Efficacy of Invasive and Non-Invasive Brain Modulation Interventions for Addiction

Addiction is a burdensome public health concern that not only affects the addicted individual but also greatly impacts his or her family members and surroundings. Though several treatment options are available, many patients show a chronic course of addiction with abstinence rates between 40% and 60% 1-year post treatment (Mclellan et al. 2000). In the last few decades there is more awareness for the problem of nonsubstance related addictions such as gambling disorder (reclassified as an addictive disorder in 5th edition of the diagnostic and statistical manual of mental disorders) and excessive internet use and gaming, increasing the importance of finding new and better interventions for addiction (Banz et al. 2016).

In recent years our neurobiological knowledge of addiction has expanded. We know more about changes in brain structure, function and its neurochemistry (Fineberg et al. 2010; Koob 2015; Koob and Volkow 2009; Volkow et al. 2012). Yet, the translation of this knowledge into new interventions or improvement of existing interventions has fallen short. The non-selective manner in which effective psychotherapeutic and pharmacological interventions induce brain changes makes it difficult to apply neurobiological knowledge to improve these interventions. Neuromodulation may close the gap by targeting directly abnormalities in brain functioning. Direct intervention in the brain in order to change addiction is not new, on the contrary, neurosurgery for addiction has a long and turbulent history (Stelten et al. 2008). Yet, the advances in our understanding of brain function and the developments of new neuromodulation techniques that are non-invasive or reversible nowadays have made this intervention method a safer and more promising option.

Repetitive transcranial magnetic stimulation (rTMS) and transcranial direct current stimulation (tDCS) have been used most frequently for treatment of addiction. Other neuromodulatory options for addiction are deep brain stimulation (DBS), and neurofeedback which can be applied with Electroencephalography (EEG neurofeedback) or functional magnetic resonance imaging (fMRI neurofeedback). These techniques have been investigated as stand-alone or add on treatment and while some have been around for decades (EEG neurofeedback), others have just emerged as promising intervention for addiction (fMRI neurofeedback). In this review we will evaluate, compare and discuss the experimental application of these neuromodulation techniques for treatment of addiction.

A complicating factor in the application of neuromodulation for addiction is its heterogeneity, with large individual differences in underlying brain abnormalities (George and Koob 2010). Therefore, we choose to not only review how neuromodulation affects primary treatment outcomes for addiction such as craving or relapse, but additionally focus on how neuromodulation affects different cognitive processes involved in addiction. Cognitive processes such as reward processing, cognitive control and emotion regulation may constitute different factors underlying addiction with a closer mapping to neural abnormalities than the currently classified symptoms of addiction. One could envision that using this approach, an individually tailored type of neuromodulation that fits the cognitive profile of the patient would become possible. For instance, a highly impulsive addicted patient may receive neuromodulation focused on enhancing cognitive control in prefrontal regions whereas deeper regions like the insula or amygdala might be targeted in a patient whose addiction is driven by harm avoidance. Although the number of studies investigating the effect of neuromodulation on these cognitive processes in addiction is very limited, we will discuss them where possible.

This review aims to discuss the efficacy of neuromodulation techniques for the treatment of addiction. Although still in an early phase of development, we additionally review neuromodulatory induced cognitive changes relevant for addiction. As addictions are perpetuated by numerous maladaptive cognitive processes that can be influenced with neuromodulation techniques, this is a highly promising area for future research and treatment development. This review extends upon existing reviews addressing rTMS, tDCS (Hone-Blanchet et al. 2015; Jansen et al. 2013; Lupi et al. 2017) with additional DBS for addiction (Coles et al. 2018; Spagnolo and Goldman 2017) by (1) specifically considering only the highest quality evidence for each technique in order to evaluate the potential for clinical efficacy, (2) expanding the definition of neuromodulation to consider fMRI and EEG neurofeedback, and (3) reviewing neuromodulatory impact on addiction-related cognitive processes and discussing potential future neurocognitive targets.

Our review focusses on studies that included adult participants who were treated for substance or behavioral addictions with neuromodulation (i.e., rTMS, tDCS, DBS, EEG & fMRI neurofeedback). With this aim we searched MEDLINE databases using PubMed for the indication terms: (addiction OR substance disorder OR craving OR gambling disorder OR binge eating disorder OR bulimia OR smoking) AND (transcranial magnetic stimulation OR repetitive transcranial magnetic stimulation OR transcranial direct current stimulation OR tDCS OR neurofeedback OR deep brain stimulation). The reference lists of relevant papers were inspected for further studies that fit the inclusion criteria. The search was conducted on the literature before August 2017 without historic limitation, and only studies written in English were included. As there are large differences in number, quality and methodologies of studies between the neuromodulation techniques we used specific inclusion criteria to identify the highest quality of available evidence for each technique where we prioritize (1) sham-controlled designs (2) outcome measures focused on clinical efficacy (3) number of sessions for non-chronic neuromodulation techniques (4) and number of participants. Because of the large differences in quality and quantity of studies between the modulation techniques, we will outline which of the criteria will be applied at the beginning of each section below.

