Retinal Function Abnormalities in Patients Treated With Vigabatrin

Objective  To evaluate central and peripheral retinal function in patients treated with vigabatrin, an antiepileptic drug associated with peripheral visual field constriction (VFC).

Methods  Six patients with epilepsy treated with vigabatrin as add-on therapy for at least 3 years were included in this observational case series. All patients underwent a clinical ophthalmologic examination, color vision testing, standard perimetry, and full-field and focal foveal cone electroretinography. Four patients, 3 of whom had VFC, completed specialized computerized static light- and dark-adapted perimetry.

Results  In 9 of 11 eyes tested, foveal cone electroretinographic amplitudes were at or below the lower limit of normal. Dark-adapted perimetry demonstrated abnormal rod-derived visual fields in the 3 patients with vigabatrin-attributed VFC, whereas rod-derived thresholds were within normal limits throughout the visual field in the patient who did not have VFC.

Conclusions  Our results suggest that vigabatrin not only impairs peripheral cone-derived function as manifested by VFC but also affects foveal cone electroretinographic amplitudes and rod-derived visual fields. The clinical dilemma regarding the use of vigabatrin therapy is further complicated since central as well as peripheral visual function seems to be adversely affected.

VIGABATRIN (the γ-vinyl analogue of γ-aminobutyric acid[GABA]) is a selective, enzyme-activated, irreversible GABA aminotransferase inhibitor.¹ It is a custom-made antiepileptic drug (AED) found to be particularly useful in the management of drug-resistant partial seizures and infantile spasms, especially those secondary to tuberous sclerosis.^2-4 The antiepileptic effect is presumably mediated by elevation of GABA levels of the brain caused by inhibition of GABA metabolism.⁵

Initially, only relatively minor adverse effects were attributed to vigabatrin use.⁶ Over the last decade, the marked efficacy of this medication and its low toxic effects prompted widespread use in Europe. Recently, visual field abnormalities were reported in adults and children treated with this AED. In most documented cases, the visual field defect seems to be a specific, bilateral, and symmetrical peripheral constriction.^7-9 The fact that most patients are asymptomatic may have contributed to the late recognition of these visual field defects that apparently occur in more than 30% of the patients.^10-13

Previous reports emphasized that vigabatrin therapy induced peripheral cone-derived visual field constriction (VFC). However, intimations of central cone dysfunction were reported by Johnson et al, ¹⁴ Nousiainen et al, ¹⁵ and Hilton et al, ¹⁶ who noted that visual acuity, color vision, contrast sensitivity, and central short wavelength automated perimetry results can also be affected. In addition, recent studies reported abnormalities not only in cone-derived but also in rod-derived electroretinographic (ERG) responses, namely, rod b waves, scotopic oscillatory potentials, or both.^11,17-21 The objectives of our study were to electrophysiologically evaluate central cone function in patients with epilepsy who were treated with vigabatrin and to examine psychophysically whether impairment of rod-derived visual fields accompanies the cone-derived VFC in these patients.

Six patients with epilepsy treated with vigabatrin as add-on therapy for at least 3 years were included in this observational case series. They were carefully chosen from among the more than 30 patients who had epilepsy treated with vigabatrin. We selected these patients for their willingness to cooperate in visual function testing and for their ability to maintain stable fixation and concentration, which are required for reliable visual field and focal ERG testing. The following data were collected: patient demographics, seizure type, other AEDs, duration of vigabatrin therapy, and dosage (Table 1). Patients and parents, where appropriate, gave their informed consent to all procedures.

A routine ophthalmologic examination including assessment of visual acuity, ocular motility, pupillary reaction, and dilated fundus as well as an examination using a biomicroscopy slitlamp was performed in all patients. Subsequently, kinetic and/or static perimetry, color vision testing, full-field, and focal foveal cone ERG were performed according to cognitive function of the patients and their degree of cooperation.

Goldmann kinetic perimetry was performed using targets V-4-e, III-4-e_, and, in some cases, I-4-e on a 10-candela (cd)/m² white background. Standard static perimetry was performed using a visual field analyzer(Humphrey Field Analyzer; Allergan-Humphrey, San Leandro, Calif) using the 120-point screening test or programs 30-2 (or 24-2) and 30/60-2 when subject cooperation permitted. In 4 patients, specialized light- and dark-adapted static threshold perimetry was performed using a modified automated perimeter according to methods previously published by Jacobson et al.²² Briefly, 71 loci (12° grid) were tested across the visual fields to identify threshold intensities of 500-nm (blue) and 650-nm (red) stimuli that, in the dark-adapted state, can be used to differentiate between rod- or cone-mediated detection. In the light-adapted state, 600-nm (orange) stimuli on a 10-cd/m² white background were used. Rod-derived (at 500 nm) and cone-derived (at 600 nm) sensitivity losses were calculated based on locus-specific normal values. Color vision testing was performed using the Ishihara 38-plate color test to identify protanopia and deuteranopia and the Farnsworth D-15 panel test that can also identify tritanopia.

