Seizures and Antiepileptic Drugs: Does Exposure Alter Normal Brain Development?

Authors

  • Eric D. Marsh,

    1. Division of Child Neurology and Pediatric Regional Epilepsy Program, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Amy R. Brooks-Kayal,

    1. Division of Child Neurology and Pediatric Regional Epilepsy Program, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    2. Departments of Neurology and Pediatrics, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A.
    Search for more papers by this author
  • Brenda E. Porter

    1. Division of Child Neurology and Pediatric Regional Epilepsy Program, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
    2. Departments of Neurology and Pediatrics, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A.
    Search for more papers by this author

Address correspondence and reprint requests to Eric D. Marsh, Division of Child Neurology, 6th Floor Wood Building, Children's Hospital of Philadelphia, 3400 Civic Center Boulevard, Philadelphia, PA 19104. E-mail: marshe@email.chop.edu

Abstract

Summary:  Seizures and antiepileptic drugs (AEDs) affect brain development and have long-term neurological consequences. The specific molecular and cellular changes, the precise timing of their influence during brain development, and the full extent of the long-term consequences of seizures and AEDs exposure have not been established. This review critically assesses both the basic and clinical science literature on the effects of seizures and AEDs on the developing brain and finds that evidence exists to support the hypothesis that both seizures and antiepileptic drugs influence a variety of biological process, at specific times during development, which alter long-term cognition and epilepsy susceptibility. More research, both clinical and experimental, is needed before changes in current clinical practice, based on the scientific data, can be recommended.

Each time clinicians make decisions regarding treatment of seizures in an infant or young child, they weigh the prevailing evidence regarding the relative effects of seizures and antiepileptic drug (AED) treatment on the developing brain. Both these topics have been discussed and debated among pediatric neurologists for decades; however, advances in developmental neuroscience, as well as recent clinical studies, now shed new light on this topic. Therefore, this review addresses both the clinical and basic science literature on the impact of seizures and AEDs on the developing brain, with the hope of contributing to an improved understanding of the consequences of the treatment decisions by clinicians caring for infants and young children with seizures.

SEIZURES AND THE DEVELOPING BRAIN

Seizures and the developing brain: clinical studies

Whether or not seizures impair brain development has been a fundamental issue in epilepsy research and clinical practice. This issue, however, has been difficult to approach in clinical research studies, as multiple factors are all potential contributors to cognitive dysfunction in children with epilepsy. In addition to the seizures themselves, AEDs, underlying brain malformations, and preexisting learning and attention problems are all potential contributors to cognitive dysfunction in children with epilepsy. The most compelling clinical literature addressing the issue of whether seizures themselves cause or exacerbate cognitive dysfunction comes from prospective cohort studies of new onset seizure patients who are followed longitudinally with formal neuropsychological testing beginning at or near the onset of their epilepsy. Early studies (five published prior to 1940 (Dodrill, 2004)) following children for 1 to more than 4 years, examined standard intelligence quotient (IQ) test data and found that 10 to 40% of children with seizures demonstrated mild to moderate reductions in IQ over time. Unfortunately, standard IQ data lack the sensitivity of more detailed neuropsychological testing, limiting the interpretation of these early studies. More recent studies had mixed results, with some finding no change in cognitive function over time (Ellenberg et al., 1986) and others finding that a subgroup of patients shows a definite cognitive decline (Bourgeois et al., 1983; Aldenkamp et al., 1990; Bjornaes et al., 2001).

Bourgeois et al. (1983) studied 72 new onset seizure patients between the ages of 18 months and 16 years with neuropsychological evaluations initiated within two weeks of treatment onset and continuing yearly for an average of four years. They additionally followed 45 siblings of the study group as controls. They found no overall difference between the patients and siblings, and no change in IQ over time in either group. A subset of epileptic patients were found to have a decrease in IQ >10 points (11%) as well as a group whose IQ increased by more than 10 points (16.7%) over the 4 year study. The patients with the IQ decline had more frequent seizures, higher drug levels, and earlier epilepsy onset. They also found that the symptomatic patients and patients with multiple seizure types had lower IQs at the initial epilepsy diagnosis regardless of age of seizure onset (Bourgeois et al., 1983). Two other prospective longitudinal studies have reached similar conclusions. Aldenkamp et al., found no overall change in cognitive outcome in the full cohort, but a subgroup of children (24%) showed a greater than nine point decline in full scale IQ over a 4.2 year follow-up (Aldenkamp et al., 1990). Another study found no change in IQ over time when the entire cohort of children and adults were analyzed together (Bjornaes et al., 2001). However, upon secondary analysis the children, but not the adults, declined an average of six points in both performance and full scale IQ during the 3.5 year mean follow-up, suggesting seizures have a greater effect on the developing brain than on the mature brain. Of note is the fact that most of the patients in these studies had active seizure disorders with a relatively high seizure burden.

