Durability of the benefit of vagus nerve stimulation in markedly treatment-resistant major depression: a RECOVER trial report

Abstract

Importance

Greater levels of treatment resistance in major depressive disorder (MDD) are associated with lower rates of initial benefit and higher rates of relapse (lower durability).

Objective

Characterize depressive symptoms, function, and quality of life (QoL) over 24 months of adjunctive vagus nerve stimulation (VNS) in participants with markedly treatment-resistant depression.

Design

Prospective, open-label, single-arm, long-term extension study (RECOVER) conducted from September 2019 to April 2025.

Setting

Outpatient.

Participants

Adults with moderate–severe MDD with ≥ 4 failed antidepressant trials in the current episode, randomized to blinded, adjunctive VNS for 12 months, who subsequently received open-label, adjunctive VNS for 12 additional months (n = 214).

Interventions

Vagus nerve stimulation and concomitant psychotropic medications and interventional psychiatric modalities (electroconvulsive therapy, transcranial magnetic stimulation, and ketamine/esketamine) were characterized over the 12-month extension.

Main Outcomes and Measures

The durability of benefit achieved at 12 months was assessed at 18 and 24 months for depressive symptoms (3 scales), daily function, QoL, a tripartite composite of all 3 domains, and the Clinical Global Impression–Improvement (CGI-I) scale (overall improvement). Loss of benefit and relapse were assessed, along with the emergence of meaningful benefit in participants without benefit at 12 months. Substantial benefit (at least 50% symptom reduction from baseline; CGI-I of 1 or 2; tripartite measures with at least 2 of 3 subscales evidencing benefit) and meaningful benefit thresholds for symptoms (at least 30% reduction from baseline), function, QoL, CGI-I, and the tripartite measure were set a priori.

Results

Most participants with substantial benefit maintained their benefit (18-month median = 78.8%; 24-month median = 79.0% across 5 measures), as did participants with at least meaningful benefit at 12 months (18-month median = 83.1%; 24-month median = 81.3% across 7 measures). Furthermore, many participants with no meaningful benefit at 12 months achieved it at 18 (median = 30.6%) and 24 (median = 37.8%) months. The strong maintenance of benefit was not accounted for by changes in psychotropic medications or interventional psychiatric modalities.

Conclusions and Relevance

Depressive symptom, daily function, and QoL benefits obtained after 12 months of adjunctive VNS were sustained in about 80% of participants continuing VNS. Approximately 30% with no meaningful benefit at 12 months accrued increased benefit over the subsequent year.

Highlights

1. Over 80% of participants with clinically meaningful benefit (as measured with 7 scales) at 12 months of adjunctive active VNS treatment maintained that benefit at 18 and 24 months of continued active VNS treatment.

2. Rates of loss of meaningful benefit were consistently low: at 18 and 24 months of active VNS, loss of treatment benefit medians were 16.9% and 18.8%, respectively.

3. Benefit with active VNS continued to emerge through the second year of treatment: among participants with no meaningful benefit at 12 months, a median of 30.6% and 37.8% achieved benefit at 18 and 24 months, respectively.

4. Across those measures having meaningful benefit (partial response) as an outcome (MADRS, QIDS-C, QIDS-SR, CGI-I, tripartite measure [combination of mood symptoms, function, and QoL]), achieving meaningful benefit at 12 months resulted in two-thirds of participants having sustained benefit at 24 months, with one-third moving to substantial (full response) benefit.

Graphical Abstract

Graphical Abstract.

Significance Statement.

Patients with markedly treatment-resistant major depression have a low likelihood of benefiting from antidepressant treatments and a high likelihood of rapid relapse if benefit is obtained. In 214 participants randomized in the RECOVER trial to active treatment with adjunctive vagus nerve stimulation (VNS) therapy, outcomes after a second year of stimulation were examined for depressive symptom severity, function, and quality of life (QoL). Participants with meaningful benefit at 12 months maintained this level of benefit or better at 24 months at high rates (median ≥ 80% across all measures). Durability was strong for symptoms, function, and QoL regardless of the threshold used to define benefit and was not attributable to changes in concomitant treatments. Moreover, a substantial proportion of participants had meaningful improvement at 24 months who did not at 12 months. In a sample characterized by profound chronicity and treatment resistance, active VNS was associated with exceptionally strong durability of benefit.

INTRODUCTION

Most patients with major depressive disorder (MDD) substantially improve with 1 to 3 antidepressant treatment attempts, but a sizable proportion (18%-33%) do not.¹ Individuals with treatment-resistant depression (TRD)² or difficult-to-treat depression³ have higher rates of relapse,¹ hospitalization,⁴ suicide attempt,⁵ medical morbidity,⁶ and disability⁷ than those with more treatment-responsive depression.

Treatment-resistant depression presents 2 fundamental therapeutic challenges: initially achieving benefit and subsequently sustaining it. The likelihood of achieving acute symptomatic improvement decreases with increasing levels of treatment resistance.^1,2,8,9 Also, when an initial benefit is achieved, the likelihood that it is sustained decreases with higher levels of treatment resistance.^1,2,10 This finding was prospectively documented in the STAR*D trial, where, following 2 inadequate treatment responses, the likelihood of achieving sustained remission for 1 year was less than 5% with subsequent medication regimens, due both to a low rate of acute improvement and poor durability of benefit.^1,11 As another example, electroconvulsive therapy (ECT) produces remission in 50%-70% of patients with TRD,¹² but approximately 50% of these patients relapse (most within 2-3 months).^13,14 Even with maintenance ECT, less than half of patients with TRD retain remission status at 12 months.^13,15 For TRD, and especially markedly TRD, an assessment of the durability of the initial benefit is critical when evaluating the clinical impact of treatment in patients for whom loss of initial benefit is highly likely.

Cervically implanted vagus nerve stimulation (VNS) has shown promise in achieving meaningful antidepressant effects^16,17 and sustaining this benefit over time.^17,18 Previous VNS trials found that approximately 40%-50% of patients with markedly TRD show meaningful improvement 3-12 months after starting stimulation,^17,19 a much longer interval to benefit than observed with other antidepressant treatments.²⁰ Critically, early VNS studies documented that a large proportion of patients sustained substantial benefit (SB) for at least 1-2 years.^18,21 Similarly, using registry data, Aaronson et al.¹⁷ found that individuals classified as responders to VNS adjunctive to treatment as usual (TAU) had a significantly longer time to relapse or recurrence compared with patients who met response criteria while treated with TAU alone. Determining whether the initial benefits of adjunctive VNS are indeed durable is essential.

In markedly TRD, in addition to durability, the breadth and nature of initial and sustained benefits require consideration.²² Traditionally, therapeutic trials in MDD examine change in depression symptom severity as the sole outcome domain to characterize short-term benefit and its consequent durability.²³ Patients with markedly TRD, however, have profound deficits in day-to-day function and quality of life (QoL), often disproportionate to symptom severity.^24–26 The initial acquisition of benefits in these domains, as well as their durability, has rarely been described in such patients.²⁷

Furthermore, the threshold defining clinical benefit has traditionally corresponded to response, typically a ≥ 50% reduction relative to baseline, or remission, the virtual absence of symptoms.^28,29 In markedly TRD, these outcomes are relatively rare and a lesser degree of improvement, corresponding to partial response (PR; eg, ≥ 30% reduction), is associated with meaningful improvement in QoL.^30,31 Partial response has also shown greater sensitivity to change over time and to treatment group differences compared with the traditional outcome categories.^32,33 However, the clinical utility of adopting lower improvement thresholds for detecting clinical benefit would be questionable if the benefits were not sustained. No previous study has examined durability of benefit in patients with MDD using meaningful benefit (MB) (partial symptom response) and SB (symptom response) levels of improvement.

