Antiepileptic drug treatment in pregnancy: Changes in drug disposition and their clinical implications


Address correspondence to Torbjörn Tomson, Department of Neurology, Karolinska University Hospital, SE 171 76 Stockholm, Sweden. E-mail:


Pregnancy is a state where pharmacokinetic changes are more pronounced and more rapid than during any other period of life. The consequences of such changes can be far reaching, not least in the management of epilepsy where the risks with uncontrolled seizures during pregnancy need to be balanced against potential teratogenic effects of antiepileptic drugs (AEDs). This article aims to review the literature on gestational effects on the pharmacokinetics of older and newer generation AEDs and discuss the implications for the treatment of epilepsy in women during pregnancy. Pregnancy can affect the pharmacokinetics of AEDs at any level from absorption, distribution, metabolism, to elimination. The effect varies depending on the type of AED. The most pronounced decline in serum concentrations is seen for AEDs that are eliminated by glucuronidation (UGT), in particular lamotrigine where the effect may be profound. Serum concentrations of AEDs that are cleared mainly through the kidneys, for example, levetiracetam, can also decline significantly. Some AEDs, such as carbamazepine seem to be affected only marginally by pregnancy. Data on pharmacokinetics during pregnancy are lacking completely for some of the newer generation AEDs: pregabalin, lacosamide, retigabine, and eslicarbazepine acetate. Where data are available, the effects of pregnancy on serum concentrations seem to vary considerably individually and are thus difficult to predict. Although large-scale systematic studies of the clinical relevance of the pharmacokinetic alterations are lacking, prospective and retrospective case series have reported an association between declining serum concentrations and deterioration in seizures control. The usefulness of routine monitoring of AED serum concentrations in pregnancy and of dose adjustments based on falling levels, are discussed in this review. We suggest that monitoring could be important, in particular when women have been titrated to the lowest effective AED dose and serum concentration before pregnancy, and when that individual optimal concentration can be used as reference.

The pharmacologic treatment of epilepsy is seldom more challenging than during pregnancy. Treatment involves drug exposure to at least two individuals: the mother with epilepsy and the unborn fetus(es). The maternal and fetal risks imposed by uncontrolled epileptic seizures need to be balanced against the possible adverse fetal effects of antiepileptic drugs (AEDs). Management is further complicated by the fact that effects on the fetus can be difficult to monitor and may be irreversible. Major convulsive seizures (generalized tonic–clonic seizures) can be harmful to the fetus, in addition to causing medical risks and psychosocial consequences for the mother with epilepsy. Fetal loss has been reported in conjunction with prolonged seizures, such as status epilepticus (Teramo & Hiilesmaa, 1982), and frequent tonic–clonic seizures during pregnancy are associated with poorer cognitive development of the child (Meador et al., 2009a). On the other hand, AEDs can be teratogenic, thereby increasing the risk of congenital malformations as well as of adverse cognitive outcomes (Tomson & Battino, 2012). Recent publications from epilepsy and pregnancy registries have suggested that AEDs differ in their teratogenic potential, but also that these adverse effects of AEDs on the fetus are dose-dependent (Meador et al., 2009a; Tomson et al., 2011; Hernandez-Diaz et al., 2012; Mawhinney et al., 2012). Therefore, the usually recommended treatment strategy is to review and possibly revise treatment well before conception, selecting the most appropriate AED for the individual woman, taking efficacy as well as teratogenic risks into account, and regardless of which drug is chosen to titrate to the lowest effective dose before pregnancy (Harden et al., 2009a; National Institute for Clinical Excellence, 2012; Tomson & Battino, 2012). The objectives of the further management during pregnancy are to maintain control of in particular tonic–clonic seizures while at the same time keeping exposure to potentially teratogenic drugs to a minimum. This demanding task is complicated by the fact that pregnancy can significantly affect the pharmacokinetics of AEDs with potential consequences for seizure control as well as drug exposure to the fetus. So, although the risk of malformations has been assessed in relation to AED doses at the time of conception (Tomson et al., 2011), or during the first trimester (Hernandez-Diaz et al., 2012; Mawhinney et al., 2012), the level of exposure to the fetus can change during the course of pregnancy. In this article, we review the literature on the effects of pregnancy on the pharmacokinetics of different AEDs and discuss the implications for the treatment of epilepsy in women during pregnancy.

