Substance Abuse and White Matter: Findings, Limitations, and Future of Diffusion Tensor Imaging Research

1.1. Diffusion Tensor Imaging

Diffusion tensor imaging (DTI) is a magnetic resonance imaging technique that estimates white matter in vivo (Bihan et al., 2001) and is increasingly being used to study the biology of addiction. DTI data can be used to infer information about white matter macrostructure, microstructure, and connectivity. White matter macrostructure is typically measured in terms of total volume, or volume of particular tracts of white matter fibers, i.e. bundles of axons. Using additional analytic techniques that model the direction of water movement, DTI can also capture the microstructural properties of white matter. The most commonly reported measure of white matter microstructure is fractional anisotropy (FA). FA is a composite measure of the extent to which water diffusion is constrained along a particular direction. Axons and their myelinated sheaths restrict water diffusion such that it is greater in the axis parallel to the main direction of axons (Soares, Marques, Alves, & Sousa, 2013). FA is often used as a general index of white matter “coherence”, with lower numbers often reported as “worse” (e.g. Kubicki et al., 2005). Changes in FA signal one or many microstructural alterations, such as Wallerian degeneration (Pierpaoli et al., 2001; Thomalla et al., 2004) or decreased neuronal membrane permeability (Jones, KnöSche, & Turner, 2013).

Several other indexes of microstructure, including mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD) are also often reported in DTI studies. MD corresponds to the molecular diffusion rate, with higher values corresponding to higher diffusivity, which is often observed in damaged tissues (Jones et al., 2013). RD has been used to infer demyelination or dysmyelination, with greater RD indicating more severe of either of the former (Song et al., 2005; Wozniak and Lim 2006). AD increases potentially indicate axonal degradation (Song et al., 2005), while AD decreases are often associated with healthy brain maturation (Tamnes et al., 2010). Together, these measures have the potential to offer critical insight into substance-related brain alterations to white matter microstructure.

1.3. Review Motivation and Aims

In this review, we survey the past decade of DTI research with the goal of examining the relationship, specificity, and directionality, of substance-related differences in white matter microstructure. We attempt to synthesize the results from following major substance categories: alcohol, cannabis, cocaine, nicotine, and opiates. To augment this review, we conduct a mini metaanalysis of the overlapping findings, and compare the relative effect sizes of various abused substances on human white matter. We organize this review into sections examining cross-sectional and longitudinal research of substance-related differences in white matter microstructure. We then discuss the possible mechanisms of these findings, the limitations of the studies reviewed, and offer guidance for future research.

1.3. Literature Selection and Exclusion Criteria

We searched the Web of Science database (http://apps.webofknowledge.com/) and PubMed (https://www.ncbi.nlm.nih.gov/pubmed/) using the following Boolean criteria: “white matter” AND (“DTI” OR “DWI” OR “diffusion” OR “dMRI”) AND “[substance of abuse]”. The substances of abuse in this case were “alcohol”, (“cannabis” OR “marijuana”), “cocaine”, “nicotine”, and (“opiates” OR “heroin”). This search included articles published through December of 2018. Inclusion criteria were: (i) employed diffusion weighted or tensor magnetic resonance imaging; (ii) published no earlier than 2008; (iii) sample size sufficiently powered to conduct at least group level statistics.

It was critical to our review that we only include research with methodologies appropriate to address our aims, leading to several exclusion criteria. First, we aim to examine if consistent relationships exist between individual substances of abuse and white matter microstructure. As studies of polydrug users obfuscate such one-to-one relationships, studies examining polydrug use were excluded. Second, studies using participants intentionally drawn from clinical populations, aside from substance abuse or addiction, were also excluded; the white matter differences observed in such studies could be due to clinical disorder, rather than substance-related. Studies that met our inclusion and exclusion criteria were then grouped by substance of abuse and details of each study were extracted (Table 1). In the following sections, we first review the cross-sectional studies, and then the longitudinal DTI research on substance abuse.

