Neurostimulation therapies for treatment resistant depression: A focus on vagus nerve stimulation and deep brain stimulation

Abstract

Antidepressant treatments, including pharmacotherapy and psychotherapy, do not result in remission for the majority of patients with major depressive disorder. The high prevalence of treatment resistant depression (TRD) poses a significant issue for patients as well as both societal and economic costs. Due to the limited efficacy of existing therapies in this sub-population, alternative somatic treatments are being explored. Both vagus nerve stimulation (VNS) and deep brain stimulation (DBS) are neurostimulation treatments for TRD. While VNS has Food Drug Administration approval as an adjunctive therapy for MDD, DBS is still in the experimental stages. This article will review the evidence supporting the clinical utility of these therapies.

Treatment resistant depression

In the last ten years, focused somatic treatments for depression have been developed for major depressive disorder (MDD), reflecting the limited therapeutic advances with conventional antidepressants and the emergence of safe and effective neurostimulation therapies. This article will review emerging clinical and preclinical evidence to support vagus nerve stimulation (VNS) and deep brain stimulation (DBS) as evidence-based therapies for treatment resistant depression (TRD).

It is difficult to assess the prevalence of TRD as there is no consensus on a universal definition. Although the majority of reports define TRD as a failure to respond to at least two adequate treatment trials within the index depressive episode (Souery et al., 1999), several other definitions have been proposed. Thase and Rush (1997) suggested a staging method, where patients failing to respond to a first adequate antidepressant trial would meet criteria for stage 1 resistance, with incremental gradations of resistance requiring failures to respond to additional treatments including tricyclic antidepressants (TCAs), monoamine oxidase inhibitors (MAOIs) and electroconvulsive therapy (ECT) (see Table I). An alternative staging method applies differential weighting based on the number of trials and modalities, with failure to respond to ECT earning a higher resistance score than failure with selective serotonin reuptake inhibitors (SSRIs) (Fava et al., 2003). However, these TRD definitions do not integrate the current strategies of sequential and combination antidepressant therapies, nor do they include emerging neurostimulation therapies.

Table I. Staging method for TRD (Thase & Rush, 1997).

TRD stage	Criteria
1	Failure of adequate trial of one class
2	Failure of two distinct classes
3	Stage 2 plus failure of third class including TCA
4	Stage 3 plus failure of MAOI
5	Stage 4 plus failure of ECT

The sequenced treatment alternatives to relieve depression (STAR*D), a large effectiveness trial of antidepressants under ‘real world’ conditions provides a recent estimate of the prevalence of TRD. During the first stage of treatment with citalopram monotherapy, 28% of MDD patients achieved remission, defined by a score of 7 or less on the 17-item Hamilton Rating Scale for Depression (HRSD-17). After three subsequent trials designed to evaluate the efficacy of various switch and ‘add-on’ strategies, 33% of patients had failed to achieve remission (Warden et al., 2007), conveying the high prevalence of TRD. The inevitable consequence of TRD is a substantial societal and economic impact, with a cost burden that is approximately twice that of non-resistant MDD (Gibson et al., 2010; Ivanova et al., 2010).

Methods

To ensure all clinical trials and case studies were retrieved, the following broad search criteria were used in PubMed: ‘vagus nerve stimulation’ or ‘deep brain stimulation’ and ‘depression’ or ‘major depression’. Further searches were done with terms related to each of the neuroanatomical targets for DBS: ‘deep brain stimulation’ and one of ‘subgenual cingulate’, ‘Brodmann Area 25’, ‘subcallosal cingulate’, ‘nucleus accumbens’, ‘internal capsule’, ‘ventral striatum’, ‘lateral habenula’, ‘inferior thalamic peduncle’. For animal and mechanism of action studies, the above search was rerun with ‘animals’ selected as a limit, and then again separately with no limit and each of ‘preclinical’, ‘animal’, or ‘rats’ included in the search criteria. All preclinical and clinical trial data related to VNS or DBS for depression were selected and reviewed. The reference list of data was also compared with relevant review papers to ensure no key publications were missed.

Neurostimulation therapies

Electroconvulsive therapy (ECT) was first used to treat psychiatric disorders in the 1930s. It remains a first line therapy for depressed patients who have severe suicidal ideation or psychotic features. ECT is also indicated for the treatment of MDD without psychotic features, typically when several antidepressants have been tried at adequate doses and for an adequate time. Despite favourable outcomes, its use is limited by patient perceptions, device and anaesthetist availability, lack of target specificity and side effects – particularly short-term amnesia, and by difficulties in ensuring sustainability of the treatment effect (Kennedy & Giacobbe, 2007).

In the last decade, alternative somatic treatments have emerged as viable options for resistant depression, including VNS and DBS. In both cases, efficacy and safety of these interventions in the treatment of neurological disorders has already been established as adjunct therapies for refractory seizure disorders and movement disorders, respectively. Guidelines for recommended use of neurostimulation therapies in the treatment of depression are available (Table II) (Kennedy et al., 2009).

Table II. CANMAT guidelines for neurostimulation (Kennedy et al., 2009b with permission from JAD).

