Prolonged Vigabatrin Treatment Modifies DevelopmentalChanges of GABAA-Receptor Binding in Young Children

with Epilepsy

*Csaba Juha´sz, *†Otto Muzik, *†Diane C. Chugani, *Chengang Shen, §James Janisse, and*†‡Harry T. Chugani

Departments of *Pediatrics, †Radiology, and ‡Neurology, Children’s Hospital of Michigan, The Detroit Medical Center, and the§Center for Health Effectiveness Research, Wayne State University School of Medicine, Detroit, Michigan, U.S.A.

Summary: Purpose:To determine whether prolonged treat-ment with vigabatrin (VGB), an antiepileptic drug (AED) thatacts by elevating braing-aminobutyric acid (GABA) levels,interferes with age-related changes of in vivo GABAA-receptorbinding in children with epilepsy.

Methods:Using [11C]flumazenil (FMZ)–positron emissiontomography (PET) imaging, 15 children (aged 1–8 years) withmedically intractable epilepsy were studied. Seven of thesechildren were treated with VGB (1,000–2,500 mg/day) for$3months before the FMZ-PET study. The remaining eight pa-tients were medicated with other drugs that are known not toact directly on the GABAergic system. Absolute quantificationof PET data was performed by using the volume of distribution(VD) of FMZ in brain tissue representing FMZ ligand binding.

Results:After controlling for age, hemispheric FMZ VDvalues were significantly lower in children treated with VGB ascompared with the non-VGB group (p4 0.012). Regional

FMZ VD values of the VGB-treated patients were significantlylower in all cortical regions and the cerebellum, whereas thedifference was not significant in the thalamus and basal gan-glia. No significant drug effect or drug-by-region interactioncould be determined when the patients were separated accord-ing to treatment with carbamazepine (p4 0.97) or valproate(p 4 0.55).

Conclusions:VGB induces a decrease in GABAA-receptorbinding in the cortex and cerebellum of the developing epilep-tic brain. A similar effect of other drugs and substances ofabuse targeting the GABAergic system may be hypothesized.Because of the important role of the GABAergic system indevelopmental plasticity, the reversibility and functional con-sequences of this age-specific drug effect should be furtherstudied.Key Words: Vigabatrin—PET—GABAA receptors—Epilepsy—Development.

g-Aminobutyric acid (GABA) is the major inhibitoryneurotransmitter in the human brain (1). GABA mediatessynaptic inhibition, and plays a key role in regulatingcentral nervous system excitability (2) and suscepti-bility to seizures (3). The action of GABA is mediatedin part by the GABAA-receptor complex, the site ofaction of numerous therapeutic pharmacologic agentsand drugs of abuse. Based on the knowledge of the roleof GABAergic mechanisms in seizure disorders, new an-ticonvulsants (AEDs) have been designed specifically totarget the GABAergic system. Among these, vigabatrin(4-amino-hex-5-enoic acid; VGB) was the first AEDrationally designed to inhibit irreversibly GABA-trans-aminase, the degradative enzyme of GABA (4). VGB ispresumed to act by increasing GABA concentrations in

the brain (5,6), and it is a drug of choice in infants andyoung children with infantile spasms (7,8) and is occa-sionally used for complex partial seizures with or with-out secondary generalization (9) outside the UnitedStates. Recent human studies, in which GABA concen-tration was measured in vivo with magnetic resonancespectroscopy, showed a two- to three-fold increase ofbrain GABA levels of the occipital cortex after therapeu-tic doses of VGB in both healthy subjects (10) and adultpatients with partial epilepsy (6,11,12). A similar effectof VGB was reported in children, although the magni-tude of GABA increase was apparently smaller (between6 and 44%) (13).

Theoretically, drugs that increase GABA concentra-tion might be expected to cause downregulation ofGABAA receptors with long-term treatment. Studies ofbenzodiazepine (BZD) binding to the receptor using[11C]flumazenil (FMZ) positron emission tomography(PET) or [123I]iomazenil single-photon emission tomog-

Revision accepted June 14, 2001.Address correspondence and reprint requests to Dr. H. T. Chugani at

Children’s Hospital of Michigan, PET Center, 3901 Beaubien Blvd.,Detroit, MI 48201, U.S.A. E-mail: hchugani@pet.wayne.edu

Epilepsia,42(10):1320–1326, 2001Blackwell Science, Inc.© International League Against Epilepsy

