Neuromodulation interventions for addictive disorders: challenges, promise, and roadmap for future research

Neuromodulatory interventions – such as deep brain stimulation, transcranial magnetic stimulation, and transcranial direct current stimulation – hold promise for the targeting and remodelling of circuit dysfunctions in addiction. Spagnolo and Goldman review the literature on neuromodulation in addiction disorders, highlighting recent developments and limitations, and topics for future studies.

Abstract

Addictive disorders are a major public health concern, associated with high relapse rates, significant disability and substantial mortality. Unfortunately, current interventions are only modestly effective. Preclinical studies as well as human neuroimaging studies have provided strong evidence that the observable behaviours that characterize the addiction phenotype, such as compulsive drug consumption, impaired self-control, and behavioural inflexibility, reflect underlying dysregulation and malfunction in specific neural circuits. These developments have been accompanied by advances in neuromodulation interventions, both invasive as deep brain stimulation, and non-invasive such as repetitive transcranial magnetic stimulation and transcranial direct current stimulation. These interventions appear particularly promising as they may not only allow us to probe affected brain circuits in addictive disorders, but also seem to have unique therapeutic applications to directly target and remodel impaired circuits. However, the available literature is still relatively small and sparse, and the long-term safety and efficacy of these interventions need to be confirmed. Here we review the literature on the use of neuromodulation in addictive disorders to highlight progress limitations with the aim to suggest future directions for this field.

Introduction

Addictive disorders are pervasive and costly. The 2013 National Survey on Drug Use and Health estimated that 20.3 million adults had a substance use disorder, representing 8.5% of the population. The adequacy of medical treatments for addictive disorders is limited, relapse rates being extremely high within the first year of treatment. Furthermore, many who are affected are never treated, resulting in a large unmet need (O'Brien, 2008).

A defining problem in the treatment of addiction is that we do not know how to restore the addicted brain (Goldman and Barr, 2002). Advances in neurophysiology and neuroimaging over the past two decades have led to insights into the mechanisms of addictive disorders. These studies strongly indicate that clinically observable behaviours that characterize addictive disorders, including compulsive consumption, impaired self-control, and behavioural inflexibility, are caused by dysregulation and malfunction of specific brain circuits. This realization has led to a reinterpretation of addictive disorders as brain disorders characterized by neurocircuit alterations (Volkow et al., 2016). Several neurobiology-based interventions have been developed to modify functions of the affected brain circuits. Pharmacotherapies such as methadone and buprenorphine, for opioid use disorders, and naltrexone, for alcohol use disorders, can be viewed as modulatory of neurocircuits, but these interventions lack spatial and temporal specificity of action.

In animal models, both optogenetic interventions, in which neurons transfected with gene constructs are activated or deactivated by light, and chemogenetic interventions, in which cells are transfected with DREADDS (Designer Receptors Exclusively Activated by Designer Drugs), have been effective in identifying and modulating circuits underlying addictive behaviours (Britt and Bonci, 2013; Kerstetter et al., 2016). These studies suggest that the applicability of circuit-based interventions for addictions in people, by whatever modalities, may be safe, effective, and practical.

A novel, promising, and immediately available approach to the treatment of addictive disorders may be represented by neuromodulation interventions, both invasive such as deep brain stimulation (DBS), and non-invasive techniques such as repetitive transcranial magnetic stimulation (rTMS) and transcranial direct current stimulation (tDCS). These techniques have been largely used both as research tools to probe affected circuits, and as therapeutic interventions for a variety of neurological and psychiatric disorders with encouraging results. However, significant challenges arise when applying these interventions to addictive disorders, including scarcity of predictive animal models, methodological issues in human studies performed to date, incomplete understanding of optimal neural targets, stimulation protocols, connectivity and networks, and the ethical challenges associated with modifying brain function with experimental treatments.

Here, we outline the current state of knowledge of neuromodulation interventions in addictive disorders, and discuss limitations and future directions. We focus on tDCS, TMS and DBS because all three methods are immediately available, and in the cases of TMS and tDCS, offer the potential to treat large numbers of patients with addictive disorders, a crucial consideration given the high prevalence of these disorders.

Neurobiology of addictive disorders and its relevance for neurocircuit-based interventions

Addictive disorders are chronic, relapsing diseases resulting from interaction between innate genetic, early developmental, and environmental factors pre-existing exposure to the addictive agent, and synaptic and circuitry changes induced by exposure to the addictive agent. Several theories have been developed to explain compulsive drug use and relapse and, while each of them has provided unique contributes, they converge on several important aspects.

A central feature in the framework of causation of addictions are that these disorders are induced and maintained by neuroadaptations in the mesocorticolimbic dopamine (DA) system and in the glutamatergic corticolimbic circuitry in which the dopamine projections are embedded (Robinson and Berridge, 1993; Nestler, 2001; Koob and Volkow, 2010). Addictive drugs have diverse primary targets, but all, with the possible exception of opioids, initially enhance reward via increased dopamine release in the nucleus accumbens (NAc) (Garbusow et al., 2016) and other areas of the limbic forebrain including the amygdala and prefrontal cortex (Kourrich et al., 2015). The repeated stimulation of dopamine pathways induces plastic changes that render these brain reward systems hypersensitive to drugs and drug-associated stimuli (Robinson and Berridge, 1993), and are mechanistic for conditioning to these cues. Formation of stimulus-response associations strengthens learned habits in drug-seeking behaviour, and this shift from goal-directed to more habitual responding reflects a transition from prefrontal cortical to striatal control over responding, and a transition from ventral striatal to more dorsal striatal subregions (Everitt and Robbins, 2005; Belin et al., 2013). Drug-specific cues acquire the capacity to evoke visceral changes, perceived as ‘feelings’ by mechanisms that depend on the insular cortex, and induce strong cravings long after acute withdrawal.

In humans, brain imaging studies confirm that drug-associated cues induce dopamine increases in the dorsal striatum, and it has been observed that displacement of PET ligands from dopamine receptors by cocaine correlates with self-reports of craving (Volkow et al., 2006). These changes in striatal dopamine function are accompanied by decreased activity in several prefrontal and associated regions. Addicted patients show depressed activity in the orbitofrontal cortex, which is involved in salience attribution and goal-directed behaviours, anterior cingulate, which is involved in inhibitory control and awareness, and the dorsolateral prefrontal cortex (DLPFC), which is involved in higher cognitive operations and decision-making (Volkow et al., 1993, 2001, 2007).

Prolonged exposure to alcohol and drugs can also lead to changes in brain reward threshold. Because addictive agents and their cues strongly activate reward circuits compared to natural reinforcers, the motivational choices of the individual exposed to addictive agents may become fixed, leading to compulsive drug use (Kalivas and Volkow, 2005), the habit itself having become incentive (Belin et al., 2013).

Neuroadaptive changes in addiction extend beyond reward, and to disturbances in brain regions other than striatum. Human imaging studies of individuals with addictive disorders find reductions in dopamine D2 receptors (Carrera et al., 2004) as well as decrease in dopamine release that may reflect hypodopaminergic functioning, and hypoactivity of the orbitofrontal-infralimbic system (Volkow et al., 2003). This attenuated dopaminergic function has been observed in response to both addictive agents and natural reinforcers (Volkow and Li, 2004). The shift in hedonic set point and a state of dysregulation of brain reward systems are associated with recruitment of anti-reward systems. Adaptations in the circuitry of the extended amygdala (central nucleus of the amygdala, bed nucleus of the stria terminalis, and NAc shell) and also in the lateral habenula, and abnormalities in neurotransmitters involved in stress response, may lead to negative reinforcement, which is characterized by withdrawal, dysphoria and distress. These negative states may powerfully motivate drug-seeking and may induce relapse even after protracted abstinence (Koob, 2009).

In summary, prolonged exposure to addictive agents results in dysfunctions in multiple circuits that maintain addiction and that are the consequences of neural adaptations with different time courses. Significant alterations can be detected in circuits implicated in reward, salience attribution, motivation, inhibitory control, learning and memory consolidation, leading to a complex phenotype characterized by different symptom dimensions, each mapping onto specific circuits. Neurocircuit-based interventions might ameliorate or disrupt addiction by targeting different neuroanatomical structures that serve as ‘nodes’ within these circuits and that are associated with particular behavioural markers. In the next sections, we will review studies that have investigated these interventions and we will primarily focus on the stimulation site and on the clinical and behavioural targets selected to discuss their efficacy.

