Bed Nucleus of the Stria Terminalis-Nucleus Accumbens Deep Brain Stimulation for Depression: A Randomized Controlled Trial and an Intracranial Physiological Biomarker Predictor

doi:10.21203/rs.3.rs-4854344/v1

Download PDF

Article

Bed Nucleus of the Stria Terminalis-Nucleus Accumbens Deep Brain Stimulation for Depression: A Randomized Controlled Trial and an Intracranial Physiological Biomarker Predictor

https://doi.org/10.21203/rs.3.rs-4854344/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Therapeutic options for refractory depression are urgently needed. We conducted a deep brain stimulation (DBS) randomized controlled trial of the bed nucleus of the stria terminalis (BNST), an extended amygdala structure, and nucleus accumbens (NAc) in 26 refractory depression patients. BNST-NAc DBS had a 50% depression response rate and 35% remission rate in the open-label phase. Stimulation improved depression, anxiety, quality-of-life and disability more than sham in the blinded controlled trial (-9.8, p < 0.001, 95% CI 6.1–13.5). Lower BNST theta, prefrontal-BNST coherence with top-down connectivity predicted better depression outcomes and quality-of-life after chronic stimulation at 3, 6 and 12 months, confirmed using separate data sets and machine learning. We identified a physiology-guided connectivity network involved dorsal anterior cingulate and lateral inferior frontal cortex tracts. These biomarkers, linked to negative emotional bias and anxiety, highlight the efficacy of BNST-NAc DBS for MDD and has potential broader clinical implications.

Health sciences/Biomarkers/Predictive markers

Health sciences/Medical research/Outcomes research

Major depressive disorder (MDD) is one of the most common disabling global mental health problems with high rates of treatment resistance¹. Therapeutic options for refractory depression are urgently needed. DBS is a neurosurgical procedure involving delivery of high frequency stimulation to deep brain structures to modify pathological activity with long-term established efficacy in Parkinson’s disease and movement disorders² and obsessive-compulsive disorder (OCD)³. Deep brain stimulation (DBS) targeting the medial prefrontal (mPFC)-limbic network associated with impaired emotion and reward processing in MDD, have demonstrated potential therapeutic efficacy^4–7.

The randomized controlled trials (RCT) for depression have had mixed results with variability related to issues including targeting precision, study design to allow individual optimization, and the heterogeneity of depression leading to the need to identify predictors of response^5,7. DBS studies targeting the subcallosal cingulate show long term open label efficacy^8,9 although a RCT study was negative⁶ with critical issues related to individualized targeting of relevant limbic tract bundles^10–12 along with potentially inclusion of a heterogeneous depression population^4,13–15. The study design is also critical as the two negative RCT studies targeting the subcallosal cingulate and the ventral internal capsule (VIC) for MDD did not have an initial open label stimulation parameter optimization prior to the RCT which might be critical given inter-individual differences in anatomy and in optimizing stimulation contacts and parameter protocols⁴. In contrast, a positive RCT targeting the VIC for depression allowed for an initial longer open label stimulation parameter optimization prior to the RCT⁷.

The bed nucleus of the stria terminalis (BNST), part of the extended amygdala, has also shown preliminary efficacy as a DBS target for MDD^16–18 and OCD¹⁹. Compelling target rationale exists that the BNST is implicated in sustained fear, stress responses, social behavior and valence surveillance, features relevant to depression^20–26. The BNST is well-placed to play an integrative role in emotional processing and regulation with extensive connections with hypothalamic autonomic, and midbrain neurotransmitter structures²¹. A recent pilot open-label study suggested potential efficacy of BNST DBS although a follow up small randomized controlled trial (n = 8) was not effective^17,27. The lack of efficacy may be related in part to the small study size, lack of individualized optimization of stimulation parameters with an initial fixed cross-over RCT of contacts at low and high frequency and stimulation of the lower two and top two contacts and single target of the BNST.

Here we conducted a DBS RCT (ClinicalTrials.gov identifier: NCT04530942) in 26 MDD patients using dual monopolar stimulation of an 8-contact electrode allowing multitargeting and stimulation of the BNST, nucleus accumbens (NAc) and VIC. To address previous study design issues, we allow at least a 6-month open label stimulation parameter optimization prior to the cross-over RCT (Fig. 1a). We further leveraged the capacity to record from intracranial physiology with high spatial precision and signal-to-noise ratio to derive subtype-sensitive objective biomarkers that predicted treatment efficacy of chronic BNST-NAc-DBS. We then confirmed with machine learning and cross-validated in separate data sets. Utilizing wireless intracranial electrophysiology recording, we extended evidence of the identified BNST dynamics tracking both stimulation effects of BNST-NAc DBS and transient individual mood fluctuations (Fig. 1c). Within an integrated multi-dimensional predictive framework (Fig. 1b), we bridged electrophysiology, neuroimaging and behavior, to interrogate relevant structural and functional connectivity, patterns of network interaction, and potential psychological correlates.

Patient characteristics and study design

Between July 2021 and April 2024, we conducted a prospective, double-blinded crossover RCT at Ruijin Hospital, Shanghai Jiaotong University School of Medicine (Shanghai, China). The primary aim was to evaluate efficacy of BNST-NAc DBS for MDD. Secondary aims included finding predictors of treatment response, establishing an adverse events profile, and assessing effects on quality-of-life. We anticipated enrolling 34 patients and considered the objective met when at least 17 patients entered the randomized double-blind crossover phase, factoring in a dropout rate of 50%.

A total of 524 patients were screened for eligibility, of which 26 were included in the study (Fig. 2). Patient demographic characteristics (88% male; mean age 31.7 years (SD 9.9); mean age of depression onset 18.9 years (SD 7.1) and mean hospitalization 2.5 times (SD 0.5) are further summarized in Extended data Table 1. During the open-label phase, three patients withdrew from the study at months 3, 9, and 10 of optimization (Fig. 2 and Supplementary information). Additionally, five patients were not randomized in the cross-over trial (Fig. 2 and Supplementary information). Consequently, 18 patients entered the crossover phase, with eight randomized to active-sham and 10 to sham-active DBS. All completed the RCT, though 8 terminated early during the sham DBS phase (Supplementary Information). We employed intention-to-treat analyses (ITT) for the outcomes of the RCT.

We obtained data from both acute externalized and chronic wireless electrophysiology to establish predictive biomarkers of treatment response. Subjects were externalized following the initial implantation. During this perioperative period, we acquired 5-minute resting state recordings of local field potential (LFP) and frontal scalp electroencephalogram (EEG) data, both with eyes closed and open. Seventeen patients (eyes-closed dataset) were included following data inspection and confirmation of target BNST localization using Lead-DBS with cross-validation in the eye-open dataset (n = 15) to exclude state bias²⁸.