1320

raphy, however, have failed to show significant changesin binding after VGB treatment of adults (10,12). Al-though no similar data are available for children, it isbeing recognized increasingly that pharmacologic agentsbehave differently in children as compared with adults.Children require larger doses than adults (on a mg/kgbasis) of agents acting at the GABAA receptor. For ex-ample, the optimal effective dose of VGB was reportedto be higher in children ($80 mg/kg/day) than in adults(the maximal recommended dose,∼40 mg/kg for a 70-kgadult) (14,15). These differences might be related to bio-distribution and other pharmacokinetic factors, but mayalso, at least in part, be attributed to developmentalchanges in subunit composition of the GABAA-receptorcomplex (16–18). We have recently shown developmen-tal differences in whole-brain and regional values ofFMZ binding, with higher values in children comparedwith adults (19). None of the subjects studied was medi-cated with drugs that are known to increase brain GABAlevels.

Based on these findings, we hypothesized that AEDsspecifically targeting the GABAergic system, such asVGB, may influence developmental changes of GABAA-receptor binding. The purpose of this study was to de-termine whether prolonged treatment with VGB altersdevelopmental changes of GABAA-receptor binding inyoung children with epilepsy.

METHODS

SubjectsFifteen children with medically intractable epilepsy

underwent FMZ-PET scanning as part of their presurgi-

cal evaluation. Seven of these children (five boys andtwo girls; mean age, 3.0 ± 2.0 years; range, 1.2–6.7years) were treated with VGB (1,000–2,500 mg/day) for$3 months before the FMZ-PET study (VGB group,Table 1). In addition, all these patients used at least oneother AED (Table 1). The remaining eight patients (sixboys and two girls; mean age, 5.7 ± 4.3 years; range,2.0–7.8 years) were receiving mono- or polytherapy withcarbamazepine (CBZ; n4 4), valproate (VPA; n4 3),lamotrigine (LTG; n4 3), gabapentin (GBP; n4 3),phenytoin (PHT; n4 1), felbamate (FBM; n4 1), andethosuximide (ESM; n4 1) at the time of PET scanning,but none of them was mediated with VGB (non-VGBgroup). None of the patients used BZDs or phenobarbital(PB) for $3 weeks before the FMZ-PET scan.

All epilepsy patients (including those with infantilespasms) were diagnosed with a unilateral seizure focusbased on seizure semiology, scalp ictal, and/or intracra-nial EEG as well as 2-deoxy-2[18F]fluoro-D-glucose PETscan. The region of the epileptic focus showed unilateral,localized decreases of FMZ binding by visual assessmentin all patients. No cortical or subcortical lesions on mag-netic resonance imaging (MRI) scans were observed inany of the subjects.

All FMZ-PET studies were performed in accordancewith the policies of the Wayne State Washington Insti-tutional Review Board, and written informed consentwas obtained.

PET scanning protocolFMZ was produced via the captive solvent methyl-

ation technique (20). FMZ-PET studies were performedusing the CTI/Siemens EXACT/HR whole-body posi-

TABLE 1. Clinical data of children who took vigabatrin (VGB) and those who used only other antiepileptic drugs

Sex/age/duration ofepilepsy (yr)

Seizuretype

Epilepticfocus

VGB dose (mg)/duration ofVGB treatment before PET

Otherdrugs

Hemispheric FMZVD (ml/g)

Patients taking vigabatrin1. M/1.2/1.0 IS L T-P 1,000/9 mo TPM 6.12. M/1.3/0.9 IS L O 1,500/6 mo CBZ 4.43. F/1.8/1.3 CP L F 1,500/6 mo CBZ, PHT 4.74. M/2.3/1.1 IS R T 1,000/4 mo TPM 4.05. F/3.3/1.6 CP L T-P-O 1,000/5 mo VPA, PHT 5.46. M/4.3/4.0 IS→GTC R F-T 2,500/3 mo VPA 4.97. M/6.7/1.7 CP L F 1,000/8 mo PHT 3.9

Patients without vigabatrin1. M/2.0/1.6 CP, GTC R F-C — CBZ, LTG 5.72. M/3.9/3.0 CP, GTC R C — VPA, FBM 5.73. M/4.6/1.7 CP L F-P — CBZ, VPA, GBP 4.94. M/6.0/3.0 SP, CP L F — LTG 5.25. M/6.7/3.0 CP, GTC R T-C — LTG, ESM 5.66. F/7.1/4.3 SP, CP, GTC R F — VPA, PHT 5.37. M/7.7/3.5 CP, GTC R F,T — CBZ, GBP 5.48. F/7.8/4.7 SP, CP L F — CBZ, GBP 4.3

The epileptic focus was determined based on seizure semiology, EEG, and glucose PET findings. [11C]Flumazenil volumes of distribution (FMZVDs) were measured contralateral to the epileptic focus.