2.1. Alcohol

Gross anatomical estimations indicate that alcohol-abusing individuals have lower white matter volume (Arnone et al., 2008; Pfefferbaum, Adalsteinsson, & Sullivan, 2006). However, such macrostructural differences are not always detectable in those with mild alcohol use disorder or a shorter history of alcohol abuse (Asensio et al., 2016), underscoring the value of the more fine-grained microstructural measures possible with DTI. The DTI studies of alcohol abuse reviewed varied in both the participant age and type of drinking behavior being studied.

A subgroup of these alcohol-related studies examined binge, or otherwise heavy, drinking in adolescent populations. Two independent research groups found that binge-drinking adolescents—compared to controls—had lower FA (Badsberg Jensen & Pakkenberg, 1993; McQueeny et al., 2009). FA also seems differentiate between high and low severity of heavy adolescent drinkers, such that those with higher Alcohol Use Disorder scores tends to have lower FA (Thayer, Callahan, Weiland, Hutchison, & Bryan, 2013). Although one large study in younger adolescents found higher FA in individuals with higher Alcohol Use Disorders scores (Cardenas et al., 2013), this result did not stand after correction for multiple comparisons.

Mirroring the general finding among adolescents, alcohol-abusing adults also exhibit lower FA than controls (Harris et al., 2008; Schulte, Muller-Oehring, Pfefferbaum, & Sullivan, 2010). Consistent with previous FA findings, measures of AD showed a positive relationship with alcohol abuse. That is, alcohol was associated with higher AD in alcohol abusing groups (Pfefferbaum et al., 2014; Topiwala et al., 2017). At variance with the aforementioned studies finding linear relationships, one recent and well-powered (n=377) study suggests that there may be an inverted-u relationship between alcohol use and white matter microstructure. Specifically, light and moderate alcohol consumption was associated with increased FA, whereas heavy consumption (28 drinks or more within a two-week period), was associated with marked decreases in FA (McEvoy et al., 2018). Together these results highlight the importance of both age and dosage in determining the relationship between alcohol and white matter.

2.2. Cannabis

Studies examining the relationship between cannabis abuse and white matter microstructure have yielded conflicting results (Cousijn et al., 2012). One preliminary study (Arnone et al., 2008) found no white matter differences between users and non-users, leading the authors to conjecture that cannabis use does not result in neurotoxic outcomes. However, several shortcomings in the methods of this study (e.g. only partial control of alcohol intake, and low magnetic field strength) should temper any strong conclusions. These limitations, and others, are addressed in more depth in the Discussion section and are summarized in Box 1.

Contrary to this null finding, studies of heavy cannabis users have tended to find significant differences in white matter microstructure. For instance, Ashtari and colleagues found that heavy users of cannabis exhibited lower FA (Ashtari, Cervellione, Cottone, Ardekani, & Kumra, 2009). This association between heavy cannabis use and lower FA was subsequently replicated by an independent research group (Gruber, Silveri, Dahlgren, & Yurgelun-Todd, 2011). In addition to the research on heavy users, recreational or “regular” cannabis use has also been studied using DTI. Two such studies found that regular cannabis to be associated with lower FA (Jakabek, Yücel, Lorenzetti, & Solowij, 2016; Shollenbarger, Price, Wieser, & Lisdahl, 2015). Contrasting with these findings of linear relationships, other research points to a quadratic relationship. Specifically, higher FA has been observed following regular initial marijuana use, but FA then decreases with extended use (Filbey et al., 2014). The relationship between RD and cannabis use is similarly mixed, with studies finding higher (Ashtari et al., 2009; Becker, Collins, Lim, Muetzel, & Luciana, 2015; Jakabek et al., 2016), lower (Filbey et al., 2014; Jakabek et al., 2016), or nonsignificant (Shollenbarger et al., 2015) differences in cannabis users relative to controls.

