Introduction

Although classically referred to as antiepilepsy drugs (AEDs), these medications provide symptomatic treatment of seizures; no medications have been shown to modify the course of epilepsy itself. The goal of epilepsy treatment, whether by medical or surgical means, was simply stated by the Centers for Disease Control and Prevention to be “no seizures, no side effects.” A drug that completely controls seizures but results in constant tiredness should not be considered an acceptable treatment. With more severe seizure disorders, the principle remains the same: to ensure the effects of chronic daily side effects of AEDs are not worse than those of the intermittent seizures.

AEDs are structurally unrelated molecules. Their pharmacokinetics and mechanisms of action are varied [1], which may be related to the variety of adverse effects and dosing schedules, which are very important in the choice of drug for a particular patient. The dramatic increase in the available agents over the past decades (Table 1), many with improved tolerability profiles (Table 2), makes seizure freedom without adverse effects a reality for increasing numbers of patients.

Table 1 Practical aspects of antiepileptic drug use
Full size table
Table 2 Advantages and disadvantages of antiepileptic drugs
Full size table

The chronic adverse effects of AEDs are also relevant. Studies have identified long-term problems with bone health and with neurodevelopment with in utero exposure, although data generally are not as available for the newer AEDs. An increase in suicidality was considered an AED class effect by the US Food and Drug Administration (FDA), which has raised controversy, and there remains confusion surrounding the bioequivalence and seizure protection properties of generic versions of AEDs.

General Principles: Selection and Use of Antiseizure Drugs

The first step in AED selection is matching for appropriate seizure type (Fig. 1). The various valproate (VPA) formulations are considered the gold standard for treatment of the idiopathic (primary) generalized epilepsies [2, 3], although their first-line use is diminishing. Levetiracetam (LVT), lamotrigine (LTG), zonisamide (ZNS), topiramate (TPM), clonazepam, clobazam, and now rufinamide (RFN) are considered broad spectrum, meaning they can treat any seizure type. These agents may be tried before VPA to avoid its potential side effects, particularly teratogenesis in women of childbearing age [4]. Other AEDs are considered narrow spectrum, as they treat only partial seizures (simple or complex) and both secondary and primary generalized tonic–clonic seizures. These agents may worsen other primary generalized seizures, such as absence or myoclonic seizures. Individual drugs may be better suited to specific seizures—for example, ethosuximide for absence seizures and clonazepam for myoclonic seizures [5].

Fig. 1
figure 1

An algorithm outlining steps in the choice of an antiepileptic drug for a patient diagnosed with epilepsy. The first step is to determine the epilepsy syndrome and seizure types. When these are unknown, broad-spectrum agents are safest. There are several choices for each seizure type, but the choices are narrowed based on the potential for intolerable side effects, of which psychobehavioral and weight effects are among the most concerning for patients. Enzyme inducers generally are considered second line for patients with long-term adverse effects, such as bone density loss, and to decrease the complexity of drug–drug interactions. ACTH—adrenocorticotropic hormone; ETH—ethosuximide; GTC—generalized tonic–clonic (seizure); LTG—lamotrigine; RFN—rufinamide; TPM—topiramate; VGB—vigabatrin; VPA—valproate

Full size image

There have not been large variations in the proven efficacy of AEDs for partial seizures. Head-to-head trials have been performed testing the effectiveness, or retention, of medications, which is affected by both efficacy and tolerability [6, 7]. There are no adequate predictors as to whether a specific AED will be efficacious in any individual patient; one may be effective when another is not. Thus, once seizure type has been taken into account, much of drug selection involves choosing medications with side effect profiles least likely to cause additional problems, or using other properties of certain AEDs to help the patient in comorbid conditions (Table 3). Examples include the use of TPM in patients who also have chronic headache, LTG in patients with depression or bipolar disease, ZNS for patients who wish to lose weight, or pregabalin (PGB) for those who wish to gain weight. Pharmacokinetic properties such as once-daily dosing may be particularly important for some college students or others in whom compliance is difficult. Finally, with most AEDs now having multiple generic versions, cost may play a role in drug selection and in the decision to stay with the brand-name version. Specific AEDs may have more published data to support monotherapy use [8•, 9•, 10], although most have been used successfully in clinical practice regardless of labeled indication.

