Understanding “Behavioral Addictions”: Insights from Research

Yvonne H.C. Yau, MSc
Sarah W. Yip, MSc, PhD
Marc N. Potenza, MD, PhD


The term “addiction” is derived from the Latin word addicere, meaning “bound to” or “enslaved by” (1). In its original formulation, the word was not linked to substance-use behaviors. As of several hundred years ago, the word became associated first with excessive alcohol and then with excessive drug use (2). By the time of the 1980s, the word was almost exclusively linked to excessive patterns of substance use, with experts involved in generating diagnostic criteria for drug dependence showing good agreement that the term applied to the condition of compulsive drug taking (35). However, by the early part of the 2000s, a growing movement to consider nondrug behaviors, such as gambling, sex, and eating (although by strict definition this last category does involve a substance—food), as addictive in nature was emerging (6). Aided by data from neurobiologic studies, this view is gaining momentum, leading to the consideration of a renaming of the “Substance-Related Disorders” diagnostic category to “Substance Use and Addictive Disorders” in the DSM-5 (1,711). Currently, only pathologic gambling (PG) has been proposed for inclusion in this category by the American Psychiatric Association (7). Other nondrug behaviors including problematic Internet use (PIU) and compulsive sexual behavior (CSB) (or hypersexual disorder) are proposed for inclusion in Section III, a section of the DSM-5 in which conditions that require further research will be included (12).

Addictions have been proposed to have several defining components: (a) continued engagement in a behavior despite adverse consequences; (b) diminished self-control over engagement in the behavior; (c) compulsive engagement in the behavior; and (d) an appetitive urge or craving state prior to the engagement in the behavior (1,13,14). If these elements are considered the core features of addictions, then excessive patterns of gambling and engagement in other non–substance-related motivated behaviors might be considered addictions. Consistently, the term “behavioral addictions” has been used recently to describe these disorders. Particularly relevant to addictions are aspects of motivation, reward processing, and decision making (1517), and these features represent potential endophenotypes, or underlying constructs that may be more closely linked to biologic processes than are diagnoses, per se, that could be pursued in biologic investigations across a spectrum of substance- and non–substance-related addictive disorders.

Many disorders termed “behavioral addictions” are currently categorized as impulse control disorders (ICDs). The goals of this chapter are to review similarities between ICDs and substance-use disorders (SUDs) and discuss the implications for treatment and theory. First, the category of ICDs in the DSM-IV-TR is reviewed. Evidence from neurobiologic research, clinical reports, and pharmacotherapy studies regarding ICDs and their relationships with SUDs is reviewed. In reviewing the individual ICDs, consideration is initially given to disorders for which there have been arguably the most data supporting similarities with SUDs (e.g., in the areas of gambling and eating), with later portions focusing on ICDs for which there are arguably less data to support a categorization as a “behavioral addiction” (e.g., pathologic skin picking [PSP]).


Many disorders that might be considered behavioral addictions are currently categorized in the Diagnostic and Statistical Manual as “Impulse Control Disorders Not Elsewhere Classified” (18). This title indicates that other disorders characterized by impaired impulse control, such as SUDs, attention deficit hyperactivity disorder (ADHD), bipolar disorder, cluster B personality disorders (i.e., antisocial, borderline, histrionic, narcissistic), among others, are categorized elsewhere in the manual. The ICDs within the “not elsewhere categorized” section include intermittent explosive disorder (IED), kleptomania, PG, pyromania, trichotillomania, and ICD not otherwise specified (NOS). Additional ICDs currently under consideration for DSM-5 include ones related to excessive Internet use, video game playing, buying or shopping, sex, and skin picking or nail biting (19). In addition, research criteria for binge-eating disorder (BED) are currently in the manual (20), and a case for inclusion of obesity has been forwarded (8), albeit debated (20a,20b). As discussed above, among issues being considered within research workgroups are whether the ICDs might be best categorized separately, with SUDs as addictions, or with obsessive–compulsive disorder (OCD) as obsessive–compulsive spectrum disorders (OCSDs) (8,10,19,21,22). Although the conceptualizations of ICDs as addictions or OCSDs are not mutually exclusive and data exist to support each formulation, these frameworks have important clinical implications with respect to prevention and treatment strategies (2325). In the following sections, we review the ICDs, their main clinical features, and what is currently known regarding their biologies. A particular emphasis is placed on PG as it is arguably the best studied of the ICDs to date. However, for a more detailed description of the clinical features of PG and its treatment, we direct you to the chapter entitled, Pathologic Gambling: Clinical Characteristics and Treatment (26).


PG has been considered a chronic, lifelong condition, although more recent data are challenging this notion (1,24,27). PG and SUDs share clinical characteristics and diagnostic criteria. Individuals with PG often experience withdrawal, craving, tolerance, and failed attempts to reduce or abate gambling behaviors—all common features of SUDs. The DSM-IV diagnostic criteria for PG reflect these similarities. A diagnosis of PG requires that the patient display five or more of the following: preoccupation with gambling; gambling with greater amounts of money to receive the same level of desired experience (tolerance); repeated, unsuccessful attempts to reduce or quit gambling; is restless/irritable when trying to stop gambling (withdrawal); gambles to escape from a dysphoric state; gambles to regain gambling-related losses (“chases” losses); lies in significant relationships about gambling; engages in illegal activity in order to fund gambling; has risked/lost a job or significant other due to gambling; and relies on others to fund gambling. Additionally, an exclusionary criterion exists to specify that the gambling is not better accounted for by manic episodes.

A Nonsubstance Addiction?

