To our knowledge, this is the first resting‐state functional connectivity (rsFC) fMRI study to examine individuals with moderate‐to‐severe CUD who have attempted to cut down or cease cannabis use, while controlling for age, sex and recent alcohol use. Consistent with prominent neuroscientific theories of addiction, CUD was shown to be associated with altered rsFC across striatal and basal ganglia subcortical regions and frontal, occipital and parietal cortical regions compared to controls. Identified regions have been implicated in reward processing, disinhibition and habitual/compulsive substance use (i.e., NAc‐frontal pole/SFG, putamen‐occipito/parietal cortex and pallidum‐occipital/intracalcarine cortex), some of which correlated with the severity of CUD scores and depression symptom scores.

The current study consistently observed greater rsFC in CUD compared to controls, particularly within fronto‐striatal and occipito‐parieto‐striatal networks. The location and direction of the findings are in line with previous work reporting greater rsFC in overlapping networks in people who regularly use cannabis compared to controls (i.e., greater NAc‐ACC cluster rsFC) [22]. The direction of rsFC change may fluctuate as a function of the intoxication cycle. Indeed, greater rsFC has been consistently reported when participants are nonintoxicated [23], whereas lower rsFC has been consistently reported during acute cannabis intoxication [24]. Therefore, chronic intoxication with THC over time may lead to greater connectivity of mesocorticolimbic regions (e.g., between frontal regions and the motor cortex, insula and temporo‐parietal regions) [23].

We found greater putamen‐occipito/parietal and pallidum‐occipital/intracalcarine rsFC in CUD compared to the control group. This pattern of rsFC may underlie habitual cannabis use and altered salience processing observed in CUD [25]. Indeed, both the putamen and pallidum have been implicated in cognitive processes reportedly altered in regular cannabis users, including habit formation/automated behaviour [25] and habitual substance use [25].

The observed correlation between greater pallidum‐occipital cortex rsFC and lower CUDIT‐R scores contributes to the literature, which suggests that rsFC changes may reflect neural adaptations linked to cannabis use severity [12]. These findings align with previous evidence implicating the pallidum and putamen in habit formation and compulsive substance use [25]. This notion is partially supported by the association identified herein between pallidum‐occipital cortex rsFC and CUDIT‐R scores. Similarly, greater activation in striatal and basal ganglia regions has emerged in cannabis users with greater severity of cannabis‐related problems, frequency of use and/or earlier age of onset across task‐based ‘cue‐reactivity’ fMRI research [26]. In contrast, task‐based fMRI studies have also detected decreased activity in the putamen and in occipital regions, in long‐term cannabis users [27]. Occipital (and some parietal) regions support attentional processing/direction towards salient information such as ‘uniquely coloured lines’, or specific‐coloured circles, presented in the participant's visual field [28], which could reflect neuroadaptations in attention/thoughts towards cannabis/salience pathways that persist during rest.

The notion that established rsFC alterations in a cannabis‐using group may be driven in part by the presence/severity of CUD [29] is supported by the negative association identified between pallidum‐occipital cortex rsFC and CUDIT‐R scores. Alterations in such neural structures are thought to underlie emotional processing, habit formation and reward‐based learning. Additionally, this correlation was largely aligned with correlations reported by Aloi et al. [12], who demonstrated a negative association between CUDIT‐R scores and amygdala‐occipital (cuneus) rsFC. Negative correlations between CUDIT‐R scores and rsFC have also been reported for the following pairings: crus‐brain stem and crus‐cerebellar lobule [30] and amygdala‐paracentral/supplementary motor area and amygdala‐temporal pole [12]. Additionally, negative associations established between pallidum‐occipital cortex rsFC and depression scores were also somewhat in keeping with previous research [31, 32]. Depression scores were previously shown to positively correlate with rsFC between frontal regions [31] and between frontal and parietal regions [32]. These findings highlight the importance of controlling for mood‐related symptomology when examining rsFC in cannabis‐using populations. It also suggests that people with comorbidities (i.e., mood disorder) may experience additional neuroadaptations and hence may be an especially vulnerable subgroup of people with a CUD.

We reported greater fronto‐striatal (NAc‐frontal pole/SFG) rsFC in people with a CUD, which may reflect increased engagement of salience pathways sensitive to THC exposure [33]. Indeed, THC increases dopamine synthesis within the NAc [34, 35], which might subsequently affect the function of the NAc, and possibly of interconnected frontal pathways implicated in salience processing. It is acknowledged that we cannot ascertain which region drives the connection or if both/all regions activate in unison, from the identified rsFC alterations. Nevertheless, increased fronto‐striatal rsFC may be a key contributor to cannabinoid reinforcement [36]. Indeed, neural projections from the NAc to the PFC may mediate the experience of ‘wanting’ cannabis and urges to use cannabis [33]. In line with this notion, the NAc plays a key role in assigning value to stimuli [37], whereas frontal regions have been linked to a loss of control over substance use [38]. Therefore, fronto‐striatal regions may be more prominently implicated in more severe forms of CUD, where a loss of control over cannabis use can be a key feature.

Overall, our finding of greater fronto‐striatal connectivity in people with more severe forms of CUD is in keeping with rsFC fMRI literature in cannabis users more broadly (for a review, see Thomson et al. [11]) and in studies in other substance use disorders (SUDs). Specifically, greater frontostriatal rsFC was observed in current, ‘chronic’ cannabis users [17], in cannabis dependent participants who abstained from cannabis for 28 days [22, 39], and in users of different substances, for example, cocaine amongst others [40]. Taken together, the findings suggest that fronto‐striatal rsFC alterations may underpin CUD, although the time course of the changes is unclear as they may predate or follow CUD (or both) and may persist following abstinence from cannabis.