Search Strategy

References for this review were identified from the authors' files and from a PubMed search (from 1966 to July 2012) using various combinations of the following terms (by searching as text words) “pregnan*,” “maternal” “mother,” “fetal,” “fetus” with the terms “*kinetics,” “disposition,” “plasma,” “utilization” and the names of individual AEDs included in the review or the terms “antiepileptic*,” “anti-epileptic*,” “anticonvuls*,” “anti-convuls*,” “barbit*.

In addition, the list of references of the retrieved articles was examined for further studies. Moreover, original research articles and previously published review articles were examined.

Only articles published in English were reviewed. The review was restricted to studies in humans.

Mechanisms of Pregnancy-Induced Alterations in Drug Disposition

The physiologic changes that take place during pregnancy can affect any level of the disposition of drugs, from absorption, distribution, metabolism, to excretion (ADME, Fig. 1). It is important to understand the mechanisms by which pregnancy alters the pharmacokinetics of AEDs, as this determines the possible clinical implications as will be discussed in more detail below. However, every pregnancy represents a singular occurrence of genetic variables and environmental factors, and AEDs may interact differently with any of these in the individual woman and also differently from pregnancy to pregnancy (Wlodarczyk et al., 2012).

Figure 1.

Pharmacokinetic changes in drug disposition during pregnancy and basic pharmacokinetic relationships. CL, total clearance; AUC, area under the curve of a drug; F, bioavailability; t1/2, half-life; k, elimination constant and Vd, volume of distribution; Css, serum concentration at steady state, and τ, dose frequency).


Generally, absorption is extensive and bioavailability high for most AEDs. The rate and extent of absorption can, however, vary with the drug formulation. Gastric pH and rate of emptying, and small intestine motility may all decrease during pregnancy, leading to a decrease in the absorption of AEDs (Pennell, 2003), and the pharmacokinetic parameters peak concentration (Cmax) and time to peak concentration (Tmax) (Fig. 1). Malaise and vomiting in early pregnancy are likely to affect the absorption of AEDs taken orally. An impaired drug absorption would result in decreased serum concentrations of AEDs and a possibly decrease inefficacy. However, to the best of our knowledge, there is only one published report of breakthrough seizures due to poor AED absorption during pregnancy. This was a single case of phenytoin malabsorption leading to status epilepticus during pregnancy (Ramsay et al., 1978). Hence, it appears that poor drug absorption is a rare cause of treatment failure during pregnancy.


During the course of the pregnancy, the blood volume progressively increases, with a decrease in the total serum concentrations of drugs in parallel. The increase in body water and in fat stores leads to an alteration of fat/water ratio that may cause an increase in the volume of distribution (Vd). The overall effects of gestation-induced alterations in Vd appear to be variable between individuals, but the net effect on active AED serum concentrations is generally limited.

Serum albumin and α1-acid glycoprotein levels decline as pregnancy progresses, and these are essential transporters for AEDs in the bloodstream. A decrease in serum albumin and protein binding capacity, and displacement of AEDs by endogenous compounds, may result in an alteration in the ratio between the total and the unbound concentration of highly protein-bound drugs, for example, phenytoin, stiripentol, tiagabine, and valproate, AEDs with a protein binding ranging from 90% to 99% (for review, see (Johannessen Landmark et al., 2012)). The total serum concentrations of valproate and phenytoin have been shown to decline in parallel with the falling albumin levels (Yerby et al., 1990, 1992). The net effect of a decrease in protein binding is a fall in the total (bound plus unbound) serum concentration of the drug, whereas the unbound drug concentration may be essentially unchanged. The unbound concentration is the pharmacologically active, and also presumably the concentration determining drug exposure to the fetus. A decrease in protein binding would thus not alter the effect of treatment, but total serum concentrations as routinely measured would be misleading and underestimating the effective concentration and the fetal exposure. Monitoring of free concentrations may be preferable in such situations (Johannessen & Tomson, 2006; Patsalos et al., 2008).


