Vigabatrin and Epilepsy: Lessons Learned


Address correspondence and reprint requests to Dr. John M. Wild, Cardiff School of Optometry and Vision Science, Cardiff University, Redwood Building, King Edward VII Avenue, Cathays Park, Cardiff CF10 3NB, United Kingdom. E-mail:


Summary: Purpose: The risk factors for visual field loss attributable to vigabatrin (VAVFL) are equivocal. This multinational, prospective, observational study aimed to clarify the principal/major factors for VAVFL.

Methods: Interim analysis of three groups with refractory partial epilepsy, stratified by age (8–12 years; >12 years) and exposure to vigabatrin (VGB). Group I comprised participants treated with VGB for ≥6 months, Group II participants previously treated with VGB for ≥6 months who had discontinued the drug for ≥6 months and Group III those never treated with VGB. Perimetry was undertaken at least every six months, for up to 36 months; results were evaluated masked to drug exposure.

Results: Based upon 563 participants in the locked data set, 432 yielded one or more Conclusive visual field examinations. For Group I, the frequency of VAVFL at the last Conclusive examination was 10/32 (31.2%) for those aged 8–12 years and 52/125 (41.6%) for those aged >12 years. For Group II, the proportions were 4/39 (10.3%) and 31/129 (24.0%). No cases resembling VAVFL manifested in Group III. VAVFL was associated with duration of VGB therapy (Odds ratio [OR] 14.2; 95% CI 5.0 to 40.5); mean dose of VGB (OR 8.5; 95% CI 2.2 to 33.2); and male gender (OR 2.1; 95% CI 1.2 to 3.7). VAVFL was more common with static than kinetic perimetry (OR 2.3, 95% CI 1.3 to 4.2).

Conclusions: The therapeutic benefit of VGB is counteracted by the progressive accrual of the risk of VAVFL with continued exposure and with increase in mean dose.

Vigabtrin (VGB) is a second generation antiepileptic drug (AED) which is used as monotherapy for infantile spasms (Chiron et al., 1997; Appleton et al., 1999; Elterman et al., 2002; Hancock et al., 2003; Mackay et al., 2004) and as adjunctive therapy for adults with partial epilepsy which is refractory to other AEDs (Ben-Menachem et al., 1989; Riekkinen et al., 1989; Kälviäinen et al., 1995). VGB acts by inhibiting GABA-transaminase and increasing the synaptic concentration of GABA in the brain (Schechter 1989).

VGB is associated with a bilateral concentric constriction of the visual field (VAVFL) (Eke et al., 1997; Krauss et al., 1998; Arndt et al., 1999; Daneshvar et al., 1999; Kälviäinen et al., 1999; Lawden et al., 1999; Miller et al., 1999; Wild et al., 1999) which is often more pronounced nasally than temporally (Lawden et al., 1999; Wild et al., 1999). The estimated prevalence of VAVFL in retrospective cohort studies ranges from 14% to 92% (Stefan et al., 2000; Besch et al., 2002). The risk factors for VAVFL are equivocal; some association has been identified with male gender (Wild et al., 1999; Hardus et al., 2000, 2001; Newman et al., 2002; Nicolson et al., 2002), cumulative dose of VGB (Miller et al., 1999; Manuchehri et al., 2000; Hardus et al., 2001; Nousiainen et al., 2001) and duration of VGB therapy (Miller et al., 1999; Manuchehri et al., 2000; Hardus et al., 2001; Nousiainen et al., 2001).

The disparity in the prevalence estimates of VAVFL, and the equivocal nature of the risk factors, are likely to arise from many causes including the relatively small sizes of the cohorts; the retrospective character of many studies; the sampling bias arising from the asymptomatic nature of the defect; the diversity of the perimetric techniques; and unfamiliarity, amongst clinicians, of the characteristics of the field loss.

The objectives of the study were to describe the frequency of VAVFL in a large and prospectively assessed cohort of adults/adolescents and children with refractory partial epilepsy and to determine risk factors for the field loss.



