Epilepsy, Antiepileptic Drugs, and Aggression: An Evidence-Based Review

1. Temporal Lobes and Hippocampus

One of the most common and studied forms of focal epilepsy is TLE. Seizures originate in the temporal lobes (either the inner surface/structures or more rarely the neocortex) and can be associated with psychiatric symptoms, including mood changes and aggressivity (Zhao et al., 2014). The temporal lobe is also important in aggression; in milestone research by Klüver and Bucy (1937), bilateral temporal lobectomy in rhesus monkeys reduced aggression. The temporal lobe is part of the limbic system that controls emotions and memory, and subjects with histories of extremely violent behavior have shown metabolic abnormalities in the temporal lobes (Seidenwurm et al., 1997); functional and/or structural abnormalities in neural networks regulating emotions have been related to an increased susceptibility for impulsive aggression and violence (Davidson et al., 2000). TLE is often associated with an extensive loss of dentate hilar neurons and hippocampal pyramidal cells—the so-called “hippocampal sclerosis” (Sloviter, 1994)—and interestingly, hippocampal pathology has been identified in pathologic aggression (as discussed later).

2. Amygdala

The amygdala has been implicated in epilepsy, particularly in TLE and epileptogenesis, since the landmark studies by Feindel and Penfield (1954) and later by Goddard et al. (1969). Since this early work, many studies have shown that different forms of epilepsy are associated with damage to the amygdaloid complex (23 distinct subnuclei in humans) and to its connectivity to the surrounding brain regions and to the cortex (Pitkänen et al., 1998; Tebartz Van Elst et al., 2002; Takaya et al., 2014). The amygdala is also implicated in aggressive behavior: amygdalectomy in both animals and humans can stop aggressive behavior (Terzian and Ore, 1955), and stereotactic amygdalotomy of specific amygdaloid nuclei (e.g., the lateral or the anteromedial group) has been shown to control behavior in highly aggressive, treatment-refractory individuals (Mpakopoulou et al., 2008).

3. Frontal Lobes

Frontal lobe epilepsy, the second most common type of focal epilepsy, is also associated with aggressive behavior during and after seizures (Gedye, 1989; Sumer et al., 2007). Damage to the frontal lobes is implicated in violent and aggressive behavior: patients with frontal ventromedial lesions consistently demonstrated higher rates of aggression/violence (especially verbal confrontations) than patients with lesions in other brain areas (Grafman et al., 1996). Motor agitation and aggressive behavior have also been shown in patients with orbitofrontal seizures (Tharp, 1972), further supporting the role of dysfunction in this brain region in aggression (Giancola, 1995).

4. Hypothalamus

Other brain areas that are implicated in the pathophysiology of aggressive behavior and are components of emotional regulation circuits include the anterior cingulate cortex, amygdala, hypothalamus, septal nuclei, and periaqueductal gray matter of the midbrain (Fig. 2). The hypothalamus is an important part of the diencephalon that is implicated in both epilepsy and aggression. Studies in cats have shown that the hypothalamus, particularly the anterior medial hypothalamus, is involved in the modulation of defensive rage behavior (Gregg and Siegel, 2001). A positron emission tomography study in humans found low hypothalamic activity in male perpetrators of domestic violence and decreased correlations between cortical and subcortical brain structures (George et al., 2004). Anterior hippocampal asymmetries have been demonstrated in antisocial and violent subjects (Raine et al., 2004), and an inverse correlation was shown between hippocampal gray matter volume and lifetime aggression in borderline personality disorder (Zetzsche et al., 2007).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Brain regions that are important in both epilepsy and aggression. Schematic rendering of brain regions and nuclei involved in the neurobiology of both epilepsy and aggression. The role of each highlighted brain region/nuclei in epilepsy and aggression is discussed in the main text.

Taken together, it is clear that neurobiological networks that are important in epilepsy are also common neural substrates implicated in aggressive behavior. A summary of the networks and brain regions that are important in both epilepsy and aggression is shown in Fig. 2.

Within these brain regions and networks, the receptors and ion channels that are implicated in both epilepsy and aggression are also the targets of AEDs and mood stabilizers. In addition, alterations at the level of intracellular signaling cascades, gene sequence, or gene expression have been shown in both epilepsy and aggression in these areas.

5. GABA

Clinical and preclinical studies have shown that seizures are liable to occur with an increase in glutamate and/or decrease in GABA neurotransmission in the brain. However, the changes in glutamate and/or GABA levels are different in different types of epilepsy, animal models, and brain regions (Engel et al., 2007). GABA_A receptors are ligand-gated chloride ion channels and are the major inhibitory receptors in the central nervous system (Olsen and Tobin, 1990). Houser et al. (2012) have extensively reviewed GABA_A receptor expression, subunit composition, and function in epilepsy. Changes in GABA_A receptor expression and function have been reported in animal models of TLE (Peng et al., 2004), and the composition and function of GABA_A receptors changes not only in epilepsy but also after prolonged exposure to GABA_A allosteric modulators such as the AEDs diazepam and phenobarbital (Raol et al., 2005).

GABA neurotransmission has been also been investigated in the pharmacology and pathophysiology of aggression (Comai et al., 2012a). The situation is complex, and there is no clear agreement on whether GABA levels are decreased or increased in aggression, or whether enhancing GABA is anti- or proaggressive. For example, Bjork et al. (2001) found a negative correlation between plasma GABA levels and aggressiveness in psychiatrically healthy people with a family history of depression, whereas Lee et al. (2009) reported a positive correlation between GABA levels in the cerebrospinal fluid and impulsivity (but not aggression) in both nonmedicated normal controls and subjects with personality disorder. Allosteric modulators of GABA_A receptors, including barbiturates and benzodiazepines, have been shown to influence aggression levels in rodents with an inverted U-shaped dose-response curve: moderate doses induce aggression, whereas low or high doses reduce aggressive behavior (Miczek et al., 2002).

6. Glutamate

Glutamate is the principal excitatory neurotransmitter in the brain. It acts through ionotropic (NMDA, AMPA, and kainite) and metabotropic receptors and plays a significant role in the initiation, spread, and maintenance of epileptic activity. Research has demonstrated that epilepsy is linked to dysfunction of the glutamate system at different levels: genetic, neurotransmitter release, and receptor expression (Bradford, 1995; Barker-Haliski and White, 2015). Studies have reported increased plasma levels of glutamate in epilepsy (van Gelder et al., 1980; Janjua et al., 1992) and sustained increases in extracellular glutamate levels during seizures in the epileptogenic hippocampus (During and Spencer, 1993). Focal brain cooling, which can suppress epileptic seizures in refractory epilepsy, has been shown to significantly decrease glutamate levels in patients who have elevated glutamate in the cortex and/or hippocampus during seizures (Nomura et al., 2014). Glutamate is cleared from synapses by the membrane glutamate transporters and is loaded into synaptic vesicles by the vesicular glutamate transporters VGLUT1, VGLUT2, and VGLUT3 (El Mestikawy et al., 2011). A recent postmortem study comparing VGLUT expression in the hippocampus found significantly decreased VGLUT2 expression and significantly increased VGLUT3 expression in patients with TLE with hippocampal sclerosis, compared with autopsy controls (Van Liefferinge et al., 2015). Glutamate receptors, both ionotropic and metabotropic, have been extensively studied for their role in epilepsy in both human and animal studies (Moldrich et al., 2003; Ghasemi and Schachter, 2011; Szczurowska and Mareš, 2013). Glutamate receptors are targeted selectively and nonselectively by several AEDs (Table 2), and novel glutamate receptor ligands are also under development as potential AEDs (Löscher et al., 2013; Nicoletti et al., 2015).

