Neurobiology of Gambling Behaviors

Abstract

For many, gambling is a recreational activity that is performed periodically without ill effects, but for some, gambling may interfere with life functioning. A diagnostic entity, pathological gambling, is currently used to define a condition marked by excessive and problematic gambling. In this review, the current status of understanding of the neurobiologies of gambling and pathological gambling is described. Multiple neurotransmitter systems (norepinephrine, serotonin, dopamine, opioid and glutamate) and brain regions (ventral striatum, ventromedial prefrontal cortex, insula, among others) have been implicated in gambling and pathological gambling. Considerations for future directions in gambling research, with a view towards translating neurobiological advances into more effective prevention and treatment strategies, are discussed.

Introduction

Anticipated Changes in Gambling Disorder Nomenclature, Definition, and Classification

Gambling, defined as placing something of value at risk in the hopes of gaining something of greater value, is a widely prevalent behavior that has been popular for millennia. While most people gamble without problems, a sizable minority (up to about 5%) develops gambling-related problems. Those with more severe gambling problems may meet criteria for pathological gambling (PG), a diagnostic entity introduced in the Diagnostic and Statistical Manual (DSM) in 1980 [1]. In anticipation of the forthcoming editions of the DSM (DSM-5), several changes to PG have been proposed. First, the name of the condition is being proposed to change to “Gambling Disorder”, as some have found the term pathological stigmatizing. Second, the threshold for diagnosing a case is proposed to shift from 5 of 10 inclusionary criteria to 4 of 9, with the “commission of gambling-related illegal acts” criterion removed. For comparison, a diagnosis of a substance-use disorder in DSM-5 may require meeting 2 or more inclusionary criteria. Third, PG is being proposed for reclassification into the category of “Addictions and Related Disorders”, a move from its current location currently in “Impulse Control Disorders Not Elsewhere Classified” [2]. This proposed reclassification is based on data from epidemiological, clinical and neurobiological domains [3, 4]. In that PG may be conceptualized as an addiction without the drug, investigations into its neurobiology have often been based on understandings of drug addictions and/or have contrasted PG with substance-use disorders [5, 6**].

Given the recently proposed changes to PG, this article will review the neurobiology of gambling and PG within the context of this changing landscape and will discuss future directions for research, with a view towards translating improved neurobiological understandings into better prevention and treatment strategies. Recent articles have reviewed in depth the neurobiology of PG and its relationship with the neurobiologies of substance-use disorders and impulse-control disorders [6**, 7**]. Here, we will summarize main findings from neurochemical and neural research of PG and subsyndromal gambling, using the tables to provide detailed information, and consider future directions.

Neurochemistry of Gambling and PG

Multiple neurochemical systems have been implicated in gambling and PG (Table 1). Adrenergic systems have been hypothesized to contribute to arousal and excitement, serotonin to impulse control, dopamine to rewarding and reinforcing aspects, opioids to pleasure/urges, cortisol to stress responsiveness, and glutamate to cognitive functioning including cognitive flexibility. Data exist consistent with these proposed roles (see Table 1), although there appears to be a more complex relationship (e.g., drugs that act as D2-like dopamine receptor agonists and antagonists have both been linked to pro-gambling motivations and behaviors, suggesting that individual differences might be particularly important to consider [8]). Arguably the most empirically validated form of pharmacotherapy for PG involves opioid antagonists, although these appear efficacious for only a portion of individuals with PG and have not been examined for long-term efficacy [9**]. As such, there exists significant need for translating information about the neurobiological underpinnings of PG into treatment advances.

Table 1.