Postoperatively, we streamed multiple 5-minute rest episodes of intracranial electrophysiological data labeled with momentary assessment of emotional states from 10 patients (8–17 episodes each) over the crossover period.

During the study, patients underwent assessments using standardized measures for depression (Hamilton Depression Scale-17 [HAMD-17], Montgomery-Asberg Depression Rating Scale [MADRS]), anxiety (Hamilton Anxiety Scale-14 [HAMA-14]), anhedonia (Dimensional Anhedonia Rating Scale [DARS]), quality-of-life (World Health Organization Quality of Life-BREF [WHOQOL-BREF], 36-Item Short Form Survey Instrument [SF-36]), and disability (Sheehan Disability Scale [SDS]) at baseline and monthly for at least 6 months during the open-label phase, and during the randomized double-blind crossover phase. An affective task was performed at baseline to capture subjective emotional bias ratings.

Clinical outcomes: BNST-NAc DBS is effective for MDD depression and anxiety

The study started with a longitudinal, open-label trial lasting at least 6 months followed by a double-blind, 4-week randomized crossover design. During the open-label phase, patients were assessed monthly clinically for symptom severity, with adjustments made to their parameters approximately every two weeks as needed. At the last follow-up (LFU), all patients received multi-site monopolar stimulation, where the volume of tissue activated (VTA) predominantly overlapped the BNST, extending to include the NAc and VIC (Extended data Fig. 1a and Extended Data Table 2). Stimulation parameters are summarized in Extended Data Table 3 and depicted in Fig. 3b.

At a cohort level, anxiety/depression and anhedonia were dissociated at baseline with HAMD, MADRS and HAMA highly correlated (p < 0.001) but not with DARS indicating potentially differential clinical phenotypes. While depression, anxiety and anhedonia were correlated in the short term postoperatively at 3-month and 6-month, only depression and anxiety were correlated at 1 year, indicating potentially a distinct long-term response to chronic BNST-NAc DBS (Extended data Fig. 1b).

At the LFU of the open-label phase (19 ± 8.5 months), stimulation led to a mean decrease in HAMD scores of 10.1 (SD 7.1) points (p < 0.001, 95% confidence interval (CI) 7.2–13.0) and in MADRS score of 13.5 (SD 7.8) points (p < 0.001, 95% confidence interval (CI) 10.4–16.6). Based on the HAMD scores at LFU, 13 of 26 patients (50%) were classified as responders, with 9 of them (35%) achieving remission, while the remaining 13 patients (50%) were classified as non-responders, with 7 of them (27%) being partial responders. Additionally, compared to baseline, there were significant reductions in HAMA scores and increases in DARS scores, although the significance of the DARS scores did not hold after multiple comparison correction. Quality-of-life measures (SF-36, WHO-BREF) and disability scores (SDS) were significantly improved (Fig. 3a and Extended Data Table 4).

The study met the primary end point of BNST-NAc DBS efficacy for MDD under blinded, controlled conditions. During the randomized double-blind crossover phase, the percentage of responders in the off-on group and on-off group, based on HAMD scores at T1, were 80% (8/10) and 38% (3/8) respectively. The mean difference in HAMD scores between active and sham stimulation was 9.8 (SD 7.2) (p < 0.001, 95% CI 6.1–13.5). No significant period or carryover effects were observed on HAMD scores. Significant treatment effects were found on MADRS, HAMA but not DARS (Fig. 3b). Notably, a strong carryover effect was observed on the MADRS scores, indicating that the mean MADRS scores of the active and sham phases were higher in the on-off group (29.7 ± 7.1) than in the off-on group (19.1 ± 7.5) (Extended Data Table 5).

The study achieved secondary efficacy outcomes for quality-of-life and disability in the crossover phase, with significant treatment effects observed on quality-of-life measures (WHO-BREF and SF-36) and disability scores (SDS) between active and sham stimulation. Additional secondary measurements, including other depression scales, sleep quality, and adverse events, are detailed in the Supplementary Information and Extended Data Table 6.

Predictive biomarker: BNST theta biomarkers predict treatment response

We then focused on intracranial electrophysiological biomarkers within the prefrontal-limbic circuitry assessing prediction of therapeutic outcomes. To achieve this, we identified spectral features and functional connectivity within the BNST, NAc, hypothalamus, and prefrontal cortex that correlated with DBS efficacy for depression (HAMD, MADRS), anxiety (HAMA) and anhedonia (DARS) at 3 months (n = 17). Significant features (unadjusted p < 0.05) indicative of therapeutic effects included BNST theta, beta and gamma power, prefrontal theta power and prefrontal-BNST theta coherence (Fig. 4a). Hypothalamic and NAc features were unrelated to therapeutic effects. Notably we only found predictive effects and no state related effects.

We then fit the 5-frequency band average power of prefrontal EEG, BNST and their connectivity into a machine learning algorithm using cross-validated penalized ridge regression model with permutation feature selection. Critically, BNST theta power and prefrontal-BNST theta coherence were the two most informative features for the training model (Fig. 4b). Either of them alone was sufficient to predict the 3-month antidepressant and anxiolytic effects of BNST-NAc DBS but not anhedonia, with lower theta power (HAMD: R² = 0.55, p < .001; MADRS: R² = 0.32, p = 0.017; HAMA: R² = 0.54, p < 0.001; DARS: R² = 0.15, p = 0.129) and coherence (HAMD: R² = 0.31, p = 0.021; MADRS: R² = 0.20, p = 0.069; HAMA: R² = 0.30, p = 0.023; DARS: R² = 0.04, p = 0.431) predicting better improvement (Fig. 4c).

Using the eye-open data set, we verified generalizability and cross-validated across a different data set. Our findings successfully replicated the predictive value of BNST theta power but not coherence in the eye-open dataset (Extended Data Fig. 2a & Extended Data Fig. 2b). We also confirmed their long-term predictive value at 6-month (n = 17) and 1-year (n = 15) (Fig. 4c & Extended Data Fig. 2b).

Theta biomarkers involve a top-down prefrontal-BNST network

We investigated the direction of functional connectivity between prefrontal and BNST using frequency resolved Granger prediction. Top-down theta band Granger causality at baseline correlated with theta power (prefrontal EEG: Spearman’s r = 0.51, p = 0.04; BNST: Spearman’s r = 0.72, p = 0.002) and coherence (Spearman’s r = 0.58, p = 0.017) and also predicted depression and anxiety 3-month improvement (HAMA: Spearman’s r = -0.53, p = 0.028; MADRS: Spearman’s r = -0.52, p = 0.031; HAMD: Spearman’s r = -0.53, p = 0.027). This was not observed for bottom-up theta band Granger causality (p > 0.05) (Fig. 5b & Extended Data Fig. 1c).