	Overall recommendation	Acute efficacy	Relapse prevention	Safety and tolerability
ECT	First-line	Level 1	Level 1	Level 1
	Second-line for treatment resistant or intolerant populations	Level 3	Level 3	Level 3
rTMS	Second-line	Level 1	Level 3	Level 1
VNS	Third-line	Level 3	Level 2	Level 2
DBS	Investigational	Level 3	Level 3	Level 3

Level 1:>2 RCTs and/or meta-analysis with narrow confidence intervals; Level 2:> 1 RCTs and/or meta-analysis with wide confidence intervals; Level 3: Non-randomized, controlled prospective studies, or case series or retrospective studies; Level 4: Expert opinion/ consensus.

Vagus nerve stimulation

The first VNS surgeries were performed in the 1980s for the treatment of epilepsy. The surgery involves wrapping an electrode around the left vagus nerve, which is connected to a subclavicularly implanted pulse generator that sends low frequency, intermittent electrical signals to the nerve. Stimulation parameters are adjusted with a peripheral programming device. A decade later, observations of enhanced mood and social functioning in patients who received the treatment for epilepsy began to emerge (Handforth et al., 1998; Rawlins, 1997). These observations were supported by preclinical studies demonstrating that activation of the vagus nerve altered brain areas implicated in depression. Particularly, afferent fibres of the vagus nerve send sensory information to the nucleus tractus solitarius (NTS), which relays this information directly and indirectly to key limbic and cortical structures, including the dorsal raphe, locus coeruleus, amygdala, hypothalamus, and cortical regions (Nemeroff et al., 2006).

FDA approval for VNS was received in 2005 as an adjunct therapy for MDD, based on a series of clinical trials (Nahas et al., 2005; Rush et al., 2000; Sackeim et al., 2001a). This was a controversial decision, due to the lack of evidence of acute efficacy in the randomized sham controlled trial (Shuchman, 2007).

Clinical data

Although response rates with VNS have been modest in the acute phase of treatment, open label extension phases suggest an increased effect over time. In the first report of VNS for TRD, 30 patients received VNS under open-label conditions (Rush et al., 2000): the response rate at 10 weeks (defined as a 50% drop on the HRSD-28) was 40% and the remission rate (defined as an HRSD-28 total item score < 10) was 17%. The most encouraging aspect of this trial was a sustained response rate of 40% and remission rate of 29% after a 9 month follow-up (Marangell et al., 2002). After 2 years, the response rate was 42–53% (Nahas et al., 2005), suggesting that the antidepressant effects of VNS require a longer trajectory. In general, this pattern of response has been replicated by other investigators in open-label trials (see Table III).

Table III. Response and remission rates across VNS open-label trials.

Study duration	Response rate	Remission rate
10 weeks	30–40^ab	17^a
3 months	14–37^cde	15–17^de
6 months	18–55^cdf	31^f
12 months	27–47^cdefghi	14–29^efgj
24 months	42–53^cej	22–39^ej

Remission based on scale used: HRSD-28≤10, HRSD-24≤9; Response based on 50% decrease in depression score from baseline on scale used to define remission.

^aRush et al., 2000; ^bSackeim et al., 2001; ^cNierenberg et al., 2008; ^dSchlaepfer et al., 2008; ^eNahas et al., 2005, ^fMarangell et al., 2002; ^gCristancho et al., 2011; ^hGeorge et al., 2005; ⁱRush et al., 2005; ^jBajbouj et al., 2010.

To date, there is only one published randomized sham controlled trial of VNS for TRD and it did not produce a positive result. In a 10-week RCT involving 235 patients (210 unipolar and 25 bipolar), response rates for active and sham treatment did not differ (15.2% versus 10%) (Rush et al., 2005). A post hoc analysis of the differences in treatment efficacy between unipolar and bipolar patients also revealed no differences at 1 and 2 years (Nierenberg et al., 2008). Patients with treatment resistant rapid cycling bipolar disorder (n = 9) have also been assessed in a 12 month open-label trial. There was a 38% improvement in overall illness (depression + mania symptom scores) (Marangell et al., 2008).

Safety data demonstrate that VNS has minimal side effects. The most common adverse events are hoarseness and voice alteration (Rush et al., 2005; Schlaepfer et al., 2008a), symptoms that may compromise the ‘blindedness’ of both patients and clinicians in the setting of an RCT. Other common symptoms reported are nausea, dizziness, shortness of breath, pain, and anxiety (Bajbouj et al., 2010; Cristancho et al., 2011).

The rate of treatment-emergent suicidality, from combined studies up until 2006, was 3.5%, with a completed suicide rate of 0.4%. However, in a subsequent trial, suicide occurred in 3% of the sample with two deaths in 74 patients, which occurred within the first year of treatment (Bajbouj et al., 2010). Importantly, while these rates are higher than those from antidepressant trials, they are lower than expected suicide rates for a severely depressed population (Bostwick & Pankratz, 2000; Guze et al., 1970; Hammad et al., 2006).