Several additional studies have compared children with epilepsy to age matched controls and concluded that the epileptic group had a higher rate of cognitive impairment; although the designs of these studies make it more difficult to differentiate the effects of seizures from the underlying etiology, preexistent cognitive/learning deficits, and/or detrimental effects of AEDs. A study by Neyens et al., (1999) of 11 children with epilepsy versus age matched controls found a trend toward the children with seizures not making equivalent gains to controls on the full scale IQ over 1.5 years of follow-up. An analysis of children whose seizures continued for over 10 years showed the biggest changes in IQ. All children with seizures were on AEDs that may have contributed to the cognitive effects. Interestingly, children who were seizure-free during the study period did not fare better than those with continuing seizures, reinforcing the idea that AEDs and/or underlying substrates in addition to the seizures may be important contributors to the cognitive delay. Schoenfeld and colleagues studied 57 children with complex partial seizures and compared them to sibling controls. They reported that seizures affected both cognitive and behavioral measures, with age at onset of seizures being the strongest predictor of cognitive functioning (Schoenfeld et al., 1999). The frequency of seizures had no predictive effect on cognition in this study, although other studies have found a correlation between higher seizure frequency and poorer cognitive performance (Bulteau et al., 2000). Additional studies have also identified early age at onset, symptomatic or cryptogenic etiology, and total lifetime seizure number as important predictors of which children with epilepsy will experience detrimental cognitive effects (Seidenberg et al., 1986; Bulteau et al., 2000; Freitag and Tuxhorn, 2005; Hoie et al., 2005). Bailet and Turk studied 74 children with idiopathic epilepsy serially over three years, excluding children with mental retardation, symptomatic and cryptogenic epilepsy, and found impairments in intelligence, psychomotor speed, memory, academic achievement, and behavior compared to sibling controls (Bailet and Turk, 2000). By excluding children with mental retardation and symptomatic epilepsy, and also finding that age at seizure onset, seizure type and seizure frequency did not correlate with neurocognitive test scores, these authors' work suggests that even children with normal intelligence and well-controlled seizures can experience learning problems (Bailet and Turk, 2000). Taken together, these studies again suggest that changes in cognitive performance may not correlate only with seizure burden, but also the underlying substrate, AEDs, and preexisting learning problems.

Not every study, however, has concluded that seizures impair cognitive development. Data collected from the large NIH Collaborative Perinatal Project (Ellenberg et al., 1986) found that patients with epilepsy had mean IQ scores that were similar to their sibling controls. Notably, on average, the study patients had very few seizures compared with the studies discussed above, which may explain, at least in part, the discrepancy with those involving patients with a higher seizure burden. Although the mean IQ did not differ between epilepsy and control groups, there were more children with mental retardation in the epilepsy group. The low IQ in this subset was present before seizure onset, did not change after the first seizures, and was attributed to underlying CNS abnormalities. Other studies have emphasized that seizures adversely affect cognitive outcomes in children with symptomatic epilepsy and underlying brain abnormalities (Shinnar and Hauser, 2002; Berg et al., 2004; Northcott et al., 2005). These and similar findings have led some authors to conclude that an underlying abnormal neural substrate is the major cause of cognitive changes in children with epilepsy, and the seizures are “concomitant rather than causal” (Lesser et al., 1986). The relative contribution of seizures, AED treatment, and underlying neurological abnormality to cognitive dysfunction in children with epilepsy is thus a complex issue that has been impossible to fully resolve with studies of patients in whom these different factors are often inextricably linked. To study the impact of seizures in isolation on brain development and cognition, scientists have turned to animal models.

Seizures and the developing brain: animal models

Animal models provide a mechanism to determine the effect of seizures on the developing brain without the complications of underlying brain pathology and medication effects. Extrapolating findings from studies in developmental epilepsy animal models, however, requires an understanding of the correlation between rodent and human neuronal development (see Table 1). Rodents have a briefer life span (approximately 2 years), shortened gestation (21 days), and postnatal development of only 1–2 months to maturity. In rodents, neural tube closure occurs within days of conception, cortical neuron birth and migration are completed shortly after birth, apoptosis and neuronal pruning occur in the first postnatal week, and the majority of brain growth occurs within the first two to three weeks of life. Thus, roughly, a newborn rat pup's development stage is comparable to a third trimester fetus, a rat at 7–10 days of age similar to a full-term newborn, and a 3-week-old rat can be considered as an older child or adolescent. These estimates are approximate and comparisons must be made with caution.