Roles for Neurochemicals/Neurotransmitters in Gambling and PG

Neurochemicals/ Neurotransmitters	Study Modality	Hypothesized Role	Main Findings
Norepinephrine	CSF, Blood, Urine, Drug, fMRI	Arousal, Excitement	Elevated levels of norepinephrine and metabolites in PG versus non-PG men (reviewed in [9**]); fMRI suggests differential amygdalar activation to yohimbine (an alpha-2 adrenergic antagonist) in PG and non-PG subjects [35]
Serotonin	CSF, Blood, Drug, RCT, PET	Behavioral Initiation and Behavioral Cessation (Impulse Control)	Diminished CSF levels of serotonin metabolite 5-hydroxyindole- acetic acid in PG; different behavioral and biochemical responses in PG and non-PG subjects to metachlorophenylpiperazine (m-CPP), a serotonergic partial agonist at 5HT1 and 5HT2 receptors; mixed responses to serotonergic drugs (serotonin reuptake inhibitors) in PG; 5HT1B receptor binding in ventral striatum/putamen and anterior cingulated linked to problem-gambling severity in PG (reviewed in [9**])
Dopamine	CSF, Blood, Urine, Drug, PET	Reward Processing, Reward-Based Learning, Reinforcement	Elevated levels of dopamine metabolites observed in PG versus non-PG samples; dopamine increases observed during gambling; D2-like dopamine antagonists linked to increased gambling motivations in PG; D2-like dopamine agonists associated with PG and other ICDs in PD; Ventral striatal differences in D2-like receptor binding observed in PD with PG versus PD without PG but not in non-PD populations; in PD, dopamine agonist treatment linked to differential brain function in PG and non-PG groups during card-game tasks; D2-like receptor binding linked to problemgambling severity and positive urgency in PG and differentially to decision-making in PG and non-PG populations ([36, 37**, 38]; reviewed in [9**, 39])
Opioid	RCT	Pleasure, Urges	Clinical trials indicate opioid antagonists (naltrexone and nalmefene) superior to placebo in the treatment of PG, and particularly helpful for individuals with strong gambling urges at treatment onset and those with a familial history of alcoholism (reviewed in [9**])
Cortisol	Blood	Stress	Elevated cortisol during gambling in PG and non-PG subjects (reviewed in [9**])
Glutamate	RCT	Compulsiveness, Cognitive Inflexibility	Preliminary data suggest that glutamatergic drugs (RCT of n-acetyl cysteine) superior to placebo in the treatment of PG, and that treatment with a glutamatergic drug (openlabel memantine) may improve cognitive function/flexibility (reviewed in [9**])

Abbreviations: CSF = cerebrospinal fluid; D2-like = D2 family of dopamine receptors (D2, D3 and D4); fMRI = functional magnetic resonance imaging; RCT = randomized clinical trial; PET = positron emission tomography; PG = pathological gambling; RCT = randomized clinical trial

Neural Biologies of Gambling and PG

Multiple cognitive processes have been linked to propensities to gamble and to the severity of gambling problems (Table 2). Although individuals with PG often score within normal range on intelligence tests and may show fewer cognitive deficits as compared to individuals with substance addictions [10], they often score highly on measures of impulsiveness and low on measures of self-control [6]. Additional behavioral measures (e.g., assessing decision-making and reversal learning) may capture deficits in PG not apparent in standard assessments of intelligence (Table 2). Differences in brain function relating to multiple cognitive domains (cognitive control, decision-making, reward/loss and “near-miss” processing, delay and probabilistic discounting, reversal learning, alternation learning, and risk-taking) have all been linked to PG or problem-gambling severity. Much of this literature supports relatively diminished activation of the prefrontal cortex (particularly ventromedial but also ventrolateral and orbitofrontal) and subcortical regions (particularly the ventral striatum) (Table 2). However, reports of relatively increased activation exist, particularly in response to gambling cues, although the findings are not uniform across studies (Table 2). In addition to the regional brain activation findings tabulated, preliminary data have identified regions of poorer white matter integrity in PG, with poorer integrity linked to behavioral tendencies in PG (e.g., fun-seeking tendencies) [11, 12]. These findings suggest that brain circuits, particularly those involving ventral prefrontal, ventral striatal and limbic regions, may contribute importantly to reward-related decision-making that drive gambling behaviors and underlie PG.

Table 2.