The effect of VNS on cognition was assessed in one trial over the domains of motor speed, psychomotor function, language, attention, memory and executive function in 27 TRD patients before and after 10 weeks of VNS. Overall, no deterioration across any of the measures tested was found. Rather, improvements on tests of psychomotor function, memory and executive function post-VNS demonstrated correlations with reductions on HRSD-28 depression scores (Sackeim et al., 2001b).

Preclinical data: Depression models

Animal models of depression provide additional theoretical support for VNS. In one study, rats were administered 30 min per day of continuous VNS for 4 days, and were then subjected to the forced swim test (FST), an established animal model for depression. VNS significantly reduced immobility time compared to unstimulated controls, reflective of antidepressant effects (Krahl et al., 2004). When compared to desipramine, a standard TCA, both VNS and desipramine significantly decreased immobility in the FST (Cryan et al., 2005). Interestingly, VNS-induced decreases in immobility were associated with increased swimming behaviour, which has been linked to a predominantly serotonergic mechanism of action (Cryan et al., 2005). In a subsequent controlled trial, rats received desipramine or active VNS for 2 h at three time points over 1 day, prior to undergoing the FST. Both active treatments were superior to the saline control (Cunningham et al., 2008). However, chronic VNS for 1 month failed to show any behavioural alterations in rats on the FST or the elevated plus maze test, in contrast to treatment with imipramine (Biggio et al., 2009)

Mechanism of action

Although the mechanism of action of VNS is still unclear, several studies suggest that its antidepressant effects may be due to changes in monoamine and neurotrophic function in specific brain areas.

Monoamines

The firing rates of norepinephrine (NE) and serotonin (5-HT) neurons, in the locus coeruleus (LC) and dorsal raphe nucleus (DRN), respectively, increase with VNS treatment (Dorr & Debonnel, 2006). However, firing rates in the LC were greater after short-term (1 h and 3 days) and long-term (14, 21, 90 days) treatments, while DRN firing increased only after 14 days, suggesting that increases in NE firing result in a secondary augmentation of 5HT firing in the DRN due to the excitatory input from the LC to the DRN. These findings have been replicated in subsequent studies (Manta et al., 2009a, 2009b) with additional data demonstrating that the increase in 5HT firing was abolished with lesioning of NE neurons (Manta et al., 2009a). Furthermore, VNS effects on 5HT firing may follow an inverted u-shape, where too little or excessive charge can lead to the loss of 5HT firing (Manta et al., 2009b). Increases in NE firing in the hippocampus, through activation of LC beta-adrenergic receptors, has also been reported (Shen et al., 2011).

Human studies of CSF metabolites have yielded findings inconsistent with the preclinical evidence. Metabolites of NE, 5HT, and dopamine (DA) were assessed before and after 24 weeks of VNS: only increases in the DA metabolite homovanillic acid (HVA) were observed, suggesting a sustained dopaminergic mechanism with VNS. Interestingly, a higher baseline ratio of HVA to 5-hydroxyindoleacetic acid, the metabolic precursor to 5-HT, predicted inferior clinical outcome (Carpenter et al., 2004).

Neurogenesis

Both acute and chronic VNS have been reported to induce persistent changes in hippocampal neurons. Rats receiving 0.75 mA VNS for 2 days showed a 50% increase in the uptake of bromodeoxyuridine (BrdU) in neurons in the dentate gyrus, reflecting progenitor proliferation (Revesz et al., 2008). This effect was not observed at 0.5 mA or 1.5 mA, suggesting a dose-dependent effect of VNS, consistent with other effects including 5-HT firing (Manta et al., 2009b). In a subsequent study, increases in BrdU were observed with acute (3 h) and chronic (1 month) VNS, while increases in BDNF cells were only observed after chronic treatment (Biggio et al., 2009). However, these authors did not demonstrate any behavioural changes suggestive of antidepressant effects.

Brain activation

Several studies have been conducted in both animals and humans to demonstrate the effects of VNS on brain activation. Using markers of short-term (c-Fos) and long-term (DeltaFosB) neuronal activation, brain regions activated by acute (2 h) or chronic (3 weeks) VNS in conscious rats were identified (Cunningham et al., 2008). Acute VNS significantly increased c-Fos staining in several key mood-regulating areas of the brain, namely the NTS, paraventricular nucleus of the hypothalamus, parabrachial nucleus, ventral bed nucleus of the stria terminalis, and LC. The cingulate cortex and DRN did not show activation; however, the time course of this study is consistent with that of the electrophysiological work by Manta and colleagues (2009b) who showed that DRN neurons are not affected in the short term by VNS. Chronic VNS significantly increased DeltaFosB and c-Fos staining bilaterally in each region affected by acute VNS as well as in the cingulate cortex and DRN.

Human brain imaging studies parallel some of these findings. In a small sample of 8 TRD patients, fluoro-deoxy-glucose uptake was measured using positron emission tomography (PET) before and after 1 year of VNS. The most significant change was a decrease in activation within the ventromedial prefrontal cortex extending from the subgenual cingulate to the frontal pole (Pardo et al., 2008). However, glucose uptake after 10 weeks of VNS in seven patients revealed increases in the orbitofrontal cortex, left amygdala, parahippocampal gyrus, insula and the right cingulate gyrus, and decreases in the cerebellum and fusiform gyrus were noted (Chae et al., 2003). These findings are consistent with preliminary PET regional cerebral blood flow (rCBF) data assessing the effect of VNS in four patients with the stimulator sequentially activated and deactivated. VNS-induced increases were observed in the orbitofrontal cortex, bilateral anterior cingulate cortex, and the right superior and medial frontal cortex (Conway et al., 2006). These studies demonstrate that short-term brain activation changes with VNS may normalize over the course of 1 year.

