Repetitive transcranial magnetic stimulation as a potential treatment approach for cannabis use disorder

Abstract

The expanding legalization of cannabis across the United States is associated with increases in cannabis use, and accordingly, an increase in the number and severity of individuals with cannabis use disorder (CUD). The lack of FDA-approved pharmacotherapies and modest efficacy of psychotherapeutic interventions means that many of those who seek treatment for CUD relapse within the first few months. Consequently, there is a pressing need for innovative, evidence-based treatment development for CUD. Preliminary evidence suggests that repetitive transcranial magnetic stimulation (rTMS) may be a novel, non-invasive therapeutic neuromodulation tool for the treatment of a variety of substance use disorders (SUDs), including recently receiving FDA clearance (August 2020) for use as a smoking cessation aid in tobacco cigarette smokers. However, the potential of rTMS for CUD has not yet been reviewed. This paper provides a primer on therapeutic neuromodulation techniques for SUDs, with a particular focus on reviewing the current status of rTMS research in people who use cannabis. Lastly, future directions are proposed for rTMS treatment development in CUD, with suggestions for study design parameters and clinical endpoints based on current gold-standard practices for therapeutic neuromodulation research.

I. INTRODUCTION

Public Health Status of Cannabis Use and Cannabis Use Disorder.

Cannabis is one of the most widely used recreational drugs in the world, with over 188 million reported users worldwide in 2019 (World Drug Report, 2020). Although at present cannabis remains illegal at the federal level in the United States (U.S.), in 2018, 43.5 million people endorsed using cannabis in the past year in the U.S. alone (>15% of U.S. population), behind only alcohol (65%) and tobacco (25%) (SAMHSA, 2019). Rates of cannabis use have increased in the past decade, paralleling changes in the social and political climate favoring legalization (Cerda et al., 2020; Hasin et al., 2019; Kerridge et al., 2017). Likewise, dramatic shifts in sociopolitical stances surrounding cannabis have coincided with a general decrease in stigmatization of cannabis use across the population. Public perception of the risks of regular cannabis use fell dramatically in the last 10-15 years, with a 50-80% decrease in perceived risk for U.S. adolescents (between 2004-2016; Hasin, 2018), a decrease from 50% to 33% for U.S. adults (between 2002-2014; Hasin, 2018), and an 11% jump in the number of adults perceiving there to be no risk at all (Compton et al., 2016). As of December 2020, 15 states plus the District of Columbia have legalized medical and recreational cannabis use; 35 states legalized medical use only; and 16 states decriminalized non-medical use, with 67% of the general public now believing that cannabis should be legal (Pew, 2019) compared to 35% in 2008 (Pew, 2010).

Contrary to these increasingly positive public views, research and clinical data have shown that heavy cannabis use is often associated with cognitive impairment, increased risk for psychotic disorders and other mental and medical comorbidities, lower educational attainment, and unemployment (Colizzi & Bhattacharyya, 2018; Sherman & McRae-Clark, 2016; Volkow et al., 2014). Although the majority of people who use cannabis do not go on to develop any major complications associated with their cannabis exposure, as many as 30% of people who use cannabis endorse experiencing sufficiently disruptive symptoms and adverse life consequences attributed to their cannabis use so as to meet standard clinical criteria for a DSM-5 Cannabis Use Disorder (CUD) diagnosis (Budney et al., 2019). Further, the number of individuals with CUD has steadily increased with expanding legalization (Hall & Lynskey, 2020; Hasin, 2018; Hasin et al., 2019), and the numbers and severity are predicted to continue to increase as cannabis becomes even more widely accepted and available and the perceived risk continues to decline (Budney et al., 2019; Hall & Lynskey, 2020; Hasin, 2018).

Need for More Effective Cannabis Use Disorder Treatment Options.

Currently, combined psychotherapeutic and pharmacological interventions are considered best practice for treatment of substance use disorders (SUDs) (Ray et al., 2020). Yet, unlike many other common SUDs (such as, Alcohol Use Disorder, Opioid Use Disorder, Tobacco Use Disorder), there are no pharmacotherapies which have reached FDA approval for the treatment of CUD (Kondo et al., 2020). While several investigational pharmacologic treatments have shown promise for CUD, none have attained sufficient support for clinical translation, to date (Brezing & Levin, 2018; Kondo et al., 2020; Sherman & McRae-Clark, 2016). Currently, the most widely implemented interventions for CUD remain psychotherapeutic approaches, including cognitive-behavioral therapy, psychosocial therapies such as family and community-based programs, and contingency management (Gates et al., 2016; Sherman & McRae-Clark, 2016), all of which have demonstrated efficacy in reducing frequency and quantity of cannabis use, but with only modest and short-lived abstinence rates (Dutra et al., 2008; Sherman & McRae-Clark, 2016). Aggregate public health reports on CUD treatment outcomes reveal that relapse rates across traditional treatment modalities remain high within the first year after treatment completion (Budney et al., 2008; Hasin, 2018). In fact, several studies have shown that most CUD patients relapse within the first few months of treatment. Moreover, meta-analyses have found lower post-treatment abstinence rates for people who use cannabis than for either cocaine, opioid, or polysubstance users (Dutra et al., 2008). Thus, individuals contending with CUD have a generally low probability of treatment success, leaving many without adequate support to overcome their problematic use patterns.

Given steadily increasing numbers of people who use cannabis, corresponding increases in the incidence and severity of CUD diagnoses, and limited efficacy of current treatment options, expanding and improving upon CUD treatment approaches represents a critical public health need. Fundamentally, to achieve greater efficacy and durability, CUD interventions should aim to mitigate the limitations or complement the strengths of current treatment models, or both.

Identifying Novel Directions for Cannabis Use Disorder Treatment Development.

In terms of etiology, research spanning the last several decades and integrated across multiple scientific disciplines has supported the understanding that substance use and behavioral disorders are fundamentally diseases of the brain, involving aberrant functioning in key areas responsible for regulating emotion, cognition, and behavior (Diehl et al., 2018; Koob & Volkow, 2016; Leshner, 1997; Verdejo-Garcia et al., 2019; Volkow et al., 2016). In fact, the complex, multifaceted behavioral manifestations of SUDs are not solely the result of dysfunction in isolated brain regions but emerge from alterations in function and organization of large-scale, spatially distributed networks of brain regions and their interconnections (Ekhtiari et al., 2016; Fatima et al., 2019; Koob & Volkow, 2016). Maladaptive changes in the structure, function, and neurochemistry within and between these core neural circuits are associated with the emotional dysregulation, cognitive impairments, and behavioral dysfunction that contribute to the development of drug-dominated cognitive-behavioral schema and likely lead to the transition from occasional, recreational use to problematic, unhealthy patterns of use characterizing SUDs (Fatima et al., 2019; Koob & Volkow, 2016; Verdejo-Garcia et al., 2019; Volkow et al., 2014; Yanes et al., 2018).

Despite our increasingly enriched understanding of the brain mechanisms underlying many key features of SUDs, the real-world implementation of this neuroscientific knowledge into clinical practice, or even into clinical research, has been slow (Ekhtiari et al., 2019; Steele et al., 2019; Verdejo-Garcia et al., 2019). High-quality brain data could potentially aid in addressing limitations of current treatment models, thereby improving treatment approaches and reducing the low treatment success rates (Ekhtiari et al., 2019; Singh & Rose, 2009; Steele, 2020; Steele et al., 2019; Verdejo-Garcia et al., 2019).

Exploring a Brain-based Treatment Approach for Cannabis Use Disorder.

Therapeutic neuromodulation approaches are an increasingly prevalent group of techniques used in neurology and psychiatry to target the neural endophenotypes of neuropsychiatric diseases in order to produce clinically relevant symptom improvement (Brunoni et al., 2019; Diehl et al., 2018; Fox et al., 2014). The most common neuromodulation approaches include neurostimulation techniques [which use external electronic tools to apply electromagnetic fields that exogenously alter activity patterns of sensitive cells; principally, neuronal cells (Chervyakov et al., 2015b)], and neurofeedback techniques [which are a group of biofeedback methods that involve learning self-regulation of cognitive and emotional states through volitional control of specific brain rhythms, typically facilitated by neurorecording tools that can instantaneously measure brain waves and neurophysiological responses to produce a real-time signal, such as a visual or auditory cue, that conveys time-locked feedback of the status and quality of attainment of a desired neurofunctional state (Martz et al., 2020)]. In general, these neuromodulation techniques operate by directly influencing neural functioning in brain regions and their associated functional circuits that then produce measurable cognitive and behavioral outputs, thereby enabling empirical demonstration of the causal links between neural manipulation and behavior change (Bestmann et al., 2008; Pascual-Leone et al., 2000; Silvanto & Pascual-Leone, 2012; Sliwinska et al., 2014; Vosskuhl et al., 2018). This contrasts with neuroimaging methods, such as functional magnetic resonance imaging (fMRI), positron emission tomography (PET), or electroencephalography (EEG), which chiefly use observational neurophysiological recording techniques during varying mental or behavioral states to infer brain-behavior associations (Sack & Linden, 2003; Silvanto & Pascual-Leone, 2012). As such, neuromodulation can uniquely be used to directly target, selectively and directionally manipulate, and, ideally, normalize aberrant neural processing circuits that contribute to specific maladaptive behaviors (Diehl et al., 2018; Dunlop et al., 2017; Guse et al., 2010; Hanlon et al., 2016; Sack & Linden, 2003; Yavari et al., 2016). Using these techniques alone or in conjunction with other established treatment modalities can provide an alternative or improvement upon SUD treatment approaches.

Among the most promising neuromodulation approaches for SUD treatment are the non-invasive brain stimulation (NIBS) techniques. In particular, repetitive transcranial magnetic stimulation (rTMS) has been the most widely investigated NIBS approach in neurology and psychiatry, although other noteworthy modalities include the transcranial electrical stimulation (tES) methods, which principally comprise transcranial direct current stimulation (tDCS), transcranial alternating current stimulation (tACS), and transcranial random noise stimulation (tRNS). While there are numerous literature reviews which have comprehensively summarized the conceptual framework and current status of therapeutic neuromodulation studies investigating NIBS across a range of SUD groups (Coles et al., 2018; Diana et al., 2017; Dunlop et al., 2017; Ekhtiari et al., 2019; Hanlon et al., 2016; Luigjes et al., 2019; Verdejo-Garcia et al., 2019; Yavari et al., 2016), the evidence presented in support of rTMS has generally outpaced what is known about tES techniques, and as such, rTMS treatment findings tend to occupy a large portion of these discussions, with tDCS research making up a close second (Ekhtiari et al., 2019). More importantly, however, none of these comprehensive literature reviews summarize and discuss the current state of NIBS research for CUD.

Thus, the objective of the present review is to discuss issues related to using NIBS to target substance use-related neural circuitry for the treatment of CUD.