Next, we asked which prefrontal regions might be implicated in the prefrontal-BNST network based on our physiology-guided tractography analyses. Using fiber filtering in Lead-DBS²⁸, we stratified based on theta power and showed that lower theta power and coherence correlated with fiber tracts projecting between BNST and dorsal anterior cingulate and lateral frontal cortex seeding from clinical BNST stimulated contacts (Fig. 5a & Extended Data Fig. 1d).

The utility of BNST theta biomarker in longitudinal dynamic tracking

Due to the predictive nature of BNST theta-band power for BNST-NAc DBS efficacy, we speculated that DBS may show therapeutic effect by modulating theta-band activity. To elucidate this, we wirelessly streamed multiple 5-minute episodes of intracranial BNST electrophysiological data from 10 patients over the double-blinded crossover phase.

We aimed to determine if high frequency active versus sham DBS could modulate BNST theta-band activity in keeping with their effects on depressive symptoms. Hence, we aligned daily LFP recordings with corresponding HAMD scores and calculated the mean BNST theta-band power for each day. During the washout period, we averaged BNST theta-band power and HAMD scores to account for potential prolonged effects of BNST-NAc DBS. VAS scores and BNST theta power were normalized to enable comparison between patients. We showed that the fluctuations in BNST theta-band activity correlated significantly with the patterns observed in HAMD scores (Spearman’s r = 0.42, p = 0.027) (Fig. 6a). Furthermore, active BNST-NAc DBS significantly reduced BNST theta-band activity compared to sham stimulation (Wilcoxon signed-rank sum test, z = -2.55, p = 0.008) (Fig. 6b).

To further demonstrate the utility and feasibility of BNST theta power to track subjective emotional dynamics, we labeled LFP episodes with momentary assessment of mood and anxiety states. Across 9 patients (110 resting state datasets), BNST theta power correlated positively with anxiety ratings (Spearman’s r = 0.24, p = 0.012) but not with mood ratings (p > 0.05), while anxiety and mood ratings positively correlated with each other (Spearman’s r = 0.369, p < 0.001) (Fig. 6c).

Neuropsychological predictive biomarker: Negative emotional bias

Patients with MDD commonly exhibit biased processing of emotional stimuli, linked to dysfunction in the prefrontal-BNST network^29,30. We hypothesized that the predictive theta biomarker may underlie this negative emotional bias in MDD. To investigate, we assessed subjective emotional bias identifying subjective emotional valence and arousal in response to visual emotional stimuli.

We reaffirmed previous findings indicating that higher neutral-controlled negative arousal correlated with more severe depression/anxiety (HAMD: Spearman’s r = 0.69, p = 0.002; MADRS: Spearman’s r = 0.56, p = 0.021; HAMA: Spearman’s r = 0.62, p = 0.008), with no significant correlation with anhedonia (p > 0.05) (Fig. 5a)^31,32.

On an exploratory basis, we found a state-related baseline behavioral measure, negative emotional arousal bias ((positive-negative)/neutral)), which was correlated with BNST theta power (Spearman’s r = -0.52, p = 0.037) and coherence (Spearman’s r = -0.53, p = 0.029) (Fig. 5a). In other words, lower negative emotional bias or lower arousal to negative versus positive images correlated with lower theta power. Furthermore, this lower bias also predicted better 3-month therapeutic outcomes of all three core symptoms (DARS: Spearman’s r = -0.48, p = 0.052; MADRS: Spearman’s r = 0.56, p = 0.02; HAMA: Spearman’s r = 0.51, p = 0.036), suggesting its comparable predictive utility (Fig. 5a). In contrast, baseline clinical questionnaires did not correlate with baseline physiology or predict treatment outcomes.

We analyzed if BNST theta power or coherence mediated the capacity for negative emotional bias to predict clinical improvement using a mediation model. BNST theta power mediated the relationship between anxiety improvement and negative emotional bias. The direct effect of negative emotional bias on anxiety improvement was insignificant. However, there was a significant indirect effect (b = 0.31, 95%CI: [0.085, 1.198]) via BNST theta power (negative emotional bias to BNST theta power: b = -0.49, p = 0.02; BNST theta power to anxiety improvement: b = -0.63, p = 0.003) (Extended Data Fig. 2b).

Quality-of-life and disability: Contributors and predictors

The cohort observed improved quality-of-life and reduced psychosocial disability after BNST-NAc DBS, along with reductions in depression, anxiety, and anhedonia. We evaluated whether these improvements contributed to overall gains in quality-of-life and psychosocial function. WHOQOL-BREF measurements were associated with better anhedonia outcomes at 3 and 6 months and with better depression, anxiety, and anhedonia outcomes at 1 year. SF-6 measurements were linked to better anxiety and anhedonia outcomes at 3 and 6 months, and better anxiety outcomes at 1 year. SDS measurements correlated with better anxiety and depression outcomes at 3 and 6 months, and improvements in depression, anxiety, and anhedonia at 1 year (Extended Data Fig. 3a).

We then investigated whether physiological theta biomarkers and neuropsychological biomarker could predict quality-of-life improvements or disability outcomes. Using stepwise regression analysis with predictor selection, we found that the neuropsychological biomarker, negative emotional bias, rather than theta biomarkers, predicted disability outcomes measured by SDS at 3 months (R² = 0.45, p = 0.003), 6 months (R² = 0.37, p = 0.009), and 1 year (R² = 0.36, p = 0.017) (Extended Data Fig. 3b). BNST theta power was associated only with 3-month SDS improvements (Spearman’s r = -0.54, p = 0.025). In contrast, baseline BNST theta power correlated with quality-of-life improvements (SF-36: Spearman’s r = 0.58, p = 0.034; WHOQOL-BREF: Spearman’s r = 0.65, p = 0.01) at 1 year, but not at 3 or 6 months.

Thus, in summary, we highlight efficacy of BNST-NAc DBS in improvement of depression, quality-of-life and disability in a randomized controlled cross-over trial. Lower BNST theta and prefrontal-BNST top-down coherence along with a negative arousal bias predicted better longitudinal therapeutic outcomes for depression and anxiety but not anhedonia. BNST-NAc DBS further decreased BNST theta activity.

We demonstrated efficacy of BNST-NAc DBS for treating refractory MDD through a phase II RCT on both antidepressant, quality-of-life and disability outcomes. We then identified state independent, ecologically valid and generalizable theta band electrophysiological biomarkers within the prefrontal-BNST network predictive of depression therapeutic outcomes. Lower BNST and prefrontal theta and prefrontal-BNST coherence along with a psychological measure of negative arousal bias predicted better longitudinal therapeutic outcomes for depression and anxiety but not anhedonia. These findings were cross-validated in a separate data set and corroborated in a longitudinal data set tracking within-subject emotion dynamics. Chronic BNST-NAc DBS further decreased BNST theta activity. Physiologically guided structural connectivity implicated a prefrontal-BNST network focusing on BNST theta power and prefrontal-BNST coherence. Thus, our findings highlight potential efficacy of BNST-NAc DBS for refractory depression particularly targeting depression with comorbid anxiety along with objective physiological and psychological predictive biomarkers and stimulation effects.

