Vigabatrin-associated retinal cone system dysfunction

Vigabatrin is an effective antiepilepsy drug (AED) in widespread use in over 50 countries for treating partial-onset seizures and infantile spasms.^1-3 Vigabatrin release in the United States, however, has been delayed because of neurotoxicity in animals. This toxicity is characterized by intramyelinic edema and brain vacuolation that occurs following vigabatrin treatment in rodents and dogs, but not in chimps and humans.^4-6 The effect may be related to vigabatrin's mechanism of increasing GABA by inhibiting GABA transaminase.⁷ Long-term human safety studies using clinical examinations,² magnetic resonance (MR) imaging,⁸ and multimodal evoked potentials⁹ have had normal results, suggesting that vigabatrin neurotoxicity may be species-specific. Recently, however, persistent visual field constriction was reported in eight English patients treated with vigabatrin.^10-14 We identified four patients treated with vigabatrin at our center who developed visual abnormalities associated with retinal cone system dysfunction.

Methods. All subjects were receiving vigabatrin therapy through a vigabatrin safety study sponsored by Hoechst Marion Roussel. Patients agreed to study treatment with consent approved by the Johns Hopkins Joint Committee on Clinical Investigation. Patients had routine neurologic examinations and screening for adverse treatment events at 6-month intervals. Patients who developed visual complaints during treatment with vigabatrin (4 of 38 patients) had ophthalmologic testing.

Pattern-evoked visual evoked potentials (VEPs) were measured to a high-contrast alternating checkerboard having mean luminance of 171 candela/m². Responses were obtained to five different check sizes that varied from 6 minutes of arc angular subtense to 100 minutes of arc, in 0.3-log increments.

The patients' pupils were dilated with 1% tropicamide, and patients were dark-adapted for 30 minutes before electroretinogram (ERG) recording. Burian-Allen bipolar electrodes were positioned in both eyes for simultaneous recording after applying proparacaine hydrochloride for topical corneal anesthesia. A ground electrode was attached to the wrist. The head was positioned facing an integrating sphere for Ganzfeld stimulation. ERGs were recorded using a UTAS-E 2000 computer-controlled electrodiagnostic system(LKC Technologies, Gaithersburg, MD). The ERG data were filtered using analog filters with a passband of 0.3 to 500 Hz.

Single-flash ERGs were recorded as a function of stimulus luminance over a range of up to 4 log units in 0.2-log steps, starting with the lowest intensity at which a b-wave was elicited up to a maximum intensity of 1.46 cd s/m². Additional single-flash data covering the range from about 14 to about 209 cd s/m² were then collected in 0.4-log steps. Pupil areas were measured and retinal illuminance in troland seconds was calculated using pupil areas. ERG amplitudes and timing were compared against data from age-matched normal subjects.

A-wave analysis. The leading edges of rod a-waves were fitted with the function P3(I,t) ∼ {1-exp[-I S (t-t_d)²]} Rm_p3 where the amplitude P3 is a function of flash energy I and time t after flash onset. S is a sensitivity parameter that scales flash energy I; Rm_p3 is the maximum amplitude, t_d is a brief delay.^15,16 The three parameters Rm_p3, S, and t_d were estimated simultaneously from waveforms elicited by stimuli ranging from about 2 to 4 log scotopic troland seconds.

B-wave analysis. ERG b-wave amplitudes were measured as the difference in µV between the trough of the a-wave and the peak of the b-wave. Naka-Rushton functions of the form (Equation 1) were fit to these data, where I, R is an intensity-response ordered pair, R_max is the asymptotic amplitude, K is the semisaturation constant, and n is related to the slope.

Results. Patient 1. Clinical. A 58-year-old man had a 45-year history of complex partial and secondary generalized seizures. He had 56 complex partial seizures per month before vigabatrin treatment, failing all standard AEDs. He began vigabatrin added to carbamazepine 3.25 years ago, following which his seizures decreased in frequency to 1.5 seizures per month. Medication doses are vigabatrin 4.5 g/day and carbamazepine 600 mg/day.