In general, there is a surprising paucity of therapeutic neuromodulation research in CUD, with, to our knowledge, only three rTMS (Prashad et al., 2019; Sahlem et al., 2018; Sahlem et al., 2020) and one tDCS (Boggio et al., 2010) interventional studies published, to date. Here we will focus solely on reviewing the rTMS-based interventional studies since there is greater clinical appraisal of rTMS for SUDs, and in psychiatry in general, enabling a more informed discussion on the relative merits of rTMS for CUD. We will assimilate the contributions of the three rTMS studies into a discourse on the potential utility of these approaches for CUD; highlight challenges unique to cannabis and CUD that might confound the implementation of rTMS in this population; integrate recent discoveries in neuroscience and technological or methodological advancements in neuromodulation techniques; and finally, propose a framework or foundation for the future development of rTMS research in CUD.

II. INVESTIGATING REPETITIVE TRANSCRANIAL MAGNETIC STIMULATION TECHNIQUES AS NOVEL, EMPIRICALLY DERIVED, NEUROBIOLOGICALLY TARGETED TREATMENT APPROACHES FOR CANNABIS USE DISORDER

Principles of Repetitive Transcranial Magnetic Stimulation (rTMS) for Therapeutic Applications.

General TMS Physiology.

TMS involves the use of electromagnetic induction to generate an electrical current in the brain that leads to the depolarization of neurons (Barker et al., 1985). In humans, TMS is delivered non-invasively in conscious and alert individuals by placing a small to moderate-sized electromagnetic coil, typically encased in plastic housing, gently against the scalp. An electric pulse generator, or stimulator, generates a rapidly oscillating electric current within the coil which results in a perpendicular magnetic field directly beneath it. This magnetic field readily penetrates the skull to reach the cerebral cortex, thus enabling the mostly painless delivery of discrete magnetic pulses to particular cortical locations with reliably high spatial specificity (Deng et al., 2013; Hanlon et al., 2013). These rapidly oscillating magnetic pulse fields then induce a secondary, electric (eddy) current in the underlying cortical tissue which, owing to the electrochemical nature of neuronal cells, is capable of modulating the excitability and activity patterns in those neurons in accordance with Faraday’s Law of electromagnetic induction (Barker, 1991; Chervyakov et al., 2015b; Lefaucheur et al., 2014; Pascual-Leone et al., 2000).

Although the most widely used TMS coils, e.g., figure-8 coils, tend to have relatively shallow stimulation depths, with induced fields typically reaching superficial cortical tissue of only ~1-4 cm² (Deng et al., 2013; 2014; Stokes et al., 2013), TMS techniques have the ability to manipulate large-scale, spatially distributed networks of brain regions by not only generating changes in activity at the direct stimulation target beneath the coil, but also by evoking reciprocal changes in activity in downstream functionally and structurally connected cortical and subcortical regions via transsynaptic modulation (Bestmann et al., 2008; Hanlon et al., 2013; 2017; Opitz et al., 2016; Pascual-Leone et al., 1998; Strafella et al., 2001; To et al., 2018; Vink et al., 2018). Specifically, when the TMS-induced electrical current in the targeted cortex is sufficiently strong, it is capable of depolarizing those underlying neurons, as well as promoting a cascade of neurotransmitter release, excitatory postsynaptic potentials, and eventually action potentials in downstream neurons receiving mono- and poly-synaptic inputs from the stimulated neurons (Chervyakov et al., 2015b). This is why direct TMS effects at relatively superficial cortical targets are so useful for non-invasively producing widespread neural effects that impact multiple interrelated circuits, including those encompassing deeper subcortical regions, which ultimately influence the complex, multifaceted psychophysiological processes regulated by these brain networks (Bestmann et al., 2008; Opitz et al., 2016; To et al., 2018).

Conventional (Fixed-Frequency) rTMS.

When TMS pulses are delivered repeatedly in rapid succession, known as repetitive TMS (rTMS), durable alterations in neuronal firing patterns along the corticospinal tract can occur through promotion of neural plasticity—an intricate set of processes resulting in sustained changes in neural activity patterns and functional and structural circuit organization, as well as the corresponding neurobehavioral consequences of these changes, which endure beyond stimulation cessation (Chen et al., 1997; Hoogendam et al., 2010; Maeda et al., 2000; Siebner et al., 2000). While the precise mechanisms governing rTMS-induced neural plasticity remain elusive and a fervent topic of investigation (Hoogendam et al., 2010; Rossi et al., 2009; 2021), extensive preclinical and clinical research point to a role of synaptic coupling efficiency, with either strengthening or weakening of neural communication pathways corresponding to the induction of either long-term potentiation (LTP) or long-term depression (LTD), respectively (see reviews by Chervyakov et al., 2015b; Platz & Rothwell, 2010; Stagg & Nitsche, 2011). Furthermore, the magnitude, direction, and duration of neuroplasticity induction are governed by intensity- and timing-specific parameters, such as the rTMS stimulation intensity (set as a given percentage of an individual’s motor threshold; described below), stimulation pulse rate, total pulse number, inter-pulse interval, duration of pulse trains, inter-train interval, etc. (Hoogendam et al., 2010; Maeda et al., 2000; Pell et al., 2011).

In general, high-frequency rTMS (HF-rTMS; ≥5 Hz pulse rate) typically produces an excitatory (LTP-like) effect, which leads to an enduring increase in cortical excitability and is employed when up-regulating activity within a targeted region or circuit is desired (Chervyakov et al., 2015b; Hoogendam et al., 2010; Pascual-Leone et al., 1994; Siebner et al., 2000). Whereas, low-frequency rTMS (LF-rTMS; ≤1 Hz pulse rate) typically produces an inhibitory (LTD-like) effect, which leads to an enduring decrease in cortical excitability and is employed when down-regulating activity within a targeted region or circuit is desired (Chen et al., 1997; Chervyakov et al., 2015b; Hoogendam et al., 2010).

Measuring Cortical Excitability.

Historically, the modulating effects of rTMS on cortical excitability have been directly evaluated on the basis of measurable shifts in the peak-to-peak amplitude of motor evoked potentials (MEPs) (Barker et al., 1986; Hallett, 1996; Rossini et al., 1994). MEPs are the physiological responses elicited in peripheral muscles when TMS is delivered over corresponding regions of the primary motor cortex (M1). Namely, when a single TMS pulse is applied to the M1 hand or leg region, a reliable contraction is induced in the contralateral hand or leg muscle (respectively), which is often visually observable and the amplitude of which can be detected using peripheral physiological recording, such as surface electromyography (EMG) (Barker et al., 1985; Bestmann & Krakauer, 2015; Rothwell et al., 1987; Day et al., 1989). The magnitude of the MEP amplitude represents the number of neurons successfully depolarized by the TMS pulse (Groppa et al., 2012), with larger MEP amplitudes indicating a greater number of neurons sensitive and responsive to stimulation and is thus a reflection of higher corticospinal excitability (Nitsche & Paulus, 2000; Ridding & Rothwell, 1997). In this way, measurement of MEPs constitutes one of the hallmark biomarkers for non-invasive quantification of corticospinal excitability levels (Bestmann & Krakauer, 2015; Chipchase et al., 2012).

Patterned rTMS.

Beyond the conventional fixed frequency rTMS varieties, there are also patterned variants of rTMS which are becoming increasingly popular. Patterned rTMS refers to protocols involving short, repeated bursts of rTMS pulses, known as pulse trains, accompanied by brief pauses between these pulse trains, which are referred to as intertrain intervals (ITIs) (Rossi et al., 2009; 2021; Suppa et al., 2016). The specific pattern of pulse burst delivery within a pulse train can alter the consequent rTMS-induced effects. For instance, thetaburst stimulation (TBS), which is the most commonly explored group of patterned rTMS variants, involves theta frequency (5 Hz) pulse rates which mimic endogenous neuronal firing patterns employed by the brain during natural learning and memory processes (Di Lazzaro et al., 2005; 2008; Huang et al., 2005; 2009). TBS follows a rhythmic bursting pattern consisting of three-pulse bursts delivered at 50 Hz every 200 ms (aka at 5 Hz) (Huang et al., 2005; Suppa et al., 2016). Depending on the precise temporal pattern of TBS pulse delivery, either excitatory, LTP-like neural after-effects (i.e., intermittent TBS, or iTBS: 2-sec train of TBS with 10-sec ITI for a total of 190 sec and 600 total pulses; Huang et al., 2005) or inhibitory, LTD-like neural after-effects (i.e., continuous TBS, or cTBS: 40-sec train of uninterrupted TBS and 600 total pulses; Huang et al., 2005) can be induced in a circuit-specific way (Dunlop et al., 2017; Hanlon et al., 2013; 2017). Further, TBS-induced changes in cortical excitability have been shown to be more robust and enduring, and with session lengths only a fraction of those required for conventional rTMS (i.e., ~15-40 min vs. ~2-5 min for conventional rTMS and TBS, respectively) (Huang et al., 2007; 2008; 2009; Ishikawa et al., 2007; Katayama & Rothwell, 2007).

Standard rTMS Session Procedures.

There is a multitude of approaches to conducting an rTMS treatment session, with the number and variety of ways increasing exponentially over time as new data and technologies emerge that further refine various aspects of the treatment process. However, at the least, a standard rTMS treatment procedure will generally involve several fundamental steps which are described below.

Motor Threshold Determination.

During a standard rTMS procedure, the first step is to determine an individual’s motor threshold (MT), which is the lowest TMS stimulation intensity required to apply over the M1 region for that individual to exhibit a reliable corticospinal response (Rossini et al., 1994), and is based on MEP measurements. MTs are known to show considerable intra- and inter-individual variability, even in healthy people (Groppa et al., 2012; Kloppel et al., 2008; Lopez-Alonso et al., 2014; Tranulis et al., 2006). Because intrinsic levels of cortical excitability vary across individuals, these differences can directly impact the intensity of stimulation a given individual needs to achieve neuronal depolarization in response to TMS, which is reflected by the magnitude of their estimated MT (Pellegrini et al., 2018; Rossi et al., 2009).

Treatment Stimulation Dose.

Consequently, the rTMS treatment stimulation “dose” (Peterchev et al., 2012), which is the therapeutic dose of rTMS applied to the clinical brain target of interest, is a personalized treatment parameter which is adjusted by MT to account for individual differences in cortical excitability. Customarily, the treatment dose is derived from the MT value estimated while at rest, known as the resting MT (rMT). The rMT is the minimum stimulation intensity of single-pulse TMS needed to be applied to M1 to produce a reliable MEP response (e.g., MEP amplitude of at least 50μV-1mV in at least 50% of a series of single-pulse TMS trials; Rossini et al., 1994) in the contralateral abductor pollicis brevis (APB) muscle in the hand (for shallow coil designs) or contralateral tibialis anterior muscle in the leg (for deeper coil designs) while these muscles are relaxed and neutral (Priori et al., 1993; Roth et al., 2014; Schecklmann et al., 2020). As such, personalized dosing of therapeutic rTMS is calibrated for each patient by constituting a relative percentage of their estimated rMT (Hanajima et al., 2007; Peterchev et al., 2012; Pridmore et al., 1998).

Scalp-to-cortex Distance.