Table 3 Psychobehavioral and weight gain/loss effects of antiepileptic drugs
Full size table

About half of patients achieve seizure freedom with the first agent attempted, but the chance decreases precipitously with subsequent trials [11]. When the first AED fails, the decision to add another medication in combination versus changing to a single new medication is debatable. Some experts advocate using polytherapy only after three monotherapy trials have failed [12]. One rationale is that polytherapy has a higher chance of causing side effects that would reduce quality of life more than continued seizures. Other authors lean toward using add-on combination therapy earlier [13], as delays to seizure freedom risk irreversible psychosocial consequences. The decision, much like the choice of AED, comes down to what is best for that individual patient.

Most older AEDs have between seven and 18 licensed generic versions. The FDA requires generics to contain the same active compounds as the branded agent. Testing in healthy individuals must show 95% confidence intervals of both peak serum concentration and total absorption falling between 80% and 125% of those of the brand-name agent. For many conditions, this amount of variability may be clinically unimportant. In epilepsy, the margin between efficacy and side effects may be small, and a relatively small decrease in the delivered dose might result in breakthrough seizures, with potentially disastrous consequences. An increase in absorption might result in toxicity. Changing from one generic with a relatively low resulting concentration (compared with the brand) to another generic with a relatively high concentration might amplify fluctuations in serum concentration from one prescription to the next. In rare cases, patients also may develop adverse effects related to the inert matrix of the tablet or capsule, such as an allergic reaction to dye. If the development of such an effect requires stopping or changing medication, even temporarily, seizures in a previously well-controlled patient may occur and result in injury, driver’s license suspension, or even death. The American Academy of Neurology published strong statements against substitution of generic for brand-name AEDs without the consent of the ordering physician [14•]. However, the random nature of seizures and the heterogeneity of epilepsy create problems in proving causality, and the debate over generic medications and therapeutic equivalence continues [15••].

Plasma drug levels are commonly monitored with older agents, and levels are available for most newer drugs as well. Most older AEDs (phenytoin [PHT], carbamazepine [CBZ], phenobarbital [PB]) have a narrow, well-defined therapeutic range, although specific patients vary in terms of a level that is effective and one that causes toxic symptoms. Plasma level monitoring is not required “routinely,” particularly for newer drugs, but may be helpful in specific clinical situations, including establishment of a drug level baseline once the patient is seizure-free, suspected toxicity, suspected noncompliance, medico-legal verification of treatment, lack or loss of therapeutic effect, and verification of level when change is suspected due to age, secondary disease, changing physiologic state (eg, pregnancy), or drug–drug interaction [16].

Update on Adverse Events

Rashes and Drug-Induced Hypersensitivity Syndrome/DRESS Syndrome

Drug rash with eosinophilia and systemic symptoms (DRESS) syndrome has been receiving greater attention. This rare but potentially fatal condition is characterized by relatively late onset; slow evolution, often mimicking an infectious illness with fever, rash, hepatic dysfunction, hematologic abnormalities, and lymphadenopathy; and reactivation of human herpesvirus [17]. When specific to AEDs, it also may be termed anticonvulsant hypersensitivity syndrome, of which PHT, CBZ, PB, LTG, oxcarbazepine (OXC), and primidone (PRM) have been implicated as a cause in the pediatric population [18]. Onset typically is within 2 months to 3 months of initiation of the implicated AED, and severity may worsen despite its discontinuation.

A retrospective database review of nearly 2,000 patients at an epilepsy center showed an overall rash rate of 2.8%, but in these patients, the risk of rash from exposure to another AED was 8.8% [19]. Cross-reactivity of rashes was generally high between AEDs, particularly between PHT and CBZ [20]. Pharmacogenomics and ethnicity also may be important: CBZ-induced Stevens-Johnson syndrome is strongly associated with HLA-B*1502 in Taiwan and Hong Kong and in those of Asian descent in Europe, but not in Japanese or other Caucasians. It appears specific to people of Han Chinese descent [21].

