Models of Pediatric Epilepsies: Strategies and Opportunities


  • Carl E. Stafstrom,

  • Solomon L. Moshé,

  • John W. Swann,

  • Astrid Nehlig,

  • Margaret P. Jacobs,

  • Philip A. Schwartzkroin

Pediatric epilepsies are among the most devastating of neurologic disorders. The developing brain is particularly susceptible to seizures, and seizure activity early in brain development can cause profound neurologic impairment, enhance subsequent seizure propensity during maturation and in adulthood, and lead to abnormalities in cognitive function (1). In infancy and childhood, two broad categories of epilepsy are of particular concern because of their intractability to treatment and association with cognitive decline: (a) The so-called “catastrophic” childhood epilepsies (including infantile spasms, Lennox–Gastaut syndrome, and the progressive myoclonic epilepsies) are characterized by numerous etiologies, by age-specific developmental windows of seizure onset, by refractoriness to medical treatment, and by progressive cognitive deterioration (epileptic encephalopathy). Seizures associated with catastrophic epilepsy syndromes tend to have bilateral/generalized manifestations: (b) Refractory partial epilepsies are often associated with dysplastic brain lesions (i.e., tuberous sclerosis complex, TSC), or severe perinatally induced injuries (i.e., perinatal stroke or hypoxia); seizures of this type may be also seen with mesial temporal lobe epilepsy, but this latter condition is more prevalent in late childhood and adolescence. Often an overlap exists between these two broad categories of medically intractable epilepsies; for example, TSC is commonly associated with generalized infantile spasms as well as with multifocal partial seizures. This overlap suggests that common features may exist in the pathogenesis of different epilepsy types during early brain development. Regardless of the epilepsy type, etiology, or syndrome, the mechanisms of epilepsies in infancy and early childhood likely differ significantly from those of epilepsies in older children and adults. These mechanistic differences have important implications for therapeutic strategy and therapeutic efficacy (2,3).

The lack of appropriate animal models is a major impediment to a more complete understanding of pediatric epilepsy and to more effective treatments. Animal models serve numerous functions. They provide opportunities to investigate and elucidate basic mechanisms, to test and/or develop new antiepileptic medications (AEDs) and other therapeutic modalities, to devise new diagnostic approaches, and to study the neurologic consequences of seizures at various stages of brain development (4–7). Given the many (and often significant) differences between human and animal (i.e., rodent) brain structure and developmental profiles, no animal model is likely to reproduce faithfully every aspect of a human epilepsy syndrome. However, the insightful use of such models can play a pivotal role in generating hypotheses about the mechanisms, pathogenesis, and consequences of seizures in the developing brain, and for testing potential therapies.

This report summarizes discussions held during a workshop on Models of Pediatric Epilepsies, held in Bethesda, MD, on May 13–14, 2004. The Workshop was sponsored by NINDS/NIH and supported by grants from the American Epilepsy Society and the International League Against Epilepsy. Whereas previous NIH workshops have examined general priorities for epilepsy research (8,9), the current Workshop focused explicitly on epilepsies in the developing brain. Invited participants included clinical pediatric epileptologists, basic epilepsy researchers, and developmental neurobiologists (Appendix A).


The general goals of the Workshop are shown in Table 1. They arose from the following questions:

