LEARNING OBJECTIVES

After reading this article you will be able to:

• • identify the neural networks involved in substance dependence and relapse

• • appreciate how neuroimaging continues to characterise different aspects of substance dependence and treatment

• • recognise how individualised therapy has potential to shape the future of treatment in substance dependence.

Addiction is a global health problem that places a huge burden on individuals and society. In the UK, it is estimated that addictions consume £15 billion of government resources yearly, and on average £4 billion of the annual National Health Service (NHS) budget is spent on alcohol and drug misuse. National Drug Treatment and Monitoring Statistics report that, for the year 2017 to 2018, approximately 270 000 adults were in contact with drug and alcohol services (Knight Reference Knight, Brand and van der Merwe2018). This is likely to be an underrepresentation of individuals who would benefit from service intervention, but nevertheless it reflects how a relatively small section of the population demand a great deal of resources and support.

Clinically, the chronically relapsing nature of addiction is a core problem, with few treatment options in abstinence maintenance. Relapse rates are high; despite usual treatment, over 20% of individuals will relapse within 1 month and a further 40% within 6 months. Longitudinal data suggest that at least 5 years of abstinence are required before recovery is well established, after which only a minority will relapse (White Reference White2012). Historically this has been a hugely neglected area of study and one with significant room for improvement and development, and the ability to predict relapse and the development of novel treatments continue to be major goals in addiction research. Understanding the neurobiological mechanisms underlying relapse will help us to identify therapeutic targets to reduce this risk, providing tools that we can use to achieve these goals.

Theories of addiction

Drug addiction has been conceptualised as a disorder that affects neural networks centred on reward, inhibitory control, stress, emotional processing and learning/memory. Interference with these networks supports the theoretical framework for the three-stage addiction cycle that promotes drug-seeking behaviour in the addicted state (Koob Reference Koob and Le Moal2008). There are also several eminent addiction theories: incentive salience (Robinson Reference Robinson and Berridge1993, Reference Robinson and Berridge2008), reward deficiency syndrome (Blum Reference Blum, Braverman and Holder2000, Reference Blum, Thanos and Gold2014), the allostatic hypothesis (Koob Reference Koob and Le Moal2005), habit formation theory (the shift from impulsive to compulsive choice) (Everitt Reference Everitt and Robbins2005, Reference Everitt and Robbins2016) and impaired response inhibition and salience attribution (Goldstein Reference Goldstein and Volkow2002).

There is overlap between all models with regard to impairments in reward processing and ‘hijacking’ of reward circuitry by drugs of misuse, such that the motivation to procure drugs overpowers the drive for non-drug-related goals. Koob's model describing the role of various brain regions in substance use and addiction is useful to understand which parts of the brain are involved in binge/intoxication, withdrawal/negative affect and preoccupation/anticipation (Fig. 1) (Koob Reference Koob and Volkow2016). For instance, the reward system in the striatum is key during binge/intoxication, with dysregulation in the amygdala mediating the low mood and emotional stress seen in withdrawal. The prefrontal cortex (PFC) is responsible for inhibitory control and a key modulator of striatal activity, i.e. reward processes. Impairments in this region manifest as impulsivity and contribute to the preoccupation/anticipation phase. In the early stages of drug misuse, positive reinforcement of drug consumption concurrent with impulsivity drives this cycle. Once the transition to substance dependence is more robust, drug-taking becomes more compulsive and negative reinforcement becomes the driver of drug-seeking and consumption (Koob Reference Koob and Volkow2010; Goldstein Reference Goldstein and Volkow2011).

FIG 1 Brain regions involved in substance use and addiction (Koob Reference Koob and Volkow2016). PFC, prefrontal cortex; ACC, anterior cingulate cortex; OFC, orbitofrontal cortex; NAc, nucleus accumbens; VTA, ventral tegmental area; GABA, γ-aminobutyric acid; dStr, dorsal striatum; Amyg(ce), central nucleus of the amygdala; CRF, corticotropin-releasing factor; BNST, bed nucleus of the stria terminalis; ACh, acetylcholine; NPY, neuropeptide Y; EC, endocannabinoids; Hippo, hippocampus; Amyg(B), basolateral amygdala.

In the past few decades, neuroimaging has transformed our understanding of the neurobiology of addiction, the impact of substances, and vulnerability and resilience factors. The focus has been on improving treatments and clinical outcomes by predicting treatment response – which enables targeted individualised therapy – and improving early diagnosis and interventions. Definitions and clinical criteria that describe and classify addictive disorders have also been changing, and we begin our discussion with an overview of these.

