Absorption

Absorption is a complex phenomenon controlled by the physicochemical properties of the drug, the formulation, and the physiologic conditions at the sites of absorption: stomach, small intestine, colon, rectum, buccal and nasal mucosa, and muscle. Most drugs are given orally as solid formulations, which must first disintegrate into small particles that are then dissolved in the gastrointestinal fluid. Only after a drug is dissolved is it available for absorption. Most drugs are absorbed by passive diffusion of the unionized form but some drugs use active transport mechanisms. Gastrointestinal absorption is influenced by age-dependent changes in physiologic variables, such as gastric emptying time, stomach and intestinal pH, absorptive area, and gastrointestinal transit time. Food intake may affect antiepileptic drug pharmacokinetics, especially absorption. Although there is concern that the ketogenic diet may alter dosing requirements, a recent study indicated that there are no significant changes in the concentrations of several antiepileptic drugs (valproate – free or total, topiramate, phenobarbital, lamotrigine, clonazepam) during the ketogenic diet [Dahlin et al., 2006].

Absorption is described in terms of rate and extent. The term absorption half-life describes the rate at which drug reaches the systemic circulation. An indirect measure of absorption rate is the T_max. This measure is useful in comparing the rates of absorption of different brands or formulations of the same drug. A lower (i.e., earlier) T_max usually is an indication of a faster rate of absorption. When all other pharmacokinetic parameters are constant, a faster rate of absorption will result in a lower T_max and higher maximum concentration (C_max).

The extent of absorption describes the amount of a drug that reaches the blood relative to the dose. It is termed oral bioavailability (F) and is the fraction of the dose that reaches the systemic circulation. Absolute bioavailability of a drug is determined by comparing the area under the concentration vs. time curve after an oral (or rectal, intramuscular, and so on) dose with the area under the concentration vs. time curve after intravenous (IV) administration of the same drug. However, because many antiepileptic drugs do not have IV formulations, an absolute biovailability cannot be determined. Rather it can be estimated by comparing available oral formulations, termed relative bioavailability. Relative bioavailability is determined by comparing the area under the concentration vs. time curve for a specific formulation to that of a reference product. For example, the relative bioavailability of a generic brand of phenytoin sodium 100-mg capsule is measured by comparing its area under the concentration vs. time curve with that for brand phenytoin (Dilantin 100-mg capsule).

Because the hepatic portal system drains the entire gastrointestinal system (except for the distal rectum and buccal and nasal mucosa) all drugs swallowed are first delivered to the liver after crossing the gastrointestinal mucosa. Thus, a fraction of these drugs may be metabolized before entering the systemic circulation. The average plasma concentration is determined by the bioavailability and the rate of elimination (clearance) by the following relationship:

Therefore, changes in bioavailability of a drug from alterations in drug formulation, alterations in gastrointestinal physiology, or interactions with inhibitors or inducers can result in clinically significant increases or decreases in drug concentrations in plasma or at the receptor site.

Descriptors of absorption, such as C_max and T_max, and the bioavailability, may vary considerably among different formulations of the same drug. The relative rates of absorption, and thus T_max and C_max, adhere to the following order of most rapid to slowest, and highest peak concentration to lowest: solutions > suspensions > capsules > tablets > extended-release formulations [Garnett and Cloyd, 1993]. Formulation-related differences in absorption characteristics are greatly influenced by the composition of the inactive ingredients (e.g., filler, particle size of drug) and nature of the capsule or tablet coating).

Volume of Distribution

Volume of distribution (V_d) is a measurement of how the drug is distributed throughout the body (tissues and blood), and is expressed in terms of liters or liters per kilogram. Mathematically, it is expressed as the following:

where V_d is the volume of distribution and C_max corresponds to the concentation immediately after IV administration, if this was the route of administration. Because many antiepileptic drugs do not have an IV formulation, the V_d of these is estimated from the concentration at T_max. V_d values may differ markedly among antiepileptic drugs because drugs distribute to tissues with different binding characteristics. For example, drugs that concentrate in fat tissue have relatively low plasma concentrations, but may have a V_d value measured in the hundreds of liters because the C_max is low. Changes in body composition, such as increases or decreases in fat, muscle, or water, may thus alter V_d values.

V_d is very useful for calculating loading doses. The purpose of a loading dose is to fill body compartments rapidly to reach a targeted plasma concentration and thus a targeted concentration at the site of action. The equation for calculating a loading dose is the following:

The change in plasma concentration term denotes the desired increase after administration of the loading dose. For example, if a patient has a plasma concentration of 5 mg/L and the target concentration is 20 mg/L, the desired change in concentration is 15 mg/L. For phenytoin, the V_d is approximately 0.75 L/kg, so an IV dose of 11.25 mg/kg is required. When loading doses are administered intramuscularly, orally, or rectally, resultant maximum plasma concentrations are usually less than after IV loading, owing to slower and incomplete absorption and the elimination of some drug during the absorption phase. Table 59-4 presents loading dose information for six major antiepileptic drugs.

Protein Binding

Several antiepileptic drugs are highly bound to one or more plasma proteins, most commonly albumin or α₁-acid glycoprotein. Drug assays usually measure the total concentration of drug, which is the sum of the bound and unbound concentrations. The free fraction is the ratio of unbound to total plasma drug concentration:

In general, only the unbound drug is able to cross the blood–brain barrier and is in equilibrium with brain and spinal fluid (Figure 59-1; see also Table 59-3). Consequently, clinical response and toxicity correlate better with the unbound drug concentration than with the total drug concentration. For drugs that have limited protein binding (less than 50 percent) or a constant ratio of unbound to total concentration, measurement of total plasma concentrations is a reliable indicator of response. However, if a drug is highly protein-bound (greater than 85 percent), measurements of total concentrations may be an unreliable guide to clinical outcomes. The fraction bound may change because of illness, age, or other drugs that compete for binding sites. In these circumstances, it may be necessary to measure the free concentration to understand the clinical effect better. Low-extraction drugs are those that have only a small portion of the drug metabolized by each pass through the liver. Most of the antiepileptic drugs are low-extraction drugs and alterations in protein binding do not affect the steady-state concentration of the unbound drug. Thus, adjustments in dose generally are not necessary when protein binding is altered because the unbound concentrations remain unchanged. Erroneously increasing the dose when total levels are low, but unbound levels have not changed, will likely result in toxicity.

Fig. 59-1 Relationship of unbound drug in vascular and nonvascular compartments.

Metabolism and Elimination

Antiepileptic drugs are eliminated from the body through metabolism and renal excretion. Clearance (Cl) and elimination half-life (t_½) are the pharmacokinetic terms that quantitatively describe drug elimination. Clearance describes the rate of elimination and defines dosage requirements, whereas t_½ is useful in determining time to steady state and dosing intervals.

Clearance is the volume of blood from which drug is removed (cleared) per unit of time. It is identical conceptually and mathematically to the more familiar physiologic parameter of creatinine clearance. For antiepileptic drugs that are eliminated by the liver, clearance is determined by the free concentration of the drug and the intrinsic ability of hepatic enzymes to metabolize the unbound drug. Although, for most drugs, clearance is proportional to dose, some drugs exhibit nonlinear pharmacokinetics in which clearance changes with concentration (Figure 59-2).

Fig. 59-2 Effect of dose on elimination kinetics.

Curve A plots nonlinear pharmacokinetics (of the Michaelis–Menten type) in which clearance decreases with increasing dose, as with phenytoin. Curve B plots linear pharmacokinetics, in which clearance remains the same with increasing dose, as with phenobarbital, felbamate, vigabatrin, tiagabine, topiramate, and valproic acid (unbound drug). Curve C plots nonlinear pharmacokinetics in which clearance increases with dose, as with valproic acid (total), carbamazepine, and lamotrigine.

The following points are important to remember:

Clearance and absorption determine the amount of drug a patient requires to maintain a targeted steady-state plasma concentration.

Changes in clearance lead to changes in concentration and in the resultant pharmacologic response.

Changes in protein binding can affect total drug concentration without affecting unbound drug concentration.

Changes in total drug concentration without a corresponding change in unbound concentration do not require a dosage adjustment.

Addition or removal of other drugs can induce or inhibit metabolism and alter hepatic clearance.

The t_½ is the time required for the plasma concentration to decline by 50 percent. Half-life is determined by both distribution volume and clearance:

Thus, t_½ is not a direct measure of Cl, but is useful in calculating the time to reach steady state or the time required for drug elimination. After initiation or modification of therapy or after any change in clearance, 5 half-lives must elapse for a new steady-state plasma concentration to be reached (see Figure 59-2). The time required to attain steady state can be explained by using the definition of t_½. The plasma concentration in the body declines 50 percent in 1 t_½; therefore, the percentage of drug remaining in the body, as indicated by the plasma concentration after 2, 3, 4, and 5 half-lives, is 75 percent, 87.5 percent, 93.75 percent, and 96.875 percent, respectively. The actual amount of drug remaining in the body and the additional amount of drug removed in subsequent half-lives are so small (after 5 half-lives) that, for practical purposes, it can be ignored. However, the rate of decline is most rapid in the first two half-lives, and the risk of seizures is greatest during this period.

Biotransformation (metabolism) of medications is accomplished through biochemical reactions. These reactions create compounds that are more polar (more water-soluble) and are eliminated from the body, usually through the kidneys. Phase I reactions (e.g., using the cytochrome P-450 system) involve oxidation and reduction. Phase II reactions (non-cytochrome P-450-mediated) couple the drug or its metabolite with endogenous substrates by conjugation or synthesis. The endogenous substrate may be glucuronic acid, glutathione, glucose, sulfate, or acetate, making the drug more polar and hence water-soluble [Capparelli, 1994]. Antiepileptic drugs may undergo metabolism by phase I or II reactions, or a combination of both. As persons mature, the pathways may change and the metabolite profile will be altered.

