A narrative review format was chosen for this review to allow for a broad assessment of the research space and analysis of available literature. In the initial stages of the review, the research question was intentionally broad to include sources on neurostimulation in SUD treatment more broadly, to ascertain the currently available literature on what was anticipated to be a sparse area of research. This was achieved using an array of search terms such as “substance use disorder*”, “cocaine use disorder”, “methamphetamine use disorder”, “stimulant use disorder”, “transcranial magnetic stimulation”, “transcranial direct current stimulation”, “craving*”, “abstinen*”, “withdrawal symptoms”. A comprehensive summary of all terms used is shown in Appendix A. The search was conducted across three databases, PubMed, OVID Medline and PsycINFO, which were chosen due to their applicability to our field of research. The publication date range was set to 2003-present, as 2003 marked the initial proposal of neurostimulation for addiction management. All sources were then uploaded into Covidence for screening purposes where screening was conducted by a single reviewer (JH). Whilst a single reviewer (JH) completed the screening process, frequent discussions took place between this reviewer and another author (RML) when any uncertainty regarding a study’s inclusion or exclusion arose to ensure all relevant literature was assessed as part of this review.

After reviewing the broader research output, we adjusted our inclusion criteria to limit our review to resources exploring non-invasive neurostimulation modalities in managing StUDs, given the limited current management options compared to other SUDs. Sources were included if they assessed non-invasive neurostimulation modalities in the management of StUDs in human participants, were a peer-reviewed systematic review, literature review, randomized-controlled trial, observational study, or case report, and assessed clinical outcomes relevant to StUDs.

Given the clinical interest of this review in noninvasive neurostimulation modalities for StUDs, sources were excluded if their primary focus was on an addictive substance other than cocaine or methamphetamines, were based upon animal studies, explored invasive neurostimulation techniques such as Deep Brain Stimulation or vagus nerve stimulation. Book chapters, letters to the editor, and commentaries were excluded as they were deemed not rigorous enough.

The same single reviewer conducted the screening and extracted and analysed evidence from included studies with assistance from a co-author (HB) to assess trends and overall findings. No quality assessments were performed so as not to hinder scoping assessment in order to gain a broader overview of an evolving field where there is limited consensus on standardized protocols, targets and impact. As such, this allowed the review to better demonstrate how the field has been evolving over the relatively short period of time. Therefore, no studies were excluded due to low quality.

The initial search returned 1,317 studies after being processed through Covidence, and manual screening removed duplicates. Using the above inclusion and exclusion criteria, 179 studies were deemed suitable for inclusion at the title and abstract screening phase, and 90 were subsequently included for extraction, as shown in the PRISMA flowchart below (Figure 1).

Across the 90 studies, 57 were primary studies with a total of 3,319 participants, and 33 were reviews. Primary studies are summarised in Tables 1–4, with a more extensive presentation of findings recorded in Supplementary Tables 1–4. Review studies are summarised in Supplementary Tables 1–4.

55 of the 57 primary studies in this review assessed clinical outcomes. 24 of these enrolled a CUD sample, and 31 a MUD sample. 35 employed a sham condition. Given the importance of an adequate control condition to assess change in clinical outcomes from interventions, results were organized according to studies with and without a sham condition. The most assessed clinical outcome was craving, although other assessed outcomes included patterns of use, withdrawal symptoms, affective symptoms, cognition, and sleep. Tables 1.1, 1.2 provide a comprehensive summary of the studies. Further details on the outcome scales and scores used are included in Supplementary Table 1.

Three primary studies compared different neurostimulation modalities in relation to clinical outcomes (11, 61, 62) (Table 2). Two compared iTBS to rTMS, with neither finding significant differences overall between the two modalities (11, 62), however one found rTMS to significantly improve symptoms of anxiety and depression whilst iTBS did not (11). Neither of the aforementioned studies had a sham condition. The final study explored iTBS in comparison to cTBS, and although no significant difference was noted in clinical outcomes, adverse effects were significantly higher in the iTBS group compared to other groups, with the main issues being tingling and sleepiness (61).

Three reviews assessed for any differences between neurostimulation modalities (6, 95, 96). The greatest improvement in craving scores was observed in participants treated with a combination of iTBS of left DLPFC and cTBS of left vmPFC (6).