Suicidality

Since 2008, the FDA has mandated the product labeling of all AEDs to include a warning about increased risks of suicidality. This decision was based on a meta-analysis of 199 randomized clinical trials of 11 AEDs, not only for conditions of epilepsy, but also for mood disorders and pain. The data showed a 1.8-fold risk of suicidal thoughts or acts in patients on active study drug versus placebo. Although suicidality, and mood disorders in general, are of definite concern in patients with epilepsy, arguments against broad class labeling include known differences in the mechanisms of action and adverse effects of individual agents, and the nature of spontaneous adverse event reporting in drug trials [22••]. Nonetheless, this labeling requires patients and practitioners to be prepared for questions. The FDA warning does not mean a patient will develop suicidality; however, patients, families, and practitioners must be vigilant for warning signs. A recent survey showed that most academic neurologists counsel their patients about neurobehavioral effects, but fewer than half specifically address suicide [23]. One approach includes a discussion of the risk of future seizures outweighing the risk of developing suicidality, which was measured at 3.4 per 1,000 patients with epilepsy. With new medico-legal implications, neurologists should document a past and current history of mood disorders and suicidality. A family history of psychiatric disorders also is related to developing behavioral side effects on AEDs. The Neurological Disorders Depression Inventory for Epilepsy (NDDI-E) is a simple six-item scale on which a total score ≥15 should raise concerns for more urgent psychiatric assessment.

Bone Metabolism

Chronic use of AEDs may adversely affect bone density, although the relationship has not been fully elucidated. Bone homeostasis is a complicated and an active process requiring parathyroid hormones, adequate serum calcium through intestinal absorption and renal reabsorption, and vitamin D. Putative mechanisms of bone loss include liver induction causing increased vitamin D breakdown, calcitonin deficiency, and effects on calcium absorption. Studies show decreased bone mineral density with the use of benzodiazepines, CBZ, PHT, PB, PRM, VPA, gabapentin (GBP), OXC, and ZNS [24••]. No worsening has been reported with LTG, LVT, or TPM. Bone mineral density testing is now recommended routinely for patients at higher risk for osteoporosis (nonambulatory, elderly, Caucasian patients), but a baseline for any patient may be considered; half the pediatric population in one study had low bone density, particularly patients with cerebral palsy, severe mental retardation, or gait impairment [25].

Pregnancy and Antiepileptic Drugs

A meta-analysis of pregnancy outcomes included 59 studies of more than 65,000 pregnancies in women with epilepsy (WWE) compared with more than 1.8 million without epilepsy [26]. Congenital malformations were significantly more frequent in WWE (7.08%) than in those without epilepsy (2.28%). The highest incidence of malformations was seen with polytherapy that included PHT, PB, or VPA (16.78%); a high rate was also seen with VPA monotherapy (10.73%). Malformations that were most significantly higher in WWE were hernia, ear/neck/face malformations, cleft lip, and spina bifida. Most of the newer AEDs were not individually assessed because of limited outcome data. The prospective Neurodevelopmental Effects of Antiepileptic Drugs study showed an additional risk of in utero VPA exposure—impaired fluency and originality—compared with children exposed to LTG or CBZ [27].

The American Academy of Neurology and the American Epilepsy Society published a series of evidence-based practice parameters regarding management issues in WWE. With respect to pregnancy-related complications, there were limited high-quality studies, yet they suggested a moderately increased (≤1.5 times) risk of cesarean delivery for WWE on AEDs, and a possible increased risk of premature contractions, labor, and delivery in WWE who smoke [28••]. There were insufficient data to comment on pregnancy-induced hypertension, bleeding complications, or spontaneous abortion. The data supported a low rate (8–16%) of seizure recurrence during pregnancy if WWE were seizure-free 9–12 months before pregnancy but could not make statements on change in seizure frequency or potential for status epilepticus. The recommendations regarding major congenital malformations (defined as having surgical, medical, or cosmetic importance) were similar: avoid VPA in monotherapy or polytherapy, specifically in the first trimester, compared with CBZ and possibly PHT and LTG [29]. The group also indicated a likely dose–risk relationship for malformations with VPA and LTG. The risks of some adverse perinatal outcomes, small size for gestational age, and a 1-minute Apgar score less than 7 were elevated. The summary of this practice parameter is that avoidance of VPA exposure and polytherapy in the first trimester of pregnancy appears to be the most important factor. Family planning is important; most women do not know they are pregnant until after 1 month of pregnancy; medication changes at this point may not prevent malformations and might introduce new risks. Despite the knowledge that VPA is clearly worse than the other first-generation medications, there are insufficient data to make comments on most of the newer medications. Encouraging WWE to enlist in the Pregnancy Registry (http://www.aedpregnancyregistry.org) is the most helpful way to continue to collect prospective data.