The study was initially designed as a prospective, multi-centre, observational study of patients in France with refractory partial epilepsy who had been treated with AEDs for at least one year. Participants were stratified by age (8–12 years; >12 years) into one of three groups (Fig. 1). Group I comprised randomly selected individuals treated with VGB (including monotherapy) for ≥6 months. Group II comprised randomly selected individuals previously treated with VGB (including monotherapy) for ≥6 months, but who had discontinued VGB for ≥6 months. Group III comprised randomly selected individuals with no previous exposure to VGB. Visual field examination in all three groups was to be undertaken every six months over 36 months. Any clinically indicated change in medication by the treating physician, including VGB, was permitted during follow-up. Participants were to be enrolled in a ratio of 4:3:3 across Groups I, II and III, respectively, with twice the number of participants aged >12 years than 8–12 years. Based upon a cumulative incidence of VAVFL of 28% (95% CI 20 to 36%) and a withdrawal rate of 15%, 1060 patients were to be enrolled. The precision associated with the prevalence estimate was ±8.4% in the smallest subgroup and ±5.3% in the largest subgroup. Ethical Committee approval had been obtained from each centre and the first patient was enrolled in March 1999.

Figure 1.

The schematic representation of the initial design of the study, showing the duration of follow-up and the frequency of perimetry, and the modifications arising from the various protocol amendments.

In October 1998, VGB was referred to the Committee for Proprietary Medicinal Products (CPMP; now the CHMP), under the European Union Article 12 Procedure, for an assessment of the risk/benefit. In May 1999, the study was extended to include centres from South Korea, Italy, Spain and Australia.

The study underwent three protocol amendments; the first was in accordance with the requirements of the CPMP, the remaining two were endorsed by the CPMP. The amendments (Fig. 1), variously, reduced the interval between visual field examinations; reduced the time in the study for some participants; created the opportunity for an incidence study of VAVFL; removed the requirement for random selection of participants (from July 2001 for those aged 8–12 years and from August 2002 for those aged greater than 12 years); terminated recruitment of participants; and enabled all participants remaining in the study to be followed for 36 months.


The baseline visit comprised verification of conformity to inclusion criteria, acquisition of informed consent, neurological and ophthalmological examinations and perimetry. At each follow-up visit, neurological examination and perimetry were undertaken.

Visual field evaluation

The results of the visual field examination for each participant at each visit were evaluated by an external investigator (JMW), masked to the participant's drug exposure. The fields were first classified as either Conclusive or Inconclusive. Conclusive results were further sub-categorized as either Normal or Abnormal. An Abnormal field, consistent with the condition recorded on the Case Report Form, was designated as field loss with an Attributable Aetiology. An Abnormal visual field, for which an appropriate cause was not identified, was designated as field loss of Non-attributable Aetiology. The latter category was further classified into VAVFL or Other Type of field loss using standard criteria (Harrington and Drake, 1990; Wild et al., 1999). No attempt was made to classify the severity of field loss. Inconclusive results were subcategorized as either Equivocal or as Technically Limited. An Equivocal result was assigned when the test result was uncertain, although the perimetric technique was appropriate, the perimetry had been adequately performed and the participant's response was deemed to be satisfactory / reliable. A Technically Limited result was assigned if the perimetric technique was inappropriate for the detection of VAVFL, the perimetry had been inadequately performed, or the participant's response was deemed to be unsatisfactory/unreliable.


The frequency for each of the possible results from the last Conclusive visual field examination was determined as a function of Group designated at the last visual field examination. Stepwise multivariate logistic regression was then undertaken to model the association between the presence of VAVFL at the last Conclusive visual field examination and each of 24 independent variables embracing demographical, medical, neurological and ophthalmological parameters (Appendix Table A1). The distributions of each of the independent variables were compared between participants with and without VAVFL using the Fisher exact test for categorical variables and the Kruskal Wallis test for continuous variables. Variables exhibiting a difference in the distribution at p < 0.10 and with <10% missing data were proposed for the logistic regression model. The stepwise selection method operated at p < 0.10 and included first order interactions.