Single photon emission computed tomography (SPECT) studies in VNS have resulted in somewhat divergent findings. In a small study involving five patients, responders to VNS had an increase in anterior cingulate, thalamus, and anterotemporal rCBF, while non-responders had no significant changes from baseline after 10 weeks (Zobel et al., 2005). A subsequent study in 15 patients before and after 10 weeks of VNS, reported increases in the left ventral prefrontal cortex and dorsomedial prefrontal cortex, and decreases in the right posterior cingulate gyrus and left insula following treatment (Kosel et al., 2011). While these findings do overlap to some extent with the PET data, the small sample sizes of the studies, differences in imaging methods, combined with different scan timing and stimulation parameters limit firm conclusions.

Emerging mechanisms

Substance P is a neuropeptide neurotransmitter that has considerable overlap with 5HT and NE networks and has been implicated in the pathophysiology of depression (Mussap et al., 1993). However, after acute (10–12 weeks) VNS in 19 TRD patients, no significant alterations in substance P levels were observed (Carpenter et al., 2008). It would have been informative if the study had been extended to include neuropeptide levels following chronic treatment.

Immune system dysfunction has also been linked to depression (Connor & Leonard, 1998; Li et al., 2011a; Miller et al., 2009). Treatment with an antidepressant decreases levels of pro-inflammatory cytokines such as interleukin-6 (IL-6) and tumour necrosis factor alpha (TNFα) (Capuron et al., 2003; Hernández et al., 2008; O'Brien et al., 2007). While it is unclear whether neurostimulation therapies for depression affect immune function, there is evidence in VNS treated epilepsy patients that pro-inflammatory cytokine levels decline with successful treatment (De Herdt et al., 2009; Majoie et al., 2011). The specific effects of VNS for 3 months on vagally mediated cytokine synthesis in 10 TRD patients were assessed by Corcoran and colleagues (2005). Paradoxically, increases in IL-1, IL-6, TNFα and TGF beta were observed, in contrast to the reductions observed with antidepressant pharmacotherapy (Corcoran et al., 2005). Clinical response was not linked to outcome in this study, suggesting that higher levels of pro-inflammatory cytokines in patients who are non-responders to treatment could be overshadowing the decreases in responders in this small sample. Furthermore, changes reported in epilepsy studies occurred after 6 months of VNS treatment (De Herdt et al., 2009).

VNS emerged from the epilepsy field and was approved as an adjunct therapy for chronic or recurrent depression that fail to respond to four or more adequate antidepressant treatments, despite evidence of lack of efficacy in acute trials. However, open-label extension trials suggest that the antidepressant effects of VNS require a longer trajectory. There is also preclinical support for the antidepressant effects of VNS from preclinical studies involving behavioural paradigms, as well as studies involving neurochemistry, electrophysiology and neuroimaging.

Deep brain stimulation

Since the early 1990s, the use of DBS has steadily increased as a treatment for chronic pain, and movement disorders (Benabid, 2007). The procedure involves bilateral implantation of electrodes under stereotactic guidance to a specific neuroanatomical target. The electrodes are then connected to a pulse generator which is surgically implanted under the clavicle. This device delivers electrical stimulation usually on a continuous basis. In a similar fashion to VNS, remote changes in electrical stimulation parameters are made through a programming controller.

The application of DBS for TRD is still at an investigational stage. The rationale stems from both the success of this treatment in movement disorders, as well as emerging evidence from functional neuroimaging studies of depression neural circuitry. Anatomical sites for DBS in clinical trials include the subcallosal cingulate gyrus (Lozano et al., 2008; Mayberg et al., 2005), the nucleus accumbens (NAcc) (Bewernick et al., 2010), and ventral capsule/ventral striatum (Malone et al., 2009). The inferior thalamic peduncle (Jimenez et al., 2005) and lateral habenula (Sartorius et al., 2010) have also been explored as potential targets.

Rationale for anatomical targets

Subcallosal cingulate gyrus

The subcallosal cingulate gyrus (SCG) is consistently implicated in mood dysregulation. Depressed subjects exhibit metabolic hyperactivity in this region in response to negative stimuli compared to healthy controls (Mayberg et al., 1999). Corroborating evidence also demonstrates that SCG activity decreases with successful antidepressant therapy including pharmacotherapy, psychotherapy, transcranial magnetic stimulation and electroconvulsive therapy (Kennedy et al., 2007; Ressler & Mayberg, 2007), and pre-treatment hyperactivity is a predictor of non-response to pharmacotherapy or psychotherapy (Konarski et al., 2009). This target was the subject of the first open trial to assess the effectiveness of DBS for TRD (Mayberg et al., 2005).