Although the use of folate is still recommended, there is no high-quality evidence to suggest a dosage or even to support its benefit [30], and it is unlikely a placebo-controlled study will ever be done. Some experts advocate 4 mg/d for those at high risk and planning a pregnancy; however, 1 mg/d for all women of childbearing age is common practice.

Stopping AEDs during pregnancy is considered a large theoretical risk, with the potential for anoxia and trauma from the physical effects of seizures, although no high-quality data exist. The Taiwan national data set found that WWE who had seizures during pregnancy were at significantly but moderately higher risk (odds ratio, 1.34) of delivering newborns small for gestational age [31]. About one third of WWE experience increased seizures during pregnancy, possibly as a result of falling AED drug levels, especially in the second half of pregnancy, because of the increased volume of distribution and clearance of the drug. AED drug levels should be followed closely during pregnancy. The perinatal risk of hemorrhage led to the recommendation that WWE who take drugs that interfere with vitamin K1 metabolism take oral vitamin K1, 20 mg/d, for a minimum of 2 weeks before expected delivery [32].

Breastfeeding

The practice parameter group found no evidence to contraindicate breastfeeding for WWE taking AEDs. There was good evidence that PHT, CBZ, PB, and VPA do not cross into breast milk in clinically significant amounts (likely because of high protein binding); LVT and PRM probably—and GPN, LTG, and TPM may possibly—penetrate in potentially clinically important amounts [30]. Despite the relatively high breast milk levels, LVT is not found in high levels in infants, possibly because of their very fast metabolism of LVT [33].

Characteristics of the Newest Antiepileptic Drugs

Rufinamide

RFN received orphan drug approval by the FDA in January 2009 as add-on therapy for seizures in patients aged 4 years and older with Lennox-Gastaut syndrome, based on a randomized double-blinded clinical trial in 138 highly refractory patients [34]. RFN was significantly better than placebo in median percent reduction in total seizures (32.7% vs 11.7%; P = 0.0015) and median “drop attack” tonic–atonic seizures (42.5% reduction vs 1.4% increase). An extension trial showed maintained response rates during the last 6 months and 12 months of the study (data on file, Eisai Inc., Woodcliff Lake, NJ). Its mechanism of action is theorized to be prolongation of the inactive state of sodium channels.

RFN also was more effective than placebo in a multicenter clinical trial of refractory partial seizures in which 3,200 mg/d was given as adjunct therapy [35]. In addition, nonblinded studies suggest positive effects on myoclonic and absence seizures [36].

RFN has a time to maximum concentration (Tmax) of 4–6 h, reaching a bioavailability of ≥85%. Absorption is improved in the fed state. The half-life elimination is 6–10 h. Hepatic hydrolysis by carboxylesterases to an inactive metabolite is the main mechanism, with less than 4% excreted as the parent compound. Protein binding is clinically insignificant. With the exception of PHT levels increasing 21% in children, the effect size of RFN on other AEDs may be clinically insignificant.

VPA increases steady-state RFN plasma levels in a dose-dependent manner, whereas coadministration of PHT, PRM, and PB decreases RFN plasma concentrations by up to 46% in children, 32% in adolescents, and 25% in adults, and coadministration of CBZ results in decreases of about 20%. Vigabatrin (VGB) coadministration also reduces RFN levels.

There is little to no effect on cytochrome P-450 (CYP) enzymes at standard dosing; however, mild induction of CYP3A4 can reduce ethinyl-estradiol and norethindrone levels, although the clinical relevance is unknown. Female patients of childbearing age should be warned of the potential reduced effectiveness of contraception.

Somnolence and vomiting are common adverse events, as are headache, dizziness, fatigue, nausea, diplopia, and tremor. The observed degree of QT shortening on electrocardiography (ECG) is without known clinical risk, but RFN should be avoided in patients with familial short QT syndrome, and caution must be exercised with other medications that can shorten the QT interval. Developmental toxicity was seen in animal studies; however, the potential for teratogenicity in humans is unknown. RFN is considered pregnancy category C.