The multicenter study was organized and funded by Aventis Pharma, now Sanofi-Aventis, Paris, France. The authors had full access to all of the data in the study; and take full responsibility for the integrity of the data, the accuracy of the analysis, and the opinions expressed in the manuscript. Each author was involved in the design, analysis and writing of the manuscript, approved the final version and had final responsibility for its submission for publication.


The cohort comprises 735 participants from 46 centres. The majority of centers (24) and patients (437) are from France. The current analysis is based upon 563 patients in the dataset locked at a cutoff date of February 2, 2005.

The Group and age categorization, at enrolment, of the 563 participants is given in Table 1. Eight patients in Group III had commenced VGB therapy soon after the baseline assessment as part of the incidence study. Seventy-nine of the 201 participants in Group I discontinued VGB during the course of the study (14 aged 8–12 years, 65 aged > 12 years). Therefore, the results for participants in Group I are given for those who remained on VGB (Group IA) and for those who discontinued VGB (Group 1B). At the time of the last Conclusive examination, those with VAVFL in Group IB had ceased VGB therapy 15.4 months (SD 9.3 months) previously and those without VAVFL 19.8 months (SD 8.8) previously.

Table 1. The Group and age categorization of the 563 participants in the locked data set. The numbers in parenthesis for Group III indicate those who commenced VGB therapy at baseline as part of the envisaged incidence arm
 Group IA (receiving VGB at last visit)Group IB (withdrawn from VGB during study)Group II (withdrawn from VGB prior to onset of study)Group III (no exposure to VGB)Total
Age 8–12 yrsAge >12 yrsAge 8–12 yrsAge >12 yrsAge 8–12 yrsAge >12 yrsAge 8–12 yrsAge >12 yrs
Completed19461144401179 (+1)32 (+1)318 (+2)
Withdrew prematurely1641 321194421 (+4)72 (+2)237 (+6)
Total358714655916130 (+5)104 (+3)555 (+8)

Two hundred forty-three (43%) of the 563 participants withdrew from the study prematurely, i.e., prior to completion of any one of the required follow-ups. The cumulative dose (p < 0.001) and duration of VGB (p < 0.001), length of participation in the study (p < 0.001) and number of visual field examinations (p < 0.001) of those who withdrew from the study were all significantly less than those who remained in the study; no statistically significant differences were found for the comparisons of the 11 other demographical variables.

The distribution of the visual field results for the 563 participants are given in Fig. 2. Of the 563 participants, 432 (76.7%) yielded a Conclusive result at one or more of the visual field examinations. One hundred and twelve of these 432 participants generated one Conclusive result, only.

Figure 2.

The distribution of the visual field results for the 563 individuals in the locked data set (VAVFL indicates vigabatrin-attributed visual field loss).

The summary measures in each Group, at baseline, for gender, age, cumulative dose of VGB at enrolment, and duration of VGB therapy at enrolment for the 432 participants yielding Conclusive visual field results are given in Table 2. The differences in the distributions of each variable between those with a Conclusive visual field result and those with either an Equivocal, or a mixed Equivocal/Technically Limited result, at all of the examinations, respectively, were not statistically significantly different. However, participants manifesting a Technically Limited outcome at all of the examinations were younger by 8.2 years than those manifesting a Conclusive outcome (mean 27.5 years SD 15.0 versus mean 19.3 years SD 10.4) (p < 0.001) and were taking a lower daily dose of VGB at baseline (mean 2.0 gm SD 0.75 versus mean 1.4 gm SD 0.4) (p < 0.001).

Table 2. The summary statistics at baseline for the distributions of gender, age, cumulative dose of VGB, and duration of VGB therapy, as a function of Group for the 432 participants producing one or more Conclusive visual field examinations
 Group IA (receiving VGB at last visit)Group IB (withdrawn from VGB during study)Group II (withdrawn from VGB prior to onset of study)Group III (no exposure to VGB)Total
Age 8–12 yrsAge >12 yrsAge 8–12 yrsAge >12 yrsAge 8–12 yrsAge >12 yrsAge 8–12 yrsAge >12 yrs
  1. Seven of the eight participants enrolled in the incidence arm in Group III yielded one or more Conclusive visual field examinations. For clarity, the summary data for these seven participants are not displayed in Tables 2 to 4; however, they are included in the data summarized in the Total column of each of these tables.

