Cerebral MRI abnormalities associated with vigabatrin therapy


Address correspondence to Phillip L. Pearl, Department of Neurology, Children's National Medical Center, 111 Michigan Avenue, NW,  Washington, DC, U.S.A. E-mail: ppearl@cnmc.org


Purpose: Investigate whether patients on vigabatrin demonstrated new-onset and reversible T2-weighted magnetic resonance imaging (MRI) abnormalities.

Methods: MRI of patients treated during vigabatrin therapy was reviewed, following detection of new basal ganglia, thalamus, and corpus callosum hyperintensities in an infant treated for infantile spasms. Patients were assessed for age at time of MRI, diagnosis, duration, and dose, MRI findings pre-, on, and postvigabatrin, concomitant medications, and clinical correlation. These findings were compared to MRI in patients with infantile spasms who did not receive vigabatrin.

Results: Twenty-three patients were identified as having MRI during the course of vigabatrin therapy. After excluding the index case, we detected new and reversible basal ganglia, thalamic, brainstem, or dentate nucleus abnormalities in 7 of  22  (32%) patients treated with vigabatrin. All findings were reversible following discontinuation of therapy. Diffusion-weighted imaging (DWI) was positive with apparent diffusion coefficient (ADC) maps demonstrating restricted diffusion. Affected versus unaffected patients, respectively, had a median age of 11 months versus 5 years, therapy duration 3 months versus 12 months, and dosage  170  mg/kg/day versus 87 mg/kg/day. All affected patients were treated for infantile spasms; none of 56 patients with infantile spasms who were not treated with vigabatrin showed the same abnormalities.

Discussion: MRI abnormalities attributable to vigabatrin, characterized by new-onset and reversible T2-weighted hyperintensities and restricted diffusion in thalami, globus pallidus, dentate nuclei, brainstem, or corpus callosum were identified in 8 of 23 patients. Young age and relatively high dose appear to be risk factors.

Vigabatrin is an antiepileptic drug whose mechanism of action is to irreversibly inhibit γ-aminobutyric acid transaminase (Fig. 1). It has been recommended as a potential first-line treatment in patients with infantile spasms (Appleton et al., 1999), especially in patients with tuberous sclerosis (Elterman et al., 2001). Initial enthusiasm for clinical use, based on generally good tolerability and a lack of drug-drug interactions (French, 1999; Granstrom et al., 1999; Prasad et al., 2001), was diminished by reports of retinal toxicity with visual field constriction in approximately 30% of patients after 1 year of therapy (Krauss et al., 1998; Spence & Sankar, 2001; Vanhatalo et al., 2002).

Figure 1.

 GABA metabolic pathway. Vigabatrin irreversibly inhibits GABA-transaminase (GABA-T), leading to elevated GABA concentrations. GHB, gamma-hydroxybutyric acid; SSADH, succinic semialdehyde dehydrogenase.

The safety of vigabatrin was initially challenged by preclinical studies showing white matter vacuolation and intramyelinic edema in rats (Butler et al., 1987). A study using 275 mg/kg/day for 12 weeks duration in rats revealed T2-weighted and diffusion-weighted magnetic resonance imaging (MRI) abnormalities in frontal and occipital cortices and cerebellum compared to age-matched controls (Preece et al., 2004). These changes were reversible. Pathologic examination confirmed intramyelinic edema and microvacuolation with reactive astrocytosis. Another study in rats utilizing 25–40 mg/kg/day found increased T2 and diffusion-weighted imaging (DWI) signal, with pathological findings including microvacuolization, axonal edema, myelin loss, and reduced oligodendrocyte-associated enzymes and myelin basic protein (Qiao et al., 2000). Two weeks after vigabatrin discontinuation, there was residual elevation of T2-weighted signal in the hypothalamus and abnormal behavioral responses, suggesting only partial recovery.