These inconsistent findings in the relationship between cannabis use and white matter may be in part due to age of onset. Specifically, lower FA values have been connected with earlier onset of among cannabis users (Gruber, Dahlgren, Sagar, Gönenç, & Lukas, 2014). Complimenting this empirical finding, a large meta-analytical study using the Human Connection Project data also found that the earlier the age of onset of marijuana use, the “worse” the white matter microstructural values (Orr, Paschall, & Banich, 2016). Together, the cross-sectional research on points to a complex relationship between cannabis use and white matter governed by an array of factors including age and age of onset, duration of use, and usage severity.

2.3. Cocaine

Previous structural MRI studies have demonstrated that cocaine use is associated with lower grey matter volume (Narayana, Datta, Tao, Steinberg, & Moeller, 2010). Several recent DTI studies have found that cocaine users had lower white matter FA relative to controls (Azadeh et al., 2016; Romero, Asensio, Palau, Sanchez, & Romero, 2010). This negative association between cocaine use and white matter microstructure is echoed by other research indicating that cocaine-dependent individuals exhibit attenuated increases in white matter volume typical during brain maturation (Bartzokis et al., 2002). Such a negative association has not always been found, however. Interestingly, cocaine abuse is associated with higher FA in certain regions of the brain, and lower FA in other areas (Morie et al., 2017; Romero et al., 2010). Although preliminary, such studies raise the possibility that the mechanism by which cocaine affects the brain may be different than that of other substances. We briefly discuss a mechanism by which cocaine might affect white matter in Section 5.

2.4. Nicotine

Heavy nicotine use in the form of smoking tobacco has been linked to neuropathy (Brody, 2006), often manifesting as prefrontal gray matter atrophy (Gallinat et al., 2006; Zhang et al., 2011). Conversely, consumption of nicotine via smoking has been associated with higher white matter volume (Gazdzinski et al., 2005; Yu, Zhao, & Lu, 2011). Studies examining nicotine use via DTI have found similarly conflicting results. In chronic nicotine users, heavy consumption has been associated with lower FA (Lin, Wu, Zhu, & Lei, 2013) and higher FA (Paul et al., 2008), as well has both lower RD (Wang et al., 2017) and higher RD (Lin et al., 2013). The results of studies examining non-chronic, regular nicotine use are similarly split. Regular nicotine use has been associated with lower FA (Huang et al., 2013; Liao et al., 2011; Zhang et al., 2011) and higher FA (Hudkins, O’Neill, Tobias, Bartzokis, & London, 2012; Wang et al., 2017). These seemingly conflicting nicotine results may be partly accounted for by the developmental stage in which it is consumed, with higher FA more commonly observed in younger nicotine users (Hudkins et al., 2012; Jacobsen et al., 2007). Alternatively, it maybe that the association between nicotine use and higher FA in adolescents is temporary, and eventually leads to microstructural declines with chronic use. Future longitudinal studies could formally address this theory.

2.5. Opiates

Opiate abuse has been associated with changes in cortical blood flow (Fu et al., 2008; Lubman et al., 2009), grey matter loss (Lyoo et al., 2006) and impaired connectivity (Schmidt et al., 2015). More recently, several studies have examined the relationship between opiate and white matter microstructure. Relative to non-using controls, opiate-abusing individuals consistently exhibit both lower FA (Bora et al., 2012; Fu et al., 2008; Li et al., 2013; Qiu et al., 2013; Wang et al., 2011) and lower AD (Bora et al., 2012; Qiu et al., 2013). Turning to RD, again the findings are consistent, with opiate users exhibiting higher RD. Within opiate users, FA has also been shown to reliably dissociate short- and long-term users, with the latter having significantly lower FA (Qiu et al., 2013). Together the results suggest that longer durations of opiate dependence are associated with more extensive and severe white matter abnormalities.

2.6. Specificity

Our review of the literature suggests that there is a relationship between substance abuse of addictive substances and white matter microstructure. A logical next question is does substance abuse broadly affect white matter or does it systematically affect certain white matter pathways?