The total clearance of AEDs that are metabolized by the liver depends on the hepatic clearance. The main route is phase I oxidative metabolism through the cytochrome P450 system (CYP). There are different CYP isoenzymes, each of which is a specific gene product with characteristic substrate specificity, where CYP3A4, CYP2C9, CYP2C19 are among the most important for the metabolism of AEDs (Johannessen & Landmark, 2010; Johannessen Landmark & Patsalos, 2010, 2012).

Phase I CYP enzymes as well as phase II enzymes (uridine glucuronyl transferases, UGTs) can be induced during pregnancy. Such enzyme induction is the clinically most important mechanism for declining serum concentrations of AEDs during pregnancy, as it will result in a fall in unbound as well as total serum concentrations. Enzyme induction appears to be most pronounced for AEDs that are metabolized through glucuronidation such as lamotrigine, oxcarbazepine (or rather its active MHD-derivative), and valproate. The extent to which pregnancy affects AED metabolism varies considerably between patients and is therefore difficult to predict (Tomson & Battino, 2007; Patsalos et al., 2008; Johannessen Landmark et al., 2012).


During pregnancy, there is a significant increase in renal blood flow and glomerular filtration rate, up to 50–80% of the prepregnancy rate (Sturgiss et al., 1994; Pennell, 2003). This process starts shortly after conception and continues throughout the second trimester and then declines in the last part of the pregnancy (Sturgiss et al., 1994). Increased renal clearance can affect serum concentrations of AED that are mainly excreted renally, such as gabapentin, levetiracetam, pregabalin, and vigabatrin. It has also been suggested as an important contributing cause for the pronounced increase in clearance of lamotrigine and its N2-glucuronide metabolite in early pregnancy (Reimers et al., 2011; Reimers & Brodtkorb, 2012).

Most documented changes can thus be considered the consequence of enzyme induction and to some extent increased renal clearance. Still, several issues are poorly investigated, for example, some isoenzymes have been investigated for only one drug, the effect of transport proteins during pregnancy remains to be clarified, as well as effects of pregnancy on drugs that are eliminated through non-CYP/UGT or multiple pathways (Anderson, 2005).

Gestation-Induced Alterations in Kinetics of Individual AEDs

Published data on some basic pharmacokinetic properties and on alterations in pharmacokinetics of different AEDs during pregnancy are summarized in Table 1. Because lamotrigine is the most extensively studied AED in this regard, data on its kinetics during pregnancy are reported separately and in more detail in Table 2. A brief account of data on individual AEDs (in alphabetical order) is provided in the following.

Table 1. Pharmacokinetic data and effects of pregnancy on serum concentrations of antiepileptic drugs
ReferencesAntiepileptic drugPregnancies with published pharmacokinetic data (n)EliminationProtein bindingEffect of pregnancy on serum concentrations
Bardy et al. (1982); Battino et al. (1985); Bernus et al. (1995); Lander and Eadie (1991); Tomson et al. (1994a); and Yerby et al. (1992, 1985)Carbamazepine157Almost completely metabolized, the most important pathway being the formation of carbamazepine epoxideAbout 75% Decrease by 0–42%
Öhman et al. (2005)Gabapentin3Unchanged via kidneyNot bound No decline
Johannessen et al. (2005); Pennell et al. (2005); Tomson et al. (2007); and Westin et al. (2008)Levetiracetam45Two thirds unchanged in urine, one-third metabolized by peripheral hydrolysisNot bound Decrease by 40–60%
Christensen et al. (2006); Mazzucchelli et al. (2006); and Petrenaite et al. (2009)Oxcarbazepine27Pre-systemic 10-keto reduction to the two enantiomers of MHD than glucuronidation of the active metabolite MHD via UGTAbout 40% Decrease by 30–38%
Bardy et al. (1987); Battino et al. (1982); Dansky et al. (1982); Lander and Eadie (1991); and Tomson et al. (1994a)Phenytoin307Mainly via hepatic microsomal-mixed function oxidase reaction, with subsequent glucuronidation and excretion of hydroxylated derivative85–90% Decrease by 56–61%
Bardy et al. (1982); Battino et al. (1984a); Lander and Eadie (1991); and Yerby et al. (1992)Phenobarbital56Hepatic oxidation, glucosidation, hydroxylation and subsequent conjugation50–60% Decrease by 50–55%
Battino et al. (1984a); and Rating et al. (1982)Primidone-derived phenobarbital23   Decreased by 70%
Öhman et al. (2009, 2002); and Westin et al. (2009)Topiramate34Mainly unchanged in urine, 20–30% hepatic biotransformationAbout 15% Decrease by 13–40%
Battino et al. (1982); Lander and Eadie (1991); Omtzigt et al. (1992); Philbert et al. (1985); and Yerby et al. (1992)Valproate68Hepatic metabolism, mainly by glucuronidation85–95% Decrease by 0–28%
Kawada et al. (2002); and Oles and Bell (2008);Zonisamide2Mainly hepatic biotransformation, 15–30% unchanged in urine40–60% Decreased by 40–50%
Table 2. Effects of pregnancy on the disposition of lamotrigine
AuthorsNo. of pregnancies on LTG monotherapy/total pregnanciesData presented asPreconceptionTrimester 1Trimester 2Trimester 3Time of sampling after deliveryPostpartum
  1. LTG, lamotrigine; PC, preconception; PP, postpartum; 2-N-GLUC, lamotrigine-2-Nglucuronide.