To the authors' knowledge at the time of writing, this is the first fMRI study to examine rsFC in current cannabis users with moderate‐to‐severe CUD who report failed attempts to cut down or quit their consumption. Findings herein lend support for the first time that the neurocircuitry of CUD overlaps with the addiction neurocircuitry implicated in reward/salience processing and compulsive use, as postulated by prominent neuroscientific theories [33, 41], which have largely been based on substances other than cannabis (e.g., cocaine, alcohol and opioids) [41]. Additionally, up until now, established neuroscientific theories of addiction have been largely validated in outdated diagnostic systems (e.g., DSM‐IV), which do not reflect the current diagnostic system relying on the DSM‐5 [42]. Furthermore, we have specifically examined cannabis users to endorse an inability to reduce or cease their use, extending our finals to a larger subset of individuals with CUD [14]. Longitudinal neuroimaging studies (e.g., Adolescent Brain Cognitive Development [ABCD] study; https://abcdstudy.org/) are required to unpack the time course of rsFC in people who develop a CUD. If indeed such changes did predate CUD‐onset, patterns of rsFC alterations identified here could contribute to the mapping of biomarkers for CUD, to identify at‐risk populations in future research, but could challenge/undermine existing evidence for fMRI differences in CUD relative to control groups.

A strength of this study is its robust methodology, including a focus on a well‐defined sample of individuals with moderate‐to‐severe CUD who have attempted to reduce or cease use. By controlling for confounding factors such as age, sex and alcohol use, the findings provide a clearer understanding of the specific neural alterations associated with CUD. First, the CUD and control groups did not significantly differ in many sociodemographic and mental health‐related variables that can also affect rsFC (e.g., age, anxiety, gender, handedness and stress). Second, confounders that showed significant group differences (i.e., past‐30‐day alcohol exposure) and variables that exert a major influence on brain function (i.e., age and sex) were controlled for in all rsFC analyses, minimising their influence on results. Third, participants abstained from cannabis use for at least 12 h prior to scanning; neither ‘duration of abstinence’ nor ‘THC‐COOH: creatinine ng/mL’ correlated with rsFC alterations in the CUD group. Therefore, it was unlikely that acute cannabis effects (i.e., intoxication) confounded results. As well as controlling for potential confounders, another strength of the study was the recruitment of heterogenous participants with a range of depression, anxiety, psychiatric symptoms scores, alcohol and nicotine use, which is representative of a population with CUD [43].

Regarding sample size, it has been posited that brain‐behaviour correlations are adequately replicable in samples ranging from 20 participants [44] to 42 participants [45]. Thus, robust neuroimaging studies with targeted samples with a CUD, such as the current experiment, are required to advance the understanding of the neurobiology of CUD. In the future, multisite consortia studies could be utilised to substantially increase power and demonstrate replicability of findings (e.g., ABCD and the ENIGMA Addiction working group; www.enigmaaddictionconsortium.com).

The altered brain networks in CUD identified herein, particularly pallidum‐occipital cortex rsFC associated with CUD severity, could be targeted by brain‐based interventions for people aiming to gain increased control over, or reduce, their cannabis use. This includes neurofeedback [46], Transcranial Magnetic Stimulation (TMS) [47] and mindfulness‐based interventions [48], known to reduce alterations of these pathways in SUD. Neurofeedback, which entails the provision of ‘real time’ feedback during an fMRI scan, regarding brain activation under certain conditions (established previously during substance cravings for alcohol, tobacco and cocaine [49]), allows for the development of personalised neural targets, which the participant can be trained to self‐regulate [50]. rsFC alterations identified herein could be used to guide neural neurofeedback targets. Furthermore, mindfulness‐based interventions have been linked to alterations of rsFC between brain regions overlapping those implicated in this manuscript and to a reduction in cigarette smoking [48]. Future research could examine if mindfulness interventions targeting core features of CUD mitigate its neural circuitry.

This study is the first fMRI study to examine rsFC in selected ROIs of individuals with moderate‐to‐severe CUD versus controls, with a robust rationale for the ROIs. We demonstrated greater rsFC across subcortical (striatal and basal ganglia) and cortical regions (frontal, occipital and parietal). The changes were partly correlated with the severity of CUD, and they may reflect neuroadaptations in salience pathways that follow the neuroadaptations within the addiction neurocircuitry. Such adaptations in nonintoxicated individuals with CUD may follow documented THC‐induced reductions in rsFC and dopamine release within the NAc [24]—a hypothesis that needs corroboration by experimental studies in CUD before, during and after THC intoxication.

It was thought that greater putamen/pallidum‐occipital/occipito‐parietal rsFC may underlie habitual cannabis use and altered salience processing observed in CUD, in part driven by CUD severity. Furthermore, fronto‐striatal rsFC increases in the CUD group may reflect the engagement of salience pathways sensitive to THC exposure, secondary to the effect of THC on dopamine synthesis within the NAc. Longitudinal neuroimaging studies are required to confirm if identified changes predate or follow CUD (or both), and multisite consortia studies could be instrumental in replicating the findings in individuals with varying CUD severity and in cannabis users without a CUD. These findings lend support to prominent neuroscientific theories of addiction insofar as their application to CUD and are likely representative of cannabis‐using populations with more severe forms of CUD and comorbid depression and anxiety. Future research could utilise the results herein to develop brain‐based interventions for people aiming to change their cannabis use, such as neurofeedback and mindfulness‐based intervention.