Table 1. Approximate equivalency between selected developmental milestones (top row) in humans and rodents. Times are listed in embryonic days postconception (E) or days post birth (P). If not stated, time is in days (Rakic and Nowakowski, 1981;Arnold and Trojanowski, 1996;Huttenlocher and Dabholkar, 1997;Clancy et al., 2001;Avishai-Eliner et al., 2002;Olney et al., 2002;Rallu et al., 2002). The last two rows denote the time periods when the current literature suggests that either AEDs or seizures have effects on the developmental processes
Developmental stageNeural tube closureForebrain expansionHippocampus: DGN formsMaximum synaptogenesisPeriod of apoptosisFull-term birthEyes openBrain mature
  1. The number of plus signs (+) denotes the level of evidence for an effect of AEDs or seizures on the corresponding developmental milestone. A question mark (?) signifies unclear or suggestive evidence, and a dash (–) marks where no evidence exists for an effect on the particular developmental milestone.

RodentE8–9E8–10E16–P15E16–P7Varied E16–P14E21(P0)P6–8P25–35
HumanE28–3535–42E260–280Regional 15mo–15 yearsE32 weeks– P13 weeksE40 weeksE26 weeksPost puberty
AEDs effects+++??+++ ++
Seizure effects?++++ ++

Early animal studies suggested that the developing brain was largely resistant to the impact of seizures on cognitive development (Stafstrom et al., 1993; Sarkisian et al., 1997; Stafstrom, 2002), but more recent studies have found definite behavioral and molecular effects that are often age-specific and quite different from the effects of seizures on the adult brain (for excellent reviews see Holmes et al., (2001; Holmes, 2004 and Velisek et al., 2002). Impaired spatial learning and memory deficits have been demonstrated in rodents after both prolonged and brief recurrent seizures on postnatal days 1–25, induced by a variety of methods including tetanus-toxin (Lee et al., 2001), corticotropin releasing hormone (CRH) (Brunson et al., 2001), lithium-pilocarpine (Liu et al., 1994; Kubova et al., 2004), kainate (Lynch et al., 2000a; Lynch and Sutula, 2000; Sayin et al., 2004), and flurothyl (Huang et al., 1999; de Rogalski Landrot et al., 2001). These studies suggest that the cognitive effects are a consequence of the seizures per se and not the specific method used to produce them. A cellular basis for these learning abnormalities is supported by findings that animals that experience early-life seizures have impaired long-term potentiation (LTP) (Lynch et al., 2000b) and dysfunction of hippocampal “place cells” that are critical for forming and maintaining spatial memories (Liu et al., 2003).

A variety of seizure-induced cellular and molecular alterations in the developing brain may contribute to impaired learning and cognition. Unlike adult animals, where prolonged seizures produce pronounced structural injury to the hippocampus including neuronal loss, particularly in CA1 and CA3 regions, and mossy fiber sprouting of dentate granule cells (reviewed in (Holmes and Ben-Ari, 2001)), most prolonged early-life seizures are not associated with significant cell loss or mossy fiber sprouting (Albala et al., 1984; Holmes and Thompson, 1988; Sperber et al., 1991; Jensen et al., 1992; Toth et al., 1998; Lynch et al., 2000a; Sankar et al., 2000; Lee et al., 2001; Bender et al., 2003; Chang et al., 2003; Zhang et al., 2004). Several acute repetitive seizure models, including kindled and pentylenetetrazol induced seizures, also result in minimal mossy fiber sprouting and cell loss in the immature animal (Golarai et al., 1992). Modest cell loss and synaptic reorganization have been seen in the hippocampal CA3 region after CRH induced seizures (Baram and Ribak, 1995; Brunson et al., 2001) and recurrent flurothyl seizures (Holmes et al., 1998; Huang et al., 1999; Liu et al., 1999) in the postnatal period. Further, neuronal damage in the mediodorsal nucleus of the thalamus (Kubova et al., 2001) has been documented following lithium-pilocarpine induced seizures as early as the second postnatal week, with the severity and extent of thalamic damage increasing with age at seizure induction (Druga et al., 2005). Overall, cell death and synaptic reorganization are absent or minimal in most models of early life limbic seizures known to produce cognitive deficits, indicating these changes may not be required for seizure-induced memory and learning deficits in the developing brain.