IS, infantile spasms; CP, complex partial seizures; GTC, secondarily generalized tonic–clonic seizures; T, temporal; P, parietal; F, frontal; O,occipital; TPM, topiramate; CBZ, carbamazepine; PHT, phenytoin; VPA, valproate; LTG, lamotrigine; FBM, felbamate; GBP, gabapentin; ESM,ethosuximide; PET, positron emission tomography.

VIGABATRIN EFFECT ON GABAA RECEPTORS 1321

Epilepsia, Vol. 42, No. 10, 2001

tron tomograph. This scanner has a 15-cm field of viewand generates 47 image planes with a slice thickness of3.125 mm. The reconstructed image in-plane resolutionobtained was 5.5 ± 0.35 mm at full-width-at-half-maximum and 6.0 ± 0.49 mm in the axial direction.Attenuation correction was performed on all images us-ing data from a 15-min transmission scan of the head.Scalp EEG electrodes were placed, and the EEG wasmonitored during the entire scan in all epilepsy subjects.A venous line was established for tracer injection, and anarterial line was placed in a precutaneous radial artery forcollection of 23 timed arterial blood samples (0.5 ml/sample; sequence: 12 × 10 s, 3 × 1 min, 2 × 2.5 min,2 × 5 min, 4 × 10min). Radioactivity in plasma wasmeasured by using a Hewlett Packard Cobra II gammacounter. Concentrations of FMZ and metabolites weredetermined with the method of Barre et al. (1991) (21).The tracer FMZ (0.1 mCi/kg) was injected as a slowbolus over a 30-s period. Beginning at the time of FMZinjection, a 60-min dynamic PET scan of the brain wasperformed (sequence: 4 × 30 s, 3 × 1min, 2 × 2.5 min,2 × 5 min, 4 × 10min). Chloral hydrate (50–100 mg/kgby mouth) or pentobarbital (PBT; 3 mg/kg, i.v.) was usedfor sedation in 12 of 15 children (in six who took VGBand in six who did not take VGB).

Data AnalysisDetailed analysis of the kinetic behavior of FMZ in

brain tissue (22) indicated that the tissue tracer pools offree, nonspecifically bound, and specfically bound tracerare in rapid exchange, allowing simplification to a singletissue compartment representing the combined tracerpools. Therefore, the mathematical description of re-gional brain activity CT is given by a two-compartmentalmodel

CT(t) 4 K1 exp(−k2 t) JCp(t) + BV Cp(t) (1)

where Cp(t) is the tracer activity in arterial plasma,Jdenotes the convolution operation, and the kinetic rateconstants K1 (ml/g/min) and k2 (L/min) describe the ex-change between the arterial plasma and the combinedtissue pool. Furthermore, BV is the blood volume frac-tion, which was fixed as 5% in all our computations. Thevolume of distribution (VD) of FMZ in brain tissue isthen defined as the quotient of the rate constants K1/k2

representing FMZ ligand binding (VD4 Bmax/KD; Bmax

is the density of binding sites, KD is the binding affinity)(22). To compute efficiently a pixel-by-pixel representa-tion of FMZ K1 and VD, we used the metabolite-corrected arterial input curve and a weighted integrallookup table approach (23).

To obtain regional values of FMZ VD, regions of in-terest (ROIs) initially were drawn at the following loca-tions: thalamus, basal ganglia, medial temporal region(including the hippocampus and amygdala), superior

frontal cortex, primary visual cortex (Brodman area 17),temporal cortex, medial prefrontal cortex, cerebellum,and the hemisphere. These regions were consistent withthose used for our previous study of age-related, region-specific changes of FMZ binding in children who did nottake VGB (19). Because high-resolution MRI with volu-metric slicing was not available in most patients, ROIswere drawn on FMZ uptake images for the majority ofthese brain regions by the same person (C.J.), who wasblinded with respect to the age or medication of thepatients. Our previous studies showed that summed ac-tivity images between 10 and 20 min after injection(FMZ uptake images) closely resemble the pattern ob-served in FMZ VD images (24). In addition, ROIs forthalamus and basal ganglia were drawn on K1 images,because these structures are not well visualized on theuptake images because of the low activity values in thesestructures. FMZ uptake and K1 images occupy the samephysical space as the FMZ VD images, so all ROIs de-fined on FMZ uptake and K1 images were directly copiedto FMZ VD images, and regional FMZ VD values weredetermined for each subject. It was reported previouslythat FMZ VD values derived from the hemisphere con-tralateral to the epileptic focus in adults with partial epi-lepsy are not different from those determined in normalcontrols (19,25). Therefore we defined ROIs contralat-eral to the epileptic focus. The hemispheric region in-cluded all supratentorial structures.