A randomized placebo-controlled MRI study of  114  dogs found that T2-weighted signal hyperintensity in the thalamus, hypothalamus, and fornix became apparent at  4  to 7 weeks of treatment, increased with duration of treatment, corresponded to intramyelinic edema histopathologically, and normalized 16 weeks after the drug was discontinued (Peyster et al., 1995). Pre- and posttreatment MRI on seven dogs treated with vigabatrin demonstrated increased T2-weighted and decreased T1-weighted signal in the fornix, hypothalamus, and thalamus versus controls (Weiss et al., 1994). This correlated with intramyelinic edema on histopathology. Similar findings with microvacuolation are produced by ethanolamine-O-sulfate, another GABA-transaminase inhibitor (John et al., 1987).

The presence of vigabatrin-associated imaging abnormalities in humans has not been confirmed. In the case of a 13-month-old infant with Down syndrome, an MRI was obtained because of ongoing infantile spasms and opisthotonus while on a vigabatrin dosage of 83  mg/kg/day. We identified new T2-weighted hyperintense signal abnormalities involving the thalamus and midbrain that resolved on a follow-up study 4 months after discontinuation of vigabatrin. This led to further study and case ascertainment of patients demonstrating MR signal abnormalities while receiving vigabatrin therapy.


Participants and procedures

Subsequent to the evaluation of the index case, all patients who had an MRI while undergoing treatment with vigabatrin at Children's National Medical Center (CNMC) and Seattle Children's Hospital and Regional Medical Center (CHRMC) were identified. These cases were retrospectively reviewed for their age at the time of MRI, diagnosis, and duration and dose of vigabatrin at the time of imaging. Data from MRI scans pre-, on, and postvigabatrin, concomitant therapies, and clinical correlation were collected. We identified a group of patients who showed T2-weighted MRI signal abnormalities on vigabatrin therapy that were not present on MRI done either before or after vigabatrin therapy. This group was compared to patients showing no MRI change while on vigabatrin therapy. In addition, we identified patients, using the CNMC epilepsy database, who had infantile spasms and MRI. The MRI abnormalities, identified on a qualitative basis, were defined as fluid-attenuated inversion recovery (FLAIR) and T2-weighted hyperintensities with associated increased signal on DWI, and decreased signal on apparent diffusion coefficient (ADC) mapping. All MRIs were interpreted independently by two pediatric neuroradiologists (L.G.V. and N.K.), who were blinded to whether the patient had been treated with vigabatrin. The study was approved by the Institutional Review Board (IRB) at CNMC and CHRMC.

All patients were scanned using a common protocol, which used sagittal T1, axial proton density, T2 spin echo, FLAIR, and axial diffusion images. Most patients had coronal T2, and all infants had axial spin echo T1-weighted images. Of the eight patients showing vigabatrin-related changes, three were on a General Electric LX 1.5 Tesla scanner and five were on a Siemens 1.5 Tesla (parameters in Table 1). Of the 15 patients treated with vigabatrin who did not develop MRI abnormalities, 11 were studied on the General Electric and 4 on the Siemens.

Table 1.   MRI scanner protocols
Scanner make, model, field strength(spin echo)(fast spin echo)FLAIRDiffusion
  1. TR, repetition time; TE, echo time; ETL, echo train length; TI, inversion time; B, diffusion encoding strength; N/A, not applicable.

General Electric LX 1.5 Tesla ScannerTR 3500TR 3600TR 10,000TR 10,000
 TE 40, 120TE 102TE 120TE 81
 ETL 8TI 2200B 1000
Siemens 1.5 TeslaN/ATR 4000TR 9000TR 3100
 TE 90TE 107TE 89
 ETL 8TI 2500B 1000

Statistical analysis

Statistical analysis used paired and unpaired comparisons in contingency table analyses to evaluate the relationship between vigabatrin use and specific MRI abnormalities. Paired comparisons based on a continuity corrected McNemar's test were used to evaluate the risk within patients comparing the frequency of the MRI changes while on versus off the medication. Unpaired comparisons based on Fisher's exact test compared specific MRI abnormalities in patients with infantile spasms treated with vigabatrin and not treated with vigabatrin. In both of these analyses, the index case that brought attention to this association was excluded to avoid biasing the results in favor of the association.