This report describes clinical outcomes after 18 and 24 months of stimulation for individuals who were randomized to blinded treatment with active VNS combined with TAU in the initial 12 months of the large multisite RECOVER trial (NCT03887715).³⁴ Outcome domains included measures of depressive symptoms, daily function, and QoL. In addition, outcomes were examined using a tripartite composite measure that combines these 3 domains into a single metric.^35,36 For all outcomes, thresholds were defined as denoting “meaningful benefit” (MB) (Table 1). Meaningful benefit corresponded to PR for depressive symptom measures and the a priori, previously empirically determined minimal clinically important differences (MCIDs) for daily function and QoL measures, to 1 or more domains with at least MB on the tripartite composite measure, and to a score of 3 or less on the Clinical Global Impression–Improvement (CGI-I) scale. A greater degree of improvement (ie, SB) was defined using the traditional threshold for response on the depressive symptom measures, 2 or more domains with at least MB for the tripartite metric, and a CGI-I score of 1 or 2. Substantial benefit was not defined for the function and QoL measures, given the absence of investigation justifying a cutoff beyond the MCID. Durability of benefit was described separately for groups that manifested meaningful or SB at 12 months. This study compared clinical outcomes at 12, 18, and 24 months and determined whether clinical outcomes improved, deteriorated, or were unchanged during the second year of stimulation. Participants were classified in terms of their outcome status at 12 months, and durability was assessed by the percentage of participants continuing to manifest benefit at 18 or 24 months. Similarly, this study also documented the percentages of participants who lost benefit (relapse) or achieved benefit (late improvement) during the second year. This study addressed the following specific questions:

Table 1.

Summary of the measures and thresholds used to define categories of benefit.

	Meaningful benefit (MB)	At least meaningful benefit (≥ MB)	Substantial benefit (SB)	Remission
MADRS	30%-49% reduction from baseline	≥ 30% reduction from baseline	≥ 50% reduction from baseline	Score ≤ 9
QIDS-C	30%-49% reduction from baseline	≥ 30% reduction from baseline	≥ 50% reduction from baseline	Score ≤ 5
QIDS-SR	30%-49% reduction from baseline	≥ 30% reduction from baseline	≥ 50% reduction from baseline	Score ≤ 5
CGI-I	Score = 3	Score ≤ 3	Score ≤ 2	Score = 1
Mini-Q-LES-Q	≥ 11.89% increase from baseline	NA	NA	NA
WPAI item 6	≥ 2-point reduction from baseline	NA	NA	NA
Tripartite metric a	Score = 1	Score ≥ 1	Score ≥ 2	NA

For symptom measures (MADRS, QIDS-C, QIDS-SR, and CGI-I), meaningful benefit is equivalent to partial response; substantial benefit is equivalent to response.

^aScore is count of outcome domains that manifested meaningful benefit.

Abbreviations: CGI-I, Clinical Global Impression–Improvement; MADRS, Montgomery–Åsberg Depression Rating Scale; Mini-Q-LES-Q, 7-item subset of the Quality of Life Enjoyment and Satisfaction Questionnaire; NA, not applicable; QIDS-C, Quick Inventory of Depressive Symptomatology–Clinician; QIDS-SR, Quick Inventory of Depressive Symptomatology–Self-Report; WPAI, Work Productivity and Activity Impairment Questionnaire.

(1) For each outcome measure, how did outcomes at 18 and 24 months compare with those at 12 months? What proportion of participants was classified as having no benefit, MB, or SB at the completion of the second year relative to the proportions after the first year of stimulation?

(2) Were the benefits obtained at 12 months durable? Did a substantial proportion of participants who benefited after 1 year of VNS sustain the benefit over the following year?

(3) Did a substantial proportion of participants lose benefit over time? What proportion of participants who achieved meaningful or SB after 1 year of VNS lost that benefit during the following year?

(4) For participants with no benefit after 1 year of VNS, what proportion developed meaningful or SB over the following year?

(5) Are the findings regarding longer-term outcomes contingent on outcome domain, or are they consistent across depressive symptom, daily function, QoL, and tripartite measures?

(6) Are the durability and relapse findings contingent on the degree of benefit at 12 months? Is MB at 12 months informative of participant outcomes at 24 months, and is MB durable?

This study described outcomes during the second year of treatment with active VNS and unconstrained TAU. To examine whether the durability findings could be attributed to change in concomitant treatment regimens, the number and classes of psychotropic medications each participant reported at 12 months and last follow-up (18 or 24 months) were compared, as was exposure to interventional psychiatric treatments (ECT, transcranial magnetic stimulation [TMS], and esketamine/ketamine). Hence, the study also addressed this question:

(7) How did the number and type of concomitant medications and use of interventional psychiatry modalities change over the course of the second year? Could such changes account for findings regarding durability of benefit?

METHODS

Study Design

The first phase of the RECOVER trial (ClinicalTrials.gov Identifier: NCT03887715) used a multicenter, randomized, triple-blind, sham-controlled design to evaluate the safety and effectiveness of adjunctive VNS over a 12-month period in participants with markedly TRD.³⁴ Participants received either adjunctive active or sham VNS therapy. Treatment as usual was continued throughout the study. All participants provided written informed consent, and the study was approved by the local Institutional Review Board of the 84 participating US sites. The study procedures were in accordance with the ethical standards of the Helsinki Declaration.

After the 12-month blinded, randomized phase, all participants were treated openly with active VNS and TAU, without constraint on changes in VNS parameters or concomitant treatments. This report describes the depressive symptom, daily function, and QoL outcomes during the second year of stimulation in the 214 participants randomized to the active VNS arm for the first 12 months and who completed outcome evaluations at 18 or 24 months. Participants initially randomized to the sham VNS arm were excluded, as they had limited exposure to active VNS, an intervention with long delay to peak effectiveness, and durability of the benefit achieved with a 12-month sham intervention has limited practical import.

Patient Sample

Participants were ≥ 18 years of age; in a major depressive episode (MDE) of MDD based on Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria³⁷; had ≥ 4 failed, adequate antidepressant treatments in the current MDE⁸; and had at least moderately severe depression at baseline (≥ 22 on the Montgomery–Åsberg Depression Rating Scale [MADRS]³⁸ on 2 occasions, 14-16 days apart). Baseline MADRS severity was the average of these 2 assessments.

A total of 249 participants were initially randomized to active VNS. Of these, the following were not included in the analytic sample for this report: 15 participants were implanted but discontinued prior to the first 3-month assessment; an additional 13 participants continued after the 3-month assessment but did not have the 12-month assessment used to classify trial outcomes; 7 participants missed the 18- and 24-month assessments. These exclusions resulted in a final analytic sample of 214, which represented 85.9% (214/249) of participants initially randomized to VNS plus TAU (VNS ON + TAU). This high retention rate suggests that differential dropout or selection bias had minimal impact on longer-term outcomes.

This is a naturalistic, observational study of longer-term outcomes of a cohort expected to have especially poor durability of benefit given the sample’s profound chronicity of illness and treatment resistance. The major analyses regarding change in outcome rates during the second year of stimulation were conducted in a subsample (n = 185) that had outcome assessments at 12, 18, and 24 months without missing observations, allowing within-group statistical comparison. The findings from the larger primary sample (n = 214), based on all available follow-up data, are reported in Table S1.

Outcome Assessments

Depressive symptom severity was assessed by the clinician-rated MADRS and Quick Inventory of Depressive Symptomatology–Clinician (QIDS-C) at the clinic visits. Both were completed by trained, off-site, telephone-based raters masked to protocol and treatment. Participants completed the Quick Inventory of Depressive Symptomatology–Self-Report (QIDS-SR) at the same clinic visits. The MADRS assesses 10 symptoms, each rated 0-6 over the previous week.³⁸ The QIDS-C and QIDS-SR assess 9 depressive symptom domains (each rated 0-3) that define an MDE based on DSM-5 criteria over the previous week.³⁹ For each symptom scale, PR and response were defined as ≥ 30% or ≥ 50% reduction in the baseline total score, respectively, based on independent studies that established these thresholds as clinically meaningful.³⁵ On these scales, the terms partial response and MB are equivalent, as are response and SB. Remission was defined as a score of ≤ 9 for the MADRS and ≤ 5 for both the QIDS-C and QIDS-SR. Table 1 defines MB, SB, and remission for each scale.