In the neonate, phase I reactions, such as hydroxylation and, to a lesser extent, dealkylation, are significantly depressed [Anderson and Lynn, 2009]. In older children, however, phase I activity generally exceeds that in adolescents or adults. As a general rule, the rate of drug metabolism reaches a peak at 0.5–2 years of age and then declines in an exponential manner, reaching adult values in early adolescence. As an example, the clearance of valproic acid, in which metabolism includes both phase I and phase II reactions, is fastest in young children and then declines to reach adult values at the age of approximately 12–15 years. Also, its metabolism shifts from phase II to phase I reactions.

For most drugs, the evaluation of maturation-related (as contrasted with chronologic age-related) enzymatic clearance has not been studied systematically. It is likely that some of the variance in drug clearance noted in children of the same age is due to differences in maturational development. As children grow, the dose may need to be increased to compensate for the increased body (and liver) mass, but later re-adjusted for the decline in phase I and II reactions.

Active metabolites formed after biotransformation of the parent compound can contribute significantly to desired and toxic effects of the parent drug. Metabolites exist in both bound and unbound states. Only the unbound metabolite is available for elimination or biotransformation to an inactive metabolite. As with unbound parent drug, the unbound metabolite concentration in the plasma most closely correlates with the concentration of metabolite at the site of action. The presence of active metabolites complicates interpretation of plasma drug concentrations and clinical response. As with the parent drug, the relationship between unbound and total metabolite concentrations can be disrupted by displacement from serum proteins or changes in metabolism. In fact, parent or metabolite concentrations may be altered independently by drug interaction or physiologic change in metabolism. Table 59-5 displays the cytochrome P-450 isoform families involved in antiepileptic drug metabolism, along with known inducers and inhibitors.

Renal elimination is accomplished by glomerular filtration and/or renal tubular secretion. Glomerular filtration rate and renal tubular function in newborns are less efficient than in adults [Linday, 1994]. After 1 year of age, glomerular filtration rate, corrected for body surface area, is the same as for older children and young adults. Renal tubular function reaches adult values, normalized to body surface areas, within several months. Decreases in clearance can be better correlated with sexual maturity than with chronologic age. Making correlations with physiologic or sexual maturation and developmental age, rather than chronologic age, may be helpful in accounting for the variability observed in clearance of drugs that are renally eliminated [Finkelstein, 1994].

Barbiturates

Phenobarbital, primidone, and methobarbital were the mainstay of the treatment of epilepsy for many years. Primidone itself is active and is metabolized to the active metabolites PEMA and phenobarbital. Because phenobarbital has a half-life measured in days and the other metbolites have relatively short half-lives, phenobarbital is considered to be the primary active agent during primidone use. Methobarbital is also metabolized to phenobarbital. Side effects of this group, including paradoxical hyperactivity in children, have markedly reduced the use of these medications.

Benzodiazepines

As a group, the benzodiazepines are rapidly (T_max of 30–180 minutes) and completely absorbed when given orally. The exception to this is midazolam, which is subject to a high first-pass effect when administered orally. Intramuscular administration of diazepam or lorazepam is not recommended because of slow or erratic absorption, but midazolam is more consistently and rapidly absorbed by this route. Diazepam is also absorbed rectally.

The benzodiazepines rapidly cross the blood–brain barrier. Brain concentrations reach equilibrium with plasma concentrations within seconds after IV administration. But the highly lipid-soluble nature of diazepam and lorazepam causes them to redistribute to other body tissues, resulting in a rapid decrease in brain concentrations and a very high V_d. This redistribution is responsible for the short duration of action after a single dose [Greenblatt et al., 1989; Leppik et al., 1983]. The benzodiazepines are metabolized by the liver and have linear elimination kinetics. Some of the metabolites exhibit antiepileptic activity and also may contribute to toxicity. Because the active metabolites have very long half-lives, repeated administration of these drugs, culminating in large total doses, may lead to long periods of sedation, usually most problematic for diazepam and chlorazepate.

Carbamates

The carbamates are organic compounds derived from carbamic acid, NH₂COOH. Some have been found effective for the treatment of epilepsy. Carbamazepine is the oldest of this class; oxcarbazepine and, more recently, eslicarbazepine acetate, have been added to this category.

Carbamazepine

Carbamazepine is a very water-insoluble, neutral compound with a slow dissolution rate in gastrointestinal fluid. The relative bioavailability is similar for all four formulations (i.e., suspension, chewable tablet, extended-release, and regular tablet), ranging from 75 to 90 percent, although marked intrapatient and interpatient variability has been observed [Faigle and Feldmann, 1975; Maas et al., 1987]. In children on maintenance therapy, the rate of absorption is much faster for the suspension than for tablet formulations, resulting in a significantly earlier T_max and higher C_max (Figure 59-3). This property occasionally produces transient concentration-dependent central nervous system side effects. Administering the same daily dose in smaller amounts and giving it more frequently may correct this problem. Sustained-release preparations may permit twice-a-day dosing, eliminating the need for administration of medications during school hours. This improves medication compliance and decreases central nervous system side effects because peak concentrations will be lower. Studies conducted by the U.S. Food and Drug Administration have demonstrated that carbamazepine tablets are significantly less well absorbed after being exposed to high humidity [Meyer et al., 1992; Wang et al., 1993]. These findings emphasize the importance of storing all antiepileptic drugs at room temperature in a dry location.

Fig. 59-3 Comparison of serum concentration–time curves after administration of carbamazepine (CBZ) tablets and suspension.

Data are from a crossover study in a child with epilepsy on steady-state, maintenance therapy.

The V_d in children for carbamazepine based on oral preparations ranges from 0.8 to 2.0 L/kg. Both carbamazepine and its active metabolite, carbamazepine-10,11-epoxide, are bound to several plasma proteins [Eichelbaum et al., 1975; Rawlins et al., 1975]. Approximately 65–85 percent of carbamazepine and 30–60 percent of carbamazepine-10,11-epoxide are bound to albumin. Both also bind to the reactant protein α₁-acid glycoprotein. Binding to α₁-acid glycoprotein increases after physiologic stress, such as from burn injuries or inflammatory diseases (e.g., juvenile rheumatoid arthritis). Increased binding to α₁-acid glycoprotein can increase total carbamazepine and carbamazepine-10,11-epoxide concentrations without necessarily increasing the unbound drug concentrations. In such situations, patients may have total concentrations above the desired range but have no side effects. Adjusting dosage is not necessary; plasma concentrations return to baseline as α₁-acid glycoprotein declines.

Carbamazepine is eliminated primarily by phase I cytochrome P-450-mediated metabolism (see Table 59-5), forming an active metabolite (10,11-epoxide). Carbamazepine displays dose-dependent elimination pharmacokinetics, in which dosage increases produce a less than proportionate increase in steady-state total concentration (see Figure 59-2). Carbamazepine metabolism undergoes autoinduction, in which clearance increases over time after exposure to the drug. Within 30 days after beginning therapy, carbamazepine clearance may increase by as much as 300 percent. Further increases in maintenance doses may result in further induction [Kudriakova et al., 1992]. In some children, induction causes plasma carbamazepine concentrations to remain unchanged or even decline, despite increasing doses up to 40–50 mg/kg per day [Curatolo et al., 1988]. Carbamazepine and carbamazepine-10,11-epoxide metabolism also can be induced by phenytoin or phenobarbital, resulting in as much as a 100 percent increase in clearance [Kerr and Levy, 1989]. Initial carbamazepine t_½ ranges from 10 to 20 hours, but the t_½ decreases because of autoinduction to 4–12 hours. Consequently, many children require three or four doses a day to maintain targeted plasma concentrations. Sustained-release preparations help overcome this problem.

Felbamate

Approximately 95 percent of a dose of felbamate is absorbed in adults after oral administration. Absorption of the drug is fairly rapid, with T_max values ranging from 1 to 4 hours [Wilensky et al., 1985]. The V_d of felbamate is approximately 0.8 L/kg [Devinsky et al., 1994; Wilensky et al., 1985]. Distribution of felbamate is fairly uniform throughout the entire body. It binds mainly to albumin (22–25 percent), and the amount of binding is dependent on albumin concentration [Perhach and Shumaker, 1995].

About 90 percent of felbamate is eliminated in the urine. Approximately 40–50 percent of the absorbed dose is eliminated as the unconjugated parent compound, and up to 58 percent is found as metabolites and conjugates [Shumaker et al., 1990]. Felbamate is cleared at a faster rate in children. Population pharmacokinetic analysis of data from 139 pediatric patients gives clearance values approximately 40 percent higher in younger children (2–12 years of age) than those in older children and adults (13–65 years).

Gamma-Aminobutyric Acid Analogues

Gabapentin and pregabalin were initally conceived as drugs that may modulate gamma-aminobutyric acid (GABA) receptors. However, this is not their main mechanism of action.

Hydantoins

The hydantoins include ethytoin, mephenytoin, and phenytoin and its prodrug fosphenytoin. Although used in the past, ethytoin and mephenytoin are not widely used now.

