Outcome of Non-Invasive Brain Stimulation in Substance Use Disorders: A Review of Randomized Sham-Controlled Clinical Trials

Among biomedical bibliographic databases, MEDLINE is the largest and the most widely used in the world.³³ Therefore, a comprehensive literature search was conducted on MEDLINE via the PubMed database up to 1st August 2016. All papers excluding editorials, reviews, practice guidelines, meta-analyses, and papers on infants and animals were evaluated using a set of inclusion criteria.

First, we used a free-text search of all relevant fields to retrieve the studies of interest. The text-word terms selected were based on the names of the most usual substances (and their derivatives) and behaviors encountered in the field of addiction. For the NIBS, our review focused on tDCS and rTMS, including the new form of rTMS called theta burst stimulation (TBS), as well as the application of electroconvulsive therapy (ECT), since some authors regard this nonfocal technique as NIBS.³⁴ We thus applied a broad search strategy including the above terms and the following formula: (“electroconvulsive therapy” OR “transcranial magnetic stimulation” OR “Theta-Burst Stimulation” OR “transcranial direct current stimulation”) AND (addict* OR “alcohol” OR “nicotine” OR “tobacco” OR “smoking” OR “opiate” OR “cocaine” OR “amphetamine” OR “cannabis” OR “hallucinogen” OR “inhalant” OR “gambling” OR “Internet”). As SUD does not apply to caffeine, we did not include this substance in the present formula (DSM-5).

Secondly, we used the PubMed filter (article types, species) to select only “clinical trials” or “randomized controlled trials” involving “humans.”

Thirdly, two independent reviewers (B.T. and S.K.) read the abstracts obtained and included studies according to their contents to check inclusion criteria. In case of disagreement, a third reviewer (S.A.) made the decision after reading the title and abstract.

Finally, the full manuscripts of the selected abstracts were read in order to confirm that these studies were in line with the inclusion criteria.

Studies had to meet the following criteria to be included:

As a sham-controlled condition was one of the inclusion criteria, for each of the selected studies we also determined whether the reliability of the sham conditions was tested.

The electronic text-word search on PubMed retrieved 592 articles. In addition, one additional study not selected using PubMed filter but meeting the inclusion criteria of our review was added.³⁵ After selecting only studies according to our selection criteria, 26 articles were identified. Among these articles, the studies used rTMS, tDCS, and TBS, but none used ECT (Figure 1).

The cortical target for all the selected studies was the DLPFC. With tDCS, the 10–20 system was used to supposedly reach the DLPFC by placing the electrodes above F3 and F4.³⁶

The largest studies were related to tobacco use disorders (N=16) and alcohol use disorders (N=7), with smaller numbers of studies for methamphetamine (N=2) and marijuana (N=1) and no studies for inhalant, hallucinogen, gambling, and Internet gaming disorders (Tables 1 and 2). No studies examined the concomitant effects of NIBS on two addictive disorders at the same time, although it is usual to find the consumption of two drugs together in clinical practice, especially the combined use of alcohol and tobacco. It should be noted that in the majority of these studies, the patients involved continued to consume (N=16) or had become abstinent (N=10). This means that the therapeutic results achieved in these studies suggest that NIBS may be considered either in a reduction or an abstinence strategy. It should also be noted that the number of stimulation sessions was very low for the vast majority of these identified studies.

In terms of target symptoms, craving reduction was the most commonly used efficacy endpoint assessed in the studies (N=24). The effect of NIBS on craving levels was measured by comparing the scores before and after NIBS, either in the change from baseline without exposure to cues associated with a given drug (N=10), or in the change in the variation of craving level after exposure to cues (N=9). Five studies included a baseline assessment of craving followed by exposure to a drug cue designed to increase or elicit craving.^24,37–40