General psychiatric status was assessed at baseline with the Clinical Global Impression–Severity (CGI-S) scale over the previous week based on behavior, symptoms, and function.⁴⁰ The CGI-I scale, completed by on-site blinded psychiatrists and psychologists, assessed the participants’ overall clinical condition compared with the baseline CGI-S score.⁴⁰ At each assessment, on the CGI-I, an MB was defined as a score of ≤ 3 (at least “minimally improved”), while an SB was defined as a score of ≤ 2 (at least “much improved”), and remission was defined as a score of 1, based on research validating these cutoffs.⁴¹

The Work Productivity and Activity Impairment Questionnaire (WPAI) measures absenteeism, presenteeism, and productivity for working participants (items 2-5).^42–44 Since most participants (74.6%) were not employed,^45,46 WPAI items 2-5 did not apply. Work Productivity and Activity Impairment Questionnaire item 6 assesses, over the previous week, the effect of depression (from 0 for “no effect” to 10 for “massive effect”) on regular daily activities (work around the house, shopping, childcare, exercising, studying, etc.). An MCID or MB on item 6 was defined as a reduction of ≥ 2 points based on research validating this cutoff.³⁵

The 7-item, self-report subset of the Quality of Life Enjoyment and Satisfaction Questionnaire (Mini-Q-LES-Q)⁴⁷ assessed QoL over the previous week using the items on the 14-item Q-LES-Q most sensitive to change during treatment of MDD: work, household activities, social relationships, family relationships, leisure time activities, ability to function in daily life, and overall sense of well-being. Each item was rated from 1 (very poor) to 5 (very good) then converted to percent maximum scores by subtracting 7 from the raw total score (range = 7-35) and dividing by 28. The MCID for the Mini-Q-LES-Q was estimated a priori as an 11.89% increase from baseline based on previous research using the 14-item Q-LES-Q.⁴⁸ Participants meeting this threshold were classified as manifesting MB in QoL, while those below were classified as having no meaningful benefit (NB).

The tripartite composite combined measures of change in symptom severity (QIDS-C), function (WPAI item 6), and QoL (Mini-Q-LES-Q) into a single metric.³⁵ Composite scoring ranged from 0 to 3, based on the number of domains in which at least meaningful benefit (≥ MB) was manifest: scores of 0 were categorized as NB, 1 as MB, and 2 or 3 as SB. The QIDS-C was selected a priori as the depressive symptom component of the tripartite measure, as it was based on blinded, off-site clinician ratings and showed superior sensitivity to symptom change during the randomized phase compared with the MADRS.³³ The tripartite metric was introduced in the RECOVER trial and has been shown to capture a wider group of patients than any one dimension alone. It correlates substantially with clinician global judgment of improvement and has shown utility in identifying the patient features predictive of VNS outcome.^35,36 Durability of benefit has rarely been examined for daily function and QoL outcomes, and never for this tripartite metric.

All baseline assessments were completed before randomization.³⁴ During the initial 12 months following randomization, depressive symptoms were assessed monthly, from months 3 through 12, and function and QoL were assessed quarterly. All 3 domains were assessed at 18 and 24 months during the second year of the trial.

Statistical Analyses

This was an observational, naturalistic study of the durability and emergence/loss of benefit in the second year of the RECOVER participants initially randomized to active VNS. Analyses were conducted on 7 outcome measures with 3 follow-up time points (12, 18, and 24 months). Confidence intervals and P-values were not adjusted for multiplicity (α = 0.05). Statistical analyses were performed with SAS 9.4 (SAS Institute, Cary, NC).

The study reports descriptive information on second-year categorical outcomes for participants with MDD who were originally randomized to the VNS ON + TAU study arm. At each assessment (12, 18, and 24 months), outcomes for each depressive symptom scale (MADRS, QIDS-C, and QIDS-SR), the CGI-I, and the tripartite metric were classified as NB, MB, SB, and remission (with exception of remission for the tripartite metric). Daily function (WPAI item 6) and QoL (Mini-Q-LES-Q) measures were classified only dichotomously, as reflecting NB or MB.

Change in Outcomes over Second Year

For each outcome measure, rates of NB, MB, SB, and remission, as well as ≥ MB (MB or SB), were computed at the 12-, 18-, and 24-month assessments. Statistical significance of the difference between the rates at the 12- and 24-month time points was tested with McNemar’s exact test, with corresponding 95% CIs estimated by the Mantel–Haenszel method for dependent proportions, in the subsample with no missing data at all 3 assessments.

Durability of Benefit

For each outcome, durability was operationalized as the proportion of participants with ≥ MB at the 12-month time point who were classified as ≥ MB at the 18- and 24-month assessments. Studies of longer-term outcomes in MDD have typically restricted follow-up to patients meeting response or remission criteria at an acute end point, including previous durability studies of VNS.^17,18 To match these thresholds, the durability analyses were repeated, restricting the sample to those who met SB or remission criteria at the 12-month time point, and documenting the percentages manifesting ≥ MB, SB, and remission at follow-up.

Loss of Benefit and Relapse

For each measure, the primary analyses of durability reflected the proportion of participants who manifested MB or SB at 12 months who retained ≥ MB at follow-up. The first assessment of loss of benefit examined the inverse, the proportion of participants with MB or SB at 12 months who had NB at follow-up. However, this analysis likely overestimated meaningful loss of benefit, as a minor change in scores could result in a shift from MB to NB. In contrast, relapse is traditionally defined as a marked increase in symptomatology in patients who previously substantially benefited from an intervention.^14,18,29,49 Therefore, we also applied a stricter definition of loss of benefit to the depressive symptom, CGI-I, and tripartite outcomes by calculating the proportion of participants with SB at 12 months who had NB at follow-up, operationalizing the traditional classification of relapse.

Emergence of Benefit

The foregoing analyses focused on the longer-term outcomes of participants who were classified with MB or SB at the 12-month assessment. Longer-term outcomes were also examined for those with NB at 12 months.

Concomitant Treatments

All ongoing medication and device-based treatments were documented at month 12 and last follow-up at 18 or 24 months and categorized according to the WHO Anatomical Therapeutic Chemical (ATC) Classification System.⁵⁰ At each assessment, the total number of psychotropic medications (ATC Level 1 Nervous System) reported by the participant and the number per class (antianxiety, anticonvulsant, antidepressant, antipsychotic, anxiolytic, hypnotic, stimulant, and other) were calculated. Supplemental Methods S1 details the methods used in classifying medications and Supplemental Methods 2 provides the complete list of psychotropic medication entries in RECOVER and their classifications. Paired t-tests determined whether overall psychotropic use and use of specific classes changed in the second year. The number of psychotropic medications added or subtracted at last follow-up relative to month 12 was computed for each class and the means tested against 0 with one-sample t-tests. Compared to other outcome measures, the tripartite metric had both the highest rate of participants manifesting benefit at 12 months and the strongest durability values. Accordingly, the analyses on change in medication use in the total sample were repeated to characterize the pattern specifically in the largest subgroup that exhibited sustained benefit. Similarly, exposure to interventional psychiatric treatments was tested for change over the study periods in the complete sample and for changes that occurred in the subgroup with sustained benefit. Similar analyses were performed using the CGI-I, which was selected because this metric also had a high rate of participants benefiting at 12 months and a high rate of sustained benefit at 24 months, and were conducted by blinded, on-site primary investigators with experience in TRD assessment.

RESULTS

Patient Sample

Table 2 details demographic, clinical, and treatment history characteristics of the participants with follow-up data at 18 or 24 months. RECOVER participants had profoundly chronic and markedly resistant depressive illness, with a mean of 13.5 lifetime antidepressant treatment trials with insufficient benefit, including a majority who had insufficient benefit from neurostimulation treatments in the current episode (52.3% TMS; 38.3% ECT). With a mean age of 55.2 years, the sample spent the majority of their lives (29.0 years; 52.6% lifetime) in depressive illness and averaged 17.1 years of sustained depression on entry into the trial. In addition, the sample was severely negatively impacted by depressive illness. A total of 72.4% were unemployed, and participants self-reported very poor QoL (mean percent max Q-LES-Q–Short Form score of 48.9%, which falls below the “severe impairment” line [50%] of the scale; a normal control/community normative score typically resides in the normal range of 75%-91% max score, with severe impairment being 2 SDs below the norm)^51,52 and were functionally impaired (WPAI item 6 mean score of 7.2, which ranks more impaired than studies of nonresistant depression [4.76],⁵³ chronic migraine [5.52],⁵⁴ and rheumatoid arthritis [3.33]).⁵⁵

Table 2.

Demographic, clinical, and treatment history characteristics of the sample treated with VNS and TAU for 2 years (n = 214).