Table 1. Workshop goals
1. Identify pediatric epilepsies to be modeled, specify clinically relevant information to guide model development
2. Review current epilepsy models, and identify obstacles to further model development
3. Discuss criteria to be used in generating and validating models of pediatric epilepsies
4. Make recommendations for the development of models for two forms of severe pediatric epilepsy:
 a. Catastrophic generalized epilepsy (infantile spasms)
 b. Intractable partial epilepsy (tuberous sclerosis complex)
5. Identify techniques and technologies that would facilitate model development in immature animals
6. Evaluate the pros and cons of model development in different species (in vitro vs. in vivo, etc.)
7. Establish mechanisms for development of networks for interlaboratory collaboration
8. Identify funding resources to support model studies
  • 1Given the wide diversity of seizure types presenting at different developmental ages, certain clinical questions must be considered (i.e., what clinical problems are most pressing and what additional information is needed to guide model development? Are clinical disorders sufficiently well defined to provide guidance for the laboratory researcher attempting to model a human clinical condition? How do we deal with the observation that, in humans, numerous etiologies can cause the same epilepsy syndrome (and, in some cases, a given etiology will result in different seizure phenotypes)?
  • 2What is presently known about existing models of catastrophic generalized epilepsy and intractable partial epilepsy? Why has it been so difficult to develop models of pediatric epilepsies? Challenges to development of animal models of pediatric epilepsy include the uncertainty of what constitutes analogous developmental ages in humans and experimental animals (especially rodents), vast interspecies differences in the rate and patterns of brain growth and neuronal maturation, and dissimilarities of seizure phenotypes and EEG manifestations seen across different species.
  • 3A goal of the Workshop was to derive recommendations on the use of current (and future) animal models, including suggestions about the appropriateness of different species, genetic background, seizure etiologies, and other clinically relevant variables. How can we generate specific modeling targets, and what criteria can we use for “validating” the models?
  • 4Although broad agreement was reached that catastrophic generalized epilepsies and intractable focal epilepsies are important targets for model development, many questions remained about how to prioritize these targets (pressing clinical need? highest likelihood of rapid returns? pathophysiologic relevance?). Given the complexity of the clinical conditions in question, what subset of experimentally approachable features should to be reproduced in useful models? What are the bases for accepting or rejecting these models as clinically relevant/useful?
  • 5What conceptual and technologic advances would facilitate our attempts to understand developmental epilepsies? Technologic developments, development of modeling approaches at different levels (computer simulations, tissue-culture preparations, intact animals), and consideration of animals species not currently exploited for model development may be key directions for future progress.
  • 6Finally, how can we establish networks of collaboration among laboratories involved in pediatric epilepsy research, and how can our limited resources be maximized?


General considerations

To facilitate model development, a variety of common strategies can and should be used. A more complete understanding of salient clinical features (e.g., epidemiology, etiology, pathology, and natural history of intractable developmental epilepsies) will undoubtedly facilitate the generation of models with features that are clinically relevant. Toward this goal, it is necessary to elucidate the developmental progression of key biomarkers (e.g., neurotransmitters, membrane ion channels, subcellular signaling cascades), behavioral phenotypes, and numerous other factors to understand their relations to age-specific epilepsies. Age-related differences in neuronal plasticity and glial function are emerging as key modulators of pediatric epilepsy. Other avenues of potential importance are just beginning to receive experimental attention, such as the role of stress in epileptogenesis, the effects of seizure-induced acute and chronic inflammation, and subtle circuit reorganization associated with neurogenesis, axonal sprouting, and molecular plasticity.

In any model of epileptic encephalopathy, a concerted effort should be made to determine long-term consequences of the neurobiologic abnormalities, including behavioral changes, cognitive dysfunction, and seizure susceptibility. In generating models of pediatric epilepsies, it is important to remember that in some children with an epileptic encephalopathy, the seizures themselves may be less impairing to the child than the associated cognitive deficits. In other cases, the seizures may exacerbate the cognitive deficits. Further, the effects of anticonvulsant drugs and other treatments on parameters of brain development and cognition must be considered in parallel with their effects on seizures themselves.

Model development should involve, as a key goal, efficient means for the testing new anticonvulsant/antiepileptic (and even antiepileptogenic) medications, as well as for evaluating novel therapies such as special diets targeted to seizure control (e.g., ketogenic diet, low-glycemic-index diet), a variety of implantable stimulation devices (e.g., vagus nerve stimulation), and novel drug-delivery systems. These models should also allow us to test the potentially beneficial effects of environmental enrichments that may reduce the morbidity of epilepsy in children. The models should be useful for investigating if/how treatments might slow (or halt) the progression of epilepsy (antiepileptogenesis).

