In this overview, we begin with a discussion of ECT. Although not a targeted neural circuit approach per se, ECT was one of the first attempts to directly affect neural function via stimulation, and subsequent efforts to optimize ECT have frequently been conceptualized within a neural circuit framework. We then discuss surgical approaches to neuromodulation, which developed contemporaneously with ECT and have been greatly refined and optimized over time. Next, we review “noninvasive” neuromodulation approaches, which have only become possible as a result of technological advances over the past 20–30 years; these interventions offer a neuromodulation strategy with fewer side effects and risks compared with ECT and surgical procedures and are therefore more scalable and disseminable. We end with a discussion of potential future directions for neuromodulation for depression and other neuropsychiatric disorders.
ECT
In 1938, Cerletti and Bini first used therapeutic electrical stimulation to elicit a seizure in a patient with mania; after 10 treatments, the former engineer who had been found wandering the streets of Rome in a delusional state was able to reunite with his wife and return to work. When he was assessed a year later, he was still employed and married (
2,
3). Since then, refinements to multiple aspects of ECT have improved its safety and tolerability while preserving its efficacy (
4), and ECT remains the single most effective treatment for depression (
5,
6). Adding the combination of a short-acting anesthetic and paralytic agent prior to the procedure in the 1950s greatly reduced anxiety and virtually eliminated the bodily injury associated with early forms of ECT. Anesthesia regimens continue to be optimized to allow the production of seizures of high quality and adequate duration with minimal side effects. The electrical stimulation itself is now delivered in a brief-pulse square wave form, using much less energy overall while still producing therapeutic seizures. This has greatly improved the cognitive side effect profile of today’s ECT compared with early forms of treatment. Continued refinement has led to the adoption of ultrabrief pulse stimulation, further reducing side effects. Routine EEG monitoring allows precision in measuring seizure length while also providing seizure quality measures—in addition to increasing safety by providing assurance that the seizure is in fact terminated after treatment. Right unilateral, rather than bilateral, stimulation was introduced in the 1970s. It is associated with good efficacy and improved cognitive side effects, although further study has shown that comparatively more suprathreshold energy is required to achieve a therapeutic effect comparable to bilateral stimulation (
7).
Like any treatment, ECT has limitations. The relapse rate for major depression is high—well over 50% if treatment is stopped completely when remission is reached; continuation and maintenance ECT, as well as various pharmacologic regimens, are often required to maintain improvements (
8). Social stigma continues to be a barrier to treatment for many patients. Despite great improvements since the early forms of treatment, today’s ECT still has cognitive side effects, particularly during the initial treatment, although overall, meta-analyses show return to or improvement over pretreatment baseline in all cognitive domains (
9). Autobiographical memory loss is problematic in a subset of patients, although this is very difficult to study or quantify. Because of the amount of energy required to cause a generalized seizure, the brain regions initially stimulated by ECT are relatively nonspecific, and a generalized seizure is by definition nonfocal. This lack of focality may represent a limitation, particularly related to the cognitive side effect burden of ECT. New forms of and variations on ECT attempt to improve on this with the hope of minimizing side effects.
Multiple forms of investigational convulsive therapies have been developed. Focal electrically administered seizure therapy, or FEAST, is a form of ECT developed in the early 2000s that focuses unidirectional electrical stimulation to initiate seizure activity in the right prefrontal cortex (
10). A recent open-label trial comparing FEAST to right unilateral ECT showed similar efficacy with decreased cognitive side effects (
11). Magnetic seizure therapy (MST) uses targeted magnetic stimulation to induce a seizure. As in ECT, general anesthesia is required. However, unlike electrical stimulation, magnetic stimulation doesn’t encounter impedance from the tissues between the stimulator and the brain, so a more focal stimulus can be delivered. MST has been studied in open-label (
12–
14) and double-blind (
15) trials and appears to have fewer cognitive side effects and faster postictal recovery and similar efficacy to ECT. FEAST, MST, and other variants of convulsive therapies show great promise in their improved focality of stimulation and thus improved side effect burden, and work is ongoing to further optimize targeting; however, these methods are not yet widely adopted. Many thousands of patients have benefited from the lifesaving potential of ECT over the past 80 years, and this treatment will remain a critical part of our repertoire for years to come (
16).
SURGICAL INTERVENTIONS
Just as delivery of ECT has been iteratively refined and is now more safely and precisely delivered than in the 1930s, surgical interventions to treat psychiatric illness have evolved over the past 100 years and are still in use today, albeit in more targeted form and with rigorous modern ethical standards. Prior to the availability of neuroimaging technology, naturalistic studies of humans with brain injuries or lesions (e.g., Phineas Gage [
17]) and studies in animal models were used to map brain structure to function. By the 1930s, it was becoming clear that the frontal lobes, via connections to other brain regions, were critical in controlling mood, thought, and behavior (e.g.,
18). Lesion studies suggested further that frontal lobe disruption could alter emotion, cognitive function, and behavior without negatively affecting other neural processes (e.g., sensorimotor function). This led to what may be considered the first circuit-based psychiatric intervention, the prefrontal leucotomy.
