A systematic electronic database search was undertaken on the 24 of May 2021 using five databases: MEDLINE, PsycINFO, Web of Science, Embase, and Scopus. The searches encompassed terms related to both “Cannabis” AND “Diffusion-weighted MRI”. Search terms were: (“Diffusion^* OR “white matter” OR white-matter OR DW-MRI OR DTI OR DTI-MRI OR dMRI OR “Fractional Anisotropy” OR Tractography OR Connectome OR Connectomics”) AND (Cannabi^* OR Marijuana^* OR hashish OR marihuana OR kush OR weed). All terms were searched in the title, abstract, keywords, and/or subject headings as appropriate. All study records found in each database were exported into Endnote, and all duplicates were removed. Any additional duplicates found in Covidence were also removed.

The search was rerun on the 7th of December 2022 to identify additional recently published manuscripts.

Inclusion criteria were:

Exclusion criteria were:

All studies were screened using the website Covidence (https://www.covidence.org) at both the title/abstract and full text stages. Screening was conducted by ER; any ambiguity in relation to the inclusion of a study was resolved in communication with VL. Studies were first screened against exclusion and inclusion criteria using titles and abstracts. Full-text articles were then further screened for inclusion in the systematic review. Finally, reference lists of (1) studies that met the inclusion criteria for this review and (2) reviews and meta-analyses on similar topics, were examined to identify any further studies that may have been eligible for inclusion in the current review.

Data extraction was conducted by ER and AC. The following information was extracted from tables, figures, and written summaries from each study. These details were summarized into nine tables. Table 1 displays information on publication characteristics (e.g., first author and year of publication); sample characteristics (e.g., sample size, sex, and age); and cannabis use levels (e.g., age of cannabis use onset and abstinence period). Table 2 outlines key definitions for technical terms regarding diffusion-MRI metrics and analyses used throughout the paper.

Table 3 includes an overall summary of differences in white matter integrity per diffusion-MRI metric in cannabis users compared to non-using controls. In Tables 4–7 we also include information about: the location, significance, and direction of group differences in white matter integrity, and their association with the level of cannabis use, psychopathology symptom scores, cognitive performance, and other variables. We summarized results from group differences and correlations as a function of the examined diffusion-MRI metrics examined. The diffusion-MRI metrics included (1) FA in Table 4; (2) MD, Trace, and apparent diffusion coefficient (ADC) in Table 5; (3) RD in Table 6 and; (4) AD in Table 7. Table 8 overviews results from associations between white matter integrity and levels of cannabis exposure and other variables. Finally, Table 9 contains information relating to studies with longitudinal findings, including follow-up time, group differences, and associations with levels of cannabis use and other key variables.

Other relevant information from each study can be found in Supplementary material. This data comprised: originally reported substance use metrics (Supplementary Figures 1–3), ethnicity and/or race in cannabis users and control groups (Supplementary Table 1); inclusion and exclusion criteria for cannabis use levels in cannabis users and control groups as some studies allowed for specific amounts/past cannabis use in their control groups (Supplementary Table 2); methods used to acquire diffusion-MRI images (Supplementary Table 3), and which variables were matched between groups or controlled for in the analyses (Supplementary Figure 4). All data were summarized by counting the number of studies endorsing specific features, and/or ranges of values and means, where relevant.

ER and AG assessed the risk of bias of the reviewed literature, via the National Institute of Health, National Heart, Lung, and Blood Institute—Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies Tool (http://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools) (Supplementary Table 4). This tool evaluates the risk of bias (i.e., present, or absent) against 14 criteria. Results from the risk of bias assessment are outlined in Section 7 of the Supplementary material.

A total of 14 manuscripts required additional handling and variations to the data extraction protocol. One manuscript (Bava et al., 2010) reported DTI comparisons that had been already presented in a previous paper (Bava et al., 2009). However, as this paper also presented the neurocognitive correlates associated with the white matter microstructural differences reported previously (Bava et al., 2010), additional correlational findings from the more recent study are also reported. As these manuscripts report on the same sample, details of these studies are reported together as necessary.