Regarding the outcomes of the 24 studies that assessed the effect of NIBS on craving, more than half of them found a reduction in craving (N=13): five studies found a reduction in baseline craving (in state, at rest),^{15,25,39,41,42} six studies a reduction in cue-induced craving,^{16,35,43–46} and two a concomitant reduction for the two craving assessments.^37,38 Studies that found no craving reduction with NIBS (N=11) used both methods to measure craving. Thus, it seems that the manner in which craving was measured did not significantly influence the results. In contrast and surprisingly, two studies in methamphetamine-dependent users found that low-frequency (LF)-rTMS over the left DLPFC and anodal tDCS over F4 increased self-reported craving during exposure to methamphetamine-related pictures.^39,47

Eight of the selected studies assessed the effects of NIBS on tobacco consumption (Tables 1 and 2). Except for Sheffer et al.’s study, which reported no significant effect of high-frequency (HF)-rTMS on cigarette consumption,²¹ and for Smith et al.’s study, which reported no significant effect of anodal tDCS on the number of cigarettes smoked in patients with schizophrenia,⁴⁰ all of the other studies found a reduction in the number of cigarettes smoked following HF-rTMS or tDCS.^{24–26,37,43,48} Three studies also found a significant reduction in the number of relapses in abstinent patients with tobacco or alcohol disorders, up to 3 or 6 months after the brain stimulation according to two of these studies.^49–51 All these results are very encouraging because whatever the mechanism implicated in the therapeutic effects of NIBS, reduced consumption and maintenance of abstinence are the ultimate therapeutic goals for patients with SUD.

In the present review, four studies examined the effects of NIBS on different tasks used to assess risk-taking impulsive behaviors and attention^21,22,26,52:

Sheffer et al. assessed the effects of NIBS on delay discounting, a cognitive process that allows the individual to compare values between the immediate and delayed consumption of a determined commodity: delay discounting is the degree to which one de-values delayed outcomes.^21,53,54 Its assessment is useful to study impulsive decision making associated with SUD and may explain self-control failure in addictive behavior. Using a delay discounting task, Sheffer et al. found that HF-rTMS in smokers decreased delay discounting of monetary gains, suggesting that neuromodulation of the DLPFC can enable individuals to make less impulsive and/or more future-oriented decisions.²¹

The risk task, a binary decision-making exercise useful to measure both impulse control and risk-taking behavior, was used in two tDCS studies.^26,55 In this task, which uses colored boxes on a screen to which an amount of money has been attributed, participants must decide whether to select the unlikely option, which could generate sizeable rewards, or the likely option, which could generate modest rewards.⁵⁶ In the first study, marijuana users demonstrated an increased likelihood to select the more-risky prospect during tDCS of the DLPFC, even though the stimulation significantly decreased craving for marijuana. In the second study, active tDCS did not significantly influence performance on the risk task in smokers, regardless of whether rewards were money or cigarettes.²⁶ However, using the ultimatum game, in which subjects can accept or reject offers of money or cigarettes, the authors found that smokers were more likely to reject offers of cigarettes after they received active tDCS than after sham.²⁶ The authors also found a significant decrease in the number of cigarettes smoked when participants received active tDCS compared with sham, suggesting that stimulation with tDCS may help smokers to control their behavior by acting on the decision-making process with cigarettes as rewards.

A Go-NoGo task was used in the latest study.⁵² This task is used to measure a participant's capacity for sustained attention and response control. In the Go-NoGo task, while participants see a salvo of stimuli presented in a continuous stream, they have to make a binary decision for each stimulus and press a button to sate it. Accuracy and reaction time are measured for each decision. Using this technique, Herremans et al. found in detoxified alcohol-dependent patients that active HF-rTMS compared with sham decreased intraindividual reaction time variability (dispersion of the reaction times), reflecting greater stability in attentional mechanisms.⁵²

Höppner et al. studied the influence of rTMS on an attentional blink (AB) paradigm to emotional and alcohol-related pictures in patients with alcohol use disorder.⁵⁷ The AB paradigm tests the limits of attentional capacities and the role of emotions on these, when rapid succession of visual stimuli are seen at the same spatial location on a screen. It is based on the phenomenon that, within a brief period after presentation of a first target stimulus (T1), attentional resources cannot be allocated adequately to a subsequent second target stimulus (T2), and thus the AB is lengthened. However, if T2 is emotionally relevant, the AB is reduced. In their study, Höppner et al. used alcohol-related pictures as T2 target stimuli.⁵⁷ They found that real rTMS compared with sham increased the AB effect, namely that patients were less able to detect the alcohol-related pictures after the real HF rTMS session above the left DLPFC, thus suggesting that the rTMS may modulate emotional regulation and attentional abilities in SUD.