The Workshop's recommendations with respect to general strategies for model development are shown in Table 2.

Table 2. Recommendations: General strategies for model development
 1. Create models based on the clinical need for better understanding and therapy but also on the potential for positive returns (i.e., focus on experimentally accessible problems)
 2. Generate models in which one can pursue key questions/testable hypotheses
 3. Generate models that reflect age-specificity of developmental epilepsies and exhibit age-specific manifestations
 4. Elucidate “normal” developmental progression for key factors (e.g., synaptic maturation) to facilitate model development
 5. Generate models that show EEG abnormalities and/or seizure predisposition; spontaneous seizures would be optimal (but not necessary)
 6. Investigate epidemiology, etiology, and natural history of intractable/catastrophic epilepsies
 7. Develop models by using “multiple-hit” strategies
 8. Consider models of epileptic encephalopathy, with a focus on long-term consequences of seizures
 9. Develop criteria for model validation (e.g., pharmacologic response profile, genetic mimicry, seizure phenotype)
10. Take advantage of models and experimental insights from investigation in related fields (e.g., ischemia, sleep, traumatic brain injury)

Criteria for an “ideal” model

Workshop participants defined an ideal model as one in which one or more aspects of human pediatric epilepsy could be studied, recognizing that inherent species differences preclude exact replication. To model a given type of pediatric epilepsy accurately, the investigator must be aware of the diversity of seizure manifestations and etiologies associated with the targeted human epilepsy syndrome. For example, what appear to be similar lesions may cause severe epilepsy in one patient and minor or no dysfunction in another. Seizures differ according to etiology, age at onset, genetic predisposition, EEG findings, treatment responsiveness, natural history, and numerous other factors. In the complex task of designing a clinically relevant animal model, two complementary approaches should be considered. An “intuitive” or “face-value” model attempts to replicate the human disease as closely as possible; this approach is likely to be high risk and may not readily yield insights into mechanism. Alternatively, a “reductionist” model isolates specific components of the pathophysiology and allows the investigator to study underlying mechanism; it may, however, provide limited data about clinically relevant behaviors and/or responses to treatment.

As a first step in establishing criteria for an “ideal” model, the definition of epilepsy as unprovoked, recurrent seizures must be considered. Seizures induced immediately by chemical agents and electrical stimulation can yield important, age-specific mechanistic information. However, for a model to be truly “epileptic,” seizures should occur spontaneously, without provocation, during an appropriate age window. Worthy of consideration, however, are models in which only EEG abnormalities are apparent, or in which enhanced seizure-propensity occurs (which is manifest when the animal is challenged in some way).

Second, the investigator should recognize that human epilepsy syndromes often involve multiple etiologies. An ideal animal model might incorporate similar etiologies, such as an acquired insult (trauma, fever) or a genetic mutation. In reality, it is unlikely that the most popular current methods used to induce seizures in animal models (e.g., kainic acid, pilocarpine, fever, hypoxia/ischemia) are “equivalent” to the factors that lead to human epilepsies. This lack of generalizability may affect our ability to “validate” a given model. However, because these factors can be experimentally manipulated, they give the investigator an opportunity to consider how such insults (or genetic factors) interact. Multiple insults, occurring at different times of development, may reflect a clinically realistic approach to generation of pediatric epilepsies.

A third criterion for an ideal animal model is that the seizure phenotype of the model should resemble the phenotype of the seizure of interest in the humans, with site-specific behavioral manifestations. Whereas the focus of many epilepsy models has been on the hippocampus and neocortex, the brainstem and other brain regions should be considered, especially because many subcortically mediated behaviors may be critical for seizure manifestations in young rodents. Realistically, it is unlikely that rodent seizures, and particularly seizures in young rat or mouse pups, will fulfill this phenotypic ideal. It is our challenge to determine what seizure phenomena in rodents—behavioral and/or electrical—correspond to salient seizure phenomena in human infants and babies.