Attention was initially directed toward antagonists of NMDA receptors, but they failed in several clinical trials (Rogawski, 2011). Interestingly, the NMDA antagonists phencyclidine (PCP) and ketamine have proconvulsant action at very high doses and anticonvulsant properties at low doses (Leccese et al., 1986). Research then moved to AMPA receptors (Rogawski, 2011), and 2012 saw the approval of the first selective AMPA receptor antagonist, perampanel, for the treatment of focal seizures. The role of kainate receptors in the pathophysiology of epilepsy is becoming better understood, particularly in TLE, and this was recently reviewed (Crépel and Mulle, 2015). Kainate receptors are located either presynaptically at both GABAergic and glutamatergic synapses, where they control neurotransmitter release and are involved in presynaptic plasticity (Schmitz et al., 2001), or postsynaptically in several regions deeply involved in epilepsy, including the cortex, hippocampus, and amygdala (Crépel and Mulle, 2015).

Few studies thus far have investigated the role of glutamate in aggression (Comai et al., 2012a); however, there is strong evidence of the involvement of glutamate in the pathophysiology and treatment of mood disorders such as depression (Kugaya and Sanacora, 2005; Niciu et al., 2014) that often coexist with aggressive behavior. Support for the glutamate hypothesis of aggression includes the demonstration of a positive relationship between cerebrospinal fluid glutamate levels and measures of impulsive aggression in both healthy human subjects and subjects with personality disorders (Coccaro et al., 2013). Studies in mice with genetic modifications at the levels of the genes encoding for the NMDA and AMPA receptors have shown altered levels of aggression in the resident intruder test (Brodkin et al., 2002; Duncan et al., 2004; Vekovischeva et al., 2004), an animal paradigm of aggressive behavior, suggesting a role for glutamate ionotropic receptors in aggression in rodents. In mice, the NMDA receptor channel blocker PCP produces a nonsignificant trend toward increased aggressiveness at low doses, whereas it seems to reduce aggression at high (near ataxic) doses (Belozertseva and Bespalov, 1999). The fact that medications that can decrease aggression (e.g., valproic acid and topiramate) have inhibitory effects at NMDA and AMPA receptors (Comai et al., 2012b) provides further supportive evidence for the involvement of glutamate in aggression; however, the association of these same AEDs with aggression in some patients with epilepsy shows that the mechanisms are complex.

7. Serotonin

Bonnycastle et al. (1957) were the first to demonstrate a link between anticonvulsant activity and 5-HT, showing that several AEDs, including phenytoin, significantly increased 5-HT levels. Pharmacological agents that elevate 5-HT levels (e.g., the selective serotonin reuptake inhibitor fluoxetine and 5-hydroxytryptophan) have anticonvulsant effects in several animal models of epilepsy (Pasini et al., 1992), whereas agents that reduce and/or deplete 5-HT (e.g., parachlorophenylalanine, a selective and irreversible inhibitor of tryptophan 5-hydroxylase) increase susceptibility to sound-induced seizures (Schlesinger et al., 1968). 5-HT receptors are expressed in almost all networks involved in epilepsy. With the exception of 5-HT₅ receptors, for which there is no evidence yet of involvement in epilepsy, Gharedaghi et al. (2014) recently reviewed and commented on the role of each 5-HT receptor subtype in epilepsy and seizure susceptibility.

Activation of 5-HT_1A receptors is anticonvulsant in various experimental seizure models. A study administering a 5-HT_1A agonist in lithium-pilocarpine–induced status epilepticus in mice showed that hippocampal 5-HT_1A receptors are involved in reducing seizure severity, whereas those located in extrahippocampal areas contribute to delayed seizure propagation (Yang et al., 2014). In patients with TLE, decreased 5-HT_1A receptor binding has been observed in the midbrain raphe, the ipsilateral thalamus, and the inferior region of the epileptogenic temporal lobe (Toczek et al., 2003). Genetic inactivation of 5-HT_2C receptors in mice produces rare spontaneous seizures that are occasionally fatal (Heisler et al., 1998).

5-HT has a central role in aggression and hundreds of articles have examined this topic since the early 1960s. It was believed that 5-HT generally inhibits aggression but recent findings challenge this hypothesis, showing that this may be limited to certain types of aggression, such as impulsive aggression, or perhaps instead to factors such as impulse control and emotion regulation (Krakowski, 2003). For example, the main metabolite of 5-HT, 5-hydroxyindoleacetic acid, is reduced in the cerebrospinal fluid of people with demonstrated auto- or heteroaggressive behavior compared with people who have never shown such behavior. These observations were made in suicidal individuals (Asberg et al., 1976), in nondepressed men with a history with aggressive behavior (Brown et al., 1979), violent men in prison (Linnoila et al., 1983), in impulsive arsonists (Virkkunen et al., 1987), and in patients with a personality disorder (Coccaro et al., 1989). The 5-HT_1A/1B agonist eltoprazine is considered a potent antiaggressive agent (Comai et al., 2012a).

8. Dopamine

The role of DA in epilepsy is still debated, although there is evidence of dopaminergic system involvement in certain animal models of epilepsy and in various forms of epilepsy in humans (Starr, 1996). In particular, alterations of subcortical dopaminergic pathways may be specifically related to the motor manifestations of certain types of seizures (Norden and Blumenfeld, 2002).

In general, DA seems to exert an antiepileptic action, as demonstrated by the fact that the nonselective D₁/D₂ agonist apomorphine, with certain limitations, has anticonvulsant properties, whereas neuroleptic drugs that act as D₁ and/or D₂ antagonists have predominantly proconvulsant actions. This is important, because some medications used to control aggression are D₁ and/or D₂ antagonists (Comai et al., 2012b). Looking at the role of specific DA receptors in epilepsy, evidence from both animal (Turski et al., 1988; al-Tajir et al., 1990) and human (Ring et al., 1994) studies has implicated D₂ receptors.

