toxin	Focal	No	Yes	Yes		No
Audiogenic seizures	Wild running (focal onset); Generalized tonic-clonic	Yes	No	No	Yes	No
Electrical kindling (corneal, hippocampal, amygdala)	Focal	Yes	Yes	Yes		Yes
Lamotrigine-resistant kindled rats	Focal	Yes	Yes	Yes		No
6 Hz electrical stimulation (32 and 44 mA)	Focal	Yes (32 mA)	Yes (44 mA)	No	No	No
Models of dysplasias:
Methylazoxymethanol-induced heterotopias (in two-hit models)	Focal	No	Yes	Yes		No
Post-status epilepticus spontaneous seizures
Chemical induction (kainic acid, (lithium)-pilocarpine)	Focal onset, limbic	Yes	Yes	Yes	Yes	Yes
Stimulation (continuous hippocampal, perforant pathway, sustained amygdala stimulation)	Focal onset, limbic	No	Yes	Yes	Yes	No
Traumatic brain injury
Cortical undercut	Focal onset	No				No
FeCl2	Focal onset	No				No
Fluid percussion	Focal onset	No	Yes	Yes		No
Controlled cortical impact	Focal onset	No				No
Early life epilepsy syndromes
Hypoxia ± ischemia	Hypoxic ± ischemic	No	Yes	Yes		Yes
Febrile seizure/Hyperthermia models	Febrile seizure	No				No
Infantile spasms models (ie. Multiple-hit)	Infantile spasms	No	Yes	Yes	Yes	Yes
Transgenic rodent models	Genetic	No		Yes		No
In vitro models		Yes		Yes	No	Yes

The purpose herein is to provide a framework for a discussion that will lead to recommendations for improved design, analysis, and reporting of preclinical AET studies. The Working Group on Recommendations for Preclinical Epilepsy Drug Discovery of the Neurobiology Commission of the International League Against Epilepsy (ILAE), the Basic Science Committee of the American Epilepsy Society (AES), and the US National Institute for Neurological Diseases and Stroke (NINDS), and the have spurred initiatives to optimize the bench-to-bedside translation of preclinical research to develop new AETs. In this document, we highlight pitfalls in translating preclinical data to successful clinical treatments for epilepsy, comparing them with the experience from other disciplines. Proof of principle experimental studies and projects aimed at identifying new mechanisms and targets for new therapies are not included in the present commentary. We will focus on preclinical assessments of treatments that are intended to be ultimately translated into human trials.

Improving reproducibility of preclinical data: lessons from other disciplines and epilepsy-related challenges

A major concern highlighted in other neurological and non-neurological disease areas is the poor reproducibility of preclinical data for compounds progressing from academic laboratories to industrial development programs and, ultimately, to clinical trials (Stroke Therapy Academic Industry Roundtable (STAIR) 1999, Ioannidis 2005, Benatar 2007, Fisher et al. 2009, Mullard 2011, Prinz et al. 2011). Given the high cost of clinical drug development, factors such as low reproducibility and translatability, or heterogeneity in study design that hinders the comparison of preclinical data are major disinsentives for investment in developing novel treatments (Mullard 2011). The reasons for these obstacles are multiple and varied, but methodological issues related to the design, execution and reporting of preclinical studies are important components. For example, methodological pitfalls identified by a meta-analysis of preclinical studies using the superoxide dismutase 1 (SOD1) transgenic mouse model of amyotrophic lateral sclerosis (Benatar 2007) included lack or insufficient study blinding, small sample sizes, initiation of treatments at a presymptomatic stage which may not be clinically relevant, publication bias favoring positive studies, measures of statistical significance with questionable clinical relevance, and failure to address issues related to translation of findings to the clinical setting.

In response to similar concerns within the spinal cord research field, the NINDS contracted a replication program where 10 replication studies were conducted in an attempt to validate the preclinical data. The majority of these data could not be replicated (Pinzon et al. 2008, Steward et al. 2008). In the stroke research field, guidelines and criteria have been proposed to identify data quality problems and offer directions for future studies (Stroke Therapy Academic Industry Roundtable (STAIR) 1999). The STAIR report outlines recommendations for preclinical study design, emphasizing the importance of sample size calculations, preset inclusion and exclusion criteria, group allocation and blinding, appropriate reporting of excluded animals as well as of conflicts of interest (Fisher et al. 2009). The report provided a set of minimal criteria to be met before selecting a candidate treatment for clinical testing. The example of the putative neuroprotectant NXY-059 which failed in a randomized multicenter clinical trial, despite successful preclinical testing in accordance with the STAIR guidelines, demonstrates that guidelines may not necessarily guarantee success in drug development (Ginsberg 2007). However, they may help in teasing out the factors that contribute to discordant results between preclinical and clinical studies, if such studies follow consistent designs and methodology.