Across substance of abuse, the majority of the studies reviewed found differences in major white matter fiber pathways such as the corpus callosum, superior and inferior longitudinal fasciculi. Table 1 includes a summary of neuroanatomical loci of changes in white matter microstructure for each substance. Overall, there was poor consistency in the neuroanatomical loci of significant microstructural differences. In interpreting these findings, however, it is important to note that many of the studies reviewed performed whole-brain analysis of white matter microstructure using TBSS. This approach is limiting in several ways that we discuss further in Section 6. Nonetheless, there did seem to be one consistent finding across substance of abuse: nearly half of the studies reviewed found at least one significant microstructural difference in the corpus callosum, as measured by FA. We therefore conducted a mini meta-analysis to compare the relationship between each substance and corpus callosum FA values, relative to control.

2.7. Mini Meta-Analysis: Corpus Callosum and Substance Abuse

In this meta-analysis, we seek to compare the relative relationship of different substances on callosal white matter microstructure, as indexed by FA. That is, we aim to address if some substances are associated with greater differences in white matter than are others. To do this, we extracted effect sizes from the studies reviewed (Table 1) in terms of R² for any significant difference in the white matter pathway most consistently implicated across substance—the corpus callosum. If effect size was given in terms of R², we recorded it. If effect size was given in another format, we converted it to R². If no effect size was given, but adequate mean FA information was provided, we calculated R². If there was more than one significant finding in the corpus callosum, we took the average of the R² values. If a study did not report sufficient information to determine effect size, we excluded it from this analysis.

None of the cocaine studies reviewed provided enough information to extract effect size. In fact, across substance fewer than half of the studied reviewed contained the necessary information to evaluate effect size in the corpus callosum¹. Accordingly, there was insufficient power to conduct between-group level statistics. However, a visual inspection (see Figure 1) of the direction and effect size of the FA differences across substances is nonetheless interesting for several reasons. First, users of alcohol, cocaine and opiates exhibited consistently lower FA in the corpus callosum. Within this group, alcohol and opiate use appeared to have a higher magnitude of effect, relative to controls. For cannabis and nicotine, the results were mixed both in terms of direction and effect size. Future research with appropriate statistical power could formally test the veracity of the trends observed here.

4.1. Causality and Tracking

In this subsection, we review longitudinal evidence and related research that examines the extent to which substance abuse may cause changes in white matter microstructure, and whether DTI can be used to track substance abuse or treatment.

For alcohol, longitudinal research hints at a causal relationship between alcohol abuse and white matter microstructure. In one corroborative study, alcoholic and control groups had their white matter and drinking habits monitored over the course of several years. While both groups showed age-related declines of FA in white matter, alcoholics exhibited lower FA irrespective of age (Pfefferbaum et al., 2014). Similarly, longitudinal research of adolescent drinkers has shown decreases in FA in major white matter pathways following the onset of alcohol use (Jacobus, Squeglia, Alejandra Infante, Bava, & Tapert, 2013). Together, these results support the notion that alcohol abuse likely induces white matter changes, not vice versa.

For cannabis, the only longitudinal study in this literature similarly suggests that cannabis use causes white matter changes, not the reverse. In this study, heavy cannabis users exhibited significant declines in white matter FA during the two year period for which they were monitored (Becker et al., 2015).

For cocaine, there is also evidence that cocaine use affects white matter microstructure. Compared to adolescents who had not been exposed to cocaine prenatally, exposed adolescents had lower FA in the cingulum and superior longitudinal fasciculus (Morie et al., 2017). The notion that cocaine induces microstructural white matter changes is corroborated by a recent preliminary longitudinal study. Ma and colleagues found that the more times an individual tested positive for cocaine between the two scanning sessions, the more an individual’s FA decreased. Similarly, they found that lifetime cocaine use negatively predicts callosal FA. Together, these studies provide fairly consistent evidence that cocaine abuse negatively impacts white matter microstructure (Ma et al., 2017). Finally, white matter FA also seems to track treatment outcome, with higher FA significantly predicting the number of successful days of abstinence from cocaine (Bell, Foxe, Nierenberg, Hoptman, & Garavan, 2011; Xu et al., 2010).