Transcranial direct current stimulation and transcranial magnetic stimulation: an overview

Transcranial DCS and TMS non-invasively modulate neural activity. TDCS uses low-amplitude direct currents, in the range of 0.5–2 mA, applied directly to the scalp through a pair of saline-soaked electrode pads connected to a battery-like device. It has been proposed that tDCS alters cortical excitability by inducing subthreshold modulation of neuronal membrane potentials (Nitsche and Paulus, 2000; Nitsche et al., 2003). The nature of neuronal modulation depends on duration, intensity, and polarity of stimulation (Nitsche et al., 2005). Generally, anodal tDCS depolarizes neurons, thus increasing cortical excitability, whereas cathodal tDCS hyperpolarizes neurons, diminishing cortical excitability (Nitsche and Paulus, 2001). When applied for a sufficient duration, tDCS induces sustained cortical excitability (Nitsche and Paulus, 2001), and can change synaptic plasticity (Brunoni et al., 2012a). Electrode positioning and configuration, as well as skull thickness and composition also alter the effects of tDCS (for a review see Klooster et al., 2016), as can current medications and context of stimulation, whether at rest or during task execution (Paulus et al., 2013). Most knowledge of tDCS safety derives from single event stimulation studies in healthy, unmedicated subjects, in whom adverse effects—skin irritation and sensation of burning—are mild (for a review, see Brunoni et al., 2012b). However, adverse effects of tDCS could be a more salient issue if tDCS is used daily or at higher current strength.

Whereas tDCS is purely neuromodulatory, TMS is both neurostimulatory and neuromodulatory. TMS generates electrical activity in targeted brain regions through the application of magnetic pulses produced by passing an electrical current through an electromagnetic coil. The magnetic stimulation can be delivered as a single pulse or as a train of pulses. Single-pulse TMS is typically used to study brain physiology and plasticity (Rossi et al., 2007; Paulus et al., 2013), whereas rTMS is commonly used to elicit neuromodulation and neuroplasticity, and can result in prolonged excitability that outlasts the stimulation period (Pascual-Leone et al., 1998; Paulus et al., 2013). Typically, the direction of neuromodulation is driven by the frequency at which stimulation is performed. High frequency rTMS (≥5 Hz) increases cortical excitability, while low frequency rTMS (1 Hz) decreases cortical excitability (Peterchev et al., 2012). However, theta burst stimulation (a variation of high frequency rTMS) can either depress or facilitate cortical excitability, depending on burst-train duration, such that intermittent theta burst stimulation increases excitability and continuous theta burst stimulation decreases excitability (Mix et al., 2010). The spatial specificity of stimulation depends on strength of the field and the shape of the magnetic coil: while the classical figure-of-eight coil provides superficial and focal stimulation, new coils (H-coils) are supposed to target brain regions to a depth of 5–7 cm, together with broad activation of overlying areas, as reviewed later. rTMS is associated with mild-to-moderate side effects such as scalp pain, headache, and transient hearing impairment, that mainly follow high frequency stimulation (Rossi et al., 2009). The most serious concern is the risk of inducing epileptic seizures, a risk minimized by prescreening for risk factors and by applying TMS guidelines (Lefaucheur et al., 2014). In a few cases, TMS has also been reported to cause acute psychiatric changes in patients with major depression or bipolar disorders (Zwanzger et al., 2002; Janicak et al., 2008; Xia et al., 2008).

The risk-benefit balance between TMS and tDCS is influenced by several considerations. tDCS has poor spatial and temporal resolution, whereas TMS has higher focality and temporal resolution (milliseconds), and is less sensitive to anatomical differences (e.g. skull thickness). Both are well tolerated but tDCS is associated with less severe side effects.

In research studies, tDCS has been shown to permit a reliable sham condition (Gandiga et al., 2006) while sham stimulation for rTMS is more challenging. In tDCS minimal or no scalp sensations accompany stimulation, and subjects tend to become habituated after a few seconds of stimulation. By contrast, rTMS induces strong scalp sensations, along with facial and scalp muscle twitches, and the device generates a loud click. Although sham TMS coils produce a similar sound artefact, they do not induce the same scalp sensations or muscle twitches. This problem may be addressed by using coils that deliver electrical stimulation to the skin in the sham condition (Okabe et al., 2003).

In view of the therapeutic applicability of these techniques, it is also important to note that TMS has been approved by the US FDA for the treatment of both depression and migraine, but tDCS is presently not approved for any indication. Both techniques may be better tolerated, and safer, than systemically administered medications for particular patients, including females who are pregnant or of child-bearing age, the elderly, subjects with medical conditions impairing the absorption, distribution or metabolism of drugs, and those taking other medications. Finally, tDCS devices are relatively inexpensive, while TMS devices are more expensive and currently available TMS systems require the use of customized pads to deliver stimulation, increasing the cost of TMS, and potentially discouraging its widespread adoption.

Transcranial direct current stimulation and transcranial magnetic stimulation in addictive disorders

A growing number of tDCS and TMS studies have been conducted for addictive disorders. Although differing in important aspects of design (e.g. number and duration of sessions, sample size), the majority of these studies have used excitatory stimulation to target the DLPFC, a node in the frontal-striatal network that governs executive control, with the aim to reduce craving, and particularly cue-induced craving, in laboratory settings (Tables 1 and 2). Several studies have also evaluated the effects of these interventions on alcohol or drug consumption and relapse, and few tDCS studies have also explored changes in executive and cognitive functions, and mood.

Table 1.

Studies of tDCS for addictive disorders: clinical and technical considerations

Studies	n	Participants	Design	Number of sessions	Stimulation site	Polarity	Current (mA)	Effects	Side effects
Nicotine
Fregni et al. (2008)	23	Smokers; >15 cigarettes per day; 1.5 h abstinence	Randomized, double-blind, sham-controlled, cross-over study	3 (1 anodal left; 1 anodal right; 1 sham)	Left and right DLPFC	A/C/S	2	Reduction of cue-induced craving after both active tDCS conditions; no effect on mood	Scalp burning, headache, itching, no difference between groups
Pripfl et al. (2014)	17	Smokers; at least 6 h of abstinence	Sham-controlled, within-subjects study	3 (1 anodal left; 1 anodal right; 1 sham)	Left and right DLPFC	A/C/S	0.45	No effect on cue-induced appraisal for both left and right anodal stimulation; attenuated negative evaluation for negative stimuli following right anodal stimulation	Itching and tingling reported for both sham and real stimulation
Xu et al. (2013)	24	Nicotine dependent; >10 cigarettes per day; >10 h abstinence	Single-blind, sham-controlled study	2 (1 Ac and 1 S)	Left DLPFC	A/S	2	No effect on cue-induced nicotine smoking following active tDCS; no effect on attention; mood improvement	Scalp burning, sleepiness
Falcone et al. (2016)	25	Smokers; at least 10 cigarettes/day; 12 h abstinence	Double-blind, sham-controlled study	2 (1 Ac and 1 S) while subjects were in a smoking room	Left DLPFC	A/S	1	Increased latency to smoke and decreased number of cigarettes smoked following active stimulation	Tingling, itching, scalp burning
Boggio et al. (2009a)	23	Smokers; at least 10 cigarettes/day; 1.5 h abstinence	Randomized, double-blind, cross-over study	5	Left DLPFC	A/S	2	Reduction of cue-induced craving and cigarette consumption	Scalp burning, headache, itching; no difference between groups
Fecteau et al. (2014)	12	Light, moderate and heavy smokers	Randomized, double-blind, sham-controlled study	5	Left and right DLPFC	A/C/S	2	Reduction in number of cigarettes smokes lasting up to 4 days following active stimulation. No effect on risk task. Effect on Ultimatum task: rejection of cigarettes as an offer.	Sleepiness, headache, pain
Meng et al. (2014)	30	Smokers; >8 cigarettes per day	Randomized, double-blind, sham-controlled study	2 (1 Ac and 1 S)	Fronto-temporal areas	C/S	1	Bilateral cathodal stimulation of the FPT areas significantly reduced the attention to smoking-related cues and daily cigarette consumption on the following day	Tingling, itching, pain
Alcohol
Nakamura-Palacios et al. (2012)	49	Alcohol-dependent; detoxified; treatment-seekers; 7 days abstinence	Randomized, sham-controlled, single-blind study	2 (1 Ac and 1 S) while subjects were exposed to visual cues	Left DLPFC	A/S	1	No effect on frontal activity as indexed by P3 in response to auditory alcohol related stimuli, except that in Lesh IV alcoholics showing increasing P3 amplitude after real tDCS	Itching and tingling reported for both sham and real stimulation
Boggio et al. (2008)	13	Alcohol-dependent; detoxified; treatment-seekers; >10 days abstinence	Double-blind, randomized, sham-controlled, cross-over study	3 (1 anodal left/cathodal right; 1 anodal right/cathodal left; 1 sham)	Left and right DLPFC	A/C/S	2	Reduction of craving by both active tDCS conditions	Discomfort, headache
da Silva et al. (2013)	13	Alcohol-dependent; detoxified; in treatment; >10 days abstinence; in routine clinical treatment	Randomized, sham-controlled, single-blind study	5 (1/week)	Left DLPFC	A/S	2	No effects of real stimulation on relapse rate; reduction in craving for alcohol and in POMS depression score	NS
Klauss et al. (2014)	33	Alcohol-dependent; detoxified; treatment-seekers; >20 days abstinence	Randomized, sham-controlled, single-blind study	10 (2 sessions/day)	Left and right DLPFC	A/C/S	2	Reduction in relapse rate at the end of the 6-month follow-up. No effect on craving, cognition and mood	Skin redness
den Uyl et al. (2015)	41	Heavy drinkers	Double-blind, randomized, sham-controlled study	3 (1 DLPFC; 1 IGF; 1 sham)	Left DLPFC and IFG	A/S	1	Reduction of inclination toward drinking by real DLPFC stimulation. No effect of IGF stimulation. No effect of real stimulation on alcohol-related attentional biases	None
Stimulants
Batista et al. (2015)	36	Crack-cocaine; average of 35 days of abstinence	Randomized, sham-controlled, double-blind study	5	Left and right DLPFC	A/C/S	2	Reduction in spontaneous craving after real stimulation	Tingling, scalp burning
Conti et al. (2014)	13	Crack-cocaine; average of 30 days of abstinence	Randomized, sham-controlled, single-blind study	5	Left and right DLPFC	A/C/S	2	Inhibitory effect on P3 drug-cued cortical activation by a single dose of active tDCS; Increase after 5 sessions	Itching and tingling reported for both sham and real tDCS
Shahbabaie et al. (2014)	32	Methamphetamine-dependent; abstinent	Randomized, double-blinded, crossover, sham-controlled study	2 (1 Ac and 1 S) at rest and during cue-inducing task	Right DLPFC	A/S	2	Reduction in spontaneous craving after 10 min stimulation induced by real stimulation; Increased in cue-induced craving by real stimulation	Drowsiness, tingling, itching
Gorini et al. (2014)	36	Cocaine-dependent; Hospitalized; average of 2 weeks of abstinence; Control group of healthy volunteers	Randomized, sham-controlled	3 (1 anodal left/cathodal right; 1 anodal right/cathodal left; 1 sham)	Left and right DLPFC	A/C/S	1.5	DLPFC stimulation of cocaine dependent subjects inconsistently altered risky decision-making depending on the task	NS
Opiates
Wang et al. (2016)	20	Heroin-dependent; in treatment; 1.5–2 years abstinence	Randomized, sham-controlled, single-blind study	2 (1 Ac and 1 S)	Fronto-temporal-parietal areas	C/S	1.5	Significant difference between pre- and post-stimulation cue-induced craving score following real stim.	None
Marijuana
Boggio et al. (2010)	25	Marijuana users; after 24 h abstinence	Randomized, single-blind study	2 (1 Ac and 1 S)	Left and right DLPFC	A/C/S	2	Reduction of craving after right anodal/left cathodal stim. Increase in risky choices	Itching and tingling reported for both sham and real stimulation