In addition to adjusting for individual differences in rMT, it is also recommended that intra- and inter-individual differences in the scalp-to-cortex (STC) distance for the therapeutic brain target be taken into consideration when establishing the treatment threshold (Kahkonen et al., 2004; Kozel et al., 2000). This is because rTMS-induced effects on cortical depolarization are directly impacted by the electrical field potential actually reaching the brain tissue, which is known to decay rapidly in proportion to the distance between the coil and cortex, a function of the physical laws governing the influence of distance on electromagnetic induction as defined by Maxwell’s equations (Kozel et al., 2000; Stokes et al., 2007). In particular, since the rMT is measured at the motor cortex, areas of the brain with cortical tissue further away from the scalp than the motor cortex will typically require stronger TMS intensities in order to achieve comparable electrical field potentials to achieve neuronal depolarization (Janssen et al., 2014; Kahkonen et al., 2004; Knecht et al., 2005). Thus, therapeutic targets within the PFC, for example, might require stimulation thresholds of 110%-130% the rMT (or 10-30% higher stimulation intensities than the motor cortex) in order to attain expected neural effects given the greater STC distances for most prefrontal regions relative to M1 (Kähkönen et al., 2004; Kozel et al., 2000; Milev et al., 2016; Stokes et al., 2007; 2013).

Moreover, there is considerable variability in the precise STC distances measured for the same brain region across different individuals (Kozel et al., 2000; Stokes et al., 2007; 2013). In fact, beyond normative inter-individual variability, many neurologic and psychiatric populations, including those with SUDs, are known to exhibit disease-related cortical atrophy which can further increase STC distance and alter MEP and MT estimates in patients relative to healthy controls (Chervyakov et al., 2015a; Groppa et al., 2012; Hanlon et al., 2019a; Iriarte & George, 2018; Klöppel et al., 2008). Thus, several tools have been developed to allow researchers and clinicians to directly measure STC distances from neuroanatomical MRI images acquired for individual patients or participants (MANGO software, http://ric.uthscsa.edu/mango/; BrainRuler software, Summers & Hanlon, 2017), which can be used to make additional personalized adjustments to rTMS treatment dose thresholds based on their unique STC distance anatomy (Kozel et al., 2000).

Individually titrated stimulation dosing not only provides a systematic and personalized way of quantifying stimulation doses for maximal efficacy, but also for staying within established safety guidelines (Rossi et al., 2009; 2021; Westin et al., 2014) since, for some individuals, stimulation intensities in excess of the doses required to effectively reach and modulate the cortex may lead to excitotoxicity or seizure induction, at worst, or unnecessary side effects, such as headaches and scalp pain, at best (Westin et al., 2014). Effectually, adherence to these procedures helps to ensure that each recipient has the best chance of achieving the desired therapeutic response while minimizing the incidence of adverse events (McClintock et al., 2018; Rossi et al., 2009; 2021; Wasserman, 1998).

Cortical Targeting & TMS Coil Positioning.

The next step after the rMT has been identified and used to establish the rTMS treatment dose, is to position the TMS coil over the therapeutic brain target for the duration of the treatment session. Coil positioning is typically determined using a standard procedure, such as a distance-based method (e.g., 5 cm rule; Pascual-Leone et al., 1996) or anatomical landmark-based method (e.g., International EEG 10-20 system; Herwig et al., 2003) which relies on previously established standardized coordinates for the expected locations and relative distances of key neuroanatomical structures. More recently, though, more advanced cortical targeting procedures which account for individual differences in brain anatomy and function have become widely available and are preferred for their accuracy and precision; e.g., methods involving personalized anatomical or functional MRI-based frameless stereotactic neuronavigation systems (Rusjan et al., 2010; Sack et al., 2009; Sparing et al., 2008). Then, to ensure consistency in positioning across repeat rTMS treatment sessions, the precise location and orientation of coil placement for each individual is documented during their first session. Depending on the particular method of cortical targeting that is implemented, these parameters are either documented through inscription on a nylon head cap which is worn at the first and each consequent session or saved and later uploaded within the neuronavigation system software.

Sham Condition & Blinding Integrity.

As with any well-designed research study or randomized clinical trial, it is expected that a blinded parallel or crossover placebo condition (for medical devices, typically referred to as a “sham”) will be included to control for any non-specific effects (Broadbent et al., 2011; Conde et al., 2019; Duecker & Sack, 2015; Flanagan et al., 2019; Loo et al., 2000; Zis et al., 2020). This is especially the case when study outcomes are based largely or in part on subjective measurements, such as the self-report endpoints common in SUD clinical trials (e.g., drug craving, withdrawal symptoms, mood, sleep quality, timeline follow-back, etc.), as these outcomes make it more difficult to disentangle placebo effects than quantifiable, objective outcome measures (Fregni et al., 2006ab; Rossi et al., 2009; Wasserman & Lisanby, 2001).

Although there are a number of different sham methods which have been developed for rTMS research, the overall goal of an effective sham is to simulate the multisensory experience of active rTMS by including the sound, pressure, and scalp sensations produced by real rTMS, but without any physiological effects on the CNS (Broadbent et al., 2011; Conde et al., 2019; Lisanby et al., 2001; Loo et al., 2000; Rossi et al., 2009). Previous studies have demonstrated that when these criteria are met, participants are generally unable to differentiate real from sham stimulation, with most people guessing at no better than chance when reporting whether they received real or sham stimulation in a given session (Borckardt et al., 2008; Broadbent et al., 2011; Flanagan et al., 2019; Hanlon et al., 2015). However, to monitor and maintain continued assurance of effective blinding, current guidelines recommend that participants should be surveyed after each session or study period to routinely assess blinding integrity (Berlim et al., 2013; Broadbent et al., 2011; Flanagan et al., 2019; Lefaucheur et al., 2014; Rossi et al., 2021) in order to maintain rigor and reproducibility.

rTMS Parameter Space.

There are a number of key factors which determine whether an individual will respond or not respond to rTMS treatment, as well as those influencing the magnitude, direction, timing, and durability of any rTMS-induced clinical effects that do emerge (Lefaucheur et al., 2014).

Number of Treatment Sessions.

Typically, a single session of rTMS (e.g., ranging from ~600-3600 pulses delivered in anywhere from 3-40 min) will produce acutely observable neurophysiological effects [such as, MEPs in the hand or leg muscles, changes in heart rate or blood pressure, changes in brain activity as measured by altered blood flow and oxygenation, glucose metabolism, dopamine signaling, etc.] starting immediately, and lasting anywhere from ~30 min (i.e., for HF-rTMS or iTBS) to several hours (i.e., for LF-rTMS or cTBS) post-stimulation (Hoogendam et al., 2010; Suppa et al., 2016; Ziemann et al., 2008). However, longer-term effects which endure for weeks, months, or years, and/or measurable alterations in more complex psychophysiological processes, such as cognitive, emotional, and behavioral correlates, typically emerge only after several rTMS sessions are delivered successively in a fairly systematic manner (i.e., repeated over several days per week for multiple weeks, or several times per day over several days, such as in accelerated protocols) (Madeo et al., 2020; Rossi et al., 2009; 2021; Song et al., 2019; Teng et al., 2017). This is because the neurobiological mechanisms purportedly underlying rTMS-induced LTP- and LTD-like plasticity are cumulative processes evoked by repeated synaptic perturbation which ultimately activate neurochemical and neurophysiological pathways that take time to manifest (e.g., changes in gene expression which facilitate the altered production of neurotransmitters, receptor proteins, and signaling molecules necessary for long-term modulation of synapse strength, neuronal architecture, and functional and structural connectivity, etc.) (Hoogendam et al., 2010; Platz & Rothwell, 2010; Rossi et al., 2009; 2021).

In fact, based on the principal study backing the 2008 FDA clearance of rTMS for treatment-resistant major depressive disorder (TRD) (O’Reardon et al., 2007), and similar studies which followed (Fitzgerald et al., 2006), rTMS treatment response was shown to accumulate over time to reach a clinically meaningful level. According to the pioneering TRD rTMS treatment protocol, daily rTMS sessions were required for at least 4-6 weeks before clinical success could be achieved, with larger consequent studies showing significant suppression of depressive relapse with the addition of a number of intermittent maintenance sessions in the weeks to months following initial treatment completion (Berlim et al., 2014; Janicak et al., 2010; Manzardo et al., 2019; Richieri et al., 2013). Similarly, a meta-analysis of NIBS studies across SUD groups showed that multi-session treatment protocols reliably show larger effect sizes for reducing drug craving and consumption than single-session protocols, with the number of sessions and total number of pulses administered proportional to the extent of drug craving reduction achieved, indicating a clear dose-response effect (Song et al., 2019).

Diverse rTMS Treatment Parameters.

That said, the precise number, duration, and spacing of rTMS treatment sessions required to achieve clinically meaningful outcomes is not always immediately apparent, nor is it uniform across rTMS applications in either research or clinical practice (Lefaucheur et al., 2014; 2020; Rossi et al., 2009; 2021; Teng et al., 2017). This is because such factors are often conditioned on the specific clinical population being addressed, rTMS modality and parameters being implemented, and any clinical and demographic variance present at the individual level (Davila et al., 2019; Lefaucheur et al., 2014; Manzardo et al., 2019; Ridding & Ziemann, 2010; Xie et al., 2013). Given the multiplicity of design decisions arising from the vast and varied rTMS parameter space, understanding the consequences of making certain parameter choices over others continues to be a rich, ongoing area of investigation in the field, with research expressly dedicated to the development, appraisal, and optimization of rTMS procedures with the ultimate goal of generating best practices for achieving and maximizing therapeutic success across various clinical groups (Ekhtiari et al., 2019; Lefaucheur et al., 2014; 2020; Rossi et al., 2021; Verdejo-Garcia et al., 2019).

Safety & Tolerability.

As a non-invasive form of neuromodulation, interventions based on rTMS are generally well tolerated and involve minimal risk of serious side effects.

Headaches, Scalp Pain & Neck Discomfort.

The most commonly reported side effects include mild, transient headaches, neck pain, or scalp tenderness at the site of stimulation (see Machii et al., 2006 for review of rTMS side effects). The headaches and scalp discomfort are thought to be a consequence of coincidental excitation of muscle and sensory nerves overlying the skull at the site of stimulation which result in muscle tension, twitching, and prickly tingling sensations (Lefaucheur et al., 2014; Machii et al., 2006). For susceptible individuals who do experience unpleasant local pain or muscle tension-type headaches, the discomfort is usually temporary and diminishes shortly after stimulation cessation. However, if headaches and pain persist, most patients will respond well to over-the-counter analgesics, which is the clinical recommendation as published in current rTMS safety guidelines (Lefaucheur et al., 2014; Rossi et al., 2009).

Importantly, rTMS-related pain and discomfort has been shown to decline fairly rapidly over the first few daily treatment sessions, with or without analgesic intervention (Janicak et al., 2008; Anderson et al., 2009). This is thought to be related to desensitization processes of the sensory nerves as predicted by the gate control theory of pain (Melzack & Wall, 1965). Thus, to facilitate desensitization and enhance tolerability, newer treatment protocols generally incorporate a “ramping” procedure which involves intentionally initiating below the target stimulation dose and gradually increasing stimulation intensity over the first few days (Rossi et al., 2009) or seconds (Hanlon et al., 2017) of treatment.