Nucleus accumbens

Anhedonia is a prominent symptom in patients with TRD (Malhi et al., 2007). In humans, pleasurable responses to stimuli such as music or faces are associated with activity in the nucleus accumbens (NAcc), ventral putamen and ventral caudate (ventral striatum) (Blood & Zatorre, 2001; Senior, 2003). There is also evidence that MDD patients have significantly attenuated responsiveness to pleasurable stimuli, particularly the ability to integrate reward reinforcement history over time (Pizzagalli et al., 2008), and they also exhibit a weaker neurological response to positive stimuli than healthy controls (Epstein et al., 2006). Furthermore, the severity of anhedonia is negatively correlated with activity in the nucleus accumbens (Hasler et al., 2008). Considering the role of the nucleus accumbens in hedonic response and depression pathophysiology, Schlaepfer and colleagues hypothesized that DBS to this site could alleviate anhedonia in TRD patients (Schlaepfer et al., 2008b).

Ventral capsule/ventral striatum

Ablative neurosurgeries for MDD (anterior cingulotomy, anterior capsulotomy, subcaudate tractotomy, limbic leucotomy) all affect the cortico-striato-thalamo-cortical (CSTC) system including the orbitofrontal cortex, basal ganglia and anterior cingulate (Rauch et al., 2003). DBS to the CSTC network has been explored in obsessive compulsive disorder (OCD). The basis to target this network in TRD comes from findings that DBS to the ventral anterior internal capsule/ventral striatum (VC/VS) in patients with OCD resulted in improvements in depressive as well as obsessive compulsive symptomatology (Greenberg et al., 2006; Nuttin et al., 1999).

Inferior thalamic peduncle

The inferior thalamic peduncle (ITP) is a bundle of fibres connecting the thalamic system to the orbitofrontal cortex, and is involved in selective attention, cortical synchronization and sleep (Velasco et al., 2006). Dysregulation of 5HT and NE secretion induces overactivity of the orbitofrontal cortex, which then affects ITP activity. Overactivity in the ITP has been shown through PET scans of MDD patients experiencing a current depressive episode, and this activity decreases with adequate treatment (Velasco et al., 2005). Currently, there are two published case reports of DBS to the ITP (Jimenez et al., 2005; Velasco et al., 2006).

Lateral habenula

The Lateral habenula (LHb) receives strong serotonergic, noradrenergic and dopaminergic innervations (Geisler & Trimble, 2008). There is evidence of lateral habenula (LHb) over-activity in depressed states (Ranft et al., 2010; Shumake et al., 2003), and strong covariation between the LHb and dorsal raphe, suggesting a convergent pathway controlling the release of 5HT (Morris et al., 1999). The LHb has also been found to play a role in reward control via the ventral tegmentum (Li et al., 2011b; Matsumoto et al., 2007). To date, there is one case study of DBS to the LHb (Sartorius et al., 2010).

Clinical outcomes

In 2011 there are no published efficacy trials of DBS in a TRD population under randomized double-blind controlled conditions, although several RCTs are ongoing. Current claims for effectiveness rely on open-label studies for MDD with a minimum time of 6 months, and in some cases, follow-up reports over several years, although trials vary in terms of specific patient criteria, follow-up schedules, and stimulation targets. The HDRS was used as the primary outcome measure and response was defined as a 50% reduction in HDRS score from baseline. Response and remission rates for the DBS trials at 1, 6, 12, 24 and 36 months are summarized in Table IV.

Table IV. Response and remission rates across open-label DBS trials.

Study duration	Response rate	Remission rate
1 month	20–35^ab	0–10^ab
6 months	40–67^abc	20–35^ab
12 months	29–63^abdef	30–50^abde
24 months	46^g	15.4^g
36 months	75^g	50^g

Remission based on scale used: HRSD-28 ≤ 10, HRSD-24 ≤ 10, HRSD-17 ≤ 7; Response based on 50% decrease in depression score from baseline on scale used to define remission.

^aLozano et al., 2008; ^bMalone et al., 2009; ^cMayberg et al., 2005; ^dBewernick et al., 2010; ^ePuigdemont et al., 2011; ^fLozano et al., 2011; ^gKennedy et al., 2011 (based on observed-case analysis data, n = 14).

Subcallosal cingulate gyrus

The first trial for DBS in TRD began in 2003 in Toronto. The initial publication reported findings from six patients who received DBS to the SCG with a Medtronic (Minneapolis, MN) device (Mayberg et al., 2005). After 6 months, four patients (66.6%) had responded. Lozano and colleagues (2008) extended the results of this trial out to one year with 20 subjects including the original six. After 1 month of active DBS, seven patients (35%) were considered responders and two (10%) were in remission (a HDRS-17 score = 7). At 6 months, 12 subjects (60%) had responded and seven (35%) had remitted. This pattern was largely sustained at 12 months, when 11 patients (55%) met criteria for a response and seven (35%) were in remission. The same population were evaluated up to 3–6 years post-DBS. At 2 years, 46% of patients had responded and 15% had remitted. After 3 years, 75% were considered responders and 50% were in remission. At the time of the last follow-up visit, 64% were responders and 43% had remitted, although two deaths were reported and were likely due to suicide (Kennedy et al., 2011). In this same group, neurocognition was tested before and 6 months after DBS in six patients. Tests of dorsolateral, superior medial, and ventrolateral/orbitofrontal behaviours were performed. Areas below average at baseline significantly improved at 6 months including dorsolateral frontal function (language), ventral prefrontal function (executive function) and orbitofrontal function (executive function). There was no evidence of neurocognitive impairment with DBS to the SCG (Mayberg et al., 2005).