The birth of dentate granule neurons throughout life is important for the formation and stabilization of memories, and alteration of this process may be a mechanism of seizure induced cognitive changes (Feng et al., 2001; Shors et al., 2001; Snyder et al., 2001; Rola et al., 2004; Snyder et al., 2005). Depending on the exact model and timing, seizures are potent inducers or inhibitors of neurogenesis in the dentate gyrus (Parent et al., 1997; Scott et al., 1998; Parent et al., 1999; Scott et al., 2000; McCabe et al., 2001; Hattiangady et al., 2004). Changes in cell birth following seizures are also accompanied by an increase in cell death of both immature and mature dentate granule neurons (Sankar et al., 2000; Ekdahl et al., 2002; Porter et al., 2004). Seizures also appear to differentially affect transcription of some neurotransmitter receptors in mature, but not immature, dentate granule neurons (Porter et al., 2006). These data suggest that seizure induced changes in dentate granule neurogenesis could contribute not only to epileptogenesis, but also to learning and memory impairment in epilepsy.

Emerging evidence suggests that early-life seizures can alter the function of neurotransmitter systems and intrinsic neuronal properties in the brain, possibly contributing to cognitive and learning impairments. γ-Aminobutyric acid (GABA) is the main inhibitory neurotransmitter in the brain, and GABAA receptors mediate most fast synaptic inhibition. Changes in inhibitory neurotransmission are known to affect learning. Enhancement of GABAA receptor function with benzodiazepines disrupts LTP and memory formation (del Cerro et al., 1992; Seabrook et al., 1997) and GABAA receptor α-subunits have been shown to be key regulators of “critical periods” for cortical plasticity (Fagiolini et al., 2004) and hippocampal dependent spatial learning (Rudolph and Mohler, 2004). Evidence exists for enhanced inhibition after early life seizures that could impair these cognitive processes. On a circuit level, increased paired pulse inhibition in the hippocampus has been seen after both hyperthermic and kainate-induced seizures in the postnatal period (P7–14)(Lynch et al., 2000b). At the cellular/molecular level, early-life lithium-pilocarpine induced seizures produce an increase in GABAA receptor expression and a selective increase in the α1 subunit in the hippocampal dentate gyrus seen both immediately and later when the animals reach adulthood (Zhang et al., 2004; Raol et al., 2006). These alterations are associated with functional changes including enhanced type I benzodiazepine augmentation of the receptor (Zhang et al., 2004). This seizure-induced increase in the immature brain is in contrast to the α1 subunit expression decrease seen in adult rats following pilocarpine-induced seizures (Brooks-Kayal et al., 1998). These findings suggest that the effects of seizures on expression of GABAA receptor subunits are age-dependent, and that increased GABAA receptor expression and resulting enhanced inhibition could contribute to cognitive deficits following early-life seizures.

Changes in excitatory neurotransmission may also contribute to learning and behavioral differences after early-life seizures. Glutamate is the primary excitatory neurotransmitter in the brain and its activity is mediated by a variety of receptor subtypes including N-methyl-d-aspartic acid (NMDA) and non-NMDA (AMPA (α -amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) and kainate) ionotropic receptors and metabotropic receptors. Excitatory signaling through both the AMPA and NMDA receptors are critical for different types of LTP and hippocampal learning (Mongillo et al., 2003; Riedel et al., 2003; Yasuda et al., 2003; Schmitt et al., 2005), and mutant mice lacking subtypes of AMPA or NMDA receptors have impaired learning (Nakazawa et al., 2002; Mead and Stephens, 2003a, 2003b). Deficits in excitatory synaptic density and in excitatory signaling through both AMPA and NMDA receptors have been found after early-life seizures. Tetanus-toxin induced seizures in the postnatal period produce a 30% decrease in dendritic spine density on hippocampal CA3 neurons (Jiang et al., 1998), and a 30–40% decrease in NMDA receptor NR1, NR2A, and NR2B subunit proteins in hippocampus (Swann, 2004; Swann et al., 2006). Decreased AMPA receptor GluR2 subunit expression has been shown after postnatal hypoxia-induced seizures (Sanchez et al., 2001) and lithium-pilocarpine-induced seizures (Zhang et al., 2004).

In addition to the effects of early seizures on neurotransmitter receptor systems, seizures also alter a number of molecules essential for intrinsic neuronal function. Recent studies have demonstrated that prolonged postnatal febrile (hyperthermic) seizures produce a profound, long-lasting enhancement of intrinsic hyperpolarization-activated membrane current, Ih, in CA1 hippocampal neurons (Chen et al., 2001) due to a decrease in hyperpolarization-activated, cyclic nucleotide-gated channel-1 (HCN1) mRNA and simultaneous enhancement of HCN2 mRNA expression in hippocampal CA1 neurons (Brewster et al., 2002, 2005). These changes are associated with persistent limbic hyperexcitability and a 35% incidence of spontaneous seizures in adulthood (Dube et al., 2000, 2009; Chen et al., 2001). These results suggest that a variety of neurotransmitter receptors and ion channels may be permanently altered following early-life seizures.