Hearing Impairment.

Further, the TMS pulse causes mechanical vibrations in the coil which produce a brief, intense sound impulse, which is often referred to as a TMS coil “click” (Counter & Borg, 1992; Counter et al., 1991; Goetz et al., 2015; Koponen et al., 2020). Exposure to this click during rTMS sessions has even been shown to produce mild hearing loss as reflected by transient increases in auditory threshold (Loo et al., 2001; Pascual-Leone et al., 1992). However, these hearing disturbances are largely or completely avoided by requiring participants and operators to wear hearing protection, such as adjustable foam earplugs, throughout the duration of rTMS sessions (Levkovitz et al., 2007; Rossi et al., 2007; Janicak et al., 2008; Schraven et al., 2013). Thus, hearing protection earplugs are the current safety recommendation to prevent TMS-induced acoustic trauma, in addition to performing auditory threshold testing before and after rTMS sessions to routinely monitor safety and health (Counter et al., 1991; Rossi et al., 2009; 2021).

Seizure Induction.

Broadly-speaking, however, the most serious potential side effect of rTMS involves the risk of seizure induction (Rossi et al., 2009). While the risk of provoking a seizure is the most concerning of the acute adverse events associated with rTMS, it remains an extremely rare occurrence, under research and clinical circumstances alike, particularly since the development, dissemination, and broad adherence to consensus clinical safety guidelines (Rossi et al., 2009; 2021). In fact, in its entire clinical and research history up to 2020, the overall risk of TMS-related seizures in humans remains <1% (Lerner et al., 2019; Stultz et al., 2020), with most reported seizures involving patients with pre-existing neurological conditions or treatment parameters outside the recommended guidelines (Chen et al., 1997; McClintock et al., 2018; Rossi et al., 2009; 2021; Wasserman, 1998). Risk factors which have been shown or are purported to increase the likelihood of seizure induction include history of traumatic brain injury/concussion, stroke, or neurologic disease with altered seizure threshold, sleep deprivation, heavy alcohol or benzodiazepine use or withdrawal, personal or family history of epilepsy or seizures, and the use of psychotropic medications which lower the seizure threshold (Lerner et al., 2019; Stultz et al., 2020). However, even under circumstances with elevated patient-risk, the relative risk of seizure is still lower than for most common psychotropic medications (Shafi, 2019), making TMS a fundamentally minimal-risk intervention.

Comparison to tES NIBS Approaches.

As with rTMS, tES techniques, such as tDCS and tACS, are also considered non-invasive, albeit they involve the principal use of electrical rather than magnetic stimulation (Bikson et al., 2016). Specifically, tES involves delivering either direct (tDCS) or alternating (tACS) low-intensity electric currents through cortical tissue via strategic placement of surface electrodes on the scalp and/or upper body (Ekhtiari et al., 2019; Woods et al., 2016). In contrast to TMS, tDCS is not believed to influence neuronal firing patterns directly, but instead, is purported to involve the subthreshold modulation of cortical excitability through the induction of a polarity-dependent shift in neuronal membrane potential which is what leads to an increased or decreased likelihood of consequent depolarization (Bindman et al., 1964; Nitsche & Paulus, 2000; 2001; Woods et al., 2016). Further, the tDCS-induced direction of membrane polarization is defined by the relative placement of the anodal and cathodal electrodes over the stimulation target of interest (Kabakov et al., 2012; Rawji et al., 2018; Woods et al., 2016). Importantly, the electrical current density in the scalp via the mechanisms involved in tES techniques is much lower than the secondary (eddy) electrical currents induced by TMS-based techniques; thus, TMS is capable of stimulating cortical tissue with a fraction of the cutaneous pain associated with tES (Rossi et al., 2009).

Potential for Effective rTMS Interventions in SUDs/CUD.

The potential for NIBS to represent a paradigm shift in psychiatry by achieving robust clinical outcomes as a targeted, non-pharmacologic, brain-centric treatment approach, while largely avoiding the drawbacks of the more invasive neuromodulation methods is a major justification for the expansion of NIBS research, particularly in complex patient populations (Chen et al., 2017; Grunhaus et al., 2000; 2003; Ren et al., 2014). Considering the often-intransigent chronic relapse cycles characterizing SUDs, individuals contending with substance use problems are, perhaps, an archetype of the complex and treatment-resistant clinical populations that might benefit most from the relatively low risk but potentially high reward NIBS-based treatment approaches (Bulteau et al., 2020). Thus far, rTMS-based protocols have been the most widely studied, effective, and well-tolerated neuromodulation-based treatment options for SUDs, making them a practical choice for rapid translation from clinical research to the clinic.

The following sections provide a preliminary exploration of the current literature assessing whether the promise of rTMS-based interventions emerging for other SUD groups might extend to individuals with CUD.

Studies Investigating rTMS Treatment Potential for Cannabis Use Disorder.

1. Sahlem et al., 2018. Feasibility and Proof-of-Principle Study of Single-session Left DLPFC HF-rTMS.

Although NIBS studies for various SUDs have consistently demonstrated the ability of DLPFC-targeted rTMS to reduce drug cue-elicited craving and/or drug consumption after multiple sessions, and sometimes even after a single session (Coles et al., 2018; Song et al., 2019; Zhang et al., 2019), Sahlem & colleagues (2018) were the first to conduct a study investigating rTMS targeted to the DLPFC in people who smoke cannabis. This double-blinded, sham-controlled, within-subject crossover design study was conducted to determine if a single session of HF (excitatory)-rTMS could be feasibly (i.e., safely, tolerably) delivered to the L DLPFC in a group of non-treatment-seeking cannabis smokers with CUD, and whether single-session active HF-rTMS could produce anti-craving effects relative to sham stimulation in this population.

Non-treatment seeking cannabis smokers meeting DSM-5 criteria for CUD (n=18) were randomized to receive a single session of either active or sham HF-rTMS (10 Hz, 110% rMT, 4000 pulses; Magventure figure-8 design coil) to the L DLPFC (F3 International 10-20 EEG system coordinate) while completing a cannabis cue exposure paradigm; one week later they received a single session of the other stimulation condition. Feasibility and tolerability were measured by the rate of retention in the study and the percentage of participants able to tolerate full-dose active rTMS, respectively. Self-reported cannabis craving was assessed before and after both stimulation sessions using the standardized Marijuana Craving Questionnaire (MCQ).

The authors reported that HF-rTMS to the L DLPFC was both safely and tolerably delivered to cannabis smokers with CUD, as 16 of the 18 participants completed the trial (i.e., 89% of participants were retained across all 3 study visits), and all 16 of the treatment completers tolerated the active rTMS at full dose without any adverse experiences. However, there was no impact of stimulation on cannabis craving, as MCQ scores following HF-rTMS did not significantly differ from baseline, relative to sham, indicating that a single session of L DLPFC-targeted HF-rTMS was not sufficient to reduce cue-elicited craving in heavy cannabis smokers.

Discussion of Limitations and Contributions.

While Sahlem et al. (2018) did not reveal an effect of rTMS on cannabis craving, it was a single session, which has only rarely been shown to affect drug craving (Ekhtiari et al., 2019; Kearney-Ramos et al., 2019; Song et al., 2019), and was thus, not a surprising finding. In fact, as mentioned previously, this is consistent with most of the FDA-cleared rTMS protocols in psychiatry which require frequent (typically daily) stimulation sessions over multiple weeks to months before clinically significant symptom improvements are anticipated (O’Reardon et al., 2007; Politi et al., 2008; Rapinesi et al., 2016; Senova et al., 2019).

2. Sahlem et al., 2020. Case Report on Feasibility and Clinical Effects of Accelerated Multi-session L DLPFC HF-rTMS Intervention.

As a letter to the Editor, Sahlem & colleagues (2020) followed up with a case series extending their single-session proof-of-principle study to a multi-session interventional study. In this open-label safety and tolerability pilot study, 20 sessions of L DLPFC HF-rTMS were administered in an accelerated format over a 2-week period (i.e., 2 sessions daily over 10 business days) to participants with CUD who reported a desire to reduce their cannabis use. Following the 2-week treatment course, participants were followed for an additional four weeks to determine whether potential treatment effects on cannabis craving and use were enduring. For each daily visit, the two rTMS treatment sessions were delivered with a 30-min break between them. Cannabis craving was assessed at the beginning of each visit using the MCQ short form (MCQ-SF). All other stimulation protocols and parameters were identical to those reported in their previous study [see above; or (Sahlem et al., 2018)]. Participants also received one session of motivational enhancement therapy per treatment week, and follow-up assessments at weeks 2 and 4 post-treatment completion.

Of 9 enrolled participants, 6 dropped out prior to treatment completion. Aside from 1 participant dropping out due to treatment-related headaches, no adverse events were reported. For the 3 study completers, the 2-week course of accelerated HF-rTMS treatment to the L DLPFC decreased both cannabis craving (mean MCQ-SF score decrease of ~16 points; from 50.3 ± 7.1 pre-treatment to 34.0 ± 26.2 post-treatment), and cannabis use (mean weekly use measured via the TimeLine Follow-Back decreased by roughly half immediately following treatment and was sustained for all subsequent follow-up weeks).

Discussion of Limitations and Contributions.

Sahlem & colleagues (2020) acknowledge that there were several critical limitations of their case report study, including the open-label design, small sample size, and low retention rate. They pointed out that their retention rate compared poorly to rTMS treatment studies in other SUD populations, which usually range from 56% to 100%, or even in comparison to other CUD treatment trials, which typically range from 36% to 71% (Sherman & McRae-Clark, 2016). The authors propose that their low retention may reflect the accelerated treatment schedule (two treatment sessions per day, each day per study week), rather than the rTMS, per se. Notably, 75% of the non-completers acknowledged that scheduling issues were the main barrier to their study completion. Furthermore, the researchers initiated a new study investigating rTMS treatment in a nearly identical study population, but with only two treatment visits per week (Clinicaltrials.gov: NCT03144232). While still ongoing, the authors report that the less time-intensive trial, so far, has a better retention rate (63%). They postulate that individuals with CUD may have particular difficulty with time-intensive interventions, often being more high-functioning than individuals from other SUD groups, which tends to correspond to a higher likelihood of being employed (which can drastically decrease scheduling availability), and to being more ambivalent about engaging in treatment, overall.

In all, while no firm conclusions can be drawn due to the aforementioned limitations, their study provides the first possible evidence that L DLPFC HF-rTMS may have a beneficial impact on cannabis craving and use in individuals with CUD, making it worthy of continued examination as a novel brain-based treatment approach.

3. Prashad et al., 2019. Mitigating Aberrant Exteroceptive Processing in CUD thru Posterior Cingulate cortex/Precuneus HF-Deep rTMS.