For nicotine, none of the studies reviewed utilized an longitudinal design. Despite the lack of longitudinal research, the current evidence suggests that FA may be a valuable measure for tracking progression of nicotine addiction. Specifically, there seems to a negative relationship between FA with daily usage (Baeza-Loya et al., 2016), duration of usage (Wang et al., 2017; Yu et al., 2011; Yuce et al., 2016), and addiction severity (Yu et al., 2011). That is, lower FA values were associated with higher and longer usage, and higher addiction severity.

Finally, for opiates, longitudinal white matter and opiate addiction research is lacking. However, as with nicotine, longer duration of opiate addiction has been associate with lower FA throughout the brain (Qiu et al., 2013) and more specifically with lower FA in SLF and frontal white matter (Bora et al., 2012). Further, FA values reliably dissociate short- and long-term users of opiates, with the latter having significantly lower FA (Qiu et al., 2013). Together the results suggest that longer durations of opiate dependence are associated with more extensive and more severe white matter abnormalities.

Does substance abuse affect particular white matter pathways?

The majority of the studies reviewed found differences in major white matter fiber pathways such as the corpus callosum, superior and inferior longitudinal fasciculus. While these observed differences between addicted individuals and healthy controls in these pathways may be valid, such findings do little to tie anatomy to addictive behaviors since the function of these white matter pathways is unclear. That is, it is unclear how lower FA in the corpus callosum would manifest in everyday life. While such white matter changes could, for example, lead to further addictive behavior, altered FA in the corpus callosum has also been associated with a host of other variables such as age (Lebel, Caverhill-Godkewitsch, & Beaulieu, 2010), gender and handedness (Westerhausen et al., 2004), autism (Hermann et al., 2007), traumatic brain injury (Kumar et al., 2009), schizophrenia (Foong et al., 2000), and bipolar disorder (Barnea-Goraly, Chang, Karchemskiy, Howe, & Reiss, 2009). Given this, an alternative possible explanation is that somefindings were restricted to the most prominent white matter pathways due to the methodological approach. We discuss some limitations of the studies we reviewed in Section 3.

Several studies did find more regionally-specific differences in white matter. These differences were often somewhere in the frontostriatal network, i.e. the white matter that connects the striatum and the frontal lobe (Chudasama & Robbins, 2006). These findings are consistent with other neuroimaging research suggesting that addiction is partly mediated by dopaminergic and glutamatergic projections among the aforementioned brain regions (Koob & Volkow, 2016). It is plausible that the decision-making observed in addiction might be partly due to, or the result of, aberrant communication caused by altered or damaged white matter between key brain regions (Hampton, Alm, Venkatraman, Nugiel, & Olson, 2017; Zhang et al., 2011).

Substances likely affect white matter via different mechanisms

It is tempting to assume that all substances interfere with the myelination process at some stage, leading to demyelination or dysmyelination. While FA and other DTI indices are associated with such changes, it is important to remember that these measures are also affected by many other factors such as transient changes in neuronal membrane permeability (Jones et al., 2013). We therefore recommend cautious interpretation of DTI findings with respect to the underlying mechanisms.

Substance-related alterations to white matter may not be permanent

Given that our the central nervous system is an intricately balanced, complex network of billions of neurons and supporting cells, some might imagine that extrinsic substances could cause irreversible brain damage. Our review paints a less gloomy picture of the substances reviewed, however. Following prolonged abstinence, abusers of alcohol (Pfefferbaum et al., 2014) or opiates (Wang et al., 2011) have white matter microstructure that is not significantly different from nonusers. There was also no evidence that the white matter microstructural changes observed in longitudinal studies of cannabis, nicotine, or cocaine were completely irreparable. It is therefore possible that, at least to some degree, abstinence can reverse effects of substance abuse on white matter. The ability of white matter to “bounce back” very likely depends on the level and duration of abuse, as well as the substance being abused. This theory could be tested in future well-controlled longitudinal studies. If true, this also underscores the need for addiction interventions that are able to induce periods of prolonged abstinence.