Among the studies included in the present review, eight assessed the effects of NIBS using rTMS or tDCS on mood and one on anxiety.^{16,35,37–39,44,51,57} This is relevant because compared with the general population, people addicted to drugs are roughly twice as likely to suffer from mood and anxiety disorders.⁵⁸ Moreover, rTMS and tDCS may be efficacious for the treatment of depressive disorders and anxiety disorders.^59–61 Da Silva et al. found significant decreases in depression scores in detoxified alcohol-dependent patients treated with tDCS. However, in all of the other studies, no changes in depression or anxiety symptoms were observed.^{16,37,38,48,57}

Da Silva et al. and Klauss et al. performed cognitive assessments before and after tDCS in patients with alcohol use disorders, but there was no statistical difference between active and sham stimulation.^44,51 In these two studies, one but not both also found an improvement in perceived quality of life.

In the majority of rTMS studies, the treatment was focused on the left-DLPFC using HF stimulation 10–20 Hz (N=8) (Table 1). This stimulation parameter seems to be particularly interesting because it leads to a large clinical effect, ranging from a decrease in craving to a reduction in consumption. We must note, however, no significant improvement in craving symptoms in patients with alcohol use disorder even though they underwent 10 HF-rTMS sessions that included up to 10,000 pulses.⁵⁷ Using the other stimulation parameters for rTMS, such as LF over the left DLPFC, and HF or LF over the right DLFPC, the results are more nuanced. In fact, even when the latest studies are included, it is difficult to identify the most effective stimulation parameter, as these studies are extremely heterogeneous with regard to TMS characteristics (i.e., number of sessions, number of pulses, laterality of treatment, percentage of motor threshold). Moreover, clinical assessments vary substantially from one study to another, in terms of the number of symptoms studied, the way they were assessed, and the duration of the follow-up (Table 1).

In tDCS studies, the issues raised are similar to those found in rTMS studies (Table 1). Studies differed, for example, for the duration of the stimulation session (10–30 minutes), the number of sessions delivered (1–5 sessions), and, above all, the laterality of the anode (F3 or F4), which revealed contradictions about the outcome on craving: some studies reported no significant difference using anodal stimulation over F4,^26,51 whereas those that compared anodal stimulation to F3 and F4 found either that only the latter was able to significantly reduce the craving⁵⁵ or that both stimulation conditions were effective in decreasing craving.^16,38

Among the 26 articles selected, only three tDCS studies tested the effectiveness of their blinding condition.^26,40,51 No rTMS studies assessed the reliability of their sham condition.