A Spanish group (Puigdemont et al., 2011) also reported findings of DBS to the SCG in 8 TRD patients using a Medtronic device and similar entrance criteria to the above TRD population (Kennedy et al., 2011). Response and remission rates at one year, respectively, were 63% and 50%.

A subsequent open-label multi-site trial was conducted across three Canadian sites, in which 21 patients received DBS to the SCG with a different device (St Jude Medical, Plano, TX). Response rates were 48% and 29% at 6 and 12 months, respectively (Lozano et al., 2011). This device delivers constant current and differs from previously reported trials using voltage controlled pulse generators. In theory, this constant current delivery should reduce the need to alter the stimulation parameters during ongoing treatment, but a randomized controlled comparison of both devices is required to confirm this potential advantage.

Nucleus accumbens

The first report of DBS to the NAcc involved three patients with severe resistant depression. Stimulation was administered in a double-blind fashion and depression ratings improved in all three patients when the stimulator was on, and worsened in all three patients when the stimulator was turned off. These effects were observed immediately (Schlaepfer et al., 2008b).

A follow-up report by the same group included data from these three patients in addition to another seven (Bewernick et al., 2010). Though this was intended to be a sham-controlled trial, the design was discontinued after the first three patients experienced severe worsening of symptoms once their stimulation was turned off. At endpoint, 50% of the patients were responders, and 30% were remitters. The Hamilton Anxiety Scale (HAMA) was a notable secondary outcome measure; after 12 months of stimulation, there was a significant reduction in the mean HAMA score from 23.3 to 14.9 across the sample. Improvements in HAMA scores were most evident in patients classified as responders.

Ventral capsule/ventral striatum

An open-label study of DBS to the ventral capsule/ventral striatum (VC/VS) was conducted in 15 patients across three clinical research sites (Malone et al., 2009). Fourteen patients had at least a 5-year history of MDD (2 or more years in current episode), and one patient had a diagnosis of recurrent bipolar I depression. More lenient co-morbidity exclusion criteria were used in this study with secondary diagnoses across the sample including personality disorders, panic disorder, eating disorders, substance dependence in remission, and post-traumatic stress disorder.

After 1 month of stimulation of the VC/VS, 20% met criteria for response. At 6 months, 40% were considered responders and 20% were in remission (HDRS-24 score = 10). At 12 months or last observation, 53% of patients were classified as responders and 40% were in remission. Response and remission were defined by the MADRS as a secondary measure, and results were comparable at the 1, 6 and 12 month marks. After one month of DBS, four patients (26.7%) had responded to treatment, based on the MADRS. At 6 months, seven subjects (46.7%) were considered responders, and four (26.7%) were in remission. Eight patients (53.3%) had responded and five (33.3%) were in remission at last follow-up (Malone et al., 2009).

Inferior thalamic peduncle

In a single case study report, DBS to the ITP yielded significant clinical benefit for MDD. Comorbid diagnoses included borderline personality disorder and bulimia nervosa (Jimenez et al., 2005). After 1 month of stimulation, the HDRS score decreased by 81%, an effect that was largely maintained up to 8 months. At that point, stimulation was discontinued to conduct a double-blind assessment. Over the next 12 months, the patient's HDRS score fluctuated and overall functioning (as assessed by the Global Assessment of Functioning scale) declined. The relapse was reversed after the stimulator was reactivated. A second case study of unilateral frequency stimulation of ITP produced improvements in clinical response (Velasco et al., 2006).

Lateral habenula

One patient, diagnosed with MDD with psychotic features was the subject of a case report on DBS to the LHb (Sartorius et al., 2010). The subject's HDRS-21 score was 45 at baseline. The voltage setting required to achieve remission was considerably higher (10 V) than with other target sites. At approximately 10 months, accidental discontinuation of stimulation resulted in a relapse which was reversed after a further 12 weeks of active stimulation (HDRS = 0).

Safety data

The most common side effects of DBS across all trials included headache, agitation, and pain at incision sites. In the SCG Toronto series, wound infection and consequent hardware removal occurred in two of the first six reported patients; however, this issue was resolved by combining electrode insertion and IPG implantation into one surgical session (Mayberg et al., 2005). In the extended follow-up report (Kennedy et al., 2011), two of the 20 study patients had suicidal attempts, and an additional two patients died by suicide (one suspected, one confirmed). There was no evidence that these deaths were related to the DBS device or parameter adjustments. The multi-site DBS trial reported nausea, tremor, superficial cellulitis at the wound site, agitation and dizziness as the most common adverse effects (Lozano et al., 2011). In this study there was also one suicide attempt and one completed suicide.