Prashad & colleagues (Prashad et al., 2019) were the first to investigate modulating the posterior cingulate cortex (PCC) and precuneus as a potential NIBS target for treatment of CUD. The PCC, which is the posterior-most region of the cingulate gyrus, is a medial limbic structure that is surrounded by and contiguous with the precuneus, which is the medial portion of the superior parietal lobule (SPL) (Fox et al., 2005; Gusnard et al., 2001). The PCC/precuneus together serve as the functional hub of the default mode network (DMN) (Fox et al., 2005; Gusnard et al., 2001), a large-scale network of brain regions comprising the PCC/precuneus, dorsomedial prefrontal cortex (DMPFC), and bilateral inferior parietal lobule (IPL) (Fox et al., 2005; Gusnard et al., 2001; Raichle et al., 2001; Utevsky et al., 2014). In healthy individuals, the DMN is reliably more active during resting wakefulness, such as when conscious but not engaged in any specific mental task, e.g., daydreaming; self-referential rumination, envisioning the future, thinking about one’s own traits, emotional states or those of others—all thought processes occurring without an explicitly established goal in mind. DMN function typically involves “deactivation” or comparatively less activity when engaged in externally directed, goal-based task performance (Fox et al., 2005; Gusnard et al., 2001) which is believed to allow for re-allocation of neural processing resources to task-active functional networks engaged in service of goal attainment (Fox et al., 2005; Gusnard et al., 2001).

The PCC/precuneus is implicated in stimulus-driven (automatic) exteroceptive processing, which is the continuous, implicit or passive monitoring (i.e., outside of conscious awareness) of salient external stimuli (Fransson, 2005; Littel et al., 2012; Lou et al., 2010; Luber et al., 2012). Its role within the DMN involves passively monitoring the surrounding environment for salient stimuli, with particular emphasis on surveilling stimuli with potential self-relevance (Davey et al., 2016; Schilbach et al., 2008; Utevsky et al., 2014). When salient external stimuli (or cues conditioned to be motivationally relevant to the individual) enter into sensory perception, the PCC/precuneus reflexively augments arousal levels, and re-orients attention to bring those stimuli from passive to conscious awareness for additional processing (Corbetta et al., 2008; Davey et al., 2016; Schilbach et al., 2008; Utevsky et al., 2014). Through this function, the PCC/precuneus, and the DMN as a whole, serve as an attentional resource dedicated to the self-referential processing needed to dynamically direct and maintain self-motivated behavior (Corbetta et al., 2008; Gusnard et al., 2001; Raichle et al., 2001; Uddin, 2015).

In people who use substances chronically, however, exteroceptive processing can be disproportionately heightened (i.e., excessive attentional resources are both implicitly and explicitly allocated to the continuous monitoring and reflexive orientation to potentially self-relevant external stimuli, such as drug cues) (Harris et al., 2018; Littel et al., 2012; Littel & Franken, 2011). Ample research has illustrated the contribution of drug cue-dominated attention in the increased vigilance, stronger attentional capture, and heightened motivational relevance of conditioned drug stimuli in long-term, heavy substance users, processes often referred to as a drug cue attentional bias and/or drug cue reactivity (Harris et al., 2018; Harris et al., 2016; Waters et al., 2003). These processes exhibit bidirectional moderating relationships with drug cue-evoked craving, both of which can trigger drug-seeking and -use behaviors and precipitate relapse in abstinent individuals (Cadet et al., 2014; Littel et al., 2012; Paulus et al., 2013; Siegel, 2005; Waters et al., 2003). Neuroimaging studies have revealed increased activity in the PCC/precuneus during drug cue exposure across a range of SUD groups (Brody et al., 2007; Claus et al., 2013; Grant et al., 1996; McBride et al., 2006; McClernon et al., 2009; Tapert et al., 2003), including heavy cannabis users (Feldstein Ewing et al., 2012; Filbey & Dunlop, 2014; Filbey et al., 2016), with elevated PCC/precuneus activation in response to cannabis cue exposure as evidence of enhanced drug cue reactivity and reward-motivated attentional bias to cannabis stimuli (Filbey et al., 2016).

In this study, Prashad & colleagues used rTMS to modulate PCC/precuneus activity in 10 people who smoke cannabis and 10 non-using healthy controls, with the goal of normalizing PCC/precuneus function in the heavy cannabis users, i.e., by reducing heightened salience and reactivity to external self-relevant and cannabis-related cues. The authors assessed rTMS-induced changes in neural response to personalized self-relevant and cannabis-related visual stimuli using event-related potential (ERP) recording, an EEG-based neuroimaging approach. Given that their aim was to stimulate the PCC/precuneus, which is a relatively deep, medial brain structure, they indicated that use of a traditional figure-8 TMS coil design, which has a distinctly more focal and shallow cortical stimulation field and depth (~2 cm), was ill-suited (Deng et al., 2013; Deng et al., 2014). Thus, the authors used the double-cone style coil, a more advanced TMS coil design enabling deeper, albeit broader, stimulation than the figure-8 coil (Hayward et al., 2007; Vanneste et al., 2012). Before and after rTMS, participants underwent a modified oddball paradigm which evoked neural response to external stimuli, and included personalized self-relevant and cannabis-related stimuli that occurred infrequently and were expected to elicit the P3, as well as the preceding P2 and N2 ERP responses, which are indices of attention to novel, self-relevant stimuli (Gray et al., 2004). The authors hypothesized that they would find a greater P2, N2, and P3 response to self-relevant stimuli during baseline compared to after HF-rTMS in both cannabis users and non-using controls, with no change following LF-rTMS (as this condition served as a negative control). Additionally, they predicted greater response to cannabis-related stimuli during baseline compared to after HF-rTMS in the cannabis group, due to the high salience of these cues in people who use cannabis, but not for controls who do not use cannabis.

The ERP results showed that, at baseline, the cannabis group exhibited greater PCC/precuneus reactivity to self-relevant cues than controls, consistent with a heightened exteroceptive processing. This elevated reactivity to external, self-relevant stimuli in the cannabis group was normalized to control levels following HF-rTMS to the PCC/precuneus. The authors interpreted these findings as: (1) support for aberrant activity in the PCC/precuneus in people who use cannabis; (2) demonstration that aberrant PCC/precuneus activity in cannabis users underlies their excessive salience attribution and reactivity to external self-relevant cues; and (3) demonstration that directly manipulating dysfunctional PCC activity, such as through neuromodulation, may provide a viable new option for mitigating exteroceptive processing abnormalities that contribute to the heightened self-relevant and drug cue-induced attentional bias, and corresponding drug cue-induced craving and seeking behaviors.

Discussion of Limitations and Contributions.

While the authors predicted a greater response to cannabis-related stimuli in people who use cannabis, their ERP results showed a lack of differences between the groups or across rTMS conditions, which was inconsistent with prior literature. The authors attributed this to the drug use history characteristics of their sample; while those in the cannabis group were near-daily cannabis users, most did not meet diagnostic criteria for CUD. They point out that most prior studies revealing significant differences in cannabis cue-related salience reactivity involved samples of daily heavy cannabis users, whereas lighter users have often been more similar to non-using controls (such as in Pope et al., 2001; Bolla et al., 2002). It is important to note, however, that results of the present study were limited by the small sample sizes in both groups (n=10 each). In addition, several demographic differences were found between the groups at baseline (such as, education, age, alcohol use), which may confound study outcomes. Nonetheless, these data provide the first investigation of a novel rTMS stimulation target, the PCC/precuneus, for the potential treatment of CUD, making the replication and extension of these findings in a larger, well-matched sample potentially worthwhile.

III. FUTURE DIRECTIONS OF REPETITIVE TRANSCRANIAL MAGNETIC STIMULATION RESEARCH FOR TREATMENT OF CANNABIS USE DISORDER.

Given the substantial heterogeneity (i.e., study design, participant characteristics, study objectives, outcome measures, etc.) across the rather small number of reports (n=3 studies), the data from the present literature on rTMS for people who use cannabis are too dissimilar to synthesize, and thus it is not yet possible to draw firm conclusions about the clinical validity of rTMS for CUD. However, the present studies reveal the general feasibility and tolerability of several rTMS-based procedures in people who use cannabis and reinforce the worth of pursuing NIBS-based approaches as novel forms of treatment for CUD in future investigations.

Below is a brief breakdown of some of the potential challenges to developing rTMS-based interventions for CUD, as well as some of the critical design decisions (i.e., study parameters and outcome measures) derived from pre-existing literature on NIBS for SUDs which should be considered when developing future rTMS treatment studies in cannabis users.

Limitations & Challenges Specific to Cannabis Use and Cannabis Use Disorder.

There are a number of factors associated with the neuropathophysiological correlates of chronic cannabis use which may confound the success of rTMS-based treatments for CUD.

Functional & Structural Neural Deficits.

For instance, ample prior research has demonstrated the considerable influence of state-dependent neural effects on rTMS outcomes, wherein the magnitude and direction of treatment response often vary as a function of the initial functional state and structural integrity of the intended brain targets (Dunlop et al., 2015; James et al., 2017; Kearney-Ramos et al., 2018b; 2019; Nicolo et al., 2015; Salomons et al., 2014; Silvanto & Pascual-Leone, 2012; Weigand et al., 2018). The precise impact that neuropathophysiological sequelae of any primary and/or co-occurring disease conditions may have on intrinsic brain states or on the brain’s capacity for rTMS-induced plasticity, is of unknown, or at best, poorly understood, consequence and should thus be a critical consideration for future therapeutic neuromodulation research in SUDs, and psychiatry in general (Iqbal et al., 2019; Lefaucheur et al., 2014; Rossi et al., 2009). For instance, interactions between disease states and stimulation may lead to illness-specific side effects, such as the increased risk of mania induction in patients with bipolar disorder, the increased risk of seizure induction in patients with cortical lesions, or the influence of brain atrophy on electrical current distribution in patients with prior traumatic brain injuries or co-morbid neurodegenerative conditions (Iriarte & George, 2018; Rossi et al., 2009). Relatedly, frequent, heavy cannabis use is known to exacerbate the risk of psychosis or schizophrenia in individuals with predispositions to these conditions (Bossong et al., 2014; Di Forti et al., 2015; 2019), though the psychotomimetic effects of cannabis exposure have also been shown in otherwise healthy individuals with no pre-existing risk factors (Bhattacharyya et al., 2012; Marconi et al. 2016). As such, the potential for mania-related side effects of rTMS represents an important potential contraindication to rTMS therapy for individuals with CUD, which should be studied further in future research, and ultimately, weighed carefully against the potential benefits of this treatment option relative to other modalities.

With regard to cannabis use and its role in functional and structural brain alterations, it is important to note that while the data are largely conflicting and many confounds remain to be addressed in the literature (opposing findings reviewed or highlighted recently in Chye et al., 2020; Colizzi et al., 2020; Ferland & Hurd, 2020; Figueiredo et al., 2020), it has generally been recognized that long-term heavy cannabis exposure is associated with functional and structural deficits across several critical brain regions (Ferland & Hurd, 2020; Zehra et al., 2018), particularly those enriched with cannabinoid type 1 receptors (CB₁Rs), such as the hippocampus, amygdala, striatum, PFC, insula, cingulate gyrus, and cerebellum (Battistella et al., 2014; Blest-Hopley et al., 2020; Bloomfield et al., 2019; Lorenzetti et al., 2016; 2019; Zehra et al., 2018). In fact, some of the hallmark cognitive-behavioral symptoms of CUD involve memory, attention, and executive function impairments (Blest-Hopley et al., 2020; Ganzer et al., 2016; Zehra et al., 2018) which have been linked to varying levels of gray matter volume and density reductions and thinning cortical thickness (Battistella et al., 2014; Chye et al., 2020; Cousijn et al., 2012; Yucel et al., 2008), as well as changes in brain activity (Batalla et al., 2013; Blest-Hopley et al., 2020; Ganzer et al., 2016; Martin-Santos et al., 2010) across these areas.