Fourth, although species differences probably preclude an exact correlation between human and animal, EEG abnormalities should be present in the animal model; preferably, those EEG abnormalities should have common features with the human condition. For example, in the case of modeling infantile spasms (see later), specific interictal versus ictal EEG patterns (e.g., hypsarrhythmia vs. electrodecremental responses) would ideally be present in the animal model.

Fifth, an ideal model would include responsiveness to syndrome-specific antiepileptic drugs (AEDs) or other treatment modalities, as seen in the human disorder.

Finally, the cognitive and behavioral consequences of seizures in an animal model should resemble those in the human syndrome. Long-standing uncertainty has existed as to whether seizures themselves cause brain damage and resultant neurologic/cognitive dysfunction, or whether underlying neurologic impairment leads to both seizures and cognitive consequences. Animal models maybe useful in sorting out these variables, because the animal-model approach offers the researcher a “normal” control brain as a starting point.


Given that catastrophic epilepsies in the immature brain are relatively rare, one might argue that modeling resources could be better used to study more common—and easier-to-target—pediatric epilepsies. The Workshop participants were unanimous, however, in supporting the importance of developing a system in which we can more effectively study and treat catastrophic syndromes. Workshop participants discussed infantile spasms as an example of catastrophic generalized epilepsy, as this syndrome carries a particularly poor prognosis, and an optimal therapy does not exist (10). Infantile spasms is an age-specific epileptic encephalopathy, occurring primarily in the 4- to 10-month range. West syndrome consists of infantile spasms, hypsarrhythmia (on interictal EEG), and mental retardation. It may appear several months after virtually any insult to the developing brain, such as infection, cerebral dysgenesis, hemorrhage, or hypoxia–ischemia.

Animal models of the human infantile spasms phenotype have been especially difficult to generate, perhaps because of our lack of understanding of the underlying causes and mechanisms. None of the recent attempts to model infantile spasms in animals has produced results that exactly “fit” the human syndrome (11). In addressing the question of what features would be important to reproduce in the animal, there was general agreement about the key characteristic factors required in an ideal model (Table 3), but considerable disagreement among Workshop participants about how to prioritize these targets. Certainly, it would be desirable to have seizures that resemble the spasms seen in infants, with sudden extension or flexion of the body, often occurring in clusters during sleep–wake transitions.

Table 3. Criteria for an ideal model of infantile spasms and West syndrome
1. Unprovoked spasms or myoclonic seizures early in postnatal development
2. EEG correlates of seizure events (ictal decremental response)
3. Abnormal interictal EEG (“hypsarrhythmia”) reflecting generalized epileptic encephalopathy
4. Response to clinical relevant treatment (e.g., ACTH and/or vigabatrin)
5. Behavioral/cognitive sequelae

Myoclonic seizures would be an alternative phenotypic target. Furthermore, an interictal EEG showing hypsarrhythmia, and an ictal electrodecrement EEG, would lend validity to the model. It was noted, however, that if hypsarrhythmia were defined to include the presence of multifocal, high-amplitude discharges, that pattern may be extremely difficult to document in a rat or mouse pup (where placement of multiple electrodes over the brain is limited by size and fragility). Therefore modeling of hypsarrhythmia may be restricted to larger animal models until technologic advances permit the development of “micro” electrode assemblies. An animal model of this disorder should also mimic cognitive stagnation or decline. Finally, the often-dramatic response in humans to administration of adrenocorticotropic hormone (ACTH) [or of vigabatrin (VGB) in a subpopulation], with resolution of both spasms and EEG abnormalities, would add credibility to a model.