Changes in neuromodulatory pathways may also contribute to learning and behavioral differences after early-life seizures. cAMP response element binding protein (CREB) is a key mediator of stimulus-induced changes in gene expression that underlies plasticity of the nervous system and phosphorylation of CREB is required for long-term potentiatioin (LTP), learning, and memory (Lonze et al., 2002). CREB phosphorylation and learning are both diminished after repetitive febrile seizures in animals (Chang et al., 2003). Corticotropin releasing hormone (CRH) is a neuromodulatory peptide released from hippocampal interneurons in response to stress. Early life seizures have been shown to enhance hippocampal CRH mRNA expression (Brunson et al., 2001), and excessive CRH (Chen et al., 2004) and early-life stress (Brunson et al., 2005) have been shown to lead to reductions in dendritic length and arborization as well as progressive cognitive deficits.

Animal studies highlight the complex role environment plays in determining the effects of early-life seizures on cognitive development. Two studies have examined the effects of timed periods of environmental enrichment on the cognitive development of rats exposed to early-life prolonged seizures (Faverjon et al., 2002; Koh et al., 2005). Although all rats that experienced seizures, regardless of environmental exposure, showed similar hippocampal damage, the rats provided with enrichment performed significantly better than normally housed rats in spatial learning and exploratory behavior tasks. These studies highlight the complicated interaction between seizures and environment that contribute to the cognitive outcome after seizures during development.

In conclusion, both clinical experience and epidemiological studies suggest that children with epilepsy are at risk for cognitive and learning deficits. Animal studies confirm long-term deficits in cognitive and behavioral function after early-life seizures and suggest a variety of seizure-induced cellular and molecular changes in the developing brain that may contribute to these effects. The clinician making treatment decisions for infants and children with epilepsy must be cognizant of these risks and weigh them alongside potential adverse cognitive effects of AEDs, which is the topic of the second half of this review.

AEDS AND THE DEVELOPING BRAIN

In accord with the clinical and basic science data presented above that seizures during early-life can have a negative impact on brain development, a recent survey found that the majority of pediatric neurologists felt that treatment of seizures in children leads to better outcomes, both acutely and in the long term (Wheless et al., 2005). There is, however, ample evidence from both clinical and animal studies that many of the AEDs used commonly to treat seizures in children may themselves interfere with normal brain development.

AEDs and the developing brain: human studies

Every year, approximately 24,000 children are born to women with epilepsy (Meador and Zupanc, 2004), raising concerns about possible drug-induced malformations as well as long-term developmental and cognitive effects of AEDs. The effects of AED exposure in utero on the occurrence of malformations in the offspring have been recently reviewed (see Dolk and McElhatton, 2002; Holmes et al., 2001; Artama et al., 2005; and Shorvon, 2002). Briefly, numerous studies of the older AEDs (phenobarbital, carbamazepine, phenytoin, and valproic acid) suggest that exposure during pregnancy increases 2 to 3 times the risk of minor malformations (e.g. subtle facial dysmorphisms, club feet) or major malformations (e.g., congenital heart defects, neural tube defects) (Yerby et al., 1992; Waters et al., 1994; Holmes et al., 2004). Studies assessing the risk of in utero exposure to the newer AEDs are limited, but the existing data suggest less overall teratogenesis with newer medications (Cunnington, 2004). This conclusion may be premature, however, since recent reports do show increased risk of cleft palate and other malformations with lamotrigine (Tennis and Eldridge, 2002; Hauser and Tomson, 2006; Meador et al., 2006) pointing to the need for continued vigilance.

In contrast to the teratogenic potential of AEDs, cognitive function in children exposed in utero to AEDs has not been extensively studied (Meador and Zupan, 2004). Existing studies have reported mixed results, but most find in utero exposure to AEDs increases the risk of cognitive dysfunction later in life. Children born to mothers with epilepsy who took an AED during pregnancy have lower mean IQs as compared to children born to mothers with epilepsy who did not take AEDs during pregnancy (Granstrom and Gaily, 1992; Koch et al., 1999; Adab et al., 2001; Adab et al., 2004). Other studies have shown limited (Leavitt et al., 1992) or no changes in developmental abilities after follow-up of less than a year (Wide et al., 2000). Well-controlled studies with large numbers of children exposed in utero to AEDs, including the newer AEDs, are needed to determine the safest drugs, associated with the best long-term cognitive outcome in their offspring, for the treatment of pregnant women with epilepsy.