The dopaminergic system is mainly implicated in behavioral activation, motivated behavior, and reward processing (Ikemoto and Panksepp, 1999), but evidence suggests that it also modulates aggressive behavior. Cerebrospinal fluid levels of homovanillic acid, the final metabolite of DA, are lower in impulsively aggressive violent offenders with antisocial personality disorder than in nonimpulsively aggressive offenders with paranoid or passive-aggressive personality disorder (Linnoila et al., 1983). DA receptor antagonists (particularly conventional antipsychotics that target D₂ receptors, such as haloperidol) have been used effectively for decades to treat aggression in psychotic patients. Even though DA has been hypothesized to be necessary for aggressive behavior to occur, reflecting the motivational aspect of violence (Nelson and Chiavegatto, 2001; de Almeida et al., 2005), the exact role of DA in aggression is still unclear (Comai et al., 2012a).

9. Noradrenaline

Endogenous NA seems to be generally antiepileptic: NA levels are decreased after seizures, depending on the seizure type and brain region (Giorgi et al., 2004), and AEDs such as valproic acid increase NA levels in the rat (Baf et al., 1994). The antiepileptic actions of NA are very likely mediated by both α₁- and α₂-adrenoceptors; indeed, α₂-adrenoreceptor agonists have been shown to suppress, and α₂-adrenoreceptor antagonists to promote, seizures in kittens (Shouse et al., 2007).

In aggression, NA seems to play a permissive role, helping to determine whether an individual elects to fight or flee in response to a challenge (Miczek and Fish, 2005). Studies have explored the effects of an experimental depletion or increase of NA levels, as well as activation or inhibition of NA receptors, on aggression. However, research is still limited and no clear positive or negative correlation has yet been demonstrated. Some studies report reduced fighting tendency after depletion of brain NA in male mice with isolation-induced fighting as well as after intraventricular injection of 6-hydroxydopamine (Crawley and Contrera, 1976). Some studies, in contrast, report increased aggression (shock-induced fighting) after NA depletion in rats (Thoa et al., 1972). The role of α₂-adrenoceptors is important in mediating the effects of NA manipulations on aggression. For example, increasing NA levels with desipramine (a NA reuptake blocker) increases isolation-induced aggression in mice in a dose-dependent manner, and the α₂-adrenoceptor blocker yohimbine dose-dependently counters this increase (Matsumoto et al., 1991). In addition, the α₂-adrenoceptor agonist clonidine, which decreases NA neuronal activity by activating NA autoreceptors, has been shown to decrease pathologic aggression in humans and is used in clinical settings in irritable autistic children, children with conduct disorder (Fava, 1997), and also in adults with aggression (Comai et al., 2012b). Clonidine is also of value in treating ADHD in children, particularly those with conduct or oppositional defiant disorder (Connor et al., 1999, 2000).

10. Intracellular Signaling Cascades, Genes, and Epigenetic Gene Regulation in Epilepsy and Aggression

The role of intracellular events (e.g., intracellular proteins and signaling cascades, genes and epigenetic modifications) is another important framework in which to consider in the neurobiology of epilepsy and AED development (Löscher et al., 2013).

The study of intracellular cascade events after receptor activation has been actively researched in epilepsy. In the pilocarpine-induced status epilepticus model, for example, increased phosphorylation of the mitogen-activated protein kinases (MAPKs) extracellular signal-regulated kinase (ERK)-1, ERK2, and p38^MAPK has been demonstrated in the acute period (up to 12 hours after status epilepticus) and protein kinase B in the latent period (5 days after pilocarpine-induced status epilepticus) (Lopes et al., 2012). These biochemical changes in the serine-threonine kinases, which are implicated in neuronal survival and differentiation as well as neuroplasticity, may in turn alter gene expression and produce long-lasting neuronal changes. However, ERK signaling has multiple roles during epilepsy-related processes; its activation can contribute to acute seizure activity and might be necessary for epileptiform synchronization (Houser et al., 2008). The mammalian target of rapamycin is a regulator of mRNA translation that is itself regulated by ERK and protein kinase B (Li et al., 2010). As reviewed by Cho (2011), there is evidence to implicate abnormal activity of signaling molecules in the mammalian target of rapamycin pathway in epilepsy.

There has been less research on the intracellular signaling cascades underlying aggressive behavior. However, because these cascades (and the G protein–coupled receptors that activate them) are targeted by medications used to treat aggression (Beaulieu, 2012), it is reasonable to hypothesize their active involvement in the pathophysiology of aggression (Comai et al., 2012a). Aggressive behaviors are common among abusers of methamphetamine, and repeated injections of methamphetamine in mice have also been shown to increase aggressive behavior (Sokolov and Cadet, 2006). In these animals, there were significant alterations in the expression of proteins involved in MAPK-related pathways in the striatum and frontal cortex (Sokolov and Cadet, 2006).

Regarding the specific involvement of genes in both epilepsy/seizure susceptibility and aggressive behavior, studies have been very limited in number. Two genes that have been implicated in both seizures and aggression are the gene encoding the MAO-A enzyme and that encoding the 5HT_1B receptor. Complex changes in seizure susceptibility are seen in MAO-A knockout mice, which have high levels of 5-HT and NA. After pentylenetetrazol kindling, the latency to seizure development is shorter in knockout mice than in wild-type mice; however, the knockout mice then have fewer seizures per day and shorter-duration seizures compared with wild-type mice, suggesting some protective effect despite the increased susceptibility to kindling (Teskey et al., 2004). Regarding aggressive behaviors, MAO-A knockout mice are more aggressive than wild-type mice, displaying decreased latency to attack in the resident intruder test (Cases et al., 1995). The only study in humans that explored links between epilepsy and polymorphisms in the gene encoding MAO-A was negative, but the study was underpowered to detect minor differences (Stefulj et al., 2010). The same study instead showed a modest association between the G861C polymorphism in the 5-HT_1B receptor gene and TLE. Notably, a 5-HT_1B polymorphism has been associated with suicide history and personality disorder in humans (New et al., 2001) and 5-HT_1B knockout mice display increased levels of aggression in the resident intruder test (Bouwknecht et al., 2001).

Epigenetic factors, including seizures themselves and AEDs, can alter neural network dynamics by interfering with signaling pathways and enzyme and receptor expression. Epilepsy can be influenced not only by alterations in genetic and environmental factors but also by a spectrum of dysfunctional epigenetic factors and processes (Qureshi and Mehler, 2010; Kobow and Blümcke, 2014). Indeed, hundreds of misregulated genes have been identified in human and experimental models of epilepsy after epigenetic chromatin modifications and DNA methylation. Modifications, such as DNA methylation, may accumulate over time after an initial injury, and a methylation hypothesis of epilepsy has been proposed (Kobow and Blümcke, 2012); such epigenetic changes may explain interindividual differences in the emergence of epilepsy, such as after brain trauma (Machnes et al., 2013). Interestingly, the AED valproic acid has been shown to inhibit histone deacetylases and thereby normalize expression of histone deacetylase–dependent genes within the epileptic dentate area in rats with kainic acid–induced seizures (Jessberger et al., 2007).