Role of the funding source

No funding source had any role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had full access to all the data in the study and all took full responsibility for the decision to submit for publication.


In addition to the index case described above, 22 patients were identified as having MRI scans during the course of vigabatrin therapy. Among these, the age range of patients at the first MRI on vigabatrin was 9 months to  18  years 7 months (median age, 4 years 5 months). Underlying diagnoses were tuberous sclerosis complex (TSC) in seven patients (32%), cortical dysplasia in five patients (23%), cryptogenic infantile spasms in five patients (23%), and a single patient each with Down syndrome and infantile spasms, cobalamin C deficiency and infantile spasms, mesial temporal sclerosis, mitochondrial disorder, and progressive myoclonus epilepsy.

Seven of the 22 patients (32%) had new-onset vigabatrin-related diffuse T2 hyperintensities within the thalami, globus pallidus, midbrain, cerebellar dentate nucleus, dorsal medulla, or medial longitudinal fasciculus (Table 2 and Figs. 2–9). The index case had additional involvement of the corpus callosum. The T2 abnormalities were accompanied by increased signal on DWI. ADC maps, available in six of the seven cases, showed decreased signal in all lesions (consistent with restricted diffusion). Of the eight patients with vigabatrin-related changes, seven had MRI pre-, during, and posttherapy; one patient had MRI done only during and posttherapy (Table 2). Using McNemar's test, we found a statistically significant association between the MRI abnormality (p = 0.023) comparing MRI results within the same patients on versus off vigabatrin therapy. The median age at the time of diagnosis with infantile spasms was 3.9 months (range 2 weeks to  6  months). The median age at the time of MRI on vigabatrin was 11 months (range 9–18 months). The median dose of vigabatrin at the time of MRI was 170 mg/kg/day (range 83–220 mg/kg/day), while the median length of therapy at the time of MRI was 3 months (range 1–11 months).

Table 2.   Patients with vigabatrin-related MRI changes
Patient no.Age at timeLength of therapyVGB dose at timeMRI abnormality by ROIConcomitant therapies
and diagnosisof MRIon VGBof MRI (mg/kg/day)ThalamusGlobus pallidusMidbrainDentate nucleiMLFDorsal medulla
  1. *Minimal residual pallidal brightness on follow-up after 18-day interval, with resolution of all other findings. VGB, vigabatrin; N/A, not applicable.

1. Down syndrome9 months1.5 months170 None
 w/infantile spasms (IS)          
 10 monthsN/A0 (POST VGB) * ACTH
2. Cobalamin C deficiency3 monthsN/A0 (PRE VGB) Phenobarbital
 disorder w/IS     
 9 months1 months90  Topiramate, phenobarbital
 17 monthsN/A0 (POST VGB) Ketogenic diet, levetiracetam,
      clonazepam, topiramate
3. Cryptogenic IS4 monthsN/A0 (PRE VGB) ACTH
 9 months1 months220Topiramate, ACTH
 12 monthsN/A0 (POST VGB) Topiramate, prednisone
4. Cryptogenic IS3 monthsN/A0 (PRE VGB) Phenobarbital zonisamide
 10 months5 months194Phenobarbital zonisamide,
           levetiracetam, ketogenic diet
 13 monthsN/A0 (POST VGB) Phenobarbital zonisamide,
      levetiracetam, ketogenic diet
5. Cryptogenic (IS)6 monthsN/A0 (PRE VGB) None
 12 months3 months113 None
 18 monthsN/A0 (POST VGB) None
6. Down syndrome5 monthsN/A0 (PRE VGB) None
 w/IS (index case)     
 13 months4 months83  Levetiracetam, topiramate
 17 monthsN/A0 (POST VGB) Valproate, carnitine
7. Cryptogenic IS7 monthsN/A0 (PRE VGB) ACTH
 13.5 months3 months125   Clonazepam, ketogenic diet,
 18 monthsN/A0 (POST VGB) Oxcarbazepine, zonisamide
8. Cryptogenic IS6 monthsN/A0 (PRE VGB) Clonazepam, topiramate
 18 months11 months190Topiramate, ketogenic diet
 34 monthsN/A0 (POST VGB) Levetiracetam
Figure 2.