  2. a

    Post hoc analyses of % change of apparent clearance: first trimester versus second trimester: p < 0.0001; first trimester versus third trimester: p < 0.0001; first trimester versus postpartum: p < 0.03; second trimester versus third trimester: p < 0.04; second trimester versus postpartum: p < 0.0001; third trimester versus postpartum: p < 0.0001.

  3. b

    Pregnancies on LTG monotherapy (or without concomitant use of other drugs known to interact with lamotrigine).

Tran et al. (2002)2/14Mean apparent clearance (L/[kg/day])±SD1.2±0.371.97±0.532.31±0.912.26±0.942–36 weeks1.22±0.54
% change apparent clearance    >65%p < 0.05 vs. PC and PP>65%p < 0.05 vs. PC and vs. PP  
Fotopoulou et al. (2009)9/9Median dose/serum concentration (25–75% percentile)39(39–41)77(68–154)92(76–167)97(74–110)On average 3 weeks35(35–36)
  197%NS236%p < 0.05248%p < 0.05
Pennell et al. (2004)a14/14Apparent clearance mg/(mg/L)52.920.88941133711711002–12 weeks (from figure)6629
% change apparent clearance  19110724914236119415069
Reimers et al. (2011)17/21Mean doses (mg/day) ± SD244±144    342±180At least 4 weeks  
% change mean doses    >34%vs. PC  
Petrenaite et al. (2005)11/11Mean ratio LTG plasma level-to-dose (μmol/l/mg) ± SD63.530.846.718.322.15.421.77.16 weeks (from figure)7010
 −27%NS−65%p < 0.0166%p < 0.01
Öhman et al. (2008b)8/8bMean 2-N-GLUC/LTG ratio      1.020.27Up to 3 months0.40.06
% change 2-N-GLUC/LTG ratio      154% >PPp < 0.01
Öhman et al. (2008a)17/17bMean dose/plasma concentrations      227.174At least 1 month66.517.9
% change mean dose/plasma concentration      250% >PP  
LTG 2-N-glucuronide/LTG concentrations μmol/L         0.3490.141
       147% >PP   
de Haan et al. (2004)12/12 Weeks 1–10Weeks 11–20Weeks 21–30Weeks 31–40 PP
Mean level-to-dose ratios (presented as percentage of baseline ± SD) for consecutive 10-week periods82%14%51%14%40%8%97%15% 48%10%


Carbamazepine is approximately 75% protein bound and is eliminated by hepatic metabolism through CYP1A2, CYP2C8, and CYP3A4 (Johannessen Landmark et al., 2012). Several studies have analyzed the effect of pregnancy on carbamazepine serum concentrations (Lander et al., 1981; Battino et al., 1985; Yerby et al., 1985; Tomson et al., 1994a,b; Bernus et al., 1995), and the results are slightly conflicting. Reported declines in total concentrations from prepregnancy to late pregnancy have ranged from 0% to 42%. The fall in unbound concentrations has been more limited: 0–28% (Lander et al., 1981; Battino et al., 1985; Yerby et al., 1985; Tomson et al., 1994a,b; Bernus et al., 1995). In fact, the largest study of women on monotherapy reported a very limited decline in total concentrations during the second and third trimesters (9–12%), whereas unbound concentrations remained essentially unchanged (Tomson et al., 1994b).


