This double-blind study was conducted to evaluate the efficacy and safety of gabapentin 1200 mg/ day and 1800 mg/day (t.i.d.) compared to placebo as an adjunctive therapy in patients with refractory epilepsy. Patients were included when they had partial seizures at least eight times during a 12-week baseline period despite treatment with one to two antiepileptic drugs. After baseline, eligible patients were randomized to gabapentin 1200 mg/day, 1800 mg/day, or placebo for 12-week treatment. The primary end-point, response ratio, was derived from seizure frequencies during treatment and baseline period based upon the seizure daily record by a patient. Of the 209 randomized patients, 86 received gabapentin 1200 mg/day, 41 received gabapentin 1800 mg/day, and 82 received placebo. A statistically significant difference was found between each of the two gabapentin groups and placebo for the primary efficacy end-point, response ratio (P < 0.005) with definite doseresponse (P < 0.001). More gabapentin patients reported moderate to marked improvement in seizure frequency and intensity/duration of each seizure than placebo patients. Treatment-related adverse events were reported by approximately 65% of patients receiving gabapentin compared to approximately 46% of patients receiving placebo; somnolence and dizziness were the most common events. Gabapentin 1200 mg/day and 1800 mg/day significantly reduced the frequency of refractory seizures compared to placebo. Favorable tolerability of gabapentin was confirmed also in a Japanese population, consistent with previous global studies.