Definitions: addiction versus dependence

Research in psychiatry generally classifies individuals on the basis of their diagnosis using one of the two main classification systems: the DSM, now in its 5th edition (DSM-5) (American Psychiatric Association 2013) is commonly used in research; the ICD is widely used clinically and is currently being released in its 11th revision (ICD-11) (World Health Organisation 2018). Although concordance between DSM-IV, ICD-10 and ICD-11 was high, this is not necessarily the case with DSM-5, particularly when diagnosing harmful or mild dependence, resulting in implications for interpreting and comparing studies using these different systems and versions (Saunders Reference Saunders, Degenhardt and Reed2019). DSM-5 considers substance use disorder as a continuum rather than separating misuse and dependence, whereas ICD-11 retains ICD-10 categories of ‘harmful use’ and ‘dependence’. It must be remembered that in pharmacological terms, dependence also refers to the neuroadaptations that underlie tolerance and withdrawal (e.g. in relation to prescribed opioids or benzodiazepines) but does not necessarily mean ‘addiction’, which is a complex behaviour involving drug-seeking and craving. Clinically this distinction must be clarified to inform appropriate management.

Functional neuroimaging in addictions

Much of what we understand of the neurobiology of addiction has been informed through research in animal models. Animals, particularly rodents, have been shown to repeatedly lever-press for drug administration, establishing that drugs of misuse act as reinforcers (for a review see Sanchis-Segura Reference Sanchis-Segura and Spanagel2006). Subsequent translational models of misuse liability, drug-taking, escalation and relapse have been developed using self-administration, conditioned place preference, drug discrimination and drug reinstatement paradigms. Such research has allowed us to identify structures and pathways that contribute to the neuropathology of addiction (Lynch Reference Lynch, Nicholson and Dance2010). However, there are limitations to animal models, such as modelling vulnerability factors and complex addictive behaviours. In an attempt to accelerate the development of drugs into the clinic, more translational approaches, such as functional magnetic resonance imaging (fMRI), have been utilised to better understand the relationship between brain mechanisms and behaviour in the human brain.

Functional neuroimaging techniques such as positron emission tomography (PET) and fMRI have been widely applied to relate biochemistry, neural substrates and networks to behaviour in humans. PET measures gamma rays emitted by radioactivity attached to a compound that binds to a known target, for example a neurotransmitter receptor or protein involved in inflammatory processes or glucose metabolism. PET is the only imaging modality to directly assess neurotransmitters such as dopamine. fMRI measures brain activity via the blood oxygenation level-dependent (BOLD) signal, providing an indirect measure of brain activity that can be examined at rest (i.e. resting-state fMRI) or in response to neurocognitive tasks. The more recently developed MRI-based techniques of magnetic resonance spectroscopy (MRS) and diffusion tensor imaging (DTI) are being applied to study addiction. MRS enables us to characterise cellular constituents such as N-acetyl aspartate or choline, as well as gamma-aminobutyric acid (GABA) and glutamate (although it reflects primarily metabolic rather than small neurotransmitter pools), and DTI reveals the function of white matter tracts.

More recently, fMRI research has focused on identifying neural ‘biomarkers’ of substance dependence, which would help clarify how and under what circumstances treatments are effective, allowing for more individualised and effective treatment of substance dependence (Moeller Reference Moeller and Paulus2018). Owing to the importance of reward processing in the manifestation of substance dependence, many fMRI paradigms have focused on delineating the neural correlates of drug and non-drug reward processing, using tasks such as cue reactivity. Nevertheless, it is important not to overlook that addiction also involves dysregulation in inhibitory control, stress and emotional processing.

Reward

The dopaminergic mesocorticolimbic pathway is commonly associated with reward and projects from the ventral tegmental area (VTA) to the striatum and PFC (Table 1). Conventional rewards such as food and sex activate this pathway to create pleasurable feelings and reinforcement of the reward. Hypofunction of this pathway underpins the ‘reward deficiency’ hypothesis of vulnerability to and perpetuation of addiction.