Gabapentin is absorbed form the gastrointestinal tract through saturable active transportation (Gidal et al., 2000), but there are no data on the effects of pregnancy on its absorption. Gabapentin is not bound to serum proteins; it is not metabolized but eliminated by renal excretion (Johannessen Landmark et al., 2012). No studies have followed the course of gabapentin serum concentrations throughout pregnancy. Available data are limited to maternal plasma concentrations at delivery and 3 weeks after delivery in three women who were taking gabapentin in combination with other AEDs (Öhman et al., 2005). These limited data do not suggest lower gabapentin levels in late pregnancy compared to 2 to 3 weeks after. The only potential anticipated effects of pregnancy on its kinetics are those related to changes in glomerular filtration. The reported few cases, however, do not suggest any major effects on the plasma concentration (Öhman et al., 2005) (Table 1).


The serum protein binding of lamotrigine is approximately 55% (Fitton & Goa, 1995). The drug is metabolized mainly by glucuronidation catalyzed by UDP-glucuronosyltransferase (UGT), UGT1A4 and 2B7 (Hussein & Posner, 1997).

Lamotrigine is by far the most extensively studied newer-generation AED in conjunction with pregnancy, and the results are summarized in Table 2. In patients taking lamotrigine, monotherapy serum concentrations usually decline markedly as pregnancy progresses. A decrease in serum concentration can be seen already in the first trimester but is most marked in the mid–third trimester (Öhman et al., 2000; Tran et al., 2002; de Haan et al., 2004; Pennell et al., 2004; Petrenaite et al., 2005; Reimers et al., 2011). On average, lamotrigine serum concentrations decrease by 50–60%, but an even more pronounced decline has been reported in individual patients (Tomson et al., 1997). The effects of pregnancy on lamotrigine disposition, however, vary considerably between individuals and are thus difficult to predict (Petrenaite et al., 2005; Franco et al., 2008; Pennell et al., 2008). The effects are much less pronounced in women taking lamotrigine in combination with valproate (Tomson et al., 2006). Although the dose-to-plasma concentration ratio of lamotrigine increased from baseline to midgestation by 295% in women on lamotrigine monotherapy; the increase was only 60% among women on lamotrigine in combination with valproate.

The decline in lamotrigine levels is probably mainly due to enhanced metabolic elimination by induction of the UGT. This hypothesis is supported by observations of increased ratios in late pregnancy between the lamotrigine-2-Nglucuronide metabolite and lamotrigine in plasma and in urine (Öhman et al., 2008a,b). However, a more detailed analysis of the ratio lamotrigine-2-Nglucuronide/lamotrigine throughout pregnancy suggests that an enhanced renal excretion may contribute in particular to the fall in lamotrigine concentrations in early pregnancy (Reimers et al., 2011). Regardless of whether the mechanism is by increased hepatic or renal clearance, the net effect is a decline on total as well as unbound lamotrigine concentrations.

Serum lamotrigine concentrations return rapidly to prepregnancy values after pregnancy. The process starts within days after delivery and is complete 2 to 3 weeks postpartum (Öhman et al., 2000; Tran et al., 2002; de Haan et al., 2004).


Levetiracetam is not bound to serum proteins and its elimination is primarily renal, although nonoxidative metabolism by hydrolysis accounts for approximately 30% of its elimination. (Patsalos, 2000; Radtke, 2001).