Various attempts to model infantile spasms in animals were discussed. One model involves intraventricular administration of picomolar amounts of corticotropin-releasing hormone (CRH) to neonatal rats (12), an interesting approach, given the peculiar response of infantile spasms to ACTH. Further, the perinatal stress caused by etiologies associated with infantile spasms has led to the hypothesis that stress may increase endogenous CRH levels in seizure-prone areas of the developing brain, leading to neuronal damage, axonal reorganization, long-term cognitive deficits, and possibly age-related seizures such as infantile spasms (13). The CRH-induced seizure phenotype (primarily “limbic”) and EEG abnormalities do not mimic features of infantile spasms. However, CRH-treated rats do display cognitive deficiencies. Although CRH-induced seizures are not responsive to ACTH, ACTH does reduce CRH gene expression in certain neuronal populations (14). Another attempt to model infantile spasms involves intraperitoneal injection of N-methyl-d-aspartic acid (NMDA) in infant rats (15,16). This agent causes a clinical seizure described as emprosthotonus, consisting of whole-body tonic flexion with back-arching. Although EEG findings are variable, these seizures are often accompanied by a decrement of the EEG amplitude, resembling the electrodecremental response in the human EEG. However, interictal hypsarrhythmia is lacking, no spontaneous seizures have been recorded, and the seizures are not responsive to ACTH.

Lennox–Gastaut syndrome is another age-specific human epileptic encephalopathy, consisting of intractable seizures of multiple types plus mental retardation. Lennox–Gastaut syndrome has also been resistant to animal modeling. The infantile spasms of West syndrome often evolve, as the infant develops, into the various seizure types associated with Lennox–Gastaut syndrome. Investigators have focused on mimicking a single seizure type frequently seen in Lennox–Gastaut syndrome: atypical absence seizures (AASs) (17). Rats treated with an inhibitor of cholesterol biosynthesis develop life-long recurrent episodes of bilateral synchronous slow spike–wave discharges (emanating from both hippocampi), characteristic of AAS; these discharges are associated with ictal clinical manifestations of staring and facial myoclonus. Later in life, these rats exhibit deficits in long-term potentiation and spatial learning and memory (18).

These modeling attempts illustrate the difficulties that arise when trying to model a seizure type—and an epilepsy syndrome—as complex as infantile spasms and West syndrome. Clearly, one of the most difficult issues is the semiology of the seizures themselves. Infant rats (and mice) exhibit a variety of jerky movements and behaviors, especially during sleep (19), that can sometimes be confused with pathologic spasms. The differential occurrence of these spasm-like movements during wakefulness or sleep adds to the difficulty of the modeling challenge. Thus to be considered epileptic, any paroxysmal behaviors resembling spasms or myoclonus should be associated with an EEG seizure discharge. Ideally, a model of these pediatric epilepsy syndromes in mouse or rat would arise from some understanding of the pathophysiology underlying the disorder, an insight not yet achieved at the clinical level.

The concept of “multiple hit” models of pediatric epilepsies was discussed in relation to model development for generalized catastrophic disorders. In some syndromes, epilepsy might occur as a consequence of two (or more) independent, but interdependent brain insults. An “acquired” insult (such as head trauma or hypoxia) may be superimposed on a background of genetic mutation; for example, mutation of the LIS1 gene (as seen in Miller–Dieker syndrome) is associated with neurodevelopmental pathology (20) and predisposes to infantile spasms. Such genetic-acquired two-hit models require further exploration.


