A comprehensive electronic database search was conducted using PsycINFO, MEDLINE, CINAHL, Scopus, and PubMed on June 10, 2022. The search strategy is outlined in Supplementary material and employed three concepts related to: (i) cannabis; (ii) functional neuroimaging; and (iii) reward processing. Medical Subject Headings (MeSH) and keywords (synonyms) were combined with Boolean OR/AND operators for each concept and were searched across the title and abstracts of the returned articles. All full-text articles from the database searches were imported into the reference manager software Covidence (www.covidence.org), and duplicates were removed.

Studies were included if they: (i) examined human participants; (ii) were written in the English language; (iii) were full-text peer-reviewed articles; (iv) assessed a sample of people who consume cannabis, as defined by each study criterion; (v) included a non-cannabis user control group, as defined by each study criterion; (vi) measured brain function during the MID fMRI task, and (vii) compared brain function between cannabis and control groups.

Studies were excluded if they: (i) examined non-human participants; (ii) were published in languages other than English; (iii) were non-peer-reviewed (e.g., conference abstracts only); (iv) were non-empirical (e.g., single case reports, dissertations, editorials, corrigendum, book chapters, letters to the editor, reviews, and meta-analyses); (v) measured brain integrity using neuroimaging techniques other than fMRI, for example: structural magnetic resonance imaging (sMRI), computed tomography scan (CT), electroencephalogram (EEG), positron emission tomography (PET), single-photon emission computerised tomography (SPECT); (vi) studies that included tasks other than the MID fMRI task; (vii) measured brain function during acute cannabis intoxication; (viii) examined a sample of participants who endorsed significant substance use other than cannabis, alcohol, and nicotine at a group level; (ix) included a sample of participants who endorsed a diagnosis of axis I mental health disorders at a group level; or (x) examined a sample with diagnoses of neurological disorders or major medical conditions that affect the central nervous system [e.g., epilepsy, human immunodeficiency virus (HIV)].

The title, abstract, and then full text of all retrieved articles were screened by two researchers (E.B. and S.A.) in accordance with the above inclusion/exclusion criteria to determine eligibility. The final list of studies was compared and resolved by the researchers via consensus; if consensus could not be reached it was resolved by the senior author (V.L.). Additionally, the reference lists of all selected studies were cross-referenced to aid the inclusion of relevant work.

The following data were extracted:

The Joanna Briggs Institute Critical Appraisal Checklist for Analytical Cross-Sectional Studies (Munn et al., 2014) was used to assess the quality of the included studies using eight distinct criteria. The criteria were: (i) Were the criteria for inclusion in the sample clearly defined?; (ii) Were the study subjects and the setting described in detail?; (iii) Was the exposure measured in a valid and reliable way—this criteria was not applicable and therefore removed from the table; (iv) Were objective, standard criteria used for measurement of the condition? (v) Were confounding factors identified? (vi) Were strategies to deal with confounding factors stated? (vii) Were the outcomes measured in a valid and reliable way? and (viii) Was appropriate statistical analysis used? Bias ratings were assessed for each included study based on the criteria above (E.B. and H.T.).

A score for each paper was generated, whereby each criterion was scored either 1 = endorsed, 0 = not endorsed, or 0.5 = partially endorsed. Subsequently, the quality of each paper was rated either high (i.e., a score of ≥8) moderate (i.e., a score between 4 and 7.5), or low (i.e., a score of ≤ 3.5).

Figure 1 illustrates the PRISMA flowchart. The initial database search produced 1,835 articles. After duplicates were removed, 979 titles and abstracts were screened based on the inclusion and exclusion criteria. Of these, 45 articles underwent full-text review, and 36 of them were not eligible for inclusion. Overall, nine studies were included in this systematic review (Nestor et al., 2010, 2020; van Hell et al., 2010; Filbey et al., 2013; Jager et al., 2013; Yip et al., 2014; Enzi et al., 2015; Karoly et al., 2015; Skumlien et al., 2022).

There were largely non-significant group differences in behavioural task performance, including accuracy and reaction times (Table 1). Only one of the nine studies reported that cannabis users were significantly more accurate than controls during loss vs. neutral trials (Nestor et al., 2020). One study found that cannabis users had a trend of slower reaction times than controls, during reward vs. neutral trials (Jager et al., 2013).

This section overviews group differences in brain function during reward, loss, and neutral conditions (Figure 3). Out of 34 contrasts reported, 16 found significant group differences.