Two reports based on 15 (Tomson et al., 2007) and 21 (Westin et al., 2008) pregnancies observed a significant decline in plasma concentrations during pregnancy by 40% to 60% at the end of pregnancy compared to baseline levels before or after pregnancy. A decrease in serum concentrations, or increase in apparent clearance, is observed already in the first trimester (Pennell et al., 2005). As with lamotrigine, the effects of pregnancy on levetiracetam disposition vary considerably between individuals (Westin et al., 2008). Serum concentrations have been found to increase rapidly after delivery to reach prepregnancy levels within the first week (Westin et al., 2008). Although the mechanisms have not been studied in detail, it is likely that the decline in levetiracetam levels is caused by a combination of an increased renal elimination and enhanced enzymatic hydrolysis.


Oxcarbazepine is a prodrug, which is almost completely metabolized to mono-hydroxycarbazepine (MHD). MHD is responsible for the effects of oxcarbazepine. The protein binding of MHD is about 40%, and it is cleared from the human body mainly by glucuronidation (May et al., 2003). Because its elimination is by the same metabolic route as for lamotrigine, significant effects of pregnancy can be expected. Three studies, comprising altogether 27 pregnancies, confirm that serum concentrations of MHD decline by on average 30–40%, hence slightly less than what is seen for lamotrigine (Christensen et al., 2006; Mazzucchelli et al., 2006; Petrenaite et al., 2009). An increase in MHD serum concentrations was observed within 7–8 days after delivery (Mazzucchelli et al., 2006).


Phenytoin is approximately 90% bound to serum proteins. It is eliminated by oxidative metabolism through CYP2C9 and CYPC19 (Johannessen Landmark et al., 2012). Pregnancy can thus affect the disposition of phenytoin by decrease in protein binding as well as by induced metabolism. The decrease of phenytoin serum concentrations generally starts during the first trimester and becomes more evident in the third when total concentrations are decreased by 55–61%, on average. However, the unbound and pharmacologically active concentrations decline to a lesser extent, by 16–31% (Battino et al., 1982; Dansky et al., 1982; Bardy et al., 1987; Yerby et al., 1992; Tomson et al., 1994a).


Phenobarbital is approximately 50% bound to serum proteins, and is eliminated through metabolism by CYP2C19 and 2E1. Phenobarbital levels have been reported to decline by 50–55% during pregnancy (Battino et al., 1984b; Yerby et al., 1990), whereas the decline has been suggested to be more pronounced, up to 70%, for primidone-derived phenobarbital (Rating et al., 1982; Battino et al., 1984b).


Topiramate is only 15% bound to serum proteins and eliminated mainly unchanged via the kidneys, but a small fraction undergoes hepatic CYP-dependent metabolism (Langtry et al., 1997).

Two studies have assessed the effects of pregnancy on topiramate serum concentrations (Öhman et al., 2009; Westin et al., 2009). These reports suggest an average decline in topiramate levels of 30% to 40% during the third trimester compared with before or after pregnancy, but with a considerable interindividual variation (Table 1). These effects are most likely explained by increased renal excretion.


Valproate is highly protein bound, approximately 90%, and is cleared through hepatic metabolism, mainly by glucuronidation through UGT1A3 and 2B7 in addition to several CYPs. Data on the effects of pregnancy on the disposition of valproate are limited. Total concentrations have been reported to fall in late pregnancy by up to 40% compared with before pregnancy, but this appears to be due mainly to decreased protein binding, since unbound valproate concentrations remain essentially unchanged (Philbert et al., 1985; Omtzigt et al., 1992; Yerby et al., 1992).


The serum protein binding of zonisamide is approximately 60%. Zonisamide is extensively metabolized by acetylation, glucuronide conjugation, and oxidation, and is eliminated also partly by renal excretion (Seino et al., 1995; Perucca & Bialer, 1996).

There are no systematic studies of the kinetics of zonisamide in pregnancy. In a single reported case, zonisamide levels were followed regularly from the fifth gestational week to the end of pregnancy. The patient received 200 mg/day of zonisamide as monotherapy until week 29, when the dose was increased to 300 mg/day, since plasma concentrations decreased from values ranging between 7.5 and 10.1 (from week 5 to week 22) to 4.4 μg/ml in week 27 (Oles & Bell, 2008).

Hence, no firm conclusions can be drawn concerning gestation-related alterations in zonisamide kinetics, but induction of its metabolism might occur.

