Vagus nerve stimulation (VNS) has demonstrated a significant anticonvulsant effect in preclinical studies, in pilot studies in humans, and in the acute phase of a multicenter, double-blinded, randomized study. After completion of a 14-week, blinded, randomized study, with 31 receiving high (therapeutic) VNS and 36 receiving low (less or noneffective) VNS, 67 patients elected to continue in an open extension phase. During the extension phase, all 67 patients received high VNS. Seizure frequency during the 3-month treatment blocks was compared with a 12-week baseline. For both groups, all periods of high VNS demonstrated a significant decrease in seizure frequency (p < 0.01 level) as compared with baseline. For the 16-18-month period of VNS, data were available for 26 of the 31 patients randomized to high VNS. This group achieved a 52.0% mean seizure frequency percentage reduction as compared with baseline. For those converted from low to high VNS, data were available for 24 of the 36 patients at the 16-18-month time period. This group reported a mean seizure frequency percentage reduction of 38.1% as compared with baseline. No significant change in the safety/side effect profile was reported during long-term follow-up. The previously reported side effects of hoarseness/voice change, coughing, and paresthesia (sensation in neck and jaw) continued to occur during VNS. These side effects were well tolerated. During the follow-up period, 1 patient died of thrombotic thrombocytopenic purpura (TTP) and 5 patients discontinued treatment because of unsatisfactory efficacy.