Statistical analysisRegional FMZ VD values were obtained in all sub-

jects. In our previous studies including children olderthan 8 years, we fitted the decline of FMZ VD with ageby using an exponential function. However, the initialpart of this decline can be well approximated by using alinear function allowing the application of a simpler, lin-ear statistical method. Initially, the subjects were sepa-rated according to their medication into a VGB and anon-VGB group. To determine if the decline with age ofFMZ VD in these two groups is significantly different aswell as to determine if a decline different between groupsis due to a drug effect, we used a repeated measuresmultivariate analysis of covariance (MANCOVA) wherethe between-subjects factor was the drug and the within-subjects factor was the ROI, with age as a covariate. Wedetermined the significance of the main effects (drug,region) as well as the drug-by-region interaction. In thecase that the drug-by-region interaction was significant,we performed simple effect tests for each individual re-gion utilizing at test with a composite error term reflect-ing the within- and between-subject sources of variation.Adjustment for multiple comparisons in the simple ef-fects test was performed using the modified Bonferronicorrection (26). The correctedp value for simple effectstests in eight ROIs (yielding seven degrees of freedom) is

C. JUHASZ ET AL.1322

Epilepsia, Vol. 42, No. 10, 2001

0.044. Subsequently, similar analyses were performedafter separating the subjects according to two other medi-cations, CBZ and VPA, which are known as not to alterbrain GABA levels and which were taken by at least fiveof the patients. All statistical analyses were performedwith the SPSS 10.0 software package. Statistical signifi-cance was assessed using a corrected p value of <0.05.

RESULTS

Figure 1 shows hemispheric FMZ VD values obtainedfrom the VGB and the non-VGB groups. An initial over-all analysis including all defined ROIs indicated thatFMZ VD values in each group significantly decline withage (average standardized beta4 −0.23, p4 0.008).After controlling for age, we determined a significantoverall drug effect (p4 0.012), a highly significant re-gion effect (p < 0.001) as well as a significant drug-by-region interaction (p4 0.02). Simple effect testsbetween groups for each individual ROI were then per-formed using a modified Bonferroni correction. Usingthis correction, the critical p value was adjusted to avalue of 0.044. Table 2 reports the results of the simpleeffect tests, indicating that after controlling for age, theVGB and non-VGB groups differ significantly in theprimary visual cortex, the superior frontal cortex, thetemporal cortex, the medial prefrontal cortex, the medialtemporal region, and in the cerebellum. In contrast, the

basal ganglia and thalamus regions did not differ signifi-cantly between the two groups. Figure 2 shows a respec-tive PET image plane obtained from a 6-year-old childmedicated with VGB and that obtained from an age-matched child not taking VGB. The pattern of FMZ VDshowed the highest values in the primary visual cortexfollowed by other cortical areas and the medial temporalregion. The lowest values were determined in the cer-ebellum, the basal ganglia, and the thalamus in descend-ing order (Table 2).

Similar analyses were performed on the FMZ VDdata, but after separating groups according to beingmedicated first with CBZ (n4 6) and non-CBZ (n4 9)groups and then with VPA (n4 5) and non-VPA (n410) groups. When comparing the CBZ versus the non-CBZ groups, no significant drug effect (p4 0.97) ordrug-by-region interaction (p4 0.59) could be deter-mined. Similarly, comparison between the VPA and thenon-VPA groups revealed no significant drug effect (p4 0.55), and no significant drug-by-region interaction (p4 0.94). Figure 3 shows the group means and SD overall ROIs after controlling for age. It demonstrates thegroup difference detected for VGB, whereas no groupdifferences were present for CBZ or VPA.

FIG. 2. Representative [11C]flumazenil (FMZ) volume of distri-bution (VD) images of two children with similar age. The figuredemonstrates considerably higher cortical (in particular occipital)FMZ binding in a 2.0-year-old child, who did not take vigabatrin(VGB) (A), as compared with a 2.3-year-old child (B) who wastaking VGB for 4 months before the positron emission tomogra-phy scan. Thalamic FMZ binding differed only slightly betweenthe two patients.