Other newer AEDs

To the best of our knowledge no published data exist on the pharmacokinetics of felbamate, pregabalin, tiagabine, vigabatrin, lacosamide, retigabine, perampanel, or eslicarbazepine acetate during pregnancy.

Clinical Implications and Conclusions

It is thus undisputable that pregnancy can affect the pharmacokinetics of AEDs and that this effect varies depending on the type of AED. For some, for example, lamotrigine, the effect is profound, whereas other AEDs, for example, carbamazepine, seem to be affected only marginally. For some of the newer generation AEDs, data are scarce or even lacking completely. However, some general assumptions can be made based on available data. A pronounced decline in serum concentrations can be expected for AEDs that are eliminated by glucuronidation (UGT). This might be relevant for eslicarbazepine acetate and possibly also for retigabine (ezogabine) (Bialer et al., 2009). Plasma concentrations of AEDs that are cleared mainly through the kidneys, for example, gabapentin, pregabalin, and lacosamide, can also be expected to decline significantly.

Perampanel is protein bound to a high degree, 96%, and metabolized through CYP3A4 (Bialer et al., 2009). A displacement can thus be expected during pregnancy, leading to declining total concentrations, whereas the active unbound concentrations could remain fairly stable. For AEDs whose pharmacokinetic properties are affected, the extent varies between individuals and is therefore difficult to predict. Polytherapy makes it even more difficult to predict the course of AED concentrations during pregnancy.

The key issue is if these alterations in serum concentrations are of clinical relevance. It appears obvious that a fall in serum AED concentrations in an individual patient is accompanied by a change in efficacy, and in general an increased risk of seizures. The usefulness of serum level monitoring and the need to adjust the dose to maintain stable serum concentrations is nevertheless often questioned (Scottish Intercollegiate Guidelines Network, 2003). Such a position would assume either that pregnancy as such makes patients less likely to have seizures or more responsive to the drug treatment, or that the woman enters pregnancy with a higher AED dose than necessary. There is no evidence whatsoever for the former, and the latter is also unlikely if the general management policy to titrate to the lowest effective AED dose before pregnancy has been followed.

Clinical observations of lamotrigine and oxcarbazepine pregnancies also seem to support that the decline in serum concentrations is associated with an increased risk of seizures. Increase in lamotrigine dose was considered to be needed in 11 of 12 pregnancies in one series (Tran et al., 2002), and dose adjustment in all 14 lamotrigine monotherapy pregnancies in another (Pennell et al., 2004). Seizure aggravation was observed in 9 of 12 cases of pregnancies from a Dutch series (de Haan et al., 2004) and in 5 of 11 pregnancies from Denmark, all of whom were taking lamotrigine monotherapy (Petrenaite et al., 2005). Low concentrations of the active oxcarbazepine metabolite were associated with emergence of seizures in two of four in an Italian series (Mazzucchelli et al., 2006). Another Danish case series found that changes in oxcarbazepine disposition were often associated with worsening in seizure frequency in a study of 13 pregnancies, 10 of which were on monotherapy (Petrenaite et al., 2009), although others have failed to find an association between changes in serum concentrations of the active metabolite of oxcarbazepine and seizure control (Christensen et al., 2006). In the prospective observational EURAP study, the dosage was increased more often in pregnancies with monotherapy with lamotrigine or oxcarbazepine compared with other treatments (The EURAP study group, 2006). The best evidence for an association between declining serum concentrations and deterioration in seizure control comes from a prospective study of 36 pregnancies with lamotrigine monotherapy, of which 39% experienced an increase in seizure frequency during pregnancy (Pennell et al., 2008). The risk was significantly increased when the lamotrigine serum concentrations dropped by >35% from the individual optimal serum concentration before pregnancy.