Vision. One year after starting vigabatrin, the patient began bumping into objects. Visual acuity at this time was 20/30 in both eyes, and visual fields showed nonspecific constriction. Four months later, the visual fields were slightly worse. Recently, the patient reported increased difficulty with constricted peripheral vision. Neuroophthalmologic examination findings were best-corrected visual acuity 20/40 in the right eye and 20/30 in the left. Color vision using Hardy-Rand-Rittler (HRR) pseudoisochromatic plates was 10/10 in the right eye and 10/10 in the left. Both kinetic and static perimetry showed nonspecific constriction of visual fields. There was no relative afferent pupillary defect (RAPD). Results of slit-lamp examination were normal. Ophthalmoscopy revealed slightly narrowed retinal arteries and an irregular appearance in both maculas, consistent with surface wrinkling retinopathy. Both optic disks had a normal appearance.

Electroretinogram and visual evoked potential. ERG demonstrated widespread bilateral loss of cone function (figure). Rod function was normal. Cone oscillatory potentials in the mixed rod and cone ERG were sharply attenuated. Pattern VEP were delayed.

The vigabatrin dosage was gradually decreased to 3 g/day with no changes in ERG or VEP after 2 months.

Patient 2. Clinical. A 22-year-old woman had complex partial seizures and a left temporal lobe oligodendroglioma resected at age 17. Postoperatively, she had transient speech difficulties and some short-term memory difficulties. Three years earlier, she was having five seizures per month and was experiencing significant side effects on standard AEDs. She therefore began vigabatrin 2 g/day as monotherapy, whereupon seizures decreased to a frequency of 0.5 per month. Recently she started valproic acid in addition to her vigabatrin, developed gastric symptoms, and was switching to topiramate during the time ERGs were done.

Vision. One year ago, and 2 years after starting vigabatrin, she developed mildly blurred vision in both eyes, right more than left. Best-corrected visual acuity was 20/30+2 in the right eye and 20/20 in the left. Color vision using HRR plates was 10/10 in the right eye and 10/10 in the left. Visual fields were full by kinetic perimetry. The pupils were normal, without RAPD. Slit-lamp biomicroscopy showed an old anterior stromal corneal scar in the right eye. Ophthalmoscopy revealed normal foveal and macular reflexes in the right eye but a slightly irregular macular reflex in the left.

Electroretinogram and visual evoked potential. VEP was normal bilaterally. The ERG showed cone system functional decline in both eyes(figure). Oscillatory potential amplitudes are also reduced, consistent with both cone system dysfunction and reduced amacrine cell response. Rod function was normal.

Patient 3. Clinical. A 29-year-old woman had complex partial seizures beginning at age 18 years. She had frequent seizures on standard AEDs and had a 1-year period of remission following resection of her left temporal lobe. Her seizures then recurred, eventually increasing to 38 seizures per month. She began vigabatrin along with carbamazepine 18 months ago. While taking vigabatrin 4.5 g/day and carbamazepine 1400 mg/day, her seizure frequency decreased to 2 per week.

Visual. Shortly after starting vigabatrin, the patient noted a gradual constriction in her peripheral vision, greater in the left eye than the right. She also experienced a flickering sensation in the temporal field of the left eye and brief oscillopsia in the left eye when she awakened in the morning. Visual acuity without correction was 20/20 in the right eye and 20/20-1 in the left. Color vision using HRR plates was 10/10 in the right eye and 10/10 in the left. Kinetic perimetry revealed slight, nonspecifically constricted visual fields in both eyes. Static perimetry gave normal results. Slit-lamp examination revealed no abnormalities. Ophthalmoscopy revealed no evidence of retinal pigmentary disturbances, optic disk pallor, or optic disc swelling.

