Objective: The purpose of this multicenter, add-on, double-blind, randomized, active-control study was to compare the efficacy and safety of presumably therapeutic (high) vagus nerve stimulation with less (low) stimulation. Background: Chronic intermittent left vagus nerve stimulation has been shown in animal models and in preliminary clinical trials to suppress the occurrence of seizures. Methods: Patients had at least six partial-onset seizures over 30 days involving complex partial or secondarily generalized seizures. Concurrent antiepileptic drugs were unaltered. After a 3-month baseline, patients were surgically implanted with stimulating leads coiled around the left vagus nerve and connected to an infraclavicular subcutaneous programmable pacemaker-like generator. After randomization, device initiation, and a 2-week ramp-up period, patients were assessed for seizure counts and safety over 3 months. The primary efficacy variable was the percentage change in total seizure frequency compared with baseline. Results: Patients receiving high stimulation (94 patients, ages 13 to 54 years) had an average 28% reduction in total seizure frequency compared with a 15% reduction in the low stimulation group (102 patients, ages 15 to 60 year; p = 0.04). The high-stimulation group also had greater improvements on global evaluation scores, as rated by a blinded interviewer and the patient. High stimulation was associated with more voice alteration and dyspnea. No changes in physiologic indicators of gastric, cardiac, or pulmonary functions occurred. Conclusions: Vagus nerve stimulation is an effective and safe adjunctive treatment for patients with refractory partial-onset seizures. It represents the advent of a new, nonpharmacologic treatment for epilepsy.