% Males47.6%51.4%63.6%45.5%61.5%51.2%71.4%48.1%51.9%
Age (yr) mean ± SD [range]10.6 ± 1.432.4 ± 13.110.6 ± 1.332.6 ± 15.310.2 ± 1.332.4 ± 13.310.6 ± 1.232.8 ± 12.527.5±15.0
Cumulative dose of VGB (kg) mean ± SD [range]2.2 ± 1.53.1 ± 2.02.2 ± 1.73.2 ± 2.40.9 ± 1.12.0 ± 2.02.3 ± 2.1
[0.3–6.4][0.0–9.1][0.7–6.5][0.0–11.8][0.0–4.7][0.0–11.8] [0.0–11.8]
Duration of VGB therapy (years) mean ± SD [range]4.2 ± 2.44.0 ± 2.34.1 ± 3.04.1 ± 2.81.8 ± 1.82.4 ± 2.13.2 ± 2.4
[0.5–8.7][0.0–8.4][1.2–11.9][0.4–11.2][0.1–6.4][0.0–10.8] [0.0–11.9]

Of the 79 participants in Group I who discontinued VGB during the course of the study, 66 (11 aged 8–12 years, 55 aged >12 years) had Conclusive results at one or more of the visual field examinations.

Of the 8 individuals (5 aged 8–12 years), allocated to Group III, who commenced VGB therapy, 7 yielded at least one Conclusive visual field result. The mean duration of VGB therapy at the last Conclusive examination in this subgroup was 9.1 months (SD 8.3; range 2.4–23.6) and the mean cumulative dose was 0.47 kg (SD 0.1; range 0.0–1.3). None of these 7 patients developed VAVFL over the course of the study.

The frequency of the different types of visual field loss at the last Conclusive outcome to the visual field examination in each group is shown in Table 3. The summary measures of the length of time enrolled in the study; the duration of VGB therapy; the total number of visual field examinations; and the numbers of Conclusive, Equivocal and Technically Limited outcomes; among those individuals in Groups I and II with one or more Conclusive results as a function of the presence or absence of VAVFL at the last Conclusive outcome is shown in Table 4.

Table 3. The frequency (absolute value; percentage; and 95% confidence interval) of the four possible outcomes of the visual field examination, as a function of Group, for the 432 participants, at the last Conclusive visual field examination
Result of visual field examinationGroup IA (receiving VGB at last visit)Group IB (withdrawn from VGB during study)Group II (withdrawn from VGB prior to onset of study)Group III (no exposure to VGB)Total 432
Age 8–12 yrs N=21Age >12 yrs N=70Age 8–12 yrs N=11Age >12 yrs N=55Age 8–12 yrs N=39Age >12 yrs N=129Age 8–12 yrs N=21Age >12 yrs N=79
  1. The frequency of VAVFL (vigabatrin-attributed visual field loss) is shown in bold.