TABLE 1 Brain regions and their involvement in reward processing

As drug use develops into regular use and possible dependence, the dopaminergic mesocorticolimbic system becomes dysregulated. Drugs of misuse are able to ‘hijack’ the reward system by increasing dopamine in the striatum, particularly limbic or ventral striatum, more so than conventional rewards (Koob Reference Koob2013) (Table 2). For example, stimulants such as cocaine and amphetamines directly increase extracellular dopamine by blocking the presynaptic dopamine transporter (DAT), and amphetamines also increase the release rate from synaptic vesicles. In contrast, drugs such as opiates, alcohol, cannabis and nicotine indirectly increase extracellular dopamine by inhibiting GABA interneurons, which in turn increase VTA dopaminergic firing. A range of inhibitory receptors modulate inhibitory GABAergic interneurons, including mu opioid, nicotinic and cannabinoid CB₁ receptors. It is important to remember, however, that, owing to the sensitivity limits of PET, findings on increases in dopamine from some substances of misuse as assessed using PET are limited and less robust (Reid Reference Reid and Lingford-Hughes2006; Nutt Reference Nutt, Lingford-Hughes and Erritzoe2015). Likewise, despite the prominent role of the dopaminergic system in addiction, neither dopaminergic agonists nor antagonists have been found to be clinically efficacious.

TABLE 2 Primary targets, acute and long-term effects of drugs of misuse (adapted from Lingford-Hughes Reference Lingford-Hughes, Watson and Kalk2010)

DAT, dopamine transporter; MAO, monoamine oxidase; Ach, acetylcholine; VTA, ventral tegmental area; GABA, γ-aminobutyric acid; NMDA, N-methyl-d-aspartate; GHB, γ-hydroxybutyric acid; CB, cannabinoid.

Aside from dopamine, other systems have also been implicated in addiction. For example, the endogenous endorphin system plays an important role in reward. PET imaging has shown that pleasure from alcohol or stimulants is associated with increased endogenous endorphin levels and stimulation of the mu opioid receptor (Colasanti Reference Colasanti, Searle and Long2012; Mitchell Reference Mitchell, O'Neil and Janabi2012; Turton Reference Turton, Myers and Mick2020). Through such mechanisms the cues that are associated with the drug become conditioned stimuli, gaining salience. Combined with a loss of top-down control, the motivational drive to seek and consume the drug of misuse becomes more intense, reinforcing drug-taking behaviour and craving (Beck Reference Beck, Schlagenhauf and Wüstenberg2009; Koob Reference Koob and Volkow2016).

Monetary incentive delay

The monetary incentive delay (MID) task is a validated task that involves periods of anticipated monetary loss, gain or a neutral outcome. fMRI measures the BOLD signal during both anticipation and outcome of monetary reward to enable examination of neural correlates of probabilistic non-drug reward and associated learning.

In a neuroimaging meta-analysis of the MID task, the BOLD response to reward anticipation showed significant activation clusters in striatal–thalamic regions of the brain (Oldham Reference Oldham, Murawski and Fornito2018). A meta-analysis comparing the BOLD response in substance-dependent populations and controls confirmed blunted striatal activation during reward anticipation tasks in substance and gambling addictions (Luijten Reference Luijten, Schellekens and Kühn2017). Interestingly, Büchel et al (Reference Büchel, Peters and Banaschewski2017) found that blunted response to anticipated monetary reward in the ventral striatum and dorsolateral PFC in novelty-seeking young teenagers was predictive of problematic drug use at age 16. This suggests a mechanism for predisposition that could be identified prior to exposure to drugs. Finally, reduced striatal activation in response to monetary reward anticipation has been shown to be predictive of relapse in individuals with cocaine dependence; this blunted response has been shown to ‘normalise’ with increasing time spent abstinent (Balodis Reference Balodis, Kober and Worhunsky2016). Together, these findings identify blunting of the reward pathway as a potential biomarker both for vulnerability to substance misuse and risk of relapse.

Cue reactivity

Although substance-dependent individuals demonstrate a hyporesponse to conventional rewards, studies have also shown hyperactivation in their neural response to drug-related stimuli, which as described gain salience during regular drug consumption.

Neuroimaging studies using cue-reactivity tasks typically demonstrate involvement of a range of brain regions encompassing reward/motivation and learning/memory-related circuitry. Meta-analyses of cue-reactivity tasks in fMRI and PET studies suggest that similar patterns of neural activation are seen across alcohol, nicotine and cocaine dependence, with increased BOLD responses observed in the ventral striatum, anterior cingulate cortex (ACC) and amygdala compared with healthy controls (Chase Reference Chase, Eickhoff and Laird2011; Kühn Reference Kühn and Gallinat2011). Greater responses to salient drug stimuli in the dorsal PFC, dorsal striatum, thalamus, ACC and posterior cingulate cortex (PCC) were associated with an increased risk of relapse across alcohol, nicotine, cocaine and opioid dependence (Courtney Reference Courtney, Schacht and Hutchison2016). There is a wealth of evidence indicating that cue reactivity is modulated by individual factors and can be predictive of substance dependence severity, risk of relapse and treatment outcomes (Jasinska Reference Jasinska, Stein and Kaiser2014) (biomarkers are discussed further below).