Clinical outcomes

Open-label trials of DBS for MDD^9,33 often report an average response rate of 50% across various targets but findings from blinded RCTs have been mixed^5–7. Here, we report a comparable response rate of 50% and remission rate of 35% in the open-label phase, and highlight significant improvement with double-blinded active stimulation versus sham on depression and anxiety and critically on clinically relevant quality-of-life and disability measures. Our findings contrast with the recent small negative RCT of BNST-NAc DBS for depression ²⁷. The differences may be related to the small sample and study design with an initial fixed stimulation RCT along with differences in stimulation targeting²⁷. Notably, we initiated the trial with a minimum six-month open-phase recognizing the potential need for longer parameter optimization before implementing an RCT⁷. In our study, most patients received more than one monopolar stimulation from an 8-contact electrode with active stimulation doses ranging from 3.5v to 5.7v thus providing a much wider spatial and amplitude variability for optimization. The choice of contacts and the stimulation dose were clinically selected and optimized over the greater than 6-month open label phase allowing for expected gradual improvements in stimulation.

Objective cortical-BNST physiological biomarker predicting treatment outcome

Here, our focus extends beyond clinical observations to include biomarker identification of outcome prediction, long-term physiological follow-up, and physiological response to stimulation. Oscillatory activity is highly relevant both mechanistically and in predicting and optimizing DBS treatment for depression^34,35. Evidence from DBS intracranial recordings in PD patients has demonstrated the reliability of aberrant beta band activity within cortical-basal ganglia circuits in discerning disease states and predicting the motor response to DBS³⁶. A recent study identified using machine learning a subcallosal cingulate physiological biomarker tracking recovery states³⁷. Intracranial recordings have also identified alpha desynchronization related to emotional imagery in the subthalamic nucleus and habenula which shift emotional valence bias with time-locked alpha specific stimulation^38,39.

This is the first study to show an objective biomarker highlighting BNST theta and prefrontal-BNST coherence predictive of therapeutic outcome following BNST-NAc chronic stimulation. The BNST acts as an integrated station to orchestrate divergent motivational, arousal and emotional states^20,21,24. Compelling evidence supports a prominent role of theta oscillation in the PFC-limbic circuit underlying depression physiopathology⁴⁰. Rodent studies suggest that theta and gamma oscillations and theta synchrony in a negative affective circuit underlie relevant processes of aversive anticipation, learned fear, and social avoidance⁴¹.

We further use physiology-guided tractography highlighting a role for ventral internal capsule tracts traversing within the stimulated area around the BNST implicating the dorsal anterior cingulate and lateral inferior frontal cortex. The terminus points of the identified tracts converge with optimal therapeutic outcomes identified previously from our MDD sample and from convergent DBS studies for OCD targeting the ventral internal capsule, subthalamic nucleus and BNST^42,43. However, our identified tracts have a more dorsal projection in the anterior cingulate than tracts identified as the optimal therapeutic outcomes for OCD DBS targeting the ventral internal capsule⁴³.

We further highlight a neuropsychological biomarker using baseline negative emotional arousal bias which predicted BNST-NAc DBS depression response. This relationship was mediated by BNST theta activity. Additionally, this biomarker showed superior performance in predicting psychosocial disability outcomes. In contrast, clinical questionnaire baseline measures neither correlated with physiology nor predicted therapeutic outcomes. Together these findings highlight a role for a prefrontal-BNST physiological network and a neuropsychological predictor of antidepressant outcomes from chronic DBS of the BNST-NAc.

The study is not without limitations. First, our findings in the BNST appear to be predictive of depressive outcomes associated with anxiety but not anhedonia suggesting potential differential effects on depression subtypes. Further larger studies are indicated to assess the effects on depressive subtypes. Second, despite patients being blinded to active and sham DBS conditions, subjects were able to explicitly assign the stimulation settings.

Our study demonstrates the efficacy of BNST-NAc DBS in MDD patients and identifies a cross-validated, generalizable predictive biomarker within the prefrontal-BNST circuit. These findings offer direct insights into the role of the prefrontal-BNST circuit in human MDD pathology. The study also has potentially broad clinical relevance. Our findings suggest a physiological and psychological prognostic biomarker with the potential to help guide patient selection, stimulation optimization and tracking symptom severity. These findings might be relevant to depression heterogeneity identifying subgroups responsive to BNST-NAc DBS. Our study also has potential implications for non-invasive stimulation protocols. It sheds light on the antidepressant benefit arising from theta burst stimulation (TBS) over prefrontal areas through top-down regulation⁴⁴. Moreover, our findings pave the way for a personalized, circuit-selective and frequency-based approach for MDD across available multiple intervention modalities. Thus, we highlight identification of predictive biomarkers for treatment with a comprehensive multimodal framework providing a generalizable and extensible approach for biomarker research in translational medicine.

Study Participants

Subjects were recruited from Ruijin Hospital, Shanghai Jiaotong University School of Medicine (Shanghai, China). Inclusion criteria for the surgery were: 1) men and women aged 18-65 years; 2) meeting the international Classification of diseases (ICD)-10 definition of non-psychotic major depressive disorder (MDD), 3) current episode ≥ 2 years duration and/or more than 4 repeated episodes with current episode ≥ 1 year duration and a minimum of 5 years since the onset of the first depressive episode⁴⁵, 4) lack of antidepressant response to a minimum of three antidepressant treatments of adequate dose and duration, including at least two medications from two different classes, failure of adequate psychotherapy and electroconvulsive therapy (ECT) (either poor response, intolerance or rejection), 5) remaining stable with the current anti-depressive medicine for the last month and 6) Hamilton Depression Scale-17 (HAMD-17) score ≥ 17. Exclusion criteria were: 1) schizophrenia or psychosis unrelated to MDD; 2) severe personality disorder, neurological disorders or medical disorders; 3) history of brain surgery; 4) contraindications for anesthesia or stereotactic surgery. We enrolled 26 patients with MDD in the study. All included patients provided written informed consent.

The study was conducted in Ruijin Hospital, Shanghai Jiaotong University School of Medicine. Approval for the trial was obtained from the Ruijin Hospital Ethics Committee, Shanghai JiaoTong University School of Medicine, under the approval number 2021-52. The research use of the implanted components authorized by the Shanghai Testing & Inspection Institute for Medical Devices and approved by China Food and Drug Administration. A total of twenty-six patients were enrolled in the study from March 29th, 2021, to July 28th, 2023.