Whereas infantile spasms represents a generalized catastrophic epilepsy of childhood, other epilepsies appear to result from focal brain abnormalities and are reflected in partial seizures (for example, traumatic brain injury and developmental migration disorders that lead to cortical dysplastic lesions) (21,22). Discussion at the Workshop focused on TSC, a multisystem disorder often associated with refractory seizures (23); thus TSC can be considered an example of an intractable partial epilepsy of childhood. Epilepsy develops in >90% of children with TSC, and the severity of neurologic impairment in TSC is correlated with early onset of seizures (24). Because of the occurrence of multiple cortical tubers in a given patient, children may have one or several seizure foci that individually or together cause multifocal partial seizures and epileptic encephalopathy. Although significant progress has occurred in understanding the genetic and molecular basis of TSC (25), still little insight into the pathogenesis of epilepsy exists in this disorder. The ideal TSC model (Table 4) would replicate the genetic defect (TSC1 or TSC2 gene mutation), the structural brain abnormalities (cortical tubers and subependymal nodules), spontaneous seizures with EEG correlates, and cognitive deficits seen in human TSC. Although several animal models of TSC are known, none provides representation of all of these TSC components. The Eker rat has a spontaneous germline mutation in which one TSC2 allele is inactivated, resulting in abnormal TSC2 gene product (tuberin). Eker rats have renal tumors and subependymal nodules, but cortical tubers are not ordinarily observed, even in two-hit models in which a somatic insult is superimposed on the genetic mutation (26,27). Eker rats exposed to a second hit do have a higher number of dysmorphic neurons and exhibit some level of cortical disorganization, but do not exhibit spontaneous seizures (although Eker rats may have a decreased threshold to induced seizures). Recently, memory alterations have been documented in Eker rats (28). In contrast, modeling attempts in mice, using conditional knockout strategies, have given rise to animals with spontaneous seizures but without the neuropathologic features typical of TSC. For example, a mouse conditional knockout of TSC1 (the gene that codes for the protein hamartin), in which the deletion is driven by the glial fibrillary acidic protein (GFAP) promoter, shows abnormal cortical excitability, astrocyte neuropathology, spontaneous seizures, and cognitive deficiencies (29,30). Complementary in vitro and in vivo studies have revealed decreased glutamate-transport currents and impaired uptake of potassium through inward rectifier channels (31), suggesting potential mechanisms for brain hyperexcitability in these mice. Still, the absence of cortical tubers weakens the validity of these mice as a model of the human disorder. Further work with the conditional genetic manipulation approach (e.g., using a neuron-specific promoter system) may provide additional insights and yield neuropathologic features that more closely resemble the pathology of TSC.

Table 4. Criteria for an ideal model of partial/focal intractable epilepsy (e.g., associated with tuberous sclerosis complex)
1. Genetic mutation of TSC1 or TSC2
2. TSC-like neuropathology, including cortical tubers and subependymal nodules
3. Unprovoked partial seizures
4. EEG correlates, including focal seizure initiation and interictal spikes
5. Response (or lack of response/intractability) to clinically relevant treatments
6. Behavioral/cognitive sequelae


Choice of animal species

A major need in developing appropriate models of pediatric epilepsy is a better knowledge base of the disorders in question. Laboratory research can contribute to that knowledge base through studies of human material from autopsy specimens as well as of surgically resected tissue. Such studies will provide a better understanding of the structural, molecular, and functional abnormalities underlying epileptic circuits. Clearly, however, this material is limited and rarely allows the experimenter to manipulate key variables. Thus animal models will continue to play a pivotal role in advancing knowledge about the developmental epilepsies (Table 5). The choice of species for these studies is critical, as species (and even strain) differences have been shown to result in very different susceptibilities to seizures and their consequences (32). The choice of rats and mice to model epilepsy phenomena has been based largely on experimental convenience and on the large background of information about brain function in these species. For studies of the genetic bases of epilepsy, mice will certainly continue to be a valuable resource (33); studies of human genetic mutations associated with pediatric epilepsy syndromes should be pursued in these versatile mouse systems. Rodents will also be helpful in assessing the roles of gender and hormones in developmental epilepsy (34). However, the sequence and rate of ontogenetic steps in rodent brain development differ in significant ways from human brain development; thus rodents may not always model epilepsy pathophysiology adequately. Consideration should also be given, therefore, to other animals such as primates and pigs, in which aspects of human brain development may be more precisely reproduced. For example, piglets are used extensively in neonatology laboratories, and experimental techniques for maintaining stable systemic parameters are well established in this field (35). In contrast, future model development should also take advantage of simple systems—such as zebra fish, flies, and invertebrates—in which large-scale, fast-throughput studies can screen for relevant genes and drug responsiveness, as well as address specific mechanistic issues (36). Workshop participants stressed the importance, particularly in these simpler systems (and in the in vitro preparations) of defining the seizure phenotype carefully. Finally, computer modeling represents a relatively untapped approach to explore certain questions of cellular and network function in pediatric epilepsies (37). In all such considerations of animal choice, the experimental question should determine the optimal approach.