Two studies utilized data from the human connectome project (HCP) dataset (Orr et al., 2016; Manza et al., 2020), although these studies used different subsets of participants, therefore mitigating any similarities between results and the results from both studies have been reported.

One paper reported that a portion of a larger sample (Jakabek et al., 2016) had been utilized by another previously published paper also included in this review (Zalesky et al., 2012). However, the results of both papers were reported, given that additional participants were utilized in the later paper. Additionally, 4 papers utilized different subsamples from a larger longitudinal study (Bava et al., 2009; Jacobus et al., 2009, 2013a,b). These papers reported on subsets of participants from the same dataset and were also authored by the same or similar groups of authors. Four other papers included in this review were also published by overlapping groups of authors and samples (Yücel et al., 2010; Yucel et al., 2016; Gruber et al., 2011, 2014). Given this similarity in samples used across these research groups, there is a potential for the generalizability of the results to be compromised.

Two studies also incorporated samples with combined binge drinking and cannabis use (Jacobus et al., 2009, 2013a), and 1 other reported on concurrent heavy alcohol and cannabis use (Jacobus et al., 2013b). However, given the high incidence of binge drinking in adolescents and young people, these studies were retained in the review. Finally, 2 papers reported on the same longitudinal cohort of participants, but they separately reported group differences at baseline (Epstein et al., 2014), and then at 18 months follow-up, and also the changes in white matter in both groups over time (Epstein and Kumra, 2015).

Ultimately, the sample consisted of 2,898 participants. Of the studies that reported a sex distribution, this included 1,234 females (43.25%) and 1,620 males (56.75%). Within the total sample, 1,457 persons (562 females and 873 males) were cannabis users with a mean age of 24.2 years (range: 16.6 to 45.0 years) and 1,441 were controls (672 females and 747 males) with a mean age of 24.0 years (range: 16.6 to 45.0 years). Female participants were included in all but four studies that recruited male-only samples (Arnone et al., 2008; Ashtari et al., 2009; Kim et al., 2011; Lichenstein et al., 2022).

Data pertaining to cannabis use levels are presented (means and SDs) in Table 1. Importantly, participants began using cannabis, reported as either age of onset or age of first use, on average at age 15.4 years (range: 13.1 to 18.1 years). Nine studies reported age of ‘regular’ use, which was found to be on average 16.1 years (range: 14.5 to 17.9 years). The average duration of cannabis use amongst participants was 8.2 years (range: 3.4 to 15.7 years).

Across the sample, the average dosage of cannabis was 343.7 cones per month, corresponding to about 115 joints/month or approximately 4 daily joints (range: 76.5 to 763.9 cones per month). This was consumed on an average of 17.2 days per month (range: 3.4 to 25.7 days) or an average of 39.0 occasions per month (range: 11.2 to 81.5 occasions). Total lifetime occasions of cannabis use were reported for 2 studies; 1 study reported 471 occasions (Jacobus et al., 2013b), and lastly, 1 study as a categorical variable ranging from 1 to 5 occasions to 100+ occasions (Orr et al., 2016).

All studies described in this review primarily used 4 different diffusion-MRI metrics of white-matter microstructure, which are described in Table 2. The diffusion-MRI metrics included: FA (27 studies), MD (also known interchangeably as ADC/Trace, 16 studies), RD (11 studies), and AD (10 studies). Novel diffusion-MRI techniques for investigating white matter microstructure were also utilized, and they are also defined in Table 2. In addition, methods of diffusion-MRI analyses that are utilized in the wider literature are also defined in the table.