FIG. 1. Hemispheric [11C]flumazenil (FMZ) volume of distribu-tion (VD) values obtained from the vigabatrin (VGB, solid sym-bols) and non-VGB (open circles) groups fitted with a linearfunction. Overall, we determined a significant decline with age inboth groups (0.008). Moreover, the rate of decline observed inthe VGB group was significantly higher than that determined inthe non-VGB group (p = 0.02). In the VGB group, patients withinfantile spasms (solid squares, n = 4) did not appear to havedifferent VD values as compared with the other three withoutinfantile spasms (solid circles).

TABLE 2. Average FMZ VD values (and SD) of thepatients treated vs. not treated with vigabatrin in differentbrain regions after controlling for the decline of FMZ VD

with age

Region non-VGB VGB p Value

Primary visual cortex 8.96 (0.38) 7.39 (0.44) p < 0.001a

Superior frontal cortex 6.53 (0.26) 5.14 (0.31) p4 0.002a

Temporal cortex 6.43 (0.28) 5.38 (0.32) p < 0.001a

Medial prefrontal cortex 5.98 (0.24) 4.60 (0.29) p4 0.002a

Medial temporal region 5.91 (0.24) 4.89 (0.28) p4 0.012a

Cerebellum 5.08 (0.22) 4.06 (0.26) p4 0.013a

Basal ganglia 4.73 (0.18) 4.29 (0.21) p4 0.23Thalamus 4.38 (0.20) 3.90 (0.23) p4 0.19

a Significant after applying the modified Bonferroni correction (26)(critical p value is 0.044).

FMZ, [11C]flumazenil; VD, volume of distribution.

VIGABATRIN EFFECT ON GABAA RECEPTORS 1323

Epilepsia, Vol. 42, No. 10, 2001

DISCUSSION

Our findings demonstrate that prolonged treatmentwith VGB, a drug that increases brain GABA concen-tration (5,6,10–13), may cause a decrease of FMZ bind-ing in children with intractable epilepsy. This suggeststhat VGB, and possibly other therapeutic agents or drugsof abuse binding to GABAA receptors, may alter thedevelopmental expression of this receptor complex pos-sibly via agonist-mediated downregulation. These find-ings differ from two recently published studies that failedto demonstrate similar changes in adults (10,12). How-ever, our recent PET study demonstrated that FMZ VD isconsiderably higher in young children than in adults, anda marked, exponential age-related decline of GABAA-receptor binding occurs in the nonepileptic hemisphereof children with epilepsy who do not use VGB (19).Subcortical regions reach adult FMZ VD values earlier(14–17.5 years) than cortical regions (18–22 years),which is consistent with previous studies of maturationalchanges of glucose metabolism in children (27). Thefindings suggested that postnatal maturation of the hu-man GABAA-receptor complex affects in vivo propertiesof the BZD binding site. Because FMZ VD is a macro-parameter reflecting both Bmax and KD, it remained un-clear whether higher FMZ binding in young children isdue to a higher receptor density, a higher affinity forFMZ related to the subunit composition of GABAA re-ceptors found early in postnatal brain development, orboth. Our present data in children younger than 8 years

are consistent with a decrease of FMZ binding with age.However, because of the small age range, we applied alinear model to our data rather than an exponential func-tion. The age-related decline of FMZ VD in young chil-dren who were medicated with VGB showed thatGABAA-receptor binding is shifted downward in com-parison to the non-VGB group, but did not reach adultvalues [as determined in our previous study (19)]. Thisraises the possibility that an increased supply of the ago-nist is able to change binding properties of the immatureGABAA-receptor complex even if this effect is not de-tected in the adult brain (10,12), perhaps because oflower baseline values of FMZ binding under normal cir-cumstances. Another possible interpretation of our find-ings is that long-term VGB treatment and the resultingincrease of brain GABA levels may have led to an ac-celerated maturation of the GABAA receptors, which isreflected by the FMZ binding shift toward adult values.

Some confounding factors in our study should be ad-dressed when interpreting our results. First, the mean ageof the VGB group was lower than that in the childrenwho were not medicated with VGB. This difference,however, was taken into account in the statistical analy-sis. Further, according to our previous findings of in-creased FMZ VD values in young children (19), the FMZVD values of the VGB group should have been higherthan in the other children, if VGB were not to influenceFMZ binding. Yet six of the seven hemispheric VD val-ues in the VGB-group were below the regression linederived from the non-VGB group (see Fig. 1), and inthree children (including two younger than 3 years), thevalues (VDs between 3.9 and 4.4 ml/g) were in the nor-mal adult range [in our previous study (19), the mean VDin the normal adult group was 4.18 ± 0.38 ml/g]. It islikely that an even more robust effect could have beendetected if the age of the two groups had matched better.