Surgical Procedure and Stimulation

All patients received bilateral implants of 8 contact electrodes (lead model SR1202-S; SceneRay, Suzhou, China) using stereotactic frame-based, magnetic resonance localization with the middle contact point in the bed nucleus of the stria terminalis (BNST) and a lower contact point in or close to the nucleus accumbens (NAc). The length of each contact is 1.5mm and the spacing between contacts is 0.5mm. Target coordinates for the electrode tip were approximately 4-8mm lateral to the midline, 1-3mm anterior to the anterior border of the anterior commissure and 5-8mm inferior to the anterior commissure. Electrodes were connected via subcutaneous extensions to a stimulator (SR1103, SceneRay).

Deep brain stimulation (DBS) settings were initiated one-week post-lead implantation, with parameter selection guided by immediate responses to stimulation and careful monitoring for transient side effects. Stimulation parameters, including electrode contacts, voltage, pulse width, and frequency, were systematically adjusted approximately every two weeks. Monopolar stimulation configuration was predominantly utilized across the patient cohort. In practice, we increased the amplitude upon observing diminished beneficial effects or limited clinical improvement, potentially reaching 6mA. Subsequently, we tailored frequency and pulse width as needed. If these adjustments were ineffective, we introduced the second and, if required, the third preselected contact for monopolar stimulation.

Study Design of Phase II Trial

The study started with a longitudinal, open-label trial followed by a randomized, double-blind crossover design. Following electrode implantation, patients underwent an initial open phase, during which they were assessed monthly for symptom severity with study questionnaires, with clinical adjustments to stimulation parameters approximately every two weeks. DBS optimization was deemed complete after parameter adjustments maintaining for at least six months.

Following the open label phase, patients were randomly assigned to either receive active DBS for two weeks followed by sham DBS for two weeks (the on-off group), or vice versa (the off-on group). Prior to each active and sham phase, a 2-day washout period was implemented. If deemed necessary by the research team or at the request of the patients, the phase was terminated, with patients progressing to the next phase upon termination. Throughout the crossover phase, medication and DBS parameters remained constant. Patients underwent evaluation at five time points: pre-randomization, following two washout phases, and after each active or sham DBS phase.

Randomization and Blinding

We employed an envelope procedure for randomization, wherein each randomly assigned group was enclosed within a sealed, opaque envelope. An independent researcher sequentially opened these envelopes and conducted the randomization by switching DBS on or off accordingly. Throughout the trial, patients, two assessors (YW, LD, psychiatrists), and other relevant researchers were kept blinded to the stimulation condition. This double-blinding protocol remained in effect until the completion of the crossover phase by the last patient.

Clinical assessments

Depression symptoms were assessed using the HAMD-17, with scores ranging from 0 to 52, where higher scores indicated more severe symptoms⁴⁶. A patient was considered a responder if they exhibited a decrease of at least 50% in their HAMD score compared to baseline. Remission was defined as achieving a HAMD-17 score of less than 8. Additionally, we utilized the Montgomery-Asberg Depression Rating Scale (MADRS), which ranges from 0 to 60, to evaluate depression symptoms⁴⁷. Anxiety severity was evaluated using the Hamilton Anxiety Scale (HAMA), which has a range of 0 to 56⁴⁸. Anhedonia was self-reported using the Dimensional Anhedonia Rating Scale (DARS), ranging from 17 to 85, with higher scores indicating lower levels of anhedonia.

Other outcome measurements included the 16-item Self-Report Quick Inventory of Depression Symptomatology (QIDS-SR16), Depression and Somatic Symptoms Scale (DSSS), World Health Organization Quality of Life-BREF (WHOQOL-BREF), 36-Item Short Form Survey Instrument (SF-36), Sheehan Disability Scale (SDS) and Pittsburgh Sleep Quality Index (PSQI).

The primary outcome measure of the trial was the difference in HAMD scores between the active and sham stimulation phases in the randomized, double-blind crossover phase. Secondary outcome measures were the difference in scores on the other scales between active and sham stimulation phases.

Affective Task

Emotional behavior exhibits a structured organization across two psychophysiological dimensions: valence, varying from negative to positive, and arousal, varying from low to high. The individual assessment of these dimensions can be achieved via an affective task, wherein patients were asked to self-report valence and arousal in response to affective imagery. The affective imagery was sourced from the well-established International Affective Picture System (IAPS) and presented in three distinct conditions: positive, neutral, and negative⁴⁹. Each condition encompassed a set of ten different images, with half rated for valence and half rated for arousal. Following the presentation of each image, participants rated their valence and arousal levels using sliding visual analog scales (VAS) that ranged from 0 to 100. A score of 0 represented very negative (valence) or not exciting at all (arousal), while a score of 100 represented very positive (valence) or very exciting (arousal). The intertrial interval lasted between 1 to 1.5 seconds. The affective task was performed at baseline visit.

Localization

DBS electrode localization for the purposes of physiological analyses utilized Lead-DBS v2 (http://www.lead-dbs.org)²⁸. In brief, postoperative CT images were first linearly co-registered to preoperative MRI and normalized into ICBM 2009b NLIN asymmetric space using the SyN approach implemented in Advanced Normalization Tools. DBS electrodes were then localized using Lead-DBS and warped into the Montreal Neurological Institute (MNI) space using the PaCER algorithm after visual review and refinement of the co-registrations and normalizations.

Volume of Tissue Activated and Connectivity Estimation

Volumes of tissue activated (VTA) were modelled following the Baldermann et al.’s approach⁵⁰. Specifically, the E-Field was estimated using a finite element method on a four-compartment mesh describing local grey and white matter, as well as electrode contact and insulating material. Subsequently, voxel-wised normative structural connectivity seeding from bilateral E-fields were estimated using normative data sets from the Human Connectome Project at Massachusetts General Hospital (32 subjects, multi-shell diffusion-weighted imaging data). E-field values were used as weights to construct connectivity profiles. For each patient, fibers passing through a non-zero voxel of the E-field were extracted from the normative connectome and mapped onto a standardized voxelized volume with 2mm resolution. Each fiber received the weight of the maximal E-field magnitude of its passage and fiber densities were weighted by these values. Spearman rank correlation analysis was then performed to examine the relationship between structural connectivity and the identified physiology in the patient cohort. The uncorrected significant fiber tracts were integrated to construct the correlated map (Rmap).

Perioperative Signal Recordings

Twenty-six patients underwent perioperative signal recording. Nine patients were excluded: three due to data noise exceeding 50%, three for incomplete 3-month follow-up during analysis, and three as we were unable to record from contacts in the BNST region although stimulation remained intact. Consequently, the reported results are based on the remaining 17 patients.

The surgery followed a staged procedure. We recorded signals between 2 and 4 days postoperatively, during the period when the leads were externalized. Resting state recordings were obtained for a minimum of 3 minutes (maximum 5 minutes) while patients sat in a relaxed and wakeful state. These recordings were conducted under two distinct conditions: with eyes open and with eyes closed.