Adverse events following DBS to the VC/VS included fainting (7%) and two documented hypomanic episodes (13%), one of which occurred in a bipolar patient. Two patients experienced suicidal ideation, unrelated to DBS (Malone et al., 2009). In the group who received DBS to the NAcc, erythema (40%), increased perspiration (30%), dysphagia (30%), and ocular oedema (60%) were reported. There was also one suicide attempt in this trial and one completed suicide (Bewernick et al., 2010).

Preclinical data

DBS has not been extensively studied in animal depression models. Currently, there are only two published reports examining DBS in preclinical models of depression reflecting DBS to the SCG (Hamani et al., 2010a; 2010b). Since there is no SCG region in the rat brain, the ventromedial prefrontal cortex (vmPFC) served as an anatomical homologue.

In the first study (Hamani et al., 2010a), FST, novelty-suppressed feeding (NSF), learned helplessness, and sucrose consumption models were explored. DBS electrodes were implanted in the vmPFC and stimulation was delivered continuously for 4 h on day 1 and 2 h on day 2. Following stimulation on day 2, rats were tested in their respective paradigms. A comparison group of rats receiving DBS to the striatum and non-implanted rats treated with and without imipramine were used as well. Findings from the FST demonstrated that rats who received DBS to the vmPFC exhibited a 45% reduction in immobility, comparable to the reduction in immobility observed in imipramine-treated rats. These antidepressant effects were not observed with striatal stimulation, or non-implanted rats. To assess whether the reduced immobility with DBS to the vmPFC was due to an overall increase in locomotor activity, vmPFC stimulated rats and non-treated controls were tested in an open field. No significant differences were observed, indicating the behavioural effect of DBS was beyond locomotor activity. In the NSF test, the latency to consume a food reward (removed the day prior to experimentation) in a novel environment was assessed. Latency to feed was 45% lower in the vmPFC stimulated group compared to controls, suggesting a reduction in anxiety related to attaining reward. No group differences in escape response were observed in the learned helplessness model. In the rats predisposed to helplessness, sucrose consumption was associated with periodic footshock. There was a 38% decrease in sucrose consumption for vmPFC stimulated animals compared to a 66% decrease in controls, indicating the vmPFC may influence hedonic response.

The second study (Hamani et al., 2010b) evaluated the effects of different stimulation parameters on the FST response. Four currents were assessed (100, 200, 300 and 400 μA) at a pulse width of 90 μs and either low or high frequency stimulation (20 Hz versus 130 Hz, respectively) compared to sham treatment. Similar to the findings of VNS stimulation, the optimal current followed an inverted u-shape, where 200 μA at high frequency produced the greatest reduction in FST immobility. Although 100 and 300 μA still produced some antidepressant effect, none was observed at 400 μA. The ineffectiveness of high current may reflect the modulation of different pathways or spill over to inactivate other structures. Notably, antidepressant effects were strongest with DBS to the left vmPFC compared to the right and the parameters used approximated clinical practice.

Mechanism of action

While the mechanism of antidepressant action of DBS remains unclear, modulation of local neuronal networks and/or long-term synaptic plasticity are proposed candidates (Lozano et al., 2002; Vitek et al., 2002). Considering there are different targets for DBS, evidence has emerged that relates to site-specific changes.

Subcallosal cingulate gyrus

DBS activity may be mediated by the 5-HT neurotransmitter; decreases in immobility in the FST generated through DBS to the vmPFC are sustained upon NE-depleting lesions, but became obsolete upon 5-HT depleting lesions. In contrast to the microdialysis study with NAcc DBS, at settings that improved behaviour, vmPFC DBS induced a significant and prolonged release of hippocampal 5-HT. Return to baseline 5HT values was observed after 2.5 h.

Disrupting neural pathways in the vmPFC of healthy rats (via radiofrequency lesions or muscimol injections) resulted in decreases in rat immobility similar to those observed in rats receiving DBS stimulation (Hamani et al., 2010a). This antidepressant-like effect was not as strong as that generated by DBS, but still suggests that DBS may induce antidepressant effects by inhibiting the vmPFC.

Imaging PET data in the first six DBS patients measured rCBF before and at 3 and 6 months and was compared to a group of healthy age- and sex-matched controls at baseline. Pre-treatment, depressed patients had significantly greater activity in the SCG and decreased activity in the PFC, premotor cortex, dorsal anterior cingulate, and anterior insula compared to controls. Differences between responders and non-responders were also observed, with responders showing less activity in the SCG, and greater activity in the PFC. At 3 and 6 months, responders also showed decreased activity in the hypothalamus, anterior insula and increased activity in the dorsal PFC, dorsal and anterior cingulate, medial PFC, premotor, and parietal regions (Mayberg et al., 2005). These findings were extended with PET measurements of glucose uptake in the next 11 patients (Lozano et al., 2008). Similar to the RCBF findings, response to treatment with DBS was associated with decreases in OFC, medial prefrontal cortex and insula, as well as increases in lateral PFC and SCG by 6 months. In a subset of the multi-site Canadian trial (Lozano et al., 2011), differences in baseline glucose uptake between responders and non-responders revealed greater activation in the precentral gyrus and decreased activation in the medial frontal gyrus, superior frontal gyrus and orbitofrontal cortex in non-responders (Rizvi et al., 2011).