Furthermore, functional and structural connectivity data acquired from a wide range of neuroimaging modalities have shown that chronic cannabis use is associated with patterns of local and global dysconnectivity in key cognitive-behavioral networks encompassing many of these cannabinoid-rich regions (Cheng et al., 2014; Filbey & Dunlop, 2014; Manza et al., 2018; Pujol et al., 2014). These neural deficits have direct implications for the potential efficacy of therapeutic neuromodulation in patients with CUD, given that aberrant functional and structural integrity and connectivity within and between key neural circuit nodes is a known barrier to adequate modulation of large-scale brain networks via rTMS-based techniques (Diekhoff-Krebs et al., 2017; Groppa et al., 2012; Klöppel et al., 2008). This is because from a network-based perspective, coordinated neural activity (i.e., resting state activity or task-related network engagement) must rely upon intact structural connections (Philip et al., 2014). Thus, pre-existing disease-related network dysconnectivity can greatly impact the transsynaptic signal propagation mechanisms by which rTMS-based techniques are purported to influence distal cortical and subcortical nodes downstream from the immediate stimulation targets, and substantially limit or preclude the clinical benefits of rTMS on large-scale cognitive-behavioral networks in that group (Opitz et al., 2016; Rossi et al., 2009).

Recently, the impact of structural integrity on rTMS-induced circuit modulation was directly illustrated in a SUD population in a study by Kearney-Ramos et al. (2018b). Using interleaved TMS/fMRI (Bestmann et al., 2008; Paus, 2005), the authors examined the TMS-evoked responses in the frontal-striatal reward-motivation circuit in chronic cocaine users. Interleaved TMS/fMRI is a multimodal imaging technique that involves the successive delivery of single-pulse TMS during the brief intervals between fMRI acquisition of individual functional brain volumes, thus enabling measurement of the real-time causal effects of TMS on functional activity in brain networks at both the direct stimulation target and indirect downstream subcortical network nodes (Bestmann et al., 2008; 2000a,b; Sack et al., 2002). In this study, 49 individuals with cocaine use disorder underwent interleaved TMS/fMRI imaging using an MRI-compatible TMS coil positioned over the ventromedial prefrontal cortex (VMPFC), or standard frontal pole EEG 10-20 coordinate, FP1 (Jasper, 1958). Blood oxygen level-dependent (BOLD) fMRI signal changes following active vs. sham single-pulse TMS were calculated and compared for the area directly beneath the coil (i.e., VMPFC) and several subcortical regions exhibiting afferent and efferent connectivity with the VMPFC (i.e., bilateral limbic and striatal regions). Diffusion tensor imaging (DTI) and voxel-based morphometry (VBM) were used to quantify white matter integrity and gray matter volume (GMV) within and between the functional nodes, respectively, followed by calculating the relationships between these neural architecture measures and TMS-evoked BOLD response.

The authors found that the size of the TMS-evoked BOLD response in the VMPFC stimulation target and downstream subcortical nodes was significantly positively correlated with GMV in the VMPFC, as well as white matter integrity (given by fractional anisotropy) in the structural tracts connecting the VMPFC to the downstream limbic and striatal regions, such that cocaine users with greater gray matter and white matter atrophy showed poorer TMS responses. This provided the first clear demonstration, in a relatively large sample of individuals with SUD, that the ability of the TMS-evoked functional effects to propagate from the superficial cortical target to distal, downstream subcortical network nodes is significantly dependent upon intact structural integrity in the circuit of interest.

Hanlon et al. (2019a) later corroborated these findings by demonstrating that, similarly, in people with alcohol use disorder, over half of the individual differences in response to active rTMS could be attributed to inherent variability in several neural architecture features, including STC distance, GMV in the cortical stimulation target, and white matter tract integrity connecting the stimulated cortical target and its downstream nodes.

Consequently, when considering the findings of these former studies, they suggest that disruptions in structural integrity within and along tracts connecting critical nodes of large-scale functional networks have an impact on functional connectivity—both of which are dysconnectivity features inherent to many neuropsychiatric disorders (Menon, 2011; Sutherland et al., 2012; Zhang & Volkow, 2019)—and are likely to reduce the ability of rTMS-induced effects to propagate to downstream targets, which may ultimately hinder how effectively rTMS is able to produce modulatory effects on plasticity and activity in the networks as a whole (Groppa et al., 2012; Klöppel et al., 2008; Nahas et al., 2004; Philip et al., 2020). These data are consistent with the impact of functional and structural integrity on rTMS outcomes in clinical populations with neurodegenerative disorders (Anderkova et al., 2015; Chervyakov et al., 2015a), traumatic brain injuries (Diekhoff-Krebs et al., 2017; Nouri & Cramer, 2011; Riley et al., 2011), and aging-related neural atrophy (Iriarte & George, 2018; Nahas et al., 2004).

Therefore, while disagreement persists in the cannabis field with regard to the existence and/or direction of causality of structural and functional brain impairments in CUD (an obvious topic independently warranting further examination; Ferland & Hurd, 2020), the potential limiting impact any such pre-existing deficits might have on the clinical benefits of rTMS as a therapy for CUD should, at the very least, elevate the clinical significance of further evaluating these factors in the design and interpretation of outcomes in future rTMS treatment studies in this population. One avenue for future work would be to incorporate measures of structural architecture, such as white matter integrity, GMV, and STC distance, into research and clinical protocols, such as for optimization of patient-specific cortical targeting and stimulation dosing, in order to personalize and maximize the clinical utility of rTMS-based treatment models in people who use cannabis.

Altered Synaptic Plasticity Induction.

A related correlate of long-term exposure to cannabis that might impact rTMS treatment response involves the sustained disruption of synaptic plasticity, as has been demonstrated preclinically (Fratta & Fattore, 2013; Hoffman & Lupica, 2013), and in humans (Fitzgerald et al., 2009). A recent study by Martin-Rodriguez et al. (2020) extended upon this by showing that long-term cannabis exposure has deleterious effects on the capacity for rTMS-induced cortical plasticity induction. In this study, 45 young adult men underwent two sessions of TBS in which either cTBS (inhibitory) or iTBS (excitatory) was applied over the primary motor cortex, followed by measurement of resulting changes in contralateral MEPs relative to baseline. Of the 45 young men, 30 had used cannabis daily over the previous 6 months (n=15 of whom met DSM-5 criteria for CUD; n=15 of whom did not have CUD diagnosis) and 15 were demographically matched non-using controls. Results showed that the non-CUD men who used cannabis and the non-using controls exhibited significant MEP inhibition following cTBS, as expected, while inhibition was not seen in the cannabis users with CUD. Furthermore, when MEP outcomes were correlated with self-reported measures of problematic cannabis use, the greatest reductions in motor cortical plasticity (aka reduced capacity for plasticity induction) were seen in participants with the most severe cannabis use problems. Conversely, no group differences were identified for iTBS-induced (excitatory) cortical plasticity outcomes. While further research involving larger sample sizes will be needed to corroborate these findings, this study provides the first evidence that CUD, and more specifically, heavy cannabis use associated with severe cannabis use-related problems, may contribute to impairments in inhibitory cortical plasticity. These data are consistent with previous work revealing cortical inhibition deficits (Goodman et al., 2017) and reductions in GABA_A-mediated inhibition for subjects with daily chronic cannabis use that also corresponded with plasma THC levels (Fitzgerald et al., 2009).

That said, the Martin-Rodriguez (2020) data also seem to indicate that plasticity deficits associated with chronic heavy cannabis use are limited to LTD-like (or inhibitory) plasticity. If that holds up in future larger studies replicating and expanding upon these findings, then when considering the potential efficacy of rTMS-based interventions, these results may be rather encouraging for people who use cannabis but do not meet criteria for CUD, as those individuals may respond well to therapeutic approaches involving inhibitory rTMS (such as, protocols explored in Hanlon et al., 2015; 2017; Kearney-Ramos et al., 2018a; 2019). Conversely, for people with more severe cannabis use, such as those with CUD who endorse cannabis-related problems, deficits in cortical inhibition induction may preclude adequate response to inhibitory stimulation protocols, perhaps making alternative treatment strategies more suitable for this group. An obvious alternative would be to focus on excitatory stimulation models (such as the majority of rTMS-based protocols tested for treatment across SUD groups; see Ekhtiari et al., 2019 for recent comprehensive review).

Additionally, methods which alter the threshold for plasticity induction through manipulation of physiology or various design parameters could be incorporated to enhance outcomes; for instance, inclusion of pharmacologic agents which favor cortical inhibition mechanisms (Schwenkreis et al., 1999; 2000; Ziemann, 2004; 2013; Ziemann et al., 1998), or via priming protocols which promote state-dependent manipulation of baseline cortical excitability to facilitate induction of cortical plasticity of a desired magnitude and direction (Rossi et al., 2009; 2021; also see Hoogendam et al., 2010; Karabanov et al., 2015; Suppa et al., 2016 reviews which discuss several of these approaches).

Notably, a recent review by Hanlon et al. (2018) summarized existing evidence of disruptions in cortical excitability and other neurophysiological properties in several SUD groups, including cocaine, alcohol, and nicotine users, and highlighted the impact these neural alterations had on their responses to rTMS. Specifically, several reports have revealed heightened MTs (Hanlon et al., 2015) and reduced MEPs in people who use cocaine chronically relative to non-using controls (Boutros et al., 2001, 2005; Sundaresan et al., 2007; Gjini et al., 2012). These lower MEP values in the cocaine group were interpreted as a reflection of lower cortical excitability (Ziemann et al., 1998; Di Lazzaro et al., 2008), which, along with higher MT levels, is a neurophysiological property known to negatively correlate with rTMS plasticity induction and corresponding clinical outcomes (Pellegrini et al., 2018). Furthermore, a TMS/fMRI study from the Hanlon Lab demonstrated that people who use cocaine heavily exhibit frontral-striatal dysconnectivity, as the application of single-pulse TMS to the VMPFC elicited markedly lower reciprocal BOLD responses in downstream ventral striatal afferents in the cocaine group compared to healthy controls (Hanlon et al., 2016).

Impairments in SUD-related functional network connectivity and integrity is known to negatively impact rTMS-induced neuromodulation of clinical brain targets, thus compromising the attainment of clinical benefit from rTMS-based interventions in associated groups. Thus, while relative cortical excitability, MT values, and TMS-evoked BOLD response, and circuit connectivity or integrity have not been comprehensively evaluated within the context of potential efficacy of rTMS-based interventions in people with CUD, the extant evidence in the literature regarding neurobiological correlates of chronic cannabis use and their likely impact on synaptic plasticity induction, provides a foundation for considering the factors that may play a role in the clinical success of NIBS treatments for CUD, and thus highlights the many potential avenues for future research to begin filling these knowledge gaps.