A = anodal; Ac = active; C = cathodal; FPT = frontal-parietal-temporal area; IFG = inferior frontal gyrus; NS = not specified; POMS = Profile of Mood States; S = sham.

Table 2.

Studies of rTMS for addictive disorders: clinical and technical considerations

Studies	n	Participants	Design	Number of sessions	Stimulation site	F (Hz)	% MT	Total pulses per session	Effects	Side effects
Nicotine
Johann et al. (2003)	11	Nicotine dependent; motivated to quit smoking	Randomized, double-blind, sham-controlled, cross-over trial	2 (1 Ac and 1 S)	Left DLPFC	20	90%	1000	Significant reduction in craving after the real rTMS	NS
Eichhammer et al. (2003)	14	Nicotine dependent; motivated to quit smoking	Randomized, double-blind, sham-controlled, cross-over trial	4 (2 Ac; 2 S)	Left DLPFC	20	90%	1000	Significant reduction in smoking in the rTMS group. No effect on craving	Mild headaches
Amiaz et al. (2009)	48	Nicotine dependent; >20 cigarettes/ day; motivated to quit smoking	Randomized, double-blind, sham-controlled trial (four subgroups: active versus sham rTMS/smoking-related versus neural picture cues). 10 daily sessions followed by a 4-week maintenance phase	10 daily sessions	Left DLPFC	10	100%	1000	Significant reduction in cue-induced craving, cigarette smoking and dependence when subjects received exposure to smoking cues followed by rTMS	NS
Li et al. (2013a)	16	Nicotine dependent; ≥10 cigarettes/day; not treatment seeking	Randomized, double-blind, sham-controlled, crossover trial. 2 h abstinence before the session	2 (1 Ac and 1S)	Left DLPFC	10	100%	3000	Significant reduction in craving in the active group. Positive correlation between the effect of rTMS and the severity of the dependence	Mild discomfort
Hayashi et al. (2013)	10	Moderate to heavy smokers	Randomized, double-blind, sham-controlled crossover study	4 (2Ac; 2S) paired with two different conditions (delayed versus immediate cigarette availability)	Left DLPFC	1	NS	1800	Significant reduction in craving for smoking during the ‘immediate availability’ condition and attenuation of concomitant fMRI signal in the medial orbitofrontal cortex	NS
Trojak et al. (2015)	37	Nicotine dependent; treatment seeking	Randomized, double-blind, sham-controlled study between-subject study. rTMS was associated to nicotine replacement therapy	10 sessions across 2 weeks	Right DLPFC	1	120%	3600	Greater abstinence rate at the end of 2 weeks following active rTMS. No effect on craving	None
Rose et al. (2011)	15	Cigarette smokers; >20 cigarettes/day	Randomized cross-over open-label study. At the beginning of each session, subjects smoked a cigarette. 1 h later, they underwent rTMS concurrently during exposure to (i) neutral (ii) smoking cues (iii) smoking a cigarette.	3 (1 Hz SFG; 10 Hz SFG; 1 Hz motor cortex)	SFG or motor cortex (side not specified)	1 and 10 Hz	90%	Greater number of pulses for the 10 Hz condition.	Combination of smoking cues exposure and 10 Hz SFG rTMS increased craving	NS
Dinur-Klein et al. (2014)	115	Nicotine dependent; >20 cigarettes/day; treatment seeking	Double-blind, placebo-controlled, randomized clinical trial. Three different stimulation condition (1 Hz, 10 Hz, sham) with or without exposure to alcohol cues	13 sessions (1 Hz; 10 Hz or sham)	Bilateral DLPFC and insular cortex	1 and 10 Hz	120%	1 Hz: 600 pulses; 10 Hz: 990 pulses	10 Hz rTMS combined with exposure to smoking cues reduced craving, cigarette consumption and dependence	Headache, nausea
Alcohol
Mishra et al. (2015)	45	Alcohol-dependent, detoxified; inpatients. Participants received zolpidem and about three quarters of them received anti-craving drugs	Single blind, sham-controlled, parallel group trial (without randomization)	10 daily sessions	Right DLPFC	10	110%	1000	Significant reduction in craving after the last rTMS session in the active group.	Seizure for patient with sham (but recently stopped lorazepam), scalp pain, headaches, anxiety
Herremans et al. (2012)	36	Alcohol dependent; detoxified; hospitalized	Single-blind, sham-controlled design	2 (1 Ac and 1 S)	Right DLPFC	20	110%	1560	Lack of anti-craving effects of a single session of high-frequency rTMS	None
Herremans et al. (2015)	26	Alcohol dependent; hospitalized	Single-blind, sham-controlled design followed by an open-label phase	2 (1 Ac and 1 S) +14 Ac sessions	Right DLPFC	20	110%	1560	No decrease in cue-induced craving, but reduce in spontaneous craving	NS
Hoppner et al. (2011)	19	Alcohol dependent; detoxified	Randomized, sham-controlled trial. Both groups were compared to a female age-matched healthy control group and were exposed to neutral, emotional and alcohol-related pictures, before the first and after the last rTMS session	10	Left DLPFC	20	90%	1000	No significant between-group difference regarding craving and mood	None
Del Felice et al. (2016)	17	Alcohol dependent Hospitalized; in treatment with disulfiram	Single-blind sham-controlled trial	4 (2Ac; 2S)	Left DLPFC	10	100%	1000	No effect on craving and alcohol consumption. Positive effects on inhibitory control and selective attention	NS
Mishra et al. (2015)	20	Alcohol dependence, treatment-seekers	Between-subject single-blind randomized study	10 (left or right side)	Left and Right DLPFC	10	100%	1000	Significant reduction in craving in both conditions	Insomnia and nightmares in one case
Ceccanti et al., (2015)	18	Alcohol dependent, treatment seekers, detoxified	Between-subject randomized double-blind sham-controlled study	10	Medial PFC	20	120%	NS	Significant reduction in craving and drinking days following rTMS	Not significant side effects
Stimulants
Camprodon et al. (2007)	6	Cocaine dependence	Randomized, cross-over study	2 (left or right side)	Left and Right DLPFC	10	90%	2000	Right but not left rTMS reduced craving	None
Politi et al. (2008)	36	Cocaine dependence	Open-label study	10	Left DLPFC	15	100%	600	Reduction in spontaneous craving	NS
Terraneo et al. (2016)	32	Cocaine dependence, treatment-seekers	Open-label, randomized study. rTMS or standard pharmacological treatment	8	Left DLPFC	15	100%	2400	Reduction in cocaine use and craving	Discomfort
Li et al. (2013a)	10	Not-treatment seeking, methamphetamine users	Single-blind, randomized, sham-controlled crossover study	2 (1Ac and 1S)	Left DLPFC	1	NS	900	Increase in craving by active rTMS	Mild scalp discomfort
Hanlon et al. (2015)	11	Not-treatment seeking Cocaine-dependent	Single-blind, randomized, sham-controlled study	2 (1Ac and 1S)	Frontal pole/MPFC	5	110%	1800	Reduced stimulus-evoked activity in MPFC and NAc, and decreased craving	Pain
Opiates
Shen et al. (2016)	20	Heroin dependence Detoxified	Randomized, sham-controlled crossover study	2 (1Ac and 1S)	Left DLPFC	10	100%	2000	Reduction in cue-induced craving by active rTMS	None

Ac = active; fMRI = functional MRI; MPFC = medial prefrontal cortex; MT = motor threshold; NS = not specified; S = sham; SFG = superior frontal gyrus.