The measurement of craving is widely used by clinicians and researchers. This is also true for research using NIBS to treat SUD, as in the present review it was the most-often used therapeutic target and experimental outcome. In one-half to two-thirds of the studies, NIBS led to craving reduction (Tables 1 and 2). However, there are a number of issues that need to be addressed regarding the assessment of craving. First, the threshold at which a desire becomes a craving was not clearly defined in the studies selected in our review. Secondly, the time frame to gauge the craving experience varies across studies. While some investigators considered craving a relatively stable experience, others viewed it as a pulsatile state with transient urges, thus raising the question of the most appropriate moment to assess it.¹⁹ According to the second perspective, the time at which craving is measured may prove to be a critical determinant of its validity, since we may miss it by infrequent measurements.¹ One way to avoid this issue is to elicit craving using experimental settings. In more than half of the studies selected for this review, participants with SUD were exposed to cues associated with drugs to elicit craving. But the natural progression of craving remains poorly understood. According to some authors, craving may be sustained for 30 minutes following exposure to cues, with a gradual decrease in intensity.⁶² It has been hypothesized that under certain conditions (for example, during an excessively long session of NIBS), craving intensity may actually diminish naturally over the course of a few minutes, and the decrease may be incorrectly attributed to the specific effects of NIBS.¹⁷ We tried to limit this bias in our review by selecting only randomized controlled trials with sham as the control group. Thirdly, although it is common in clinical research to assess levels of craving after exposure to cues, little is known about the value of this cue-reactivity to predict increases in addictive behavior or relapse.^18,63,64 Fourthly, none of these studies took into account the severity of the SUD, although a positive correlation was reported between reduced craving and the severity of dependence, in particular in nicotine dependence.⁴⁶ Fifthly, two different ways were used to assess craving in the selected articles. The first, based on single-item rating using Visual Analog Scales (VAS), was used in nearly half of the selected studies (Tables 1 and 2). Even though this method is easy to administer and score, and is suitable for frequent and repeated measurements, it does not reflect the multidimensional nature of craving. Moreover, this subjective measurement fluctuates over time and is subject to the influence of other variables.¹⁷ The second, based on multi-item self-report questionnaires, remedies this drawback by including multiple content domains, and the fact that it contains multiple items tends to boost the reliability of the scale.¹⁹ Multiple-item scales also have the advantage of being specifically suited to each type of drug. Different multi-item self-report questionnaires were used in 14 of the 23 selected studies and, in two cases, in combination with VAS.^43,49 However, although multiple-item questionnaires have the advantage of providing a craving assessment by the type of drug with better sensitivity than single-item measurements, most of them have been made over the past two decades with the dominant concept at that time. They thus represent different conceptualizations of craving.^19,65 The Obsessive-Compulsive Drinking Scale, which reflects compulsive characteristics of drinking-related thought and was used in three of the selected studies, is one such example.^44,57,66 Moreover, using multi-item questionnaires can be disadvantageous if we posit that certain items are more accurate reflections of craving than others, since the global calculated score would be lower by including items that may prove to be a weak indicator of craving.¹⁹ In addition, the time required to complete the questionnaire should also be taken into account: long self-report craving questionnaires compared with the “immediate” response in single-item methods may exacerbate or even create a craving experience in patients deprived of a drug.¹⁹ Finally, there are numerous types of self-report craving questionnaires that differ in several parameters. It is still unknown whether a reliable correlation exists between self-report single-item rating and multi-item questionnaires to measure craving and which one is the more useful to predict improvements in the addictive behavior.¹⁹ The present review shows great diversity in the instruments used to measure craving and in the use of cues to elicit craving. On the one hand, this may be a limitation of our study because we cannot objectively compile all the results in order to draw conclusions on an overall effect of NIBS in craving reduction. On the other hand, as there is no perfect way to measure craving, the broad variety of the methods used in these studies suggests that NIBS has the potential to reduce craving in many of its dimensions and components.

The outcomes measured in the selected studies were not limited to craving reduction. Among the other assessments, substance consumption was the most widely studied outcome, as were cognitive functions and decision-making processes. Indeed, several impaired decision-making patterns, especially risk taking and impulsivity, have been observed in patients with SUD, and they may represent one of the key symptoms of addiction.⁶⁷ Furthermore, converging evidence from neuropsychological, neuroimaging, and animal studies suggests that decision-making processes are supported by a complex neural network that includes the orbitofrontal cortex, the anterior cingulate cortex, the thalamus, parietal cortices, the caudate nucleus, and more particularly, the DLPFC, identified as a key element of this network.^68,69 It is uncertain whether the neural impairment of these regions identified by neuroimaging studies is a preexisting condition that led to the SUD or whether they are a consequence of repeated drug use.⁶⁸ In any event, data from the selected studies indicate that both rTMS and tDCS, by modulating cortical excitability of the DLPFC, can transiently modulate processes involved in decision making for SUD.^21,22,26