SUDs and PG share phenomenologic features. As with SUDs, PG often begins in adolescence and young adulthood, and prevalence estimates of PG tend to decrease across the life span (28). A common feature of SUDs and PG is “telescoping”—the phenomenon whereby the time between initiation and problematic engagement in the addictive behavior is shorter in females than in males (28). These commonalities suggest that there may be shared developmental vulnerabilities for both disorders. In particular, the high prevalence estimates in adolescents (29) suggest common vulnerability factors (e.g., impaired impulse control) for both SUDs and PG. There additionally exist high comorbidity rates for PG and SUDs, with co-occurrence estimates as high as 39.0% for comorbid substance abuse in clinical populations (30) and elevated odds (as high as 76.3%) observed in community samples as well (31,32).


Neurocognitive measures provide insight into neurobiologic functioning. For example, neurocognitive research suggests a dysregulation of the ventromedial prefrontal cortex (vmPFC) and orbitofrontal cortex (OFC) in individuals with PG (33,34). Individuals with PG display impaired performance on the Iowa gambling task (IGT), a task involving risk/reward decision making (35,36). Other populations with impaired performance on the IGT include individuals with SUDs, schizophrenia, or vmPFC lesions. Goudriaan et al. (37) reported that individuals with PG and individuals with a history of alcohol dependence both had impaired performance on neurocognitive tasks involving inhibition, time estimation, cognitive estimation, and planning tasks in comparison to normal controls and to individuals with Tourette syndrome, who only had an impaired performance on inhibition tasks. Neurocognitive measures of disinhibition and decision making are also positively associated with problem-gambling severity (38) and may predict the relapse of PG (39). Both individuals with problem gambling and those with alcohol dependence display alterations in risky decision making and reflection impulsivity—processes involving the vmPFC—in comparison to matched controls (40). These data highlight some of the clinical similarities between PG and SUDs and suggest similarities in underlying neurobiologic deficits. As such, neurocognitive research may be a useful tool in the identification of brain regions of interest warranting further investigation via more direct imaging-based modalities.

Delay Discounting

Individuals with PG often make disadvantageous decisions, selecting small immediate rewards over larger delayed rewards (e.g., “I’ll go gambling one last time.”). This rapid temporal discounting of rewards has been termed delay discounting, as rewards are more steeply discounted as a function of delay duration (41). Poorer performance on delay discounting measures (i.e., discount rewards more rapidly) has been observed in multiple populations, including individuals with SUDs (42). A dose-dependent relationship has been found between level of alcohol use and patterns of delay discounting (42). A similar additive effect has also been reported among heroin users in relation to needle sharing: users who engaged in needle sharing scored more highly than did non– needle-sharing heroin users on delay-discounting measures (42). Together, these data suggest that multiple risk factors may contribute to rates of delay discounting. Individuals with comorbid SUDs and PG discount rewards more rapidly than do PG individuals without SUDs (41,43). It is unclear whether comorbid SUDs and PG promote delay discounting or whether delay discounting is a vulnerability factor for comorbidity.

Although some data exist to suggest that abstinent substance users perform better (display less delay discounting) than do current substance users on delay-discounting measures, other data suggest no significant differences (42). A recent study suggests that delayed discounting did not differ between PG individuals before treatment and 1 year after treatment (44). To our knowledge, this is the first study that investigates delayed discounting among abstinent PG individuals. Further research into the effect of abstinence on delay discounting is needed.

Delay discounting involves aspects of reward evaluation, and multiple brain regions contribute to reward processing in humans (45,46). Amongst the most widely implicated brain regions in subjective reward valuation is the nucleus accumbens (NAcc), situated in the ventral striatum (4749). In healthy controls, the anticipation of working for or receiving monetary rewards is associated with ventral striatal activation, whereas an increase in vmPFC activation is associated with the processing of actual reward outcomes during performance of a monetary incentive delay task (50,51). An effective functional balance between these reciprocally connected neural regions may mediate appropriate and advantageous behavioral responses to varying reward contingencies. A recent functional magnetic resonance imaging (fMRI) study reported increased delay discounting among pathologic gamblers as well as an association between decreased activations in the ventral striatum and OFC and probability discounting compared to matched controls (46). Preclinical research suggests that diminished serotonin (5-HT) activity in the forebrain may hypersensitize animals to delays, influencing patterns of delay discounting (52). Further research is needed to determine the relationship among 5-HT regulation, prefrontal cortical activation, and delay-discounting behavior in humans.


Corticostriatal and forebrain neuromodulatory systems have been implicated in delayed and probabilistic reward (i.e., delay discounting) (52). Research suggests that dysregulation of the mesocorticolimbic dopamine (MCL DA) system may contribute to PG. The MCL DA system, often referred to as the “reward pathway,” has long been implicated in SUDs (53,54). Given the phenomenologic (e.g., craving, tolerance) and neurocognitive (e.g., delay discounting) similarities between PG and SUDs, it is possible that these disorders may share similar neurobiologic abnormalities, and current investigations are examining this hypothesis.

There have been several fMRI studies of PG that suggest that specific brain regions contribute to the pathophysiology of PG. For example, Potenza et al. (55) have reported reduced frontal cortical, basal ganglia, and thalamic activations in response to gambling videos—during the period prior to the subjective onset of emotional or motivational response—among PG subjects (vs. controls). This finding differed from those observed in similar studies involving subjects with OCD in which relatively increased activation of these regions was observed in the patient group. In particular, when viewing the portion of the videotapes during which the most robust gambling stimuli (e.g., video clips of a gambling public service announcement involving slot machines or an advertisement for a casino in which table gambling is shown) were presented, individuals with PG displayed relatively less activation of the vmPFC. Diminished activations of the vmPFC among individuals with PG as compared to control subjects have also been observed in other studies across a range of fMRI tasks, for example, the Stroop color–word interference task (56), a monetary incentive delay task (57), and a simulated gambling task (53). Similarly, individuals with SUDs either with or without co-occurring PG show relatively diminished activation of the vmPFC during performance of the IGT (58), in comparison to control subjects. Together, these data indicate an important role for the vmPFC in PG.