Transcranial direct current stimulation effects on craving and cue reactivity, consumption and relapse

Craving is a major clinical feature of addiction and is associated with poor clinical outcome and repeated relapse behaviours. As mentioned above, among the several neural substrates that have been implicated in both spontaneous and cue-elicited craving, including the anterior cingulate, orbitofrontal cortex and DLPFC, amygdala and insula (Wilson et al., 2004; Wang et al., 2007), the DLPFC has been often selected, also given that it can be easily targeted with external, non-invasive stimulation.

The application of tDCS over the DLPFC to modulate craving has yielded mixed results. Studies investigating the effects on spontaneous craving have showed that both single and multiple sessions of tDCS reduced craving for marijuana (Boggio et al., 2010) and for crack cocaine (Batista et al., 2015), regardless to treatment status and abstinence duration. Conversely, in another study, higher dose of tDCS (10 twice-daily sessions) did not affect craving in treatment-seeking alcohol-dependent subjects but reduced relapse rate (verbally reported and considered as return to the previous drinking pattern), over a 6-month follow-up (Klauss et al., 2014).

With regard to cue-induced craving, usually elicited using visual stimuli but also by cue manipulation procedures, four studies reported a reduction in craving for different addictive agents following tDCS (Boggio et al., 2008, 2009b; Fregni et al., 2008; da Silva et al., 2013). Among these studies, two targeted the left DLPFC using multiple sessions (Boggio et al., 2008; da Silva et al., 2013), while in the other two studies both hemispheres were stimulated, sequentially (Fregni et al., 2008), or concurrently (Boggio et al., 2009a).

In their study, Boggio and colleagues (2009a) also found that active tDCS induced a decrease in cigarette consumption, although it was not clear whether this was only self-reported. Conversely, da Silva and colleagues (2013) reported increased alcohol consumption, but decreased craving in the group receiving active stimulation. Other studies reported no effect on craving following a single session of anodal tDCS of the left or right DLPFC in smokers and methamphetamine users (Xu et al., 2013; Pripfl et al., 2014). Shahbabaie and colleagues (2014) instead found an increase in craving for methamphetamine during a cue-challenge procedure in subjects receiving stimulation of the right DLPFC.

Two further studies focused on the effects of DLPFC tDCS on drug-taking behaviours. Fecteau and colleagues (2014) reported reduced cigarette consumption, but not carbon monoxide levels, following multiple bilateral stimulations, while Falcone and colleagues (2016) showed that a single session of active stimulation was associated with a small decrease in number of cigarettes smoked during the experimental session (17% less, compared to sham), while self-reported consumption over the following day was unaffected. Interestingly, ability to resist smoking, measured as the latency to smoke, was slightly increased by active tDCS. Therefore, it appears that tDCS of the DLPFC acutely decreases drug consumption, while more prolonged effects so far have only been seen for self-reported measures. Similarly, effects on craving were modest and commonly assessed immediately after stimulation was delivered, thus preventing evaluation of their durability.

While all the studies described above targeted DLPFC, three studies stimulated other regions. den Uyl and colleagues (2015) found that alcohol craving was reduced in response of tDCS of the left DLPFC but was not affected by tDCS of the right inferior frontal gyrus, an area involved in response inhibition (Jacobson et al., 2011; Ditye et al., 2012). Two studies reported that inhibition of frontal-parietal-temporal areas decreased self-reported cigarette consumption (Meng et al., 2014), and cue-induced heroin craving (Wang et al., 2016).

In conclusion, tDCS studies on craving and consumption of addictive agents have yielded inconsistent, and, at best, weakly positive findings. The small number of sessions delivered in the studies reviewed may explain these modest effects, as suggested by both tDCS and TMS studies in depression indicating that a therapeutic course of 20–30 sessions is necessary to observe significant clinical changes (Lam et al., 2008; Brunoni et al., 2016). Also, the use of different outcome measures limits the possibility to compare results from different trials, and to fully extrapolate tDCS efficacy.

In the absence of strong clinical effects, lack of imaging markers in these studies also prevents the evaluation of whether tDCS modulated the activity or connectivity of the target site or other brain regions. In this regard, it is important to consider the results of three studies, which investigated the effects of tDCS on cue reactivity via changes in electrophysiological activity in the DLPFC and other prefrontal areas (Nakamura-Palacios et al., 2012; da Silva et al., 2013; Conti et al., 2014). Specifically, these studies examined amplitude and current density of P3, an event-related potential component associated with attention arousal and motivational engagement in response to a stimulus. Two studies found that single and multiple stimulations of the left DLPFC increased P3 amplitude in response to drug-related stimuli (Nakamura-Palacios et al., 2012; da Silva et al., 2013). Conversely, Conti and colleagues (2014) reported an inhibitory effect on P3 drug-cued cortical activation following a single dose of bilateral tDCS, while multiple sessions produced opposite effects. Enhanced P3 evoked by drug cues has been reported in subjects with addictive disorders, and is positively correlated to craving and relapse risk (Petit et al., 2015). Hence, these preliminary findings seem to suggest that tDCS elicits activity in DLPFC in response to drug cues, and this may represent an increased motivated attention to drug-related stimuli, but may also reflect an attempt to exert cognitive control over craving. In fact, congruent with DLPFC activation in response to drug-related cues, human imaging studies have reported deficits in executive function that are reflected by decreases in prefrontal cortex activity (for a review see Goldstein and Volkow, 2011). In the next section, we will review in more detail the impact of tDCS on cognitive processing in subjects with addictive disorders.

Transcranial direct current stimulation effects on cognitive and executive functions

Dysfunctions in cognitive and executive functions such as attention, working memory, decision-making and response inhibition have been showed to play a critical role in driving and maintaining compulsive drug-seeking and drug use (Dalley et al., 2011; Goldstein and Volkow, 2011).

The effects of tDCS on these functions have been studied both in healthy volunteers and patients undergoing stimulation of DLPFC and have been shown to be mixed and controversial (Tortella et al., 2014; Tremblay et al., 2014), if not absent (Horvath et al., 2015). With regard to addictive disorders, two studies have evaluated effects of tDCS on frontal executive functions and cognitive mental status in alcohol-dependent subjects, also with contradictory results (da Silva et al., 2013; Klauss et al., 2014). In one study, multiple sessions of stimulation of the left DLPFC improved cognitive performances (da Silva et al., 2013). Conversely, Klauss and colleagues (2014) reported that repetitive bilateral stimulation did not induce any effect in alcohol-dependent subjects. Negative were also the results of a study evaluating tDCS effects on attention performances in smokers (Xu et al., 2013). Also, DLPFC tDCS in heavy drinkers did not influence alcohol-related associative processes, or biases due to them, measured by the implicit association test (den Uyl et al., 2015).

Risky decision-making has been implicated in risk and maintenance of addiction. Three studies specifically investigating whether DLPFC tDCS modulated risky decision-making have yielded mixed results (Boggio et al., 2010; Fecteau et al., 2014). Boggio and colleagues (2010) reported that a single session of bilateral stimulation in marijuana users increased propensity for risk-taking during the Risk Task (Rogers et al., 1999). However, in cigarette smokers, multiple sessions of tDCS over left DLPFC did not influence performance in the same task (Fecteau et al., 2014). Gorini and colleagues (2014) found that DLPFC stimulation of cocaine-dependent subjects inconsistently altered risky decision-making depending on the task: the game of dice task (GDT), which provides subjects with explicit rules for gains and losses (Brand et al., 2005), or the balloon analogue risk task (BART), in which participants have no information about probability of reward or punishment (Lejuez et al., 2002). Active stimulation of DLPFC (left and right) reduced risky behaviours in the BART task. For the GDT task risky behaviour increased after left DLPFC stimulation, but decreased following tDCS of the right DLPFC. Differences among these studies may be task-related. Some decision-making tasks involve risk ambiguity, and others do not, and differential neural substrates have been associated with these processes (Hsu et al., 2005; Weller et al., 2007).

Finally, only one study investigated a different target, namely the fronto-parietal-temporal area, and found that cathodal bilateral stimulation was associated with reduced attentional biases to smoking cues (Meng et al., 2014).