Variable
Demographics
Age, yr, mean (SD)	55.2 (14.3)
Sex, n (%)
Female	146 (68.2)
Male	68 (31.8)
Employed, n (%)	59 (27.6)
Baseline clinical presentation
Single-episode vs recurrent MDD, n (%)
Single-episode MDD	95 (44.4)
Recurrent MDD	119 (55.6)
MADRS total score (0-60), mean (SD)	34.1 (4.9)
CGI-S score (1-7), mean (SD)	5.1 (0.7)
QIDS-C total score (0-27), mean (SD)	17.2 (3.1)
QIDS-SR total score (0-27), mean (SD)	17.4 (3.7)
Function and quality of life assessments
WPAI item 6, mean (SD)	7.2 (2.0)
Q-LES-Q total score (14-70), mean (SD)	34.2 (7.3)
Mini-Q-LES-Q total score (7-35), mean (SD)	15.0 (4.1)
Mood disorder history, mean (SD)
Age at onset of first symptoms of depression, years	20.9 (12.1)
Age at definitive diagnosis of any major mood disorder, years	27.1 (12.6)
Duration of current MDE, years	17.1 (15.3)
Duration of lifetime MDEs, yearsa	29.0 (16.5)
Percentage of lifetime in MDEb	52.6 (24.4)
Lifetime number of prior hospital admissions for mood disorder	2.3 (4.3)
Lifetime history of suicide attempts, n (%)	89 (41.6)
Antidepressant treatments, lifetime
Lifetime number of inadequate responses to antidepressant treatments, mean (SD)c	13.5 (9.2)
Lifetime number of inadequate responses to antidepressant pharmacotherapies, mean (SD)	11.5 (8.6)
Lifetime history of ECT, n (%)
Responded	25 (11.7)
Did not respond	76 (35.5)
Antidepressant treatments in current MDE
Any pharmacotherapy, n (%)	214 (100)
Esketamine	48 (22.4)
Any psychotherapy, n (%)	162 (75.7)
Any neuromodulation, n (%)	144 (67.3)
TMS	112 (52.3)
ECT	82 (38.3)

^aDuration of lifetime MDEs is the number of years of previous and current major depressive episodes. If both previous and current episodes were missing, then duration of lifetime episodes was considered as missing. If either of the 2 was not missing, then duration of lifetime episodes was the sum of the non-missing number. The maximum duration of estimated lifetime episodes in years was set as age at baseline. If the number of hospital admissions or of lifetime episodes was more than 10, the number was capped at 10.

^bPercentage of lifetime in MDE was calculated as “duration of lifetime MDEs” divided by “age at baseline.”

^cIncluding pharmacotherapies, ECT, TMS, and psychotherapies.

Abbreviations: CGI-S, Clinical Global Impression–Severity; ECT, electroconvulsive therapy; MADRS, Montgomery–Åsberg Depression Rating Scale; MDD, major depressive disorder; MDE, major depressive episode; Mini-Q-LES-Q, 7-item subset of the Quality of Life Enjoyment and Satisfaction Questionnaire; QIDS-C, Quick Inventory of Depressive Symptomatology–Clinician; QIDS-SR, Quick Inventory of Depressive Symptomatology–Self-Report; Q-LES-Q, Quality of Life Enjoyment and Satisfaction Questionnaire; TAU, treatment as usual; TMS, transcranial magnetic stimulation; VNS, vagus nerve stimulation; WPAI, Work Productivity and Activity Impairment Questionnaire.

Overall Patient Course

Table 1 details the definitions of MB, ≥ MB, SB, and remission for all applicable metrics. Table 3 presents the proportion of participants classified as having NB, MB, SB, and remission at 12, 18, and 24 months for the depressive symptom measures and CGI-I; the proportion of participants having NB, MB, and SB for tripartite outcomes; and those with NB and MB in function (WPAI item 6) and QoL (Mini-Q-LES-Q) outcomes for participants with assessments at all 3 time points (12, 18, and 24 months; n = 181). At 12 months, there were considerable differences among the outcome measures in classification of MB or at least MB (≥ MB), with 43.6% achieving ≥ MB on the MADRS and 80.0% on the tripartite composite measure. The increase from month 12 to 24 in the percentage achieving ≥ MB was significant for the MADRS and CGI-I measures, and marginally so for the QIDS-C. The specific increase in SB status was significant for the QIDS-C and CGI-I, and marginally so for the MADRS. Notably, the increase from month 12 to 24 in the proportion of participants achieving remission was significant for the MADRS. Table 3 shows that for all outcome measures other than the QIDS-SR, the proportion of participants with NB decreased over the course of the second year, while the rates of participants achieving ≥ MB and SB increased. Table S1 presents the rates at the 12-, 18-, and 24-month assessments across the sample.

Table 3.

Rates of each clinical outcome at 12, 18, and 24 months in participants with no missing values at all 3 time points.

Outcome classification	12 months	18 months	24 months	%Change (95% CI) months 12-24	P a
MADRS
n	181	181	181
≥ MB	43.6% (79)	49.2% (89)	53.6% (97)	9.9% (2.9%, 17.0%)	0.007
SB	28.7% (52)	32.6% (59)	34.3% (62)	5.5% (−0.7%, 11.8%)	0.086
MB	14.9% (27)	16.6% (30)	19.3% (35)	4.4% (−2.7%, 11.6%)	0.228
Remission	15.5% (28)	18.2% (33)	21.5% (39)	6.1% (0.3%, 11.8%)	0.041
NB	56.4% (102)	50.8% (92)	46.4% (84)	—	—
QIDS-C
n	181	181	181
≥ MB	55.8% (101)	58.6% (106)	62.4% (113)	6.6% (−1.0%, 14.2%)	0.090
SB	33.7% (61)	37.6% (68)	40.9% (74)	7.2% (0.7%, 13.7%)	0.033
MB	22.1% (40)	21.0% (38)	21.5% (39)	−0.6% (−8.7%, 7.6%)	0.895
Remission	19.3% (35)	23.2% (42)	24.3% (44)	5.0% (−1.4%, 11.3%)	0.128
NB	44.2% (80)	41.4% (75)	37.6% (68)	—	—
QIDS-SR
n	181	181	181
≥ MB	61.3% (111)	56.9% (103)	59.1% (107)	−2.2% (−9.2%, 4.8%)	0.537
SB	38.1% (69)	39.2% (71)	41.4% (75)	3.3% (−3.5%, 10.1%)	0.343
MB	23.2% (42)	17.7% (32)	17.7% (32)	−5.5% (−13.0%,1.9%)	0.149
Remission	24.3% (44)	21.5% (39)	23.8% (43)	−0.6% (−7.7%, 6.5%)	0.879
NB	38.7% (70)	43.1% (78)	40.9% (74)	—	—
CGI-I
n	185	185	185
≥ MB	68.6% (127)	80.5% (149)	77.3% (143)	8.6% (1.0%, 16.3%)	0.029
SB	41.1% (76)	48.6% (90)	51.9% (96)	10.8% (3.8%, 17.8%)	0.003
MB	27.6% (51)	31.9% (59)	25.4% (47)	−2.2% (−9.8%, 5.5%)	0.579
Remission	17.3% (32)	17.3% (32)	21.6% (40)	4.3% (−1.2%, 9.9%)	0.131
NB	31.4% (58)	19.5% (36)	22.7% (42)	—	—
WPAI item 6
n	175	175	175
MB	57.1% (100)	65.7% (115)	61.1% (107)	4.0% (−3.8%, 11.8%)	0.317
NB	42.9% (75)	34.3% (60)	38.9% (68)	—	—
Mini-Q-LES-Q
n	172	172	172
MB	64.0% (110)	69.2% (119)	65.7% (113)	1.7% (−5.9%, 9.4%)	0.655
NB	36.0% (62)	30.8% (53)	34.3% (59)	—	—
Tripartite metric
n	165	165	165
≥ MB	80.0% (132)	83.6% (138)	82.4% (136)	2.4% (−4.1%, 8.9%)	0.465
SB	60.6% (100)	67.3% (111)	66.1% (109)	5.5% (−2.6%, 13.6%)	0.189
MB	19.4% (32)	16.4% (27)	16.4% (27)	−3.0% (−11.3%, 5.3%)	0.475
NB	20.0% (33)	16.4% (27)	17.6% (29)	—	—

Data presented as % (n).

%Change (95% CI): Difference in proportions (dependent) and 95% CI estimated by Mantel–Haenszel method.

^aSignificance level of McNemar’s test comparing the dependent proportions at 12- and 24-month assessments. Bolded values represent significance at alpha level of 0.05.