Although there seems to be an association between declining serum concentrations of AEDs and deterioration in seizure control, many women remain seizure free despite lower drug concentrations. Some therefore advise against drug level monitoring and also against increasing the dose unless the patient has had deterioration in seizure control (Scottish Intercollegiate Guidelines Network, 2003). This position rests on the assumption that risks associated with an increase of the AED dose outweigh seizure-related maternal and fetal risks. We question this assumption. A dose increase should aim primarily at maintaining the woman's individual optimal serum concentration, not increase it. In general, organogenesis is already completed at the stage of pregnancy when such dose adjustments are considered, and a dose adjustment would not increase the risk of birth defects. It can be argued that drug exposure later in pregnancy can be of relevance for cognitive development. However, adverse effects of AEDs on cognitive outcomes have been shown convincingly only for valproate (Meador et al., 2009a), and this is an AED where serum concentrations of the active unbound drug is normally unchanged. Seizures probably do not induce birth defects, and a single seizure is unlikely to cause significant harm to the fetus. However, five or more tonic–clonic seizures during pregnancy have been associated with poorer cognitive development of the child (Adab et al., 2004; Meador et al., 2009b), and a single seizure in a previously seizure free woman can have a major impact on that woman's everyday life. In addition, although rarely, a single tonic–clonic seizure can even be fatal. The Confidential Enquiries into Maternal Deaths in the United Kingdom noted that of 261 maternal deaths, 14 women died of epilepsy, 5.4% of all maternal deaths (Cantwell et al., 2011). The potential consequences of unnecessary seizures are thus significant.

Admittedly, the evidence for an association between pharmacokinetic alterations and deterioration in seizure control during pregnancy is circumstantial. Class I evidence would require a randomized controlled trial. However, based on the available data and on common sense, it is the opinion of the present authors that it would be unethical to randomize pregnant women to having their AED dose guided by clinical judgment only or by clinical judgement supported by serum concentration data in settings where drug level monitoring is available. We find it difficult to see the true equipoise in the two management alternatives.

The evidence-based practice parameter update of the American Academy of Neurology and American Epilepsy Society concluded that monitoring of lamotrigine, carbamazepine, and phenytoin levels during pregnancy should be considered; monitoring of levetiracetam and oxcarbazepine (as MHD) levels may be considered, and although there is insufficient evidence to support or refute changes in phenobarbital, valproate, primidone, or ethosuximide levels, this lack of evidence should not discourage monitoring during pregnancy (Harden et al., 2009b).

We agree that the usefulness of drug level monitoring in pregnancy depends on the AED, but also on the individual characteristics of the patient and her epilepsy. Therefore, monitoring appears less important for drugs such as carbamazepine and valproate, where alterations in active drug levels are usually minor. On the other hand, monitoring is more important for drugs with marked but unpredictable alterations such as lamotrigine, levetiracetam, oxcarbazepine, and possibly topiramate. The value of monitoring during pregnancy largely depends on whether the patient's optimal serum concentration has been determined before pregnancy. The sampling frequency during pregnancy depends on the AED as well as on knowledge of the individual woman's prior sensitivity to dose changes. Sampling monthly could be justified for a woman on lamotrigine, who has been carefully titrated to the lowest effective dose (and serum concentration) prior to pregnancy, and where a lower dose in the past has been associated with major seizures. For such women, we would also propose a proactive approach with dose adjustments before breakthrough seizures, aiming at maintaining the optimal prepregnancy levels. Less frequent sampling could on the other hand be considered in a woman who is on a drug with less pronounced gestational changes in kinetics, who has been seizure free for a long period prior to pregnancy, and for whom it is possible that the used dose could be higher than necessary for seizure control. In such one would also be less inclined to increase the dose based only on a fall in serum concentration. If the AED dose has been increased during pregnancy, one needs to be aware of the rapid reversal of the kinetics after delivery: dose reductions may be necessary during the first few days after delivery to avoid toxicity.

In conclusion, pregnancy can have profound effects on AED pharmacokinetics. Declining serum concentrations is likely to have an impact on the efficacy of the treatment, although for obvious reasons, class I evidence is lacking. Drug level monitoring should nevertheless be considered and decisions on dose adjustments made on an individual basis. If monitoring is considered for highly protein-bound AEDs, such as phenytoin and valproate, free concentrations are preferred. The value of monitoring relies on comparison with the individual optimal serum concentration established before pregnancy rather than with general so-called reference ranges.


TT has received speaker's honoraria or research funding from Bial, Eisai, GSK, UCB, Sanofi-Aventis, and Novartis. CJL declares no conflicts of interest. DB has received speaker's honoraria from Eisai and UCB Pharma. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.