For children aged 4 years and older, the recommended initial starting dosage is 10 mg/kg per day in divided doses. The daily dosage may be increased by 10 mg/kg every other day until the target dosage of 45 mg/kg per day or 3,200 mg/d, whichever is less. In adults, the recommended starting dosage is 400–800 mg/d, given twice daily, and may be increased as quickly as 400–800 mg/d every other day until the maximum dosage of 3,200 mg/d is reached, although this rapid titration has been difficult for some patients to manage.

Lacosamide

Lacosamide (LCM) was released in mid-2009, as the FDA finalized its categorization as a schedule V controlled substance. It has a novel mechanism on the sodium channel, functionally enhancing its slow inactivation state, theoretically limiting rapid sequential firing. It also modulates CRMP-2, although it is unknown whether this will have clinical impact. One clinical trial found that adjunct use of LCM significantly reduced partial-onset seizure frequency at 200 mg/d (35.3%) and 400 mg/d (44.9%) versus placebo (25.4%) [37], and another study found reductions at 200 mg/d (26%), 400 mg/d (39%), and 600 mg/d (40%) [38] versus placebo (10%). The highest dosage, 600 mg/d, did not offer significantly greater benefits and caused more adverse events, so this dosage was not included in the product labeling.

LCM is rapidly and completely absorbed, with plasma levels peaking between 1 h and 4 h following oral administration [39]. The elimination half-life is 13 h, primarily through the kidney, and protein binding is less than 15%. There are no known drug–drug interactions with other AEDs.

The starting dosage is 50 mg twice daily and is increased by 50 mg twice daily each week to the target dosage of 200 mg twice daily. Some patients cannot tolerate the target dosage. Dizziness, headache, diplopia, somnolence, fatigue, ataxia, tremor, and nausea were reported as ≥10% in some clinical trials. A dose-related increase in PR interval on ECG may occur (4.2–4.6 ms at 400 mg/d).

The intravenous (IV) formulation is FDA approved for use as a substitution for the oral form when medically necessary. The IV formulation is indicated when oral mediation cannot be taken or may not be well absorbed, but its availability lends itself to intensive care where it has been used off-label to treat refractory status epilepticus [40]. Research is ongoing into its effect on chronic pain and anxiety.

Vigabatrin

In August 2009, VGB was given orphan drug status by the FDA for use in infantile spasms and as adjunctive therapy for adult patients with refractory complex partial seizures who have inadequately responded to several alternative treatments and for whom the potential benefits outweigh the risk of vision loss. VGB is an analogue of γ-aminobutyric acid (GABA) but is not an agonist. Instead, it irreversibly inhibits GABA transaminase, thereby increasing GABA levels. Visual loss may be a result of its efficient transport to the retina. Once detected, the loss is permanent. In 30% or more of adult patients, progressive bilateral concentric visual field constrictions occur, often from the normal of 90° to 65°, but the range is from mild to as severe as 10° of visual fixation. There are some reports of central retinal damage and loss of visual acuity. There is no safe period beyond which retinal damage will not continue to occur.

As part of the FDA approval, the medication must be obtained from a central pharmacy that requires physician registration and patients must undergo quarterly visual assessments by an ophthalmologist, which includes electroretinograms in a percentage of patients, some whom require anesthesia. VGB should be withdrawn as early as possible if substantial clinical benefit is not observed; guidelines are 2–4 weeks for infantile spasms and 3 months in adult patients.

Lamotrigine XR

LTG was approved in the United States in 1995 for use as add-on therapy in partial seizures with or without secondary generalization. There have been further indications to include: adjunct use in partial seizures and Lennox-Gastaut syndrome (chewable tabs) to as young as 2 years old, as monotherapy conversion from valproate or enzyme-inducing medications in partial epilepsy, and for bipolar I disorder. In June 2009, the extended-release formulation (LTG-XR) was approved.

LTG-XR tablets are enteric coated to delay the drug’s absorption until gastric emptying and also use a matrix formulation to limit the rate of dissolution for approximately 12–15 h. This agent was more effective than placebo as an adjunct in once-daily dosing in a double-blinded study [41]. Pharmacokinetic studies in an open-label study switching patients (n = 44) from immediate-release LTG (LTG-IR) to LTG-XR showed a similar area under the curve (AUC) for 0–24 h, except in those taking concomitant inducing agents, in whom the AUC was 21% lower in those receiving the XR formulation [42]. This study showed a decrease in the peak-to-trough fluctuation in those on LTG-XR, and that Tmax increased from 1–1.5 h in the LTG-IR group to between 4 h and 11 h in the LTG-XR group, depending on whether patients were on inducing or inhibiting agents. The XR formulation may help avoid peak dose adverse events.