Abbreviations: CGI-I, Clinical Global Impression–Improvement; MADRS, Montgomery–Åsberg Depression Rating Scale; MB, meaningful benefit; Mini-Q-LES-Q, 7-item subset of the Quality of Life Enjoyment and Satisfaction Questionnaire; NB, no meaningful benefit; QIDS-C, Quick Inventory of Depressive Symptomatology–Clinician; QIDS-SR, Quick Inventory of Depressive Symptomatology–Self-Report; SB, substantial benefit; WPAI, Work Productivity and Activity Impairment Questionnaire.

Missing data had minimal impact, as most participants remained in follow-up through 24 months. Discontinuation rates were similar regardless of month 12 outcomes, with a median dropout of 15.5% versus 17.7% for those with and without MB, respectively (Table S2).

Durability of Benefit

Table 4A and Figure 1 provide the percentage of participants manifesting MB or SB (≥ MB) at the 12-month assessment who also manifested MB or SB at the 18- and 24-month assessments. At the 18-month assessment, these percentages ranged from 78.0% to 89.2% across all outcome measures. At 24 months, these percentages ranged from 79.2% to 89.6%. The median durability percentage across the 7 measures was 83.1% at 18 months and 81.3% at 24 months. Thus, across all outcomes, there was substantial retention of benefit during the second year of stimulation and little loss of durability from 18 to 24 months. A conservative analysis of all participants that imputed all missing data points as representing lost benefit found similar substantial retention of benefits (Table S3).

Table 4.

Percentage of participants who maintained (A) meaningful benefit or substantial benefit and (B) remission at 18- and 24-month follow-up.

A. Durability of meaningful or substantial benefit at 12 months
Maintenance of ≥ MB in participants with ≥ MB at 12 months
Outcome measure (n at month 12)	≥ MB at 18 months	≥ MB at 24 months
MADRS (97)	83.0% (78/94)	82.9% (68/82)
QIDS-C (122)	81.5% (97/119)	80.8% (84/104)
QIDS-SR (129)	78.0% (99/127)	79.6% (90/113)
CGI-I (147)	89.0% (129/145)	85.3% (110/129)
QoL (Mini-Q-LES-Q; 130)	87.5% (112/128)	81.3% (91/112)
Function (WPAI item 6; 119)	83.1% (98/118)	79.2% (80/101)
Tripartite metric (160)	89.2% (140/157)	89.6% (121/135)
Maintenance of ≥ MB in participants with SB at 12 months
Outcome measure (n at month 12)	≥ MB at 18 months	≥ MB at 24 months
MADRS (62)	95.0% (57/60)	92.6% (50/54)
QIDS-C (75)	87.8% (65/74)	90.3% (56/62)
QIDS-SR (82)	92.6% (74/80)	87.3% (62/71)
CGI-I (86)	95.3% (81/85)	92.2% (71/77)
Tripartite metric (122)	93.3% (111/119)	94.2% (97/103)
Maintenance of SB in participants with SB at 12 months
Outcome measure (n at month 12)	SB at 18 months	SB at 24 months
MADRS (62)	75.0% (45/60)	74.1% (40/54)
QIDS-C (75)	74.3% (55/74)	79.0% (49/62)
QIDS-SR (82)	78.8% (63/80)	74.6% (53/71)
CGI-I (86)	80.0% (68/85)	83.1% (64/77)
Tripartite metric (122)	84.9% (101/119)	80.6% (83/103)
B. Durability of remission at 12 months
Maintenance of ≥ MB in participants with remission at 12 months
Outcome measure (n at month 12)	≥ MB at 18 months	≥ MB at 24 months
MADRS (35)	97.1% (33/34)	89.7% (26/29)
QIDS-C (42)	95.1% (39/41)	88.9% (32/36)
QIDS-SR (53)	98.0% (50/51)	95.7% (44/46)
CGI-I (36)	94.3% (33/35)	100% (33/33)
Maintenance of SB in participants with remission at 12 months
Outcome measure (n at month 12)	SB at 18 months	SB at 24 months
MADRS (35)	82.4% (28/34)	79.3% (23/29)
QIDS-C (42)	80.5% (33/41)	80.6% (29/36)
QIDS-SR (53)	86.3% (44/51)	78.3% (36/46)
CGI-I (36)	88.6% (31/35)	97.0% (32/33)
Maintenance of remission in participants with remission at 12 months
Outcome measure (n at month 12)	Remission at 18 months	Remission at 24 months
MADRS (35)	58.8% (20/34)	65.5% (19/29)
QIDS-C (42)	65.9% (27/41)	61.1% (22/36)
QIDS-SR (53)	62.7% (32/51)	47.8% (22/46)
CGI-I (36)	62.9% (22/35)	69.7% (23/33)

Abbreviations: CGI-I, Clinical Global Impression–Improvement; MADRS, Montgomery–Åsberg Depression Rating Scale; MB, meaningful benefit; Mini-Q-LES-Q, 7-item subset of the Quality of Life Enjoyment and Satisfaction Questionnaire; QIDS-C, Quick Inventory of Depressive Symptomatology–Clinician; QIDS-SR, Quick Inventory of Depressive Symptomatology–Self-Report; QoL, quality of life; SB, substantial benefit; WPAI, Work Productivity and Activity Impairment Questionnaire.

Figure 1.

For each outcome measure, the percentage of participants who had at least meaningful benefit (≥ MB) at 18 and 24 months of those who had at least meaningful benefit (≥ MB) at 12 months. Abbreviations: CGI-I, Clinical Global Impression–Improvement; MADRS, Montgomery–Åsberg Depression Rating Scale; MB, meaningful benefit; QIDS-C, Quick Inventory of Depressive Symptomatology–Clinician; QIDS-SR, Quick Inventory of Depressive Symptomatology–Self-Report; QoL, quality of life.

To determine whether durability was impacted by the extent of clinical benefit at 12 months, the follow-up data were reexamined, restricting the sample to participants with SB status at 12 months (ie, ≥ 50% reduction from baseline in symptom measures; Table 4A). The percentage of participants who manifested MB or SB at 18 and 24 months was higher than when a lower threshold of improvement (MB) was used to define benefit at 12 months. Across all outcomes, for participants who manifested SB at 12 months, the durability percentages ranged from 87.8% to 95.3% at 18 months (median = 93.3%) and from 87.3% to 94.2% at 24 months (median = 92.2%). Thus, nearly all these participants retained ≥ MB through 24 months. As also seen in Table 4A, durability of benefit remained high even when requiring that SB status be maintained to qualify as durable, with medians of 78.8% and 79.0% at 18 and 24 months, respectively. Therefore, attaining MB and/or SB at 12 months was associated with high likelihood of maintaining benefit over the second year of stimulation.

Table 4B describes the percentage of participants manifesting ≥ MB, SB, or remission at 18 and 24 months who achieved remission at 12 months. Across all outcome measures, among participants who achieved remission at 12 months, 58.8%-65.9% maintained that level of benefit at 18 months (median = 62.8%) and 47.8%-69.7% at 24 months (median = 63.3%). Few participants achieving remission at 12 months lost all benefit, with retention of SB having medians of 84.4% and 80.0% at months 18 and 24, respectively, and retention of ≥ MB in nearly the entire sample, with medians of 96.1% and 92.7%.

Loss of Benefit and Relapse

Loss of benefit was defined as NB status at follow-up among participants who had achieved MB or SB (≥ MB) status at 12 months (Table 5). At 18 months, this loss of ≥ MB was observed across outcomes in 10.8%-22.0% of participants, with a median of 16.9%. At 24 months, this loss of benefit characterized 10.4%-20.8% of participants, with a median of 18.8%. When the extent of clinical benefit at 12 months was increased to SB, reflecting traditional definitions of clinical response and relapse,^28,29 rates of loss of benefit (relapse) were consistently lower at both follow-up time points, with medians of 6.7% and 7.8% at 18 and 24 months, respectively. Similar findings in loss of benefit were observed among participants who achieved remission at 12 months, with medians of 3.9% and 7.3% at 18 and 24 months, respectively.

Table 5.

Percentage of participants with loss of benefit at 18- and 24-month follow-up.