In an attempt to address some of these issues with preclinical studies of drugs for neurological diseases, the “Rigor in Science Working Group” of the NINDS has recently released an essential list of “Points to consider for improving the quality of NINDS-Supported Preclinical and Clinical Research through Rigorous Study Design and Transparent Reporting” (http://www.ninds.nih.gov/funding/transparency_in_reporting_guidance.pdf) (Table 2).

Table 2. NINDS recommended points to consider for “Improving the Quality of NINDS-Supported Preclinical and Clinical Research through Rigorous Study Design and Transparent Reporting”.

(http://www.ninds.nih.gov/funding/transparency_in_reporting_guidance.pdf)

Experimental design	Rationale for the selected models and endpoints (animal and/or cellular) Adequacy of the controls Route and timing of intervention delivery/dosing Justification of sample size, including power calculation Statistical methods used in analysis and interpretation of results Entrance criteria for compounds being screened
Minimizing bias	Methods of blinding (allocation concealment and blinded assessment of outcome) Strategies for randomization and/or stratification Reporting of data missing due to attrition or exclusion Reporting of all results (negative and positive)
Results	Independent validation/replication, if available Robustness and reproducibility of the observed results Dose-response results Verification that interventional drug or biologic reached and engaged the target
Interpretation of results	Alternative interpretations of the experimental data Relevant literature in support or in disagreement with the results Discussion of effect size in relation to potential clinical impact Potential conflicts of interest

Epilepsy differs from many other neurological conditions due to its extreme heterogeneity in etiologies and phenotypes. Even in clinical trials, the reproducibility of efficacy results can vary across studies targeting the same seizure type, possibly due to population differences in placebo response, and genetic, societal or biological factors, including etiologic and diagnostic heterogeneity (Sperling et al. 2010). To optimize translation of preclinical results to the clinical setting, it would be helpful if the precisely defined terminology and criteria were used in both phases of development (Tables 3 and 4).

Table 3.

Comparison between preclinical and clinical trials

	Preclinical trials	Clinical Trials
Subjects
Species	Rodents (less frequently, and for anti-seizure drug studies: cats, dogs and monkeys)	Humans
Genetic background	Less heterogeneous	Very heterogeneous
Sex	Not always accounted for, often only male	Both
Age	Homogeneous within-study Heterogeneous across studies	More heterogeneous within study Heterogeneous across studies
Etiology/pathology	Etiology is more uniform Pathology is more uniform (within same study) Post-insult observation/treatment often not done/reported (e.g. oxygenation, pH, CO2, electrolytes, T, etc.) Usually after initial precipitating event (unless genetic or inbred strain)	Etiology can be uniform or variable Often variable underlying pathology Post-insult observation/treatment can be available Identifiable initial precipitating event or genetic susceptibility are less common
Living conditions	Uniform	Variable
Past medical history	Often considered as normal prior to seizure induction, unless there is a genetic predisposition or prior lesion	Variable
Co-administered drugs	Typically none	Variable
Seizure/epilepsy history
Age at onset	Homogeneous	Heterogeneous
Duration	More homogeneous	Heterogeneous
Epilepsy history	New onset seizures Epileptic animals (with or without history of drug-resistance)	Variable
Prior/concomitant AET history	Usually none Prior exposure in some models of pharmacoresistance	Patients typically receive or have received other AEDs
Experimental design
Comparator	Vehicle controlled Active control	Placebo controlled Active treatment controlled Historical controls
Presence of other AEDs	No (Monotherapies)	Yes (Add-on therapies) No (Monotherapies: de novo or conversion)
Blinding	Inconsistent	Common
Timing of administration	Pre-symptomatic phase (most often) Symptomatic phase (rare; early or late)	Usually symptomatic phase (often late)
Outcome assessments	Variable (model, study, design, mechanism-related) Often different from clinical outcomes	More standardized
Power analysis details	Rarely included	Generally included
A priori inclusion/exclusion criteria clearly written	Rarely reported	Generally provided
Replication studies	Not as frequent Methodology may vary	Frequent
Publication bias
Negative vs positive outcomes	Bias towards publishing positive studies	Bias towards publishing positive studies
Across-studies comparisons	Difficult due to heterogeneous design and outcomes Meta-analyses essentially impossible	More common Meta-analyses sometimes feasible

Table 4.