 Thirteen-month-old boy on vigabatrin for 4 months (83 mg/kg/day at time of MRI; index case, Patient 6 in Table 2). (A) T2-weighted MRI prior to vigabatrin (normal). (B) T2-weighted MRI on vigabatrin; increased signal in thalami (arrows). (C) ADC confirms increased restricted diffusion (on vigabatrin) (arrows). (D) T2-weighted image 4 months following cessation of vigabatrin; resolved.

Figure 3.

 Twelve-month-old on vigabatrin for 3 months (113 mg/kg/day at time of MRI, Patient 5 in  Table  2). (A) T2-weighted MRI previgabatrin; normal, age 6  months. (B) ADC map previgabatrin; normal, age 6 months. (C) T2-image on vigabatrin shows increased thalamic signal (arrows). (D) ADC map on vigabatrin; restricted diffusion in thalami (arrows). (E) T2-image done 6  months following cessation of vigabatrin; resolution. (F) ADC map done 6 months following cessation of vigabatrin; resolution.

Figure 4.

 Nine-month-old on vigabatrin for 6 weeks (170 mg/kg/day at time of MRI, Patient 1 in Table 2). (A) T2-weighted MRI with increased signal in globus pallidi (arrowheads) and thalami (arrows). (B) DWI;  increased signal in globus pallidi (arrowheads) and thalami (arrows). (C) T2-image 18-days following cessation of vigabatrin shows resolution in thalami and minimal residual pallidal brightness. (D) DWI 18-days  following  cessation of vigabatrin shows resolution.

Figure 5.

 Nine-month-old on vigabatrin for 1 month (220 mg/kg/day at time of MRI, Patient 3 in Table 2). (A) T2-weighted MRI with increased signal in the medulla (arrowheads) and dentate nuclei (arrows). (B) T2-weighted MRI with increased signal in the globus pallidi (arrowheads) and thalami (arrows).

Figure 6.

 Ten-month-old on vigabatrin for 5 months (194 mg/kg/day at time of MRI, Patient 4 in Table 2). (A) T2-weighted MRI with increased signal in dentate (arrow) and medial longitudinal fasciculi (MLF; arrowheads). (B) Corresponding ADC map; restricted diffusion in dentate nuclei (arrows) and MLF (arrowheads). (C) T2-weighted MRI with increased signal in thalami (arrows) and globus pallidi (arrowheads). (D) Corresponding ADC map; restricted diffusion in thalami (arrows) and globus pallidi (arrowheads).

Figure 7.

 Thirteen and one-half-month-old on vigabatrin for 3 months (160 mg/kg/day at time of MRI, Patient 7 in Table 2). (A) T2-weighted MRI previgabatrin; normal, age 7 months. (B) Increased signal in dentate nuclei on vigabatrin (arrows). (C) T2-weighted MRI previgabatrin; normal, age 7 months. (D) Increased signal in MLF on vigabatrin (arrows).

Figure 8.

 Nine-month-old on vigabatrin for 2 months (90 mg/kg/day at time of MRI, Patient 2 in Table 2). (A) T2-weighted MRI shows increased signal in ventrolateral thalami (arrows). (B) ADC map; restricted diffusion in thalami (arrows).

Figure 9.

 Eighteen-month-old on vigabatrin for 11 months (190 mg/kg/day at time of MRI, Patient 8 in Table 2). (A) T2-weighted MRI shows increased signal in thalami (arrows). (B) ADC map; restricted diffusion in thalami (arrows). (C) Increased T2-signal in midbrain (arrow).