If prenatal exposure to AEDs can diminish IQ and school performance, what are the effects of AED exposure during early childhood? For ethical and practical reasons, most studies of the AEDs' effects on cognitive development have been performed on children with epilepsy. An exception is a study of children with febrile seizures who were randomized to placebo or phenobarbital and underwent neuropsychological testing while undergoing treatment. The study found a decrease in the mean IQ of the phenobarbital treated group (Farwell et al., 1990). Long-term (3 to 5 years) follow up of the cohort that showed decreased IQ due to phenobarbital exposure found a persistent difference in reading comprehension tests, but not on more general cognitive testing paradigms (Sulzbacher et al., 1999). These findings suggest that exposure to phenobarbital in early childhood may result in mild but lasting differences in some cognitive modalities.

To tease out the specific effects of AEDs on cognition and learning in patients with epilepsy, different approaches have been used to control for the underlying epilepsy. These strategies have included testing cognition in patients before and after starting a medication (Farwell et al., 1990; Chen et al., 1996, 2001), testing medication(s) as add on therapy (Meador et al., 2003), crossover studies between drugs (Vining et al., 1987), and testing cognition at different drug levels (Trimble and Thompson, 1984; Bourgeois, 1998; Meador et al., 2001; Drane and Meador, 2002; Loring and Meador, 2004; Ortinski and Meador, 2004). One study comparing a small group of children sequentially treated with phenobarbital and valproic acid found that phenobarbital treatment reduced cognitive performance, decreased IQ, slowed reaction times, and had negative effects on behavior (Vining et al., 1987). These negative effects on cognition and behavior seen with phenobarbital have been born out in most but not all studies (Camfield et al., 1979; Calandre et al., 1990; Chen et al., 1996, 2001). Of note is the fact that the studies that show a relative impairment from phenobarbital treatment, compared with other older AEDs, often used a crossover design; these studies suggest that at least some of the impairments with phenobarbital are transient. Valproate has been shown to have variable effects, with the most severe cognitive changes associated with polytherapy (Adab et al., 2001).

There are essentially no data using formal neuropsychologic testing of children taking AEDs approved since 1990. The available relevant data are either from adult studies or are conclusions gleaned from the symptomatic complaints during clinical trials of the newer AEDs in children (Loring and Meador, 2004). In particular, there are several studies in adults suggesting topiramate impairs psychomotor function, verbal fluency, and attention. The side effect profiles reported in pediatric clinical trials of topiramate would suggest this might also be a concern for children (Meador et al., 2003, 2005; Reith et al., 2003). Other newer medications, such as lamotrigine and gabapentin, appear to have fewer cognitive effects in adults (Meador et al., 2005). However, all AEDs have somewhat worse effects than placebo (see reviews by Meador et al., (2001) and Drane et al., (2002). In conclusion, clinical studies show that many of the older AEDs have at least a transient (and possibly longer) effect on cognition and behavior in children. More clinical research is needed on the newer medications to understand their influence on cognition and brain development. As with the literature on seizures and development, animal studies are able to address issues of the cognitive effects of AEDs in a more controlled fashion, yielding insightful information.

AEDs and the developing brain: animal studies

To study the teratogenic potential of AEDs, pregnant rats and mice are exposed to drugs to look for effects on the offspring. The exposed animals can be followed into adulthood to identify long-term effects of in utero AED exposure on growth, health, reproduction, learning, and behavior. Studies have found deleterious effects of in utero exposure to phenobarbital, benzodiazepines, and carbamazepine, though in all animal studies of AED exposure equivalences to human drug levels can only be estimated. For example, half the pups from timed pregnant rats fed phenytoin (200 mg/kg) between prenatal days E7 to E18 (Schilling et al., 1999) showed diminished viability and an abnormal circling behavior. Further, offspring tested between postnatal day 50 and 70 on straight channel swimming performance and on versions of the Morris water maze had lasting reference-based spatial learning deficits. Other studies have shown similar effects on various aspects of working memory and spatial learning after prenatal exposure to phenobarbital, benzodiazepines, and carbamazepine (Vorhees, 1986; Vorhees et al., 1990, 2000; Weisenburger et al., 1990; Mantovani and Calamandrei, 2001). Similar studies of the newer AEDs are lacking and are needed to assess the impact of in utero exposure on brain development and cognition.