Some pitfalls in preclinical study design and translation to clinical trials

Parameter	Preclinical study design	Implications for transition to clinical trials
Anti-seizure studies
Timing of administration	Before seizure induction in normal animals.	First seizure is not predictable. Although existing seizure models have demonstrated value in anti-seizure development, models allowing post-treatment should also be considered.
	After epilepsy onset in epileptic animals.	Consistent with clinical practice.
Pharmacokinetics	Plasma concentration of parent compound and/or active metabolites.	Can therapeutic levels be identified? What is the minimally effective/maximally tolerated dose? Can toxicity be predicted?
	Tissue concentration of parent compound and/or active metabolites.	Does it correlate with plasma levels? What are the implications for plasma level monitoring?
	Variable doses and duration of drug across studies.	Uniformity across studies would facilitate validation.
Drug-to-drug interactions	Done late in preclinical development with selected tests (i.e. interactions with cytochrome P450 or transporters).	Could interacting drugs (i.e. enzyme inducers) limit the efficacy of the anti-seizure therapy in clinical trials?
Efficacy outcomes	Comparisons with vehicle-controlled group.	The efficacy of the drug may be different in epilepsy patients already treated with other anti-seizure drugs.
	Vehicle-controlled group outcomes may vary across studies.	Placebo-controlled group outcomes may also vary across clinical studies. Sample size estimates should be adjusted. Replication studies may be useful.
Disease modification: Anti-epileptogenic studies
Timing of administration	Before epilepsy induction in normal animals with a recognizable initial insult.	Initial insult is not predictable or may not always be identifiable in humans.
	Before epilepsy onset in genetically predisposed animals.	Feasible in humans. Biomarkers for therapeutic window would be desirable to reduce unnecessary exposure.
	After initial precipitating event in latent period.	Feasible in humans. Biomarkers for therapeutic window would be desirable to reduce unnecessary exposure.
	After onset of epilepsy.	Consistent with clinical practice.
	Variable across studies.	Uniformity in doses across studies would facilitate validation of the anti-epileptogenic treatments.
	Is there a therapeutic time window?	Are there relevant timepoints/biomarkers to identify a similar time window in humans?
Pharmacokinetics	Plasma concentration of parent compound and/or active metabolites.	Can they predict efficacy/toxicity? What is the minimally effective/maximally tolerated dose? Are active metabolites similar in humans?
	Tissue concentration of parent compound and/or active metabolites.	Does it correlate with plasma levels? What are the implications for plasma level monitoring and determination of “wash-out” period?
	Variable doses and duration of drug treatments across studies.	Uniformity across studies would facilitate validation.
Drug-to-drug interactions	Done late in preclinical development with selected tests (i.e. interactions with cytochrome P450 or transporters).	Could interacting drugs (i.e. enzyme inducers) limit the efficacy of the anti-epileptogenic treatments in clinical trials?
Seizure outcomes	Comparisons with vehicle-controlled group.	The efficacy of the drug may be different in epilepsy patients already treated with other anti-seizure drugs.
	Lack of uniformity in seizure definitions.	More uniform criteria would facilitate validation. Do they correspond to the clinical criteria? More uniform and clinically-relevant criteria need to be agreed upon for seizure detection/classification.
Anti- comorbidity studies (symptomatic or disease modification)
Cognitive/neurodevelopmen tal outcomes.	Limited	Extensive. Are the selected testing batteries relevant to human outcomes?
Handling, living conditions	More uniform.	In humans, environmental/societal influences are usually extensive, with potential to further modify outcomes.
Outcome assessment	Cognitive and neurodevelopmental assessment batteries are customary to lab practices.	More uniform protocols would facilitate validation. How do they relate to human clinical testing methods?
	Variability in timepoints of outcome assessment across studies.	More uniform protocols would facilitate validation.
	Variability in phenotype (across study, experimental protocol, strain etc).	May hinder validation of disease modifying treatments. How does it relate to human phenotype?
	Lack of uniformity in pathology endpoints.	More uniform criteria would facilitate validation. Do the specified pathology endpoints in preclinical studies have clinically-applicable biomarkers to monitor progression?
All 3 study types
Model of seizures/epilepsy	Variable method of induction.	Which seizure/epilepsy does the model target? Is the AET effect sustainable across same seizure models?
	Variability in phenotype across studies.	Does it influence negative/positive outcomes? Sample size estimates should be adjusted. Modifiers of phenotype should be accounted for.
	Usually after initial precipitating event (unless genetic or inbred strain)	Initial precipitating event or genetic predisposition is not always identifiable. Does this alter outcomes?
Adverse effects	Often limited.	More detailed reporting is encouraged. Human adverse effects are not always predicted by preclinical data.