Normal14; 66.7%;41; 58.6%;7; 63.6%;20; 36.4%;34; 87.2%;80; 62.0%;20; 95.2%;73; 92.4%;296; 68.5%;
Identified etiology1; 4.8%;5; 7.1%;06; 10.9%;1; 2.6%;8; 6.2%;04; 5.1%;25; 5.8%;
Nonidentified etiology: VAVFL6; 28.6%;23; 32.9%;4; 36.4%;29; 52.7%;4; 10.3%;31; 24.0%;0; 0.0%;0; 0.0%;97; 22.5%;
Nonidentified etiology: other pattern01; 1.4%;000/4610; 7.8%;1; 4.8%;2; 2.5%;14; 3.2%;
Table 4. The summary measures of the length of time enrolled in the study, the duration of VGB therapy, the total number of visual field examinations, and the number of Conclusive, Equivocal and Technically Limited results, for the 432 individuals in Groups I, II and III who yielded one or more Conclusive visual field results
 Group IA (receiving VGB at last visit)Group IB (withdrawn from VGB during study)Group II (withdrawn from VGB prior to onset of study)Group IIITotal
8–12 yr of ageN=6N=15N=4N=7N=4N=35N=2197
 Duration in study (yr), mean ± SD [range]2.6 ± 0.51.6 ± 1.22.8 ± 0.62.6 ± 0.51.3± 1.22.0 ± 1.11.8 ± 1.21.9 ± 1.1
 Duration of VGB therapy (yr), mean ± SD [range]7.1 ± 2.55.6 ± 2.26.7 ± 4.24.3 ± 1.72.5 ± 1.51.7 ± 1.8-3.38 ± 3.0
[3.0–11.0][2.9–10.4][2.0–12.0][2.5–7.6][0.9–4.3][0.1–6.4] [0.0–12.0]
 Number of visual field examinations, mean ± SD [range]5.7 ± 1.53.7 ± 2.35.8 ± 1.56.0 ± 1.22.8 ± 1.74.4 ± 2.14.4 ± 2.64.5 ± 2.2
 Number of Conclusive results, mean ± SD [range]3.3 ± 2.32.9 ± 2.04.0 ± 0.83.0 ± 1.72.5 ± 1.73.0 ± 1.83.5 ± 2.23.2 ± 1.9
 Number of Equivocal results, mean ± SD (median) [range]2.3 ± 2.30.2 ± 0.61.0 ± 0.82.4 ± 1.60.0 ± 0.00.9 ± 1.20.4 ± 0.70.8 ± 1.3
 Number of Technically Limited results, mean ± SD [range]0.0 ± 0.00.7 ± 1.00.8 ± 1.00.6 ± 1.00.3 ± 0.50.6 ± 0.90.5 ± 1.30.5 ± 0.9
>12 yr of ageN=23N=47N=29N=26N=31N=98N=79335
 Duration in study (yr), mean ± SD [range]2.1 ± 1.11.6 ± 1.22.6 ± 0.52.4 ± 0.72.2± 0.81.8 ± 1.11.8 ± 1.22.0 ± 1.1
 Duration of VGB therapy (yr), mean ± SD [range]6.3 ± 2.55.7 ± 2.76.3 ± 3.13.9 ± 2.63.2 ± 1.92.2 ± 2.1-3.96 ± 3.0
[1.7–11.5][1.3–11.1][1.4–12.5][0.8–11.4][0.8–7.6][0.0–10.7] [0.0–12.5]
 Number of visual field examinations, mean ± SD [range]4.7 ± 2.63.9 ± 2.45.7 ± 1.35.2 ± 1.75.2 ± 1.84.6 ± 2.34.3 ± 2.54.6 ± 2.3
 Number of Conclusive results, mean ± SD [range]3.3 ± 2.22.9 ± 2.03.9 ± 1.84.0 ± 2.03.5 ± 2.23.5 ± 2.13.4 ± 2.13.5 ± 2.1
 Number of Equivocal results, mean ± SD [range]0.9 ± 1.30.7 ± 1.11.5 ± 1.90.7 ± 0.81.3 ± 1.80.7 ± 1.10.5 ± 1.10.8 ± 1.3
 Number of Technically Limited results, mean ± SD [range]0.5 ± 0.70.3 ± 0.80.3 ± 0.60.5 ± 0.90.4 ± 0.60.3 ± 0.80.3 ± 0.70.4 ± 0.7

Sixteen participants (10.2%) in Group I (6 aged 8–12 years, 10 aged >12 years) appeared to have developed VAVFL during the course of the study (i.e., they each exhibited a Normal outcome followed by one or more VAVFL outcomes). In contrast, only 6 participants (3.6%) in Group II (one aged 8–12 years, 5 aged >12 years) were deemed to have developed VAVFL during the course of the study. The fields of one participant with VAVFL, who discontinued VGB during the study, reverted to normality during follow-up. However, subsequent re-evaluation of the fields for this individual suggested that the reversal of the field loss was more likely to have been due to the perimetric learning effect.