Table 5. Recommendations for resource development
Animal species
 1. Take advantage of human material (autopsy, surgical resection)
 2. Continue to use rodent models, particularly for studies of genetic contributions to complex epilepsy syndromes
 3. Consider the use of higher vertebrates in which brain development more closely parallels that in humans (e.g., pigs, monkeys)
 4. Exploit simple systems (e.g., zebra fish) for drug screening, genetic testing
Technologic advances
 1. Develop animal neonatal intensive care units (“NICUs”)
 2. Create behavioral monitoring and testing laboratories for infant rodents
 3. Miniaturize EEG recording and telemetry equipment
 4. Adapt advanced, noninvasive imaging techniques for neonatal rodents
 5. Apply high-throughput genomic and proteomic techniques
 1. Take advantage of insights, techniques, and models developed in other (related) areas of pediatric research
 2. Develop specialized service centers (e.g., neonatal intensive care units for rat pups)
 3. Develop specialized centers for monitoring (video-EEG, imaging)
 4. Develop collaborative networks (clinicians and basic scientists), via web-based interactions and program project-like cooperation
 5. Share resources (websites to share information, centers for tissue distribution, distribution of unique animal models)

Technologic development

Extensive discussion was held regarding technical advances that might facilitate our study and understanding of pediatric epilepsies. At the top of the list (Table 5) was the development of reduced-scale (e.g., for rat and mouse pups) “neonatal intensive care units” that would allow concurrent monitoring of several aspects of neurologic function: video-EEG, motor activity, and blood pressure and respiration parameters. Metabolic parameters in young rodents, both during and between seizures, have been underemphasized in experimental work in this field; in particular, temperature measurement and its regulation are critical parameters that require assiduous attention in neonatal rodents (38). Another critical need is miniaturized systems for EEG recording, especially in freely moving animals. In the last several years, technical advances have made it possible to carry out both short- and long-term electrophysiologic monitoring of spontaneous seizures in adult rodents. However, it remains very difficult to record EEG reliably in rats younger than age postnatal day 10, and even more challenging to record from immature mice. Capabilities for small animal neuroimaging (in modalities such as MRI, functional MRI, and PET) will add significantly to the understanding of structural and functional substrates of epilepsy during development. Large-scale molecular screening techniques, including proteomics and gene profiling with microarrays, must be incorporated into experimental protocols for pediatric epilepsy studies. Finally, general agreement was reached about the urgent need to develop more precise methods for behavioral characterization and cognitive testing, to assess the long-term effects of seizures in (and on) the developing brain (39).

Resource development

One feature of the current Workshop was the participation of investigators working on pediatric models outside the epilepsy field (e.g., in genetics and neurogenesis, in brain development and plasticity, in sleep and mental retardation, and in pathologies associated with ischemia and traumatic brain injury). Investigators in these fields have developed modeling approaches and insights that should be incorporated into the epilepsy research arena. The sharing of such approaches, across fields and across laboratories, remains an untapped but potentially rich resource (Table 5). Innovative approaches to networking and collaboration, including web-based interactions, are needed. Such approaches would include resource “centers” that might be shared across institutions, such as rodent “neonatal intensive care units” (with video-EEG capability and apparatus for behavioral testing), neuroimaging facilities, and genetic profiling centers. Collaboration between networks of investigators, spanning the spectrum of clinical and basic research, is essential for substantial progress in the pediatric epilepsy field. The recent study by Dzhala et al. (40), which identified a novel treatment for neonatal seizures, represents an example of fruitful collaboration between institutions, involving the integration of data from human autopsy material and animal investigations (both in vivo and in vitro). The ready availability of web-based information, tissue-distribution resources, and genetic models make interinstitutional projects not only necessary but obviously feasible.