This article briefly discusses some of the challenges that this goal will present. It is important to note that, unlike other neurological disease areas mentioned above, the failure of AETs introduced over the last two decades to significantly impact unmet clinical needs in the epilepsy population does not necessarily represent a failure in translating preclinical results to the clinical setting. Recent drugs have shown substantially similar anti-seizure profiles to those of older generation drugs in preclinical models, and similar efficacy profiles in clinical trials. Nonetheless, valuable lessons learned from other neurological diseases have important implications for the preclinical translation of potential new AETs.

Preclinical AET development: the need for rigorous standardized study practices

Crucial aspects in preclinical AET development include: (a) study design and execution, (b) unbiased reporting of both positive and negative outcomes, (c) replication studies, and (d) translation of data into clinical trials. There is no ideal design and protocol that is suitable for all purposes. Protocols may need to be adapted to account for the mechanism of action of the AET (where known), etiology and features of the seizure/epilepsy syndrome studied, specific targeted population, and the utilized endpoints. However, standardizing designs, testing paradigms and outcome assessments across preclinical AET studies for the same indication, and anticipating their clinical applicability, will be needed to assess the replicability of data. Certain aspects of these steps are common for all treatment goals, whereas others relate to the specific objectives. Below, and in Tables 3 and 4, we outline problems related to these issues. Our objective is not to provide definitive guidelines, but rather a framework for discussion leading to a future set of recommendations that will be widely shared by the scientific community.

Issues to be considered in AET development, in general

Species/strains/biological and genetic factors

Accurate prediction of a drug’s efficacy and toxicity from preclinical data can be limited by inter-species/strain differences in seizure threshold, pharmacodynamics and pharmacokinetics (Frankel et al. 2001, Baillie and Rettie 2011). Consideration should be given to whether preclinical studies should examine effects in both sexes and different age groups, as deemed appropriate for the specific epileptic syndrome. Genetic background may influence not only the susceptibility to seizures and epilepsy, but may also modulate the consequences of an initial precipitating insult. When considered clinically relevant, the effects of different genetic substrates could be investigated in preclinical AET development.

Models

Ideally, animal models used for preclinical drug development should: (a) allow identification of the components of the seizure type and epilepsy syndrome that are relevant to the human condition, (b) address the age, sex or etiology-specific features of the human syndrome, (c) manifest comorbidities or pathologies that are characteristic of the human condition and relevant to goals of the study, and (d) allow for monitoring of outcomes using reliably quantifiable and clinically-relevant endpoints or disease biomarkers. There are no animal models that fulfill all of these characteristics. However, studies carried out on existing models can still yield useful proof of principle data, provided that the model displays phenotypic features that the proposed treatment is expected to address. Discussion of how the advantages and limitations of any proposed model provide information that will be useful in the design of future clinical studies would be helpful.

Minimization of bias

To minimize bias, randomization and blinding are desirable. In some circumstances, such as presence of specific abnormal behavioral phenotypes or tissue pathology, complete blinding is not possible and this should be discussed when reporting the data. Inclusion and exclusion criteria may influence applicability of the results, and thus may also be a source of bias if they are not predefined and applied by an investigator blinded to the group assignment. Attention to a rigorous, predefined, statistical analysis plan, including sample size calculations, can also help minimize bias.