The MRI abnormalities showed complete resolution following withdrawal of vigabatrin in all patients except for minimal residual pallidal brightness in a single patient studied only 18-days following discontinuation of therapy (Table 2, Patient 1). The thalamus and midbrain were affected in seven of the eight patients, and the globus pallidi and dentate nuclei were affected in six patients. The medial longitudinal fasciculi and dorsal medulla were each affected in four patients. MR spectroscopy was obtained using single voxel imaging on six affected patients  (Table 2, Patients 2, 4, 5, 6, 7, and 8). This revealed decreased N-acetylaspartate (NAA)/creatine ratios in three patients  [Table 2, Patients 5, 6, and 7 (50%)]; all showed improvement on follow-up MR spectroscopy off vigabatrin.

All patients showing vigabatrin-related MRI changes had infantile spasms. Concomitant therapies were adrenocorticotropic hormone (ACTH), clonazepam, ketogenic diet, levetiracetam, phenobarbital, prednisone, topiramate, and zonisamide. Of three patients treated with the ketogenic diet, the diet was continued in one (Patient  4) at the time of the postvigabatrin MRI, and the imaging abnormalities resolved. Another patient (Patient 2) showed signal abnormalities on vigabatrin, with resolution on follow-up imaging off vigabatrin but started on the ketogenic diet. One patient (Patient 8) cotreated with topiramate was imaged twice on vigabatrin. The topiramate was discontinued at the time of the second MRI, but the abnormalities did not resolve until a later MRI was obtained following discontinuation of vigabatrin. Another patient cotreated with topiramate had resolution of abnormal signal when imaged off vigabatrin while remaining on topiramate (Patient 7).

Patients 1 and 3 (Table 2) were the only patients  reported  to have clinical worsening associated with the period of vigabatrin therapy. Patient 1 presented with a decrease in activity level when the vigabatrin dose was increased from 150 to 170 mg/kg/day. Examination  revealed mild lethargy and increased hypotonia, which resolved upon discontinuation of therapy. Patient 3, an 18-month-old girl with cryptogenic infantile spasms, developed a single episode of complex partial status epilepticus at 9  months of age, after  1  month of vigabatrin therapy. MRI showed changes in the basal ganglia, thalamus, and brainstem which resolved following discontinuation of therapy. This patient has had no other episodes of complex partial seizures or status epilepticus. Six of the eight patients with MRI abnormalities had no associated clinical worsening.

The remaining 15 patients (Table  3) (median age 5 years; range 11 months to 18 years 7 months) had no new imaging abnormalities during vigabatrin therapy. Each had stable MRI findings attributable to their underlying diagnosis. None of these patients were treated for infantile spasms. The median dose of vigabatrin at the time of MRI in this group was 87 mg/kg/day (range 15–200), and the median length of therapy at the time of the first MRI on vigabatrin was  12  months (range 1 month to 3.7 years) (Table  4).

Table 3.   Patients without vigabatrin-related MRI changes
 Age at timeLength of VGB therapyVGB Dose at the time Concomitant therapies
Patient no. and diagnosisof MRI on VGBat time of scan(s)of MRI scan(s) (mg/kg/day)MRI findingsat time of MRI on VGB
  1. VGB, vigabatrin.