These diffusion-MRI metrics are derived from the tensor model of the diffusion-MRI and vary in a way that provides specific interpretations of the white matter (Pierpaoli and Basser, 1996; Alexander et al., 2011). FA is highly sensitive to microstructural changes and provides a summary measure of white matter characterization at the microstructural level but does not assign the changes to specific features of the tissue microstructure without further assumptions. Alternatively, MD provides an inverse measure of membrane density and fluid viscosity providing a more biological measure of white matter characterization. Finally, RD and AD provide direct measures of more macrostructural elements of the white matter including axonal density (i.e., RD) and overall axonal caliber not influenced by myelin (i.e., AD). These diffusion-MRI metrics are often complementary to one another and taken together, they can provide an overall metric of the integrity of the brain's white matter.

This section summarizes white-matter differences between cannabis users and controls, as a function of the diffusion-MRI metric used (i.e., FA, MD, RD, and AD). Within each section below, findings are synthesized by brain region and direction of the difference (e.g., lower, higher, and both). A visual summary of the most consistent results is given in Figure 2. Specifically, Figure 2 focuses on the most consistent findings in the literature, which are prevalent for the FA and MD metrics only. As seen in the figure, there are consistent findings of significantly lower FA in cannabis users compared to controls in the SLF and the corpus callosum, followed by the uncinate fasciculus, inferior fronto-occipital fasciculus, internal capsule, and anterior thalamic radiations. MD is consistently found to be significantly greater in cannabis users compared to controls in the corpus callosum.

In addition to the figure, the overall findings for FA, MD, RD, AD, and non-tract-specific metrics (e.g., global network metrics) are summarized in Table 3. As can be seen, results are primarily detected in FA and MD metrics. The following section outlines further detail in relation to the findings from Table 3 and outlines the group differences in (1) FA; (2) MD; (3) RD; (4) AD; and (5) global network metrics. Within each section, there is an outline of the most consistent white matter tracts implicated in cannabis use for each diffusion-MRI metric, followed by a summary of the overall findings (if applicable). Details for consistent results within the literature (i.e., tracts implicated in at least 4 studies for 1 diffusion-MRI metric) are presented below. For tracts implicated in 3 or fewer studies, see Section 5 of the Supplementary material for further descriptions.

As discussed previously, FA is highly sensitive to microstructural changes and provides a summary measure of white matter characterization at the microstructural level (Pierpaoli and Basser, 1996). Overall, 17 of 27 studies found group differences in FA in cannabis users compared to controls (Table 4). For 11 studies, cannabis users had significantly lower FA than controls. Interestingly, 6 studies found higher FA in cannabis users in partially overlapping areas. Of these, 1 study reported higher and lower FA in multiple fibre tracts (Bava et al., 2009). Overall, the consensus of these findings is that cannabis users have lower FA compared to controls, except for in a few studies.

Multiple tracts showed significant group differences in FA. The most consistent findings were within the arcuate/SLF, corpus callosum, and internal capsule, which were shown to be significantly different in cannabis users in at least 4 studies. In addition, there were multiple tracts implicated in 3 or fewer studies that were not as consistent in the literature.

This section will further disentangle which white matter tracts were consistently found to have differences in FA (i.e., in 4 or more studies). This includes the (1) Arcuate/SLF; (2) corpus callosum; and (3) internal capsule. For tracts implicated in 3 or fewer studies, see Section 5 of Supplementary material; however, it is important to note, that these findings are inconsistent within the literature.

As outlined in Table 1, MD provides an inverse measure of membrane density and fluid viscosity providing a measure of white matter characterization that is complementary to FA [which does not provide biological specificity (Pierpaoli and Basser, 1996)]. Sixteen studies compared MD between cannabis users and controls, and of these 10 studies found differences (Table 5). The most consistent finding was higher MD in cannabis users compared to controls in 7 studies. In contrast, 2 studies reported lower MD in cannabis users compared to controls (Delisi et al., 2006; Sweigert et al., 2020). Interestingly, 1 of these studies also found lower MD across different white matter tracts, in addition to the higher MD found (Bava et al., 2009).