This section overviews group differences during the anticipation and receipt of rewards (Tables 2–4, respectively).

This section overviews group differences in brain function during the anticipation and receipt of losses (Table 4).

This section overviews group differences during anticipation and receipt of neutral outcomes (Table 6).

Overall, the most consistently reported finding was altered NAcc activation in two of seven studies during reward anticipation vs. neutral anticipation (van Hell et al., 2010; Karoly et al., 2015), followed by the ventral striatum during loss anticipation vs. neutral anticipation (two of three studies; Nestor et al., 2010; Yip et al., 2014), and the caudate while participants received neutral outcomes (two of two studies; Nestor et al., 2010; Enzi et al., 2015).

The findings from the reviewed literature need to be interpreted in light of methodological limitations. First, the reviewed evidence is cross-sectional, and longitudinal work is required to determine if altered incentive salience predates or follows cannabis use onset. Second, to date, only nine fMRI studies used the MID task to examine brain reward function in people who use cannabis. The low number of studies precluded the running of a meta-analysis, which requires at least 17–20 studies employing an unbiased whole-brain approach (Muller et al., 2018). Instead, in the literature reviewed herein, the most consistently examined contrasts were reported only by five studies (i.e., reward anticipation vs. neutral anticipation) and three studies (e.g., reward feedback vs. neutral feedback; loss anticipation vs. neutral anticipation; and loss feedback) or <2 studies. Therefore, replication work is required to confirm how reward/loss processing plays a role in the neurobiology of cannabis use; and to explore which variables moderate such associations (e.g., presence of CUD/dependence, greater withdrawal, and exposure to cannabis and nicotine). Additionally, given the low number of studies, we could not systematically quantify differences in brain reward function between adolescents and adults. Furthermore, within the existing literature, there was no emerging evidence of consistent patterns of brain reward function between age groups. Moreover, one study which did directly compare adolescents vs. adults, found no differences in brain reward function (Skumlien et al., 2022). As such, future research is required to determine if age moderates group differences in brain reward function.

Third, assessment of CUD/cannabis dependence was lacking in half of the studies. Future work is required to demonstrate if processing non-drug related rewards affect the mesocorticolimbic circuitry differently in CUD/dependence vs. non-dependent use, as shown in other measures of neural integrity in cannabis users (Chye et al., 2019; Lorenzetti et al., 2020; Rossetti et al., 2021), and as postulated by neuroscientific theories of addiction (Robinson and Berridge, 2001; Koob and Volkow, 2016). Finally, assessment of metrics of exposure to cannabis/other substances and in relation to brain function is lacking. This issue precludes the understanding of which mechanisms may drive changes in reward brain function in people who use cannabis. To address this gap, future work should use robust metrics of substance use e.g., THC Unit, iCannToolkit (Freeman and Lorenzetti, 2020; Lorenzetti et al., 2022) and related problems using tools with diagnostic cutoffs (e.g., CUDIT; Myers et al., 2023).

The review focused on a single measure of reward processing i.e., the MID fMRI task. Perhaps, integrating other reward processing tasks, could have included other aspects of reward processing relevant to the neurobiology of cannabis use (e.g., reversal of learning contingencies with a reward learning task). Yet, the MID was the most consistently used fMRI reward processing task to date in cannabis users (and in normative samples; Oldham et al., 2018), therefore the focus on this task enabled the systematic integration of the findings. Future studies should use varied fMRI tasks to create a body of work examining how different facets of reward and loss processing are affected in people who use cannabis. Another limitation of the review is the exclusion of samples with comorbid mental health problems (e.g., schizophrenia, depression). While this approach enables the examination of cannabis-specific effects, the findings cannot be generalised to the most vulnerable people who use cannabis (Hasin and Walsh, 2020).

Overall, there exists largely non-significant evidence of brain alterations in cannabis users compared to controls, examined with the MID fMRI task. A subset of results reporting significant findings consistently identified significantly different striatal function during the anticipation of rewards and losses; and mixed results supporting associations between brain function and chronicity of cannabis use. Replication longitudinal neuroimaging studies of cannabis users are warranted to use robust metrics of substance use/mental health, and in relation to different types of rewards e.g., monetary, cannabis, and natural rewards (e.g., food). Such new evidence is required to identify with precision the neurobiology of reward processing in cannabis users and to enable comparison of the evidence in cannabis users with prominent neuroscientific theories of addiction.