Phenytoin

Phenytoin is a poorly water-soluble compound and is often formulated as a sodium salt. It is available as a suspension, chewable tablet, capsule, and a parenteral formulation. The nonlinear, saturation pharmacokinetics of phenytoin can make even small differences in dose clinically significant. The chewable tablet and the suspension contain phenytoin acid, whereas the capsules and the IV formulation contain phenytoin sodium, which is equivalent to only 92 percent of the acid (see Table 59-1). Thus, two 50-mg chewable tablets deliver 8 percent more phenytoin than a 100-mg phenytoin sodium capsule. The suspension is available in a 125-mg/5-mL strength. In the past, the reliability of the suspension has been questioned owing to its tendency to settle. It has been demonstrated that some agitation is sufficient to permit delivery of the specified amount of drug [Sarkar et al., 1989]. Unexpected fluctuations in plasma phenytoin concentrations associated with the use of suspensions likely have been the result of pharmacokinetics rather than formulation problems. Intramuscular injection of the parenteral solution (with a pH of 11) is contraindicated because it results in precipitation of drug that is then slowly and erratically absorbed [Kostenbauder et al., 1975]. It is also painful and causes tissue necrosis [Serrano and Wilder, 1974]. Phenytoin capsules are available in both prompt-release and extended-release formulations. Patient and family education about the differences in phenytoin dosage forms is important to maintain optimal therapy.

Bioavailability, T_max, and C_max are altered when large amounts of phenytoin are given as a single oral loading dose, presumably because of the formation of a multicapsule concretion. In a study involving six healthy adults, absorption was reduced by 10–15 percent and T_max increased from 18 to 32 hours, when phenytoin sodium capsules were given in single dose of 10 mg/kg or more [Jung et al., 1980]. Dividing the total loading dose into 5-mg/kg doses given every 2–3 hours overcomes this problem. The most rapid way to attain therapeutic plasma concentrations after oral loading is to use prompt-release phenytoin sodium capsules or suspension [Goff et al., 1984].

Some reports have suggested that phenytoin is poorly absorbed in newborns because of irregular gastric emptying time, relatively high gastric pH (greater than 4), or decreased absorptive surface [Morrow and Richens, 1989; Morselli, 1983; Painter et al., 1978]. Other evidence indicates that formulation-related differences in absorption, nonlinear pharmacokinetics, and age-dependent changes in elimination account for much of the variability in plasma concentrations [Dodson, 1984].

Phenytoin V_d in children and adolescents is approximately 0.7 ± 0.1 L/kg [Koren et al., 1984]. The relatively small variability in phenytoin V_d permits precise calculation of loading doses. For example, a 15-mg/kg loading dose given intravenously will result in a postinfusion mean concentration of 20 mg/L, with a range of 15–25 mg/L. Phenytoin is also highly protein-bound to albumin.

Phenytoin undergoes extensive hepatic metabolism. A small fraction, 5 percent or less, is excreted unchanged [Battino et al., 1995]. It is metabolized by a saturable cytochrome P-450 enzymatic pathway. Phenytoin follows Michaelis–Menten or saturable elimination kinetics (see Figure 59-2). As the plasma concentration rises, phenytoin clearance decreases, owing to saturation of enzymatic metabolism. Therefore, increases in dose result in disproportionately greater increases in steady-state plasma concentration.

Phenytoin half-life also varies with plasma concentration. As plasma concentration rises, the half-life increases. Half-life ranges from 8 to 12 hours at low concentrations to more than 60 hours at concentrations above 20 mg/L. At concentrations within or above the therapeutic range, 1–4 weeks may be required to reach steady state.

Phenytoin metabolism in neonates is reduced compared with that in older children. Within days to weeks after birth, the rate of metabolism accelerates, reaching a maximum in the 0.5- to 3-year age range. Children in this age group require the highest mg/kg daily doses (Figure 59-4). Thereafter, metabolism and dosage requirements decline gradually, reaching adult values during adolescence. Figure 59-5 illustrates that even minor alterations in dosage produce disproportionately large changes in plasma concentration. Therefore, small adjustments in daily dose are indicated as the plasma concentration approaches or exceeds K_m, which usually falls in the range of 5–10 mg/L. Use of the scored 50-mg chewable tablet, the 30-mg phenytoin sodium capsule, or the suspension facilitates making such adjustments in dose.

Fig. 59-4 Phenytoin dosing requirements in children.

Curves were derived from data of maximum capacity of the enzyme system to metabolize phenytoin (V_max) and the plasma concentration at which the rate of metabolism is 50 percent of maximum capacity for children (K_m) (see Table 59-2).

Fig. 59-5 Schematic of steady-state plasma concentrations of antiepileptic drugs when a particular antiepileptic drug is given alone or in combination with other medications.

(We wish to thank Nina Graves, Pharm.D., for her contribution to this figure.)

Lacosamide

Lacosamide (known as harkoside in earlier studies) was approved for use in the United States during 2009. It is completely and rapidly absorbed with a bioavailability of 100 percent [Hovinga, 2003]. Lacosamide appears to follow linear pharmacokinetics and its C_max is between 1 and 4 hours after an oral dose [Hovinga, 2003]. There does not seem to be a food effect on lacosamide pharmacokinetics [Kellinghaus, 2009]. Protein binding has been shown to be less than 15 percent. Lacosamide is metabolized to an inactive desmethyl metabolite, which is then cleared along with the parent drug via the kidney. Lacosamide is not approved for use in young children at this time; however, there is at least one clinical trial that is investigating the safety and pharmacokinetics in children with partial seizures between the ages of 2 and 17 years. Doses being studied are 8–12 mg/kg/day.

Lamotrigine

Lamotrigine has a bioavailability approaching 100 percent. Its T_max occurs 2–3 hours after administration of a single oral dose. With doses in the range of 15–240 mg, T_max and area under the concentration–time curve are directly proportional to dose [Yuen and Peck, 1987]. The V_d of lamotrigine is approximately 1.5 L/kg in children [Chen et al., 1999]. It is distributed uniformly throughout the body, and approximately 55 percent of lamotrigine is bound to plasma proteins [Garnett, 1997].

Lamotrigine is extensively metabolized to an inactive glucuronide by an inducible uridine diphosphate (UDP) glucuronyltransferase reaction [Lamictal, 2004]. Clearance is greater in younger children [Battino et al., 2001]. Infants (younger than 2 months of age) metabolize lamotrigine more slowly, and metabolic rates increase during the first year of life [Mikati et al., 2002]. Estrogen-containing oral contraceptives may cause dramatic cyclical fluctuations of blood levels (decreasing during the estrogen phase and rebounding in the placebo week), and levels decrease dramatically during pregnancy [Pennell et al., 2008]. There may be an impact of age on lamotrigine concentrations [Reimers et al., 2007]. However, data for weight were not available in a majority of the patients and the effect of age could be due to the fact that older children weigh more than younger children [Reimers et al., 2007].

Levetiracetam

Levetiracetam is extremely water-soluble and has an oral bioavailability of close to 100 percent. Its C_max is within 2 hours of dose administration. When the drug is administered with food, the rate of absorption is decreased, but the extent of absorption is unchanged. Therefore, levetiracetam can be administered without regard to timing of meals. Levetiracetam also exhibits linear pharmacokinetics after single doses ranging from 250 up to 5000 mg [Patsalos, 2000]. The V_d for levetiracetam is 0.5–0.7 L/kg, and protein binding is less than10 percent. Two-thirds of a dose is eliminated unchanged and the rest is eliminated as an inactive metabolite in the urine [Keppra package insert, 2004]. The clearance is 30–40 percent higher in children than in adults [Chhun et al., 2009; Glauser et al., 2007; May et al., 2003; Pellock et al., 2001]. Results are now available from studies in children as young as 2 months old and they confirm that the half-life of levetiracetam in younger children is approximately 5–6 hours, shorter than that seen in adults [Glauser et al., 2007]. An IV formulation and a sustained-release formulation are now on the market. It may be possible to give levetiracetam rapidly. Patients (4–32 years of age) who temporarily lost the ability to take medications by mouth were given an IV loading dose of levetiracetam via rapid infusion (given within 6 minutes). Levetiracetam concentrations ranged from 14 to 189 μg/mL and no significant changes in blood pressure or significant abnormalities in the electrocardiogram were experienced [Wheless et al., 2009]. There were also no local infusion site reactions noted; however, the formulation was diluted 1:1 with normal saline or dextrose 5 percent in water (D5W).

Drug–drug interaction information in children indicates that the concentrations of carbamazepine, valproic acid, topiramate, or lamotrigine are similar between study groups who receive levetiracetam or placebo [Otoul et al., 2007].

Rufinamide

Rufinamide is slowly absorbed due to the fact that it is lipophilic and dissolves slowly in aqueous solutions. The bioavailability is nonlinear and decreases at higher doses, with doses of 1600 mg producing less than proportional concentrations. Steady-state doses of 200, 400, and 800 mg produced morning trough concentrations of 2.81, 5.23, and 11.0 μmol/L, whereas 1600-mg doses resulted in a mean trough concentration of 19.6 μmol/L (4.7 mg/L) [Elger et al., 2010; Limite, 2005; NoAuthors, 2005]. Dissolution is the rate-limiting step in absorption; therefore, changes in gastrointestinal environment, such as with the presence of food, could have an effect on the absorption of rufinamide. It is suggested that rufinamide be administered with food. Rufinamide exposure is increased by 31–34 percent and the C_max by 36–56 percent when the drug is given after high-fat meals [Perucca et al., 2008]. Rufinamide is metabolized to inactive metabolites via the carboxylesterase pathway. Although rufinamide does not appear to be metabolized via the cytochrome P-450, the carboxylesterase enzymes can be induced by inducers of the CYP450 enzymes. Rufinamide also appears to be a weak inhibitor of CYP2E1 and a weak inducer of CYP3A4.