Of the 7 studies that found higher MD in cannabis users, a total of four studies found higher MD in the corpus callosum (Arnone et al., 2008; Gruber et al., 2011, 2014; Rigucci et al., 2016). There were multiple studies that found individual tract differences in MD, see Section 5 of the Supplementary material for a discussion of these tracts.

Overall, the literature shows a consistent finding of higher MD in cannabis users compared to controls. Although there was heterogeneity in the findings linked mostly to single tracts, 57% of studies implicated the corpus callosum with poorer white matter integrity in cannabis users.

As described earlier, RD provides a metric related to the integrity of white matter macrostructure (Alexander et al., 2011). There were no consistent findings in RD across all studies when measuring white matter integrity of cannabis users compared to controls. Less than half of the studies that examined RD found group differences (i.e., 5 out of 11, Table 6). The direction of the differences was mixed, 2 studies found higher RD in various tracts (e.g., arcuate fasciculus, internal capsule/thalamic radiation) (Ashtari et al., 2009; Rigucci et al., 2016) and another 3 studies showed lower RD in cannabis users compared to controls in the corpus callosum (Becker et al., 2015), forceps minor (Filbey et al., 2014), and middle cerebellar peduncle (Sweigert et al., 2020). Overall, this metric (1) was not as frequently used in the literature compared to other diffusion-MRI metrics, such as FA and MD and (2) showed large heterogeneity across findings in the white matter tracts and directionality of the results, with no clear outline on higher or lower white matter integrity of cannabis users when compared to controls.

AD also provides a measure of white matter macro-structural integrity, as presented in Table 1 (Alexander et al., 2011). There were no consistent findings in AD across the studies when measuring white matter changes in cannabis users compared to controls. Only 2 of 10 studies that examined AD, found group differences (Table 7). One study found cannabis users had higher AD in the temporal lobe, internal capsule/thalamic radiation, and motor tracts (Ashtari et al., 2009), and 1 study found cannabis users had lower AD in the corpus callosum (Rigucci et al., 2016). Overall, these findings indicate that AD: (1) is not as frequently used compared to other diffusion-MRI metrics such as FA and MD; (2) did not detect differences in the majority of studies; and (3) there is heterogeneity in the findings across the 2 studies and differences in the directionality of the findings, with no clear consistency in the findings.

It is important to note that not all studies measured tract-specific quantifications of white matter integrity. A minority of studies focused on other metrics including (1) global network metrics of the white matter pathways; (2) metrics on the number of streamlines in the white matter bundles; and (3) fibre bundle length. The descriptions of these single studies can be found in Section 5 of the Supplementary material. Overall, it is important to note that these metrics are not widely used in the cannabis use literature, and there are no consistent findings reported.

Table 8 shows that 15 of 30 studies investigated correlations between diffusion-MRI metrics (i.e., FA, MD, RD, AD, and other non-tract specific metrics) and indices of cannabis use (e.g., age of onset, duration, dosage, frequency, cannabis dependence severity, and abstinence), as well as other key cognitive, alcohol use, and mental health-related variables.

Overall, 2 indices of cannabis use were consistently measured to determine their associations with white matter integrity. This includes (1) age of onset; and (2) duration of use. There were two main implications of these findings. First, although FA, MD, RD, and AD were measured frequently to determine their associations with age of onset and duration of use, ~50% of the studies did not detect any significant associations. Secondly, most of the significant findings indicated that reduced white matter microstructure across all measures was associated with (1) lower age of onset and (2) longer duration of use. These important findings are comparatively rare in the literature as not all potential correlations are investigated.

The findings which were not consistently reported within the studies reviewed (i.e., in 3 or fewer studies per metric) are described in Section 6 of the Supplementary material. These findings have several implications pertaining to (1) the lack of consistency in running correlations for these outcomes and diffusion-MRI metrics; (2) the lack of associations detected; and (3) the need to further include these analyses in future studies to ensure robust comparisons can be determined.