Role of Brain-derived Neurotrophic Factor (BDNF) Deficits in Plasticity Potential.

Long-term cannabis use has also frequently been associated with reductions in neural compounds involved in nerve health, including nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF) (Jockers-Scherübl et al., 2004). BDNF is a neurotrophin that is involved in the genesis, differentiation, survival, and repair of neuronal cells (Binder & Scharfman, 2004; Chao et al., 2006; Cheeran et al., 2008), and is thus a key modulator of neuroplasticity and adaptive processes underlying experience-dependent learning and memory (Egan et al., 2003; Hariri et al., 2003; Ninan et al., 2010; Pezawas et al., 2004; Yamada et al., 2002). Numerous studies have found reduced levels of BDNF in people who regularly use cannabis (D’Souza et al., 2009; Lisano et al., 2019; Miguez et al., 2019), with some indication that reduced BDNF levels may play a role in the learning and memory deficits and general cognitive impairments characteristic of long-term cannabis exposure (D’Souza et al., 2009; Miguez et al., 2019; Yucel et al., 2016). BDNF expression is, in part, regulated by CB1R-mediated activation of signaling pathways which result in the transcription of BDNF mRNA (Berghuis et al., 2005; Derkinderen et al., 2003). Thus, cannabinoid receptors play a regulatory role in modulating the levels of BDNF molecules circulating in the brain and periphery. Animal and PET studies in people who use cannabis heavily have consistently revealed substantial downregulation and desensitization of CB₁Rs after long-term cannabis use (Colizzi et al., 2016; Gonzalez et al., 2005; Hirvonen et al., 2012; Lichtman & Martin, 2005; Weinstein et al., 2016). Consequently, reduced CB₁R function and corresponding dysregulation of CB₁R-mediated BDNF signaling, may represent one of the potential mechanisms by which BDNF expression is blunted in this population (D’Souza et al., 2009).

Relatedly, with regard to the potential success of rTMS-based techniques for the treatment of CUD, it is critical to note that BDNF levels have a direct impact on the potential for rTMS-induced plasticity induction processes (Cheeran et al., 2008; Figurov et al., 1996). Even in healthy individuals, there exists a single nucleotide polymorphism (SNP) producing a valine-to-methionine substitution at codon 66 (Val66Met) in the human BDNF gene which influences human cortical plasticity induction (Egan et al., 2003; Kleim et al., 2006; Ninan et al., 2010; Soliman et al., 2010) and, consequently, neural response to rTMS (Cheeran et al., 2008) and other brain stimulation techniques (Antal et al., 2010; Riddle et al., 2020). In fact, this SNP is relatively common in the general population (65% Val66Val vs. 35% Val66Met in Caucasians), making any of its functional consequences potentially significant to the ultimate efficacy of rTMS treatment in the population at large. Current research has been demonstrated that Met allele carriers have impaired LTP and LTD plasticity potential, as they exhibit markedly blunted, or even absent, responses to both excitatory and inhibitory rTMS, as opposed to their homozygous Val counterparts who exhibit expected plasticity responses (Cheeran et al., 2008; Cirillo et al., 2012; Hwang et al., 2015).

Thus, while not yet directly evaluated in people with CUD, chronic cannabis use-related suppression of BDNF levels may present similar challenges to rTMS-induced plasticity as those caused by general variability in BDNF genotype, thereby further contributing to uncertainty with regard to the potential clinical efficacy of rTMS-based interventions in CUD. In general, illnesses or mitigating conditions which alter the neurochemistry and neurophysiology underlying plasticity induction are likely to block or substantially suppress the neural response to rTMS-based techniques, and thus should be considered when designing future studies evaluating the clinical merits of rTMS for CUD (Rossi et al., 2009). That said, it is also encouraging to note that preclinical (Muller et al., 2000) and human studies (Zanardini et al., 2006; Yukimasa et al., 2006) have shown that rTMS is also capable of enhancing BDNF expression, particularly following longer treatment protocols. Thus, it may be that for long-term cannabis users with BDNF-related plasticity impairments to exhibit successful clinical responses to rTMS, the number of treatment sessions must be extended beyond the standard 4-6 weeks of daily sessions recommended under current clinical guidelines. Perhaps this added time and exposure to rTMS will serve to facilitate the intended neuromodulatory outcomes by providing sufficient opportunity for the underlying mechanisms leading to BDNF normalization to be activated, thereby surmounting key barriers to plasticity. However, data also exists calling into question whether rTMS produces its clinical effects through alterations in BDNF, although such reports have so far been for other patient populations (Brunoni et al. 2015; Jiang & He, 2019), and not yet assessed in individuals with CUD or other SUDs.

As such, while it is helpful to consider the unanswered questions and ways to alter the design of rTMS protocols to meet the potential needs of individuals with CUD, until future research is conducted to provide further and stronger preliminary evidence of rTMS efficacy in people who use cannabis, it is difficult to definitively predict whether and which factors will limit rTMS outcomes in this population, which also precludes the value of developing concrete courses of action for maximizing as of yet indeterminate treatment success. Put more succinctly, there are a lot of unknowns, which is all the more reason continued high-quality research is critically needed in this area.

Future Design Considerations for rTMS Studies in Cannabis Use Disorder.

A. rTMS Brain Targets.

Dorsolateral Prefrontal Cortex (DLPFC).

The majority of investigations into NIBS for SUDs, including those focused on either rTMS or tDCS designs, have focused on targeting the DLPFC (either the L or R DLPFC, or both simultaneously) making it the most widely studied NIBS target for SUDs, to date (Ekhtiari et al., 2019; Jansen et al., 2013; Song et al., 2019). First, as the central hub of the executive control network which functions in a regulatory role across a majority of cognitive, behavioral, and emotional processes, the DLPFC is an obvious neurocognitive target for NIBS-based treatment approaches as it has the greatest potential for far-reaching therapeutic impact (Diehl et al., 2018; Ekhtiari et al., 2019; Goldstein & Volkow, 2011). Second, the DLPFC was the first neural treatment target to amass sufficiently compelling evidence to achieve the world’s first FDA clearance for a neuromodulation-based treatment approach in psychiatry (i.e., rTMS for TRD which was FDA cleared in 2008; O’Reardon et al., 2007; Berlim et al., 2014). Since then, the DLPFC has been one of the most common preliminary targets investigated as a proof-of-principle and feasibility litmus test for neuromodulation treatment in a wide variety of neurological and psychiatric disease populations (Berlim et al., 2014; Dunlop et al., 2017; Fox et al., 2014; Hanlon et al., 2017). In addition, given its relatively superficial cortical location, it is one of the few core SUD-related brain regions with easy accessibility to the widest assortment of NIBS stimulation devices, including those limited by shallow stimulation depths.

It is noteworthy to also highlight the recent ground-breaking first FDA clearance of a NIBS-based treatment approach for application in a SUD group (BrainsWay Press Release, August 24, 2020); namely, involving HF-rTMS for treatment in tobacco cigarette smokers. In the principal study leading to this FDA clearance (Dinur-Klein et al., in press), delivery of 18 near-daily sessions of HF deep rTMS to the bilateral lateral PFC and insula via the BrainsWay H4-coil was shown to aid in tobacco cigarette smoking cessation by improving the continuous quit rate, reducing cigarette craving, and decreasing the average number of cigarettes smoked per week. While the broad swath of cortex stimulated by the H4-coil is not limited solely to the DLPFC, instead also involving stimulation of the bilateral insula and other contiguous parts of the lateral PFC, the L and R DLPFC comprise a substantial proportion of the targeted brain tissue, and thus contribute greatly to the purported neurobehavioral success of this technique. Thus, this newly approved approach contributes further to the existing literature supporting the value of the DLPFC as a viable treatment target, while it also highlights the potential merits of using a NIBS-based treatment protocol which broadly modulates multiple SUD-related neural circuit nodes in order to produce or enhance clinically meaningful outcomes.

Given the growing body of high-quality evidence supporting the success of DLPFC-targeted NIBS interventions across SUD groups (Dinur-Klein et al., in press; Ekhtiari et al., 2019; Jansen et al., 2013; Song et al., 2019), the DLPFC remains a compelling treatment target for CUD that deserves continued investigation in future research studies focused on the development of novel, brain-based therapeutic tools for problematic cannabis use.

Other Brain Regions.

There are, however, a number of other logical NIBS brain targets (Ekhtiari et al., 2019), some of which have shown substantial promise across a growing number of SUD groups [e.g., the medial prefrontal cortex/frontal pole (Hanlon, 2019; Kearney-Ramos et al., 2019), anterior cingulate cortex (Martinez et al., 2018), insula (Ibrahim et al., 2019), etc.], as well as some which have newly emerged as potentially of interest [e.g., PCC/precuneus, as investigated for the first time by Prashad et al. (2019); discussed in the present review]. However, evidence for these other novel targets has not yet reached the level of support in SUD groups as has been demonstrated for the DLPFC; thus, additional research is critical to ascertaining the potential clinical relevance of these alternative targets for treatment in people who use cannabis.

B. rTMS Clinical Specificity.

Biomarker-based Validation of rTMS Clinical Endpoints.

An increasingly important proof-of-principle feature of many NIBS treatment development studies is the use of fMRI or other neurorecording tools (such as, EEG, PET, TMS/fMRI, etc.) to demonstrate treatment target engagement by characterizing the ability of NIBS to strategically manipulate the intended neural biomarkers (Caulfield & George, 2018; Dunlop et al., 2017; Fox et al., 2012; 2013; Hanlon et al., 2013; 2016; James et al., 2017; Siebner et al., 2009; Weigand et al., 2018). Furthermore, these neural biomarkers can facilitate a clearer understanding of the neurobiological mechanisms underlying NIBS treatment effects on CUD-related clinical endpoints (e.g., by correlating brain changes to changes in biologically determined drug use, self-reported drug use, craving, cue reactivity, cognitive and behavioral markers of SUD-related problems, etc.).

For instance, despite there being a lack of anti-craving effect after single-session active rTMS for both Sahlem et al. (2018) and Prashad et al. (2019), it could still be of added value to know whether Sahlem et al.’s HF-rTMS protocol was capable of engaging and modulating activity in the stimulation target (L DLPFC) and its associated circuitry in their group of cannabis smokers by demonstrating rTMS-specific changes in either fMRI or EEG-based recordings as evidence of successful neural outcomes. The inclusion of biomarker-based outcome measures for the validation of clinical target engagement and elucidation of the neurobiological mechanisms underlying NIBS treatment efficacy are increasingly being recognized as best practice for well-designed NIBS-based clinical trials (Dunlop et al., 2017; Ekhtiari et al., 2016; 2019; Fox et al., 2012; Hanlon et al., 2016; James et al., 2017; Kearney-Ramos et al., 2019; Liston et al., 2014; Medaglia et al., 2020; Siebner et al., 2009).