Electroretinography and visual evoked potential. VEP was normal in both eyes. ERG was remarkable for reduced cone system function in both eyes (see figure) and reduced rod photoreceptor function in the left eye. Oscillatory potential amplitudes were also reduced bilaterally, consistent with losses in electrical activity of both the cone system and amacrine cells. The reduction in cone function was greater in the left eye than in the right.

Patient 4. Clinical. A 67-year-old man developed complex partial seizures 6 months after a left parietal meningioma was resected in 1991. Surgery was complicated by bleeding and a right hemiparesis and aphasia. He did not tolerate standard anticonvulsants: phenytoin and carbamazepine caused allergic rashes, valproic acid caused thrombocytopenia, and phenobarbital produced emotional lability. He was treated with ethotoin and continued to have monthly seizures along with sedation. Four years ago he began vigabatrin 4 g/day as monotherapy. Seizures decreased to a frequency of one per 6 months, with one seizure in the past 12 months. Six months ago, he had mild slurred speech along with his aphasia, which improved when vigabatrin was reduced to 2500 mg/day.

Vision. Five months ago, and 3.5 years after starting vigabatrin, he noted difficulty with central vision in the right eye. An ophthalmologist found a best-corrected vision in the right eye of 20/50 and in the left eye of 20/20. A retinal specialist observed surface wrinkling retinopathy in the right eye unassociated with a posterior vitreous detachment. He continues to have stable, but blurred, central vision in the right eye. Best-corrected visual acuity was 20/40 in the right eye and 20/15 in the left. Color vision using HRR plates was 9.5/10 in the right eye and 10/10 in the left. Visual fields showed moderate peripheral constriction, greater temporally than nasally by both automated and kinetic perimetry. The pupils were normal, without RAPD. Slit-lamp biomicroscopy showed mild cataracts in both eyes. Fundus examination revealed normal-appearing optic discs and mild surface wrinkling retinopathy temporal to the macula of both eyes. There was no abnormal retinal pigmentation nor were there any abnormalities of the retinal vessels.

Electroretinogram and visual evoked potential. Latencies of recordable VEP waveforms were normal; however, the VEP was difficult to record, suggesting possible decreased P1 amplitude. The ERG showed cone amplitude and implicit time delay in both eyes (seefigure) and rod photoreceptor loss in the left. Oscillatory potential amplitudes are also reduced, consistent with both cone system dysfunction and reduced amacrine cell response.

Discussion. Our patient's symptoms of visual constriction and blurring following vigabatrin treatment are similar to those reported by eight patients in the United Kingdom^13,14 and four in Switzerland and Germany.¹⁷ Preferential cone system dysfunction is an uncommon finding in the general population, and thus it is unlikely that these patients had reduced cone responses before starting therapy with vigabatrin. Vigabatrin effects on retina may not be surprising, because vigabatrin increases brain gamma-aminobutyric acid (GABA) levels.¹⁸ GABA is an established inhibitory neurotransmitter in the vertebrate retina, and occurs in retinal horizontal and interplexiform cells as well as in many types of amacrine cells.^19-21 Although mainly found in the inner (postreceptor) retina,²² GABA has also been implicated as a regulator of cone synaptogenesis in neonatal rabbits.²³

Responses from horizontal, interplexiform, amacrine, and other retinal cells can be measured noninvasively using the ERG. The leading edge of the ERG a-wave reflects activity of the photoreceptors, whereas the b-wave reflects activity from a variety of inner retinal cells, including the bipolar, horizontal, and amacrine cells. ERG oscillatory potentials arise from a feedback loop involving mainly the amacrine cells, but also the ganglion and bipolar cells.

GABA selectively affects components of the ERG in a dose-dependent fashion. In rabbits, low concentrations of extracellular GABA enhanced the a- and b-wave amplitudes, whereas high concentrations reduced b-wave amplitudes and had no effect on photoreceptor potentials.²⁴

The antiepileptic drugs carbamazepine and phenytoin have also been shown to decrease ERG b-wave and oscillatory potential amplitudes.¹⁴ In a study of 67 patients with epilepsy, Bayer and colleagues found paradoxically that the addition of vigabatrin to the medical regimen of these patients promoted recovery of the b-wave amplitude.²⁵ Two of our patients were also receiving carbamazepine during the period visual symptoms developed; however, two patients were receiving vigabatrin alone.