In the same way, a number of studies using behavioral, neurobiological, and imaging techniques have confirmed a strong association between impulsivity and addictive behaviors.⁷⁰ DLPFC disruption in SUD could affect self-control and behavior-monitoring processes and may lead to impulsivity, compulsivity, risk taking, and impaired self-monitoring,⁷¹ hence the idea to act on compulsive drug taking by using NIBS with the intention to reverse activity alterations in this cortical region. For clinical studies using NIBS, delay discounting can be used to assess changes in impulsivity in patients after NIBS, as observed in Sheffer et al.’s study and in other studies that could not be included in the present review.^21,72,73

Four SUDs were studied in the selected articles of the present review (tobacco, alcohol, methamphetamine, and marijuana). All of the studies obtained a concrete clinical response to the brain stimulation. This result is in line with our finding of common dysfunctions involving brain reward, motivation, and memory among the various SUDs. The shared mechanisms that underlie drug addictions involve the mesolimbic dopamine pathway, which includes dopaminergic neurons in the ventral tegmental area and their targets in the limbic forebrain, as well as several addiction-related brain areas that interact with this dopamine circuitry, such as the amygdala, hippocampus, hypothalamus, and several regions of the frontal cortex, the latter being the usual targets for NIBS.⁷⁴ However, even though we can argue that as all drugs involved in abuse implicate similar neural mechanisms during the addiction process, and as a consequence we can expect brain stimulation to be effective in various drug addictions, we must recognize that the results obtained in NIBS studies in addictions are unclear. Indeed, therapeutic results were obtained for both rTMS and tDCS with different stimulation parameters, and sometimes these results appear contradictory (i.e., reduction in symptoms with an increase or a decrease in the excitability of the cortical target). Moreover, some results are nonreproducible from one study to another even though the stimulation parameters are similar, suggesting that other factors should be considered when interpreting findings. Among these parameters, we can evoke the accuracy of the targeting methods used, as well as intensity levels of the stimulation and, a parameter that is often ignored in NIBS, the scalp to cortex distance, which may affect the efficacy of brain stimulations.^12,75,76 In addition, although the most common target for therapeutic NIBS is the DLPFC, other cortical areas, such as the dorsal anterior cingulate cortex and the insula, may be of interest for SUD and potentially accessible to NIBS. Specific cell types of the cortico-striatal circuitry could also be targeted in future research.^77,78

Independently of the stimulation parameters and the methods used to assess the outcomes, stages of addiction also deserve to be considered in NIBS studies. Indeed, the development of addiction can be perceived as a succession of neuroadaptive changes in the brain: it begins with changes in the mesolimbic dopamine system and continues with a cascade of neuroadaptations from the ventral striatum to the dorsal striatum and orbitofrontal cortex and eventually dysregulation of the prefrontal cortex, cingulate gyrus, and extended amygdala.⁷⁹ As the neurobiology varies during the addiction process, the impact of NIBS on the brain may also vary over time, depending on the underlying brain changes at the moment of the treatment. We can suggest this hypothesis to explain the various results with NIBS in addiction.

Some caution is warranted in interpreting the results of this review. First, we limited our review by selecting only randomized controlled trials in order to have the highest level of evidence. However, numerous studies with relevant results in subjects have been done, even though the level of evidence was lower. Secondly, a large majority of the selected studies were preliminary, with small sample sizes, without testing their sham condition and without follow-up of the patients, and there was considerable heterogeneity in terms of sample population, study design, TMS or tDCS parameters, and outcome measurements. This field is still recent, and most studies delivered a small number of NIBS sessions, since 12 of the 21 selected studies estimated the efficacy of NIBS from only one or two stimulation sessions with both rTMS and tDCS, which encouragingly showed no serious adverse events. Future studies should now include a greater number of NIBS sessions in order to induce greater and more sustained changes in brain activity to reduce craving and consumption and to improve cognitive functions. Thirdly, although NIBS has been studied in different SUDs, not all types of addiction were taken into account.