Multiple studies also implicate the striatum in PG. For example, Reuter et al. (53) observed significantly less vmPFC and ventral striatal activation in PG participants compared to control subjects during a simulated gambling task. Both right ventral striatal activation and vmPFC activation were inversely correlated with severity of gambling symptomatology in PG subjects, indicating that the less the activation of these brain regions, the greater the gambling pathology. Other fMRI studies during gambling-cue exposure have similarly observed diminished activation in the ventral (33) and dorsal striatum (59) in PG participants (vs. controls). As dopamine (DA) is an important neurotransmitter for both vmPFC and ventral striatal functioning, these findings appear consistent with the hypothesis of a dysregulation of DA among individuals with PG (53). However, as these studies did not directly measure DA levels, further investigation into the relationship between DA and PG in relation to the reward pathways is needed to test this hypothesis.


In recent years, multiple studies have investigated the neurostructural correlates of psychiatric disorders using either diffusion tensor imaging (DTI) or voxel-based morphometry (VBM) to assess white and gray matter structures, respectively. Findings from DTI studies indicate similar alterations in white matter microstructures encompassing regions of callosal, association, and projection fiber tracts between PG (60,61) and SUDs (6265). Among individuals with PG, a history of previous alcohol abuse or dependence is associated with a greater magnitude of white matter micro-structural alterations within regions of the corpus callosum, and white matter integrity within the corpus genu is associated with increased levels of self-reported impulsivity (61). Together, these data provide preliminary evidence for white matter involvement in the pathophysiology of PG.

To our knowledge, there have been only two published VBM studies conducted among individuals with PG (66). In both studies, no significant differences were reported in gray matter volumes between individuals with and without PG. Van Holst et al. additionally reported significant increases in gray matter volumes among individuals with PG, in comparison to individuals with alcohol dependence. These data therefore do not support the general hypothesis of altered gray matter macrostructures in PG; however, future studies employing larger samples should examine further gray matter contributions to PG and explore the relationship between gray matter volumes and clinical variables among individuals with PG.


Serotonin (5-HT)

Serotonin neurons project from the raphe nucleus of the brain stem to multiple brain regions including the hippocampus, amygdala, and prefrontal cortex (PFC). It has been hypothesized that dysregulated 5-HT functioning may mediate behavioral inhibition and impulsivity in PG (34,67,68). Data from studies of cerebrospinal fluid (CSF), pharmacologic challenge studies, and preclinical investigations together suggest a role for 5-HT in PG.

Low CSF levels of the 5-HT metabolite 5-hydroxyin-doleacetic acid (5-HIAA) have been reported in individuals with PG (69). Low CSF levels of 5-HIAA in humans have been associated with violence, suicidality, and impulsive aggression (52) and observed in other psychiatric disorders including other ICDs and alcohol abuse/dependence. Evidence from preclinical research has also identified a correlation between risk-taking behaviors and lowered CSF levels of 5-HIAA in monkeys (52) and rats (70). Low levels of platelet monoamine oxidase (MAO) activity, considered a peripheral marker of 5-HT activity, have been reported among males with PG (71,72). Consistent with findings of 5-HIAA CSF levels, lowered levels of platelet MAO have additionally been reported in both suicidal and risk-taking individuals (73).

5-HT receptor sensitivity has been investigated via the administration of the 5-HT1/5-HT2 receptor partial agonist meta-chlorophenylpiperazine (m-CPP). Individuals with PG, like those with SUDs, report a euphoric or “high” response to the m-CPP, whereas control subjects report an unpleasant response (74), along with an enhanced prolactin response (75). In PG, this differential response was associated with severity of PG symptomatology, with higher scores on the Yale-Brown Obsessive–Compulsive Scale Modified for PG, a PG severity rating scale, significantly correlating with increased prolactin responses. In response to the 5-HT1B/1D receptor agonist sumatriptan, a blunted growth hormone and prolactin response was observed among PG individuals (vs. controls), a response that suggests decreased 5-HT receptor sensitivity (76), similar to that observed among alcohol-dependent individuals (77). Furthermore, disturbances in 5-HT function, as reflected by blunted prolactin response to m-CPP, appear associated with severity of drug use among cocaine-dependent individuals (78). Taken together, these studies suggest that responses to 5-HT agonist administration are similar to those reported for substance-related addictions.

Other pharmacologic studies support the hypothesis that there is a dysregulation of the 5-HT system in PG. Among individuals with PG, 5-HT1B receptor availability within both the ventral striatum/pallidum and the anterior cingulate is positively correlated with problem-gambling severity (as assessed using the South Oaks Gambling Screen) (79). In some studies, selective serotonin reuptake inhibitors (SSRIs), such as fluvoxamine and paroxetine, have been found to improve social functioning and reduce gambling behaviors and thoughts about gambling in PG (80). However, clinical trial findings involving SSRIs have generally been negative or mixed (8084). As such, the precise role for SSRIs in the treatment of PG remains an active area of investigation (68,85).