 1. Tuberous sclerosis complex (TSC)11 months6 months110Consistent with TSC, no acute changeNone
 29 months24 months110No change 
 2. Tuberous sclerosis complex12 months5 months 50Consistent with TSC, no acute changeNone
 3. Progressive myoclonus epilepsy23 months3 months130Cerebellar atrophy and white matter loss delayed cortical myelination. No acute findingsClonazepam zonisamide
 4. Tuberous sclerosis complex1.4 years5 months200Consistent with TSC, no acute changeTopiramate
 5. Mitochondrial disease1.7 years3 months 65Mild global atrophy, no acute changeTopiramate
 6. Tuberous sclerosis complex2.8 years1.9 years130Consistent with TSC, no acute changeTopiramate
 3.1 years2.2 years130No change 
 4.6 years3.7 years130No change 
 7. Tuberous sclerosis complex3.2 years26 months114Consistent with TSC, no acute changeCarbamazepine levetiracetam
 4.3 years3.3 years120No change 
 8. Tuberous sclerosis complex4 years12 months 60Consistent with TSC, no acute changeNone
 9. Cortical dysplasia with complex partial seizures4.5 years6 months 86PVL, white matter gliosis, atrophic corpus callosum; no acute findingsCarbamazepine valproate
 5 years 12 months 50No change 
 5.8 years1.8 years 48No change 
 6.5 years2.5 years 45No change 
10. Cortical dysplasia4.9 years3 months100Cortical dysplasia (posterior temporal lobes), no acute findingsLamotrigine valproate
 5.2 years6.6 months121No change 
 5.9 years15 months 70No change 
 6.9 years27 months 94No change 
11. Tuberous sclerosis complex9.4 years6 months 80Consistent with TSC, no acute findingsGabapentin
 10.5 years18 months 40No changePhenytoin
12. Cortical dysplasia with complex partial seizures10 years 12 months 15Atrophic left thalamus, hippocampus, and cerebellum; no acute changePhenytoin
13. Cortical dysplasia with partial complex seizures11.4 years4 months 70Atrophic left amygdala, hippocampus; no acute findingsLamotrigine
 11.8 years9 months 70No change 
14. Mesial temporal sclerosis with complex-partial epilepsy16.9 years1 month 88Ventriculomegaly, hippocampal sclerosis; no acute changeLamotrigine
 17.4 years 6 months 50No change 
 18.3 years17 months 50No change 
15. Cortical dysplasia18.6 years 6 months100Hypomyelination, right amygdala dysplasia, no acute changeNone
Table 4.   Median data on patients on vigabatrin therapy
 Patients with MRI changesPatients without MRI changesInfantile spasm patients
 on vigabatrin (n = 8)on vigabatrin (n = 15)not on vigabatrin (n = 56)
  1. VGB, vigabatrin; N/A, not applicable.

Age at time of MRI scan11 months5 years5.6 months
 (range 9–18 months)(range 11 months to 18.6 years)(range 3 days to 3.1 years)
Length of vigabatrin therapy at time of MRI scan3 months12 monthsN/A
 (range 1–11 months)(range 1 month to 3.7 years) 
Dose of vigabatrin at time of MRI scan (mg/kg/day)17087N/A
 (range 83–220)(range 15–200) 

Given that all affected patients were treated for infantile spasms, we sought to compare this series to patients with infantile spasms that were not treated with vigabatrin. The CNMC epilepsy database identified 56 of these patients (29 males, 51.8%; 27 females, 48.2%) since 2000 who had MRI. The median age of this group was 5.6  months (range 3 days to 3.1 years) and the median  duration of spasms by the time of imaging was 1 month (range 9  days to 2.6 years). The findings revealed 23 patients with  migrational  anomalies (41%), 11 patients with cerebral atrophy (20%), 2 patients with tuberous sclerosis, 1 patient with Dandy Walker malformation, and 19 patients with no structural anomalies (40%). In MRI studies evaluated independently by two blinded reviewers, none of these patients were found to have the specific MRI findings under study (Fisher's exact test, p = 0.00007).


Clinical reports of MRI abnormalities potentially attributable to vigabatrin are limited. Kim and colleagues (1999) reported demyelination of the splenium of the corpus callosum associated with antiepileptic drug treatment in six patients, three of whom received vigabatrin in addition to phenytoin. One patient who had a normal pretreatment MRI developed a callosal lesion after 7 days of vigabatrin. A second patient had no pretreatment MRI scan, but 4 months after discontinuation of vigabatrin had resolution of a lesion seen during treatment. A third patient with a lesion on treatment had no posttreatment MRI. None of these patients had a series of scans taken before, during, and after treatment. The methodology and conclusions of the paper have been challenged (Tennison, 1999). Another study reported T2-weighted hyperintensity in the basal ganglia of three patients with infantile spasms treated with vigabatrin (Desguerre et al., 2003). The abnormality was presumed to be a possible source of the infantile spasms and present prior to treatment. There were no pre- or posttreatment scans.

