A systematic review and meta-analysis of neuromodulation therapies for substance use disorders

While pharmacological, behavioral and psychosocial treatments are available for substance use disorders (SUDs), they are not always effective or well-tolerated. Neuromodulation (NM) methods, including repetitive transcranial magnetic stimulation (rTMS), transcranial direct current stimulation (tDCS) and deep brain stimulation (DBS) may address SUDs by targeting addiction neurocircuitry. We evaluated the efficacy of NM to improve behavioral outcomes in SUDs. A systematic literature search was performed on MEDLINE, PsychINFO, and PubMed databases and a list of search terms for four key concepts (SUD, rTMS, tDCS, DBS) was applied. Ninety-four studies were identified that examined the effects of rTMS, tDCS, and DBS on substance use outcomes (e.g., craving, consumption, and relapse) amongst individuals with SUDs including alcohol, tobacco, cannabis, stimulants, and opioids. Meta-analyses were performed for alcohol and tobacco studies using rTMS and tDCS. We found that rTMS reduced substance use and craving, as indicated by medium to large effect sizes (Hedge’s g > 0.5). Results were most encouraging when multiple stimulation sessions were applied, and the left dorsolateral prefrontal cortex (DLPFC) was targeted. tDCS also produced medium effect sizes for drug use and craving, though they were highly variable and less robust than rTMS; right anodal DLPFC stimulation appeared to be most efficacious. DBS studies were typically small, uncontrolled studies, but showed promise in reducing misuse of multiple substances. NM may be promising for the treatment of SUDs. Future studies should determine underlying neural mechanisms of NM, and further evaluate extended treatment durations, accelerated administration protocols and long-term outcomes with biochemical verification of substance use.

Reinforcing effects of substances are primarily mediated by mesocorticolimbic systems, which include midbrain dopamine (DA) projections to prefrontal cortex (PFC) and ventral striatum [nucleus accumbens (NAc)] [10,11].Substance misuse is associated with mesolimbic hypodopaminergia [12], and dysfunction of dorsolateral prefrontal cortex (DLPFC) and dorsal anterior cingulate cortex (dACC), which are involved in decision-making and self-control.Moreover, the ventral PFC, including the orbitofrontal cortex (OFC) and ventral anterior cingulate cortex (vACC), is involved in limbic arousal and emotional processing [13].Hence, dysfunction in these systems has been associated with SUDs [14].Furthermore, left DLPFC mediates reward-based motivation, while right DLPFC is involved in withdrawal-related behaviors and inhibition [15].Thus, use of NM to stimulate right DLPFC may strengthen executive functions by inhibiting the left DLPFC to counterbalance hemispheric imbalance, which may contribute to reduction of substance consumption and craving [16,17].Invasive and/or non-invasive NM may be promising brainbased approaches since they modulate SUD-related mesolimbocortical circuitry [8,9,18].Such interventions include repetitive transcranial magnetic stimulation (rTMS), transcranial direct current stimulation (tDCS), and deep brain stimulation (DBS).
Repetitive transcranial magnetic stimulation (rTMS) rTMS is a non-invasive NM technique that has shown utility for neurological and psychiatric disorders [19].Application of alternating magnetic fields to the scalp through a copper wire induces temporary electrical currents and modulates cortical excitability in localized brain tissue [20] (Fig. 1a).Numerous studies have demonstrated enduring functional and structural neuroplastic changes in target regions [21,22], and increased DA release in the mesolimbic system [23][24][25][26].
Stimulation parameters vary significantly with respect to stimulus intensity, frequency and total number of pulses, which can produce differential effects [27].Typically, low frequency (LF; ≤1 Hz) stimulation produces local inhibitory effects while high frequency (HF; ≥5 Hz) stimulation produces local excitatory effects on motor cortex [28,29].rTMS primarily alters motor cortical excitability and inhibition, with indirect effects on craving or motivation.Frequency-dependent rTMS effects on regional brain activity may have implications for clinical therapeutics in neuropsychiatric disorders [30,31].Coil type can also modulate effects; while traditional TMS employs a figure-8 coil design and can only reach depths of 0.7 cm, deep TMS, wherein a threedimensional H-coil helmet design is used, can stimulate a deeper and broader brain area, reaching a depth of 3.2 cm [32].
Two robust rTMS adaptations have emerged wherein bursts of magnetic pulses, referred to as theta burst stimulation (TBS), are applied.In intermittent theta burst stimulation (iTBS), a two second train of TBS bursts is repeated every ten seconds, inducing long-term potentiation and cortical excitability [33,34].Contrastingly, continuous theta-burst stimulation (cTBS) applies trains of uninterrupted TBS bursts and induces long-term depression and inhibitory effects [34].rTMS appears safe when administered according to recommended guidelines [35].There is little risk beyond local discomfort at the site of stimulation and other minor side effects (e.g.mild headache, dizziness) [36].Importantly, a deep-TMS system was recently cleared by the Food and Drug Administration (FDA) for smoking cessation [37].However, long-term effects of repeated rTMS sessions are unknown [38].