DA has been implicated in rewarding and reinforcing processes in drug addiction. There have been several recent ligand-based studies investigating DA functioning among individuals with PG versus healthy controls. For example, Linnet et al. (86) have reported in a very small sample a positive association between self-reported excitement levels during IGT performance and DA release within the ventral striatum among individuals with PG. Joutsa et al. (87) reported that whereas individuals with PG and healthy controls showed similar levels of DA release during slot-machine-task performance, DA release was positively correlated with problem-gambling severity among PG individuals. Similarly, there have been two recent reports of no alterations in overall DA receptor binding within the striatum among individuals with PG (88,89), despite significant correlations between D3 receptor binding and both problem-gambling severity and impulsivity within the dorsal striatum (88), as well as between mood-related impulsivity and D2/D3 receptor binding within the striatum (89) among individuals with PG. Taken together, these data suggest that the clinical presentation of PG may be partially mediated by DA (e.g., DA release is associated with gambling-related excitement and PG severity) but do not support the hypothesis of a general alteration in DA receptor availability among individuals with PG.

Psychopharmacologic data suggest that the DA system may influence impulsive behavior, although the precise manner is not completely understood. Stimulants such as amphetamine increase DA release and prevent DA uptake within the synaptic cleft and lead to improved impulse control in individuals with ADHD (52). However, amphetamine administration in PG has been associated with the priming of gambling motivations (90,91). Whereas DA agonists have been associated with PG and other ICDs in the treatment of Parkinson disease (PD) (9294), the DA D2-like receptor antagonist haloperidol has been reported to enhance the rewarding and priming effects of gambling in PG (90), though individual differences also appear important (95). Investigations of CSF DA levels in PG have also yielded equivocal findings. Decreased CSF levels of DA and increased levels of the DA metabolites homovanillic acid (HVA) and 3,4-dihydroxyphenylacetic acid have been reported in PG (96). However, these findings were no longer present when correcting for CSF flow rate (69). A more recent investigation reported increased levels of CSF HVA among individuals with PG (97). The authors additionally reported enhanced levels of 5-HIAA, a finding different from their earlier investigations of CSF 5-HT metabolite concentration levels (69). These data highlight the need for studies using larger carefully controlled samples, while accounting for mediating factors such as comorbid pathologies and CSF flow rate.

Norepinephrine and Arousal in PG

Dysregulation of norepinephrine (NE)—a neurotransmitter implicated in arousal-, attention-, and sensation-seeking behaviors—has been reported in individuals with PG. Individuals with PG have elevated urinary concentrations of NE, as well as elevated CSF levels of a metabolite of NE (metabolite 4-hydroxy-3-methoxyphenyl glycol) (98). Research supports dysregulated hypothalamic–pituitary–adrenal axis regulation in PG. Meyer et al. (99) measured the neuroendocrine responses to “real-life” casino gambling in problem gamblers and found that problem gamblers had higher heart rates and elevated NE and DA levels in comparison to controls. In a separate study, PG individuals (vs. controls) showed a higher growth hormone peak response to clonidine—an adrenergic agonist used to assess NE function—the magnitude of which was positively correlated with problem-gambling severity (100). Together, these data suggest that there may be an elevation of NE activity in PG that may be potentiated by gambling behaviors.


Pharmacologic challenge studies suggest a dysregulation of the opioid system in PG. Naltrexone and nalmefene, both opioid receptor antagonists, have been found to reduce gambling-related thoughts and behaviors in individuals with PG (101104). Kim et al. (101) conducted a double-blind study of naltrexone in a sample of 89 individuals with PG and found that naltrexone (vs. placebo) administration at a mean dosage of 188 mg/d reduced subjective craving reports. Consistent with previous research demonstrating naltrexone’s dose-dependent hepatotoxicity, approximately 20% of participants displayed liver function test abnormalities subsequent to naltrexone treatment (101,102). A randomized double-blind study of nalmefene, an opioid antagonist that has not been associated with hepatotoxicity, found significant improvement in PG symptoms subsequent to lose-dose (25 mg/d) treatment in comparison to controls (102); moreover, a post hoc study suggests that medication dosage appears to be important in achieving symptom control (104). Together, these data suggest an important role for opioid antagonists in the treatment of PG as there is in alcohol and opiate use disorders.

Population Genetics

Family- and twin-based studies of addiction indicate that genetic factors are important in the development of drug and alcohol abuse. Family studies of PG suggest a significant parental influence on the development of offspring gambling behaviors (105). Gene/environment data suggest that the cross-generation transmission of drug addiction involves both environmental and genetic factors. Data derived from the Vietnam Twin Era Registry found that between 35% and 54% of the probability of meeting criteria for a DSM-III-R diagnosis of PG were attributable to heritable factors (106). Data from the same sample reported that 34% of the probability of developing drug dependence was attributable to inherited factors, suggesting similar degrees of heritability for PG and SUDs (107). Another study using the same population found that PG and alcohol dependence shared genetic and environmental contributions (108). Shared genetic and environmental factors may not be limited to SUDs, with both contributing to externalizing conditions like conduct and antisocial personality disorders as well as internalizing conditions like anxiety disorders, although genetic factors may contribute more prominently to the co-occurrence of PG and internalizing disorders (109).

Emerging data suggest that genetic factors may contribute substantially to the development of disordered gambling (DG)—a term that refers to the full continuum of gambling-related problems, including PG. Slutske et al. (110) reported that 49% of the probability of developing DG can be attributed to genetic influences and argue that there was no evidence for shared environmental influences. Moreover, correlations between DG symptoms and lifetime major depression, cigarette smoking, and alcohol and nicotine dependence are largely genetically determined (83%) (111).