In conclusion, studies evaluating tDCS effects on cognitive and executive functions are few in number, and primarily inconclusive. They are characterized by important differences in terms of stimulation protocols, tasks, and severity of disease, all of which may affect these functions and their refractoriness to treatment. These findings, together with effects of tDCS on craving, bring into question whether tDCS may reduce drug-seeking behaviours by enhancing top-down cognitive control. Only two studies reported that improvement in cognitive processing in subjects receiving tDCS was associated with reduced craving (da Silva et al., 2013; Meng et al., 2014), another study reported decrease in craving but increase in risk-taking behaviours (Boggio et al., 2010), while in the other studies reviewed above, there was no relation between effects on craving or drug use and effects on cognitive processing. Of note, many studies were not designed to investigate this relation on the basis of specific hypotheses. It is important to recognize that ability to inhibit strong urges, such as craving, involves a complex prefrontal cortex-subcortical machinery (Volkow et al., 2008). TDCS of the DLPFC has not yet been demonstrated to be capable of modifying function of this circuitry in a specific way.

Transcranial direct current stimulation effects on mood

Dysphoria, anhedonia and increased stress reactivity are commonly observed in subjects with addictive disorders. These symptoms are related to dysfunctions in reward and stress circuits and also reflect impaired cognitive control over stress and emotion responsiveness (Davidson et al., 2002). An increasing number of studies has evaluated the antidepressant effects of tDCS of the DLPFC in psychiatric populations, with encouraging results (for a review see Kekic et al., 2016), while very few studies have specifically investigated effects on mood in subjects with addictive disorders. Three studies reported transient effects of anodal stimulation of the left DLPFC on subjective mood rating scales, with decreases in anxiety score (Batista et al., 2015), and depression scores while two other studies failed to observe any effect on mood (Fregni et al., 2008; Klauss et al., 2014). Overall, the effects on mood appear to be disjointed from effects on drug-seeking and taking behaviours, with only two studies reporting that improvement in mood was associated with decreases in craving (da Silva et al., 2013; Batista et al., 2015). Further studies are required to investigate the potential mood-altering action of tDCS in subjects affected by addictive disorders, as well as by comorbid depression and/or anxiety disorders.

Repetitive transcranial magnetic stimulation effects on craving, cue reactivity and consumption of addictive agents

Investigation of rTMS as a therapeutic intervention for addictive disorders began over a decade ago. As mentioned above, the majority of these trails have evaluated the anti-craving effects of TMS, leaving other symptom domains largely unexplored.

The earliest publications on rTMS effects on craving and consumption of addictive agents were two pilot studies conducted in smokers (Eichhammer et al., 2003; Johann et al., 2003). Johann and colleagues (2003) found that a single session of active rTMS of left DLPFC significantly reduced spontaneous nicotine craving, while Eichhammer and colleagues (2003) reported that active rTMS reduced the number of cigarettes smoked during an ad libitum session, but did not affect craving. Extending the treatment to 2 weeks, Amiaz and colleagues (2009) found that exposure to smoking-related cues followed by high frequency stimulation of the left DLPFC transiently reduced cigarette consumption (confirmed by urine cotinine levels), and cue-induced craving.

Subsequent open and blinded studies of rTMS of the left DLPFC followed, but results have varied. Two studies reported that a single session of high frequency stimulation decreased spontaneous craving in smokers not seeking treatment (Li et al., 2013a), and reduced cue-induced craving in subjects with heroin addiction (Shen et al., 2016). Two further studies failed to find any effect on craving and consumption in recently detoxified alcohol-dependent subjects receiving single (Del Felice et al., 2016), or multiple rTMS sessions (Hoppner et al., 2011). In two open-label pilot studies, multiple sessions of rTMS were reported to decrease spontaneous craving for cocaine (Politi et al., 2008; Terraneo et al., 2016), and also increased abstinence rates, as assessed by the number of cocaine-free urine drug tests (Terraneo et al., 2016).

Few studies have also investigated the effects of low frequency stimulation of the left DLPFC, with mixed results (Hayashi et al., 2013; Li et al., 2013b). Li and colleagues (2013b) demonstrated that 1 Hz rTMS increased craving in methamphetamine users not seeking treatment. In the study by Hayashi and colleagues (2013), low-frequency rTMS of the DLPFC and functional MRI were combined to obtain correlational and causal evidence of the role of the DLPFC in craving. First, the investigators demonstrated that cue-induced craving for cigarettes was dramatically increased when subjects were told that they would be given the opportunity to smoke immediately after testing (compared with 4 h later). Craving and expectancy of using drugs has been shown to activate DLPFC (McBride et al., 2006). To test the causal role of the increased activity of the DLPFC in craving, Hayashi and colleagues (2013) used rTMS to inactivate the DLPFC during exposure to the cues and showed that this prevented craving. This result alone is critical because it demonstrates that factors such as drug availability, expectancy and treatment status modulate DLPFC activity in response to cues.

Due to Hayashi et al.’s (2013) findings, it is clear that exposure to drug-related cues prior to or during rTMS may influence DLPFC activity differently on the basis of the perceived drug use opportunity, and this may lead to different responses to TMS. In addition to this, Hayashi and colleagues (2013) demonstrated that rTMS-induced decrease in DLPFC activity was associated with reduced craving-related signals in the medial orbitofrontal cortex, anterior cingulate, and ventral striatum, therefore providing some insights into the mechanisms of action of DLPFC stimulation.

While the majority of the studies reviewed above have investigated the effects of rTMS delivered to the left DLPFC, in few cases the right hemisphere has also been targeted. Mishra and colleagues (2015) reported reduced craving in subjects receiving multiple sessions of high frequency stimulation. Conversely, Herremans et al. (2012) reported lack of anti-craving effect following a single session of high frequency stimulation. The same authors also compared the effects on craving of a single session (sham-controlled) followed by multiple rTMS sessions (open-label), in hospitalized alcohol-dependent patients. Furthermore, they also used functional MRI to evaluate brain response to alcohol and neutral cues prior to and after rTMS. Cue exposure did not elicit activation of DLPFC, as well as of other regions associated with craving, including NAc, amygdala, anterior cingulate. Furthermore, changes in cue-induced craving ratings were modest and clinically not significant, both at the baseline and after rTMS was delivered. Considering that participants in this study were hospitalized and aware of the fact that alcohol was not available may explain these findings, as DLPFC is a region most influenced by perceived drug use opportunity, as discussed above (Wilson et al., 2004; McBride et al., 2006).

Also negative were the results of the only study investigating the anti-craving effects of low frequency rTMS of the right DLPFC in combination with nicotine replacement therapy, although the investigators reported that multiple sessions induced an increase in abstinence rates in nicotine-dependent subjects (Trojak et al., 2015).

Few studies have attempted to directly compare the effects of rTMS of left versus right DLPFC on drug-seeking and taking behaviours. In an open-label pilot study, Camprodon and colleagues (2007) found that only stimulation of the right DLPFC reduced craving in cocaine-dependent individuals, whereas Mishra et al. (2015) demonstrated that rTMS of both left and right DLPFC resulted in lower craving ratings in alcohol-dependent subjects.

In conclusion, the studies conducted to date suggest that rTMS of the DLPFC transiently reduces drug-seeking and taking behaviours, although only a small number of studies have investigated the effects of this intervention on consumption of addictive agents. Yet, after a decade of experimentation, these data do not provide a firm answer regarding the efficacy or the durability of rTMS of the DLPFC as a treatment for addictive disorders. Systematical replication is lacking, and effect sizes have been small and variable. However, prior to settling on a particular modality of DLPFC rTMS, important questions remain with regard to efficacy of low frequency stimulation, differential effects of right versus left hemisphere treatment, or unilateral versus bilateral stimulation. More importantly, the selection of the DLPFC as the primary target seems premature, given that several cortical and subcortical areas are implicated in addictive disorders.

Only in a few cases, have the effects rTMS on targets other than DLFPC been examined, and we are far from determining whether these significantly improve response. Rose and colleagues (2011) conducted a small pilot study targeting the superior frontal gyrus, a region implicated in cue-induced craving, using both 1 Hz and 10 Hz stimulation. Craving induced by smoking cues was elevated in the 10-Hz superior frontal gyrus condition, whereas craving was reduced after neutral cues. Hanlon and colleagues (2015) observed that continuous theta burst stimulation (a long-term depression-like form of TMS) of the left frontal pole/medial prefrontal cortex reduced stimulus-evoked activity in the ventral medial prefrontal cortex and in the NAc, and decreased craving in cocaine users. This study is remarkable for the use of a different frequency of stimulation and for suggesting the possibility to affect craving via modulation of medial prefrontal-striatal circuit. Data from this study indicate that TMS with traditional figure-of-eight coils, which can stimulate to depths of 2–3 cm, may modulate activity in subcortical structures via connectivity between the cortical region stimulated and the subcortical target. This is critical when considering that the use of figure-of-eight coils restricts TMS to superficial cortical targets, such as the DLPFC, while alternative, and maybe more efficient, targets for addictive disorders may include non-superficial (∼3–5 cm depth) brain areas such as medial frontal and orbitofrontal cortices, as well as deeper (∼6–8 cm depth) brain areas such as NAc, amygdala, insula and lateral habenula, among others. Of note, the incongruence of stimulating the DLPFC across different behavioural and clinical manifestations has already been observed with regard to TMS for depression and obsessive-compulsive disorder (OCD), more than a decade ago (Wassermann and Lisanby, 2001). Therefore, advances in understanding the neurobiology of these disorders impact on the technical characteristics of neuromodulation interventions, which limit the selection of more effective targets, and ultimately the efficacy of neuromodulation.

