**1. Introduction**

Major depressive disorder (MDD) is one of the most common psychiatric diseases, and while a number of therapies are available, many patients remain symptomatic despite treatment [1,2]. Well-established treatment modalities for MDD include psychotherapy, medication, and electroconvulsive therapy (ECT) [3–9]. While ECT is e fficacious for many patients resistant to medication and therapy, there are significant adverse e ffects associated with ECT, including cognitive and memory dysfunction [3]. Furthermore, there are patients who are refractory to multiple available therapies, including ECT. Patients who fail to improve following treatment with two or more therapies are considered to have treatment-resistant depression (TRD) [5,8,10–12]. Due to the considerable number of treatment non-remitters (30–40% of patients with MDD), developing novel therapies for TRD represents a major unmet need.

Deep brain stimulation (DBS) is a technique that uses implanted intracranial electrodes to modulate neural activity. It is currently a well-established, FDA-approved treatment for movement disorders such as Parkinson's disease (PD) and essential tremor (ET) [13,14]. In addition to movement disorders, DBS has been explored as a treatment modality for psychiatric conditions. Multiple human trials have explored the e fficacy of DBS for TRD. Anatomic targets have included the ventral anterior limb of the internal capsule (vALIC) [15], ventral capsule/ventral striatum (VC/VS) [16], subcallosal cingulate (SCC) [17–22], inferior thalamic peduncle (ITP) [23], medial forebrain bundle (MFB) [24,25], and lateral habenula [26]. Reports regarding the e fficacy of DBS for TRD have been mixed, with some studies demonstrating encouraging results, while others have shown a lack of e fficacy relative to sham stimulation. To leverage all of the available data, we performed a meta-analysis to determine the efficacy of DBS for TRD. We then performed a meta-regression to compare stimulation targets. While prior meta-analyses have been undertaken [27–29], here we included only studies that compared active to sham stimulation in a blinded fashion. Furthermore, our analysis includes more recent studies.
