This study characterized the durability of improvement in patients who responded early or late while receiving vagus nerve stimulation (VNS). In both a pilot and pivotal study, patients were identified who had at least a 50% reduction in symptom scores 3 months (early responders) or 12 months (late responders) after starting VNS. Probabilities were determined for maintenance of response at 12-month and 24-month time-points. Consistency of improvement throughout the 24-month study period was evaluated, testing for change in serial depression ratings. In the pilot study, 30.5%, 23.7% and 45.8% were early responders, later responders, and non-responders, respectively. These rates were 14.6%, 19.5%, and 65.9% in the pivotal trial. The potential confound of alterations in antidepressant treatment was examined in the pivotal trial. In the pilot study, 72.2% and 61.1% of early responders (n=18) were responders at 12 and 24 months, respectively; 78.8% of late responders (n=14) were responders at 24 months. In the pivotal trial, of early responders (n=30), 63.3% and 76.7% maintained response at 12 and 24 months, respectively; of late responders (n=40), 65.0% maintained response at 24 months. Early and late responders had fewer changes in medication than non-responders across the pivotal study period. In both studies, analyses of serial depression ratings showed stable improvement in early and late responders. These samples had exceptional levels of chronicity and treatment resistance. Yet patients who showed substantial clinical benefit maintained the improvement at remarkably high rates. This durability of benefit was not attributable to alterations in other treatments.