The epigenetic modifications involved in aggressive behavior have not been well studied. Evidence thus far demonstrates a relationship between epigenetic modifications in the 5-HT system and aggression. In a recent study, we found decreased MAO-A activity in antisocial offenders, resulting from epigenetic modifications in MAO-A gene methylation levels (Checknita et al., 2015). Methylation levels of the 5-HT transporter gene promoter (SLC6A4) were also shown to be altered in individuals displaying physical aggression during childhood (Wang et al., 2012).

1. Paradoxical Proaggressive Effects of Enhancing GABA Neurotransmission

The GABA system has the potential for both inhibitory and excitatory actions. In the neonatal brain, GABA is excitatory and not inhibitory, due to the high intracellular concentration of chloride (Ben-Ari and Holmes, 2006). In the mature brain, rhythmic afterdischarge of pyramidal CA1 cells after cessation of the stimulus relies on a powerful GABA_A-mediated excitation mechanism (Fujiwara-Tsukamoto et al., 2003). In adults with TLE, the morphology, hippocampal expression, and subcellular distribution of GABA_A receptor subunits is markedly altered (Loup et al., 2000) and, in mature neurons, recurrent and prolonged seizures may trigger a pathologic re-emergence of immature excitatory features of GABA_A receptors (Galanopoulou, 2008). Moreover, TLE is often characterized by hippocampal sclerosis and thus depletion of GABAergic interneurons and glutamatergic hilar mossy cells, by mossy fiber (with colocalization of GABA and glutamate receptors) sprouting and synaptic reorganization in the dentate gyrus (Jinde et al., 2013). We hypothesize that in the epileptic brain, many modifications in GABA receptors and GABA-ergic neurons can occur, such that agents enhancing GABA neurotransmission (which decrease aggressive behaviors in subjects without epilepsy) can, depending on the brain region, have the opposite effect of triggering, rather than decreasing, aggression.

2. Dose-Dependent and Opposite Effects of NMDA Receptor Antagonists on Epilepsy and Aggression

The glutamatergic system, like the GABA system, may have dose-dependent opposite effects. Depending on the dose, NMDA receptor blockers such as PCP appear to have both pro- and antiepileptic as well as pro- and antiaggressive properties (Leccese et al., 1986; Belozertseva and Bespalov, 1999). In particular, PCP at low doses is antiepileptic but proaggressive, whereas the opposite is observed at high doses. Increased aggression might consequently occur when AEDs that have an NMDA receptor–inhibitory component to their action are used at doses that have minimal NMDA receptor antagonism. Some subtypes of epilepsy, such as TLE, can be associated with a reshaping of glutamatergic neurons and receptors in the hippocampus; consequently, drugs blocking glutamatergic receptors may have a different or opposing effect in the epileptic brain than in nonepileptic brains.

3. Genetic Predisposition and Decreased Dopaminergic Activity

A recent association study set out to explore whether there was a genetic basis that predisposes some patients to develop behavioral AEs during levetiracetam treatment. Helmstaedter et al. (2013) identified several polymorphisms, all of which were associated with reduced dopaminergic activity (variations in DA-β-hydroxylase, catechol-O-methyltransferase, and an intracellular D₂-receptor binding protein), which seem to predispose patients with epilepsy treated with levetiracetam to develop aggression. It is not yet known whether this result is specific to levetiracetam, or whether reduced dopaminergic activity will predispose patients with epilepsy to behavioral AEs with other AEDs

4. Forced Normalization

Finally, the sometimes controversial forced normalization theory is an attempt to explain the occasional observation of a paradoxical inverse relationship between epileptiform abnormality in EEG and psychiatric symptoms. It was first described in 1953 by Heinrich Landolt, who observed that EEGs paradoxically normalized and seizure activity was inhibited during psychotic episodes (Landolt, 1953). The forced normalization phenomenon is similar to but distinct from the concept of the “alternative psychosis” or “reciprocal psychosis” (that psychosis is better when seizures are worse and psychosis is worse when seizures are better/controlled, as discussed later). This hypothesis was supported by epidemiologic studies that found lower seizure frequency in patients with epilepsy and psychosis, and studies reporting relatively few cases involving comorbid schizophrenia and epilepsy (Schmitz and Trimble, 1992). This observation can be extended to psychiatric phenomena other than psychosis, such as hypomania/mania, aggression, depression, and anxiety (Trimble and Schmitz, 1998). Therefore, the psychiatric adverse effects seen during AED treatment may not be a direct adverse psychotropic effect of the AED, but rather a consequence of the suppression of seizures. This inverse link between seizure control and psychiatric symptoms (Flor-Henry, 1983) deserves further research, because it has far-reaching implications. Studies are needed to determine whether psychiatric symptoms seen with AEDs in patients with epilepsy are side effects with no prognostic or clinical value, whether they are a necessary consequence of seizure control, or whether they could be a stage of a complex and progressive phenomenon leading to a chronic psychiatric disorder (Mula et al., 2007). This is an important distinction, and future research should focus on clarifying the complex relationships between seizures and aggressive behavior.

1. Acetazolamide

No evidence was found.

2. Carbamazepine

No evidence was found.

3. Clobazam

We identified one double-blind, randomized trial with clobazam that included behavior-specific measures (Paolicchi et al., 2015). Using data from a trial of clobazam to treat Lennox–Gastaut syndrome, Paolicchi et al. (2015) carried out a post hoc analysis of all randomized pediatric patients (aged ≤18 years) treated with at least one dose of study drug or placebo. Of the 146 clobazam-treated patients, aggression-related AEs were seen in 23 (15.8%), compared with 8.3% of placebo patients. In patients taking high- and medium-dose clobazam, most of the aggression-related AEs occurred during the 3-week titration period, whereas they were evenly distributed in the low-dose and placebo groups during the 15-week study. The aggression-related AEs included the following MedDRA-preferred terms: aggression, irritability, abnormal behavior, perseveration, and negativism (although no patients fell into the perseveration category and only one fell into the negativism category). Three patients discontinued clobazam because of aggression-related AEs. There was no significant difference between clobazam and placebo in the behavior item scores on the Achenbach Child Behavior Checklist (CBCL), although there was a trend toward worsening scores in the aggression domain with clobazam in patients with a history of aggressive behavior (Supplemental Table 2). The authors concluded that the overall rate of aggression was low with clobazam, was dose dependent, resolved by study end, and was independent of the history of aggression/behavioral problems (Paolicchi et al., 2015). The other clobazam studies we retrieved did not use behavior-specific measures but reported behavior-related AEs in varying degrees of detail (Supplemental Table 2). In a small, randomized, blinded (double dummy) study, Bawden et al. (1999) compared 24 patients taking clobazam with 17 receiving standard monotherapy (9 received carbamazepine and 8 received phenytoin). Three of the patients taking clobazam exhibited externalizing behavioral AEs, compared with three who received standard monotherapy; three taking clobazam exhibited internalizing behavioral AEs, compared with two taking standard monotherapy (Bawden et al., 1999). The rate of behavioral AEs was very similar between the two groups but it is difficult to draw any conclusions from such small numbers. Sheth et al. (1994, 1995) reported that 7 of 63 children (11%) treated with clobazam for refractory epilepsy developed severe behavioral disturbance. “Aggressive agitation” was reported in the seven children (mean age 6.4 years). The aggression was described by parents as being “animal-like” and included biting, kicking, head-banging, tantrums, and hyperactivity, all of which were said to be out of character (Sheth et al., 1994, 1995).