Cocito conducted MRI scans in 11 adults with epilepsy treated with vigabatrin for more than 5 years and observed no detectable areas of abnormal signal in the white matter (Cocito & Maffini, 1995). Chiron reported normal MRI findings in children exposed to therapeutic doses of vigabatrin for an average of 11 months (Chiron et al., 1989). A large review failed to identify any definite case of  intramyelinic   edema in 412 humans treated with  vigabatrin   (Cohen et al., 2000). The data reviewed were  derived  from Hoechst Marion Roussel's global clinical trials and postmarketing surveillance databases and worldwide published literature, although the age-related demographics are not provided and it is unclear whether infants were included. Jackson and colleagues reported a quantitative MRI study of 45 patients with refractory partial seizures during a prospective, randomized, double-blind trial of vigabatrin, followed by open treatment and found no evidence of vigabatrin-related white matter changes after 20 or 35 weeks of treatment (Jackson et al., 1994).

Desguerre et al. (2008) recently reported on transient magnetic resonance diffusion abnormalities in West syndrome in five patients with infantile spasms imaged at a mean age of 13 months with follow-up  6  to 18 months later. Symmetrical T2-weighted and DWI changes were described in the globus pallidus, thalamus, brainstem, and dentate nucleus, which normalized on the second MRI. All patients were treated with vigabatrin during both MRI studies, and the authors did not attribute the changes to vigabatrin therapy. The mean dose, however, was  61  mg/kg/day during the follow-up MRI as opposed to  100  mg/kg/day on the initial MRI. Previgabatrin scans were not available.

Here we report a series of 22 patients, excluding the index case, treated with vigabatrin who underwent MRI scans during treatment. This is a retrospective analysis from our epilepsy database, wherein we identified patients who received MRI scans during vigabatrin therapy. Seven patients were found to have new-onset MRI abnormalities. MRI findings in all eight, including the index case, were  reversible  following discontinuation of vigabatrin. The  structures  involved include thalamus, globus pallidus, cerebellar dentate nucleus, midbrain, medial longitudinal fasciculus, and corpus callosum. DWIs were consistent with reversible cytotoxic edema. To our knowledge, this is the first report of patients with vigabatrin-associated MRI abnormalities who were imaged before, during, and following treatment. The affected patients were imaged between 9 and  18  months of age. The median duration of vigabatrin therapy was 3 months, and the median dosage at the time of imaging was 170 mg/kg/day. This data contrasts with the 15 patients treated with vigabatrin having no treatment-related MRI changes. The age range of this cohort was  11  months to 18 years 7 months, with a median age of 5 years. The median duration of vigabatrin therapy was  12  months, and the median dosage at the time of imaging in this group was 87 mg/kg/day.

New MRI abnormalities on vigabatrin were not seen in our patients with tuberous sclerosis, who comprised approximately one-third of our study population. Subsequent to studying our population, we have learned of a five month old infant with tuberous sclerosis who showed very similar T2- and diffusion-weighted MRI findings of the globus pallidus, midbrain, and dorsal pons on a dosage of  150  mg/kg/day (personal communication, S. Leber, M.D.).

All of our patients with the defined neuroimaging abnormality were treated with vigabatrin for infantile spasms. We therefore identified 56 patients with infantile spasms who received MRI and were not treated with vigabatrin. None had a similar imaging abnormality.

Although a single case report (Erickson et al., 2002) suggested that the ketogenic diet may produce basal ganglia changes on MRI, the T2-signal abnormality was limited to the posterior putamen and did not have the widespread involvement seen in our patients. Of three  infants in our series treated concomitantly with vigabatrin and the ketogenic diet, one remained on the diet and still showed resolution of signal abnormalities off vigabatrin. Another showed new-onset MRI changes while on vigabatrin, which cleared on follow-up MRI off vigabatrin, while having been subsequently treated with the diet. Hence, the imaging abnormalities do not appear to be related to the ketogenic diet. We additionally note that the brother of Patient 2 (Table 2) was also affected with cobalamin C deficiency and seizures and had a normal MRI.