Stimulation of different, deeper brain structures may be achieved using H-coils that have slower electric field attenuation with depth, although at the expense of reduced focality and increased intensity of stimulation of superficial areas (Deng et al., 2014).

With regard to addictive disorders, two studies with H-coils have examined the effects of stimulating deep targets on craving and consumption of addictive agents. Ceccanti and colleagues (2015) found that multiple sessions of high frequency deep rTMS of the medial prefrontal cortex in alcohol-dependent subjects significantly reduced craving and alcohol consumption. The second was a large sham-controlled study in which low versus high frequency rTMS was delivered with or without cue exposure (Dinur-Klein et al., 2014). One of the novel aspects of this study was the use of an H-coil designed to stimulate multiple targets, specifically the DLPFC and insular cortex, bilaterally. The investigators found that high frequency stimulation versus sham significantly decreased cigarette consumption, as supported by urine cotinine levels. These effects were enhanced when subjects were exposed to cues prior to stimulation. The study by Dinur-Klein and colleagues (2014) introduced the possibility of stimulating multiple targets to address different subcircuits, a compelling point given that addictive disorders are caused by dysfunctions in multiple neural circuits. Furthermore, stimulation of deeper targets, as the insula, a region implicated in interoception, craving and self-awareness, may be valuable when treating disorders as addictive disorders, as well as OCD or depression. Nevertheless, further studies using imaging and/or behavioural markers of target engagement are necessary to evaluate how activity in deeper regions is modulated and in which direction compared to the overlying areas.

Finally, a limited number of studies have considered whether rTMS affected other clinical and behavioural manifestations of addictive disorders and whether this was associated with reduced drug-seeking and drug use. Hoppner and colleagues (2011) reported that high frequency stimulation of the left DLPFC increased attentional blink effect to alcohol-related pictures and did not affect mood and craving. A further study found no effect of high frequency rTMS of the right DLPFC in response inhibition performances, as measured by a Go/No-Go task (Herremans et al., 2013). Recent work suggests that in subjects with comorbid addictive disorders and dysthymia high frequency rTMS of the bilateral DLPFC reduced craving and improve depressive symptoms compared to standard pharmacological treatment (Girardi et al., 2015).

Summary and future directions

Studies investigating the therapeutic potential of tDCS and rTMS for addictive disorders are generally small in size, variable in design, and—as may be related—conflicting in outcomes. Other sources of variability across studies include differences in stimulation parameters, concomitant therapies, and patient sample characteristics.

In terms of stimulation paradigms, in tDCS studies duration of single sessions varied from 10 to 30 min, and intensity of stimulation has ranged from 0.45 to 2 mA. Similar observations can be made for rTMS studies. Duration of rTMS sessions ranged between 10 and 20 min. The majority of the studies used high frequency stimulation, but overall the frequency varied from 1 to 20 Hz. The intensity used, as a percentage of resting motor threshold, ranged from 90% to 120% and the motor threshold was usually determined visually and only in a few cases using EMG, which is more accurate and reproducible. The total number of pulses, an essential parameter for rTMS efficacy (Gershon et al., 2003), varied and comprised between 1000 and 2000 in most trials. In rTMS studies, the identification of the target site, generally the DLPFC, was usually done using scalp landmarks, a method that could lead to inaccurate targeting compared to MRI-based neuronavigation systems (Fitzgerald et al., 2009). Therefore, simple and economical methods for precise and reliable coil placement are needed, as this factor is important for effectiveness (Wassermann and Lisanby, 2001; Grall-Bronnec and Sauvaget, 2014).

It is also important to consider that a large number of studies used a cross-over design, which may lead to misinterpretation of findings, because delayed and progressive effects have been described for both tDCS (Fregni et al., 2007) and rTMS (Hallett, 2007), and these effects are usually identified by parallel group trials. Also several studies failed to have a sham control session, or the sham procedure used did not allow blinding of participants, especially in rTMS studies.

Another important factor concerns the number of sessions received by study participants, which ranged from 1 to 15, and the inclusion of follow-up visits. The studies reviewed above provide a limited evidence on the persistence of rTMS addiction treatment effects: it appears that a single session of high frequency rTMS of the right DLPFC reduced spontaneous craving for less than 4 h (Camprodon et al., 2007), whereas multiple sessions were associated with more prolonged effects (Dinur-Klein et al., 2014; Terraneo et al., 2016). Therefore, it is crucial to investigate and develop maintenance strategies if non-invasive neuromodulation interventions are to move to clinic.

Interpretation of the results of rTMS and tDCS is also limited by the size and the characteristics of the sample. Small sample sizes characterized the majority of trials, especially in tDCS studies. Furthermore, the populations studied were clinically heterogeneous. Participants showed different patterns of consumption, did or did not wish to quit, or were already receiving pharmacological or other treatment. Duration of abstinence prior to treatment ranged from 1.5 h to more than 12 h. We have already noticed that these factors can crucially affect craving, particularly cue-induced craving, which constituted the outcome measure of the majority of trials (Wilson et al., 2004). In this regard, a meta-analysis recently conducted to evaluate the effects of tDCS and rTMS on craving (Jansen et al., 2013) found that non-invasive neuromodulation significantly reduced craving compared to sham. This finding should be interpreted in light of the limitations we have discussed. Furthermore, anti-craving effects, or the lack thereof, do not automatically translate into reduction in drug consumption or harm reduction. In several trials effects on craving and consumption were disjointed. Whether this reflects lack of causal linkage or difficulties in assessment is an open question as very few studies have tried to investigate the mechanisms of action of neuromodulation interventions. This knowledge will help identify optimal targets and refine stimulation protocols. For example, stimulation of the DLPFC has been associated with change in dopamine binding in the caudate/dorsal striatum (Strafella et al., 2001), decrease in EEG delta power (Pripfl et al., 2014), decreased activity in the medial orbitofrontal cortex, anterior cingulate, and ventral striatum (Hayashi et al., 2013), and change in brain connectivity (Jansen et al., 2015).

Lack of significant therapeutic effects, with regard to craving and consumption, as well to other symptom domains of addictive disorders, may also be related to the choice of the target area. We have previously discussed how the majority of studies evaluating non-invasive neuromodulation have targeted the DLPFC area. Although amplifying executive control circuits in an attempt to improve cognitive control can be a valid therapeutic strategy, the neurobiology of addictive disorders suggests that stimulation of several other nodes within the same or other circuits can be clinically useful. An important limitation of non-invasive neuromodulation interventions is the depth of penetration of current by tDCS and current-inducing fields by TMS. Novel TMS coils have been designed to stimulate deeper areas but these coils also activate broad regions of overlying cortex, and other structures (Deng et al., 2014). One way to modulate deep targets is to identify cortical windows, which can be specific to the individual patient, that preferentially activate these areas. Another alternative would be to specifically alter the magnetic susceptibility of the target tissue, making it especially easy to activate with alternating magnetic fields. Finally, there has been a surge of interest in focused ultrasound (FUS). This was initially proposed as a lesion technology similar to stereotactic radiosurgery (Elias et al., 2013) or as a way to deliver chemotherapy to tumours (Liu et al., 2012). More recently, it has been shown that low intensity FUS can increase or decrease neural activity, in focused brain regions, without damaging the tissue (Yoo et al., 2011; Legon et al., 2014).

In conclusion, both the efficacy and real indications for the use of tDCS and rTMS in addictions remain to be confirmed. Although studies are heterogeneous, many critical answers remain open, including the optimal stimulation parameters, whether broader or bilateral stimulation should be preferred to focal and/or unilateral stimulation, the appropriate patient population and the opportunity to combine other treatments. More work and larger studies are needed, but these studies should be based on specific, testable pathophysiological hypotheses and robust electrophysiological effects (Table 3). Physiological measurements, including EEG, PET and functional MRI, should be associated with clinical outcomes to provide mechanistic insights and to unravel effects on different behaviours. Furthermore, together with clinical and genetic measures, these measures may constitute predictors of treatment outcomes. For example, severity of the disease or genetically-driven variability in cortical excitability has been linked with different responses to rTMS (Sturgess et al., 2011; Li et al., 2013a).

Table 3.