An analysis assessing the effects of electrode placement and optimal target selection on response in the same population of DBS patients was also conducted (Hamani et al., 2009). The authors reported that there was no more than a 1.5 mm deviation in electrode placement across responders and non-responders. However, in responders, contacts used for stimulation were within a 70% distance from the anterior commissure and 26% percentile distance from the inferior corpus callosum, relatively clustered in the SCG. This indicates that optimal target selectivity can play an important role in clinical outcome.

In order to further elucidate the connectivity patterns of SCG white matter, diffusion tensor imaging (DTI) was conducted in 13 non-depressed patients, and the patterns were compared to the anterior limb of the internal capsule, another DBS target for TRD. The analysis revealed considerable overlap in regions implicated in antidepressant response including the frontal pole, amygdala-hippocampal complex, nucleus accumbens, hypothalamus, dorsal thalamus, and dorsal brainstem. Interestingly, the SCG and internal capsule have distinct trajectories to these areas, potentially reflecting differences in overall connectivity (Gutman et al., 2009).

Nucleus accumbens

One of the mechanisms of action of NAcc DBS could be changes in monoamine release. Animals receiving 300 or 400 μA of unilateral stimulation to the NAcc core for 5 h underwent micro dialysis to measure 5HT, NE and DA before and after. No significant effects of stimulation on monoamine release were observed. Considering Hamani and colleagues (2010b) found the optimal stimulation was at 200 μA, it could be that the parameters used in this study were not optimal for this neuroanatomical site. However, it could also be the case that DBS affects sites distal to the target due to effects at axons and not cell bodies of neurons (Nowak & Bullier, 1998). For example, a study of NAcc DBS in rats showed reduced cell firing in the OFC (McCracken & Grace, 2007). These changes were only observed with high frequency (130 Hz) and not low frequency (10 Hz) DBS. A follow-up study examined the effects of low and high frequency stimulation on the NAcc on a circuit comprising the medial prefrontal cortex, lateral OFC, medial dorsal thalamus, and NAcc regions. Findings revealed selective changes based on low or high frequency, whereby high frequency DBS enhances rhythmicity and synchronous inhibition particularly in the OFC (McCracken & Grace, 2009).

Data from PET imaging in patients also clarifies the brain regions involved in the effectiveness of NAcc DBS. Glucose uptake measured by PET was investigated in 7 NAcc DBS patients before surgery and 6 months post-operatively. At 6 months, there were reductions in metabolic activity in prefrontal subregions including the OFC, and SCG regions, and the posterior cingulate cortex, thalamus and caudate nucleus. Increased metabolic activity was present in the precentral gyrus. PET images in patients with DBS to the NAcc also showed significant reductions in amygdala activity of responders compared to non-responders; the amygdala is an area implicated for its hyperactivity in anxiety disorders (Bewernick et al., 2010).

DBS was first evaluated as a treatment for chronic pain and movement disorders. The first publication in TRD was in 2005 and since then, the published literature includes several open-label trials involving three neuroanatomical sites in approximately 100 patients. While no RCTs have been published to date, these trials suggest a sustained response in approximately 50% of patients at 1 year or longer and are strengthened by the inclusion of functional brain imaging in several studies. There are limited data on animal models of DBS and the use of analogous brain areas for the SCG limits the conclusions of these findings. In addition to an expanding literature on DBS for refractory OCD, the technique is also being explored for treatment of bipolar disorder (Lipsman et al., 2010).

Conclusion

Both VNS and DBS affect the neurocircuitry implicated in depression. In general, preliminary results from the open trials involving different stimulation targets have yielded remarkably similar results. While VNS has undergone more rigorous assessment, including sham controlled double blind trials, there is no evidence to confirm efficacy during the acute phase of these trials and subsequent increases in effectiveness did not occur under double blind conditions. Although the antidepressant effects of DBS appear to occur earlier than effects of VNS, results from ongoing randomized controlled trials are needed to confirm the efficacy of DBS. Investigating DBS in preclinical models is particularly difficult, since a brain homologue is used, which may not generalize to human data. Overall, both treatments appear to affect similar neurochemical cascades as pharmacotherapy. These targeted treatments provide an excellent opportunity to explore emerging neurobiological disturbances in depression. Functional neuroimaging combined with other biomarker tests have the potential to identify biologically distinct phenotypes within the spectrum of TRD with implications for prediction of treatment outcome. In summary, both treatments have promise as an alternative therapy for treatment resistant patients.

Declaration of interest: Dr. Sidney Kennedy has received honoraria or grant funding from AstraZeneca, Biovail, Boehringer-Ingelheim, Eli Lilly, GlaxoSmithKline, Janssen-Ortho, Lundbeck, Merck Frost, Pfizer, Servier, and St. Jude Medical. Dr. Giacobbe has received honoraria or grant funding from AstraZeneca, Eli Lilly and St. Jude Medical. Other authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.