Among our four patients taking vigabatrin, three had well-documented peripheral field constriction (one severe), two had mild-to-moderate reductions in visual acuity, and all four had reductions in the cone system and oscillatory potential components of the ERG. VEPs of two patients were normal, consistent with normal optic nerve function and normal or near-normal macular function. VEPs only evaluate the central 10 degrees of visual fields and may miss more peripheral dysfunction. None of the patients had baseline ERG or VEP before vigabatrin treatment.

The extreme reduction in oscillatory potentials in consistent with impaired function of the highly GABAergic amacrine cells. The reductions may also reflect diminished cone system activity, given that our recording conditions emphasized generation of cone- versus rod-driven oscillatory potentials. The source of the reduction in the cone ERGs is probably the inner retina, given the domination of inner retinal potentials in photopic and flicker ERGs²⁶ and the well-documented activity of GABAergic cells in the inner retina. Albino rats have photic-induced damage to the outer nuclear layer of the retina following vigabatrin treatment,²⁷ but a relationship between this injury and our patients' cone dysfunction is unclear. Our patients' retinal disturbances are dissimilar to the optic neuropathy recently reported in a child in Ireland.²⁸

The one patient in our study with VEP abnormalities also had symptoms of the longest duration (2 years) and greatest severity, severe ERG cone system and oscillatory potential reductions, peripheral vision loss on perimetry, and evidence of a bilateral retinopathy on ophthalmoscopic examination. These findings suggest that vigabatrin-associated retinal changes may progress with length of treatment.

The prevalence and course of retinal cone system changes in patients treated with vigabatrin remains to be determined, particularly in patients without visual complaints. The patients with retinal dysfunction reported in this paper are 4 of 38 patients currently being treated with vigabatrin at our center. The remaining 34 visually asymptomatic patients have not yet had a neuro-ophthalmologic examination or ERG to determine if they have asymptomatic clinical or electrophysiologic evidence of retinal dysfunction. Vigabatrin has been used in Europe for 12 years and relatively few patients are reported with the severe visual field constrictions our one patient experienced. This suggests that severe visual complications with vigabatrin are not common. Our experience, however, shows that less severe visual symptoms may occur more frequently than previously recognized.

Our patients' visual symptoms may reflect selective vulnerability to relatively common GABA-mediated physiologic changes in retina. This would be consistent with a recent report that vigabatrin alters visual perception(increased tritanopia thresholds and critical flicker fusion) acutely in normal volunteer patients²⁹ (patients with epilepsy treated chronically with vigabatrin and other anticonvulsants had complex, less-specific alterations in visual perception).³⁰ Additional evidence of this possibility is that symptoms of blurred vision in one patient (Case 2) improved as her vigabatrin was tapered from 2 to 0.5 g/day and then discontinued. She is scheduled for follow-up ERG.

The vigabatrin-associated retinal changes may be dose-dependent; therefore, it may be possible to continue vigabatrin at reduced doses, as was verified in one patient by Wong.¹² Careful follow-up of patients would be necessary to titrate the vigabatrin dosage needed for clinical effect within the narrow 2 to 5 g/day dose range while reducing or eliminating retinal effects. Vigabatrin is an effective anticonvulsant, and two of our patients have extremely severe epilepsy that improved only with vigabatrin. For some patients, risks from uncontrolled seizures may outweigh risks to retina, particularly because visual symptoms are relatively mild and appear to progress slowly. One of the patients with extremely severe epilepsy and visual field constriction (Case 1) continues vigabatrin treatment at a reduced dose with ophthalmology and ERG examinations scheduled at 6-month intervals to screen for injury progression.