Optimal candidates for VNS as a treatment for refractory epilepsy have not been identified. In this retrospective two-center study, we used the Engel classification for evaluating seizure outcome, and tried to identify predictive factors for outcome by means of subgroup analysis. The medical records of patients who have been treated with VNS for at least one year at Dartmouth-Hitchcock Medical Center and Ghent University Hospital were evaluated. Seizure frequency outcome was assessed using the Engel classification for the study population as a whole, and for patient subgroups with regard to mental functioning, seizure type, predisposing factors for developing epilepsy, age at time of VNS implantation and epilepsy duration. 189 patients (102M/87F) were included in the study (mean FU: 41 months). 6% had a class I outcome (seizure-free), 13% a class II outcome (almost seizure-free), 49% a class III outcome (worthwhile improvement) and 32% had a class IV outcome (no improvement). When patients were divided into specific subgroups, a statistically significant better outcome was found patients with normal mental functioning (p=0.029). In our series, results for VNS are clearly inferior to resective surgery, but comparable to other treatment modalities for refractory epilepsy. With combined class I and II outcomes around 20%, and another 50% of patients having worthwhile improvement, VNS is a viable alternative when resective surgery is not feasible.