Five of the 30 studies included were longitudinal studies that examined white matter integrity, with details outlined in Table 9. Three studies reported significant group-by-time effects in several white matter pathways (Becker et al., 2015; Epstein and Kumra, 2015; Lichenstein et al., 2022), and additional interesting findings emerged. Overall, there was no consistency across these significant findings across the literature in terms of direction of findings and white matter tracts implicated, longitudinally. Two of these studies reported a lower increase in white matter integrity over time in cannabis users compared to healthy controls (Becker et al., 2015; Lichenstein et al., 2022), however these were in different white matter tracts [i.e., SLF, cingulum, and Corticospinal tract (CST) (trend)]. Interestingly, FA was the most sensitive metric to longitudinal changes in group-by-time effects, being implicated in all three studies, with RD only being implicated in one study (i.e., Becker et al., 2015).

For the effects of time, only two studies found significant differences, which were both decreases in white matter integrity in cannabis users (Jacobus et al., 2013a; Becker et al., 2015). These differences were only found in the FA metric and were implicated in a wide array of tracts, not consistent between both studies.

Finally, there were no consistent findings of brain-behavior correlations between measures of change in white matter integrity and cannabis-related metrics. Two studies showed negative correlations between cannabis-related metrics (e.g., cannabis hits/past year, total days cannabis use over time) and white matter integrity (Becker et al., 2015; Epstein and Kumra, 2015). However, the white matter tracts implicated were not consistent across the studies.

The SLF was one of the most consistent pathways with white matter differences between groups. The SLF, along with the arcuate fasciculus, is a major association pathway in the brain, connecting the frontal lobes with the ipsilateral parietal, occipital, and temporal lobes (Schmahmann et al., 2008). It has been implicated in executive functioning, including sustained attention (Clemente et al., 2021), and the frontal mediation of attention and executive function (Baker et al., 2013). Importantly, 2 studies found correlations between differences in the SLF, and cannabis exposure metrics, a positive association between FA and age of onset, and a negative association between RD and age of onset, in addition to a positive correlation between FA and frequency of use (hits/past 3 months). Furthermore, in a study of children and adolescents, FA in the SLF was positively correlated with cognitive set-shifting, an important domain of executive functioning (Urger et al., 2015). As such, it could be suggested that decreased FA within the SLF of cannabis users may also be associated with decreased executive function, especially considering that early onset (Gruber et al., 2012), and exposure to high-potency cannabis (Ramaekers et al., 2006) has been associated with decreased executive function. Based on the evidence above, we may infer a potential cumulative relationship between cannabis use and differences in SLF microstructure. However, given how few studies indicated significant associations between SLF microstructure, and cannabis exposure metrics, it is difficult to ascertain if these differences are neuroadaptations associated with the effects of cannabis on the brain or, are instead differences in integrity, perhaps related to executive function, that predate cannabis use onset. Longitudinal diffusion-MRI research with careful assessment of cannabis exposure metrics is warranted to elucidate the role of cannabis exposure on SLF microstructure.

Differences in white matter microstructure were also consistently seen in the corpus callosum, with differences in both FA and MD. The corpus callosum is a major commissural tract that allows for inter-hemispheric communication (Standring and Gray, 2021). Across multiple studies, correlations were shown between white matter differences in the corpus callosum and cannabis exposure metrics. Associations were seen between microstructure of the corpus callosum and age of onset (Gruber et al., 2011, 2014; Orr et al., 2016), duration of use (Arnone et al., 2008; Gruber et al., 2011), and frequency of use (Rigucci et al., 2016).