Genetic Polymorphisms in PG

Molecular genetic research has identified specific genetic alleles associated with PG. Genetic polymorphisms related to genes encoding for the DA-related moieties (DRD1Ddel, DRD2 Taq I A, DRD4 [exon III]) (112116) have been implicated in PG, though negative results have also been observed (117). Variants of the serotonin transporter gene promoter region (5HTTLPR) (73) and MAOA enzymes (MAO-A [intron I], MAO-A [promoter], MAO-B [intron II]) (71,72) have been reported in association with PG. Similarities with respect to allelic distributions have been reported in PG and drug and alcohol dependence; for example, variations in DRD2 and MAOA genes have been linked to PG and alcohol abuse and dependence, and DRD4 variants have been linked to PG and alcohol, cocaine, and heroin abuse and dependence (118). These and other molecular genetic studies of PG should be considered preliminary in nature given relatively small samples, incomplete subject characterization, and frequent absence of stratification by race/ethnicity (119,120), particularly as some early results have not replicated in studies using alternate designs (e.g., discordant sibling pairs) (113).


Although PG was initially included in the DSM in 1980, the precise neurobiology of PG remains incompletely understood. However, growing evidence from neurocognitive, neurochemical, neuroimaging, and genetic research suggest that PG may share similar pathophysiologies with disorders involving poor impulse control, such as SUDs. There are important treatment implications inherent in these similarities. For example, the biologic similarities between SUDs and PG could help to guide treatment development.


Although specific drugs of abuse have different influences on eating behaviors (e.g., weight loss with amphetamine and increased food intake with marijuana), data suggest that both substance-use and eating behaviors may be modulated by the same motivational neurocircuitry leading to the conceptualization of “foods as drugs” (121). The DSM-IV-TR category of eating disorders includes anorexia nervosa (AN), bulimia nervosa (BN), and eating disorder not otherwise specified. Although not specifically included in the main portion of the DSM-IV-TR, BED—distinct from BN as it does not include compensatory behaviors such as purging—is associated with obesity and other negative sequelae. The recent rise in the prevalence of obesity within the population has led to more research on obesity and related disorders like BED. Given that BED has an important element of episodic behavioral dyscontrol similar to the formal ICDs, this section focuses primarily on the neurobiology of this disorder. For a review of AN and BN, see Ref. (122). Preclinical data relevant to BED and obesity will also be described.


Previously listed as an eating disorder NOS, BED is to have its own diagnostic code in the DSM-5 (123). The proposed diagnostic features are very similar to the diagnostic features of SUDs and ICDs: recurrent episodes, impaired control, and marked distress in relation to binge eating. The DSM-IV-TR also notes that individuals may make repeated unsuccessful attempts to stop binge eating and that they may report that their binge eating has detrimental social and occupational effects—two important criteria for substance dependence. Over a quarter of clinicians report often or always using addiction-based therapies for BED (124), further suggesting phenomenologic and clinical similarities between BED and SUDs.

Obesity, a common consequence of binge eating, is increasingly common. Defined as “abnormal or excessive fat accumulation that may impair health” (125), it is estimated that up to 35.7% of the U.S. population meets the criteria for obesity (126). Globally, estimates from 2008 suggest that 1.4 billion adults were overweight and that at least 200 million men and 300 million women were obese (125). Associated medical conditions include type II diabetes, stroke, osteoarthritis, heart disease, and cancer (125,127). Frequently co-occurring psychiatric disorders include depression, anxiety, personality disorders, and lifetime SUDs (128). Features of PG among individuals with BED are associated with decreased self-esteem and increased substance-use problems (129).

While there are strong links between BED and obesity (130), it is important to note that obesity and addiction are not necessarily related and that for some individuals, overeating is a relatively passive event that takes the form of liberal snacking and eating large portions (131). BED may be a specific subtype of obesity driven by a biologically based hyperreactivity to the hedonic properties of food, coupled with an enhanced motivation to encourage appetitive behaviors that is related to neurobiologic differences in reward and control circuitry (132134).

Conceptualization and Treatment

Along with the recent rise in the prevalence of obesity, there has been a concurrent rise in the number and diversity of treatment interventions. Such interventions range from preventative interventions, such as the incorporation of nutrition classes into school curriculums, to pharmaceutical interventions, such as the administration of appetite-suppressing drugs and use of surgical innervations like gastric bypass surgery. Because some treatment interventions for obesity are highly invasive (e.g., gastric bypass, jaw wiring), it is important to examine and assess not only efficacy but also tolerability and impact on quality of life. It is also important to understand the pathophysiology underlying the disorder, particularly if individual differences contribute to selection of effective interventions.

The Biology of Eating Behaviors

The hypothalamus (via regulatory neuropeptides such as leptin, cholecystokinin (CCK), and ghrelin) is an important site for the maintenance of energy homeostasis. Lesions to the ventromedial hypothalamus result in hyperphagia and obesity, whereas lesions to the lateral hypothalamus (LH) result in hypophagia and weight loss. This finding led researchers to conceptualize the LH as a “feeding center” and the ventromedial nucleus as a “satiety center.” In this “dual-center” model of eating behavior, the hypothalamus plays a central role and hypothalamic function contributes significantly to homeostatic regulation within a larger motivational network (135).


Leptin, an adipose-derived hormone, is a chemical modulator involved in the maintenance of energy homeostasis and feeding behaviors (136). Leptin is also implicated in other reward-seeking behaviors (136,137) including SUDs (138). Leptin acts as a peripheral metabolic cue within the central nervous system to modulate neuronal activity in brain areas involved in appetite control, including the hypothalamus. Administration of exogenous leptin increases energy expenditure and reduces hyperphagia and obesity in genetically leptin-deficient mice and humans (ob/ob) (139). However, leptin deficiency syndrome is extremely rare in humans (140). A recent fMRI study investigated three adults with genetically mediated leptin deficiency 5 and 6 years post– leptin replacement treatment. A longer duration of replacement treatment was associated with increased activation in a ventral portion of the posterior cerebellum in response to food images, whereas decreases in body mass were associated with decreased activation in this area (141). Another fMRI study similarly presented visual food and nonfood stimuli during a fasting state and a fed state, both before and after leptin treatment, to two adolescents. Results yielded significant behavioral and neural response changes between conditions. When participants rated their level of preference for specific foods presented, leptin administration was associated with decreased preference ratings during the fed condition. Whereas activation in the NAcc and caudate nucleus of the striatum was positively correlated with preference ratings in both the fed and fasting conditions prior to leptin treatment, subsequent to leptin treatment striatal activation was only positively associated with preference ratings in the fasting condition (142). These data demonstrate extrahypothalamic action of exogenous leptin and suggest that leptin may help to encode palatability.