Pharmacokinetic data in children are beginning to appear in the literature. The majority of the data concerning rufinamide is on file with the manufacturer; however, data on these trials have been summarized in Perucca et al. [2008]. A dosing regimen for children has been simulated from several phase II and III clinical studies involving patients with partial or generalized seizures and one study with Lennox–Gastaut syndrome patients. The simulation included 119 children (aged less than 12 years) and 99 adolescents (12–17 years) [Marchand et al., 2010], and was used in supporting material presented to the European Agency for the Evaluation of Medicinal Products. These studies involved patient populations and the presence of inducing and inhibiting polytherapy. Children who weigh less than 30 kg had greater variability in simulated rufinamide exposure measures than those children who were heavier than 30 kg. In addition, the co-administration of valproate resulted in higher rufinamide plasma concentrations. This study proposed a maximum daily dose of 600 mg for children who also receive valproic acid, and a maximum daily dose of 1000 mg for those children who do not also receive valproic acid. Indeed, valproic acid increased rufinamide steady-state plasma concentrations from 54.9 percent in male children and 70.1 percent in female children in a population pharmacokinetic study that is on file at Eisai [Perucca et al., 2008]. The increase in concentrations when valproic acid was a co-medication was approximately 3 times higher in children than in adults. (Adult concentrations were 15 percent higher.) Population studies involving children have also identified increases in rufinamide’s apparent clearance values due to the presence of inducing co-medications. Known antiepileptic drug inducers, such as carbamazepine, phenobarbital, phenytoin, or primidone, as well as vigabatrin, decreased rufinamide steady-state plasma concentrations in children from 23 to 46 percent [Perucca et al., 2008].

Tiagabine

Oral bioavailability of tiagabine is greater than 90 percent, and the bioavailability of tablets and that of the oral solution appear to be similar [Mengel et al., 1991a]. The absorption rate for tiagabine is reduced up to 65 percent when the drug is administered with food, although the extent of absorption is not affected [Mengel et al., 1991b]. Tiagabine has a V_d of 0.74–0.85 L/kg [Gustavson et al., 1997], and protein binding is approximately 96 percent [Gustavson et al., 1997; Ostergaard et al., 1995]. Tiagabine has a half-life of 5–9 hours [Gustavson and Mengel, 1995]. Tiagabine undergoes extensive metabolism by means of the cytochrome P-450 3A isoform, but no active metabolites have been identified. As observed with other drugs, clearances in children are faster than in adults; adjusting both V_d and clearance values according to body size gives values similar to those seen in adults [Gustavson et al., 1997].

Topiramate

Oral tablets of topiramate have a relative bioavailability of approximately 80 percent that of an oral solution. It is rapidly absorbed, with a T_max of approximately 2 hours [Easterling et al., 1988]. Overall absorption is not affected by co-administration with food [Doose et al., 1992]. The protein binding of topiramate is less than 15 percent [Ben-Menachem, 1995a], and its V_d is 0.6–0.8 L/kg [Johannessen, 1997]. Elimination is mainly through the kidney, although a portion of the dose is removed by the liver [Ben-Menachem, 1995a]. Preliminary observations in pediatric patients indicate that the mean oral topiramate clearance is 44–55 percent higher in children, and they may need a higher dose than adults to achieve a similar plasma concentration [Adin et al., 2004; Glauser, 1997].

Valproic Acid

Valproic acid appears to be completely absorbed from all the available products (except for the extended-release formulation), but the rate of absorption varies significantly with dosage form (see Table 59-1). Its T_max is 0.5–1 hour after ingestion for the syrup, 0.5–2 hours after the capsule, 1–6 hours for the enteric-coated tablet, and 3–6 hours for the sprinkle capsule [Cloyd et al., 1983, 1985, 1992]. Absorption after ingestion of the enteric-coated tablet is delayed for several hours; during this lag phase, plasma concentrations may continue to decline. Food further delays, but does not decrease, valproic acid absorption from the enteric-coated tablet [Fischer et al., 1988]. The extended-release tablet has an absolute bioavailability of 90 percent [Dutta et al., 2004]. The extended-release tablets and the enteric-coated tablets are not interchangeable. The extended-release tablets need to be prescribed at a dose that is 8–20 percent higher than the enteric-coated dose to attain similar plasma concentrations [Dutta and Zhang, 2004]. An IV formulation is also available and has been studied in children. The IV formulation should not be given intramuscularly due to toxic muscle necrosis [Gallo et al., 1997].

Valproic acid distribution volumes vary more than those of carbamazepine, phenobarbital, or phenytoin, and range from 0.12 to 0.49 L/kg. When loading doses are calculated, a volume of 0.2 L/kg is recommended. Valproic acid protein binding to albumin is saturable. At concentrations below 40–50 mg/L, the bound fraction averages 90–95 percent. As total concentration rises to the upper boundary of the therapeutic range, the percentage bound falls to 80–85 percent. Saturable protein binding results in a nonlinear relationship between total valproic acid concentration and dose. Unbound valproic acid concentrations, however, remain linear with dose [Cloyd et al., 1993; Riva et al., 1983]. The percentage of unbound valproic acid can be unpredictable and unusually high after rapid infusion in children who are critically ill [Birnbaum et al., 2003].

Valproic acid is extensively metabolized to both active and inactive metabolites. Principal pathways include glucuronidation, β-oxidation in the mitochondria, and oxidation through cytochrome P-450. The monounsaturated metabolites 2-ene-valproic acid and 4-ene-valproic acid possess anticonvulsant activity; the 2-ene-valproic acid achieves pharmacologically relevant concentrations in the brain. 2-Ene-valproic acid accumulates in the brain more slowly than does valproic acid, which may account for the slow onset and cessation of valproic acid anticonvulsant activity. Although the diunsaturated metabolite 2,4-diene-valproic acid, is a potent hepatotoxin, its role in valproic acid hepatotoxicity remains unclear.

The elimination half-life of valproic acid is short. In the first few weeks after birth, half-life ranges from 17 to 40 hours, but rapidly declines from 3 to 20 hours in infants and young children. The full antiepileptic effect of valproic acid is obtained days to weeks after steady state is reached, and this effect continues for some time after the drug is discontinued. The lingering effect may be partly the result of its inhibitory effect on the enzymes responsible for the degradation of GABA, or may be due to the accumulation of 2-ene- valproic acid. This prolonged duration has prompted some investigators to propose single daily dosing [Stefan et al., 1984]; however, with other proposed mechanisms of action, seizure control is related to plasma valproic acid concentrations. Under these circumstances, dosing at least every half-life is indicated to minimize fluctuations in plasma concentrations. When valproic acid is given as monotherapy, the half-life may be long enough to maintain therapeutic concentrations for up to 24 hours in some children. Such children may be able to take valproic acid once a day [Cloyd et al., 1985].

Vigabatrin

Vigabatrin is a 50:50 racemic mixture of R(−) and S(+) enantiomers, with the S(+) enantiomer responsible for the drug’s pharmacologic activities and toxic effects [Haegele et al., 1983; Rey et al., 1990]. Bioavailability is approximately 75 percent in adults when the drug is given orally and is similar for both the tablet and the solution [Hoke et al., 1991]. Administration of vigabatrin with food does not significantly affect vigabatrin pharmacokinetics [Frisk-Holmberg et al., 1989]. Vigabatrin has been studied in children [Rey et al., 1989, 1990], and bioavailability appears to be somewhat lower in children than in adults [Rey et al., 1992]. Vigabatrin does not bind to plasma proteins, and it has an apparent V_d of 0.8 L/kg [Ben-Menachem, 1995b]. Clearance or the dose to concentration ratio of vigabatrin appears to be greater in children than in adults, indicating that a higher dose is needed in children to achieve a particular concentration when compared to adults [Armijo et al., 1997].

Most of a vigabatrin dose appears unchanged in the urine. Studies in infants and children show that the half-life of the inactive R(–) enantiomer, but not the half-life of the active S(+) enantiomer, increases with linear age. Vigabatrin exerts its pharmacologic effect by irreversible inhibition of GABA transaminase, causing increases in synaptic GABA concentrations by inhibiting its breakdown. Because the half-life of GABA transaminase is much longer than that of vigabatrin itself, the pharmacologic effect is determined by the half-life of the enzyme [Ben-Menachem, 1995b; Jung et al., 1977].

Zonisamide

Absolute bioavailability of zonisamide has not been assessed owing to the lack of an IV formulation. The drug reportedly is nearly completely absorbed, as indicated by early studies in which radioactive drug was measured in the urine and bile of rats [Matsumoto et al., 1983]. Human data indicate that, after multiple doses, 65 percent of zonisamide is recovered in urine (62 percent) or feces (3 percent). Food does not affect bioavailability [Zonegran Product Information, 2004]. Zonisamide is only 40 percent bound to plasma proteins, and its apparent V_d is approximately 1.45 L/kg [Zonegran Product Information, 2004]. Zonisamide is highly bound to erythrocytes, resulting in whole-blood concentrations that are higher than plasma concentrations. Binding to erythrocytes causes nonlinearity in plasma concentrations after a dose of greater than 400 mg [Matsumoto et al., 1989]. Zonisamide is eliminated by renal and hepatic routes. The metabolism of zonisamide is mediated in part by cytochrome P-450 3A4 [Zonegran Product Information, 2004]. Zonisamide–carbamazepine interaction was identified in a study including 12 children (5–16 years of age); trough concentrations decreased by 37 percent and carbamazepine-10,11-epoxide by 13 percent when carbamazepine was added to zonisamide monotherapy. The mean trough concentrations observed in 72 children undergoing zonisamide monotherapy were 27.0 ± 9.4 mg/L for a dose of 11.1 ± 2.5 mg/kg/day [Miura, 2004].