Full-field ERGs were recorded using corneal electrodes and a computerized system (Cyberscan model 4000; Microshev Systems, Efrat, Israel; or UTAS 3000; LKC Technologies Inc, Gaithersburg, Md). In the dark-adapted state, 2 responses were acquired—a rod response to a dim blue flash (Wratten filter No. 47b) and a mixed cone and rod response to a white flash (2.35 cd × s/m²). In the light-adapted state, a background light of 21.5 cd/m² was used to suppress rods; the cone response to flashes of white light(9.4 cd × s/m²) presented at 1 and 30 Hz was acquired. All ERG responses were filtered at 0.3 to 500 Hz and signal averaging was used. Focal foveal cone ERGs were performed using Burian-Allen contact lens electrodes(Hansen Ophthalmic Development Laboratory, Iowa City, Iowa) and a commercial computerized system (Maculo Scope Spectrum; Doran Instruments Inc, Littleton, Mass). A 42-Hz flickering stimulus subtending 4° on the retina within a 12° anulus of bright light (to suppress stray activation) was directed on the fovea or parafoveally using a Maxwellian view handheld system. Because of the difficulty of positioning the stimulating beam precisely and steadily on the fovea during recording, accurate and reliable testing with the focal ERG system is a challenging task and is highly dependent on the operator and subject. All tests in this study were performed by 1 of 2 operators (E.B. or A.O.), who had the experience of performing such testing in more than 540 patients (>1000 eyes) during the last 5 years. Recording was immediately stopped if fixation on the fovea was unstable and was resumed only when the beam was again well aimed. In addition, at least 4 sets of responses (each averaging 2500 waveforms) were collected in each eye. "Clustering" of the results of the different sets (ie, repeatability) helped to assess reliability of the recording. Focal foveal cone ERG testing was also performed in a control group of 6 consecutive epileptic patients with disease of comparable severity (age range, 14-27 years) who were being treated with multiple AEDs but have not received vigabatrin.

All patients underwent focal foveal cone ERG testing, using a 42-Hz flickering stimulus. The results of 8 test sets (averaging 2500 stimuli per set) in the right eye of patient 2 are shown in Figure 1A. Amplitudes of the averaged responses were borderline low while implicit times were within normal limits. Goldmann perimetry in this eye shows the typical constriction associated with the toxic effect of vigabatrin therapy (Figure 1B). In 9 of 11 eyes tested, foveal cone ERG amplitudes were similar at or below the lower limit of normal (Figure 1C; patient 4 allowed only 1 eye to be tested). Foveal cone implicit times were normal in all eyes. Amplitudes of the focal foveal cone ERG responses were within the normal range in 10 of 12 eyes tested in a control group of 6 epileptic patients with disease of similar severity who were not being treated with vigabatrin. In 1 eye each of 2 patients, amplitudes were reduced (0.14 µV and 0.09 µV; lower limit of normal, 0.18 µV).

On full-field ERG, which measures the mass response across the entire retina, cone 30-Hz flicker amplitudes were normal in patient 1, slightly below the lower limit of normal in patient 3, and markedly reduced in patients 4 through 6. Implicit times were borderline in patient 4 and delayed in patients 5 and 6 (Table 2).

Light-adapted (cone-derived) visual fields showed characteristic vigabatrin-associated peripheral constriction in patients 2, 4, 5, and 6 (patient 5 also had postsurgical right superior quadrianopsia). Patient 1 had left superior quadrianopsia (following surgery) but no constriction in other quadrants. Patient 3 had preserved visual fields. To assess peripheral rod-derived visual fields, dark-adapted perimetry was performed in 4 patients in addition to light-adapted perimetry. Patients 4 through 6 had cone- and rod-derived thresholds that were significantly elevated, ie, more than 2 SDs higher than the mean, in many peripheral loci (testing results of patients 4 and 5 are shown in Figure 2). Note that full-field cone flicker and rod b-wave ERG amplitudes were borderline or reduced in these patients (Table 2). Cone- and rod-derived thresholds in the right eye of patient 1 were mildly elevated but still within 2 SDs of the mean in most loci outside of the scotoma that developed following intracranial surgery (Figure 2).

Visual acuity ranged from 6/5 to 6/9 (Table 2). Color vision tests revealed that patient 3 had deuteranopia; patient 2 had a normal finding from the Ishihara examination but on the Farnsworth D-15 panel had 2 protanopic lines in the right eye and 2 nonspecific mistakes in the left eye. In all patients, no characterizing abnormalities of the fundus were identified.