Leptin and DA-Regulated Reward Processing

Mesolimbic reward pathways play an important role in eating behaviors. Increased activation of brain regions implicated in substance use and dependence—such as the OFC, insula, striatum, and midbrain—have been observed subsequent to the consumption of palatable food (143). Differential striatal activation in response to food has also been reported in obese individuals in comparison to healthy control subjects. An fMRI study of 13 obese females and 13 female controls found that obese (vs. control) females selectively activated the dorsal striatum in response to images of high-calorie foods (144). Moreover, in response to high-calorie food stimuli, body mass index (BMI) was predictive of dorsal striatum, anterior insula, claustrum, posterior cingulate, and postcentral and lateral OFC activation (144). A separate study found that obese versus lean individuals showed elevated corticostriatal responses to favorite-food cues, with thalamic activation mediating the relationship between insulin resistance and food craving in the obese but not in the lean individuals (145). Based on previous reports of lowered DA levels in AN (146), one study investigating potential striatal differences found that whereas healthy control subjects displayed differential ventral striatal activation for rewards versus losses in a monetary reward task, women who had recovered from anorexia showed no activation differences for wins and losses, suggesting a reduced experience of reward (147). A more recent study reported increased striatal activations to both pleasant and aversive food stimuli among women recovered from anorexia (vs. healthy controls), further suggesting that alterations in reward processing may be implicated in the pathophysiology of disordered eating behaviors (148).

Preclinical research using pharmacologic and genetic knockout strategies suggests that D2 receptor blockade attenuates the acute hypophagic effect of leptin in fasted mice (149). A reduced availability of DA D2 receptors in the striatum has been reported in individuals with SUDs (150,151) and in obese individuals, with receptor availability negatively correlated with BMI (152,153). Preclinical research has demonstrated that knockdown of striatal DA D2 receptors rapidly accelerates the development of addiction-like reward deficits and the onset of compulsive-like food seeking in rats (154). Together, these data suggest dys-regulated striatal function in eating disorders that at least in part reflects dysregulated DA function.

Whereas leptin-modulated hypothalamic activity has been well documented, research suggests that leptin acts directly on other brain regions, including the substantia nigra pars compacta (SNc) and ventral tegmental area (VTA) of the midbrain. DA neurons in the VTA and SNc project to the striatum and are implicated in reward, motivation, and addiction. Preclinical data (155) have demonstrated that DA neurons within the VTA express mRNA encoding for both leptin and insulin receptors. Exogenous leptin administration to the VTA results in a decrease in food consumption, and intravenous exogenous leptin administration reduces the firing of VTA DA neurons (156). The expression of mRNA encoding the long form of the leptin receptor (ObRb) has additionally been reported in regions including the hippocampus, brain stem, cortex, thalamus, cerebellum, and substantia nigra (157). Leptin has additionally been reported to enhance synaptic plasticity in brain regions such as the hippocampus (158).

Preclinical research using self-administration of electrical brain stimulation of reward (BSR) technology suggests that leptin influences drug-taking behaviors. An investigation by Fulton et al. (159) found increased BSR in response to fasting conditions, which was attenuated by direct exogenous leptin administration. A different study found that leptin-deficient ob/ob mice have a lessened locomotor response to amphetamine administration, which is corrected by leptin administration (159). These data suggest that, in addition to its metabolic function, leptin may help to modulate mesolimbic reward circuits that may relate to both palatability and substance use.


Partially modulated by adipose-derived hormones such as leptin and ghrelin, the hypothalamic neuropeptides orexin-A and orexin-B—also referred to as hypocretin 1 and hypo-cretin 2—are important modulators of eating behavior and help to maintain energy homeostasis. The hypothalamus is the primary site of hypocretin-containing neurons, though these neurons project to other brain regions (160162). Orexin administration has been demonstrated to increase feeding behaviors (161), while orexin antagonists have been shown to impair operant responses to food reinforcers (163) in preclinical populations.

Preclinical research has demonstrated that administration of orexin-A reinstates cocaine-seeking behaviors in a dose-dependent manner (164). Administration of orexin-A was also associated with increases in BSR, suggesting a negative regulation of reward circuits (164), consistent with orexin neurons’ activation in chronic morphine administration and precipitated morphine withdrawal (165). Similar research has demonstrated that the administration of an orexin receptor agonist abolishes reinstatement of cue-induced alcohol-seeking (166) and heroin-seeking (167) behaviors in rodents, further implicating orexin in substance-seeking behaviors.

Preclinical research also indicates a significant increase in hypothalamic orexin-containing neurons subsequent to preference conditioning for food, cocaine, or morphine, suggesting important similarities between the development of food and drug preferences (168). Whereas the number of orexin-containing neurons was positively correlated with increases in preference, no such correlations were found for any other type of hypothalamic neurons. This study additionally reported reinstatement of morphine preference subsequent to direct orexin administration to the VTA, one of the brain regions receiving projections from hypothalamic orexin neurons (168).

