Background: Use of the antiepileptic drug vigabatrin is associated with an elevated risk of visual field loss. Objective: To determine the frequency of, and risk factors for, vigabatrin-attributed visual field loss (VAVFL) in the setting of a large-scale, multinational, prospective, observational study. Study design: A comparative, open-label, parallel-group, multicentre study. Setting: Hospital outpatient clinics at 46 centres in five countries. Patients: 734 patients with refractory partial epilepsy, divided into three groups and stratified by age (8–12 years; >12 years) and exposure to vigabatrin. Group I comprised patients treated with vigabatrin for ≥6 months. Group II comprised patients previously treated with vigabatrin for ≥6 months who had withdrawn from the drug for ≥6 months. Group III comprised patients never treated with vigabatrin. Patients underwent perimetry at either 4- or 6-month intervals, for up to 36 months. Visual field outcome was evaluated masked to drug exposure. Intervention: Perimetry. Main outcome measure: The visual field outcome at each of four analysis points: (i) at enrolment (i.e. baseline, all patients); (ii) for patients exhibiting a conclusive outcome at the initial visual field examination; (iii) for patients exhibiting at least one conclusive outcome to the visual field examinations; and (iv) at the last conclusive outcome to the visual field examinations. Results: Of the 734 patients, 524 yielded one or more conclusive visual field examinations. For Group I, the frequency of VAVFL at the last conclusive examination was 10/38 (26.3%) for those aged 8–12 years and 65/150 (43.3%) for those aged >12 years. For Group II, the respective frequencies were 7/47 (14.9%) and 37/151 (24.5%). One case resembling VAVFL was present amongst the 186 patients in Group III at the last conclusive examination. The frequency of VAVFL in Groups I and II combined was 20.0% for those aged 8–12 years and 33.9% for those aged >12 years. VAVFL was associated with duration of vigabatrin therapy (odds ratio [OR] up to 15.2; 95% CI 4.4, 51.7), mean daily dose of vigabatrin (OR up to 26.4; 95% CI 2.4, 291.7) and male gender (OR 2.51; 95% CI 1.5, 4.1). VAVFL was more frequently detected with static than with kinetic perimetry (OR up to 0.43; 95% CI 0.24, 0.75). Conclusions: Since the probability of VAVFL is positively associated with treatment duration, careful assessment of the risk-benefit ratio of continuing treatment with vigabatrin is recommended in patients currently receiving this drug. All patients continuing to receive vigabatrin should undergo visual field examination at least every 6 months for the duration of treatment. We recommend two-level (three-zone), gradient-adapted, suprathreshold static perimetry of the peripheral field together with threshold perimetry of the central field out to 30° from fixation. The frequency of ophthalmological and perimetric examinations should be increased in the presence of VAVFL.