Generic Formulations

Absorption may differ among similar formulations (such as tablets) from various manufacturers (e.g., carbamazepine) [Kauko and Tammisto, 1974; Pynnonen et al., 1978; Sillanpaa, 1981]. Significant changes in serum phenytoin concentrations have been observed when the same maintenance doses of medication are taken from different manufacturers [Burkhardt et al., 2004]. Eight adult patients (ages 34 ± 49) had mean total phenytoin concentration of 17.7 ± 5.3 mg/L while on branded drug, which decreased to 12.5 ± 2.7 mg/L with generic and then increased to 17.8 ± 3.9 mg/L after the branded drug was re-introduced. Brand and generic phenytoin do not yield equivalent concentrations in some patients and substitution should not be permitted without prescriber and patient notification. For drugs with nonlinear pharmacokinetics such as phenytoin, such increases can lead to clinical toxicity or seizures [Mikati et al., 1992]. Generic medications can create significant cost savings for patients; however, the risks associated with possible changing drug concentrations need to be considered against the monetary advantages. Substitution of drugs made by differing manufacturers (brand to generic, generic to brand, or generic to generic) may lead to problems, especially when drugs with narrow therapeutic ranges or nonlinear pharmacokinetics are used. Therefore, counseling patients to keep taking medication prepared by the same manufacturer, whenever possible, usually is advisable [Nuwer et al., 1990]. Problems usually arise within 2 weeks of formulation change, so patients and parents should be particularly vigilant during this period. Most of the data available concerning generic substitution of antiepileptic drugs consists of surveys, retrospective database collection, or case studies, and are not prospectively designed to compare brand against generic medications directly.

Dose-Response or Concentration-Response Concept

Substantial evidence exists that both the number of patients experiencing seizure control increases and the degree of seizure control for a given patient improves with higher antiepileptic drug concentrations [Porter, 1989; Schmidt and Haenel, 1984]. The minimum effective concentration defines the concentration at which the desired pharmacologic response is first observed in some patients. The intensity of response increases in direct proportion to drug concentration, until a response is maximized. This result is the maximum pharmacologic effect for that drug. With some drugs, such as phenytoin, further increases in concentration may reduce response (e.g., phenytoin-induced exacerbation of seizures) [Troupin and Ojemann, 1975]. It is not known whether other antiepileptic drugs share this property but there are anecdotal reports of seizures increasing as antiepilpetic drugs are stepped up.

The incidence and severity of side effects also increase with drug concentration. The concentration at which side effects appear is usually, but not always, greater than the concentrations needed to achieve seizure control. The difference between concentrations that produce desired responses and those that produce toxic responses defines the therapeutic range. Optimal ranges, representing population averages, for commonly used antiepileptic drugs are listed in Table 59-6.

Table 59-6 Optimal Target Ranges for Commonly Prescribed Antiepileptic Drugs

PB, phenobarbital.

* Range is not defined, and values are concentrations found in clinical trials.

Tolerance

Tolerance is a phenomenon in which pharmacologic or toxicologic effect diminishes with chronic use, even at the same or increasing plasma concentrations. Different mechanisms for the development of tolerance have been recognized [Froscher and Engels, 1986; Koella, 1986]. One mechanism is the result of biochemical adaptation (e.g., upregulation or downregulation of receptor sites). For example, the effectiveness of benzodiazepine therapy diminishes over time, necessitating even larger doses. Animal studies have demonstrated that chronic exposure to benzodiazepines causes a downregulation of benzodiazepine receptors, which reduces their anticonvulsant effect.

Another mechanism of tolerance is behavioral. The patient becomes progressively insensitive to drug effect without any apparent biochemical change [Siegel, 1986]. For example, many patients initiated on phenobarbital or primidone exhibit mild neurotoxicity even at low plasma concentrations but later, at the same concentration, have no symptoms [Leppik et al., 1984].

Tolerance can produce changes in clinical response that mimic certain drug interactions. For example, loss of seizure control may result from either tolerance, as in the case of clonazepam, or enzyme induction, which decreases concentrations of many antiepileptic drugs. Differentiating between these phenomena is important in considering adjustments in drug therapy.

General Considerations

Antiepileptic drug pharmacokinetics and dosage requirements are influenced by the changing physiology of children as they age. Table 59-7 summarizes the effect of various developmental stages on absorption, protein binding, metabolism, and excretion. The physiologic changes associated with maturation produce marked alterations in antiepileptic drug pharmacokinetics. The variability with which children reach developmental milestones, along with genetic and environmental factors, causes antiepileptic drug pharmacokinetics to vary substantially, even in children of the same age. As a result of the differences between children and adults, children require more frequent and larger doses relative to body size to attain targeted plasma concentrations. Although important age-related, quantitative differences in antiepileptic drug pharmacokinetics are recognized, the pattern of drug-specific pharmacokinetics, such as saturation of absorption, protein binding, metabolism, and enzyme induction, is similar for children and for adults.

A number of physiologic and pathophysiologic processes can alter antiepileptic drug pharmacokinetics. Gastrointestinal disorders, particularly those that increase transit time, can alter the endothelial lining and decrease absorption. Antiepileptic drugs that are slowly absorbed, such as carbamazepine and phenytoin, are most likely to be affected. Physiologic stress, such as that associated with myocardial infarction, burns, traumatic injury, chronic inflammation, and surgery, increases α₁-acid glycoprotein, resulting in greater binding with carbamazepine and carbamazepine-10,11-epoxide [MacKichan, 1992]. Febrile illnesses may produce increases in phenytoin clearance that persist for several weeks, resulting in lowering of plasma concentrations by as much as 50 percent [Leppik et al., 1986].

Renal failure can affect the excretion of antiepileptic drugs for which the kidney is the primary route of elimination. Both clinical response and plasma antiepileptic drug concentrations should be carefully assessed in patients with renal disease. Severe liver disease can alter the metabolism of antiepileptic drugs, resulting in accumulation of the parent drug [Kutt, 1983]. Antiepileptic drug metabolism appears to be well preserved in mild to moderate liver disease. Clinical response may vary, depending on the particular antiepileptic drug.

Plasma antiepileptic drug concentrations can fluctuate over a 24-hour cycle. A day–night cycle, in which both valproic acid and carbamazepine concentrations decrease at night, exists [Lockard and Levy, 1990; Pisani et al., 1990]. The mechanism for this alteration is presumed to be a circadian rhythm in clearance. Absorption of enteric-coated valproic acid tablets may be reduced at night [Cloyd, 1991]. In either case, considering the effect of night-time plasma antiepileptic drug concentrations when designing dosing strategies for patients with poorly controlled nocturnal seizures may be helpful.

Neonates

Gastric emptying time is irregular, absorption area is reduced, and biliary function is underdeveloped for several months or more after birth. These factors can contribute to erratic drug absorption [Morselli, 1983]. In addition, gastric pH is elevated, which may reduce the absorption of certain drugs, such as clonazepam [Meyer and Straughn, 1993]. In contrast with the oral route, rectal absorption is reliable and efficient. Protein binding is reduced because of lower plasma albumin concentrations, whereas greater extracellular water and less adipose tissue can either increase or decrease V_d. Neonatal drug metabolism and renal excretion are reduced. Reactions mediated by hepatic cytochrome P-450 enzymes reach and then exceed adult values within a few weeks after birth. By contrast, renal elimination of drugs and active metabolites may be lower than in adults until the age of 6 months.

Infants and Children

Gastric emptying time and intestinal motility are increased as children grow. Absorptive area, microbial flora, and biliary function begin to approach those in the adult. Also, gastrointestinal blood flow is greater than in adults. These physiologic characteristics contribute to faster absorption of most drugs, resulting in earlier T_max. Plasma albumin levels are lower than adult values, particularly in infancy, resulting in an increased free fraction. Infants and children are capable of synthesizing α₁-acid glycoprotein, which will alter the free fraction of drugs bound to this protein. Metabolism remains elevated for the first 2 years of life; in some cases, drug clearances may be 2–3 times the values seen in adults. Thereafter, metabolism slowly declines, reaching adult values at puberty. After the first 6 months of life, renal excretion is comparable with that in adults.

Absorption

Concomitant therapy may influence absorption of antiepileptic drugs. For example, continuous tube feedings interfere with phenytoin absorption [Maynard et al., 1987]. Calcium- and magnesium-containing antacids appear to complex with phenytoin, decreasing its rate and extent of absorption. They should not be administered concurrently [Cacek, 1986; Fischer et al., 1988]. Gastric pH affects the rate and extent of benzodiazepine absorption; thus, use of antacids and histamine H₂ blockers may decrease the plasma concentration of these drugs [Meyer and Straughn, 1993].

Protein Binding

Valproic acid and phenytoin compete for the same binding sites. Concurrent administration may decrease valproic acid and phenytoin binding. When this occurs, unbound concentrations briefly rise, potentially leading to clinical toxicity. Total phenytoin and valproic acid concentrations decrease; however, unbound concentrations rapidly re-equilibrate to the original concentration. In general, phenytoin and valproic acid dosing does not have to be adjusted in these circumstances. [Scheyer and Cramer, 1990]. Aspirin and other highly protein-bound drugs can have a similar effect, and measuring unbound concentrations in these circumstances can be very helpful in guiding treatement.