Cognitive and Neural Underpinnings of Gambling and PG

Cognitive Process	Measures	Groups	Main Findings
Simulated Gambling	fMRI, PET	PG, non-PG; PD with PG, PD without PG	PG versus non-PG subjects show diminished vmPFC and striatal activation during simulated gambling, with degree of activation correlating inversely with problem-gambling severity [40]; differential patterns of thalamic, IFG and temporal gyral activations in PG versus non-PG [41]; in PD subjects with PG versus PD subjects without PD, greater baseline availability of D2-like dopamine receptors in ventral striatum and greater displacement during simulated gambling/decisionmaking task [42]
Cue Exposure	fMRI	PG, problem gamblers, non- PG	PG versus non-PG subjects show relatively diminished activation of vmPFC, ventral striatum, cingulate, insula, IFG and other regions during gambling cue exposure [5, 43]; self-reported gambling urges correlate with activations in the temporal pole in PG [44]; increased cortical and subcortical responses to gambling cue exposure in PG [45] and increased activation of cortical (visual processing) and subcortical (amygdala, parahippocampus) regions in problem-gambling subjects [46]
Cognitive Control	fMRI	PG, non-PG PG	versus non-PG subjects show relatively diminished activation of vmPFC [47]; activation of vmPFC, ventral striatum, and other regions correlate with treatment outcome in PG [29*]
Decision-making	fMRI, PET	Problem- gambling/SUD, SUD, non- problemgambling/SUD; PG, non-PG	Diminished vmPFC and other cortical activation in problem- gambling/SUD group compared to non-problem-gambling/SUD group [48]; striatal dopamine release during decision-making in PG associated with excitement in PG [49]; different relationships between striatal dopamine release and decision-making in PG and non-PG subjects [50]; see also “simulated gambling” above
Reward/Loss Processing	fMRI	PG, non-PG, OCD	PG versus non-PG subjects show relatively diminished activation of ventral striatum, insula, vmPFC and other regions [51*]; problem gamblers showed relatively greater (compared to non-PG subjects) activation of ventral striatum to high-value versus low-value reward [52*]; relatively diminished striatal activation in PG versus non-PG and OCD subjects [53*]
Delay and Probability Discounting	fMRI	PG, non-PG PG	versus non-PG show differences in discounting that relate to neural activations in “reward” regions (vmPFC, ventral striatum, and others) and problem-gambling severity [54**]
Reversal Learning	fMRI	Problem- gambling, non- PG	Problem-gambling versus non-PG subjects showed increased response perseveration and reduced activation of right ventrolateral prefrontal cortex [55]
Alternation Learning	fMRI	PG, non-PG	In PG, problem-gambling severity correlates with cortical (prefrontal and orbitofrontal) activations during alternation learning [56]
Loss-Chasing	fMRI	Non-PG	Decisions to chase losses versus decisions to quit associated with greater activation of dorsal anterior cingulate and less activation of vmPFC [57]
Near-Miss	fMRI	PG, Non-PG	In non-PG, near-miss events recruited “win” circuitry (insula, ventral striatum) and rostral anterior cingulate activation varied with a measure of control [58]; problem-gambling severity related to midbrain responses to near-misses [59]; PG as compared to non-PG subjects showed differences in midbrain activation and activation patterns similar to wins [60]
Risk-taking	fMRI	PG and other ICDs in PD, PD	PD subjects with ICDs as compared to PD subjects without ICDs showed perfusion of the ventral striatum and less activation of the ventral striatum during risk-taking [61]

Abbreviations: D2-like = D2 family of dopamine receptors (D2, D3 and D4); fMRI= functional magnetic resonance imaging; ICDs = impulse control disorder; IFG = inferior frontal gyrus; PET = positron emission tomography; PD = Parkinson’s disease; PG = pathological gambling; SUDs = substance-use disorders; vmPFC = ventromedial prefrontal cortex

Important Next Steps: Translating Neurobiological Findings into Prevention and Treatment Advances