Jan and Shaabat (2000) reported a small, open, uncontrolled study of 31 children with refractory epilepsy, aged 2 months to 15 years, in whom clobazam was added to the existing antiepileptic medication in dosages of up to 2 mg/kg per day. Seven of the 31 children had adverse effects, including behavioral change in two (Jan and Shaabat, 2000). The clobazam had to be withdrawn in three children because of repeated vomiting or behavioral changes but it was not clear, from the report, whether it had to be withdrawn in both of the children with the behavioral change.

Klehm et al. (2014) reported a large retrospective chart review in 300 children (mean age 9.1 years) with refractory epilepsy who were prescribed clobazam. The children had a variety of seizure types (many had multiple seizure types), and the majority of patients (97%) were taking at least one other AED. The median starting dosage was 0.2 mg/kg per day and the average dosage at last follow-up was 0.73 mg/kg per day (range, 0.05–3.3 mg/kg per day). The median seizure reduction was 80%; the 50% responder rate was 67.7% and 84 patients (28%) were seizure free at the last follow-up. Twenty-three patients (7.7%) were reported as having AEs related to mood or behavior change (Klehm et al., 2014).

4. Clonazepam

None of the clonazepam studies used behavior-specific measures; two observational studies and two reviews of previous studies were included (Supplemental Table 2). Mikkelsen et al. (1976) carried out a single-blind, placebo-controlled trial of clonazepam in 10 patients with simple absence seizures and 10 patients with myoclonic-atonic seizures. Most of the patients (17 of 20) were aged <18 years. Clonazepam had to be withdrawn from two of the patients because of severe irritability, dysphoria, and aggressiveness in one case and somnolence, behavioral disturbance, and lack of efficacy in the second case. Further details were sparse in this report.

Lander et al. (1979) reported that 22 of 40 patients treated with clonazepam for refractory epilepsy had “undesirable effects” attributable to the clonazepam, the commonest of which were drowsiness, loss of concentration, irritability, and aggression.

Kalachnik et al. (2002) reviewed behavioral adverse effects of benzodiazepines, including clonazepam, diazepam, and lorazepam, commenting that these are easily overlooked and under-recognized. They stated that behavioral adverse effects occurred in 13% of 446 individuals with mental retardation prescribed these AEDs for behavioral, psychiatric, or medical conditions. The rate of behavioral disturbance in the 208 individuals who had epilepsy was 15.4%.

In a review of clinical trials of clonazepam, Browne (1976) commented that behavioral disturbance occurred in a minority of patients, usually children. This represented an exacerbation of the previous disorder in some cases, but not in others. The behaviors were variously described as irritable, aggressive, excitable, irrational, antisocial, temperamental, violent, disobedient, noisy, and hard to discipline. In 10 studies, the percentage of behavioral disturbance ranged from 2% to 50% (median 17%). The behavioral disturbance sometimes resolved with a reduction in dosage but required discontinuation of the clonazepam in other cases.

5. Eslicarbazepine Acetate

In a prospective, open-label study of 29 children aged between 2 and 17 years, aggressive behavior was reported in one child and one other child had “aggression aggravated” (Almeida et al., 2008).

6. Ethosuximide

The only evidence retrieved for ethosuximide was two case studies. Yamamoto et al. (2001) reported that an 11-year-old boy with refractory myoclonic epilepsy and severe psychomotor delay had complete control of myoclonic seizures with ethosuximide but had “behavioural changes, more of the manic type.” This was attributed to the forced normalization phenomenon because the EEG was said to be almost normal during the episode (Yamamoto et al., 2001).

A case study by Chien (2011) described a 10-year-old boy who developed acute mania with psychotic symptoms and suicidal ideation with ethosuximide.

7. Gabapentin

Two observational studies that reported aggression from AE data and two case studies were identified (Supplemental Table 2). Khurana et al. (1996) carried out a chart review of 32 children treated with gabapentin. Behavioral AEs occurred in 15 patients. Physician intervention was required in four children: one became more withdrawn and the other three became more hyperactive and more aggressive with violent outbursts and mood swings (Khurana et al., 1996). In three of the four children, the behaviors returned to baseline after the gabapentin was stopped.

Lee et al. (1996) reported on seven children who developed behavioral adverse effects in association with gabapentin. It is not clear how these children were selected from the larger pool of 55 children treated at this center (Lee et al., 1996). In some cases, the behaviors were present before treatment but were exacerbated when the child was treated with gabapentin. New behaviors included oppositional defiant disorder (58%) and conduct disorder (33%). Again, because of the small numbers, it is difficult to draw any firm conclusions.

The results of studies carried out by one of the current authors and his colleagues on 14 teenagers, using a standard instrument for monitoring behavior (the Rutter Behavioral Scales) before and during treatment with gabapentin, did not confirm these results. There were no significant behavioral changes with gabapentin: two subjects moved from the “nondisturbed” to the “disturbed” behavioral range and one moved from the disturbed to the nondisturbed range. In a further behavioral study on the same pool of patients, 6 female and 10 male subjects were matched for sex as well as for age and other parameters (as closely as possible). Examination of the changes in the Rutter scale scores revealed no significant differences between the gabapentin-treated group and the comparison group (Besag, 1996).

A case study by Tallian et al. (1996) reported two children who had aggressive behavior with gabapentin. The first case was a 16-year-old boy in whom the seizures were fully controlled but his appetite decreased and he had problems sleeping. He then developed aggressive behavior with biting, slapping, scratching, growling, and “acting like an animal.” The gabapentin was discontinued and the behavioral disturbance resolved. When the gabapentin was recommenced and the dose increased, the aggression became marked. When the gabapentin was subsequently discontinued, the problem resolved again. The second patient was a 6-year-old girl with intellectual disability and ADHD but no history of aggressive behavior. When gabapentin was commenced, she stopped interacting socially; she became physically aggressive when confronted socially and she assaulted other children. A moderate improvement in behavior occurred when the gabapentin dose was decreased; it was continued because her parents judged her behavior as being tolerable and she remained seizure free.