In addition to age, the two groups in our study showedmore differences in their seizure types (see Table 1). Inthe VGB group, four of seven patients had infantilespasms during the treatment period or by history,whereas the non-VGB group did not include any patientswith infantile spasms. Although the number of subjects issmall, it appears that within the VGB-group in Fig. 1,there were no clear differences between patients withinfantile spasms (denoted by squares) and those without(denoted by solid circles). Nevertheless, it cannot be ex-cluded that differences in FMZ binding between the twogroups are partially attributed to different seizuresmechanisms in addition to different medication. Involve-ment of GABAergic mechanisms in the pathophysiologyof infantile spasms was suggested by studies showingdecreased GABA concentration in the cerebrospinalfluid of children with infantile spasms (28,29). Similarchanges were found in unmedicated children with gen-eralized tonic–clonic seizures (29). Nevertheless, the ef-

FIG. 3. Group means (with standard deviations) of [11C]fluma-zenil (FMZ) binding (volume of distribution; VD) over all brainregions after controlling for age in the 15 patients grouped ac-cording to prolonged treatment with three different antiepilepticdrugs [vigabatrin (VGB; n = 8), valproate (VPA; n = 5); carba-mazepine (CBZ; n = 6)]. A significant group difference could bedetected only for VGB (p = 0.012), whereas no group differenceswere present for VPA (p = 0.55) or CBZ (p = 0.97).

C. JUHASZ ET AL.1324

Epilepsia, Vol. 42, No. 10, 2001

fect of seizure type (infantile spasms vs. partial orgeneralized tonic–clonic seizures) on thein vivo FMZbinding in children remains to be determined.

Further confounding factors in our study include thepossible effect of sedation applied during the PET scan-ning and the use of various AEDs in addition to VGB bypatients in both groups. It has been reported previouslythat PTB (used for sedation in some children) has nosignificant effect on FMZ binding (30), but it remains apossibility that chloral hydrate might cause allostericchanges affecting FMZ affinity. However, such changeswould be expected to result in an increased affinity,which would not explain a decreased FMZ binding in ourVGB group. Further, in our previous study, we showedthat chloral hydrate did not significantly affect whole-brain FMZ VD values (19). In addition, test–retest stud-ies in adult epilepsy patients at our institution showed nosignificant differences between regional patterns of FMZVD with and without chloral hydrate (unpublished data).

In the present study, the majority of our patients weremedicated with more than one drug; thus none of ourgroupings according to drug treatment resulted in homo-geneous groups in terms of medication. Still, separationaccording to VGB resulted in robust group differences,whereas no similar effects based on grouping accordingto VPA or CBZ were found. Nevertheless, prospectivestudies on patients with monotherapy would be moresensitive to examine drug effects on GABAA-receptorfunction.

Some AEDs taken by our patients might act throughalteration of brain GABA levels. In particular, topira-mate (TPM) was recently shown to elevate brain GABAin both healthy subjects and patients with epilepsy(31,32). Although two patients in our VGB group usedTPM as well, their FMZ VD values were not among thelowest in this group (one of them had the highest hemi-spheric value that was thought to be due to his youngage; see Table 1), suggesting that TPM treatment mighthave a minimal, if any, effect on the lowered mean FMZVD value of the VGB group. In addition, three childrenin the non-VGB group used GBP, which was also re-ported to increase brain GABA levels (33). FMZ VDvalues of these patients also did not appear to consis-tently deviate from the others. However, it remains to bedetermined whether TPM, GBP, and other AEDs that donot directly alter brain GABA level [but possibly act atthe GABAA receptors through allosteric changes in thereceptor (34)] interfere with the effect of VGB, whencombined.

Our findings of age-specific effect of VGB on corticalGABAergic receptor binding may have implications forthe long-term consequences of prolonged GABAergicdrug use in humans during neurodevelopment. It hasbeen recognized recently that inhibitory neurotransmit-ters play an important role not only in neural signaling,

but also may have toxic effects on neurons under certaincircumstances. For example, there is emerging evidencethat excessive amounts of endogenous GABA can beneurotoxic, and GABAA-receptor overactivation postna-tally may cause a gradual cell loss in the hippocampus ofrats (35). Recentin vivo data also demonstrated the pos-sible toxic effect of VGB and increased GABA underdepolarizing conditions (36). A developmental window,coinciding with the period of synaptogenesis, for suscep-tibility to overinhibition-induced neuronal death wassuggested based on animal studies (37). A similar effectmay be related to the recently suggested impairment ofGABAergic cells located outside the brain, such as theamacrine cells of the retina (38), providing a possibleexplanation for the visual impairment occasionally ob-served with VGB treatment (39–41).