The pursuit of clinical neurobiomarker characterization in NIBS research will serve as a critical step toward developing and incorporating personalized medicine frameworks into NIBS intervention models (Cocchi & Zalesky, 2018; Ekhtiari et al., 2016; Verdejo-Garcia et al., 2019). Ideally, such innovations would facilitate the integration of clinical, demographic, and brain-based tools for “deep phenotyping” of individual patients, enabling advancements in disease classification, treatment selection, and outcome prediction (e.g., informed diagnosis, treatment prognostication, monitoring, and adjustment) (Gordon et al., 2017; Medaglia et al., 2020; Philip et al., 2014; Poldrack et al., 2015; Singh & Rose, 2009). However, for these data rich approaches to work, they must be informed by empirically validated biomarkers reflecting the fundamental mechanisms contributing to and/or predicting individual differences in treatment success or failure (Cocchi & Zalesky, 2018; Dunlop et al., 2017; Hanlon et al., 2016; Hawco et al., 2017; James et al., 2017; Liston et al., 2014; Medaglia et al., 2020; Nicolo et al., 2015; Philip et al., 2014; 2018; Rossi et al., 2021).

Precision-oriented Cortical Targeting.

The clinical purpose of NIBS in psychiatry is to normalize aberrant brain function corresponding to psychopathology. Thus, NIBS-based clinical outcomes fundamentally depend on accurate and reliable coil placement during stimulation at the intended neurocognitive targets, and even minor deviations in coil positioning can lead to large variations in the underlying brain regions that are stimulated (Brasil-Neto et al., 1992; Fuhr et al., 1991; Herbsman et al., 2009; Kraus et al., 1993). Mounting evidence has established the advantages of incorporating personalized cortical targeting and coil positioning techniques during rTMS administration using state-of-the-art (f)MRI-guided neuronavigation procedures (Bashir et al., 2011; Ekhtiari et al., 2019; Kim et al., 2014; Medaglia et al., 2020; Rossi et al., 2021; Rusjan et al., 2010; Sack et al., 2009; Sparing et al., 2008). Neuronavigation systems are auxiliary devices that are used in conjunction with pre-existing rTMS setups to facilitate high-precision stereotactic positioning and orientation of TMS coils during stimulation sessions, using targets tailored to an individual’s own anatomical or functional neuroanatomy (i.e., as derived from their baseline functional and/or structural MRI images). Personalized positioning using their own brain anatomy can then be established, monitored, and adjusted in real-time to maintain target accuracy and maximize full dose delivery of stimulation to intended neural treatment foci, in addition to allowing for reproducible targeting over separate visits across multi-session courses of treatment (Bashir et al., 2011; Medaglia et al., 2020; Schönfeldt-Lecuona et al., 2005).

Further, MRI-guided cortical targeting should be considered in individuals with known or potential structural abnormalities (such as, those with past traumatic brain injuries, stroke histories, aging-related atrophy, or other diseases known to impact structural integrity, including SUDs), as using their unique neuroanatomy to guide coil placement can help to identify optimal hot spots (Ahdab et al., 2010; Ameli et al., 2009; Bradfield et al., 2012; Taylor et al., 2018), while circumventing lesion locations (Gugino et al., 2001; Julkunen et al., 2009). This is because accidental stimulation of brain lesions may induce electrical currents that follow unpredictable paths through damaged cortical tissue and cerebrospinal fluid, running the risk of producing undesirable or even unsafe psychological (e.g., confusion or memory deficits) or behavioral (e.g., motor impairment) side effects, including seizures (Sack et al., 2009). Thus, neuronavigation provides a personalized approach to cortical targeting which improves safety without sacrificing clinical benefit (Caulfield et al., 2017; Nahas et al., 2004; Rossi et al., 2021; Wagner et al., 2008). The safety advantages are also apparent in the use of personalized positioning for adjustment of stimulation intensity as a function of differences in STC distance, since the increased accuracy of personalized targets usually results in better stimulus response and, consequently, estimate significantly lower stimulation doses (Gugino et al., 2001; Schmidt et al., 2009). Lower stimulation doses have the obvious advantage of reducing the potential for seizure induction or other adverse events (Deng et al., 2014; Rossi et al., 2021). From a research design perspective, it has also been suggested that enhanced precision of coil placement can improve the effect sizes of rTMS treatment outcomes, thereby decreasing the number of participants needed for a given study or clinical trial (Sack et al., 2009).

That said, while there are many critical advantages to MRI-guided techniques, particularly for ensuring rigor and reproducibility in highly controlled research study designs or adding clinical precautions in complex patient populations, much evidence supports the suitability of the more traditional cortical targeting methods (such as BeamF3; clinicalresearcher.org/software.htm; Beam et al., 2009), as they are generally able to provide reasonable approximations to the advanced techniques in a majority of participants (Moghtadaei et al., 2015; Nikolin et al., 2019). This is significant because it supports the continued merits of conventional techniques, particularly under broader circumstances, such as if/when rTMS interventions expand beyond academic and specialized medical research environments and into SUD treatment facilities in community clinics or rehab centers, which often do not readily have access to the expensive, complex neuroimaging technology or expertise necessary for (f)MRI-guided approaches (Brunoni et al., 2019; Nikolin et al., 2019).

Cue Exposure/Induction for Neurocognitive Specificity.

Another gold-standard brain stimulation practice derived from prior NIBS research, involves the inclusion of an active priming procedure (such as a cue induction/cue exposure paradigm) which has consistently been shown to substantially enhance the neurobehavioral outcomes of NIBS-based interventions. This process works by enhancing the specificity of the induced neuroplastic changes, primarily through activating and behaviorally engaging the precise target circuit proximal to the time of modulation (Dinur-Klein et al., 2014; Hanlon et al., 2017; Hoogendam et al., 2010; Kearney-Ramos et al., 2018a).

Specifically, during rTMS in cannabis users, specificity of neural circuit engagement and maximal manipulation of target circuit activity can be facilitated through exposure to multisensory cannabis cues which involve visual, tactile, and/or olfactory sensory stimulation [such as those incorporated by (Sahlem et al., 2018; 2020)], and/or narrative cue exposure [such as those involving personalized narratives recounted by each individual regarding their own drug use contexts, co-users, and sensory memories which elicit craving and drug-seeking and, thereby, actively engage relevant drug use circuitry].

IV. CONCLUSIONS

The present review provides very early encouraging but not conclusive evidence that rTMS-based therapies are feasible and tolerable treatment avenues for individuals with CUD. The public health need for novel CUD interventions has increased in parallel with shifting social, political, and legal circumstances surrounding cannabis use. The consequences of these converging factors are evident in the ongoing global increases in cannabis use problems and use disorders, thus elevating the importance of conducting high-quality, empirically supported CUD treatment development trials for rapid clinical translation. While the limited number and heterogeneity of reports examining rTMS for CUD, to date, clearly highlight the need for robust expansion of research investigating the efficacy and optimization of NIBS-based strategies for treating cannabis use behaviors, the growing interest in this area is encouraging and illustrates a clear impetus for establishing clinically meaningful studies, with design parameters and endpoints most relevant to generating valid clinical evidence to move rTMS treatment forward for those with CUD.

TABLE 1.

rTMS studies for treatment of Cannabis use disorder

Author (s)	N	Sample Characteristics				Study Design	Sham Method	Coil Type	Brain Target	Target Localization Method	# of Sessions	Stimulation Intensity (in %rMT)	Stimulation Frequency (in Hz)	Total Pulses per Session	Primary Outcome Measure	Secondary Outcome Measure(s)	Results Summary
		Participants	Sex	Age	Average Baseline Cann use
Sahlem et al., 2018	18	Non-treatment seeking participants with CUD (via MINI)	13 M (81.3%)	26.0 (± 17.9)	1.3 (± 1.3) grams/day; 23.5 (± 4.3) Cann use days/month	Randomized, double-blind, sham-controlled, within-subject crossover design, single-session study	Integrated sham with scalp electrodesa	Figure 8	L DLPFC	F3 EEG Coordinate using Beam F3 Method	2 (1 active; 1 sham)	110	10 Hz	4000	Feasibility (retention rate); bTolerability (% able to tolerate full dose rTMS)	Craving (via MCQ)	16 (three women) completed the trial (89% retained for the three study visits; 2 rTMS sessions). All of the treatment completers tolerated rTMS at full dose without adverse effects. There was not a significant reduction in the total MCQ when participants received active rTMS as compared to sham rTMS.
Sahlem et al., 2020	3	Participants meeting DSM-5 criteria for ≥ moderate CUD who were interested in reducing their Cannabis use	2 M (67%)	38.3 (± 11.2)	31.7 (± 15.1) Cannabis use-sessions/week	Open-label pilot safety and tolerability trial	Integrated sham with scalp electrodesa	Figure 8	L DLPFC	F3 EEG Coordinate using Beam F3 Method	20 over 10 days (2 per day)	120	10 Hz	4000	Feasibility (retention rate); Tolerability (% able to tolerate full dose rTMS)	Craving (via MCQSF)	3 out of 9 enrolled participants completed the full study. No adverse events were reported, except for 1 participant dropping out due to treatment-related headaches. For the 3 study completers who received the full 2-week course of treatment, Cannabis craving and Cannabis use was reduced (respectively, mean MCQ-SF score decrease of ~16 points and mean weekly Cannabis use measured via TLFB decreased by half)
Prashad et al., 2019	20	10 participants who regularly use Cannabis; 10 age- and sex-matched non-using controls	5 M (50%); 5 M (50%);	27.1 (± 4.5); 33.9 (± 14.1)	76.7 (± 18.1) Cann use days in previous 90 days; 0 Cann use days in previous 90 days	Randomized, double-blind, active placebo-controlled, crossover design, single-session study	1 Hz comparison condition as negative control	Double-cone	PCC/precuneus	4-cm posterior to motor strip in parietal cortex	2 (1-10 Hz experimental condition; 1-1 Hz negative control)	80	10 Hz vs. 1 Hz	2000	ERP response to self-relevant cues before and after rTMS	ERP response to Cannabis cues before and after rTMS	At baseline, Cannabis group exhibited heightened PCC/precuneus salience reactivity to external self-relevant stimuli (as reflected by increased ERP amplitude in P3, and faster latencies in the P3, N2, and P2 ERP components in response to self-relevant cues during the Modified Oddball Task, relative to non-using controls), which normalized to control levels following HF-rTMS.

^aSham treatments were delivered using an established electronic sham system (Borckardt et al., 2008). The sham system consists of a coil that mimics the appearance and sound of rTMS, combined with a transcutaneous electrical nerve stimulator (TENS) device which produces a small electrical stimulus delivered to the scalp just below the hairline, mimicking the feeling of active rTMS (Borckardt et al., 2008). Both the participant and the administrator remain blind to stimulation condition.

^bFor tolerability, intensity was ramped up 10% of rMT every train from 60% rMT (200 pulses delivered sub 100%).

Highlights.

• Rise in cannabis use and CUD elevate public health need for effective CUD treatment

• Very early preliminary evidence of feasibility and tolerability of rTMS in CUD

• rTMS clinical efficacy in CUD unclear due to disparate study designs and outcomes

• While evidence promising, limitations highlight need for more high-quality research

• Implementation of gold-standard rTMS research designs may enhance future outcomes