A case series described by Wolf et al. (1995) reported behavioral problems in three children with learning disability when treated with gabapentin. The behaviors described included unprovoked outbursts of anger in case one, episodes of hyperactivity and oppositional behavior in case two, and outbursts characterized by throwing food, screaming, and fighting in case three.

It is difficult to assess the causative role of the gabapentin itself in all of these reports, not only because of the small numbers and open nature of the studies but also because of confounding factors such as drug interactions and changes in seizure control.

8. Lacosamide

Our searches for lacosamide data in children retrieved only small observational studies (Supplemental Table 2).

Gavatha et al. (2011) reported a study in 18 children (10 male and 8 female, aged 3–18 years) with intellectual disability. No aggression was specifically reported but there were two cases of irritability. No further details were provided (Gavatha et al., 2011).

Guilhoto et al. (2011) analyzed results from a retrospective study of 16 young people with treatment-resistant epilepsy (7 male and 9 female, aged 8–21 years, mean age 14.9 years). Lacosamide was withdrawn in one boy (aged 8.4 years) because of severe behavioral outbursts (Guilhoto et al., 2011).

Heyman et al. (2012) carried out a retrospective study of medical records of 17 children (10 male and 7 female, aged 1.5–16 years, mean age 8 years) with epilepsy taking lacosamide. Restlessness was reported in two patients but no other behavioral disturbance was reported (Heyman et al., 2012).

Kim et al. (2014) described a retrospective study of medical records that included 21 children (16 male and 5 female, aged 1.2–17.9 years, median age 13.9), 2 of whom discontinued lacosamide because of AEs (aggressive behavior and depression).

9. Lamotrigine

Two observational studies were retrieved (Supplemental Table 2). In a long-term, open-label extension study, Piña-Garza et al. (2008) followed 204 infants (aged 1–24 months), treated with lamotrigine after phase III studies. The only AE that they considered could be reasonably attributable to lamotrigine in >2% of patients was irritability, which occurred in 10 patients (5%) (Piña-Garza et al., 2008).

Cardenas et al. (2010) carried out a retrospective review of patients who developed “neurobehavioural adverse reactions to lamotrigine.” They identified nine children (seven male and two female, mean age 5 years) who became hyperactive and agitated, over a wide range of doses from 0.7 to 14 mg/kg per day. Five patients developed self-injurious and violent behaviors, two had severe insomnia, and the most affected patient (a 6-year-old boy) developed “extremely volatile mood and affect” with visual and auditory hallucinations together with insomnia. All nine patients improved markedly after discontinuation or dose reduction of the lamotrigine. These authors said that severe, reversible neurobehavioral disturbance associated with lamotrigine therapy had not previously been reported in the literature.

It should be noted that lamotrigine is now widely acknowledged as an AED that can improve mood significantly; large studies have established this in adults and the positive psychotropic effects of improving mood/behavior in young people with or without epilepsy have also been confirmed (Frye et al., 2000; Cramer et al., 2004; Biederman et al., 2010).

10. Levetiracetam

There are several reports of aggression in children with epilepsy treated with levetiracetam, including two RCTs with behavior-specific measures; however, there are also several reports of improved behavior (Supplemental Table 2).

Our searches identified one study that specifically explored the behavioral effects of levetiracetam in a randomized, placebo-controlled study, using standardized measures including the CBCL and the Child Health Questionnaire–Parent Form 50 (de la Loge et al., 2010). Patients received adjunctive levetiracetam (N = 64) or placebo (N = 34) for 12 weeks. The CBCL separates scores into a total problems score and a total competence score. Among the per-protocol population (levetiracetam, N = 46; placebo, N = 27), there was no difference between treatment groups in the total competence score but a significant difference in the total problems score (P = 0.020) between levetiracetam (worsening) and placebo (improvement). In the problems component of the CBCL, there was a significant worsening of aggression with levetiracetam versus placebo (P = 0.013), which drove the overall difference in the problems score. In the competence component of the CBCL, there was a small improvement with levetiracetam versus placebo in the activities subscale (P = 0.049). The Child Health Questionnaire–Parent Form 50 score showed little change during treatment and no significant between-group differences. A long-term extension that included 80 patients who continued from this study and 23 additional patients found no significant change from the baseline CBCL score (or aggression subscale scores) with levetiracetam (Schiemann-Delgado et al., 2012). From AE reporting in this study, aggression was seen in 7.8% of patients, irritability in 7.8%, and abnormal behavior in 3.9%.

One other study with levetiracetam specifically studied aggressive behavior but did not use standardized measures. In 12 children with epilepsy with continuous spikes and waves during slow sleep and pervasive developmental disorder, parents reported the frequency of seizures and frequency of episodes of panic or aggressive behavior. In the eight patients whose seizures improved with the addition of levetiracetam to existing AEDs, six patients also had a ≥50% reduction in the frequency of episodes of panic or aggression; there was no change in episode frequency in the other two responders (Kanemura et al., 2014).

There have been many RCTs and observational studies with levetiracetam in children and adolescents; rather than review them all here, the results of a recent systematic review and meta-analysis of the behavioral effects of levetiracetam are presented (Halma et al., 2014). These authors included studies in children aged from 1 month to 18 years with a diagnosis of epilepsy, who were taking oral levetiracetam as monotherapy or add-on therapy, with follow-up of at least 2 weeks and which reported behavioral side effects. They excluded case studies or case series with fewer than 10 patients, studies of neonatal convulsions, and studies that included adults and did not have separate subgroup analyses for patients aged <18 years. Their review identified 13 studies in 727 patients. In the three RCTs they identified, the most frequent behavioral AEs with add-on levetiracetam (N = 165) were hostility (7.3%), nervousness (6.1%), and aggression (4.9%). A meta-analysis of these studies revealed a statistically significant increased risk of behavioral AEs with levetiracetam (risk ratio 2.18 versus placebo; 95% confidence interval, 1.42 to 3.37). Ten observational studies met their selection criteria and these reported both worsening and improvement of behavior with levetiracetam. Levetiracetam add-on therapy was associated with behavioral AEs of irritability (4.7%), hyperexcitability (4.4%), and aggression (2.7%); monotherapy was associated with behavioral problems in general (19%) and irritability (2.6%). No meta-analysis was possible across the observational studies.

Our searches identified the majority of the studies included by Halma et al. (2014) in their systematic review and meta-analysis, as well as a number of additional studies that were excluded by their selection criteria. Details of all of these studies can be found in Supplemental Table 2; the rates of behavioral AEs they report are broadly in line with those reported by Halma et al. (2014).