Loss of benefit in participants with ≥ MB at 12 months
Outcome measure (n at month 12)	NB at 18 months	NB at 24 months
MADRS (97)	17.0% (16/94)	17.1% (14/82)
QIDS-C (122)	18.5% (22/119)	19.2% (20/104)
QIDS-SR (129)	22.0% (28/127)	20.4% (23/113)
CGI-I (147)	11.0% (16/145)	14.7% (19/129)
QoL (Mini-Q-LES-Q; 130)	16.9% (20/128)	20.8% (21/112)
Function (WPAI item 6; 119)	12.5% (16/118)	18.8% (21/101)
Tripartite metric (160)	10.8% (17/157)	10.4% (14/135)
Loss of benefit in participants with SB at 12 months
Outcome measure (n at month 12)	NB at 18 months	NB at 24 months
MADRS (62)	5.0% (3/60)	7.4% (4/54)
QIDS-C (75)	12.2% (9/74)	9.7% (6/62)
QIDS-SR (82)	7.5% (6/80)	12.7% (9/71)
CGI-I (86)	4.7% (4/85)	7.8% (6/77)
Tripartite metric (122)	6.7% (8/119)	5.8% (6/103)
Loss of benefit in participants with remission at 12 months
Outcome measure (n at month 12)	NB at 18 months	NB at 24 months
MADRS (35)	2.9% (1/34)	10.3% (3/29)
QIDS-C (42)	4.9% (2/41)	11.1% (4/36)
QIDS-SR (53)	2.0% (1/51)	4.3% (2/46)
CGI-I (36)	5.7% (2/35)	0% (0/33)

Data presented as % (n).

Proportions were calculated using all available data at each visit.

Abbreviations: CGI-I, Clinical Global Impression–Improvement; MADRS, Montgomery–Åsberg Depression Rating Scale; MB, meaningful benefit; Mini-Q-LES-Q, 7-item subset of the Quality of Life Enjoyment and Satisfaction Questionnaire; NB, no meaningful benefit; QIDS-C, Quick Inventory of Depressive Symptomatology–Clinician; QIDS-SR, Quick Inventory of Depressive Symptomatology–Self-Report; QoL, quality of life; SB, substantial benefit; WPAI, Work Productivity and Activity Impairment Questionnaire.

Emergence of Benefit for Those without Benefit at 12 Months

Table 6 describes the longer-term outcomes of participants who manifested NB at 12 months. At both 18 and 24 months, and for each outcome measure, a substantial number of participants manifested clinically meaningful benefit (≥ MB). At 18 months, this characterized 23.7%-54.7% of participants across outcome measures, with a median of 30.6%. At 24 months, this improvement ranged from 26.0% to 60.7% across outcome measures, with a median of 37.8%.

Table 6.

Percentage of participants with no meaningful benefit at 12 months who had no meaningful benefit, meaningful benefit, substantial benefit, and meaningful benefit or substantial benefit at 18- and 24-month follow-up.

	18 months				24 months
Outcome measure	NB (%)	MB (%)	SB (%)	≥ MB (%)	NB (%)	MB (%)	SB (%)	≥ MB (%)
MADRS, n = 116	76.3	16.7	7.0	23.7	70.2	16.3	13.5	29.8
QIDS-C, n = 91	73.0	20.2	6.7	27.0	62.2	25.6	12.2	37.8
QIDS-SR, n = 84	75.3	21.0	3.7	24.7	74.0	15.1	11.0	26.0
CGI-I, n = 67	45.3	39.1	15.6	54.7	39.3	29.5	31.1	60.7
QoL (Mini-Q-LES-Q), n = 76	69.4	30.6	–	30.6	59.1	40.9	–	40.9
Function (WPAI item 6), n = 89	60.0	40.0	–	40.0	64.6	35.4	–	35.4
Tripartite metric, n = 40	48.6	24.3	27.0	51.4	47.2	25.0	27.8	52.8

Abbreviations: CGI-I, Clinical Global Impression–Improvement; MADRS, Montgomery–Åsberg Depression Rating Scale; MB, meaningful benefit; Mini-Q-LES-Q, 7-item subset of the Quality of Life Enjoyment and Satisfaction Questionnaire; NB, no meaningful benefit; QIDS-C, Quick Inventory of Depressive Symptomatology–Clinician; QIDS-SR, Quick Inventory of Depressive Symptomatology–Self-Report; QoL, quality of life; SB, substantial benefit; WPAI, Work Productivity and Activity Impairment Questionnaire.

Stability of Meaningful Benefit Outcomes

Figure 2 provides an alluvial diagram graphically conveying the changes in outcomes over time for the CGI-I and tripartite metric. Figures S1 and S2 provide the alluvial diagrams for the remaining outcome measures. Of the 5 outcome measures that provided MB and SB outcomes at 24 months, 28.6%-40.5% (median = 35.7%) of participants classified as MB at 12 months improved to SB at 24 months, 25.0%-35.7% (median = 33.3%) lost benefit (NB), and 26.2%-46.2% (median = 31.0%) maintained MB a year later. Thus, approximately twice as many participants who achieved MB status at 12 months maintained or improved benefit at 24 months than lost benefit.

Figure 2.

Alluvial diagram for changes in (A) CGI-I and (B) tripartite composite metric outcomes across months 12, 18, and 24. Abbreviation: CGI-I, Clinical Global Impression–Improvement.

Concomitant Treatments

Table S4 summarizes the comparisons of the number and type of concomitant psychotropic medications reported at the 12-month and last follow-up (18 or 24 months) assessments. Across and within medication classes, paired t-tests were used to identify changes in number of concomitant psychotropics. These analyses were conducted in 3 samples: total sample (n = 214), participants with durable MB on the tripartite metric (n = 142; maintaining ≥ 1 on the tripartite), and participants with durable SB on the CGI-I, a more conservative metric (n = 70; maintaining score of 1 or 2). None of the 24 comparisons (8 classes × 3 samples) reported in Table S4 achieved statistical significance. There was no evidence of consistent change in number of psychotropics administered over the follow-up period, either across or within classes. The mean number of psychotropics was 4.6 at month 12, and 4.6 at the last follow-up time point. This pattern held specifically in the subgroups that manifested durable benefit, regardless of definition.

Exposure to interventional treatments (ECT, TMS, and esketamine/ketamine) over the follow-up period was also examined in the 3 samples (Table S5). Overall, use of these interventions was quite limited, particularly ECT and TMS. For example, for the tripartite-durable subsample, while several participants continued maintenance ECT, TMS, or esketamine, no participants started a new course of ECT or TMS during the follow-up period, and 2 started a new course of esketamine/ketamine. Similar results were obtained in the CGI-I–durable subsample, as no participants received a new course of ECT or TMS, and 1 participant received a new course of esketamine/ketamine. Hence, the addition of interventional treatments was rare in the subgroups that manifested durable benefit, regardless of definition.

DISCUSSION

The study participants had marked chronicity of depressive illness and profound treatment resistance with impairments of daily function and QoL. These clinical features have been consistently linked to poor prognosis in MDD.^2,56,57 Therefore, achieving and sustaining meaningful clinical improvement would be considered unlikely in these patients. Nonetheless, this study documented strong durability of MB in this markedly TRD sample treated with adjunctive VNS for 2 years. Across depressive symptoms, function, QoL, and the tripartite metric, and across off-site and on-site clinician and self-ratings, approximately 80% of participants with ≥ MB at 12 months maintained this status at 18 and 24 months (Figure 1). Those participants with greater benefit at 12 months (SB or remission) maintained ≥ MB at higher rates at each follow-up (Table 4, A and B) compared to those with less robust benefit at 12 months. The rates of loss of benefit were notably low and mirrored the durability findings. Relapse was rare, as participants with more extensive improvement (SB or remission) at 12 months were especially unlikely to exhibit NB at each follow-up (Table 5). Across outcome measures, participants who had MB at 12 months were twice as likely to maintain this level of benefit or improve to SB than manifest NB at the 24-month follow-up (Figure 2; Figures S1 and S2). This finding underscores the utility of including partial response when classifying outcomes in markedly treatment-resistant MDD.^3,35 Thus, strong durability was observed regardless of the threshold used to define benefit. Finally, though remission is a relatively rare outcome in markedly TRD, the data indicated that a meaningful subset (21.5%) of participants moved into this level of benefit on the MADRS at 24 months—a significant increase vis-à-vis the 12-month assessment (Table 3).