At the moment, little evidence supports the hope that neuromodulation will soon become a new effective intervention for addiction. This may be due to a lack of well-powered systematic and controlled studies. In addition, the reviewed studies indicate that it may too simplistic to assume that neuromodulation targets specific brain regions. Non-invasive neurostimulation and EEG neurofeedback are targeted at broader brain regions or networks, as well as DBS, where specific regions are stimulated and distant brain regions are critically involved in the clinical effects. Moreover, due to the heterogeneity of addiction, neurobiological knowledge may not be easily translated into treatment targets. In sum, it is important to realize the complexity of both the neurobiology of addiction and the effects of neuromodulation when applying neuromodulation. And one way forward could be to use study designs that integrate and test more complex models of neuromodulation.

Because of the different number and quality of studies available for each neuromodulation technique we used separate inclusion criteria. This means that we cannot directly compare the efficacy of these studies and therefore we will discuss them separately with respect to the stage of development they are in. The studies with DBS and fMRI neurofeedback are either pilot or proof-of-principle studies that do not warrant any conclusions about the efficacy as addiction interventions. This is not surprising since these interventions as indication for addiction are still new, require a high degree of expertise and are expensive. As Table 4 shows, the fMRI neurofeedback studies are all published in the last five years and have developed over time in terms of quality and samples size, therefore it is to be expected that larger controlled trials investigating the effect on drug use will follow. With DBS, difficulties with the inclusion of patients have hampered larger controlled DBS studies, while such trials have been published for the treatment of depression and obsessive-compulsive disorder. It is unknown whether future developments in the field will overcome these difficulties.

With rTMS and tDCS, most studies have focused on immediate effects of craving. Yet there are a few studies that looked at (immediate and longer-term) effects on drug use, with promising results indicating the potential of these techniques for treating addiction. In this regard, the current state of research is not dissimilar to the early years of experimental rTMS for major depression, which eventually has become an effective treatment. Whether rTMS will ultimately be an effective treatment for addiction as well, awaits larger trials delivering longer treatment courses, incorporating objective and long-term outcome assessments, and systematic investigation of stimulation protocols.

On the other hand, EEG neurofeedback has been investigated for decades and although one of the first studies included a sham feedback control condition (Jones and Holmes 1976), very few studies afterwards have used an appropriate control with either sham-feedback or where at least the time spend on the neurofeedback sessions was matched with another activity. This makes it difficult to draw conclusions about the specific effect of EEG neurofeedback, since any effects might be the result of the extra time and care patients received for the neurofeedback sessions. In one study that did include such a control condition, the control condition was not used as comparison in the final analysis (Schmidt and Martin 2016). Fortunately, there are two exceptions with control condition (time-matched activity) (Keith et al. 2015; Scott et al. 2005) suggesting that the combination of SMR and alpha-theta neurofeedback can have a positive effect on impulsivity, and on the period of time spend in rehabilitation. However, after decades of investigating EEG neurofeedback the lack of well-designed studies is worrisome and raises the question whether this reflects the lack of potential of EEG neurofeedback to treat addiction.

Until now all fMRI neurofeedback studies and most rTMS/tDCS studies have focused on short-term outcome assessments. Yet, in order to apply these neuromodulation techniques as intervention for addiction a crucial question is whether it is possible to bring about lasting changes in well-established addictive behavior. To answer this question, we should not only assess long-term outcomes but also investigate the number of treatment sessions required for these lasting behavioral changes. In most studies short treatment courses were used: fMRI neurofeedback studies (max 3), tDCS (max 10), rTMS (max 16), only in EEG neurofeedback four studies used 30 sessions or more. This might have contributed to the lack of long-term effects found by several studies: Da Silva et al. (2013) with 5 sessions, Amiaz et al. (2009) with 16 sessions, Trojak et al. (2015) with 10 sessions and Lackner (2016) with 12 sessions. It is likely that in order to change neural programs and their associated behavior, extensive practice is needed using more sessions. Therefore, future research in this area should address the dose of neuromodulation requires to achieve persistent long-term reduction in addiction severity.

Moreover, the second main concern of the reviewed studies is that many of them are underpowered. Two fMRI neurofeedback studies (Canterberry et al. 2013; Kim et al. 2015), five rTMS studies (Bolloni et al. 2016; Ceccanti et al. 2015; Dinur-Klein et al. 2014; Höppner et al. 2011; Walpoth et al. 2008) and three tDCS studies (Conti and Nakamura-Palacios 2014; da Silva et al. 2013; Fecteau et al. 2014) used less than 10 participants per condition. Four DBS only case reports or series have been published. This warrants caution with the interpretation of these results, which have a higher change of type II error and inflated effect sizes. However, in all fields except for EEG neurofeedback, increasing the quality of studies for our review by excluding those with less than 10 participants per condition would not have weighed up against the low number of studies left with (0 dB, 6 fMRI neurofeedback, 4 tDCS and 5 TMS). Fortunately in recent years sample sizes are increasing (e.g., Gay et al. 2016; Hartwell et al. 2016; Ljubisavljevic et al. 2016), and in 2017 the largest tDCS trial in addiction was published. In this study, 91 patients were randomized into three groups with the active intervention group receiving a single session tDCS to boost a cognitive bias modification training. No effects of tDCS were found on craving, approach bias or abstinence after three months, but there was a trend towards lower relapse in the tDCS condition after one year (den Uyl et al. 2017). In sum, while there is still a lack of well-powered studies the field seems to be progressing in the right direction and future studies should aim for larger samples sizes.