The results of the model generated by the multivariate logistic regression, at the last Conclusive examination, are given in Table 5. The risk of VAVFL progressively increased with increase in duration of therapy and with increase in mean dose. Those treated with VGB for more than 5 years were approximately 14.2 times more likely to exhibit VAVFL than those receiving VGB for less than one year. Similarly, those who had received a mean dose of 3 kg or more were 8.5 times more likely to exhibit VAVFL than those who had received less than 1 kg. Male gender was approximately 2.1 times more likely to be associated with VAVFL than female gender. The identification of VAVFL was also facilitated by a visual field examination comprising static threshold perimetry for the central field (i.e., within 30° from fixation) combined with two level suprathreshold perimetry for the periphery (i.e., beyond 30°) compared to kinetic perimetry with the Goldmann perimeter.

Table 5. Factors influencing the probability of VAVFL at the last Conclusive visual field examination derived by stepwise multivariate logistic regression analysis
Effectp-valueComparisonOdds Ratio95% CI
Duration of VGB therapy< 0.0001>5 yr vs. < 1 yr14.25.0–40.5
> 3 yr to < 5 yr vs. < 1yr8.26.3–24.0
> 1 yr to < 3 yr vs. < 1yr4.03.4–12.2
Mean dose of VGB0.0063>3,000 mg vs. < 1,000 mg8.52.1–33.2
>2,000 mg to ≤3,000 mg vs. < 1,000 mg4.91.7–14.0
>1,000 mg to ≤2,000 mg vs. < 1,000 mg4.91.8–12.9
Gender0.0060Male vs. Female2.11.2–3.7
Perimetric technique0.0043Static perimetry vs. Peripheral kinetic perimetry8.31.0–66.7
Static perimetry vs. Peripheral and Central kinetic perimetry2.31.3–4.3


This study is the largest prospective evaluation of the frequency of VAVFL. A Conclusive visual field result was obtained in approximately three-quarters of the participants and, among those with Conclusive results, VAVFL was identified, overall, in 29.8% (95% CI 24.9% to 35.1%) of cases associated with ongoing or previous VGB exposure. Duration of VGB therapy, mean dose of VGB therapy, and male gender, were identified as risk factors for VAVFL. There was a progressive accrual of the risk of VAVFL with continued VGB exposure and with increase in mean dose.

The overall frequency of VAVFL was in the low to middle part of the range of previous prevalence estimates [Stefan et al., 2000; Kälviäinen and Nousiainen, 2001; Besch et al., 2002). Investigators were requested to enrol participants at random and to disregard any knowledge about visual field status; however, particularly due to the abandonment of the randomization procedure, the possibility of selection bias cannot be fully excluded and, therefore, the observed frequencies may not reflect true prevalence rates.

The analysis was based upon the last Conclusive examination. Alternative options were to analyse the results at baseline or at the time of the first Conclusive examination. Since most participants had never undergone perimetry, analysis at baseline would have excluded a substantial number of individuals who had yielded an Inconclusive or Technically Limited result at the first examination. Additionally, an analysis at baseline, or at the first Conclusive examination, would not have accounted for the potential subsequent development of VAVFL in Group I participants. Analysis based upon the last Conclusive examination accounted for participants who developed VAVFL during follow-up, maximized the number of participants with informative results and provided a standardized end-point across the three groups. However, such analysis is not related to a specific time point during follow-up. An analysis based upon the time to detection of the VAVFL was not feasible since no equivalent time point existed for those without VAVFL.

It has been suggested that the frequency of VAVFL is greater in adults than in children (Vanhatalo et al., 2002). Whilst there was a clear trend for the frequency of VAVFL to be lower in those aged 8–12 years than in the older participants (Table 3), the statistical power of these comparisons was limited by the small size of the subgroups. Moreover, in the group who were receiving VGB at the last Conclusive examination, the proportion of participants with VAVFL was only marginally lower for the younger age group (28.6%) than for the older age group (32.9%), suggesting that pediatric age, at least within the 8 to 12 years range, confers little protection against the visual toxicity of VGB. In addition, age was not identified as a risk factor for VAVFL.