All T2 abnormalities were accompanied by increased signal intensity on DWI and restricted ADC (ADC maps available in six of seven patients). Decreased diffusion can be seen with failure of the Na+/K+ ATP pump (e.g., acute infarction, ischemia, hypoglycemia, mitochondrial disorders, seizures), dense cell packing [e.g., lymphoma, primitive neuroectodermal tumor (PNET)], tissue vacuolation/spongiform change (e.g., Creutzfeld Jacob disease, some types of demyelination), and in tissues containing highly proteinaceous or viscous material (e.g., abscess) (Sagar & Grant, 2006). In our patient population, the explanation of the diffusion abnormalities is likely a toxic effect with transient failure of the Na+/K+ ATP pump or  possibly transient vacuolation. The distribution within the brainstem, cerebellum, thalamus, and globus pallidus likely reflects a specific glial or neuronal vulnerability to vigabatrin, based on regional variations of GABA metabolism.

In summary, these observations strongly suggest that the edema, microvacuolization, and astrogliosis seen in radiographic and histopathological animal studies may have a correlate in patients treated with vigabatrin. The affected patient group had abnormalities noted within a relatively short time frame, comparable to the therapy duration used in the animal studies. Our data suggests that infants are a high risk group for this complication, and there also appears to be a dose-related effect. These findings support our earlier observations (Pearl et al., 2006a, 2006b). Our study is a case series and would ideally be followed by a prospective, age-matched, placebo-controlled study of imaging in patients treated with vigabatrin. The pathophysiology of these changes is unknown, although the pharmacological effect of increasing GABA concentrations via vigabatrin may inadvertently lead to neuronal excitotoxicity due to the depolarizing effects of GABA in immature brain (Ben-Ari, 2006). Our findings highlight the potential for neurotoxicity in patients on vigabatrin, especially infants and those on relatively high doses, and suggest consideration of medication-related causes of clinical syndromes or MRI abnormalities that are reversible with drug discontinuation.


Ian Miller, M.D. contributed background research for the manuscript. All coauthors have been involved in the study and/or preparation of this manuscript. All have read and approved this manuscript. No undisclosed groups have had a primary role in this study. P.L.P. and R.P.S. contributed to protocol design and data interpretation. R.M. contributed to protocol design and statistical analysis. L.G.V., E.M.-W., A.H., S.T., and W.D.G. contributed to data interpretation. All authors participated in data extraction and writing of the report.

Conflict of interest: 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. P.L.P. is supported by the National Institutes of Health (NIH), Pediatric Neurotransmitter Disease Association, and Delman Family Fund for Pediatric Neurology Research, and received a consulting fee from Ovation Pharmaceuticals. R.S. is supported by the Mitochondrial Research Guild at Children's Hospital and Regional Medical Center. He has been the principal investigator in phase 2 and 3 clinical trials of antiepileptic medications developed by Marinus (ganaxalone), UCB Pharma (levetiracetam), GlaxoSmithKline (lamotrigine), and NIH (Childhood Absence Epilepsy study). He is on the speaker's bureau for GlaxoSmithKline and has been a consultant for Abbott and Ortho McNeil. A.H. is supported by Children's Research Institute GCRC, grant no. 5-MO1-RR-020359-02 and Delman Family Fund. H.P.G. receives support from the NIH and has received a consulting fee within the last 3 years from MedImmune, Inc. J.A.C. has been the principal investigator in phase 2 and 3 clinical trials of antiepileptic medications developed by Ovation Pharmaceuticals (clobazam), Marinus (ganaxalone), Eisai (rufinamide and zonisamide), Pfizer (pregabalin), UCB Pharma (levetiracetam), Ortho McNeil (Topiramate), GlaxoSmithKline (lamotrigine), and King Pharma (Vanquix, injectable diazepam). She has been a consultant for Ovation Pharmaceuticals. W.D.G. is supported by the NIH.