Comprehensive discussion about funding resources yielded a broad array of possibilities. We, as investigators, must convince traditional funding mechanisms (including NIH and private foundations) of the urgency of research on pediatric epilepsy disorders. In addition to the hypothesis-driven NIH R01 grants, model-driven grants through the R21 mechanism should be considered. Sources of funding for pilot grants include private foundations. A listing of these potential funding sources can be found under Resources in Epilepsy Research on the American Epilepsy Society website (


Workshop participants enthusiastically supported the idea of specific model development in the pediatric epilepsies. Even in this era of limited funding, advances in the field are not only scientifically feasible but also crucial for the optimal care of children with epilepsy. Over the long term, investments of time, resources, and scientific effort should be cost-effective in terms of health and human needs. Some of the issues discussed in this workshop stemmed from the NIH-Curing Epilepsy Conference report (developed in 2000), which identified specific gaps in our understanding of epilepsy and epileptogenesis, and set specific benchmarks of attainable goals ( Many of these benchmarks focus specifically on a more complete understanding pediatric epilepsies and their underlying bases, so that we can develop better treatment options—and eventually “cures”—for these often devastating disorders.


Table APPENDIX A.. Workshop participants
 Margaret P. Jacobs (NINDS/NIH)
 Solomon L. Moshé, M.D. (Albert Einstein College of Medicine)
 Astrid Nehlig, Ph.D. (INSERM/University of Strasbourg, FRANCE)
 Philip A. Schwartzkroin, Ph.D. (University of California, Davis)
 John W. Swann, Ph.D. (Baylor College of Medicine)
 Scott C. Baraban, Ph.D. (University of California, San Francisco).
 Tallie Z. Baram, M.D., Ph.D. (University of California, Irvine)
 Mark S. Blumberg, Ph.D. (University of Iowa)
 Amy Brooks-Kayal, M.D. (University of Pennsylvania/Children's Hospital of Philadelphia)
 Gregory L. Holmes, M.D. (Dartmouth-Hitchcock Medical Center)
 Frances E. Jensen, M.D. (Harvard University/Boston Children's Hospital)
 Michael Johnston, M.D. (Johns Hopkins University/Kennedy Krieger Institute)
 Bryan Kolb, Ph.D. (University of Lethbridge, CANADA)
 Gary W. Mathern, M.D. (University of California, Los Angeles)
 Jeffrey L. Noebels, M.D., Ph.D. (Baylor College of Medicine)
 Manisha N. Patel, Ph.D. (University of Colorado)
 Margaret Ross, M.D., Ph.D. (Weil Medical College of Cornell University)
 Russell M. Sanchez, Ph.D. (University of Texas, San Antonio)
 Raman Sankar, M.D., Ph.D. (University of California, Los Angeles)
 Faye Silverstein, M.D. (University of Michigan)
 O. Carter Snead III, M.D. (University of Toronto/Hospital for Sick Children)
 Roberto Spreafico, M.D., Ph.D. (Istituto Nazionale Neurologico “C.Besta,” ITALY)
 Carl E. Stafstrom, M.D., Ph.D. (University of Wisconsin)
 Libor Velisek, M.D., Ph.D. (Albert Einstein College of Medicine)
 Claude G. Wasterlain, M.D. (University of California, Los Angeles)
 Karen S. Wilcox, Ph.D. (University of Utah)
NIH Staff and Guests
 Daofen Chen, Ph.D. (NINDS)
 Brandy Fureman, Ph.D. (NINDS)
 Meenaxi Hiremath, Ph.D. (NINDS)
 Deborah Hirtz, M.D. (NINDS)
 Diane Howden (NINDS)
 Barbara Kelley, B.A. (Citizens United for Research in Epilepsy)
 Michael A. Rogawski, M.D., Ph.D. (NINDS)
 Randall Stewart, Ph.D. (NINDS)
 William H. Theodore, M.D. (NINDS)
 Christina Vert (NINDS)
 Ljubisa Vitkovic, Ph.D. (NICHHD)
 Alan Willard, Ph.D. (NINDS)