The associations of male gender and dose and duration of VGB therapy with VAVFL are in agreement with previous findings (Miller et al., 1999; Wild et al., 1999; Hardus et al., 2000; Manuchehri et al., 2000; Hardus et al., 2001; Nousiainen et al., 2001; Nicolson et al., 2002; Ardnt et al., 2002).

Due to the design of the study, participants with VAVFL were treated with VGB for between 11.4 months and 15.2 years prior to detection of the field loss. Similarly, in clinical practice, the time to detection of VAVFL is usually not the time to onset due to the insidious and asymptomatic nature of VAVFL and to the inability to detect the disorder before the age of at least 8 years.

The absence of apparent VAVFL in patients never exposed to VGB (Group III) indicates the robustness of the characteristics of VAVFL and also the lack of association of this, or any other, type of field loss with other AEDs.

One of the objectives of the study was to evaluate the incidence of VAVFL in patients commencing de novo VGB therapy. The drastic reduction in the prescribing of VGB after the discovery of the associated visual toxicity prevented the enrolment of a sizeable cohort for this purpose.

The present analysis does not address the issue of deterioration of VAVFL during follow-up. Any longitudinal assessment was made difficult by the fact that the cohorts contained different numbers of participants with different durations of follow-up and examination intervals, with a resultant loss of statistical power. The loss to follow-up was greatest in Groups IA (46.7%) and III (70.9%) compared to Groups IB (30.3%) and II (28.6%), and the impact of this was more pronounced for those participants with an Inconclusive outcome to the previous visual field examinations. The relatively high frequency of VAVFL among patients withdrawn from VGB during the study (Group IB) is consistent with the notion that VAVFL is an irreversible condition (Nousiainen et al., 2001). Although there was a trend amongst those who withdrew from VGB prior to the study to exhibit a lower frequency of VAVFL than patients remaining on VGB, the difference may simply reflect the shorter duration of exposure to VGB in the former group.

VAVFL was approximately 2.5 more common with automated static perimetry than with manual kinetic perimetry; such a finding is not surprising given that static perimetry is more sensitive than kinetic perimetry for the detection of field loss, in general (Wild et al., 1984).

Approximately 25% of patients with partial refractory epilepsy undergoing repeated visual field examinations were unable to produce a Conclusive result at any of their study visits. This figure is similar to other estimates for patients with epilepsy (Harding et al., 2000) and represents a major problem for the assessment of such individuals. The greatest number of participants exhibiting an Inconclusive result originated from Group IA (28.6%) compared to Groups IB (15.1%), II (16.8%) and III (21.0%). The number of Inconclusive results was proportionately greater in Group IA for the younger age group (34.3%) than for the older age group (17.2%), but this age difference did not emerge in the other subgroups. Alternative techniques for the diagnosis of VAVFL have been suggested such as the wide-field multifocal electroretinogram McDonagh et al., 2003), the 30Hz flicker electroretinogram (Harding et al., 2000; Westall et al., 2002), the H-stimulus which generates simultaneously recorded central and peripheral visually evoked potentials (Harding et al., 2002), and ocular imaging of the retinal nerve fibre layer (Wild et al. 2006).

In conclusion, at least one third of participants aged 8 years and older exposed to VGB exhibited a characteristic type of visual field loss that was not encountered in unexposed participants. The confirmation of a number of risk factors provides additional information for a correct appraisal of the risk/benefit ratio in children and adults with refractory partial epilepsy in whom use of VGB is contemplated. The study did not include infants with infantile spasms, who represent the current primary indication of VGB, and the findings cannot be extrapolated to the risk assessment in this specific age group. Equally so, the implications from the study for the alternative use of the drug, such as the short-term use of vigabatrin in the treatment of substance abuse (Fechtner et al., 2006), are also unknown.