Transcranial direct current stimulation (tDCS)
Using two or more electrodes (i.e., anodal, cathodal), tDCS delivers a low intensity current (0.5-2.0 milliamps [mA]) to a targeted brain region for several minutes (Fig. 1b).This allows for polaritydependent modulation of the neuronal resting membrane potential and cortical excitability.Cathodal current decreases while anodal current increases cortical excitability [39,40].Similar to rTMS, tDCS protocols can vary with respect to numerous parameters such as current strength, electrode size and placement, stimulation duration and frequency [41].
tDCS is an accessible, low-cost stimulation method that is welltolerated, though minor side effects such as scalp irritation are reported [42].Similar to rTMS, tDCS has been used to effectively treat neuropsychiatric conditions such as Parkinson's disease, chronic pain, and major depression [43].Although underlying mechanisms for tDCS are not fully understood, induction of neurochemical changes in targeted brain tissue is being investigated for SUD treatment.

Deep brain stimulation (DBS)
DBS is an invasive NM technique used to treat Alzheimer's disease, Parkinson's disease, and obsessive compulsive disorder [44].It involves a neurosurgical procedure wherein implanted electrodes deliver electrical pulses directly to targeted brain regions, which modulates neural circuitry and subsequently alters neuroplasticity (Fig. 1c).While rTMS and tDCS use lower frequencies to induce excitation or inhibition of neurons, DBS blocks neural transmission with high-frequency stimulation [45].Implanted electrodes are connected to an implantable pulse generator placed under the skin of the chest wall, allowing for continuous stimulation at a preset frequency [46].Thus, stimulation parameters can be modulated as a patient's condition changes.
Unlike other surgical interventions, DBS does not damage brain tissue [47], but given its invasive nature, is associated with infection, seizures or stroke.DBS is well-tolerated once the patient has recovered from the primary surgical procedure [48].Focal stimulation of deep brain regions involved in addiction neurocircuitry (e.g.NAc) may facilitate SUD treatment.
We conducted a systematic review and meta-analysis to determine the efficacy of NM for improving addiction outcomes (e.g., drug craving, consumption, and relapse).As significant progress has been made in this area, a systematic review and meta-analysis building on previous narrative reviews [8,9] with quantification of NM effects in SUDs is warranted.

Search strategy
A comprehensive literature search by two authors (D.M. and A.P., trained on Covidence) was conducted using Medline, PubMed and PsycINFO databases, in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [49] (Supplementary Fig. 1), through October, 2023.Articles published after 2000 in peerreviewed journals were considered.A list of keywords and search terms for four key concepts (SUD, rTMS, tDCS, DBS) was applied (See Supplementary Table 1 for Search Strategy).Reference lists of relevant reviews were also screened for applicable articles.The review was registered at PROSPERO (CRD42023475165).