Metabolism

Drugs interact in many ways, such as induction and inhibition of liver enzymes. Information on how drugs are metabolized may provide insight into the potential for a drug interaction. For example, if a drug is known to be an inducer or an inhibitor of a particular P-450 isoenzyme, it may be possible to predict if a medication will affect the plasma concentrations of a second drug that is metabolized by that P-450 particular isoenzyme when the drugs are given concomitantly. This approach may be particularly useful in predicting the interaction between the many new antiepileptic drugs being developed and other possible concomitant medications (see Table 59-5).

Intravenous and Intramuscular Administration

Intravenous administration is a good alternative to oral therapy when patients are hospitalized. Phenytoin is very insoluble in water and thus is formulated with ethanol and propylene glycol, and adjusted to a pH of around 11. This formulation should not be given intramuscularly. Fosphenytoin, a phenytoin prodrug, is not formulated in propylene glycol and is maintained at close to physiological pH. It can be given intramuscularly, and the drug can be infused more rapidly than is possible with phenytoin, with fewer adverse effects [Leppik et al., 1989]. Other antiepileptic drugs that are available for intravenous administration include valproic acid, diazepam, lorazepam, lacosamide, levetiracetam, and phenobarbital. Rapid infusion (less than 15 minutes) of some antiepileptic drugs has been shown to be safe, making them more suitable for loading doses, as well as for bridging therapy.

Rectal Administration

When administered as a solution, diazepam is rapidly absorbed rectally, reaching the T_max within 5–20 minutes in children [Lombroso, 1989]. By contrast, rectal administration of lorazepam is relatively slow, with a T_max of 1–2 hours [Graves et al., 1987]. Table 59-9 provides a comprehensive list of antiepileptic drugs available for rectal administration and recommendations for use. Only one rectal formulation is available commercially (Diastat). The bioavailability of carbamazepine oral suspension diluted with an equal amount of water given rectally is similar to that of the oral tablet [Graves and Kriel, 1987]. The bioavailability of topiramate tablets crushed in water and given rectally is similar to that of oral tablets [Conway et al., 2003]. The availability of lamotrigine tablets crushed in water and given rectally is about half that of oral tablets [Birnbaum et al., 2000, 2001].

Some drugs cannot be administered rectally owing to factors such as poor absorption or poor solubility in aqueous solutions. The relative rectal bioavailability of gabapentin, oxcarbazepine, and phenytoin is so low that the current formulations are not considered to be suitable for administration by this route [Clemens et al., 2007; Fuerst et al., 1988; Kriel et al., 1997; Kvan and Johannessen, 1975; Moolenaar et al., 1981]. The dependence of gabapentin on an active transport system, and the much-reduced surface area of the rectum compared with the small intestine, may be responsible for its lack of absorption from the rectum.

Intranasal Administration

Intranasal administration is a potentially attractive alternative route when intravenous administration is impossible or impractical. Nasal access is readily available, even when a patient is having a seizure; requires little, if any, caregiver training; and, under the right circumstances, allows rapid drug absorption. Although the absorptive area of the nasal cavity is approximately the same as that of the rectum, the volume of solution that can be instilled is 0.5 mL or less [Romeo et al., 1998]. Several studies comparing intranasal midazolam with intravenous diazepam have been carried out in children. In summary, both drugs had similar onset of effect and seizure control, although intranasal midazolam has been used for preoperative sedation. Several disadvantages are associated with intranasal midazolam administration:

The rapid elimination half-life may explain the reports of seizure breakthrough within a few hours after intranasal delivery of midazolam [Scheepers et al., 2000]. Absorption of diazepam has been demonstrated after intranasal delivery [Lindhardt et al., 2001]. There are currently no commercial intranasal formulations available.

Other Routes of Administration

Lorazepam is absorbed after sublingual administration and can be useful in treatment of serial and prolonged seizures [Yager and Seshia, 1988]. Delivery of antiepileptic agents directly into the cerebral subarachnoid or intraventricular space is being investigated. Spinal intrathecal administration of baclofen for control of spasticity is currently used for persons with spinal cord disease and cerebral palsy, and can be administered by a programmable continuous pump surgically placed in the abdomen [Albright et al., 1993; Penn et al., 1989].

Clinical Monitoring of Efficacy

Clinical assessment of patients is of paramount importance in the assessment of the effectiveness of therapy. The ideal response would be the total elimination of seizures without any adverse effects of therapy. A satisfactory clinical response for many patients, however, might be substantial reduction of seizures with elimination of or decrease in adverse effects.

One of the most common clinical errors is the premature interpretation of the clinical response. Assessment of seizure control requires an accurate record of seizure frequency and severity before and during therapy. Seizures may be eliminated, reduced, or increased in frequency, and more or less severe during therapy. Patients with epilepsy have varying seizure frequency rates before therapy. The length of time required to assess clinical response varies with the frequency of seizures observed before beginning therapy. Patients with infrequent seizures need a longer period of observation on therapy to permit assessment of clinical response.

Statistical interpretation of the change in seizure frequency based on modified sequential analysis has been proposed [Leppik et al., 1989]. In addition, achievement of steady state with the new drug regimen is necessary to determine if a medication has been given an adequate trial at any specific dose. The time to achieve a steady state varies considerably from drug to drug, depending on the drug’s half-life. In every case, 5 half-lives must elapse before a steady state is reached.

Interpretation of clinical response in patients in whom polypharmacy is necessary is especially difficult. Clinical improvement or deterioration after the addition of a second antiepileptic drug could be due to numerous factors. Such factors include the additive or synergistic effect of the two medications, the effect of the second drug alone, and the effect of the second drug on the metabolism and/or displacement of the initial drug. It is important to make clinical conclusions only after a steady state is reached with the new drug regimen. If improvement is obtained after the addition of the second medication, the clinician should attempt to taper and discontinue the first medication, to determine whether the clinical improvement is an effect of synergistic therapy or that of the second medication itself. Finally, it is necessary to remember that withdrawal of co-medication might have an effect on the disposition of the remaining drug(s), which could influence clinical response (see Figure 59-5).

Clinical Monitoring of Adverse Effects

The clinical state of the patient is more important than laboratory testing for the assessment of adverse effects. It is possible to monitor clinical states continuously, whereas laboratory tests generally are obtained at arbitrary times; therefore, impending organ dysfunction often may be detected earlier by clinical changes, rather than by laboratory tests. The most frequent adverse effects are dose-related and often involve the nervous system. Common manifestations of adverse effects are cognitive changes and drowsiness, impaired attention, incoordination, ataxia, and diplopia. In addition to central nervous system adverse reactions, the liver also may produce signs of toxicity. Early signs of liver failure include anorexia, nausea, vomiting, lethargy, and abdominal pain.

Monitoring of Drug Concentrations

The ability to measure blood levels of antiepileptic drugs has been a major advance in the understanding of drug metabolism and disposition, and has led to more rational drug therapy. The influence of periodic monitoring of blood levels may reduce the number of therapeutic failures by 50 percent [Kutt and Penry, 1974]. As expected, drug concentrations generally are related to the dose administered; however, unexpected results may occur, especially in cases of patient noncompliance or with drugs that follow nonlinear kinetics.

When to Obtain Drug Concentrations

Although enthusiasm for routine monitoring of drug concentrations has moderated, this assistance from the laboratory is helpful in clinical management when used appropriately. Laboratory specimens collected for measurement of drug concentrations should be obtained after the drug is at steady state. The time after dose of when the sample was drawn should be recorded in order to interpret where the concentration would be on a pharmacokinetic profile. Recommendations with regard to specific clinical situations in which determination of drug concentration is useful are listed in Box 59-1 [Glauser and Pippenger, 2000]. It is useful to measure the antiepileptic drug concentration after the patient has reached steady state dosing and optimal clinical outcome has been obtained. This level could be considered the “optimal drug concentration” or the “target level” for the individual patient. Then a level should be obtained when a breakthrough of the seizure pattern occurs (i.e. a single seizure in a person usually well controlled, or a doubling of the usual seizure rate). If the level is lower than the “target value,” one needs to consider noncompliance or alteration in pharmacokinetics. Noncompliance is by far the most common cause of deviation from the “target value.” In these cases, the dose should not be changed, but steps should be taken to increase compliance (pill box, education, and so on). Illnesses affecting absorption or clearance can also be detected by comparing the current value with the “target level.” Also, levels should be measured after changes in co-medication once a new steady state has been reached to determine if any changes in the levels have occurred. If a patient experiences an increase in seizures but the current concentration is at the “target,” it is possible that the seizure threshold has changed or was underestimated, and changes in dose may be needed and a new “target level” should be established.