Portuguese neurologist Egas Moniz developed the prefrontal leucotomy in the 1930s (and won the Nobel Prize in Physiology or Medicine in 1949 for this accomplishment) using a device called a leucotome to disrupt white matter connecting anterior frontal regions to the rest of the brain, effectively interrupting connections between cortical and subcortical regions. In the absence of any effective psychopharmacologic therapies, the leucotomy was applied to a wide range of psychiatric conditions, and improvement was reported in 25%−50% of cases, more commonly in patients with affective disorders (
19,
20). Modification of the leucotomy by Freedman and Watts to even more completely separate the prefrontal lobes from the rest of the brain was called the prefrontal lobotomy (
21), and this procedure was, perhaps too rapidly, adopted in the United States to treat a myriad of psychiatric disorders (
19). During the 1950s, the adverse effects of these procedures became better understood, and ethical issues came to light around consent, proxy decision making, and inhumane anesthetic techniques. Around the same time, chlorpromazine and the first antidepressant medications were introduced. Thus, the leucotomy and lobotomy appropriately fell out of favor. However, since the newly introduced psychopharmacologic agents helped some patients with mental illness, but by no means all, further refinement of neurosurgical techniques continued. The cingulotomy, now employing image-guided stereotactic surgery, is still used in some cases of treatment-resistant depression and OCD, with few complications or side effects (
22). Image-guided capsulotomy, in both surgical and radiosurgical forms, is well tolerated in severe, refractory OCD (
23), and innovative methodologies, including focused ultrasound capsulotomy, continue to be developed (
24).
Developments in neuroimaging greatly advanced our understanding of the circuity of psychiatric disorders, and this work in part led to the first trial of deep brain stimulation (DBS) for depression in 2005. In an initial proof-of-principle study, chronic focal stimulation was delivered to the subcallosal cingulate region in six patients with severe, treatment-refractory depression, leading to remission or near remission in four patients (
25,
26). Prior to this, the only widespread use of DBS was in movement disorders (
27). Since 2005, several other neuroanatomic regions have been targeted in controlled trials of DBS for depression, including the ventral capsule/ventral striatum (
28), the anterior limb of the internal capsule (
29), and the superolateral medial forebrain bundle (
30); other regions, such as the habenula (
31), are emerging targets. DBS has the obvious drawbacks of requiring surgery and the accompanying remote risk of intracranial bleeding and infection, as well as the maintenance requirements of the implanted device. Overall, however, DBS is well tolerated and generally safe. The timeline of peak antidepressant response to DBS ranges from months to several years. Ongoing stimulation appears to provide some protection from relapse (e.g.,
30,
32). Refinement in targeting through advances in diffusion tensor imaging tractography has led to ongoing improvements in electrode placement, and thus in outcomes (
33). It has become clear that targeting one or more white matter tracts precisely, rather than gross anatomic targeting, may be crucial for treatment success. Several large double-blind placebo-controlled trials published in the past few years have not shown reliable separation of active treatment from placebo (
34,
35). This may be for several reasons beyond simple lack of efficacy, including imprecise targeting, selection of primary outcomes (including both time point and manner of measurement), and study design. Nevertheless, a robust body of work demonstrating promise and safety for DBS in several hundred patients with depression remains (
36,
37). As targeting and stimulation modeling continue to be refined, as study design parameters are optimized, and as predictive biomarkers become available to improve patient selection, it is likely that DBS will continue to have a place in treating some of the most refractory cases of depression.
Epidural cortical stimulation targets superficial areas of the prefrontal cortex with implanted electrodes. Several small trials have shown promising results (
38,
39). Improvements appear to be maintained over at least 5 years (
40). Adverse effects were relatively modest and included scalp infection and device malfunctions associated with worsening of the underlying psychiatric condition. Overall, these initial findings are encouraging, although relatively little research on epidural cortical stimulation is currently under way.