Eligibility criteria
Using PICOS [50], studies were included if they satisfied the following criteria -Population (P): Studies recruiting participants (18+ years of age) diagnosed with SUD/dependence of alcohol, tobacco, cocaine, methamphetamine, opioids, or cannabis, according to standardized criteria (e.g.,

Study selection
Two authors (D.M. and A.P.) independently screened titles and abstracts obtained on Covidence to determine eligibility for full-text review, and subsequently reviewed the full text of the screened studies.Disagreements were resolved by consensus, and review with the senior author (T.P.G.).

Data extraction and risk of bias
For included studies, two authors (D.M. and A.P.) extracted author information, sample size, study design, stimulation parameters, primary substance use outcomes (craving and consumption), and any secondary outcomes.Effect sizes (Hedge's g) of substance use and other outcomes were calculated for each study using post-treatment data of active and control (sham and/or no treatment) groups, respectively (see Tables 1-4).Due to the heterogeneity in follow-up periods across studies, treatment effects were determined using end-of-treatment data, unless otherwise stated.For DBS studies with no control conditions, within-subject (pre-post treatment) effect sizes were calculated.
The Cochrane Risk-of-Bias Tool (RoB-2) [51] assessed quality of included RCTs.Studies with a high risk of bias were subsequently excluded if at least four domains were considered of moderate risk, or if two or more domains were flagged as high risk.The Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) [52] tool assessed risk of bias in non-randomized studies (DBS Studies); all extracted DBS studies were included in this review.

Meta-analysis
To quantify NM effects, we performed meta-analyses on rTMS and tDCS studies investigating alcohol and tobacco use disorders.Acute versus repeated stimulation were independently evaluated.Meta-analyses were conducted when three or more studies evaluated a synonymous outcome (craving, cue-induced craving, and/or consumption).
We utilized standardized mean difference (SMD; Hedge's g) with 95% confidence intervals (CI's) in each selected meta-analysis to calculate the effect size of NM-related changes in alcohol and tobacco craving, cueinduced craving, and/or consumption (p ≤ 0.05, two-tailed).Randomeffects models pooled individual SMDs, and used data from studies that reported end-of-treatment substance use data from active and control treatment arms.Negative values indicated that active stimulation produced greater reductions in craving, cue-induced craving, and/or consumption compared to sham treatment.The I 2 statistic estimated between-trial heterogeneity; I 2 of ≤40% was considered low heterogeneity, 40-60% moderate heterogeneity, and >60% high heterogeneity [53].Meta-analyses were performed using R version 4.3.1 [54] with package metafor [55].