In a previous meta-analysis of white matter microstructure, differences in FA in the corpus callosum have been noted between persons who use substances (e.g., cannabis, alcohol, nicotine, and opiates) and controls (Hampton et al., 2019). Such group differences in white matter microstructure might reflect neuroadaptations from addiction processes shared across different substances (Hampton et al., 2019). Additionally, a recent study of medicinal cannabis use, found increased FA in the corpus callosum after 3 and 6 months of administration (Dahlgren et al., 2022). Although medicinal and recreational cannabis use differ, this finding nonetheless suggests that cannabis use may alter corpus callosum microstructure.

Studies into CB1 receptor density in rats have shown that high receptor density in white matter tracts, including the corpus callosum, is present early in development, before decreasing and becoming denser in gray matter areas into adulthood (Romero et al., 1997). Adolescence is a period characterized by large-scale neurodevelopment and major brain maturational changes (Asato et al., 2010). It is also, concurrently, a time during which cannabis use may first commence (Richmond-Rakerd et al., 2017; AIHW, 2020). Further, the endocannabinoid system—which comprises CB1 and CB2 receptors, multiple endogenous lipid derivatives which activate them, as well as enzymes which control the levels of the lipid derivatives—plays a key role in typical neurodevelopment (Malone et al., 2010; Hourani and Alexander, 2018). Therefore, white matter differences observed in cannabis users may reflect the influence of exogenous cannabis exposure on the endocannabinoid system that play a key role in neurodevelopment.

The exact mechanisms by which cannabis use may be associated with differences in white matter integrity are largely unclear. Correlations between white matter and cannabis dosage suggest neuroadaptations from exposure to cannabis as postulated by proponents of neuroscientific theories of addiction (Koob and Volkow, 2016; Zehra et al., 2018); possible neurotoxic effects preliminarily shown in animal studies (Scallet, 1991; Sarne et al., 2011); while correlations between the age of cannabis use onset and FA may reflect developmental processes affected by cannabis exposure through brain maturation, possibly via cannabinoids affecting the endocannabinoid system that regulates neurodevelopmental processes (Meyer et al., 2018; Farrelly and Vlachou, 2021). Yet, the exact mechanisms that underlie such associations are yet to be clarified with human and preclinical studies. While there is some evidence to suggest that corpus callosum microstructural differences are associated with cannabis exposure, longitudinal studies that capture brain maturation prior to the onset of cannabis use, are necessary to understand which parameters drive white matter changes in cannabis users.

A minority of 5 studies did not find significant cross-sectional group differences in DTI metrics (Jacobus et al., 2013b; Orr et al., 2016; Yucel et al., 2016; Knodt et al., 2022; Cousijn et al., 2022). One study reported only trend level differences in white matter microstructure (Gruber and Yurgelun-Todd, 2005). In addition, another study found non-significant differences in diffusion-MRI metrics, finding significant differences in fibre bundle length, but not the DTI metrics FA, MD, RD, or AD (Levar et al., 2018). The lack of group difference may be due to moderate levels of cannabis exposure in the control groups [e.g., up to 50 cannabis use occasions (Cousijn et al., 2022)], low levels of cannabis exposure in the cannabis samples [e.g., < 10 times in 50% of the sample (Orr et al., 2016), or recreational levels of use (Levar et al., 2018)]. It also cannot be excluded that the examined samples endorsed additional features that are protective of cannabis-related neuroanatomical changes, such as youth age (Solowij et al., 2016a; Lorenzetti et al., 2021), unmeasured lifestyle and physiological indices, known to affect neuroanatomy [e.g., exercise and body composition (Kandola et al., 2016; Den Ouden et al., 2018; Kakoschke et al., 2019)].

In understanding these heterogeneous findings, it is important to note that many of the reviewed studies were subject to several limitations. This section will detail the limitations of these studies, which include and are not limited to small sample sizes; presence of cannabis use in control groups and low cannabis use in cannabis groups; heterogeneity of measures and analyses; limited exploration of key variables correlating with white matter microstructure and the intrinsic limits associated with diffusion-MRI metrics due to the lack of biological specificity in DTI metrics.