Vagus nerve stimulation (VNS) involves surgical implantation of a stimulator on the cervical region of the right vagus nerve, attached to a control unit and battery implanted in the chest. VNS had been used in refractory epilepsy since the 1990s, and anecdotal data showing improved mood in some patients led to trials in patients with depression. VNS was approved by the U.S. Food and Drug Administration (FDA) for the treatment of unipolar and bipolar depression in 2005. While not specifically targeting a particular brain region, VNS is thought to exert antidepressant effects largely through stimulation of its projections to the raphe nuclei, locus ceruleus, insula, thalamus, and prefrontal cortex (
41). Common side effects are hoarseness and stridor, which are caused by vagus nerve activation in the muscles of the larynx and pharynx. Like DBS, VNS appears to be a slow-acting treatment, taking months to years to reach full effect in some patients; critically, though, it has been shown to reduce relapse rate (
42). A less invasive form of VNS, transcutaneous VNS, acts by stimulating the auricular branch of the vagus nerve, which is located near the surface of the skin of the ear. Multiple studies suggest that transcutaneous VNS may also be an effective treatment for depression, although this indication is still investigational (
43,
44).
Surgical interventions for depression have become more precise and technologically advanced since the prefrontal leucotomy, and evidence bases for modern surgical interventions, particularly DBS and VNS, are growing. Continued advances in image-guided targeting and study design, as well as improvements in the devices themselves, increase the probability that these interventions will become more widely available.
“NONINVASIVE” NEUROMODULATION
Previously, nonsurgical stimulation of the brain was only possible using the direct application of electricity, which encounters impedance from the skull and other tissues between scalp surface and cortex. The amount of electricity required to do this is associated with imprecise and potentially painful stimulation. In 1985, Barker et al. (
45) reported the first use of TMS, in which a single pulse of magnetic stimulation over different parts of the motor cortex induced movement in the corresponding contralateral leg or hand. Initial applications for TMS were for neurologic and neurosurgical assessment. As the technology advanced to allow for multiple magnetic pulses to be delivered in a controlled fashion, repetitive TMS (rTMS) as a therapeutic modality became possible. In 1993, Höflich and colleagues (
46) published the first study of rTMS in treating depression in two patients. They stimulated the “central cortex” using a very low repeated pulse frequency, given the limitations of existing equipment (i.e., the time needed for the device to recharge between pulses). Over the next several years, multiple groups showed that open-label daily frontal rTMS was associated with improvement of depressive symptoms (
47–
49). Advances around the same time in the burgeoning field of neuroimaging contributed to more precise target selection, and most trials over the next decade focused on stimulating the left dorsolateral prefrontal cortex (DLPFC).
rTMS uses rapidly alternating magnetic pulses that pass through the skull, inducing an electric current in the underlying cortex. The induced current in turn can depolarize cortical neurons and alter the excitability of neural tissue. High-frequency (up to 20 Hz) rTMS has been associated with increased excitability, and low frequency (≤1 Hz) with decreased excitability, at least when applied to the motor cortex (
50,
51). Coil configuration determines the extent of tissue affected: rTMS using a figure-of-eight coil targets a 1–2 cm region of cortex directly under the treatment coil, whereas an H-coil delivers broader and deeper stimulation. Multiple other coil configurations are in various stages of development.
Several controlled studies in the first decade of the 2000s led to initial FDA clearance of an rTMS device to treat depression in 2008, and clearance of several other devices followed over the next several years. Initially, cleared treatment protocols used high-frequency stimulation delivered over the left DLPFC at 120% of a patient’s resting motor threshold, 30–40 minutes per day, 5 days per week, for 4–6 weeks (
52). Low-frequency stimulation of the right DLPFC also has evidence for efficacy but has not been explicitly cleared by the FDA (
53,
54). Advances in technology allowed for the development of theta burst rTMS, which delivers a series of very high frequency (e.g., 50 Hz) bursts of pulses; the frequency of the series of bursts is thought to mimic the brain’s intrinsic theta frequency (
55). Based on its demonstrated noninferiority to existing FDA-cleared rTMS protocols, an intermittent theta burst stimulation protocol delivered in approximately 3 minutes per daily treatment was recently cleared by the FDA (
56).
rTMS is generally well tolerated. It does not require anesthesia and has no cognitive side effects, so patients can drive themselves to and from treatments. Current protocols do not require neuroimaging to target treatments. There are drawbacks, however. Presenting to the rTMS clinic 5 days per week for a month or more can be cumbersome for patients. Common side effects, seen in up to half of patients, are headaches and discomfort at the stimulation site; these are usually relieved by over-the-counter analgesics. Seizures are a potentially dangerous, but exceedingly rare, side effect at approved treatment parameters; patient selection for rTMS does consider propensity to seizures. As in other modalities, relapse is a concern after successful rTMS treatment; maintenance or continuation therapy appears to sustain improvement, but no consensus has been reached regarding timing and density of ongoing treatment (
57).