Monitoring and outcomes

Outcomes that are appropriate for the study goals and can be assessed by quantitative, unbiased methods have obvious advantages. Video-EEG or equivalent biomarkers are currently considered the standard methods for the assessing seizure outcomes in models of epileptogenesis and spontaneous epilepsy, but not necessarily for anti-seizure studies utilizing electrical or chemically induced seizures, where the threshold required to trigger a seizure is a primary outcome measurement, nor in neonatal rodents when the skull size may not permit the placement of EEG electrodes. The optimal duration and timeline of monitoring to assess a treatment’s effect is also variable, and may depend on the specific objectives. Consideration or discussion of possible confounding variables such as age, sex, or other experimental parameters would be important, as well as reporting adverse outcomes (mortality, toxicity etc).

Dose-response experiments

Assessment of dose-response curves, preferably using routes of administration that could be easily applied in humans, and identification of minimally effective and maximally tolerated doses is helpful for the design of future clinical studies. Investigating pharmacokinetic/pharmacodynamic (PK/PD) relationships, including the rate and extent of brain penetration, can also facilitate the design of proof-of concept studies in the clinic. For non-pharmacological AETs, dose-response experiments may need to be adapted to the specific modality.

Replication, across model validation, and reporting

In principle, replication studies using the same model or different models of the same epilepsy syndrome or seizure type are essential for determining whether to proceed to human investigation. For example, the demonstration that a new AET shows efficacy in more than one model relevant to the same seizure or epilepsy type or comorbidity would strengthen the evidence supporting a decision to progress to clinical development. Preclinical replication studies could be facilitated by organizing a system whereby independent preclinical AET testing centers are supported to undertake replication of promising preclinical data from academic laboratories. This system would need to address issues such as funding sources, protocol standardization, intellectual property rights, financial and non-financial conflicts of interest, objectivity in testing and analysis, as well as scientific merit (Table 4). Publication bias is arguably even more of a problem for preclinical than clinical studies, as it is very difficult to publish studies in animal models that produce negative results. A system of registration for preclinical studies, as currently exists for clinical trials, would be helpful in minimizing this bias.

Regulatory agencies requirements

Familiarization with the preclinical regulatory requirements, as established by the regulatory agencies of the country where preclinical research is conducted, can strengthen preclinical AED development research and anticipate issues to be addressed in an Investigational New Drug (IND) application. Relevant guidelines relate to good laboratory practices (GLP), good manufacturing practices (GMP), methods and procedures for animal PK/PD and toxicology studies, as well as assays for quantitation of drugs and metabolites. The investigators are referred to publications of the Food and Drug Administration (FDA) in the U.S.A., the European Medicines Agency (EMA) in Europe, the Japanese Regulatory Agency, or specific regulatory bodies of the individual country where research is conducted.

Conflicts of interest

Potential intellectual property and financial conflicts of interests for investigators involved in preclinical studies should be stated in publications.

Issues related to development of new anti-seizure treatments

There is a pressing need for new anti-seizure treatments that can control drug-resistant seizures and/or offer a better tolerability/efficacy profile for specific target patient populations. To date, preclinical anti-seizure drug development has primarily been carried out using a few well-characterized acute or chronic animal models of seizures or epilepsy (Table 1). Yet, the reiteration of the same algorithms and models might create a bias towards identifying anti-seizure therapies with indications and efficacy profiles similar to those of currently clinically available drugs (Coatsworth 1971, Perucca et al. 2007, Perucca 2010, Loscher and Schmidt 2011). Validating an anti-seizure therapy in an adult animal may not necessarily guarantee its efficacy in neonates or infants, where brain physiology and connectivity are still immature and the underlying pathophysiology may affect anti-seizure efficacy. Expanding the range of seizure models used in preclinical testing to include models of drug-resistant seizures or specific populations (i.e. age, gender, epilepsy syndrome etc) could be useful.

Minimizing differences in preclinical and clinical AET study designs could also be important in improving the clinical translation of preclinical study results (Tables 2 and 3).

Issues related to development of anti-epileptogenic therapies

Compared with preclinical anti-seizure drug screening, studies on anti-epileptogenesis effects are expensive and time-consuming because they require prolonged video-EEG monitoring and large sample sizes to account for significant inter-animal variability in seizure frequency, latency to onset, and other phenotypic endpoints. Consequently, it is imperative to ensure that the study design and utilization of data are appropriate, not only to minimize costs but also to facilitate comparisons across studies. Optimizing the design may also reduce the number of inconclusive studies and ultimately reduce the number of laboratory animals required.