Scalp electroencephalogram (EEG) and local field potential (LFP) data were simultaneously recorded using a BrainAmp MR amplifier (Brain Products) at a 500 Hz sample rate. EEG was obtained from 7 frontal electrodes (Fp1, Fp2, F3, F4, F7, F8, Fz) using the 10-20 placement system. To ensure that the desired signals were not contaminated by blinks or saccades, electrooculogram (EOG) recordings were performed. The data was reference online using a left mastoid electrode. The ground was placed at the apex nasi.

Wireless Data Streaming

Ten patients participated in the daily in-lab electrophysiological recordings and momentary self-report behavioral assessments during the randomized, double-blind crossover phase. One patient was excluded due to insufficient data available during the active stimulation phase, leaving a total of nine patients included in the final analysis.

Patients completed 1-3 trials per day during the washout period or in the post-active and post-sham stimulation phases, while stimulation was turned off. In each trial, participants engaged in 5 minutes of wireless data streaming, followed by self-reporting on their mood or anxiety states using a Visual Analog Scale (VAS). Patients maintained an open-eye resting state throughout the trial. The intertrial interval was more than 4 hours.

Wireless data streaming was sampled at a rate of 415Hz or 1000Hz (for P3). LFPs were sensed in bipolar configuration where a pair of sensing contacts were recorded with one contact referenced to the other. The contact pairs within (defined as less than 1mm of distance from the target) or closest to the BNST were preselected for recording. Patients rated their mood or anxiety states on the VAS immediately after data streaming, where 0 corresponds to “very happy” or “not anxious at all”, 100 corresponds to “very sad” or “very anxious”, and 50 corresponds to “moderate states”.

Signal Processing

All data were analyzed offline using MATLAB R2023a with EEGLAB toolbox v2022.1 and custom-written scripts.

Perioperative LFP data applied offline bipolar referencing to reduce volume conduction effects. The data were high-pass filtered at 1Hz using a 2nd order Butterworth filter (pop_basicfilter) and band-stop filtered at 50Hz and its harmonics (pop_eegfiltnew). Independent Component Analysis (ICA) and then visual inspection were utilized for artifact rejection. Data epochs or electrodes containing artefactual activity (e.g., blinks, saccades, or cardiac activity or line noises) identified during visual inspection were omitted from further analyses. Following artefact rejection, we analyzed a total of 17 patients' eye-closed datasets, yielding a mean duration of 230 ± 55s (range: 140-308s); and a total of 15 patients' eye-open datasets, yielding a mean duration of 227 ± 52s (range: 94-282s).

Based on the atlas-defined lead reconstruction in the MNI space using Lead-DBS v2 toolbox, the LFP contact pairs within or closest to the one of the three brain structures (BNST, NAc, hypothalamus) was selected from each recorded hemisphere for final analysis. The BNST and hypothalamus contacts (n = 17) were all within or <1mm from the structures. The NAc contacts (n = 11) were within <2mm distance. To facilitate subsequent individual-level analysis, we computed the LFP activity in the BNST, NAc, and hypothalamus by averaging across both hemispheres. EEG data were averaged across 7 channels.

Wireless LFPs followed a preprocessing procedure akin to perioperative LFPs, with the addition of band-stop filtering at 40Hz and its harmonics (pop_eegfiltnew) to cancel out the sensing noise. To maintain consistency, wireless data obtained from P3 were down-sampled at 415Hz.

Spectral analysis was performed using the Welch’s power spectrum density (PSD) estimation (pwelch). Non-overlapping windows of 512 data points were analyzed, affording a spectral resolution at ~1Hz. LFP power at frequencies < 4Hz was excluded from further analysis to eliminate the movement artifacts and the influence of 1/f nature of the signal. Our primary area of interest was power within the 4-90Hz range. To mitigate the effects of individual variance in targeting and impedance, we computed the relative (%) powers by expressing the power within this range as a percentage of the total power. We summed power into canonical frequency bands: θ (4–8Hz), α (8–12Hz), lβ (13–20Hz), hβ (21-35Hz), γ (40–90Hz).

We fit spectral power features of regions or functional connectivity between regions into a cross-validated penalized ridge regression model with permutation feature selection to identify the electrophysiological features predictive of treatment response. The modelling pipeline consisted of built-in 3-fold cross-validation for model hyperparameter optimization and repeated 10000 iterations for stability and better performance. Permutation feature importance was then computed for the model based on the decrease in a model score when a single feature value is randomly shuffled.

Non-directed spectral connectivity was computed using coherence analysis. The continuous data were divided into 10-second epochs with a 50% stepwise overlap. The magnitude-squared coherence was then calculated for each epoch using the Welch's overlapped averaged periodogram method with blocks of 512 samples, a Hanning window, a 50% overlap, and frequency resolution of ~1Hz in 4-90Hz range (mscohere) and scaled by 100 for expressing percentages. The coherence values were summed across frequency bins and then averaged across epochs. These averaged values were subsequently grouped into five predefined frequency bands.

Directional spectral connectivity (estimates of both A -> B and B -> A connectivity) was estimated using spectrally resolved Granger causality⁵¹. This analysis utilized the MVGC multivariate Granger causality toolbox. To maintain data stationarity, LFP signals were initially segmented into 4-second time-windows with 50% overlap. A multivariate vector autoregressive (MVAR) model was fit to the full “universe of data” time series, with model order optimized using Bayesian information criterion (BIC). The corresponding MVAR model parameters for the selected model order was estimated and the resulting model was checked for stability. Then, the frequency conditional spectral causalities were computed for signal pairs and the significance of the resulting causalities was tested via nonparametric permutation testing. The subsequent analysis focused on the summed frequency-based connectivity within the theta (4-8Hz) band, based on the prior findings.

Statistical Analysis

The sample size was based on the assumption that a mean reduction of 9 points on HAMD score with a standard deviation of 8 when compared active DBS with sham DBS⁷. Consequently, it was estimated that 17 patients in the crossover phase would be sufficient to evaluate the efficacy of the trial with a type I error rate of 0.05 and a type II error rate of 0.9.

For the open-label outcome analysis, participants missing one or two assessments at months 3, 6, or 12 had data from adjacent month visits used. In the crossover phase outcome analysis, participants who terminated early during sham stimulation included data at the time of exit. For participants transitioning directly from the sham DBS phase to active DBS without a washout, scores during the second washout phase were considered equivalent to those during the sham DBS phase.

Data are presented as mean (standard deviation, SD) in tables and as mean (standard error, SE) in figures, and statistical significance was set at a 2-tailed 5% level. Data normality was determined by Shapiro-Wilk test. Correlational analyses were performed utilized Spearman rank correlation or Pearson correlation. Group differences were inferred using Wilcoxon signed-rank sum test or paired student t-tests. Linear regression was used to model the relationship between a dependent variable and one or more independent variables using the stepwise method. We applied false discovery rate (FDR) correction to account for multiple comparisons where necessary.