Surprisingly, only a small body of literature exists detailing the effects of AED exposure during early life. Phenobarbital worsened, but topiramate and gabapentin had no effect on spatial learning in water maze testing when given for several weeks during the first or second month of postnatal development, provided the drug was stopped prior to testing (Mikati et al., 1994; Cilio et al., 2001; Cha et al., 2002). A few studies have specifically tried to address the effect of concurrent AED exposure on memory tasks, but these tests only been performed on adult animals. In a study by Shannon and Love adult (10 week old) male rats were exposed to a number of older and newer AEDs. The investigators found that on working memory tasks GABA modulators such as phenobarbital and triazolam, or ethosuximde produced significant disruptions in performance, whereas carbamazepine and topiramate produced more modest effects, and tiagabine, lamotrigine, valproic acid, and gabapentin had no effects on working memory (Shannon and Love, 2004). Other studies showed similar findings in different cognitive modalities; triazolam, phenobarbital, chlordiazepoxide, and carbamazepine disrupted attention (Shannon and Love, 2005), and midazolam, chlordiazepoxide, and pentobarbital interfered with spatial learning (Kant et al., 1996; Keith et al., 2003). Similar studies are needed in younger animals to determine the effects of these AEDs on memory and learning in the immature brain.

Several studies exploring the effects of AEDs on memory or other cognitive tasks have treated animals following an induced seizure. Treatment with topiramate following a series of early life seizures resulted in a mild improvement in water maze testing as compared to animals treated with saline. Rats treated with valproic acid or gabapentin, following a kainic acid induced seizure, showed improvements in water maze testing and had less aggression and hyperactivity; however treatment with phenobarbital and phenytoin had an opposite effect (Mikati et al., 1994; Zhao et al., 2005). These findings suggest that some AEDs, such as phenobarbital, may have deleterious effects, while others may be beneficial for cognition in animals experiencing seizures.

There are several potential mechanisms by which AEDs may affect the developing brain including chronic alterations in gene regulation and protein expression. Several groups found that rats exposed to phenobarbital or benzodiazepines in utero or during postnatal development had changes in GABA neurotransmitter receptor subunits, synthesis enzymes, or transporter expression when assessed in adulthood (Rothe and Bigl, 1989; Gruen et al., 1990; Chesley et al., 1991; Tseng et al., 1994; Holt et al., 1997, 1997; Raol et al., 2005). These studies support the idea that drugs that alter GABA neurotransmission during development may permanently alter the molecular machinery that controls the neuronal responses to GABA throughout life.

Valproic acid is another AED that may change cellular physiology. Valproic acid alters gene expression patterns, with one mechanism being inhibition of histone deacetylase, an enzyme important for proper chromatin unfolding which is a critical determinant of transcriptional regulation. Valproic acid is known to regulate the expression of genes important in neuronal development such as homeobox genes and genes involved in neurotransmission (Faiella et al., 2000; Huang et al., 2002). The role of valproic acid in regulating gene expression may explain, at least in part, the high incidence of neural tube defects and neurocognitive deficits in infants exposed in utero to valproic acid (as discussed above).

During development, neurotransmitter receptor signaling plays a role in neuroblast migration, proliferation, and differentiation. Alterations in cortical development have been attributed to GABA and NMDA receptor modulators in in vitro models of neuronal migration (Kriegstein and Owens, 2001; Represa and Ben-Ari, 2005). For example, GABA has been shown to alter directed migration and random movement of cultured neuroblasts (Behar et al., 1996, 1998) and GABA antagonists have been shown to alter migration in slice culture (Barker et al., 1998; Behar et al., 2000). NMDA receptors are expressed on reelin positive Cajal-Retzius cells and their activation alters radial neuronal migration (Monti et al., 2002; Chen et al., 2005). This observation suggests that stimulators or inhibitors of these receptors during development may have profound effects on normal cortical development. Altered neuroblast migration during development may contribute to the cognitive changes that have been reported after in utero exposure to AEDs such as GABAergic modulators. Neuroblast formation and migration is completed in most brain regions by term, but in a few select areas it continues throughout life (Tozuka et al., 2005; de Graaf-Peters et al., 2006). Thus the effects of AEDs on neuroblast formation and migration are most relevant in the fetus, in preterm infants and for brain regions such as the dentate gyrus and ventricular zones where neuronal migration continues well into the postnatal period and beyond.