Conflict of interest

JMW, JB, CC, US and EP have served as paid consultants to Aventis or Sanofi-Aventis. US and EP have also received grants from Aventis, or Sanofi-Aventis, for other research or activities not reported in this article.


Karine Ostrowsky, Lyon, France; Philippe Kahane, Grenoble, France; Simone Garrel, Grenoble, France; Laurent Vercueil, Grenoble, France; François Mauguière, Lyon, France; Hervé Vespignani, Nancy, France; Jean-Luc Schaff, Flavigny-sur-Moselle, France; Bernard Echenne, Montpellier, France; Michel Baldy-Moulinier, Montpellier, France; Pierre Thomas, Nice, France; Jacques Motte, Reims, France; Pascal Sabouraud, Reims, France; Josette Mancini, Marseille, France; Pierre Genton, Marseille, France; Dominique Broglin, Marseille, France; Philippe Derambure, Lille, France; Louis Vallée, Lille, France; Arnaud Biraben, Rennes, France; Delphine Taussig, Rennes, France; Patrick Contis, Colomiers, France; Cécile Marchal, Bordeaux, France; Jean-Michel Pedespan, Bordeaux, France; Alexis Arzimanoglou, Paris, France; Catherine Chiron, Paris, France; Olivier Dulac, Paris, France; Vincent Desportes, Paris, France; Marie-Anne Barthez, Tours, France; Pierre Castelnau, Tours, France; Sylvie Nguyen The Tich, Nantes, France. SangAhm Lee, Seoul, South Korea; ByungIn Lee, Seoul, South Korea; SangKun Lee, Seoul, South Korea; Sung Hwan Kim, Suwon, South Korea; Seung Bong Hong, Seoul, South Korea; Luigi Specchio, Bari, Italy; Carlo Andrea Galimberti, Pavia, Italy; Giuliano Avanzini, Milan, Italy; Dante Besana, Alessandria, Italy; Carlo Cianchetti, Cagliari, Italy; Federico Vigevano; Rome, Italy; Pier Gaetano Garofalo, Vicenza, Italy; Salvatore Striano, Naples, Italy; Ettore Beghi, Monza, Italy; Roberto Biondi, Catania, Italy; Emilio Fernández, Esplugues de Llobregat, Spain; Llibert Padrò, Barcelona, Spain; Jaime Campos-Castello, Madrid, Spain; José-Luis Herranz, Santander, Spain; Juan José Zarranz, Vizcaya, Spain; Sam Berkovic, Melbourne, Australia; Andrew Bleasel, Westmead, Australia; Terence O'Brian, Melbourne, Australia; Mark Cook, Melbourne, Australia; Mathew Kiernan, Sydney, Australia.


Table A1. The independent variables proposed for the stepwise multivariate modeling of VAVFL
DemographicsDuration of antiepileptic drug therapy
Gender< 1yr, 1–< 3yr, 3−< 5yr, > 5yr
CountryDuration of Carbamazepine therapy
< 1 yr, 1-< 3 yr, 3-< 5 yr, >5 yr
History of infantile spasmsSystemic factors
Type of epilepsyAlcohol consumption
 Generalized epilepsyTobacco consumption
 Localization-related epilepsyPresence of hypertension
Type of seizuresPresence of cerebral, cardiac or peripheral vascular ischemia
Number of seizures in the last monthOcular factors
 0,1–2, 3–6, 7–20, >20Visual acuity
Age at onset of epilepsyIris color Light, dark
Duration of epilepsyIntraocular pressure (max)
Vigabatrin therapyCup:disc ratio
Cumulative dosePerimetric Methodology (in decreasing order of sensitivity)
Mean doseRecommended (Central field: static threshold perimetry. Peripheral field: Three Zone age-corrected suprathreshold perimetry)
Other antiepileptic drug therapy (ranked from 1 to 5 in decreasing order of arbitrarily assignedKinetic perimetry: central and peripheral field
magnitude of GABAergic activity)Kinetic perimetry: peripheral field
1. Tiagabine
2. Benzodiazepines, retigabine
3. Valproate, phenobarbitiol, primidone
4. Any other antiepileptic drug
5. None