Indications for further neuromodulation studies for addictive disorders

Non-invasive interventions (rTMS and tDCS)
Disease selection	Alcohol Opioids Stimulants Others
Population selection	Patients with: Mild ADs Moderate ADs Severe ADs	No comorbidity Comorbid ADs Comorbid ADs and other neuropsychiatric disorders
Target selection	Test targets other than the DLPFC
	Evaluate single versus multiple targets
Study design	Develop sham- controlled, randomized, double-blind studies, based on larger sample groups. Avoid cross-over design
	Increase study follow-up to examine the long-lasting effect of rTMSReplicate trials
Stimulation parameter selection	Choose and evaluate different stimulation parameters according to the target site, on the basis of physiopathological hypotheses
	Describe stimulation parameters accurately and justifying them
Identification and quantification of outcomes measures	Couple subjective assessment and objective and physiological measurements (fMRI, EEG, etc.)
	Include disease related outcomes other than craving
	Assess effects on mood, cognition and quality of life
	Measure neurophysiological and neuroimaging variables regularly throughout studies
	Develop clinical and other predictors of treatment outcome
	Investigate mechanisms of action of neuromodulation
Context of stimulation	Evaluate whether neuromodulation is more efficacious while delivered during task or at rest
	Investigate the effect of neuromodulation as an add-on therapy
Invasive intervention (DBS)
Efficacy	Further evaluate DBS efficacy in animal models, testing different targets and parameters, and considering issues related to translatability
	Collect data on drug-seeking and drug-use in patients undergoing DBS for other diseases (essential tremor, dystonia)
	Investigate whether DBS for addictive disorders should be associated to other therapies
Feasibility and safety	Consider issues related with patients’ recruitment and retention in study (severity of illness, proof of treatment-refractory, psychosocial and familiar support)
	Evaluate issues associated with study drop-outs in terms of safety, device maintenance and/or removal

AD = addictive disorder; fMRI = functional MRI.

Deep brain stimulation: an overview

DBS is a neurosurgical procedure that involves the stereotactic implantation of unilateral or bilateral electrodes connected to a stimulator device, which is implanted subcutaneously below the clavicle (Schlaepfer and Lieb, 2005). During surgery microelectrode unit recordings are used to confirm accuracy of electrode placement, with the opportunity of recording activity from single or populations of neurons, and more recently also neurochemical release, from single or populations of neurons, at rest or during task execution (Shah et al., 2010). This has generated data that have improved our understanding of how different neural targets are involved in diverse functions (Hutchison et al., 1998; Zaghloul et al., 2009; Sheth et al., 2012). Nevertheless, microelectrode mapping comes at the cost of increased risk of intraoperative haemorrhage and stroke (Sansur et al., 2007) and therefore is being replaced by interventional MRI, which allows for real-time imaging-guided implantation of DBS electrodes.

Once electrodes are implanted, electrical stimulation can be finely adjusted in terms of frequency, pulse width and choice of bipolar or monopolar stimulation to maximize clinical benefits and avoid unwanted adverse effects. Stimulation can be delivered continuously, can be rapidly cycled on and off for fractions of a second to several seconds and recently, in patients undergoing DBS for movement disorders, recordings from electrodes have also been used to optimize the stimulation parameters to best manage symptoms (‘closed loop’ stimulation), a promising approach for treating psychiatric disorders.

The specific mechanisms of action of DBS are still under investigation. First, DBS affects multiple neural elements, but foremost myelinated axons. Second, experimental evidence indicates that effects of DBS are via stimulation-induced modulation of brain activity (Montgomery and Baker, 2000; McIntyre et al., 2004; Kringelbach et al., 2007; McIntyre and Hahn, 2010), rather than competing hypotheses such as synaptic inhibition (Dostrovsky and Lozano, 2002), depolarization blockade (Beurrier et al., 2001) or synaptic depression (Urbano et al., 2002). In addition to functional changes in neural circuits, DBS also appears to induce neuroanatomical remodelling at the cellular level (Chakravarty et al., 2016). DBS effects can be both local and remote across brain networks, and may be immediate, such as tremor cessation following thalamic stimulation, or delayed and progressive as in depression or dystonia, suggesting that multiple mechanisms with different kinetics are involved (Lozano and Lipsman, 2013). Also, DBS effects vary with the stimulation parameters, with the intrinsic physiological properties, and with the interaction between the electrode and the geometric configuration of the surrounding neural tissue and specific anatomy of the targeted region (Kringelbach et al., 2010).

In terms of side effects, DBS is associated with important operative and perioperative risks, while long-term effects represent an active area of investigation. Among short-term adverse effects, the most serious, such as brain haemorrhage, are rare, occurring in less than 1–2% of patients, with less serious, typically reversible events, such as wound infection and stimulation-related side effects, occuring in up to 9% of patients (Hamani et al., 2008). Other potential risks associated with DBS range from unexpected psychiatric side effects (e.g. mania, impulsivity) to more global changes (e.g. changes in personality).

Both from a research and clinical perspective, DBS presents unique challenges. Despite being reversible and non-ablative, DBS is an invasive procedure associated with several risks, and availability is related to the number of centres and neurosurgeons capable of installing the devices and delivering the aftercare. Also, DBS is an expensive procedure that imposes maintenance costs averaging several thousand dollars per year (Carter and Hall, 2011). It has been reported that financial burden and fear of adverse effects are key reasons for hesitancy or refusal of subthalamic nucleus (STN) DBS among eligible candidates with Parkinson’s disease (Kim et al., 2016). On the other hand, this intervention has been shown to be highly effective in patients with severely impairing conditions, not responding to pharmacological or other treatments. Currently, DBS has been approved by the FDA for the treatment of essential tremor and Parkinson’s disease, and a humanitarian device exception has been granted for dystonia and obsessive-compulsive disorders.

Considering exclusively the research perspective, DBS posits several challenges, including target identification, patient selection and recruitment, need of long follow-up phase, and necessity to provide support and/or device removal to subjects that withdraw from DBS studies.

Deep brain stimulation in addictive disorders

The published literature on DBS for the treatment of addictive disorders is sparse, and includes several case reports of patients with comorbid psychiatric disorders (Kuhn et al., 2007a,b, 2009; Mantione et al., 2010; De Ridder et al., 2016). In the first, a patient was treated unsuccessfully for agoraphobia by bilateral DBS of the NAc but his comorbid alcohol dependence was ameliorated (Kuhn et al., 2007a). The same group reported smoking cessation in 3 of 10 patients who underwent DBS of the NAc, for Tourette’s syndrome, OCD or anxiety (Kuhn et al., 2011). A similar observation was described by Mantione and colleagues (2010). Here, a patient who had undergone NAc DBS for treatment-refractory OCD quit smoking and lost weight post-surgery.

Although in some of the cases described above decrease in drug-seeking and taking seems attributable to the successful treatment of the primary psychiatric condition, these observations supported the investigation of NAc DBS in patients with severe, treatment-resistant addictive disorders (Muller et al., 2009, 2016; Kuhn et al., 2011, 2014; Zhou et al., 2011; Valencia-Alfonso et al., 2012). Kuhn and colleagues (2011) reported that following NAc DBS an alcohol-dependent patient significantly reduced his alcohol consumption and achieved abstinence at 1-year follow-up. The investigators also found that electrophysiological markers related to error-processing were changed following DBS, and hypothesized that DBS affected decision-making processes. The same group also investigated the effects of NAc DBS in two patients with opiate addiction and concomitant use of other substances (Kuhn et al., 2014). Following DBS, patients reported abstaining from heroin except on a few occasions, as confirmed by urinary drug screen, but they continued using other psychoactive substances. Interestingly, the authors felt that continued use was motivated by persisting difficult life conditions and lack of adequate coping skills, suggesting the importance of a multidisciplinary approach in treating addictive disorders.

Effects of NAc DBS were further evaluated in another subject with a 5-year history of heroin dependence (Zhou et al., 2011). During the following 6 years, the patient did not relapse and, in addition, markedly reduced his nicotine consumption. These effects persisted not only when stimulation was interrupted 24 months after surgery, but even after complete removal of the entire DBS system. Positive results were also reported by Valencia-Alfonso and colleagues (2012), who performed DBS of the NAc and anterior limb of the internal capsule in a patient with heroin addiction. Interestingly, in this study EEG recordings during presentation of drug-related and neutral cues were used to set stimulation parameters. The patient reportedly relapsed on several occasions 6-months post-surgery. Notably at the time of recruitment, this patient used heroin daily while receiving methadone (20 mg/day). Potentially, this patient was not treatment-resistant because the methadone dose was far from the standard dose necessary to achieve anti-craving (WHO, 2009).

Muller and colleagues (2016) recently reported results of the only pilot study of NAc DBS in alcohol addiction. The first three cases were previously described (Muller et al., 2009), and subsequently two additional patients were enrolled, all receiving bilateral NAc DBS. One patient was reported to be abstinent for 8 years and a second was followed for 6 years, during which he did not relapse. At the end of this period, he dropped from the study and the battery of his device went out of power. The other three patients relapsed at various intervals after receiving DBS (the longest abstinence period being 20 months). Two of these patients died, most likely due to severe alcohol withdrawal.