Our present study represents preliminaryin vivo find-ings suggesting that VGB may induce a decrease of hu-man GABAA receptors in the developing central nervoussystem. A similar effect of other drugs and substancestargeting the GABAergic system can be postulated. Suchinfluences may interfere with developmental plasticity ofthe human brain and, therefore, deserve further attentionwith regard to their potential reversibility and ultimateconsequences.

REFERENCES

1. Lambert N, Grover L. The mechanism of biphasic GABA re-sponses.Science1995;269:928–9.

2. Sivilotti L, Nistri A. GABA receptor mechanisms in the centralnervous system.Prog Neurobiol1991;36:35–92.

3. Petroff OAC, Rothman DL, Behar KL, et al. Low brain GABAlevel is associated with poor seizure control.Ann Neurol1996;40;908–11.

4. Jung MJ, Palfreyman MG. Vigabatrin: mechanisms of action. In:Levy RH, Mattson RH, Meldrum BS, eds.Antiepileptic drugs.4thed. New York: Raven Press, 1995:903–13.

5. Petroff OAC, Rothman DL, Behar KL, et al. Human brain gamma-aminobutyric acid levels and seizure control following initiation ofvigabatrin therapy.J Neurochem1996;67:2399–404.

6. Petroff OAC, Rothman DL, Behar KL, et al. Human brain GABAlevels rise after initiation of vigabatrin therapy but fail to risefurther with increasing dose.Neurology1996;46:1459–63.

7. Appleton RE, Peters AC, Mumford JP, et al. Randomised, placebo-controlled study of vigabatrin as first-line treatment of infantilespasms.Epilepsia1999;40:1627–33.

8. Cossette P, Riviello JJ, Carmant L. ACTH versus vigabatrintherapy in infantile spasms: a retrospective study.Neurology1999;52:1691–4.

9. Bruni J, Guberman A, Vachon L, et al. Vigabatrin as add-ontherapy for adult complex partial seizures: a double-blind, placebo-controlled multicentre study: The Canadian Vigabatrin StudyGroup.Seizure2000;9:224–32.

10. Weber OM, Verhagen A, Duc CO, et al. Effects of vigabatrinintake on brain GABA activity as monitored by spectrally editedmagnetic resonance spectroscopy and positron emission tomogra-phy. Magn Reson Imaging1999;17:417–25.

11. Mattson RH, Petroff O, Rothman D, et al. Vigabatrin: effects onhuman brain GABA levels by nuclear magnetic resonance spec-troscopy.Epilepsia1994;35(suppl 5):S29–32.

12. Verhoeff NPLG, Petroff OAC, Hyder F, et al. Effects of vigabatrinon the GABAergic system as determined by [123I]iomazenilSPECT and GABA MRS.Epilepsia1999;40:1433–8.

VIGABATRIN EFFECT ON GABAA RECEPTORS 1325

Epilepsia, Vol. 42, No. 10, 2001

13. Novotny EJ Jr, Hyder F, Shevell M, et al. GABA changes withvigabatrin in the developing brain.Epilepsia1999;40:462–6.

14. Dulac O, Chiron C, Luna D, et al. Vigabatrin in childhood epi-lepsy.J Child Neurol1991;suppl 2:S30–7.

15. Armijo JA, Cuadrado A, Bravo J, et al. Vigabatrin serum concen-tration to dosage ratio: influence of age and associated antiepilepticdrugs.Ther Drug Monit1997;19:491–8.

16. Brooks-Kayal AR, Pritchett DB. Developmental changes in humangamma-aminobutyric acidA receptor subunit composition.AnnNeurol 1993;34:687–93.

17. March D, Shaw CA. Fetal, neonatal and adult expression of ben-zodiazepine receptor subtypes in human visual cortex.Eur J Phar-macol1993;236:333–6.

18. Reichelt R, Hofmann D, Fodisch HJ, et al. Ontogeny of the ben-zodiazepine receptor in human brain: fluorographic, immuno-chemical, and reversible binding studies.J Neurochem1991;57:1128–35.

19. Chugani DC, Muzik O, Juha´sz C, et al. Postnatal maturation ofhuman GABAA receptors measured with positron emission tomog-raphy.Ann Neurol2001;49:618–26.