Box 59-1 Clinical Indications for Obtaining Antiepileptic Drug Concentrations

Establishing baseline effective concentration

Change in dosage

Addition or elimination of another antiepileptic drug

Evaluation of lack of efficacy (noncompliance or fast metabolizer)

Evaluation of toxicity (excessive dose, drug interaction, slow metabolizer, renal or liver failure)

Evaluation of loss of efficacy (compliance, drug interaction, change in formulation, pregnancy)

Estimation of the “room to move”

Laboratory Tests for Idiosyncratic Reactions

In an effort to identify patients in whom serious or potentially life-threatening adverse effects may develop, obtaining complete blood counts and chemistry profiles at routine intervals has become common practice. The cost for this routine monitoring can exceed the costs of medication [Hart and Easton, 1982]. Moreover, it is very debatable whether routine laboratory monitoring can actually identify patients at risk for serious reactions or can identify significant adverse events better than clinical monitoring. Identification of life-threatening reactions is rarely made by routine laboratory screening of asymptomatic children [Camfield et al., 1986]. Fulminant or irreversible hepatic failure during valproic acid therapy is not reliably predicted by laboratory monitoring [Willmore et al., 1991]. Many reports, therefore, indicate that routine laboratory screening of asymptomatic patients is of little value [Camfield et al., 1989; Pellock and Willmore, 1991]. However, complete blood counts and chemistry profiles before, and then after, initiation of antiepileptic drug therapy are reasonable [Harden, 2000]. It is much more effective to caution patients and parents to obtain these tests immediately whenever there are signs of any possible reaction, such as unusual bleeding or bruising, significant loss of appetite, jaundice, and so on. Testing for inborn errors of metabolism may be useful for identifying young children at risk for hepatic dysfunction from valproic acid.

Central Nervous System Adverse Reactions

The central nervous system adverse reaction profile for children is similar to that for adults. Most of the older antiepileptic drugs – phenytoin, phenobarbital, primidone, carbamazepine, and valproic acid – and their active metabolites have similar side-effect profiles, comprising of cognitive impairment, sedation, dizziness, diplopia, ataxia, headaches, and effects on somnolence. The effects of antiepileptic drugs on cognitive function generally are concentration- (and dose-) dependent. Drug therapy is known to contribute to cognitive deficits, especially treatment with multiple drugs [Thompson and Trimble, 1982]. Removal of antiepileptic drugs frequently results in improved cognitive function and motor skills [Duncan et al., 1990]. It is important to recognize that reaction times, disordered attention, and impulsivity, even in untreated children with epilepsy, differ from those in control subjects. The differences in attention and reaction time, however, are small between children with mild as opposed to severe seizure disorders [Mitchell et al., 1992].

Side effects frequently occur in children who are given phenobarbital. Phenobarbital has been associated with a significant depression of cognitive function when it is used for treatment of febrile seizures in children. In a randomized, placebo-controlled, prospective study, children had lower mean intelligence quotient (IQ) scores both during and 6 months after stopping therapy with phenobarbital [Farwell et al., 1990]. Clinical tolerance does develop to some degree; however, an unacceptably high frequency of adverse effects from phenobarbital commonly is observed in children [Farwell et al., 1990; Wolf and Forsythe, 1978]. The adverse effects are reminiscent of complaints present in attention deficit disorders and include overactivity, aggressiveness, inattention, and irritability. The behavioral disorders are observed in 20–50 percent of children receiving phenobarbital for febrile seizures, and result in discontinuation of therapy in 20–30 percent of children [Herranz et al., 1984].

The presence of pre-existing behavior problems, especially hyperactivity, is strongly predictive of adverse reactions during phenobarbital therapy. In a study of children receiving phenobarbital for febrile seizures, 80 percent of those with abnormal behavior before drug therapy reported aggravation of pre-existing hyperactive behavior with phenobarbital, versus only 20 percent for children with normal preseizure behavior [Wolf and Forsythe, 1978]. In some cases, these adverse effects may resolve with continuation or with a lower dose of phenobarbital; however, discontinuation of therapy or alternative medication should be considered.

The newer antiepileptic drugs – gabapentin, felbamate, topiramate, lamotrigine, tiagabine, oxcarbazepine, levetiracetam, vigabatrin, and zonisamide – are thought to have fewer and less severe adverse effects; however, their adverse central nervous system effects are similar to those of the older antiepileptic drugs. Although gabapentin has a good safety profile and seems well tolerated in adults, significant behavioral side effects are observed in some children. In a nonblinded, uncontrolled series, adverse behavioral effects were reported in 47 percent of children given gabapentin for add-on therapy; however, only 12 percent of children were withdrawn from gabapentin because of these problems [Khurana et al., 1996]. In other reports, aggression, hyperactivity, temper tantrums, and increased oppositional behavior developed in 12 children taking the drug [Lee et al., 1996; Tallian et al., 1996; Wolf and Forsythe, 1978]. Most, but not all, of these children had pre-existing cognitive dysfunction, such as mental retardation, autistic features, or behavioral difficulties. The adverse cognitive changes resolved with discontinuation of gabapentin; if administration of the drug was resumed, the adverse cognitive changes tended to recur. In one 6-year-old female, the drug was clinically effective in controlling seizures as monotherapy and was well tolerated for 6 months before the aggressiveness developed [Tallian et al., 1996].

In limited pediatric trials, 36 percent of children taking tiagabine have experienced treatment-emergent adverse effects, most commonly somnolence [Gustavson et al., 1997]. In approximately 15 percent of children, decreased functioning in school, aggression, and other cognitive changes occurred within the first 4 months of initiation of topiramate therapy [Gerber et al., 2000]. A previous history of behavioral changes and concurrent lamotrigine therapy were associated factors. Central nervous system-related adverse effects from topiramate tend to decrease over time [Ritter et al., 2000]. Of greater concern are reports of the development of major depression, schizophrenic-like reactions, and organic psychoses in adults taking vigabatrin [Ferrie et al., 1996; Ring et al., 1993]. A reversible acute psychosis has been reported in a child taking vigabatrin [Canovas Martinez et al., 1995]. Development of hyperkinesia, somnolence, and insomnia in children taking vigabatrin has been reported by a number of investigators [Aicardi et al., 1996; Dalla Bernardina et al., 1995; Dulac et al., 1991]. Generally, these effects have not led to discontinuation of the drug. Meador et al. recently reported that, when compared with other antiepileptic drugs, valproic acid can significantly lower, in a dose-related fashion, the IQ of children of 3 years of age who have been exposed to valproic acid in utero [Meador et al., 2009].

Gastrointestinal Effects

Gingival Hyperplasia

Gingival hyperplasia occurs in 10–20 percent of persons treated with phenytoin and is more frequent in young persons. It is observed more frequently in persons taking higher dosages, often appears 2–3 months after beginning therapy, and reaches its maximum severity in 12–18 months [Babcock, 1965; Little et al., 1975]. Gingival hyperplasia generally resolves within 2–5 months after discontinuation or a reduction in medication [Little et al., 1975]. A vigorous program of good oral hygiene started before and continued during phenytoin therapy can be effective in minimizing gingival enlargement [Modeer and Dahllof, 1987; Pihlstrom et al., 1980]. Drugs other than antiepileptic drugs, including nifedipine and cyclosporine, have been associated with the development of gingival hyperplasia [Butler et al., 1987].

Increased Seizures

Some patients, especially those with generalized or absence seizures, may experience an increase in seizure activity when carbamazepine or phenytoin is added [Horn et al., 1986; Perucca et al., 1998; Snead and Hosey, 1985]. Vigabatrin also has been reported to increase seizure activity and even lead to status epilepticus [de Krom et al., 1995]. Aggravation of seizures occurred in 3 percent of children with generalized seizures, most often in those with nonprogressive myoclonic epilepsy, and was extremely unusual in children with partial epilepsies and infantile spasms [Lortie et al., 1997]. Oxcarbazepine has been reported to increase myoclonic jerks and absence seizures [Gelisse et al., 2004]. Phenytoin and carbamazepine may increase seizures associated with the Lennox–Gastaut syndrome, and gabapentin and pregabalin can increase or induce myoclonic seizures.

Osteomalacia

Osteomalacia (rickets) may occur in patients taking antiepileptic drugs. In a survey of institutionalized patients, 52 of 144 patients with a developmental cognitive disability ranging in age from 3 to 26 years, who were on antiepileptic drug therapy for at least 2 years, had elevations of serum alkaline phosphatase levels greater than 2 standard deviations above normal. Most patients were taking phenytoin or phenobarbital [Hunt et al., 1986]. The cause of osteomalacia during antiepileptic drug therapy appears to be multifactorial [Farhat et al., 2002]. Many antiepileptic drugs presumably enhance hepatic conversion of 25-hydroxyvitamin D to biologically inactive metabolites. Clinically significant rickets seldom develops in ambulatory children on antiepileptic drug therapy, or in institutionalized children who are not taking medication. On the other hand, clinically overt rickets was observed in 10 percent of institutionalized children on phenobarbital and/or phenytoin. In these patients, sunlight was more important for the maintenance of serum 25-hydroxyvitamin D than were supplements of vitamin D [Morijiri and Sato, 1981].

Current evidence indicates that antiepileptic drugs may decrease bone mineral density in adults. In adults, correlation with 25-hydroxyvitamin D levels was poor [Farhat et al., 2002]. In children, no correlation was observed between duration of antiepileptic drug therapy and 25-hydroxyvitamin D levels. Bone mineral density measurements tended to be lower in patients on older antiepileptic drugs than in patients on newer antiepileptic drugs, but were lower in all patients on any antiepileptic drug than in the control population. It was previously thought that older antiepileptic drugs resulted in induction of vitamin D metabolism, resulting in decreased bone mineral density; however, the lack of correlation between vitamin D levels and bone mineral density suggests that other mechanisms are operative.

Some authors recommend routine bone mineral density monitoring for patients on chronic antiepileptic drug therapy [Farhat et al., 2002]. Consensus is lacking on treatment of patients who have low bone mineral density secondary to antiepileptic drug therapy. Other factors may contribute to decreased bone mineral density, such as lack of exercise and sunlight.