Nine studies compared target sites, as seen in Table 3. Most of these compared variations of DLPFC targeting, whether left to right, or various electrode montages over this region. Only a single study reported the right DLPFC to be significantly superior (59), with the most recent of the studies finding both left and right DLPFC high-definition tDCS groups to show similar significant reductions in craving compared to sham tDCS, lasting up to one month (37, 58). In addition to these findings, one study reported targeting the left DLPFC but not the right DLPFC, produced a significant improvement in anxiety, whilst targeting either the left or right DLPFC significantly improved depression (61).

In one study comparing various electrode montages, two of the montages were found to be significantly more effective in improving specific cognitive outcome measures (16). Finally, in a study that assessed vmPFC as an additional target site, a greater significant effect on depressive and withdrawal symptoms, as well as sleep, was observed in participants who received neurostimulation targeting both the DLPFC and the vmPFC compared to either site alone (38).

Three reviews comparing target sites were assessed, both reporting the DLPFC as the primary target and finding it effective for reducing cravings in StUDs (97–99).

17 primary studies and one review assessed the effects of neurostimulation treatment, accompanied by neuroimaging. A summary of results can be found in Tables 4.1-4.2. It should be noted that the following neuroimaging findings are promising but preliminary, and that casual inferences should be avoided at this stage.

A relatively even split of studies with MUD and CUD patient samples were included in this review. Cravings were the most frequently assessed outcome, which is consistent with previously published reviews (90). Two-thirds (66%) of the studies measuring cravings found active neurostimulation therapy to produce significantly better reductions in craving scores than comparator groups, or to significantly reduce baseline levels of cravings. Contrastingly, two studies reported an increase in at least one measure of craving in active groups compared to sham groups (36, 48). Furthermore, as mentioned in the reporting of results, a number of studies found non-significant differences between active and sham groups. However, despite this, craving reduction remains the most reproducible outcome across studies, with most of this evidence coming from sham-controlled trials. In interpreting these results and their reliability it must be noted that all except one study used subjective measures of cravings (37), highlighting a gap in objective craving data, such as heart rate. Findings across other clinical outcomes including use patterns affective symptoms, and cognition are more variable and are frequently reported by methodologically weaker studies.

In terms of follow-up, five studies completed a one-month period (24, 26, 28, 37, 55), three a three-month period (20, 29, 66), two a six-month period (25, 56) and finally a single study followed-up participants for a total of 12-months which was the longest period observed across studies (67). Their outcome of interest was abstinence (67). Reductions in cravings were observed to last up for a maximum of three months after treatment across all reviewed literature (25). This poses the question of how long clinical improvements actually last for following an acute phase of neurostimulation treatment, something that has already been asked and dicsused in previous reviews (89).

This leads into the important real-world application of the role that maintenance neurostimulation sessions play in long term effectiveness of treatment. Although a handful of studies employed maintenance treatment regimes, the volume of data on their efficacy remains extremely limited, and further consensus is required to accurately determine how these can be best optimized. The only study that explored the indications for maintenance therapy in depth was a case report with a sample size of one, reducing its generalisability to broader target populations (56).

Overall results appear to show promise in neurostimulation managing affective symptoms, although the variation in significance of improvements, and scales used, must be acknowledged. Out of all the clinical outcomes assessed in this review, affective symptoms were the most commonly improved in both active and sham groups, potentionally revealing a greater susceptibility to placebo effect (25–27, 34, 57). The most frequently assessed affective outcomes were depression and anxiety. In addition to neurostimulation significantly improving affective symptoms, one of the studies also found this improvement to be even greater compared to sham if participants received more than 40 neurostimulation sessions and/or were taking psychotropics concurrently (22). A similar relationship was found by a study assessing sleep quality improvement, whereby participants who had completed a greater number of neurostimulation sessions in the days preceding the completion of sleep scale scores, experienced greater benefit (52). These correlations could be worth further exploration given it’s clinical application for more holistic and comprehensive regimes.

