Background: Vagus nerve stimulation (VNS) is a form of neuromodulation that has been used to treat individuals with drug-resistant epilepsy (DRE). Although many retrospective studies have explored the predictors of VNS treatment efficacy, the data have been inconsistent, and no VNS studies to date have used the 2017 ILAE epilepsy classification system. Methods: In this single-center retrospective study, we reviewed data from 77 patients who underwent VNS implantation between 2011 and 2019. Data regarding patient gender, epilepsy duration, age at implantation, numbers of antiepileptic drugs, MRI findings, history of craniotomy, epilepsy classification, stimulation current intensity, stimulation duration, and responder status (responders were those with ≥ 50 % seizure frequency reduction) were analyzed via univariate analysis or receiver operating characteristic (ROC) curves to select possible related factors and classification cutoffs. Then, binary logistic regression analysis was used to identify the variables that significantly predicted responder status. Results: In the inclusion group, the mean stimulation duration, mean seizure reduction, and responder rate was 47.2 months, 42.6 %, and 50.6 %, respectively. The univariate analysis indicated that seizure types and MRI findings might be related categorical variables (P < 0.05). After performing binary logistic regression and creating ROC curves for the continuous variables, epilepsy duration was chosen as a related variable, with 12.5 years as the classification cutoff (P = 0.002). A binary logistic regression of risk factors showed that epilepsy duration was significantly related to responder status (P < 0.05). Conclusion: Epilepsy duration is an independent predictor for responders to VNS. VNS is more likely to be efficacious in those with an epilepsy duration < 12.5 years, especially those with a duration of 5–12.5 years. © 2020 Elsevier B.V.