Tremor and Movement Disorders

A benign essential type of tremor has been observed with use of valproic acid, beginning within 1 month of initiating therapy. It is dose-dependent, occurring in some patients with blood valproic acid levels greater than 40 mg/mL. Reduction of tremor may occur with reduction of dose [Karas et al., 1982]. Tremor also has been seen with use of lamotrigine and gabapentin [Schmidt and Kramer, 1994]. Oculogyric crises and choreoathetotic movements have developed in several adults while taking gabapentin [Buetefisch et al., 1996; Reeves et al., 1996].

Other Effects

Hair changes have been associated with antiepileptic drug treatment. Hair changes (thinning or wavy hair) are seen in 12 percent of children during valproic acid therapy [Clark et al., 1980]. Additional dose-related adverse effects of phenytoin include hirsutism, coarse facial features, and acne, especially in children and teenagers. Pancreatitis has been associated with valproic acid therapy [Coulter and Allen, 1980; Wyllie et al., 1984]. Visual field changes, which may be irreversible, may occur with vigabatrin treatment [Krauss et al., 1998]. Topiramate has been reported to cause acute myopia and secondary angle closure glaucoma. Patients present with decreased visual acuity and ocular pain. In a majority of cases, immediate discontinuation of topiramate results in a resolution of symptoms [Sankar et al., 2001; Thambi et al., 2002]. Hyponatremia occasionally is encountered in patients taking carbamazepine and oxcarbazepine [Borusiak et al., 1998; Holtmann et al., 2002; Koivikko and Valikangas, 1983]. Hyperthermia and oligohydrosis have been reported in several children taking topiramate and zonisamide. Clinical signs included fever, decreased sweating, and exercise intolerance, which resolved in a majority of cases when topiramate or zonisamide was discontinued [Arcas et al., 2001; Knudsen et al., 2003]. Topiramate and zonisamide may cause nephrolithiasis. The risk is increased with topiramate or zonisamide treatment and the ketogenic diet [Kossoff et al., 2002].

Anticonvulsant Hypersensitivity Syndrome in Children

Prevention

Routine laboratory monitoring (i.e., with complete blood counts or liver function tests) will not prospectively identify which patients are at high risk for the development of an acute hypersensitivity reaction. Nevertheless, certain factors place some patients at high risk. Clinical profiles that identify patients at high risk for hypersensitivity reactions differ for specific antiepileptic drugs (valproic acid, felbamate, and lamotrigine) (Box 59-2). When one of these drugs is being considered for a patient at high risk, one should proceed with caution. Alternative therapy, if possible, should be considered. Additional laboratory tests to screen for inborn errors of metabolism may be indicated when use of valproic acid is under consideration in children younger than 2 years of age. Families need to be advised that the child is in a high-risk category and counseled on the early recognition of the syndrome.

Box 59-2 Clinical Profiles for Patients at High Risk for Idiosyncratic Reactions to Valproic Acid, Lamotrigine, and Felbamate

Valproic Acid: Hepatotoxicity

Children younger than 2 years of age

Multiple concomitant antiepileptic drugs

Underlying metabolic disease

Developmental delay

Felbamate: Aplastic Anemia

Caucasians

Adults more than children

Females more than males

Previous cytopenia

History of antiepileptic drug allergy or toxicity

History of immune disorder

On felbamate less than 1 year

Lamotrigine: Stevens–Johnson Syndrome and Toxic Epidermal Necrolysis

Children more than adults

Concurrent valproic acid use

High starting dose

Rapid titration

On lamotrigine less than 1 year

(From Glauser TA. Idiosyncratic reactions: New methods of identifying high-risk patients. Epilepsia 2000;41 [Suppl 8]:S16.)

Although the identification of a high-risk clinical profile is the most inexpensive and practical way available to reduce the risk of hypersensitivity syndrome, various biomarkers also have been identified [Glauser, 2000]. These biomarkers may indicate that some patients have genetic susceptibility to the accumulation of toxic metabolites due to deficient detoxification pathways. The biomarkers that have been investigated include deficient epoxide hydrolase activity and deficiencies in free radical-scavenging enzyme activity. Studies in patients of Chinese ancestry indicate a strong association between the risk of developing toxic epidermal necrolysis and Stevens-Johnson Syndrome and the presence of HLA-B* 1502. Clinicians may want to avoid use of carbamazepine in this population.

Managing Adverse Effects

The clinician and the patient are faced with adverse effects approximately 30 percent of the time when antiepileptic drugs are used, especially during initiation of therapy. In many instances, this rate is acceptable, especially when the patient and the physician mutually agree on which problems are most important to avoid. When no urgency exists, many adverse effects can be avoided during initiation of therapy by gradually increasing doses. In general, management of adverse effects depends on the problem. Most adverse effects are concentration-dependent, and are less prominent at lower doses and blood concentrations. Seizure control may or may not be maintained at the lower dose; if not, the clinician and the patient need to consider whether a compromise is acceptable with regard to either increased seizures or increased adverse effects, or if alternative medication should be considered.

At times, toxicity is encountered only with high drug concentrations and is seen only at peak times; for these patients, it may be possible to avoid toxicity by more frequent and lower doses or by the use of extended-release formulations. Gastric irritation generally is dose-dependent and may be managed in several ways. The amount of drug can be reduced; the effective dose to the gastric mucosa can be lowered by use of enteric-coated formulations; or the patient can be protected by use of H₂ receptor antagonists.

Idiosyncratic reactions pose another problem. Mild, asymptomatic elevations of liver enzymes do not mandate discontinuation of therapy if not in excess of 2–3 times normal values. On the other hand, Stevens–Johnson syndrome and toxic epidermal necrolysis, clinically symptomatic hepatotoxicity, pancreatitis, and most rashes are indications for immediate cessation of the responsible medication. If the rash was mild, however, and especially if the patient had an otherwise favorable clinical response to the drug, a cautious rechallenge might be considered. Successful resumption of therapy has been reported in a number of patients, especially with valproic acid after that drug was discontinued, when a more gradual dose escalation was used [Schlumberger et al., 1994; Tavernor et al., 1995].

Benefits of Drug Discontinuation

The potential benefits of drug discontinuation are numerous. Most obvious is the immediate reduction in cost to the family and third-party payer with elimination of drug therapy and usually the reduction in associated expenses incurred by laboratory tests and physician and clinic charges. In addition, in many patients, the adverse, sometimes previously unrecognized effects of antiepileptic drug therapy are reversed. Patients who had been on phenobarbital, phenytoin, valproic acid, and topiramate had subtle improvement in various psychometric tasks, such as memory, vigilance attention, and visual motor performance [Gallassi et al., 1992; Lee et al., 2003], and in psychomotor speed [Aldenkamp et al., 1993], when therapy was discontinued.

Risks of Drug Discontinuation

The obvious risk of stopping antiepileptic drug therapy is the recurrence of seizures. In most cases, seizure control can be regained with resumption of the previous drug therapy. Response to therapy after a relapse may be rapid and satisfactory [Arts et al., 1988]. Chadwick and colleagues observed that approximately 90 percent of patients who experience seizures after antiepileptic drug discontinuation have a 2-year remission after therapy is resumed [Chadwick et al., 1996]. Thus, control was usually but not always regained despite increasing doses of a previously successful drug regimen.

A guideline has been published to assist the physician and the family considering discontinuation of antiepileptic drug therapy. The conclusions were based on a MEDLINE search [Camfield and Camfield, 2008].

Children who were seizure-free for 2 or more years while on antiepileptic drugs, who had a single type of partial or generalized seizure, who had a normal IQ and normal findings on neurologic examination, and a normalized electroencephalogram (EEG) on therapy had a 69 percent chance of discontinuing antiepileptic drug therapy without seizure recurrence. A similar conclusion was reached by another meta-analysis, in which 75 percent of “relatively unselected” patients were seizure-free 1 year after discontinuation, and 71 percent 2 years after discontinuation of antiepileptic drug therapy [Berg and Shinnar, 1994]. Children with remote symptomatic seizures and those with seizure onset in adolescence (rather than earlier) were at greatest risk of recurrence.

Although relapse was observed in every study, in one prospective series, relapse rates were slightly higher in children who had epileptiform activity on the last EEG before antiepileptic drug discontinuation [Andersson et al., 1997]. The authors concluded that the persistence of 3-Hz spike-wave activity during therapy also was an unfavorable prognostic sign. Children more frequently had recurrences of seizures when antiepileptic drugs were stopped after 1 year versus 3 years of therapy. In another randomized, prospective study of 6 versus 12 months of antiepileptic drug therapy in children who became seizure-free on antiepileptic drug therapy, no difference between relapse rates was observed [Peters et al., 1998]. That report again presented the predictive variables for relapse, which were partial epilepsy, a remote symptomatic etiology, an older age at onset (12 years and older), and an epileptiform EEG during therapy.

Families should discuss withdrawing antiepileptic drugs with their neurologist once children have become free of seizures for 1 year or longer while on antiepileptic drug therapy. Children whose EEGs have become normal have a somewhat better chance of successful withdrawal of therapy. However, there are critical life transitions when the risk of discontinuing therapy outweighs continuation. These are when children have been seiure-free and are eligible to begin driving, and when they are leaving home and transitioning to college or independent living. Finally, for some patients, withdrawal of antiepileptic drugs is rarely successful – for example, those with abnormalities on neurologic examination or with certain seizure syndromes such as juvenile myoclonic epilepsy, Lennox–Gastaut syndrome, or infantile spasms [Peters et al., 1998; Shinnar et al., 1994].