Normal programmed cell death during development, a process also known as apoptosis, is essential for normal network formation in the brain and has been found to be influenced by several AEDs. Olney, Ikonomidou, and their colleagues have shown that early perinatal exposure to some AEDs, alcohol, and anesthetic agents results in widespread enhancement of apoptosis during early postnatal development in rats (Ikonomidou et al., 2000; Bittigau et al., 2003; Jevtovic-Todorovic et al., 2003). Administration in 1–2 week old rats of phenytoin, phenobarbital, diazepam, clonazepam, valproic acid, and vigabatrin at doses thought to be clinically relevant resulted in wide spread neuronal cell death throughout the brain as determined by light and electron microscopy and increased DNA nicking (Bittigau et al., 2002). These results were relatively specific for these agents, as they were not found with administration of topiramate, leviteracitam or cholinergic, dopaminergic, AMPA, or calcium channel antagonists (Ikonomidou et al., 2000; Bittigau et al., 2002; Glier et al., 2004; Manthey et al., 2005). The AEDs had dose dependent effects, and there was a window of time between P0 and P14 when the drugs were most injurious to specific brain regions. While the mechanisms by which AEDs increase apoptosis are not known, there is a correlation between the down regulation of the neurotrophin brain derived neurotrophic factor (BDNF) and its signaling pathway and the increased apoptotosis seen histologically (Bittigau et al., 2002). These studies show significant increases in apoptosis after drug exposure but only a small (7–9%) decrease in brain weight one week after AED exposure at postnatal day 7. However, the exposed animals have not been extensively studied into adulthood (Bittigau et al., 2002) making it impossible to determine if the drugs change the overall amount, or just the rate, of the normal programmed cell death process. Similar studies have found apoptosis in developing granule cells of the cerebellum after phenytoin exposure at P2–P4 with associated minor deficits in motor control, on rotorod testing (but not on grip strength or straight walking) (Ohmori et al., 1999). In humans the major window for programmed cell death is completed around term, again suggesting that in utero and preterm exposure to AEDs are most worrisome for altering programmed cell death. In addition, these animal studies must be interpreted with caution as the dose equivalence between rodents and humans is not well established for all drugs; further these experiments were performed on normal (seizure naive) animals.

Formation, stabilization, and pruning of synapses occur throughout life, but these processes peak early in postnatal development (Lippman and Dunaevsky, 2005; Waites et al., 2005). The timing and strength of synaptic transmission fine tune synapses, stabilizing some and pruning others. AEDs alter synaptic transmission and therefore could theoretically impact synapse formation and stability. For example, benzodiazepines alter the timing and stabilization of synapses that contribute to ocular dominance columns during postnatal development (Fagiolini et al., 2004). Further studies are needed to understand how AEDs alter these critical developmental processes occurring throughout childhood.

In summary, there is ample evidence that AED exposure during early life in rodents especially with older drugs such as phenobarbital, phenytoin, and valproic acid alters neuronal development including gene expression, neuronal migration, differentiation, and survival. What is lacking is a better understanding of the long-term cognitive and memory consequences of exposure to AEDs during different developmental periods, especially those that correlate with human term infants and young children.

THE TREATING PHYSICIANS' QUANDARY

This review summarizes the recent clinical and animal literature on the impact of AEDs and seizures on the developing brain. No clear consensus can be generated from these data, but there appears to be some degree of detrimental effects of both seizures and AEDs during the fetal period, infancy, and early childhood. The extent of cognitive impact and the degree of exposure necessary to put a fetus or infant at risk, however, is unknown. The effects of seizures and AEDs during development, including an assessment of the full impact of exposure is most clear in the animal literature where the consequences of seizures and AEDs can be separated and shown to independently alter molecular expression, cell survival, and the neuronal circuitry for learning and memory formation. However, as in patients, the animal literature has not been able to separate out completely the effects of underlying brain malformations, seizures, and AEDs that all can influence behavior and learning.

So where does this leave a practicing neurologist or pediatrician? Clearly more data are needed so that physicians can maximize the developmental potential of children with epilepsy. Neurologists must continue to generate well-controlled clinical studies with large numbers of patients examining the long-term cognitive outcome of children with epilepsy. Neuroscientists need to generate additional basic science studies addressing the mechanisms of epilepsy' s influence on the developing brain and identifying novel therapeutics. In addition, immediate safety and quality of life concerns for the child and/or mother must be considered alongside or ahead of the concerns regarding long-term effects on cognitive development that this review engages. In the interim, there are several reasons to be hopeful: the newer AEDs appear to be generally less harmful to cognitive function. There is also a small amount of literature that suggests that in the face of epilepsy, AEDs may improve cognition and behavior. To be sure, there also are large gaps in our knowledge about the long-term effects of the newer AEDs in animals and in humans that need to be addressed before we can say with confidence which drugs are safest, least likely to cause cognitive impairment acutely and chronically, and which might alter the negative impact of seizures during development.

Ancillary