The results of this study raise several important questions. First, although all patients reported that cue-induced craving vanished following DBS, three relapsed and these episodes were generally triggered by negative emotional states. This seems to indicate that successful treatment for alcohol addiction, as well as for addictive disorders in general, should address but not be limited to reduction of craving and cue responses. Second, it appears extremely challenging to recruit patients with addiction in DBS studies and also to perform long-term follow-up, as previously noted (Luigjes et al., 2015). Finally, the two reported deaths due to severe alcohol withdrawal challenge the efficacy of DBS, at least in significantly reducing alcohol consumption. In particular, the NAc may not represent the most effective target for patients in whom habit learning and emergence of negative states, which underlie alterations in several other neurocircuits, mainly drives alcohol and/or drug consumption.

In this regard, two recent case reports have evaluated the effects of modulation of different target areas to improve addiction-related symptoms. In the first, DBS was performed in a subject with refractory cocaine dependence and the target was the posterior-medial part of the anterior cingulate and the bed nucleus of the stria terminalis (Goncalves-Ferreira et al., 2016). After 2.5 years, including a blind on/off period, DBS resulted in a significant decrease of both subjective and objective measures of cocaine dependence, but not in full remission. Notably, no differences were detected between on/off blind periods suggesting the emergence of DBS delayed effects or placebo effects. Also, it was not investigated or reported whether targeting brain regions involved in modulation of emotional responses to stimuli, stress and anxiety, affected these domains that are critically involved in craving and relapse.

In the second case report, De Ridder and colleagues (2016) described the results of DBS of the dorsal anterior cingulate/supplementary motor area in a subject with alcohol dependence associated with anxiety and agoraphobia. It is not clear whether the patient presented with treatment-resistance also with regard to therapies for anxiety. Before surgery, functional MRI was used to assess blood oxygen level-dependent responses to alcohol and neutral cues and resting state EEG data were recorded to evaluate baseline activity. Imaging data showed increased activity of the dorsal anterior cingulate cortex (dACC) during presentation of alcohol cues, and 1 Hz TMS of the dACC using a double-cone coil during a 2-week period resulted in a transient reduction of alcohol craving. Such procedures were used to identify the target area as well as to verify whether modulating dACC activity would show clinical effects, suggesting that imaging techniques and non-invasive stimulation can play a key role in guiding and probing target selection. In this patient, DBS of the dACC resulted in progressive decrease in self-reported craving, alcohol abstinence and remission of anxiety and agoraphobia during the follow-up period (18 months). Despite the obvious limitations of a single case report, these findings suggest that a single target could be effective among different disorders, raising the question whether it could be appropriate to study in isolation disorders that share common pathways and behavioural alterations. More specifically, the question would be whether selection of DBS targets should be based exclusively on diagnostic criteria or should also be guided by specific behavioural phenotypes linked to the circuits and ‘nodes’ within brain networks we are implanting. The implication of such a transdiagnostic, neuroscience-based approach is represented by the possibility to use the same brain targets for different diseases, but more importantly to use different targets among subjects affected by the same disorder but presenting with clinical and behavioural alterations in some domains instead of others.

Summary and future directions

Research investigating the potential role of DBS in the treatment of addictions has produced mixed results that mainly derive from single case reports. This limitation can be addressed by larger trials but this proposal raises several issues (Table 3). The feasibility of DBS studies in addictive disorders appears difficult, as shown by several unsuccessful attempts to conduct clinical trials (Luigjes et al., 2015; Muller et al., 2016). There are a number of possible reasons that can explain why patients with addictive disorders are more difficult to recruit than patients with OCD or Parkinson’s disease. First, patients with addictive disorders frequently have impaired self-awareness and insight (Goldstein et al., 2009) that may lead to the false belief that one is in control over drug-taking behaviour, leading to underestimation of the severity of illness and need for treatment. This may also affect the willingness to participate in long-term studies requiring invasive procedures. Also, the concept of addictive disorders as a medical condition and a brain disorder is still questioned, both by the general public and by clinicians.

This aspect also has implications for patient selection. Given the invasive nature of DBS, and the risks associated with it, this procedure is generally performed in subjects with diseases that are refractory to standard treatments. For addictive disorders, it can be difficult to define treatment resistance because many patients with addictive disorders have not been adequately treated whether because of availability of treatment, negative expectations of the clinicians, or because the patient perceives seeking treatment as stigmatizing (Knudsen et al., 2011; Oliva et al., 2011; Spagnolo et al., 2015).

Another relevant issue concerns DBS efficacy. This is related to the choice of target area, the optimization of stimulation parameters and the possibility to combine other therapies. So far, the NAc has been targeted in all the studies on addictive behaviours, with two exceptions (De Ridder et al., 2016; Goncalves-Ferreira et al., 2016). Findings from preclinical studies indicate that other targets may be potentially effective, including the medial prefrontal cortex (Parthoens et al., 2014), the lateral habenula (Friedman et al., 2010), the amygdala (Langevin, 2012), and the subthalamic nucleus (Baunez et al., 2005; Rouaud et al., 2010; Lhommee et al., 2012). For this latter target, there are also human data showing that STN DBS is associated with reduced abuse of dopaminergic medications and improved decision-making under risk in patients with Parkinson’s disease (Lhommee et al., 2012; Boller et al., 2014), and with decrease in the compulsive component of the disease in patients with OCD (Mallet et al., 2008). With regard to the amygdala, DBS of this target (specifically the basolateral nucleus) has been reported only in two cases to treat autism and related self-injurious behaviour (Sturm et al., 2012), and refractory post-traumatic stress disorder (Langevin et al., 2016). Both patients experienced symptoms improvement and only transient side-effects, but DBS of the amygdala is far from an established practice. Its safety and therapeutic efficacy needs to be confirmed, and the outcomes of targeting a structure critically involved in both positive and negative reinforcement are still elusive. In fact, although data from animal studies identify the amygdala or other structures as potential effective DBS targets to reduce addictive behaviours, these findings may not be readily transferable to apparently related conditions in humans. Anatomical and functional variation between animals and humans, shorter duration and larger amplitude of stimulation used in animal studies, and also lack of animal models of addiction with high predictive value, can represent important constraints to translational research of DBS in addiction. Furthermore, the complexity of the human brain architecture, the techniques currently available, as well as electrode size and configuration, may limit the selection of certain targets. Therefore, identifying functionally homologous areas, or sites that fall within the same network and which stimulation can indirectly modulate activity in other target regions may represent a successful strategy.

DBS efficacy is also crucially related to the choice of stimulation parameters able to suppress symptoms without causing side effects. Currently, DBS for psychiatric disorders is mainly a static therapy. Stimulation can be titrated to suppress symptoms, but once the ‘dose’ has been identified, usually it is continuously applied for weeks at a time. Addictive disorders, meanwhile, are not static. Craving, anxiety and stress, which all play a key role in inducing relapse, are often transient, lasting from minutes to days. Therefore, DBS may benefit from systems that operate by tracking an electrical signal in the brain and changing the stimulation parameters to drive the tracked signal to a desired range. Such closed-loop systems have already shown to succeed in patients with neurological disorders, as mentioned above (Rosin et al., 2011; Beuter et al., 2014). Their applicability to patients with psychiatric disorders and with addictive disorder critically depends on the identification of biomarkers, intended as brain signals that can provide the necessary feedback and symptom state detection, allowing for more nuanced and individually tuned neural modulation.

Finally, several observations seem to point to the necessity of combining DBS with other interventions to refine its effects, and increase its efficacy. This is suggested by a recent animal study in which low-frequency DBS, refined by selective blockade of dopamine D1 receptors, abolished behavioural sensitization to cocaine and restored normal synaptic transmission similarly to optogenetic interventions (Creed et al., 2015). In addition to increasing DBS specificity, pharmacological and other interventions may be integrated with DBS treatments to target other neurocircuit dysfunctions that maintain addictive behaviours. It is also interesting to note that DBS appears to improve response to therapies to which patients were previously refractory. For example, adding structured cognitive behavioural therapy (CBT) for OCD patients whose degree of response had plateaued from NAc DBS resulted in symptoms improvement (Widge et al., 2016). This supports the notion that modulation of pathological activity within brain networks represents an essential therapeutic mechanism of DBS.

Conclusions

Compelling preclinical studies in animals, and neuropsychological and neuroimaging studies in people, identify addictions as brain diseases for which interventions at the neurocircuit level are feasible, and probably more specific and effective than any therapy currently available. Evidence from small clinical trials seem to support this hypothesis, although these findings are still preliminary. Nevertheless, the hope of developing drug-free interventions that remodel dysfunctional neural circuits sustains this work despite the difficulty of studying incompletely understood procedures with several dosing parameters and technical limitations, and often without clear hypotheses about their mechanisms of action. The future of neuromodulation for addictive disorders and for other psychiatric disorders depends on designing studies that will provide answers for these crucial questions unless we wish to keep ‘waiting for Godot’.

Funding

This work has been funded by Z99 AA999999/Intramural NIH HHS/United States.

Glossary

Abbreviations

DBS

deep brain stimulation

DLPFC

dorsolateral prefrontal cortex

NAc

nucleus accumbens

OCD

obsessive-compulsive disorder

rTMS

repetitive transcranial magnetic stimulation

tDCS

transcranial direct current stimulation