Orexin may directly influence DA neurons in the VTA. Pharmacologic blockade of orexin-1 receptors (Ox1r) within the VTA—but not within the paraventricular thalamus— dose-dependently attenuated cue-induced reinstatement of cocaine seeking (169). Direct administration of orexin-A or orexin-B produces locomotor-enhancing effects in mice that were prevented by prior administration of a DA receptor antagonist (170). The same study additionally demonstrated a lack of hyperlocomotion and a significantly lessened increase in DA in response to morphine (170). Increases in PFC DA subsequent to orexin-A administration to the VTA have also been reported (171). Together, these data suggest that orexin may directly influence mesolimbic DA pathways implicated in reward and drug addiction. However, further research investigating the relationship between orexinergic and dopaminergic signaling in clinical populations is needed.


Ghrelin is a gastrointestinal hormone that helps to maintain energy homeostasis and may contribute importantly to the initiation of eating. Unlike leptin and orexin, which are anorexigenic, ghrelin is orexigenic and increases food intake and body weight (172). One study measured ghrelin levels in the plasma of healthy controls during a 24-hour period and found significant increases in ghrelin levels prior to meal initiation followed by significant decreases after consumption (173). Reduced levels of circulating ghrelin have been reported in obese individuals, with an inverse correlation between BMI and ghrelin levels observed (174). Interestingly, preclinical research suggests that whereas ghrelin administration increases the motivation to eat, it does not alter perceived food palatability (175).

Although ghrelin is primarily synthesized in the stomach, recent research suggests that it may also mediate feeding behaviors via direct action on certain brain regions. Preclinical research has identified a ghrelin receptor, growth hormone secretagogue 1 receptor (GHSR), in the hypothalamus and VTA. Ghrelin has been linked to increased synapse formation and DA turnover in the NAcc (176). Administration of exogenous ghrelin in the VTA prompted feeding behavior, and GHSR antagonist administration reduced feeding subsequent to food deprivation (176). Direct administration of ghrelin into the VTA, but not into the NAcc, was reported to motivate behavior for sucrose reward in an operant conditioning paradigm in rats, suggesting that ghrelin signaling within the VTA contributes to incentive-motivated behavior for a food reward (177). Ghrelin administration–induced increases in motivation to eat are eliminated following pretreatment with a DA D1 receptor antagonist, suggesting that the orexigenic effects of ghrelin are mediated by DA signaling (175). Conversely, ghrelin-deficient mice display attenuated responses to chronic and acute cocaine administration, indicating that dopaminergic neurotransmission is disrupted by deletion of the ghrelin gene (178). Overall, these findings suggest that ghrelin may help to modulate the warding properties of food via interaction with DA neurons.

The Nucleus Accumbens: Opioid and Endocannabinoid Encoding of Palatability

Research suggests that opioid receptors in the NAcc region of the ventral striatum may be particularly important for the encoding of food palatability. Preclinical research suggests that DA release in the NAcc modulates excitatory NAcc neuronal activity in response to previously learned reward-associated cues, thereby increasing the likelihood of cue responsiveness (e.g., performance of behaviors learned via operant conditioning) (179). Stimulation of NAcc opioid receptors has been found to increase food intake (180). Administration of opioid receptor antagonists, such as naloxone or naltrexone, extinguishes previously established preferences for sweetened versus unsweetened water in rats, whereas morphine has been demonstrated to enhance palatability and preference for sweet food (180,181). Importantly, such preclinical data demonstrate opioid involvement in the palatability—or reward value—encoding of food that does not appear to directly affect caloric intake. Rather, the opioid system may be involved in general reward processing, as opposed to specific appetitive control. These data nonetheless suggest shared neurobiologic mechanisms in eating and substance-use behaviors. Despite evidence implicating the opioid system in eating behaviors, pharmaceutical challenge studies of opioid receptor antagonists such as naltrex-one have yielded equivocal results in human populations, and further research is needed to establish appropriate pharmacotherapies for the treatment of BED.

Preclinical research has identified two distinct populations of NAcc neurons responsive to food consumption: a group of neurons displaying primarily excitatory responses to increased palatability (i.e., increased sucrose) and a separate group of neurons displaying primarily inhibitory responses prior to initiation of feeding behaviors that does not appear to be related to palatability (182). This finding is consistent with other preclinical research demonstrating increases in nonappetitive food intake subsequent to NAcc inhibition (183). One possible interpretation of these data is that—in addition to neurons encoding for reward—a group of NAcc neurons may also be implicated in general habit formation or basic motor control of feeding behaviors, independent of palatability, although this hypothesis requires further testing. These data further support the hypothesis that the NAcc is an important brain region for the encoding of palatability as well as the initiation of feeding responses, and highlight similarities between eating and substance-use behaviors.

Human and animal studies also implicate the endocannabinoid system (composed of cannabinoid receptors, endocannabinoids, and associated enzymes) in eating behaviors (184,185). The endocannabinoid system is composed of two endogenous ligands, anandamide (arachidonoylethanolamide) and 2-arachidonoylglycerol (2-AG), and two cannabinoid receptors, CB1 and CB2 (186). Cannabinoid receptors in the NAcc have been implicated in appetitive behavior.