Models

Most of the currently used chronic epilepsy models are models of adult genetic generalized or acquired limbic epilepsy (Table 1). Epileptogenesis however is unlikely to be a single process across epilepsy syndromes. More models representing other important epilepsy syndromes and developmental ages should be developed and rigorously characterized across multiple laboratories to facilitate the identification of syndrome-appropriate anti-epileptogenic therapies.

Early life epilepsies

There are major unmet needs in early life epilepsies. Most preclinical development has focused on adult animal models, partly to avoid confounding effects of developmental processes. Yet human epilepsies often begin at very young ages and, by the time these children mature into adults, the effects of seizures, therapies, and comorbid conditions have altered brain biology and functioning, and consequently the way epileptogenesis has progressed. How applicable for young patients are therapies discovered using adult-onset epilepsy models? Validation and utilization of the emerging early life models of epileptogenesis and epilepsies will be necessary for the identification of more effective age-specific anti-epileptogenic therapies.

Initial insult

Some, but not all, models are created using an inciting insult. Factors expected to modify the epileptogenic response to the initial insult or induction method may be investigated as to their potential impact on antiepileptogenic efficacy. These include the severity or duration of the initial epileptogenic insults, as well as the impact of different pathologies and genetic background. Pre-existing genetic mutations may alter the response not only to the initial insult but also to anti-epileptogenic therapies (Glasscock et al. 2007, Chiu et al. 2008). Studies that account for etiology and severity of the initial insult may help to characterize etiology-specific, anti-epileptogenic therapies, thereby guiding eligibility criteria for future clinical trials. It is important to differentiate anti-seizure effects from anti-epileptogenic effects if the initial insult is an induced seizure, as the former may confound the assessment of the latter (e.g. kindling or post-status epilepticus models).

Therapeutic time window

AETs may have different effects if administered before or after the onset of seizures, during different stages of the disease, and if given for short or long periods (Goodkin and Kapur 2003, Buckmaster et al. 2009, Jozwiak et al. 2011). Since in many clinical situations treatment can only be initiated after the initiating epileptogenic insult (e.g., after a stroke or head injury) or after the occurrence of spontaneous recurrent seizures, it is important to demonstrate preclinical efficacy under similar conditions. Application of therapies prior to an epileptogenic insult may provide limited relevant information. If treatments are effective only when given at the presymptomatic stage, biomarkers or surrogate endpoints may need to be identified to guide the timing and duration of anti-epileptogenic therapy prior to onset of clinical symptoms. The identification of clinically relevant biomarkers for treatment administration, progression or epileptogenesis and treatment-related side effects and confirmation of prevention or cure could greatly increase the efficacy/safety profile of the antiepileptogenic therapy and will reduce the costs and efforts invested in these studies (Roch et al. 2002, Glauser 2007, Engel 2008, Shinnar et al. 2008, Bragin et al. 2010, Dube et al. 2010). The STAIR experience with tissue-type plasminogen activator (tPA) is an encouraging example, because the therapeutic window has been found to be about the same in animals and humans (Papadopoulos et al. 1987, Ding et al. 2004, Hatcher and Starr 2011).

Outcomes

For antiepileptogenesis studies, a full reporting of the methods, duration, and frequency of seizure detection and analysis is important. Blinded analysis revealing the sensitivity achieved for seizure detection, information regarding the number of data gaps (i.e. periods without EEG monitoring), and calculations of the seizure frequencies in a manner that accounts for the duration and timing of monitoring as well as seizure clustering can facilitate interpretation of preclinical data and subsequent clinical development (Williams et al. 2009). Determination of the optimum observation period to assess the outcomes (e.g. spontaneous seizures) should consider the expected frequency and time of appearance in the model being used. If the study aims at evaluating the persistence of an anti-epileptogenic effect following withdrawal of treatment, the time taken to wash out the effects of the treatment needs to be considered.