20. Jewett DM, Mangner TJ, Watkins GL. Captive solvent methods forfast, simple carbon-11 radioalkylations. In: Emram AM, ed.Newtrends in radiopharmaceutical synthesis, quality assurance andregulatory control.New York: Plenum Press, 1991:387–91.

21. Barre L, Debruyne D, Abadie P, et al. A comparison of methodsfor the separation of [C-11]Ro 15-1788 (flumazenil) from its me-tabolites in the blood of rabbits, baboons and humans.Appl RadiatIsot 1991;42:435–9.

22. Koeppe RA, Holthoff VA, Frey KA, et al. Compartmental analysisof [C-11]flumazenil kinetics for the estimation of ligand transportrate and receptor distribution using positron emission tomography.J Cereb Blood Flow Metab1991;11:735–44.

23. Alpert NM, Eriksson L, Chang JY, et al. Strategy for the measure-ment of regional blood flow using short-lived tracers and emissiontomography.J Cereb Blood Flow Metab1984;4:28–34.

24. Niimura K, Muzik O, Chugani DC, et al. [11C]flumazenil PET:activity images versus parametric images for the detection of neo-cortical epileptic foci.J Nucl Med1999;40:1985–91.

25. Henry TR, Frey KA, Sackellares JC, et al. In vivo cerebral me-tabolism and central benzodiazepine-receptor binding in temporallobe epilepsy.Neurology1993;43:1998–2006.

26. Keppel G.Design and analysis: a researcher’s handbook.3rd ed.Englewood-Cliffs, NJ: Prentice-Hall 1991.

27. Chugani HT, Phelps ME, Mazziotta JC. Positron emission tomog-raphy study of human brain functional development.Ann Neurol1987;22:487–97.

28. Ito M, Mikawa H, Taniguchi T. Cerebrospinal fluid GABA levelsin children with infantile spasms.Neurology1984;34:235–8.

29. Loscher W, Siemes H. Cerebrospinal fluid gamma-aminobutyricacid levels in children with different types of epilepsy: effect ofanticonvulsant treatment.Epilepsia1985;26:314–9.

30. Bertz RJ, Reynolds IJ, Kroboth PD. Effect of neuroactive steroidson [3H]flumazenil binding to the GABAA receptor complex invitro. Neuropharmacology1995;34:1169–75.

31. Kuzniecky R, Hetherington H, Ho S, et al. Topiramate increasescerebral GABA in healthy humans.Neurology1998;51:627–9.

32. Petroff OA, Hyder F, Mattson RH, et al. Topiramate increasesbrain GABA, homocarnosine, and pyrrolidinone in patients withepilepsy.Neurology1999;52:473–8.

33. Petroff OA, Rothman DL, Behar KL, et al. The effect of gaba-pentin on brain gamma-aminobutyric acid in patients with epi-lepsy.Ann Neurol1996;39:95–9.

34. Granger P, Biton B, Faure C, et al. Modulation of the gamma-aminobutyric acid type A receptor by the antiepileptic drugs car-bamazepine and phenytoin.Mol Pharmacol1995;47:1189–96.

35. Xu W, Cormier R, Fu T, et al. Slow death of postnatal hippocampalneurons by GABA(A) receptor overactivation.J Neurosci2000;20:3147–56.

36. Lukasiuk K, Pitkanen A. GABA(A)-mediated toxicity of hippo-campal neurons in vitro.J Neurochem2000;74:2445–54.

37. Ikonomidou C, Bittigau P, Ishimaru MJ, et al. Ethanol-inducedapoptotic neurodegeneration and fetal alcohol syndrome.Science2000;287:1056–60.

38. Krauss GL, Johnson MA, Miller NR. Vigabatrin-associated retinalcone system dysfunction: electroretinogram and ophthalmologicfindings.Neurology1998;50:614–8.

39. Lawden MC, Eke T, Degg C, et al. Visual field defects associatedwith vigabatrin therapy.J Neurol Neurosurg Psychiatry1999;67:716–22.

40. Gross-Tsur V, Banin E, Shahar E, et al. Visual impairment inchildren with epilepsy treated with vigabatrin.Ann Neurol2000;48:60–4.

41. Harding GF, Wild JM, Robertson KA, et al. Electro-oculography,electroretinography, visual evoked potentials, and multifocal elec-troretinography in patients with vigabatrin-attributed visual fieldconstriction.Epilepsia2000;41:1420–31.

C. JUHASZ ET AL.1326

Epilepsia, Vol. 42, No. 10, 2001