The proposed change in threshold for making a diagnosis of a gambling disorder in DSM-5 raises questions about where to “draw the line” for making a diagnosis and whether clinically relevant levels of problem-gambling severity exist. Different thresholds have been proposed, ranging from non-gambling to PG (Figure 1) [13]. Different thresholds for defining groups based on problem-gambling severity have been identified using data-driven methodologies (e.g., latent class analysis [14]) or creation of groups based on numbers of inclusionary criteria for PG acknowledged (e.g., [15]), and data indicate that the different levels of problem-gambling severity shown in figure 1 have clinical relevance (e.g., with respect to non-gambling psychiatric conditions [16]). Data also indicate that people transition through different stages based on levels of problem-gambling severity [17], although the factors that promote transitions from lesser to greater levels of problem-gambling severity or vice-versa are not well understood and represent an important and needed area of investigation.

Figure 1. Proposed Levels of Problem-Gambling Severity and Possible Factors Contributing to Problem-Gambling Severity and Gambling-Related Transitions.

Schematically diagrammed are different stages of problem-gambling severity ranging from non-gambling to pathological gambling. Many factors likely influence decisions not to gamble as well as the nature and extent of gambling, including participation in specific levels of gambling falling along a spectrum of problem-gambling severity (pictured on top). Domains of factors with specific examples of factors within each domain are provided (pictured on bottom). Domains may operate in main or interactive manners (indicated by a main bidirectional arrow that contains within it multiple bi-directional arrows), and may do so in ways that vary across development. Domains may influence any of the identified severity levels or stages, as schematized with the multiple bi-directional arrows.

Multiple factors likely influence these transitions. For example, women have been found to demonstrate faster transitions from gambling initiation to gambling problems (a process termed “telescoping” [18]), although the biological factors underlying the telescoping process (or sex differences in the neurobiology of PG) have yet to be examined systematically. Life experiences may also influence transitions in both detrimental and protective fashions. For example, resiliency factors may include being married, having strong social networks, being more educated, having larger incomes, participating in religious services utilizing protective cognitive and behavioral strategies [19]. On the other hand, risk factors may include trauma and stress (including early childhood adversity), early exposure to and initiation in gambling, truant or delinquent behaviors, impulsive tendencies, substance-use behaviors, and other psychiatric or medical conditions [19-21]. Importantly, these risk and protective factors may vary and operate biologically in developmentally sensitive fashions, making the consideration of people’s developmental statuses important in understanding gambling transitions and trajectories [22**].

These factors may also interact in important manners. Commonly occurring allelic variants may demonstrate clinically relevant interactions with environmental factors to confer risk or protection in manners that could influence gambling transitions. For example, the gene coding for the serotonin transporter (SLC6A4) has commonly occurring short and long variations in the promoter region that lead to differential protein expression and are associated with different patterns of amygdala responsiveness and vulnerability to depression based on childhood maltreatment [23, 24]. The identification of specific epigenetic factors underlying such vulnerability and resilience and the extent to which such factors may influence gambling transitions and trajectories warrant direct investigation. In order to investigate such phenomena more precisely, a better understanding of the genetic underpinnings of gambling and PG is warranted. Although data suggest that 50% or more of the variance for PG is related to genetic factors [25, 26], candidate gene analyses have generated mixed results and more systematic investigations (e.g., genome-wide analyses) have yet to be published [7**].

Multiple constructs (e.g., impulsivity or facets thereof) have been proposed as important intermediary phenotypes or potential endophenotypes for gambling and PG [6**]. Specific aspects of impulsivity (e.g., relating to rapid response and choice) may be dissociable, and within each domain, self-report and behavioral measures may relate differently to clinically relevant measures like treatment outcome [22**]. These constructs may also be sensitive to environmental factors (e.g., stress, reward), and thus contextual considerations warrant examination. Impulsivity-related constructs may be multi-faceted and associate in specific fashions with gambling transitions and trajectories at different developmental epochs [22]. As temperamental and related behavioral characteristics at early ages have been linked prospectively to multiple behaviors and phenomena (substance use, academic performance, and brain responses) [27], the extent to which such factors relate to gambling behaviors should be examined directly as such factors may influence prevention and policy efforts (e.g. targeting children with high impulsiveness).