Neurofeedback, rTMS and tDCS have very few side effects, and are transient; therefore, they are considered safe interventions (Priori et al. 2009). DBS is invasive and side effects such as transient hypomanic episodes have been reported. Additionally, there is a small risk of hemorrhage, infection or headaches. Yet, studies of DBS targeted at the ventral striatal area for other psychiatric disorders have proven to be safe and modulate neurocognitive dysfunctions that are also relevant for addiction (Luigjes et al. 2013). For neuromodulation and non-invasive neurostimulation, future research will have to investigate the number of sessions needed for an optimal effect, and whether patients will benefit from booster sessions after treatment. DBS gives chronic stimulation; however, one case report in a patient treated for opioid addiction showed that the patient remained abstinent even with cessation of stimulation (Zhou et al. 2011).

One question that remains to be investigated is whether neuromodulation techniques can function as stand-alone treatments or whether they are more effective as an add-on treatment. In the majority of EEG neurofeedback studies, the treatment is provided as add-on to either pharmacological treatment or psychotherapy with the rationale that EEG neurofeedback may augment the effects of treatment as usual. Moreover, two tDCS studies used concurrent therapy (de Silva et al. 2003; Klauss et al. 2014) and one rTMS study (Trojak et al. 2015) used additional nicotine replacement treatment. Kuhn et al. (2014) proposed that additional behavioral therapy may be helpful for patients to learn to cope with stress and boredom after DBS treatment which are often triggers for relapse. The combination of DBS and cognitive behavioral therapy has been successful in the treatment of obsessive-compulsive disorder (Mantione et al. 2014). Since addiction is a disorder that affects many aspects of a patient’s life it is likely that additional cognitive behavioral treatment or psychosocial support can play an important role in the success of neuromodulation interventions especially for the most severely affected patients. Alternatively, neuromodulation could enhance the success of add-on treatment. rTMS and DBS have substantial efficacy in mood disorders (Kisely et al. 2018; Perera et al. 2016), and beneficial effects on mood in addiction could enhance motivation for change and ability to engage in complementary therapies. Moreover, combining neuromodulation with other treatments may result in beneficial synergistic effects at cognitive and neural levels. For example, combining neuromodulation and cognitive training has been shown to be beneficial for major depression and Alzheimer’s disease, when the training involves exercising a cognitive skill required for recovery. For patients with addictions, using neuromodulation techniques to induce plasticity whilst exercising key cognitive skill such as response inhibition, extinction learning, or cognitive re-framing, may have greater potential to modify neural networks and cognitive processes driving addictions, and potentially induce greater long-term effects (Rabey et al. 2013; Segrave et al. 2014).

A way forward in the application of neuromodulation for addiction might be to tailor these interventions to individual patients or subgroups of patients based on cognitive or neural profiles. First, this can be done by targeting specific cognitive candidates that are likely to map more closely to neural abnormalities than general addiction outcomes as craving or drug use. For instance, an overreliance on negative-reinforcement may indicate modulation of anti-reward brain network, whereas high impulsive patients may benefit more from targeting regions involved in cognitive control. Moreover, profiling based on neural functioning might be another way to investigate for which patients’ specific types of neuromodulation are most effective. Although few studies in the addiction field investigated or found effects on specific cognitive processes and none used profiling, studies in other fields show reason to be optimistic about this approach. A study that focused on rTMS for depression found that based on MRI resting state scans depressed patients can be divided in four subtypes related to specific symptom profiles that predict the response on rTMS treatment (Drysdale et al. 2016). Additionally, for neurofeedback approaches there are indications that certain participants are better able to modulate their brain activity. Better prediction of which patients will be successful can lead to better targeted neurofeedback treatment which can be especially important for the more expensive fMRI neurofeedback. A possible cheap method of predicting patients’ success is selecting patients that are able to modulate their temperature with biofeedback (Hartwell et al. 2013)

Moreover, there are encouraging developments in other psychiatric indications for neuromodulation that show how the interaction between neurobiological science and the application of neuromodulation can mutually advance each other. For instance, a German research group found a more optimal new target for the treatment of DBS in depression, the medial forebrain bundle, based on translational neurobiological research (Bewernick et al. 2017; Coenen et al. 2011). A different group found that scans informing about structural brain connectivity could help predict the optimal DBS target for individual patients with depression (Riva-Posse et al. 2018). These studies in other psychiatric disorders show a possible way forward to improve future application of neuromodulation in addiction treatment.