Across the sample, clinical outcomes generally improved during the second year of stimulation. Vagus nerve stimulation is known to have a slow onset of clinical improvement in both epilepsy and MDD,^17,19,58,59 often taking 6-18 months to manifest peak improvement. In the RECOVER trial, VNS electrical parameters were fixed after 2 months titration. It is possible that more accelerated responses could be observed if more flexible dosing were allowed in the first year. In this study, clinical outcomes at 24 months were generally superior to those at 12 months. Up to 60% of participants (median of ~38% across measures) with NB at 12 months achieved MB at 24 months (Table 6). This finding demonstrated that a significant subset of participants with markedly TRD required more than a year to achieve MB with VNS therapy. Extent of treatment resistance may not only predict the likelihood of acute benefit and the durability of obtained benefit but also the intensity or duration of treatment needed to achieve benefit. For instance, the number of ECT sessions needed to achieve response has shown a sequential increase over time as the intervention has been used with more resistant patients, and longer courses of ECT (ie, > 12 sessions) are now commonly required.⁶⁰ Similarly, there are considerable individual differences in the time to benefit with VNS, and some patients may require at least 2 years of VNS to determine their responsivity.

The findings regarding durability of benefit across multiple domains (depressive symptoms, daily function, and QoL) support the validity of the conclusions, which are consistent with previous studies of VNS in TRD.^17,18 RECOVER is among the first of the long-term MDD treatment trials to examine durability, not only for reductions in symptom measures but also for improvements in daily function and QoL. In general, improvements in function and QoL were equally or slightly more durable than improvements in symptom measures. The tripartite metric that evaluated outcomes across the 3 domains identified 80% of the sample as having MB or better (≥ MB) at 12 months, and approximately 80% of those participants maintained this level of benefit at 24 months. Thus, by this measure, approximately two-thirds of the sample benefited by 1 year and had sustained benefit during the second year of stimulation. In contrast, after inadequate response to 2 medications in the STAR*D trial, with subsequent pharmacological interventions less than 5% of patients achieved sustained remission at 1 year. The RECOVER VNS sample was likely the most resistant ever studied in a randomized controlled trial, having not benefited sufficiently from more than 13 lifetime antidepressant treatment trials, as well as interventional treatments with established efficacy in TRD (ECT, TMS, and esketamine). In this context, the durability of benefit was exceptional.

Importantly, this study documented strong durability of benefit in the group of patients treated with active VNS and TAU for 2 years. It did not demonstrate that VNS was necessary in achieving this sustained benefit. Such a demonstration would require a randomized, masked, discontinuation design where patients who benefited from VNS are assigned to maintenance treatment with active or sham VNS. Given the nature of the VNS population, discontinuation studies of this type are unlikely on ethical grounds. There is documentation of a lead break followed by relapse in a patient with markedly TRD responding well to VNS, which was reversed upon lead repair.⁶¹

One major alternative explanation for observed durability findings could potentially be changes in concomitant treatment during the second year, during which there were no constraints on concomitant treatment. However, examination of the numbers and classes of medications added and subtracted during the second year, and specifically the receipt of ECT, TMS, and esketamine, did not yield a pattern suggesting impact on durability of benefit. These findings are consistent with the view that active VNS was an essential component to the durability of benefit.

It is not known why VNS results in strong durability in patients with markedly TRD. It has been suggested that antidepressant interventions with slower onset to clinical benefit display greater durability than those of faster-acting interventions, perhaps reflecting differences in neuroplastic processes.¹¹ In addition, VNS, like deep brain stimulation (DBS), involves continuous neurostimulation with compliance ensured. The beneficial effects of DBS on motor symptoms of Parkinson’s disease often persist for 10 years or more, despite continued degeneration of dopamine-producing cells in the substantia nigra, and DBS in markedly TRD appears to have substantial durability.^62,63 The strong durability of VNS in individuals with markedly TRD could be related to its slow onset of peak benefit and/or the use of continuous neurostimulation.

In summary, this study examined durability of benefit in a sample of patients with markedly TRD who were profoundly disabled, chronically ill, and had histories of inadequate antidepressant benefit from numerous trials of medication, psychotherapy, and interventional treatments. A substantial subset of participants treated with VNS and TAU achieved benefit at 12 months across measures of symptoms, function, and QoL. About 80% of those who benefited after 12 months of VNS maintained or improved upon these benefits over the ensuing year. In addition, another subset (~38%) of participants who did not achieve MB at 12 months did so by the end of the second year.

Author contributions

Charles R. Conway (Conceptualization, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing [lead]), Augustus John Rush (Conceptualization [lead], Investigation [equal], Methodology [lead], Supervision [equal], Validation, Visualization, Writing—original draft, Writing—review & editing [lead]), Scott T. Aaronson (Conceptualization [lead], Investigation [equal], Methodology [lead], Supervision [equal], Validation, Visualization, Writing—original draft, Writing—review & editing [lead]), Mark T. Bunker (Conceptualization, Methodology [lead], Supervision [equal], Validation, Visualization, Writing—original draft, Writing—review & editing [lead]), Charles Gordon (Conceptualization, Data curation, Formal analysis, Methodology, Software, Validation, Visualization, Writing—original draft, Writing—review & editing [lead]), Mark S. George (Conceptualization [lead], Investigation [equal], Methodology [lead], Validation, Writing—review & editing [equal]), Patricio Riva-Posse (Investigation, Validation, Writing—review & editing [equal]), Rebecca M. Allen (Investigation, Validation, Writing—review & editing [equal]), Ziad Nahas (Conceptualization [lead], Investigation [equal], Methodology [lead], Validation, Writing—review & editing [equal]), Christopher L. Kriedt (Investigation, Validation, Writing—review & editing [equal]), John Zajecka (Investigation, Validation, Writing—review & editing [equal]), David L. Dunner (Investigation, Validation, Writing—review & editing [equal]), João Quevedo (Investigation, Validation, Writing—review & editing [equal]), Yvette Sheline (Investigation, Validation, Writing—review & editing [equal]), Walter Duffy (Investigation, Validation, Writing—review & editing [equal]), Brian J. Mickey (Investigation, Validation, Writing—review & editing [equal]), Mary Stedman (Investigation, Validation, Writing—review & editing [equal]), Gustavo Alva (Investigation, Validation, Writing—review & editing [equal]), Lucian Manu (Investigation, Validation, Writing—review & editing [equal]), Quyen Tran (Data curation, Investigation, Project administration, Validation, Writing—review & editing [equal]), Charles F. Zorumski (Investigation, Validation, Writing—review & editing [equal]), Matthew Macaluso (Investigation, Validation, Writing—review & editing [equal]), Michael Banov (Investigation, Validation, Writing—review & editing [equal]), Cristina Cusin (Investigation, Validation, Writing—review & editing [equal]), Jeffrey I. Bennett (Investigation, Validation, Writing—review & editing [equal]), Hunter Brown (Investigation, Validation, Writing—review & editing [equal]), Jeffrey Way (Data curation [equal], Funding acquisition, Project administration [lead], Validation, Writing—review & editing [equal]), Olivia Shy (Data curation, Formal analysis, Software, Validation, Visualization, Writing—review & editing [equal]), Ying-Chieh (Lisa) Lee (Data curation, Formal analysis, Software, Validation, Visualization, Writing—review & editing [equal]), Richard Hamish McAllister-Williams (Validation, Writing—review & editing [equal]), Roger S. McIntyre (Validation, Writing—review & editing [equal]), and Harold A. Sackeim (Conceptualization, Methodology [lead], Supervision [equal], Validation, Visualization, Writing—original draft, Writing—review & editing [lead]).

Funding

This work was supported by LivaNova, PLC, the developer and manufacturer of the Vagus Nerve Stimulation therapy system. Conducting the study, analyzing the data, and drafting the report were supported by LivaNova, PLC. Conducting the study was also supported by the Centers for Medicare & Medicaid Services. Final approval of the content of this manuscript and the decision to submit it were determined solely by the authors.

