**4. Discussion**

MDD is a very prevalent neuropsychiatric condition. Although there are a number of existing treatment modalities, many patients remain symptomatic despite adequate treatment protocols. There is an urgen<sup>t</sup> need for additional treatment options for TRD, since it is a significant source of morbidity and mortality.

Pathological neural activity (either hyperactivity or hypoactivity) may lead to neurological or psychiatric disease, and non-pharmacological neuromodulatory techniques may be used to ameliorate these conditions. DBS is a neuromodulatory approach that is an FDA-approved treatment for movement disorders, so multiple investigators have trialed DBS as a therapy for TRD. Initial open-label reports regarding the efficacy of DBS for TRD were encouraging [36,45], and subsequently, multiple randomized trials were initiated and completed [16,17]. The large-scale trials, however, did not reveal a significant difference between the control and treatment groups. A number of reasons have been posited to explain these negative results, including choice of stimulation target, electrode placement technique, patient selection, and short-term follow-up of patients [43,46–52].

Due to the discrepancy between earlier reports and larger trials, we undertook a meta-analysis to investigate the efficacy of DBS for TRD. In agreemen<sup>t</sup> with prior analyses [27–29], we found a significant effect of DBS on TRD. Specifically, active stimulation was associated with an improvement in depression scores relative to sham stimulation. The effect size of this treatment is medium to large, as indicated by the SMD of −0.75 [53]. This effect size is much larger than the small effect sizes (~0.3) seen with pharmacologic antidepressant treatment of patients with MDD [54]. Our findings extend those of prior analyses by including all currently available studies and by only including blinded studies that utilized a trial design in which active stimulation was compared to sham stimulation. Importantly, analyzing the data in this manner allows for the control of placebo effects. While some patients are able to detect active stimulation due to side effects such as visual disturbances, these effects are largely absent with a stimulation parameter adjustment [24,25]. With most studies including an optimization phase prior to the blinded assessments, it is reasonable to propose that comparing active stimulation to sham stimulation controls for placebo effects.

We conducted a meta-regression to determine if the data would reveal an optimal stimulation target. Due to the limited number of available studies, our analysis did not demonstrate an optimal target. With further research, the answer to this question may be elucidated in the future; however, a significant limitation is the heterogeneous nature of MDD neurobiology (i.e., symptoms of depression as a common manifestation of multiple brain functional abnormalities). It may be that instead of one optimal stimulation target for all patients, the optimal stimulation target varies across individuals [46,55–58].

As with any therapy, there is a risk of publication bias since studies with positive results and large effect sizes are more likely to be published than studies with small effect sizes or negative results [59]. To investigate publication bias in DBS for depression, we plotted the studies in a funnel plot and used Egger's test to assess for asymmetry. We did not find evidence of publication bias in this series of studies.

DBS is a well-tolerated treatment for movement disorders, but it is important to critically evaluate potential side effects of DBS for depression. The thoroughness of adverse event reporting varied between the studies. Nonetheless, the majority of adverse events were transient. Furthermore, many side effects were relieved by a stimulation parameter adjustment, as is seen with DBS for movement disorders. Patients with MDD are at significant risk for suicide, and patients with TRD have an even higher risk of suicide [60,61]. Currently available data do not demonstrate an increased risk of suicide with DBS, since suicide was rare in these studies and occurred at a lower rate than in patients with severe MDD not receiving DBS [62]. Moreover, the completed suicides occurred in non-responders or were deemed unrelated to DBS. As an invasive therapy, DBS may be perceived by patients as a "last resort" for recovery. Therefore, non-responders may represent a particularly high-risk group for suicide. Psychoeducation and adequate discussion of post DBS treatment options to reframe those

perceptions may decrease the risk of suicide. In summary, published trials have demonstrated the safety of DBS for TRD.

While the results of this meta-analysis are encouraging, additional large-scale clinical trials demonstrating the efficacy of DBS for TRD are essential. Crucial to these efforts will be careful consideration of future trial design [24,63]. The large-scale clinical trials of DBS for TRD conducted to date have utilized a parallel trial design, whereas many of the smaller trials employed a crossover design (Table 1). Given the negative outcomes of the large-scale trials, a crossover design may be optimal to investigate the efficacy of DBS for TRD. As a therapy, DBS is unique in that sham stimulation and active stimulation may be compared within individuals, thereby facilitating crossover trial design. Crossover studies require fewer patients to achieve significance, so this design optimizes statistical power [64]. Some studies have used customized trial designs with a variable optimization period [15,65]. Ensuring proper optimization prior to randomization may be necessary to determine the efficacy of DBS for TRD. Another important factor in trial design is the length of time that the patient undergoes therapy. Long-term open-label studies have demonstrated that the efficacy of DBS for TRD improves over time, so longer trials may be necessary [17,18,38,66,67]. Finally, depression severity waxes and wanes throughout a patient's disease course. Therefore, trial designs that compare groups at specific time points may not be optimal. Integrating scores over predefined time periods may enhance outcome assessment.

In addition to trial design, proper electrode targeting is important for the efficacy of DBS for the treatment of depression. Instead of targeting brain nuclei, as is routinely done, targeting fiber tracts using patient-specific tractography may be important for therapeutic efficacy [49–52,58,68]. Furthermore, stereotactic accuracy of electrode placement is essential. For example, intraoperative phenomena, such as brain shift, should be accounted for [48]. Patient selection is also a critical consideration for future studies. The available data have demonstrated that there is a subgroup of patients that respond to this therapy and a subgroup that does not. Determining patient-specific factors (e.g., anatomical or symptom-based) that predict response to DBS for TRD may enable targeted selection of patients for whom DBS would be therapeutic [43,69–75]. These patient-specific factors could then be used as study inclusion criteria to enhance the probability of study success.