Visual field constriction, which occurs in more than 30% of the treated patients, has recently emerged as a significant toxic adverse effect of vigabatrin therapy.^10,12,13 The pathologic features of VFC have been attributed mainly to the loss of peripheral cone–mediated vision, ^11,23,24 but recent findings have intimated that there may be a generalized decrease of cone sensitivity across the central and peripheral visual fields.¹⁴ Our findings confirm these initial observations. Using focal foveal cone ERG, we found that central (foveal) cone function was impaired. In addition, rod-derived visual fields, as tested by dark-adapted perimetry, were constricted in patients with cone-mediated VFC. This may be the psychophysical correlate to reduced scotopic ERG b-wave amplitudes and abnormal scotopic oscillatory potentials reported in some of the patients treated with vigabatrin.^11,17-21 While all of our patients were receiving additional AEDs, the results of Coupland et al¹⁷ support the notion that these signs of toxic reactions are vigabatrin related.

Foveal cone dysfunction is a particularly worrisome finding since the consensus has been that central vision, as measured by visual acuity, perimetry, and color vision, was preserved in patients treated with vigabatrin.^9,11 This was reassuring information for patients, parents, and physicians who were called on to decide between seizure control and the risk of VFC. However, preliminary psychophysical evidence that visual acuity, color vision, central short wavelength automated perimetry, and contrast sensitivity may also be impaired has recently been published.^10,14-16 There are only a few electrophysiological studies assessing foveal function (all using the multifocal ERG technique) with conflicting results: in 3 studies, the multifocal ERG revealed changes in the responses recorded from the periphery while central responses were considered normal.^24-26 Only 1 study showed diffuse perturbation of cone function in 3 patients.¹⁴ Multifocal ERGs were also performed in 8 additional patients with known vigabatrin-attributed VFC.²⁷ The issue of central involvement was not directly commented on in this fourth study. In some cases there was good concordance between the visual field defects and the multifocal ERG abnormalities, but in others the ERG abnormality was more diffuse than the visual field defect.

In our study we used focal foveal cone ERG, a different technique that allows assessment of foveal cone-derived electrophysiological function under direct visualization. The abnormal results obtained in most patients in conjunction with the preliminary results of Johnson et al¹⁴ support the contention that vigabatrin therapy does not spare central cone function. Based on this finding, we would like to propose that generalized retinal dysfunction can be induced by vigabatrin therapy. The electrophysiological abnormality on focal foveal cone ERGs could be part of the cone flicker amplitude reductions previously reported using full-field ERGs.^10,14,24 It is possible that the relative preservation of visual acuity and color vision is a function of the high density of cones in the macula that affords a "sparing effect" despite the electrophysiological abnormalities. However, compromise of central vision may ultimately occur with prolonged use or high accumulative doses of vigabatrin.

Rod-derived visual field impairment was not previously reported in patients treated with vigabatrin. The toxic effect of this medication is, thus, not restricted to the cone-derived system. This finding is in accord with the reports of reduced scotopic ERG b-wave amplitudes^11,17,21 also observed in 3 of our patients. However, a longitudinal study in children treated with vigabatrin did not find a substantial change in rod-derived ERG responses over an 18-month follow-up period.²⁸ The site and mechanism of the toxic effects of vigabatrin therapy are unknown. Multiple retinal cells are GABAergic, including several subtypes of cone and rod bipolar cells as well as many subtypes of amacrine cells.^29,30 Animal studies demonstrated retinal outer nuclear layer destruction with a peripheral disposition.³¹ However, most studies tend to attribute the ERG changes caused by vigabatrin to effects on the inner retina. Amacrine cell involvement has been implicated, ^21,23,32 as well as a possible toxic effect on Müller cells.^9,11,17 The elevated thresholds on psychophysical testing in the present study could point to impairment of bipolar cell function.³³ In addition, a postmortem study suggests that vigabatrin may be toxic to ganglion cells, resulting in loss of nerve fibers in the optic nerves, chiasm, and tracts.³⁴

The long-term outcome of vigabatrin-induced central cone dysfunction is as yet unknown. Vigabatrin-induced VFC persists, at least for the short-term, following discontinuation of the medication in most cases.^14,35 On the other hand, there is some evidence that in patients with minimal VFC, visual acuity and color vision deficits may be reversible with improvement in some of the electrophysiological alterations (such as the electro-oculogram Arden ratio and ERG oscillatory potentials).^{11,14,23,25,32} We have not yet had the opportunity to examine foveal cone ERGs serially in such cases, which would help assess the significance of the electrophysiological finding. The small number of patients thus far tested also precludes drawing conclusions regarding possible correlation between degree of VFC, visual acuity, and foveal cone ERG amplitudes. However, in the complex dilemma of whether to use vigabatrin therapy in patients for whom this is the only effective AED, the risk for eventual central visual impairment should, in our opinion, be considered.

Corresponding author: Eyal Banin, MD, PhD, Department of Ophthalmology, Hadassah-Hebrew University Hospital, PO Box 12000, Jerusalem 91120, Israel(e-mail: banine@md.huji.ac.il).

Submitted for publication April 26, 2002; final revision received February 11, 2003; accepted February 20, 2003.

This study was supported in part by grants from Yedidut Research, Mexico City, Mexico, and the Beyrard Family Foundation, Paris, France (Dr Shalev).