RESULTS
We identified a total of 94 studies that met our inclusion criteria, with a total of 4306 participants.
Dinur-Klein et al. [84] and Zangen et al. [37] also applied deep-TMS to the lateral PFC and insula using the H-ADD and H-4 coils respectively, and found significant reductions in tobacco consumption and craving [37,84].Importantly, Dinur-Klein et al. [84] applied both 1 Hz (LF) and 10 Hz (HF) stimulation to the lateral prefrontal cortex (PFC) and insula, finding that cigarette consumption decreased significantly only in the 10 Hz condition.These studies were amongst the largest studies of NM for SUDs, with sample sizes of 115 and 262 respectively.The study by Zangen et al. [37] is the only multisite clinical trial in the addiction NM field, and led to FDA clearance of the H-4 coil for smoking cessation.
Notably, while Trojak et al. [81] reported positive results, findings were not maintained at follow-up (12 weeks), signifying a lack of durability in long-term outcomes, though this was the only study to apply LF stimulation (1 Hz) exclusively.
Additionally, two studies [77,84] investigated cue-induced provocation prior to stimulation, and found that presentation of smoking cues reduced cigarette consumption and cue-induced craving, respectively.
Cannabis.Only two RCTs [87,88] examined the use of rTMS for cannabis use disorder (CUD).Sahlem et al. [87] used a randomized, sham-controlled, crossover design to investigate therapeutic effects of a single 10 Hz stimulation session applied to left DLPFC, finding no significant differences in cannabis craving compared to sham.Kozak-Bidzinski et al. [88] applied 20 sessions of 20 Hz rTMS to bilateral DLPFC using a parallel groups design in participants with CUD and schizophrenia.Nonsignificant reductions in cannabis consumption were noted post-TMS versus sham (60 versus 5%), and trends towards reductions in urine toxicology (carboxy-tetrahydrocannabinol) and craving were observed.
Cocaine.Six studies [59,[89][90][91][92][93] investigated rTMS for cocaine use disorder.Two studies demonstrated a significant decrease in cocaine craving following multiple sessions of 15 Hz rTMS to the left DLPFC.Martinez et al. [91] applied both 1 Hz and 10 Hz stimulation to mPFC and ACC using the H-7 coil, finding no significant effect on cocaine craving, though a reduction in cocaine self-administration was present in the 10 Hz condition versus 1 Hz rTMS and sham.Conversely, Bolloni et al. [89] found no significant effects of deep TMS on cocaine consumption when targeting the PFC with H-1 coil, though there was a trend for decreased consumption between baseline and 6-months post-TMS in the active group.Hanlon et al. [59] applied a single stimulation session, finding no treatment-related effects on craving following cTBS to the left frontal pole.
Opioids.Four studies [102][103][104][105] evaluated outcomes in opioid use disorder (OUD) patients following multiple HF rTMS sessions  targeting the left DLPFC.Three studies [103][104][105] reported significant improvements in opioid craving and/or cue-induced craving, with the exception of Tsai et al. [102] who evaluated treatment effects in participants receiving concurrent methadone maintenance therapy.Although there was no significant impact on opioid craving or consumption, an improvement in depressive symptoms was present post-treatment.Li et al. [105] also observed improvements in depressive symptoms, in conjunction with reduced opioid craving, though their participants received concurrent occupational therapy.Liu et al. [104] applied both 1 Hz and 10 Hz stimulation to the left DLPFC, finding that both conditions produced similar reductions in cue-induced opioid craving compared to no treatment.

Transcranial direct current stimulation (tDCS)
Thirty-six studies investigating tDCS as treatment for SUDs, with 1582 participants receiving either active or control treatment (sham stimulation or no treatment; Table 2).
Results showed that longer durations of tDCS resulted in the highest abstinence rate at 6 months post-treatment (25.7%).
Opioids.Three studies [139][140][141] were conducted on tDCS treatment efficacy for OUD.Two studies [139,140] applied ten sessions of 2.0 mA tDCS to the DLPFC for 20 min.Taremian et al. [140] evaluated opioid craving and depressive symptoms in participants receiving methadone, and compared right anodal DLPFC stimulation with sham.Active tDCS significantly reduced opioid craving and depressive symptoms, compared to sham, and methadone alone.Eskandari et al. [139] compared left anodal DLPFC stimulation with right anodal DLPFC stimulation and sham, observing a significant reduction in craving in all groups; active groups exhibited greater effects.Wang et al. [141] applied a single stimulation session targeting the fronto-parietal-temporal area at 1.5 mA for 20 min.Despite these differences, a significant decline in heroin craving was observed, which persisted with the presentation of opioid-related cues.

Deep brain stimulation (DBS)
Seven studies investigated DBS as SUD treatment, with 48 participants receiving active or sham stimulation (Table 3).
Alcohol.Four studies [142][143][144][145] investigated effects of DBS on AUD by targeting the NAc.All studies observed significant decreases in alcohol consumption and/or craving post-treatment.Notably, Bach et al. [145] (N = 12) was the first to compare active and sham DBS and found significant improvements in substance use and craving following 6-months of active stimulation.
Tobacco.One study examined the use of DBS on TUD by targeting the NAc.Kuhn et al. [146] found that 3/10 TUD participants in their study quit smoking post-treatment, while the remaining seven participants showed a significant decline in tobacco craving and cigarette consumption.
Opioids.Two studies [147,148] examined effects of DBS treatment in heroin-dependent participants and reported significant reductions in opioid craving and an increase in opioid abstinence.While Kuhn et al. [147] targeted the NAc exclusively, Chen et al. [148] applied simultaneous stimulation to anterior limb of the internal capsule (ALIC) and NAc.