This review includes several limitations that need to be considered. For example, the reviewed studies might have incorporated participants endorsing major mental health problems or significant exposure to substances other than cannabis. Indeed, we screened for studies endorsing these exclusion criteria at a group level. Thus, it cannot be excluded that the literature findings might have been partially affected by other variables that also affect brain integrity, including mental health problems or exposure to substances other than cannabis or both (Brambilla et al., 2003, 2005; Squarcina et al., 2017; Prunas et al., 2018; Mackey et al., 2019; Chye et al., 2020b; Delvecchio et al., 2020; Navarri et al., 2020; van Velzen et al., 2020; Cao et al., 2021; Rossetti et al., 2022; Zovetti et al., 2022). This issue might have biased the findings, or else might have rendered the findings more representative of persons who use cannabis. Indeed, some groups of cannabis users—such as those endorsing a cannabis use disorder—experience mental health problems and polysubstance use (Lawn et al., 2022).

Further, we included samples exposed to alcohol, in particular, 2 manuscripts included persons who used cannabis and engaged in binge drinking (Jacobus et al., 2009, 2013b). Importantly, alcohol exposure can affect white matter integrity independently and in interaction with cannabis (Rossetti et al., 2022). Indeed, binge drinkers with vs. without cannabis use showed differences in selected white matter tracts [e.g., uncinate fasciculus (Jacobus et al., 2013b)]. Given that cannabis and alcohol use often co-occur (AIHW, 2022), the findings may reflect neurobiological changes of exposure to both substances. Future work is warranted to measure alcohol exposure and account for it in analyses of white-matter integrity in cannabis users, to unpack the impact of co-occurring alcohol and cannabis exposure.

Finally, the review did not include samples who consumed synthetic cannabinoids, due to their different psychopharmacological profile of synthetic cannabinoids vs. cannabinoids (van Amsterdam et al., 2015). Synthetic cannabinoids can also affect white matter integrity (Zorlu et al., 2016; Gokharman et al., 2020) and are becoming increasingly diversified and available. Therefore, further research is required to examine how synthetic cannabinoids affect white matter microstructure.

The reviewed evidence warrants recommendations for future work. First, the evidence reports heterogeneous metrics and findings (e.g., distinct measures of brain integrity and brain pathways). Future studies are warranted to apply precise and robust analytical frameworks in order to identify subtle changes to white matter in cannabis users. For example, FBA holds promise as a robust tool to map with precision white matter integrity as outlined above (Raffelt et al., 2017). Second, we warrant pre-registration of future empirical studies to encourage transparency in the reporting of the analyses and findings and reduce the risk of “cherry-picking” significant results and publication bias. Thus far, only 1 of the reviewed studies was pre-registered (Knodt et al., 2022).

Third, future studies should report cannabis use in a systematic and standardized fashion (Lorenzetti et al., 2016a,c; Solowij et al., 2016b; Lopez-Pelayo et al., 2021). They include standard THC units (Freeman and Lorenzetti, 2019, 2021) which have received support from researchers and health organizations internationally (Filbey, 2020; Hammond and Goodman, 2020; Volkow and Weiss, 2020; NIDA, 2021), as well as the iCannToolkit (Lorenzetti et al., 2021, 2022a,b), also endorsed by researchers internationally (Kuhns and Kroon, 2021; Weiss and Volkow, 2021). The use of standardized tools to measure cannabis exposure will enable the understanding of how specific parameters of cannabis exposure (e.g., age of onset, dosage, and duration) and levels of exposure might be risky for white matter integrity. Fourth, longitudinal multimodal neuroimaging studies research are necessary to determine when white matter changes begin to appear (e.g., before cannabis use onset, at specific times after cannabis use commences), how they change over time and if they dissipate after cessation of cannabis use. These research questions may be achieved via using data from large-scale longitudinal consortia initiatives (e.g., ABCD study).