We performed intention-to-treat analyses (ITT) to results in this prospective randomized controlled trial (RCT). In the open-label phase, outcome measures were analyzed using paired t-tests with a Bonferroni correction applied. Descriptive statistics included the percentage of responders, non-responders, and remission cases. In the crossover phase, analyses of the primary efficacy outcome involved testing three effects: carryover, period (T3 and T5), and treatment (active and sham) effects. This was accomplished using a general linear mixed model analysis, with two-paired HAMD scores as the dependent variables, group (on-off and off-on group) and treatment as independent variables, and HAMD scores at T1 as a covariate⁵². Carryover effects were assessed examining the between-group main effect. Period effects were assessed through the interaction term of treatment × group effects. Additionally, treatment effects were assessed using the within-group main effects. A similar procedure was applied to analyze other measures.

Mediation model was used to test the indirect effect of negative emotional bias on clinical outcomes via theta biomarkers. It was performed using Model 4 in the PROCESS Marco (SPSS). Note that the negative emotional bias did not follow a normal distribution and therefore a pre-applied rank-based inverse normal transformation was performed. The mediation model investigated two paths: the indirect path from (a) negative emotional bias to theta biomarkers and (b) theta biomarkers to clinical outcomes, as well as the direct path from (c’) negative emotional bias to clinical outcomes. Model fit was evaluated based on a non-significant X² (p > 0.05). The indirect effects were estimated using 5000-sample bootstrap procedure with bias-corrected confidence intervals.

Data analyses were performed using SPSS 26.0, Python 3.11, or Matlab R2023a. The figures were generated using R version 4.3.1.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The data supporting the results of this study are available upon reasonable request from the corresponding authors. The protocol and statistical analysis plan for the clinical trial have been uploaded to ClinicalTrials.gov.

Code availability

The code used to produce the results in this paper is available at https://github.com/Fallenworm/BNSTDBS.

Acknowledgements

This study is funded by the STI 2030-Major Projects (No. 2021ZD0200407 [to VV]), the National Natural Science Foundation of China (Grant No. 82271515 [to BS], 81971294 [to DL] and T2250710686 [to VV]), the Science and Technology Commission of Shanghai Municipality (Grant No. 20410712000 [to DL]), Medical Research Council Senior Clinical Fellowship (Grant No. MR/W020408/1 [to VV]), the SJTU Trans-med Awards Research (Grant No. 2019015 [to BS]), the Scientific and technological innovation action plan of Shanghai (Grant No. KY20211478 [to BS]), the Shanghai Municipal Science and Technology Major Project (Grant No. 2021SHZDZX [to BS]), and the Nursing Development Program of Shanghai Jiao Tong University School of Medicine (Grant No. SJTUHLXK2022 [to XQ]).

Author contributions

BS and VV initiated this work. BS and VV supervised the study. LB, YZ and VV drafted the manuscript. LW, YZ, QD, YZ, YW, LD, KY, JH, XZ, PH, XQ, DL, SZ, WL, and YP collected data for the study, VV, LW, BS, YL, XL and JL designed the study. LW and YZ conducted the statistical analysis and verified the data reported in the manuscript. LW and DQ performed the imaging analysis. LW, AM and VV designed the behavioral task. LW and LM set up the perioperative and wireless signal recording platform. All authors contributed to finalizing the manuscript and reviewed the final version. BS, LW, YZ and YW are equal contributors listed as first coauthors, YL, BS, LW and VV are corresponding authors.