An important line of research in addictions involves understanding the brain mechanisms underlying change during treatment, and behavioral and pharmacological interventions may have shared or distinct mechanisms of action [28**]. Integrating neuroimaging measures into randomized clinical trials offers the opportunity to understand the mechanisms underlying effective treatments and their active ingredients. Such data could provide a stronger foundation for modifying and adapting therapies or combinations thereof in order to improve their efficacies. While research in this area is further along in substance addictions than in PG, mechanisms for active ingredients of behavioral therapies for PG have been proposed, along with the hypothesized neural systems that may underlie their actions [29**]. Further research is needed to examine these hypotheses and to translate the findings into improved treatments.

Future Directions

As an increasingly large and diverse array of biotechnologies permit the acquisition of data from multiple domains, it is increasingly important to gather and synthesize information from multiple areas in order to understand gambling behaviors. Arguably, this work has progressed further in understanding non-gambling conditions (e.g., those relating to depression or substance-use disorders) than it has in understanding gambling behaviors. Important areas to consider include the integration of genetic and neuroimaging measures and the concurrent and integrative use of multiple imaging modalities (e.g., diffusion tensor imaging, high-resolution structural scanning, functional magnetic resonance imaging, and positron emission tomography). Critically important in this work is the inclusion of careful phenotypic assessments (diagnostic, behavioral, and self-report) that can be linked to imaging measures in order to understand how individual differences might relate to biological measures underlying gambling behaviors.

Advanced analytical techniques also deserve attention, particularly those that consider connectivity or identify brain circuitry. Independent component analysis has been used in preliminary examinations to identify functionally integrated activations (or brain circuits) that relate to cognitive control differently in cocaine-dependent and non-addicted comparison subjects; within the cocaine-dependent subjects, relationships with treatment retention and drug abstinence were also examined [30*]. The identification of circuitry relating specifically (and perhaps differentially) to pathology and treatment outcome should improve our understanding of treatment mechanisms and lead to better therapies.

Translational efforts involving animal models of gambling behaviors have only begun to be examined in gambling. Rat models examining near-miss and gambling task phenomena have been developed and used to examine neurochemical contributions to these processes [31, 32]. Given advances in targeted gene expression in rodents [33], such techniques may help better understand how specific molecular entities within specific cells and neuronal circuits influence gambling behaviors. As animal research permits alternative and complementary avenues of investigation, it should help promote a better understanding of the pathophysiology of PG and identify novel targets for treatment development.

Using neuroscience to influence policy remains a goal for gambling. While specific avenues of research have direct policy implications (e.g., findings linking lottery-ticket-gift receipt to problem-gambling behaviors and more permissive attitudes towards gambling [34]), investigations into the biological factors relating to PG and gambling transitions may hold implications as well. Important in this regard is understanding better the factors that relate gambling, particularly PG, to psychiatric, neurological and medical illnesses, in addition to the biological factors that may promote transitions to levels of greater/lesser problem-gambling severity.

Conclusions

While the past decade has witnessed a significant advance in our understanding of the neurobiology of gambling and PG, there remain many poorly understood aspects. Important gaps exist with respect to integrating information from multiple domains and understanding how specific factors relate (perhaps in interactive fashions) to gambling transitions and trajectories across the lifespan. While empirically validated treatments for PG have varying degrees of support, little is known about their mechanisms of action or how specific therapies might work better for specific individuals. Generating improved prevention, treatment and policy efforts from a better understanding of the neurobiology of gambling and PG seems an attainable future goal.

Highlights.

1. Noradrenergic, serotonergic, dopaminergic and other systems may influence gambling.

2. Ventral prefrontal cortical and ventral striatal systems relate to gambling.

3. Translating biological understandings of gambling to interventions is important.