Modulation of brain indices and biomarkers

Prediction of relapse

As observed in the cue-reactivity literature, neuroimaging markers of non-drug reward processing may also have utility in predicting clinical outcome. For example, in a non-drug reinforcement-learning task, methamphetamine users who had relapsed were observed to exhibit lower striatal activation in response to task outcomes relative to abstainers at 1 year (Stewart Reference Stewart, May and Poppa2014). Reduced striatal activation in response to monetary reward anticipation was similarly shown to be predictive of relapse in a longitudinal study involving participants with cocaine dependence and, interestingly, this blunted response was shown to ‘normalise’ with increasing time spent abstinent, highlighting the potential of this biomarker to track progress during recovery (Balodis Reference Balodis, Kober and Worhunsky2016).

Using a similar approach but with decision-making tasks, a role for the insula in decision-making deficits observed in addiction has been demonstrated (Gowin Reference Gowin, Mackey and Paulus2013). In an early study, baseline insula activation correctly predicted the relapse status of >90% of methamphetamine-dependent individuals (Paulus Reference Paulus, Tapert and Schuckit2005). In follow-up studies using a monetary reward task that incorporates a decision-making component, differential insula and dorsal striatal response to winning large risky, relative to small safe rewards could reliably predict relapse with better sensitivity and specificity than clinical variables (Gowin Reference Gowin, Harlé and Stewart2014, Reference Gowin, Ball and Wittmann2015).

Together these data suggest that attenuated response to non-drug reward and risk processing in striatal and/or insula areas may increase vulnerability to relapse. With further validation, such markers could potentially serve as targets for future therapeutic interventions and could help identify those most susceptible to relapse or those in most need of early intervention or additional resources to improve outcomes.

Clinical management: neurobiological translation and current issues

In the UK, effective treatments for substance dependence include psychosocial interventions and pharmacological therapy. Recommended management involves a combination of these, with multidisciplinary team involvement. Pharmacological therapies vary depending on the drug of dependence but can be broadly categorised for use during acute intoxication, substitution, detoxification (detox) or relapse prevention. Table 3 summarises the current treatment options and their mechanisms. For more information about prescribing, see the British Association of Pharmacology (BAP) guidelines (Lingford-Hughes Reference Lingford-Hughes, Welch and Peters2012).

TABLE 3 Current pharmacological interventions and their mechanisms of action for drugs of dependence

GABA, γ-aminobutyric acid; NMDA, N-methyl-d-aspartate; GHB, γ-hydroxybutyric acid; GBL, γ-butyrolactone; nAChR, nicotinic acetylcholine receptor.

Pharmacological treatment should be seen as adjunctive to psychosocial therapy for most substance use disorders. However, for specific substances, pharmacological therapy is highly likely to be required, such as benzodiazepines for medically assisted alcohol detoxification to reduce the symptoms of withdrawal and prevent complications of seizures, or management of opioid dependence with opiate substitution therapy. It should be noted that for stimulants (cocaine, amphetamine, etc.), cannabinoids, nitrous oxide, methylenedioxymethamphetamine (MDMA), hallucinogens and other novel psychoactive substances, there are no specific pharmacological therapies, but treatment during any phase of intoxication, detoxification or withdrawal should be supportive and medication for symptomatic relief (i.e. antipsychotics, anxiolytics, etc.) can be administered if it is safe to do so.

As previously described, substance dependence is multifactorial and extremely complex, with environmental, genetic, social and physical factors all playing a role in an individual's susceptibility and treatment needs. Moreover, abstinence rates remain low and relapse rates high for all substances of dependence, with a lack of specific and evidence-based relapse prevention options available. Randomised controlled trials (RCTs) investigating pharmacotherapy in addictions produce mixed results, or efficaciousness cannot be generalised to all patient subsets even for individuals using the same psychoactive substance, owing to likely heterogeneity of underlying neurobiology.

As mentioned above, alcohol consumption leads to the release of endogenous opioids that bind to mu opioid receptors, which in turn interact with the dopamine system of the brain, causing pleasurable and reinforcing effects. Naltrexone competitively binds to the mu opioid receptor and modulates the dopaminergic cell firing in the VTA, which would be consistent with it reducing the rewarding effects of alcohol. Several meta-analyses of RCTs comparing naltrexone have found a small but significant reduction in drinks per day, days of heavy drinking and reduction in craving, but there is no effect on improving continuous abstinence rates. Therefore, naltrexone appears to benefit dependent drinkers who want to remain abstinent but lapse and may also be beneficial for those who are non-dependent heavy drinkers who want to reduce their consumption.