In studies using LTG-IR, the increased rate of glucuronidation with higher estrogen and other sex hormones leads to significant reductions in LTG plasma levels; therefore, frequent monitoring of serum levels is now advised during pregnancy. Use of oral contraceptives induces its metabolism, with cessation of these agents associated with an 84% increase in LTG in plasma within 1 week of the change [43], suggesting that deinduction of LTG glucuronidation is faster than that of CYP.

Levetiracetam XR

LVT is indicated as adjunctive therapy in the treatment of a) partial-onset seizures with or without secondary generalization in adults and children aged 4 years and older with epilepsy, b) myoclonic seizures in adults and adolescents aged 12 years and older with juvenile myoclonic epilepsy, and c) primary generalized tonic–clonic seizures in adults and adolescents aged 12 years and older with idiopathic generalized epilepsy. LVT-IR was approved in 1999 for adults as adjunctive therapy; it is available in an oral solution and an IV form. LVT-XR was FDA approved in late 2008 based on a multinational comparison study [44]. It uses a matrix system to control the release, increasing Tmax from just under an hour to 4 h, with an AUC comparable with that of LVT-IR [45]. The Tmax can be extended further with intake of a high-fat, high-calorie meal. Behavioral issues such as irritability have not been formally compared with those in LVT-XR, although they were reported in 6% of 77 patients in a placebo-controlled trial [46], a rate lower than the accumulated rate of 13.5% in clinical trials and postmarketing trials of LVT-IR. Continued seizures were a factor in developing negative mood, as some patients who were seizure-free on LVT-IR reported a positive change in mood [47]. In other studies, discontinuation of LVT-IR because of behavioral abnormalities was statistically associated with a faster titration rate, a history of psychiatric disorder, a diagnosis of symptomatic generalized epilepsy, learning disabilities, status epilepticus, and febrile convulsions. Concomitant use of LTG may improve these symptoms. In children, an early study using pyridoxine at 100–200 mg/d also showed some potential to alleviate behavioral abnormalities [48]. LVT showed promise in a placebo-controlled trial for chronic pain in multiple sclerosis [49].

Comparison with First-Generation Medications

Notably, the first-generation medications, including PB, PHT, PRM, CBZ, and VPA, all significantly affect the liver enzymes. Although the newer medications are not clearly more efficacious against seizures, they generally are better tolerated [6, 7]. The potential problems of chronic enzyme induction include loss of bone density and significant elevation of cardiac disease markers [50].

Conclusions

A wide variety of AEDs are now available. Figure 1 and Table 3 demonstrate the thought process involved in choosing among them. For partial seizures, virtually any agent (except ethosuximide) may be considered. For the idiopathic generalized epilepsies, only broad-spectrum agents would be expected to treat all seizure types. The next major factor in choosing a medication is tolerability. The most successful medication in terms of seizure freedom and long-term retention is one that a patient finds tolerable at the very least, and at best, goes unnoticed. Matching the “personality” of the medication to the personality of the patient may lead to better successes. Underlying mood disorders and the wish for weight gain or loss are the most common issues that arise, with tendencies of the medications noted in Table 1. However, using TPM or ZNS in patients with a tendency for chronic headache and PGB or GBP in those with chronic pain or anxiety may be effective for the population with epilepsy. Another key to compliance is the dosing schedule. Once-daily dosing is convenient for patients with busy lives and is becoming common with the addition of XR formulations of older medications.

The shift in attention to tolerability includes the chronic adverse effects of AEDs. The first-generation medications cannot be considered first line because of their associations with osteoporosis and their tendency to increase cholesterol and other assays with negative cardiovascular connotations.

The Pregnancy Registry remains our best hope for determining the true risks of congenital malformations, perinatal complications, and neurodevelopment with use of the various AEDs. For now, it seems clear that VPA should be avoided as monotherapy and most definitely in polytherapy. It is important to define the risks, but it also is worthwhile to stress the positive: WWE have a nearly 95% chance of having a perfectly healthy baby.