Conflicts of interest

C.R.C. has received research support from the American Foundation for Suicide Prevention, Assurex Health, August Busch IV Foundation, Barnes-Jewish Hospital Foundation, LivaNova, National Institute of Mental Health, and the Taylor Family Institute for Innovative Psychiatric Research. He has also consulted for LivaNova. A.J.R. has received consulting fees from Beckley Psytech Inc., Better Up, Inc., Compass Inc., Curbstone Consultant LLC, Emmes Corp., Evecxia Therapeutics, Inc., Holmusk Technologies, Inc., ICON, PLC, Johnson & Johnson (Janssen), LivaNova, MindStreet, Inc., Neurocrine Biosciences Inc., and Otsuka-US; speaking fees from LivaNova and Johnson & Johnson (Janssen); and royalties from Wolters Kluwer Health, Guilford Press, and the University of Texas Southwestern Medical Center, Dallas, TX (for the Inventory of Depressive Symptomatology and its derivatives). He is also named co-inventor on 2 patents: US Patent No. 7795033: Methods to Predict the Outcome of Treatment with Antidepressant Medication, Inventors: McMahon FJ, Laje G, Manji H, Rush AJ, Paddock S, and Wilson AS; and US Patent No. 7906283: Methods to Identify Patients at Risk of Developing Adverse Events During Treatment with Antidepressant Medication, Inventors: McMahon FJ, Laje G, Manji H, Rush AJ, and Paddock S. S.T.A. is a consultant to Genomind, Janssen, LivaNova, Neuronetics, and Sage Therapeutics and has received research support from Compass Pathways and Neuronetics. M.T.B. is a former employee of and current consultant for LivaNova. C.G. is an employee of LivaNova and holds LivaNova stock. M.S.G. has received research support from Abbott, LivaNova, Neurolief, and Magnus Medical. He consults for Abbott, Hospital Corp of America, the Jacob Zabara Family Foundation, Neurolief, and Sooma. P.R.-P. is a consultant for LivaNova, Janssen Pharmaceuticals, Motif Neurotech, and Abbott Neuromodulation. R.M.A. has received research support from LivaNova, Compass Pathways, MindMed, Transcend Therapeutics, Wave Neuroscience, Magnus Medical, Janssen, Kernel, Usona Institute, and Alto Neuroscience. She has served on the advisory board for LivaNova and consulted for Starfish Neuroscience. Z.N. is a consultant to LivaNova, Magnus Medical, and Motif and has also received research support from LivaNova. C.L.K. has no conflicts to disclose. J.Z. receives research support from Boehringer Ingelheim, Compass Pathways, Hoffman-LaRoche, Johnson & Johnson (Janssen), LivaNova, Otsuka, Neurocrine Bioscience, and Sage Therapeutics and has received consulting fees from Alfasigma USA and Johnson & Johnson (Janssen). D.L.D. receives payment for clinical services for a former research patient from LivaNova, is a speaker for Janssen (esketamine nasal spray), and conducts forensic consultations, independent medical evaluations, and legal testimony for various firms. J.Q. has clinical research support from LivaNova, Neumora Therapeutics, and Johnson and Johnson; has been a consultant for LivaNova; and receives copyrights from Artmed Editora, Artmed Panamericana, and Elsevier/Academic Press. Y.S. has no conflicts of interest to declare. W.D. has received research support from Abbott Nutrition, AbbVie, Acadia, Akili, Alkermes, Allergan, Alto Neuroscience, AriBio, Axsome, Biohaven, Bionomics, Clexio, Compass Pathways, Corcept, Corium, Denovo Biopharma, Emalex, GlaxoSmithKline Biologicals, Hoffmann-LaRoche, Intra-Cellular, Ironshore, Janssen, Jazz, LivaNova, Lumos, Merck Sharp & Dohme, Neurocrine Biosciences, NRx, Otsuka, Sage Therapeutics, Sanofi Pasteur, Shire, Sirtsei, Spark Neuro, Sumitomo, Sunovion, and Supernus. He is on a speakers bureau or advisory board or is a consultant for Abbott Neuromodulation, Corium, and LivaNova. B.J.M. received research support from NIH, NSF, Wellcome Leap, PCORI, Health Rhythms, LivaNova, Compass, and Abbott and consulting fees from Inside Edge, VML, Atheneum, Guidepoint, Kx Advisors, and S2N Health. M.S. states the following disclosures: LivaNova, Compass Pathfinder Limited, Neurocrine Biosciences, Neumora Therapeutics, Lilly, and Eisai. G.A. receives research support from AbbVie, Accera, Axsome, Axovant, Biogen, Eisai, Eli Lily, Neurotrope, Genentech, Intra-Cellular, Janssen, Lundbeck, Neurim, Novartis, Otsuka, Roche, Sage, Suven, and TransTech and is on the speakers bureau of and a consultant for AbbVie, Acadia, Alkermes, Axsome, Biogen, Janssen, Idorsia, Lundbeck, Myriad, Neurocrine, Nestle, Otsuka, Sage, Sunovion, Teva, and Takeda. L.M. has received research funding administered through Stony Brook University from Janssen Pharmaceuticals (developer of esketamine) to conduct clinical trials with esketamine, received research funding administered through Stony Brook University from LivaNova (the developer of vagus nerve stimulation [VNS] technology) to conduct clinical trials with VNS, and served as a consultant and member of the Spravato speakers bureau for Janssen Pharmaceuticals. Q.T. is an employee of LivaNova. C.F.Z. served on the Scientific Advisory Board of Sage Therapeutics and had equity in the company. He has received royalties from Oxford University Press and research support from the Taylor Family Institute for Innovative Psychiatric Research and the Bantly Foundation. M.M. discloses the following over the last 24 months: (1) Received research support from Alto, Boehringer Ingelheim, LivaNova, Janssen, Merck, Neurocrine, Otsuka, SAGE, PCORI, and NIH/NIMH. All clinical trial payments were made to the University of Alabama at Birmingham. (2) Served as a paid consultant for the CME Institute, NuSachi Labs, PharmaTher, Residents Medical, Tactical Mind Solutions, and the University of Missouri. (3) Received royalties from Springer Nature for textbooks published. M.B. has the following disclosures to declare: AbbVie, Intra-Cellular, Axsome, Janssen, and Teva. He is also a consultant for LivaNova. C.C. has received research support for conducting clinical trials from AFSP, Clexio, ATAI, and Janssen and consulting fees from Compass Therapeutics and Boehringer. J.I.B. receives research support from Teva Pharmaceutical Industries Ltd, Intra-Cellular Therapeutics, J&J Innovative Medicine, and Relmada Therapeutics for clinical trials, administered through the Southern Illinois University School of Medicine. He has received research support from Janssen Research and Development to conduct clinical trials with esketamine, which are also administered through the Southern Illinois University School of Medicine. H.B. has no conflicts to disclose. J.W. is a former employee of LivaNova and holds LivaNova stock. O.S. is an employee of LivaNova and holds LivaNova stock. Y.-C.(L.)L. is an employee of LivaNova and holds LivaNova stock. R.H.M.-W. reports acting as TSC chair for the NIHR HTA-funded SNAPPER trial and DMC chair for the EU-funded PReDicT study. He is Director of Education for the British Association for Psychopharmacology, receives support for meetings via Janssen-Cilag, and receives payments/consultation fees from LivaNova, Janssen-Cilag, Sage Therapeutics, P1Vital, Takeda, and Lundbeck. R.S.M. has received research grant support from CIHR/GACD/National Natural Science Foundation of China (NSFC) and the Milken Institute; speaker/consultation fees from Lundbeck, Janssen, Alkermes, Neumora Therapeutics, Boehringer Ingelheim, Sage, Biogen, Mitsubishi Tanabe, Purdue, Pfizer, Otsuka, Takeda, Neurocrine, NeuraWell, Sunovion, Bausch Health, Axsome, Novo Nordisk, Kris, Sanofi, Eisai, Intra-Cellular, NewBridge Pharmaceuticals, Viatris, AbbVie, and Atai Life Sciences. H.A.S. serves as a scientific advisor and receives consulting fees from Cerebral Therapeutics, Holmusk Technologies, LivaNova, MECTA Corporation, NeuroInsights, Neurolief, Neuronetics, Parow Entheobiosciences, and SigmaStim; receives honoraria and royalties from Elsevier and Oxford University Press; is the inventor of nonremunerative US patents for focal electrically administered seizure therapy, titration in the current domain in electroconvulsive therapy (ECT), and the adjustment of current in ECT devices, each held by the SigmaStim Corporation; and is also the originator of magnetic seizure therapy.

Data availability

The statistical analyses that are reported in the article will be shared upon reasonable request to the corresponding author. The raw data are not publicly available.