Mbizvo et al. (2014) performed another meta-analysis of levetiracetam studies, which was also discussed earlier (section V.B.11). This meta-analysis only included two RCTs in children/adolescents, whereas Halma et al. (2014) included three; when all of the captured behavioral AE terms were combined across these two studies for the child/adolescent populations, an incidence of 40.6% (versus 21.4% with placebo) was found. This seems remarkably high, especially considering the fact that the analysis excluded behavioral AEs that were not reported in the original publications because they were below the reporting thresholds (≥5% or ≥10%); however, it is broadly consistent with the doubling of risk (i.e., a risk ratio of approximately 2 versus placebo) as reported by Halma et al. (2014). The terms included by Mbizvo et al. (2014) were hostility, personality disorder, nervousness, depression, aggression, agitation, emotional lability, psychomotor hyperactivity, irritability, abnormal behavior, altered mood, anxiety, and dissociation. It was not clear, from their article, whether they excluded “double counting” (e.g., adding together reports of AEs of aggression and agitation when these occurred in the same subject); if they did not exclude double counting, this could explain the very large percentages of behavioral AEs in both the levetiracetam and placebo groups.

11. Oxcarbazepine

Very little evidence of any aggression-related features was found for oxcarbazepine, and just one open-label study was identified (Supplemental Table 2). Northam et al. (2005) carried out a prospective, open-label study of 24 young patients (aged 2–45 months). Oxcarbazepine was associated with irritability in 5 of 24 patients (21%) in the treatment phase (up to 30 days) and in 7 of 20 (35%) in the 6-month extension phase (Northam et al., 2005). Oxcarbazepine was discontinued in one of the patients with irritability (who also had fatigue and ataxia). The usual comments about the reliability of findings from small, open, uncontrolled studies apply.

12. Perampanel

Our searches identified one study with perampanel in adolescents that used standardized measures to assess behavior, one pooled analysis of phase III clinical trials, and several observational studies (Supplemental Table 2). Lagae et al. (2014) reported a phase II, randomized, placebo-controlled trial of add-on perampanel in 133 adolescents with refractory focal-onset epilepsy; behavior was assessed with the CBCL. There was no significant difference between perampanel and placebo in the change from baseline in CBCL total problem or total competence scores, or in any of the subscales (e.g., aggression). AEs related to hostility or aggression occurred in 17.6% (n = 15) of the perampanel-treated subjects (aggression, n = 7; irritability, n = 6; anger, n =2; and laceration, n = 1) and 4.2% (n = 2) of the placebo group (aggression, n = 1; irritability, n = 1).

Rosenfeld et al. (2015) reported pooled AE data for the 143 adolescents in the three phase III trials of perampanel. The most common AEs included aggression in 8.2% of adolescents versus 0% for placebo; this was more frequent than in the overall perampanel-treated population (1.6%). Furthermore, aggression was reported as being one of the most common reasons (6.6%) for interruption or dose adjustment of the perampanel among adolescents during the extension phase (Rosenfeld et al., 2015).

The other perampanel reports come from observational studies. Biró et al. (2015) published a retrospective analysis of 58 children (mean age 10.3 years; range, 2–17 years) treated with perampanel. Aggression was reported in eight patients (13.8%). In a retrospective study of 18 patients (age range, 4–19 years), Philip et al. (2014) reported behavioral change in only 1 patient (5.6%).

13. Phenobarbital

Four studies were identified for phenobarbital (Supplemental Table 2). Willis et al. (1997) carried out a neuropsychological and EEG study of 11 children with epilepsy aged 7–14 years treated with phenobarbital and mephobarbital. They stated that parents reported clear behavioral changes in 6 of 11 subjects, including irritability, oppositional attitude, and overactivity. In four of the six patients, the changes were relatively mild and the barbiturate was not discontinued.

In a large observational study that used parent questionnaires, Domizio et al. (1993) compared 197 children (116 male and 81 female, mean age 5.3 years) treated with phenobarbital, with 103 children (66 male and 37 female, mean age 6.4 years) who were treated with other AEDs. In the phenobarbital group, 150 children (76.1%) had one or more behavioral disturbances, compared with 32 (31%) in the other group (P < 0.0001). Hyperactivity was the most frequent behavioral disorder (Domizio et al., 1993).

In a randomized study of four AEDs (phenobarbital, phenytoin, carbamazepine, and sodium valproate), de Silva et al. (1996) discontinued the phenobarbital arm after 6 of the first 10 children taking this drug had unacceptable AEs, which were primarily behavioral. In contrast, Pal et al. (1998) found no behavioral effects of phenobarbital (N = 47) in a study in 94 children in India.

14. Phenytoin

Although anecdotally phenytoin is associated with behavioral disturbance in young people, no firm published evidence for this was found (Supplemental Table 2). In the study in India referred to above, Pal et al. (1998) also found no behavioral effects of phenytoin in 47 children. Krishnamoorthy et al. (1983) reported three cases of phenytoin-induced choreoathetosis associated with agitation/restlessness in three young children: two were aged 2 years and one was aged 15 months.

15. Pregabalin

No evidence was found.

16. Primidone

Despite the fact that primidone is partly metabolized to phenobarbital, no evidence of aggression in children and teenagers with primidone was found.

17. Retigabine

We identified only one study. Groening et al. (2012) treated 17 patients (aged from 1 year and 10 months to 19 years) with retigabine for pharmacoresistant seizures (Supplemental Table 2). Results were analyzed for 12 patients, and AEs included hallucinations, agitation, and personality changes. Further details were not provided in this conference abstract (Groening et al., 2012).

18. Rufinamide

In a recent consensus paper on rufinamide in childhood epilepsy by Coppola et al. (2014), aggression and related behaviors were not reported among the adverse effects. There have been few other reports of behavioral adverse effects with rufinamide (Supplemental Table 2).

In a multicenter, prospective, add-on, observational study of rufinamide in 43 children, adolescents, and adults with Lennox–Gastaut syndrome (26 male and 17 female, aged 4–34 years, mean 15.9 years), Coppola et al. (2010) reported irritability/aggressiveness in 3 patients (6.9%). In a separate observational study, the effects of rufinamide in encephalopathies other than Lennox–Gastaut syndrome were reported in 38 patients (19 male and 19 female, aged 4–34 years, mean age 13.7 years). Irritability/aggressiveness was seen in two patients (5.3%) (Coppola et al., 2011).

In a prospective, open-label, add-on trial in refractory epilepsy in 69 children and adolescents, Cusmai et al. (2014) reported irritability in 11 patients (15.9%). In a retrospective study in 23 patients in Korea (age range, 4–22 years), Lee et al. (2013) reported aggressive behavior in 2 patients (8.7%).

19. Stiripentol

One observational study was identified (Supplemental Table 2). Inoue et al. (2009) reported clinical results using stiripentol to treat Dravet syndrome in patients aged 1–22 years. Of 23 patients, 6 had hyperactivity/irritability early in treatment, which resolved with continued treatment or dose reduction. One patient (a 15-year-old girl) discontinued stiripentol because of early irritability (Inoue et al., 2009).

20. Tiagabine

No evidence was found.