DISCUSSION
We systematically reviewed the cumulative literature on the efficacy of NM (rTMS, tDCS, DBS) for SUD treatment (Table 4).Findings were inconsistent across each stimulation methodology, and varied significantly with respect to SUD.This may be attributed to variations in treatment parameters, symptom severity across SUD participants, use of adjunctive treatment interventions and population heterogeneity, including the presence of comorbid psychiatric disorders, age, sex, and treatment history.
Nonetheless, findings from rTMS and tDCS studies demonstrated several commonalities.For rTMS, positive outcomes when treating tobacco, stimulant and opioid use disorders were observed, as indicated by post-treatment reductions in subjective and cue-induced substance craving and/or consumption when compared to sham treatment.Accordingly, effect sizes were clinically relevant (Hedge's g > 0.5) but highly variable, consistent with heterogeneity of the published literature [9].Furthermore, meta-analyses found that multi-session active versus sham rTMS was particularly effective in reducing tobacco consumption, but effects on tobacco craving were non-significant.Interestingly, effects of rTMS on AUD were less consistent, with 7/16 studies demonstrating significant improvements.Subsequent metaanalyses found that multi-session rTMS produced significantly greater reductions in alcohol craving and consumption.tDCS studies were promising in the treatment of tobacco, alcohol, stimulant, and opioid use disorders, as suggested by medium effect sizes (Table 2).However, meta-analyses of tDCS trials for AUD and TUD found that both single-and multi-session stimulation were not superior to sham stimulation in reducing craving or consumption, suggesting that rTMS may be superior to tDCS for these SUDs.
DBS produced reductions in craving, consumption and/or abstinence in alcohol, tobacco, and opioid use disorders.Available data is limited to case-series making it difficult to calculate effect sizes (Table 3), with the exception of one randomized shamcontrolled study in AUD [145].Sample sizes in DBS studies were low (ranging 2-12, averaging 6.9 ± 3.1 participants), suggesting the need for larger samples and randomized controlled trials.