Issues related to symptomatic and disease modifying treatments affecting comorbidities

Preclinical development of symptomatic and disease modifying treatments affecting comorbidities, such as neurocognitive and developmental impairments, neuropsychiatric conditions and cardiovascular events, shares many of the challenges described for anti-seizure and anti-epileptogenic studies. The importance of early intervention to minimize comorbidities has been advocated in several early life human epilepsy syndromes, such as infantile spasms and tuberous sclerosis (Lux et al. 2005, Jozwiak et al. 2011). Identifying the therapeutic time window and appropriate clinically relevant biomarkers may facilitate the design of future clinical trials. The major challenge in the designing these studies is the delineation of appropriate and easily quantifiable outcomes, including cognitive, behavioral, cardiorespiratory, and neurological outcomes, that are relevant to the respective comorbidities in people with epilepsy. The likelihood of some of those outcomes, i.e cognitive/behavioral, to change as a result of experience, handling, biological, genetic or other epigenetic factors requires special care and standardization in the monitoring of the experimental animals. In particular, developmental studies face the challenge in interpreting these outcomes as a function of age, sex, maturation rate, in addition to treatment and the previously mentioned modifiers.

Potential value of biomarkers and surrogate endpoints

A biomarker can be defined as an objectively measured characteristic of a normal or pathological biological process, or a biological response to a therapeutic intervention (Biomarkers Definitions Working Group 2001, Engel 2011). A surrogate endpoint is a biomarker that can substitute for a clinical endpoint (Biomarkers Definitions Working Group 2001) and can therefore provide an indirect measure of disease presence or progression (Engel 2011). Diagnosis and treatment of epilepsy suffer from the lack of reliable biomarkers and/or surrogate endpoints for the presence and severity of an epilepsy condition, epileptogenesis or comorbidities. Biomarkers or surrogate endpoints appropriate for the epileptic syndrome under study would be of great benefit in selecting the patient populations likely to benefit, in guiding treatment timing and selection and in monitoring clinical outcomes (Engel 2011). According to the goals, different modalities may be used to identify clinically relevant biomarkers, i.e. electrophysiological measures, biochemical markers, magnetic resonance imaging, positron emission tomography, genetic and epigenetic markers, and cardio-respiratory monitoring (Roch et al. 2002, Glauser 2007, Engel 2008, Shinnar et al. 2008, Bragin et al. 2010, Dube et al. 2010). There is specifically a need for clinically applicable biomarkers that: (a) differentiate epileptogenic changes from reactive changes or epiphenomena, especially at the early stage following the initial epileptogenic insult, to guide selection and early implementation of symptomatic and disease modifying treatments; (b) permit early and sensitive monitoring of disease progression, prior to appearance of clinical symptoms; (c) identify accurately the boundaries of the epileptogenic focus to allow implementation of more selective therapies (i.e. resection surgery); (d) reliably predict the time of seizure recurrence, the response or resistance to AETs, or the occurrence of AET-related adverse effects, (e) identify the risk for developing comorbidities of epilepsy and (f) document prevention or cure. Assessment of the limitations in translating preclinical markers to the clinical arena should be encouraged, including whether and how selected markers could be applied in the design of a clinical trial.

Conclusions

The discovery of better treatments for patients with epilepsy requires improved ways of evaluating preclinical models, more robust protocols, and a more consistent assessment of results. The present commentary should not be considered as a definitive list of recommendations but, rather, a framework for the development of specific guidelines. Such guidelines could also inform the grant review process, to ensure support of those proposals which are most likely to lead to clinically relevant results. Future work could involve definition of a hierarchical list of preclinical evidence recommended to progress to formal clinical testing, plus a separate list of optional, complementary information. This would allow the comparison of different AETs in terms of translational value, attributing different degrees of likelihood of clinical success to each, based on evidence deriving from preclinical studies.

Regular systematic reviews of published preclinical AET data would be helpful. Such reviews should address appropriateness and best utilization of animal models, tools for behavioral or outcome assessment, methodological design, and critical, comparative evaluation of the preclinical efficacy data for specific seizure types or syndromes. This would allow adaptation of guidelines to the evolving needs of the field. A Cochrane-like collaboration may be useful to pursue this aim (www.cochrane.org). To objectively evaluate the promise of new AETs, it would be useful to provide a forum to publish not only positive but also negative studies.

Abbreviations

AED

antiepileptic drug

AET

anti-epilepsy treatment

AES

American Epilepsy Society

EMA

European Medicines Agency

FDA

Food and Drug Administration

GLP

good laboratory practices

GMP

good manufacturing practices

ILAE

International League Against Epilepsy

IND

Investigational New Drug

NINDS

National Institute for Neurological Diseases and Stroke

PK/PD

pharmacokinetic/pharmacodynamic

STAIR

Stroke Therapy Academic Industry Roundtable