21. Topiramate

Two large retrospective studies, together with several smaller studies and case studies, were identified (Supplemental Table 2). Reith et al. (2003) reported data in 159 patients with epilepsy aged <18 years who were taking topiramate. Follow-up was possible in 127 patients (0.5–17.9 years); aggression/psychosis was a treatment-limiting AE in 10 of these patients (7.9%) (Reith et al., 2003). Grosso et al. (2005) treated 59 children aged ≤2 years with topiramate for localization-related and generalized epilepsies. Irritability was listed as one of the most frequent AEs, but precise rates were not given (Grosso et al., 2005). During topiramate treatment, Lee et al. (2011a) reported that 4 of 28 infants (14.3%) aged 2–18 months with West syndrome developed irritability, and Endoh et al. (2012) reported that 5 of 33 children (15.2%) aged ≤12 years with epileptic spasms developed irritability.

Metabolic acidosis is a known adverse effect of topiramate, and this can be associated with irritability/aggression (Ko and Kong, 2001). In patients taking topiramate who present with hyperpnoea or mental status change, metabolic acidosis must be excluded as a possible cause.

22. Valproic Acid

Two studies were identified with sodium valproate in adolescent epilepsy populations (Supplemental Table 2). In a retrospective study of 100 children with epilepsy treated with sodium valproate, Egger and Brett (1981) reported aggressive behavior in 4 (4%).

Ronen et al. (2000) studied eight children (six male and two female, aged 6–12 years) without clinical seizures but with abnormal EEGs and significant developmental learning disorder. While taking sodium valproate, the children were more distractible, had increased delay in response time, and had lower memory scores. Their parents also reported higher internalizing scores on the CBCL while the children were taking valproate (Ronen et al., 2000). In contrast, sodium valproate and divalproex sodium are used extensively as mood-leveling drugs in both adults and children in nonepilepsy populations; for example, Hollander et al. (2000) reported that impulsivity and aggression were decreased in 14 patients with autism taking divalproex sodium as a psychotropic medication.

23. Vigabatrin

We found no RCTs of vigabatrin with behavior-specific endpoints, and we found one observational study in children and adolescents that focused on behavior (Supplemental Table 2). Sheth et al. (1996) reported that in 31 patients aged 1–22 years (mean age 12.6 years) with refractory focal and generalized seizures, add-on vigabatrin was associated with negative behavior change in 6 children (based on parent reports), with discontinuation in 1 patient with severe aggressive agitation and positive behavior change in 1 patient.

The remaining studies we identified relied on AE reporting from five observational studies, three case reports, and six studies in babies with infantile spasms (Supplemental Table 2). Gobbi et al. (1999) carried out a prospective study of vigabatrin monotherapy in the treatment of focal epilepsies in 40 children (mean age at last visit 7.5 years) compared with 40 children treated with carbamazepine monotherapy. They stated that tolerability was good in the vigabatrin group but 4 of 37 patients had mild irritability at the end of the trial (versus none in the carbamazepine-treated group). Raucci et al. (1994) studied 61 children with various types of epilepsy treated with vigabatrin, 12 as monotherapy and 49 as add-on therapy. Vigabatrin was discontinued in six children because of adverse effects, including irritability.

In the first of three case studies, Weber et al. (2012) reported psychosis associated with vigabatrin in an adolescent girl with refractory symptomatic epilepsy after an early middle cerebral artery insult. The onset of the psychosis was 7 weeks after the vigabatrin was commenced, when she had been seizure free for 2 weeks. They attributed this to the phenomenon of forced normalization, although it should, more accurately, be termed alternative psychosis or reciprocal psychosis (the situation when the psychosis is likely to occur when the seizure control has improved; see earlier). Two of the teenage patients of one of the current authors also developed psychosis with vigabatrin, which resolved with dose reduction or discontinuation (F. Besag, personal communication). Cánovas Martínez et al. (1995) published a case of a 7-year-old boy with refractory epilepsy who developed an acute psychosis 3 days after rapid introduction of vigabatrin. The psychosis resolved within 48 hours of discontinuing the vigabatrin. Recommencement of the vigabatrin 2 months later, using a slower dose escalation, was well tolerated with no return of the psychosis. Chiaretti et al. (1994) also published a single case report of psychosis with vigabatrin in a child.

The data in infantile spasms are not presented here, because AEs such as irritability in young infants cannot be interpreted as aggression-related behaviors, but studies are listed in Supplemental Table 2.

24. Zonisamide

We identified one study with behavior-specific outcomes and five additional studies reporting behavioral AEs (Supplemental Table 2).

Eun et al. (2011) studied the neuropsychological and behavioral effects of low-dose and high-dose zonisamide, using a Korean version of the CBCL, in children aged 2–16 years receiving zonisamide monotherapy for newly diagnosed epilepsy. Data were available for 63 patients (27 and 36 receiving low-dose and high-dose treatment, respectively) and were presented as group data, so it is not possible to determine whether some patients became more aggressive and others less aggressive. Overall, Eun et al. (2011) saw a significant improvement (P < 0.05) in various parameters, including aggressive behavior, in the low-dose group and a nonsignificant improvement in the high-dose group. No aggression-related or behavior-related AEs were reported.

Cross et al. (2014) carried out a pooled analysis of 17 zonisamide studies in patients aged 16 years or younger, including 4 randomized, double-blind trials. Irritability was reported in 5.8% of the 391 zonisamide-treated patients but was not among the AEs commonly leading to discontinuation. Irritability was somewhat more common (7.5%) in patients aged 6–11 years than in patients aged 12–16 years (<5%).

A phase III study by Guerrini et al. (2013) was included in the meta-analysis by Cross et al. and is thus not discussed in detail here. No aggression-related AEs were reported in the zonisamide-treated group, but one patient in the placebo group discontinued due to aggression (Guerrini et al., 2013). In the subsequent long-term extension study, there were no behavioral AEs in patients continuing on zonisamide from the phase III study but there were two cases of aggression (2.8%) in patients who switched from placebo to zonisamide during the extension (Guerrini et al., 2014).

Coppola et al. (2009) carried out a prospective, add-on, open-label study of 82 young patients (45 male and 37 female, aged 3–34 years, mean age 13.1 years). Irritability was reported in nine patients (11.0%) but resolved in most cases with dose reduction. No other behavioral adverse effects were reported (Coppola et al., 2009).

Hirai et al. (2002) treated 27 children who had idiopathic epilepsy with zonisamide in a prospective study, observing behavioral disturbance in 2 children (7.4%). One of the cases developed an obsessive-compulsive disorder. The other case, a 14-year-old girl with focal seizures whose seizures were treated effectively with zonisamide from age 6 years, developed selective mutism, violent behavior, and lack of concentration at age 10 years. Decreasing the zonisamide dose was said to have maintained adequate seizure control while resolving the behavioral disturbance. Because of the long time interval between the prescription of zonisamide and the development of the violent behavior, the direct role of the AED is questionable in this case; a more likely explanation might be that the violent behavior was an interaction between seizure control and developmental factors.