Treatment parameters
Variability in treatment efficacy across NM studies may be attributed to differences in stimulation parameters (e.g., stimulation target, frequency, intensity, treatment duration and sample size/demographics).For both rTMS and tDCS studies, multi-session protocols are more effective than single-sessions protocols, as indicated by larger effect sizes and the number of positive outcome studies (see Tables 1-4).This is consistent with previous reports in the addictions neuromodulation literature [149].However, total number of sessions needed to produce  long-lasting effects is unclear and requires further investigation.For rTMS, the most commonly used paradigm across substances was 10-20 sessions once daily.In contrast, studies investigating TMS in depression suggest ≥30 sessions are needed for treatment durability [150].While studies demonstrated persistent effects, including post-TMS reductions in 3-month alcohol [71] and cigarette consumption [82] after only 10 sessions of rTMS, durability of these effects remains uncertain as there is lack of long-term follow-up and biochemical verification beyond 1-month.Amiaz et al. [77] found that reductions in cigarette consumption after 10 sessions of rTMS were not maintained at 6-months.Similarly, number of tDCS sessions needed remains unclear due to lack of long-term follow-up.tDCS protocols were also considerably shorter, with all but two studies [129,138] applying ≤10 sessions overall.Interestingly, Ghorbani Behnam et al. [129] applied 20 total sessions and found that when these sessions were distributed over a longer period of time (12 versus 4 weeks), tobacco abstinence was considerably higher at 6-month follow-up.Accordingly, session frequency may also play an important role.Moreover, potential effects of an accelerated stimulation paradigm (e.g. more than one session daily) should also be further investigated.Studies in depression have found that accelerated protocols are safe and welltolerated, and perform comparably to standard once-daily rTMS [151][152][153].Martinotti et al. [93] conducted the only randomized sham-controlled addictions study to adopt such an accelerated stimulation approach, but reported unfavourable cocaine use outcomes following twice daily stimulation.Nonetheless, Steele and colleagues [154] have found that three iTBS sessions/day for 10 days was tolerable and reduced cocaine consumption.
The need for maintenance sessions following initial stimulation treatment should be further evaluated to increase durability [155].Two studies incorporated weekly reminder sessions following 15 daily HF deep-TMS sessions, and found that reductions in alcohol consumption [68] and tobacco craving [37] persisted 3-months post-treatment.However, Amiaz et al. [77] found that improvements in tobacco use outcomes following 10 HF rTMS sessions and 8 maintenance sessions did not persist at 6-months; this may reflect the effects of the coil (Figure-8 vs. H-coil) or the number of initial sessions (10 versus 15).
Four rTMS studies [83,84,91,104] compared the effects of LF (1 Hz) and HF (10 Hz) stimulation and found that 10 Hz rTMS significantly reduced substance craving and/or consumption, suggesting that HF rTMS stimulation parameters have greater therapeutic potential in comparison to LF stimulation.Accordingly, most rTMS studies used HF stimulation (e.g., ≥5 Hz) regardless of SUD.For tDCS studies, the effects of stimulation intensity (1 mA vs. 2 mA) were less clear.However, tDCS outcomes were more promising when stimulation sessions were of longer duration (>15 min).
Cue-exposure prior to rTMS may activate craving-related neurocircuitry, and subsequent stimulation could then disrupt drug-related memory consolidation [156].Accordingly, Dinur-Klein et al. [84] incorporated smoking cue exposure prior to HF deep TMS and found that it reduced cigarette consumption.Amiaz et al. [77] evaluated differential effects of both neutral and smoking cues prior to HF rTMS, finding that smoking cues reduced cueinduced tobacco craving.This expands on previous findings in both PTSD [157] and OCD [158], wherein provocation using brief cue exposure prior to treatment alleviated symptoms compared to no cue provocation.Future studies should determine whether cue exposure should be utilized in all rTMS and tDCS protocols.
There were inconsistencies for rTMS in AUD treatment, with positive outcomes reported in 44% of studies.Nonetheless, deep TMS was effective when compared to rTMS using a Figure-8 coil, suggesting that the H-coil may be advantageous when treating AUD due to targeting of deep brain structures (e.g., insula, nucleus accumbens).Subsequent meta-analyses did find positive effects of multi-session rTMS on alcohol craving and consumption.However, given that there are several evidence-based treatments available for AUD [159], we suggest that neuromodulation treatment development should be focused on SUDs with a lack of evidence-based biological treatments, such as cannabis and stimulants.