Cognition was variably affected by neurostimulation therapy, with just over half the studies assessing it finding significant improvements. However, cognitive outcomes were measured with several different tasks, which assessed varied components of cognition, making direct comparisons and conclusions difficult. Although this variation means specific findings are less directly replicable across current literature, it does allow the broad impact of neurostimulation on multiple facets of cognition to be seen. In studies that found significant cognitive improvements in active groups compared to sham, effects were seen to last until as long as one-month follow-up (24, 28). Only a single previously published systematic review included in this narrative review conducted a meta-analysis for cognitive outcomes (72). The effect of neurostimulation on various cognitive tasks assessing episodic and working memory, problem solving and visuospatial learning, amongst other domains, was analysed (72). Significant improvements in the active group participants compared to sham were noted in all but two domains, working memory (P = 0.08) and emotion recognition (p=0.28) (72). Of the domains that were significantly improved by neurostimulation intervention, the degree of overall effect was highly variable, episodic memory (p=0.0007); problem solving, reasoning and visuospatial learning and memory (p=0.001) and visuospatial learning and associations (p=0.02) (72). Although more frequently reaching significance, the variability in the meta-analysis’ findings is in line with the varied effects uncovered in this review. These more convincing meta-analysis findings may also be explained by the stricter inclusion criteria leading to the analysis of only five studies, all of which had to be sham-controlled trials to be eligible for inclusion. Ultimately, it is a reassuring finding that meta-analysis of methodologically strong studies supports neurostimulation in improving cognitive outcomes in MUD patients.

Whilst the above points support potential clinical applicability of neurostimulation in StUD management, it is important to recall that a handful of studies saw clinical improvements in at least one outcome in both intervention and control group participants (17, 22, 25–27, 33, 34, 38, 53, 57), suggesting a degree of placebo effect within this space. Furthermore, only nine studies assessed neurostimulation with current addiction treatment (20, 21, 24, 25, 46, 53, 55, 57, 66), despite this being the most likely way in which the technology would be introduced into the clinical space, bringing into question the extent of generalisability, and the need for future studies assessing combination treatment. Of the studies that did assess both neurostimulation and standard treatment, most found that neurostimulation improves outcomes, with only one study reporting this improvement to be non-significant (66). Although small in number, these studies provide a good starting point and insight into how using neurostimulation as an adjunct to standard therapy might apply to real-world patient management. Furthermore, only a single study explicitly explored the impact of having a comorbid psychiatric diagnosis of ADHD on the effectiveness of neurostimulation treatment in addiction (51). Although this study found no significant differences in the improvements post neurostimulation between participants with ADHD and CUD or CUD alone, there is insufficient evidence to predict whether this is the case with other psychiatric diagnoses at this time.

In general, a trend of good tolerability was observed across all of the studies that commented on this, with mostly only minor side effects reported, tingling of the scalp being the most frequent. This augurs well for the safety of implementing such therapy in the clinical space.

Only a single review study found one neurostimulation type to be superior to another (76), although others highlighted important differences between types and their parameters, which may be of use clinically. iTBS was as effective as traditional rTMS (11, 62), which is important given the practicality of its shorter session times. This may make treatment easier to adhere to, especially given that compliance can be challenging in target populations. Comparisons between high and low frequency have yielded mixed results, with one study reporting no difference in outcomes (58), whilst a review declared high-frequency to be superior (6). Furthermore, multiple treatment sessions produced superior clinical outcomes compared to single treatment sessions (75, 91). Aside from these few direct comparisons between parameters, stimulation protocols were vastly heterogeneous across the studies, making determining which factors produce the most efficacious treatment regime almost impossible at this stage.

The most established target site is the DLPFC, with most studies finding it an effective target. This can be explained by the known role of the PFC in the pathophysiology of addiction, with changes in its activity as demonstrated by studies employing fMRI, being associated with improved executive functioning and reductions in cravings. The vmPFC has also been addressed as a target site, with its role in emotional regulation and decision-making, justifying this. One of the reviewed studies found that targeting both the DLPFC and the vmPFC was superior in terms of clinical outcomes compared to targeting either site alone (25, 38). This may suggest that having multiple target sites could improve the clinical efficacy of neurostimulation in StUDs, although evidence remains sparse.

Imaging studies that used MRI and/or EEG have assisted in hypothesising and understanding the underlying mechanisms of neurostimulation and the brain regions likely to be most efficacious when targeted, however, they are still relatively small in number when it comes to StUDs and present some conflicting results.