Many innovations in rTMS devices and protocols are under investigation in the treatment of depression. “Accelerated” rTMS may have the potential to improve symptoms more quickly than current protocols by providing multiple sessions in a single day (
58–
60). While it is clear that a number of discrete treatments over a period of some days is required, optimal intertreatment interval and number of treatments per day to treat depression most efficiently are not yet established. Targeting of brain regions other than the DLPFC, particularly more medial frontal regions, is an active area of research (
61,
62). Synchronized TMS, which delivers stimulation at an individual patient’s alpha frequency as determined by EEG, has some early evidence of efficacy (
63,
64). Neuronavigation, or image-guided coil placement, and functional MRI (fMRI) connectivity-based approaches for target identification are being assessed but have not yet consistently demonstrated improved effectiveness in the clinic over current DLPFC targeting methods (
65).
Although individual-level neural predictors have not yet changed the way rTMS is routinely delivered, the elusive biomarkers to guide and even personalize treatment are moving closer to reality. For example, the efficacy of TMS in treating major depression has been associated with pretreatment patterns of resting-state functional connectivity within the default mode network, a network that, interestingly, does not include the DLPFC (
66). Drysdale et al. (
67) used pretreatment fMRI resting state patterns to classify a large sample of patients with depression into four subtypes with high sensitivity and specificity; these subtypes, which were not distinguishable by clinical characteristics alone, were shown to predict response to TMS.
Other forms of noninvasive neuromodulation are available and are being investigated. Transcranial direct current stimulation (tDCS) uses small currents (1–2.5 mA) passed between a cathode and an anode attached to the scalp. Most tDCS studies target the DLPFC. Positive results have been found in some controlled studies and meta-analyses (
68), although a recent large international multisite sham-controlled trial was negative (
69). tDCS has not been FDA-cleared for any psychiatric indication, and there is a small risk of burns to the scalp (
70). Cranial electrical stimulation, another form of noninvasive neuromodulation, uses proprietary fluctuating electrical waveforms. Because devices similar to currently available cranial electrical stimulation devices were available prior to FDA regulation of devices in 1976, they were adopted as cleared at that time despite little evidence of efficacy.
FUTURE DIRECTIONS
Neuromodulation, defined broadly, has established itself as an important treatment approach for psychiatric disorders that complements existing medications and psychotherapies. Further development of neuromodulation is as much an engineering problem as it is a neuroscience problem. Technological innovation will continue to allow the development of more targeted, safer, less invasive, and better tolerated treatments. Some examples of new technologies poised to begin having an impact on depression treatment are closed-loop feedback-responsive DBS systems (
71), multiarray DBS responsive feedback systems (
72), directional and multi-lead DBS systems (
73), multiarray TMS devices (
74,
75), and transcranial focused ultrasound (
76).
As with other treatments in psychiatry, matching patients to an appropriate neuromodulation strategy remains a trial-and-error process. Symptom profiles and demographic characteristics appear to have limited utility in this regard. However, it is hoped that biomarkers to predict treatment response, potentially including MRI and EEG phenotypes, can be identified. To accomplish this, large, carefully designed studies will be critical. One such study is currently under way to examine pretreatment and early-treatment-course fMRI markers of the antidepressant response to rTMS (
77). A complementary study assessing potential EEG biomarkers of rTMS response is also ongoing (
78).
Improving neuromodulation treatment approaches will also require improving our understanding of the precise mechanisms of these interventions. Combining clinical trials with mechanistic studies, such as neuroimaging, not only will help establish potential biomarkers of response but also could delineate the neural circuit changes associated with effective treatment. This work could help both optimize our current neuromodulation approaches and identify new treatment targets. Beyond this, more preclinical research specifically focused on neuromodulation approaches applied to animal models of psychiatric disorders will help advance the field (
79).
The three paradigms guiding treatment development in psychiatry are complementary and not mutually exclusive. An exciting area for future study is the combination of neuromodulation treatments with other modalities to improve outcomes, short term and long term. For example, taking advantage of the enhanced neuroplasticity engendered by rTMS by providing psychotherapy within a specific time frame has the potential to synergistically improve efficacy and extend the effect of both treatments (
80). Alternatively, neuromodulation could be used to enhance cognitive mechanisms involved in psychotherapy, as has been initially demonstrated in the treatment of PTSD (
81).
Neuromodulation represents both an old and a new strategy for treating psychiatric disorders. Older approaches, based on an extremely limited understanding of the brain regions involved in psychopathology, were crude and (at best) appropriate only for the most severely ill patients. Advances in technology have provided a more sophisticated understanding of the neural circuitry of disorders of mood, thought, and behavior as well as more nuanced ways of interacting with and modulating these circuits. Continued advances will ensure that neuromodulation remains an integral and complementary treatment strategy for patients with neuropsychiatric illness.