Target brain region
Substance use outcomes with NM are influenced by targeted brain region, as well as the subsequent bilateral or unilateral stimulation of regions of interest.Most rTMS studies for SUDs have targeted the DLPFC (38/50 studies).rTMS targeting the left DLPFC produced predominantly positive effects and clinically relevant effect sizes when treating tobacco, stimulant and opioid use disorders, while those stimulating the right or bilateral DLPFC were less effective (Table 1).In contrast, studies in AUD were not responsive to left DLPFC rTMS, though right and bilateral DLPFC stimulation was effective when multiple sessions were conducted.Alternative regions were less commonly studied.Notably, the mPFC/frontal pole (with or without concurrent stimulation of ACC) emerged as a novel therapeutic target, particularly with a deep TMS protocol with H-coil technology, as indicated by studies with alcohol [63,68] and cocaine [91].Targeting bilateral PFC and insular cortex with deep TMS may also be effective in alcohol and tobacco treatment [37,66,84,86].
Both DLPFC and mPFC have emerged as leading rTMS targets; much remains unknown about the mechanism by which rTMS induces its therapeutic effects in SUDs.An understanding of rTMSinduced alterations in SUD-related brain circuitry is limited as very few studies have incorporated neuroimaging.Furthermore, there is much uncertainty surrounding optimal target locations, both for specific SUDs and individual patients, as there have been no direct head-to-head comparisons of different active rTMS targets.Consequently, it is possible that alternate targets may be required for distinct SUDs.Interestingly, there is evidence that the Default Mode Network may be a SCZ-specific network of tobacco dependence [160].It is critical that rTMS clinical trials include brain-based measures (e.g., MRI, EEG) in order to elucidate mechanisms of action and identify optimal treatment targets.
With respect to tDCS, right anodal DLPFC stimulation appears to be most efficacious across all substances.However, right anodal DLPFC studies had considerably more stimulation sessions (≥5 sessions) than those applying left anodal DLPFC (≤5 sessions) stimulation.Thus, observed differences may be related to treatment duration, and future studies should explore longer durations of left anodal DLPFC tDCS.
Importantly, stimulation sites for rTMS and tDCS are conventionally identified using the 10-20 EEG system or by measuring distances from predefined external landmarks.While this one-size-fits-all approach produces approximate targeting of specified regions, it does not consider inter-individual differences in brain morphology and network architecture.Neuronavigationguided NM with magnetic resonance imaging (MRI) may achieve greater precision with personalized targets.rTMS studies in depression have demonstrated the benefits of such an approach and found that clinical outcomes were significantly improved when patients were stimulated closer to fMRI-personalized targets [161].Selected rTMS studies integrated MRI-neuronavigation [56-58, 60, 75, 81, 90], though the number of studies was insufficient to distinguish its effectiveness in comparison to nonpersonalized targeting.No tDCS studies were present.Consequently, future randomized control trials are warranted to assess the clinical potential of neuronavigation-guided personalized rTMS and tDCS.Most DBS studies targeted the NAc, and were consistently positive.

Alternate neuromodulation modalities
Other NM methods that are less frequently used and excluded from this review include Electroconvulsive Therapy (ECT) [162], Magnetic Seizure Therapy (MST) and Transcranial Alternating Current Stimulation (tACS) [163].Studies examining their effects on SUDs are limited.We also excluded invasive ACC stimulation; ACC implants have shown positive effects, particularly for AUD, although adverse events have been reported [164].

Psychiatric comorbidities
Only a few studies have tested neuromodulation interventions in populations with comorbid psychiatric disorders.Notably, 3/4 of rTMS studies that examined TUD participants with co-occurring SCZ observed significant reductions in tobacco craving and consumption [78,80,85] (Table 1).Prevalence of tobacco use in SCZ is 60-80% and contributes to a 25-year decreased life expectancy in SCZ [165], emphasizing the therapeutic potential of rTMS for this comorbidity.Moreover, SCZ patients have high rates of cannabis misuse [166].Kozak-Bidzinski et al. [88] studied rTMS in outpatients with SCZ and CUD (N = 19).Although the difference in cannabis use was not statistically significant, larger reductions (~60%) were observed in the active (n = 9) versus sham (n = 10) group, highlighting its treatment potential.Ultimately, these NM methods show promise in treating co-occurring SUD and psychiatric disorders, warranting further research in clinical trials with larger sample sizes.

Strengths and limitations
This comprehensive systematic review and meta-analysis contributes substantially to the literature on NM for SUDs for the following reasons: (1) We calculated effect sizes for each study across all three stimulation modalities, and where applicable, conducted a meta-analysis of the published data, to compare and contrast these treatment outcomes.This is the first comprehensive systematic review of the addiction NM literature to include metaanalytic comparisons; (2) We evaluated the treatment efficacy of each stimulation technique, with respect to each SUD and the stimulation parameters applied, to identify their differential effects across substances; (3) We included several new studies that have been published since the reviews by Salling and Martinez [8] and Coles and colleagues [9].
However, there were some limitations.First, there was significant variability in the number of studies for each SUD and NM methodology.Many of these studies were also preliminary (sample size <40 participants).Second, studies were not balanced for sex, with an emphasis on males.Thus, sex-related differences in treatment outcomes are unclear.Third, there was variability in outcomes evaluated (e.g., craving vs. consumption) and in methods used to measure them (e.g., biochemical verification versus self-report).Fourth, as substance use was the primary outcome of interest, associated outcomes such as psychiatric symptoms and cognition were secondary and not always reported.Finally, treatment effects were quantitively assessed using end-of-treatment data due to heterogeneity in follow-up periods.Thus, enduring effects of NM interventions cannot be adequately determined.

