Back

The extent and durability of improvement in depressive symptoms, quality of life, and daily function with three years of vagus nerve stimulation in markedly treatment-resistant depression: A RECOVER study report

Conway, C. R.; Aaronson, S. T.; Rush, A. J.; Lee, Y.-C.; Shy, O.; Bunker, M. T.; Gordon, C.; Riva-Posse, P.; Reeves, K.; George, M. S.; Zajecka, J.; Nahas, Z.; Dunner, D. L.; Figee, M.; Mickey, B. J.; Allen, R. M.; Bohnenkamp, D.; Kriedt, C. L.; Hristidis, V. C.; Quevedo, J.; Zorumski, C. F.; Macaluso, M.; Duffy, W.; Sheline, Y.; Alva, G.; Cusin, C.; Bennett, J. I.; Tran, Q.; McIntyre, R. S.; McAllister-Williams, R. H.; Sackeim, H. A.

2026-08-04 psychiatry and clinical psychology
10.64898/2026.07.31.26358854 medRxiv
Show abstract

Background: Management of markedly treatment-resistant depression is characterized by low initial benefit and poor durability. Treatments with sustained benefits are needed. This report summarizes clinical outcomes and durability of both the active treatment arm (Early-Active) and the initially sham treatment arm (Delayed-Active) over the second and third years of the multicenter, prospective, implanted vagus nerve stimulation (VNS) RECOVER trial. Methods: A total of 436 participants (N=221 Early-Active and N=215 Delayed-Active, 65.8% females) were studied. Within each group, analyses of change in benefit (depressive symptoms, clinical impression, quality of life [QoL], daily function, and a composite measure) occurred with assessments at 12, 18, 24, 30, and 36 months. Two within-group methods of benefit appraisal over time were conducted: 1) a comparison of the degree of benefit change, and 2) a comparison of proportions of participants achieving benefit categories. Additionally, the degree of durability of benefit was assessed, comparing 12-24 months, 24-36 months, and 12-36 months. Results: For the Early-Active group: The 12-24-month and 12-36-month periods, but not the 24-36-month period, demonstrated significant improvement in benefit categories for depressive symptoms and clinical impression measures. Similarly, statistically significant increases in the proportions of participants achieving benefit during Year 2 for depressive and clinical impression measures occurred. For the Delayed-Active group: The 12-24-month (first exposure of this group to active VNS) and 12-36-month periods, but not the 24-36-month period, demonstrated statistically significant improvements in benefit categories for depressive symptoms and clinical impression measures. Additionally, statistically significant increases in the proportions of participants achieving benefit during Year 2 for depressive symptoms, clinical impression, QoL, daily function, and the composite measure were observed. Averaged across the depressive symptom and clinical impression measures, the Early-Active group demonstrated continued progression to higher benefit categories during Years 2 and 3, whereas the Delayed-Active group was characterized by the emergence of new benefit during Year 2 followed by further progression to higher benefit categories during Year 3. Durability: Robust durability of response was observed for both groups across all time intervals, with a median of 71.1% and 69.0% maintaining or improving benefit from 12 to 36 months for Early-Active and Delayed-Active, respectively. Conclusions: In a highly chronic and markedly resistant depressed sample, active VNS produced benefits that often emerged gradually, sometimes beyond one year after initiation, continued to improve in degree of benefit over time, and were highly durable. The time-associated benefit patterns observed in the sham group (Delayed-Active) closely resembled those of the initially active group (Early-Active) but were delayed by approximately one year, consistent with the delay in therapy activation.

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.