In comparison, acamprosate, a glutamatergic N-methyl-d-aspartate (NMDA) receptor antagonist, attenuates the hyperglutamatergic state resulting from chronic alcohol use. Unlike naltrexone, there are limited neuroimaging studies of acamprosate. However, in a study using MRS, Umhau et al (Reference Umhau, Momenan and Schwandt2010) reported that acamprosate reduced glutamate levels in alcohol dependence, supporting the preclinical data. Several systematic reviews of RCTs conclude that, compared with placebo, acamprosate moderately increases duration of abstinence after detoxification. However, studies that included individuals with lower severity of dependence have found no effect on duration of abstinence, indicating that acamprosate might only be beneficial for those with severe dependence. Other factors, such as delay in starting acamprosate after detoxification, high physiological dependence at baseline, negative family history of alcoholism and serious anxiety or craving at baseline may predict that a response to acamprosate is less likely (for a review see Lingford-Hughes Reference Lingford-Hughes, Welch and Peters2012).

The dearth of clinically useful predictors for treatment and the limited range of medications highlight the substantial gaps that remain in our knowledge and ability to identify alcohol-dependent individuals who would respond best to different relapse prevention medications. Indeed, arguably the situation is worse for other substance dependences with fewer, or no, clinically efficacious medications available (for a review see Lingford-Hughes Reference Lingford-Hughes, Welch and Peters2012). In addition, being able to identify the best treatment and response for individuals dependent on more than one substance or with a coexisting psychiatric disorder would improve abstinence rates and quality of life, while decreasing the cost and burden of addictions on society as a whole.

Future directions

Functional neuroimaging has played a key role in delineating the neurobiology of addiction in humans, as it has done more broadly within psychiatric research. Our understanding of fMRI, higher spatial and temporal resolution, better image processing and statistical modelling, along with an increase in number of participants in studies, have resulted in greater confidence in interpretation and replicability of findings. Many of these issues also apply to PET; however, the critical limiting factor here is the lack of available robust tracers to probe targets of interest such as glutamate receptors. There is also now an emphasis on preregistration of neuroimaging studies, which involves committing to an analysis plan prior to the research being conducted. This avoids what are known as ‘fishing expeditions’, where hypotheses are formulated after results are known. Preregistration enables better quality of analysis and transparency of results. Important for future functional neuroimaging studies is a focus on the ‘three Rs’: replication, reliability and reproducibility.

Most fMRI studies currently use a region of interest (ROI) approach to test hypotheses concerning brain regions. Regarding addiction as a problem involving single brain regions greatly oversimplifies what is known to involve a complex interaction of multiple brain circuits. Recently, analysis techniques have evolved to include modelling of networks, requiring greater computational power and mathematical expertise, but with the potential to regard addiction as a problem of dysregulated networks rather than isolated brain regions. Future research in this area is likely to involve the formation and growth of multicentre cohort studies to provide insight into these questions, such as ENIGMA (http://enigma.ini.usc.edu/) and IMAGEN (https://imagen-europe.com/).

Although fMRI continues to provide insight into the underlying neural mechanisms involved and how these are modulated with treatment, several questions remain unanswered. There is growing evidence that neural circuitry plays a role in one's predisposition to addiction, and this needs to be explored in more detail. Additionally, many changes in neural activity associated with chronic substance use are not well characterised. For example, direct causality and effect of such neural changes and the involvement of environmental and genetic factors are still to be determined. It is also unknown whether changes in neural activity ‘normalise’ during abstinence and whether there is a role for pharmacotherapy to facilitate this process. Addressing these questions may confer a therapeutic benefit.

The frequency of polysubstance use is a factor that is likely to continue to be a barrier to the investigation of individual substances and their isolated effects. Continued investment in large-cohort longitudinal studies is likely to aid in answering these questions, in particular those that include well-defined addiction outcomes, as historically there have been inconsistencies in alcohol and substance use data collected by large multicentre studies, making research with these resources more difficult to employ.

Finally, in the UK the present segregation of addiction services from the NHS and systematic cuts in funding are ongoing barriers to successful clinical management that should not be overlooked (Bowden-Jones Reference Bowden-Jones, Sinclair and Lingford-Hughes2020). With most services currently being run by third-sector organisations, this has eliminated training posts for psychiatrists, including academic trainees, and inhibits general interaction with healthcare trusts and research bodies. The adverse impact of this on research must be countered if we are to optimise development of novel treatment and prediction tools to improve outcomes.