Conclusions and future directions
There is considerable promise for the use of NM therapies in SUDs.Nonetheless, further research is required to determine clinical safety and efficacy.Future studies should focus on optimizing stimulation parameters and regimens for these NM methods, with emphasis on stimulation duration, number of treatment sessions needed to produce enduring effects, accelerated treatment paradigms, stimulation frequency and intensity and targeted brain region.Assessment of enduring effects of NM treatment using biochemical verification at extended time-points and the need for maintenance sessions following treatment cessation to optimize clinical outcomes should be emphasized.Neuroimaging data (fMRI) should be acquired prior to, during, and following treatment to elucidate the underlying neural mechanisms mediating treatment effects.Moreover, MRI-neuronavigation may address potential discordance between coil/electrode placement and region of interest, potentially improving treatment efficacy.
Finally, greater emphasis on co-occurring psychiatric disorders is needed.rTMS may be a promising intervention for patients with SCZ and concurrent SUDs, warranting larger randomized shamcontrolled trials.Finally, the potential of adjunctive psychotherapeutic and/or pharmacological intervention should be determined, which may improve substance use outcomes [81].While some studies have implemented concurrent pharmacological interventions [78], few have parsed the clinical impact of each therapy for augmentation of NM outcomes.

Fig. 2
Fig. 2 Meta-analyses of AUD studies using rTMS.Forest plots of studies evaluating (A) alcohol craving following a single-session of rTMS (B) alcohol craving following multi-session rTMS (C) alcohol consumption following multi-session rTMS.

Fig. 3
Fig.3Meta-analyses of TUD studies using rTMS.Forest plots of studies evaluating (A) tobacco cue-induced craving following a singlesession of rTMS (B) tobacco craving following multi-session rTMS (C) tobacco consumption following multi-session rTMS.

Fig. 4
Fig. 4 Meta-analyses of AUD studies using tDCS.Forest plots of studies evaluating (A) alcohol craving following a single-session of tDCS (B) alcohol craving following multi-session tDCS (C) alcohol consumption following multi-session tDCS.

Fig. 5
Fig. 5 Meta-analyses of TUD studies using tDCS.Forest plots of studies evaluating (A) tobacco craving following a single-session of tDCS (B) tobacco consumption following a single-session of tDCS (C) tobacco craving following multi-session tDCS (D) tobacco consumption following multi-session tDCS.
Bold values have been used to highlight the percentage of studies with positive outcomes, as well as the substance use disorder investigated, for improved clarity as well.

Table 1 .
continued Intervention employing either rTMS, tDCS, or DBS; Comparison (C): Studies including either sham stimulation, a control group receiving no intervention or an active control arm were included.DBS studies were exempted considering the ethical constraints on the use of control groups with invasive brain surgery/stimulation; Outcomes (O): Studies investigating substance-related outcomes (consumption, craving, cue-induced craving, abstinence, relapse) as primary or secondary outcomes of interest using a validated measurement tool (e.g. DSM-IV or DSM-5); Intervention (I):

Table 1 .
continued Bold values have been used to highlight the outcome of interest and the brain region targeted, to improve clarity.Substance use disorder investigated is also shown in bold.D.D. Mehta et al.

Table 2 .
continued Bold values have been used to highlight the outcome of interest and the brain region targeted, to improve clarity.Substance use disorder investigated is also shown in bold.

Table 3 .
Deep Brain Stimulation (DBS) [Total N Bold values have been used to highlight the outcome of interest and the brain region targeted, to improve clarity.Substance use disorder investigated is also shown in bold.

Table 4 .
A Summary of End-of-Treatment Substance-use Outcomes in Neuromodulation for Substance Use Disorder Studies.[N = 4036, Participants; 94 Studies].