Competing interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

World Health Organization. Depression and Other Common Mental Disorders: Global Health Estimates. (WHO, 2017).
Deuschl, G. et al. A Randomized Trial of Deep-Brain Stimulation for Parkinson’s Disease. New England Journal of Medicine 355, 896–908 (2006).
Visser-Vandewalle, V. et al. Deep brain stimulation for obsessive–compulsive disorder: a crisis of access. Nat Med 28, 1529–1532 (2022).
Johnson, K. A., Okun, M. S., Scangos, K. W., Mayberg, H. S. & de Hemptinne, C. Deep brain stimulation for refractory major depressive disorder: a comprehensive review. Mol Psychiatry (2024) doi:10.1038/s41380-023-02394-4.
Dougherty, D. D. et al. A Randomized Sham-Controlled Trial of Deep Brain Stimulation of the Ventral Capsule/Ventral Striatum for Chronic Treatment-Resistant Depression. Biol Psychiatry 78, 240–248 (2015).
Holtzheimer, P. E. et al. Subcallosal cingulate deep brain stimulation for treatment-resistant depression: a multisite, randomised, sham-controlled trial. Lancet Psychiatry 4, 839–849 (2017).
Bergfeld, I. O. et al. Deep Brain Stimulation of the Ventral Anterior Limb of the Internal Capsule for Treatment-Resistant Depression: A Randomized Clinical Trial. JAMA Psychiatry 73, 456–64 (2016).
Kennedy, S. H. et al. Deep brain stimulation for treatment-resistant depression: follow-up after 3 to 6 years. Am J Psychiatry 168, 502–10 (2011).
Mayberg, H. S. et al. Deep Brain Stimulation for Treatment-Resistant Depression. Neuron 45, 651–660 (2005).
Riva-Posse, P. et al. A connectomic approach for subcallosal cingulate deep brain stimulation surgery: prospective targeting in treatment-resistant depression. Mol Psychiatry 23, 843–849 (2018).
Riva-Posse, P. et al. Defining Critical White Matter Pathways Mediating Successful Subcallosal Cingulate Deep Brain Stimulation for Treatment-Resistant Depression. Biol Psychiatry 76, 963–969 (2014).
Lujan, J. L. et al. Tractography-Activation Models Applied to Subcallosal Cingulate Deep Brain Stimulation. Brain Stimul 6, 737–739 (2013).
Figee, M. & Mayberg, H. The future of personalized brain stimulation. Nat Med 27, 196–197 (2021).
Drysdale, A. T. et al. Resting-state connectivity biomarkers define neurophysiological subtypes of depression. Nat Med 23, 28–38 (2017).
Chen, D. et al. Neurophysiological stratification of major depressive disorder by distinct trajectories. Nature Mental Health 1, 863–875 (2023).
Raymaekers, S., Luyten, L., Bervoets, C., Gabriëls, L. & Nuttin, B. Deep brain stimulation for treatment-resistant major depressive disorder: a comparison of two targets and long-term follow-up. Transl Psychiatry 7, e1251–e1251 (2017).
Fitzgerald, P. B. et al. A pilot study of bed nucleus of the stria terminalis deep brain stimulation in treatment-resistant depression. Brain Stimul 11, 921–928 (2018).
Neumann, W. J. et al. Different patterns of local field potentials from limbic DBS targets in patients with major depressive and obsessive compulsive disorder. Mol Psychiatry 19, 1186–1192 (2014).
Mosley, P. E. et al. A randomised, double-blind, sham-controlled trial of deep brain stimulation of the bed nucleus of the stria terminalis for treatment-resistant obsessive-compulsive disorder. Transl Psychiatry 11, 190 (2021).
Jennings, J. H. et al. Distinct extended amygdala circuits for divergent motivational states. Nature 496, 224–228 (2013).
Marcinkiewcz, C. A. et al. Serotonin engages an anxiety and fear-promoting circuit in the extended amygdala. Nature 537, 97–101 (2016).
Li, H. et al. Neurotensin orchestrates valence assignment in the amygdala. Nature 608, 586–592 (2022).
Kim, S. Y. et al. Diverging neural pathways assemble a behavioural state from separable features in anxiety. Nature 496, 219–223 (2013).
Giardino, W. J. et al. Parallel circuits from the bed nuclei of stria terminalis to the lateral hypothalamus drive opposing emotional states. Nat Neurosci 21, 1084–1095 (2018).
Yang, B., Karigo, T. & Anderson, D. J. Transformations of neural representations in a social behaviour network. Nature 608, 741–749 (2022).
Mei, L., Yan, R., Yin, L., Sullivan, R. M. & Lin, D. Antagonistic circuits mediating infanticide and maternal care in female mice. Nature 618, 1006–1016 (2023).
Fitzgerald, P. B. et al. No Consistent Antidepressant Effects of Deep Brain Stimulation of the Bed Nucleus of the Stria Terminalis. Brain Sci 14, 499 (2024).
Horn, A. et al. Lead-DBS v2: Towards a comprehensive pipeline for deep brain stimulation imaging. Neuroimage 184, 293–316 (2019).
Disner, S. G., Beevers, C. G., Haigh, E. A. P. & Beck, A. T. Neural mechanisms of the cognitive model of depression. Nat Rev Neurosci 12, 467–477 (2011).
Fan, X. et al. Brain mechanisms underlying the emotion processing bias in treatment-resistant depression. Nature Mental Health 2, 583–592 (2024).
Wenzler, S. et al. Intensified emotion perception in depression: Differences in physiological arousal and subjective perceptions. Psychiatry Res 253, 303–310 (2017).
Disner, S. G., Beevers, C. G., Haigh, E. A. P. & Beck, A. T. Neural mechanisms of the cognitive model of depression. Nat Rev Neurosci 12, 467–477 (2011).
Figee, M. et al. Deep Brain Stimulation for Depression. Neurotherapeutics 19, 1229–1245 (2022).
Scangos, K. W. et al. Closed-loop neuromodulation in an individual with treatment-resistant depression. Nat Med 27, 1696–1700 (2021).
Scangos, K. W., Makhoul, G. S., Sugrue, L. P., Chang, E. F. & Krystal, A. D. State-dependent responses to intracranial brain stimulation in a patient with depression. Nat Med 27, 229–231 (2021).
Kühn, A. A. et al. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain 127, 735–746 (2004).
Alagapan, S. et al. Cingulate dynamics track depression recovery with deep brain stimulation. Nature (2023) doi:10.1038/s41586-023-06541-3.
Mandali, A. et al. Acute Time-Locked Alpha Frequency Subthalamic Stimulation Reduces Negative Emotional Bias in Parkinson’s Disease. Biol Psychiatry Cogn Neurosci Neuroimaging 6, 568–578 (2021).
Sonkusare, S. et al. Power signatures of habenular neuronal signals in patients with bipolar or unipolar depressive disorders correlate with their disease severity. Transl Psychiatry 12, 72 (2022).
Smart, O. L., Tiruvadi, V. R. & Mayberg, H. S. Multimodal Approaches to Define Network Oscillations in Depression. Biol Psychiatry 77, 1061–1070 (2015).
Likhtik, E. & Paz, R. Amygdala-prefrontal interactions in (mal)adaptive learning. Trends in Neurosciences vol. 38 158–166 Preprint at https://doi.org/10.1016/j.tins.2014.12.007 (2015).
Lai, Y. et al. Structural and functional correlates of the response to deep brain stimulation at ventral capsule/ventral striatum region for treatment-resistant depression. J Neurol Neurosurg Psychiatry 94, 379–388 (2023).
Li, N. et al. A Unified Functional Network Target for Deep Brain Stimulation in Obsessive-Compulsive Disorder. Biol Psychiatry 90, 701–713 (2021).
Solomon, E. A. et al. Theta-burst stimulation entrains frequency-specific oscillatory responses. Brain Stimul 14, 1271–1284 (2021).
Bewernick, B. H. et al. Nucleus Accumbens Deep Brain Stimulation Decreases Ratings of Depression and Anxiety in Treatment-Resistant Depression. Biol Psychiatry 67, 110–116 (2010).
Zheng, Y. et al. Validity and Reliability of the Chinese Hamilton Depression Rating Scale. British Journal of Psychiatry 152, 660–664 (1988).
Montgomery, S. A. & Åsberg, M. A New Depression Scale Designed to be Sensitive to Change. British Journal of Psychiatry 134, 382–389 (1979).
Hamilton, M. The Assessment of Anxiety States by Ratings. British Journal of Medical Psychology 32, 50–55 (1959).
Lang PJ, Bradley MM & Cuthbert BN. International Affective Picture System (IAPS): Technical Manual and Affective Ratings. (NIMH Center for the study of Emotion and Attention, 1997).
Baldermann, J. C. et al. Connectivity Profile Predictive of Effective Deep Brain Stimulation in Obsessive-Compulsive Disorder. Biol Psychiatry 85, 735–743 (2019).
Barnett, L. & Seth, A. K. The MVGC multivariate Granger causality toolbox: A new approach to Granger-causal inference. J Neurosci Methods 223, 50–68 (2014).
Dı́az-Uriarte, R. Incorrect analysis of crossover trials in animal behaviour research. Anim Behav 63, 815–822 (2002).

There is NO Competing Interest.

Download PDF

Version 1

posted

You are reading this latest preprint version

Bed Nucleus of the Stria Terminalis-Nucleus Accumbens Deep Brain Stimulation for Depression: A Randomized Controlled Trial and an Intracranial Physiological Biomarker Predictor

Status:

Version 1

Abstract

Figures

Main

Results

Patient characteristics and study design

Clinical outcomes: BNST-NAc DBS is effective for MDD depression and anxiety

Predictive biomarker: BNST theta biomarkers predict treatment response

Theta biomarkers involve a top-down prefrontal-BNST network

The utility of BNST theta biomarker in longitudinal dynamic tracking

Neuropsychological predictive biomarker: Negative emotional bias

Quality-of-life and disability: Contributors and predictors

Discussion

Clinical outcomes

Objective cortical-BNST physiological biomarker predicting treatment outcome

Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1