What appears to be very sparse are more studies that frequently compare brain activity in patients to that in healthy controls, although what we have observed so far suggests neurostimulation may be able to normalise patients towards this baseline (45, 63). A fascinating, novel concept identified in this review is individualised montages, and this approach, although still in its early development, has shown promise in not only clinical results but also in predicting response to treatment on a per-patient basis (100). If this develops, it would likely be useful for clinicians in selecting which patients they wish to manage in this way.

EEG studies primarily focused on P3, theta and beta activity, with mixed results, making definitive conclusions difficult to draw. Despite this, all EEG studies have found that neurostimulation alters activity in some way.

The strengths of this review include the number and variety of studies included. It is one of the most extensive reviews on this topic, certainly in StUDs specifically. This meant we were able to assess multiple relevant outcomes and variables when it comes to neurostimulation in the addiction medicine space. However, this review has limitations, including the use of a single reviewer for inclusion and exclusion criteria, and the absence of quality or bias assessments or meta-analyses. As such, it is recommended that readers weigh the results obtained from sham-controlled study designs more heavily than those with less rigorous methods. Furthermore, although our scope adjustment allowed us to assess the literature in the StUDs space more comprehensively, it does mean we could not evaluate the literature available for other SUDs as initially intended.

In terms of the included studies, a growing number of randomised controlled trials were identified, as well as those employing a sham condition, which increases the usefulness of this review. However, only just over half of the studies reporting on clinical outcomes employed a sham, and a handful of other primary studies still failed to even employ a control group, comparing only to baseline levels. Many sample sizes were also relatively small, with most employing under one hundred participants, with more male than female representation. These small sample sizes have been predicted to be the cause behind results not reaching significance, with one study stating significance would be reached should the study have had more participants (26). It was noted that female representation in samples is increasing in more recent publications, however the discrepancy in male to female representation could be a potential hinderance to the generalizability of presently obtained results to female patients. One of the reviews also commented on the possibility of publication bias in the literature thus far (72).

One of the main gaps identified in the literature that would be a valuable focus of future research is the limited generalisability of study samples to the real-world target population. None of the included studies clearly aimed to assessed neurostimulation in patients with diagnosed comorbid affective disorders, which commonly exist in patients with SUDs. Only a single study evaluated outcomes in patients with comorbid CUD and ADHD (51). Furthermore, none of the studies had a sample with polysubstance use diagnoses; the only exception to this was a single study that commented on how the use of other substances was also reduced in their participants who underwent active iTBS for CUD (65).

Additional gaps included a lack of extended follow-up periods, with few studies following participants past one to three months. Given the relapsing and remitting nature of addiction disorders, this is a considerable limitation to the current literature and how it can be applied to a clinical setting. Further research is needed on the duration of results following initial treatment and the role that maintenance therapy should play in the ongoing management of these patients. Ideally, establishing an optimal range of number and frequency of sessions would be highly valuble. As briefly discussed above, the heterogeneity of stimulation parameters was extensive across the reviewed literature (7, 74, 76, 80, 84, 88). As such, more studies that aim to compare various parameters such as frequency, duration and number of sessions may be valuable in eventually constructing future clinical guidelines.

A final area of recommendation would be to further explore the personalised medicine approach and investigate possible predictors of treatment response so that clinicians can better determine patient suitability for this treatment. This area is particularly relevant as medicine moves towards a more individualised patient care model. Data sharing initiatives for non-invasive brain stimulation techniques, such as Big NIBS data (103–105), may help to address field-wide issues associated with low sample sizes and poor statistical power to detect significant effects, as was evidence in some studies included in this review.

Other important considerations include funding, cost, ethics, and the availability of this treatment, especially considering the social complexities and vulnerability of many patients in addiction medicine clinical practice. Future exploration into the logistics of delivering neurostimulation treatment to vulnerable patients would be of value, including equity of access.

To conclude, this review is a comprehensive overview of the currently available literature on non-invasive neurostimulation in the treatment of StUDs. Whilst neurostimulation is a promising upcoming treatment modality in StUD management, with current results generally positive, at present it should be considered experimental or adjunctive, rather than a standalone evidence-based treatment. We also recognise the variation and occasional contradictions in the research findings and suggest ongoing investigation into these limitations moving forward.