Preclinical research suggests that CB1 receptors partially mediated by leptin (187) are involved in the presynaptic modulation of release for the neurotransmitters GABA, glutamate, DA, noradrenaline, and serotonin (184,188) and has identified cannabinoid receptors (CB1/2) in the limbic forebrain, striatum, and NAcc (reviewed in Ref. (189)). Colocalization of opioid and CB1 receptors in the striatum has also been reported (reviewed in Ref. (189)). Preclinical investigations have additionally reported increases in feeding behaviors subsequent to administration of delta9-THC and anandamide (an endogenous cannabinoid neurotransmitter) (reviewed in Ref. (189)). Cannabinoids have long been associated with rewarding psychotropic effects and are additionally associated with increases in food intake. In addition to cannabinoid-induced increases in feeding behaviors, studies have implicated endogenous cannabinoids in the experience and encoding of food-associated reward. For example, direct anandamide administration to the NAcc shell has been found to significantly enhance “hedonistic” reward and increase feeding behaviors in rats (189). Elevated anandamide blood levels have been reported in individuals with AN and individuals with BED, whereas no such elevation has been reported in individuals with BN (186). Interestingly, a recent study suggests that endocannabinoid administration increased gustatory nerve response to sweeteners in a concentration-dependent manner without affecting responses to other tasting compounds (e.g., salty, sour, bitter, and savory) (190). The cannabinoid 1 receptor (CB1R) is thought to be a promising drug target for antiobesity medication, and the CB1R antagonist/inverse agonist, rimonabant, has been shown to significantly reduce body weight in humans; however, it is associated with psychiatric side effects including anxiety, depression, and suicidal tendencies (191,192). Recent preclinical research suggests that the neutral CB1R antagonist NESS0327 is as effective as rimonabant in reducing weight gain and food intake and lacks potentially harmful effects on anxiety and motivation (193).

Prefrontal Cortex

Research from neuroimaging, neurocognitive, and lesion studies implicates prefrontal cortical modulation of eating behaviors. Frontotemporal dementia (FTD), a degenerative disorder involving atrophy of frontal, insular, and temporal cortical regions, is characterized by behavioral changes in eating and sexual behaviors (194) as well as deficits in insight, empathy, and social interaction (194,195). Clinically reported changes in eating behaviors associated with FTD include increases in weight, food cravings/obsessions, and gluttony (195). Similarly, a recent study found that individuals with FTD meeting criteria for “gluttonous” overeating during “all-you-can-eat” 1-hour meal sessions had significantly increased atrophy of the OFC, right ventral insula, and striatum (196).

Evidence derived from positron emission tomography (PET) research additionally suggests a relationship between frontal lobe activity and eating behaviors. In weight loss studies, “successful” as compared to “nonsuccessful” weight-loss dieters/maintainers had significantly greater activation in dorsal prefrontal cortex (DPFC), dorsal striatum, and anterior cerebellar lobe brain regions following meal consumption (197) and food visualization (198). Conversely, nonsuccessful dieters had significantly greater OFC activation following meal consumption (197). These data suggest differential modulation of eating behaviors by prefrontal regions. Further investigation is required to fully understand interactions between PFC regions in relation to eating behaviors.

Consistent with the finding that greater dietary restraint is negatively correlated with OFC activation and positively correlated with DPFC activation, one randomized, double-blind, parallel group study using repetitive transcranial magnetic stimulation found reduced self-reported craving sensations in response to exposure to craving-inducing foods subsequent to left dorsolateral prefrontal cortex (dlPFC) stimulation (199). dlPFC stimulation has also been reported to be effective in reducing cravings for nicotine, alcohol, and cocaine, suggesting potentially similar neuro-biologic mechanisms for food and drug craving (reviewed in Ref. (200)).

Neurocognitive research additionally implicates frontal lobe involvement in binge eating. Similar to impulse control–related disorders (e.g., SUDs, PG), deficits in decision making have been reported in BED and obesity. In a sample of 41 healthy adult women, a tendency to overeat in response to stress and higher BMI both significantly predicted poorer IGT performance (201). In other studies, obese (vs. nonobese) individuals also performed significantly worse on the IGT and showed no improvement in performance over time (201203).

Similar decision-making deficits have been reported in both AN and BN. Individuals with AN have been reported to perform significantly worse on the IGT and displayed significant reduction in skin conductance response in comparison to healthy controls and recovered AN patients (ANR) (204). In contrast, Liao et al. (205) reported that while BN individuals similarly performed poorly on the IGT, unlike individuals with AN, they showed no reduction in skin conductance response. Similar performance deficits among individuals with BN were also reported in a study using the Game of Dice Task (206), a decision-making assessment that, unlike the IGT, provides explicit information of reward-loss contingencies (207).

Serotonin (5-HT)

Increases in both exogenous and endogenous serotonin (5-HT) are associated with a reduction of food intake and weight gain, and an increase in energy expenditure (208). In relation to eating behaviors, research has focused on 5-HT and the hypothalamus. 5-HT neurons located in the dorsal raphe nucleus receive direct projections from hypothalamic orexin neurons and express orexin-A and orexin-B receptors (reviewed in Ref. (209)). Hypothalamic 5-HT may in part mediate the experience of satiety. Medial hypothalamic 5-HT is implicated in the temporal management of eating behavior, in particular with meal termination, as opposed to initiation. Preclinical research has demonstrated that d-fenfluramine (d-FEN)—also known as “Fen-Phen” when combined with phentermine—an exogenous agent that increases 5-HT release while also blocking reuptake, may exert its anorexigenic effects via 5HT2C receptor activation of proopiomelanocortin (POMC) neurons in lateral hypothalamic regions (210). Further research is needed to establish the precise role of the central melanocortin system in the regulation of food intake. Such data may help to develop alternative pharmacotherapies for BED and other eating disorders (210).

5-HT is also thought to be implicated in food preference. For example, preclinical studies have demonstrated that the injection of either exogenous 5-HT or drugs that increase 5-HT availability (such as fluoxetine) into the medial hypothalamus selectively inhibits carbohydrate intake but has no significant effect on fat or protein intake (211). Conversely, elevated levels of tryptophan (TRP), a 5-HT amino acid precursor and hypothalamic 5-HT, are associated with high-carbohydrate intake (reviewed in Ref. (208,212)).

SSRIs have been shown to be efficacious in treating eating disorders with fluoxetine approved by the Food and Drug Administration for the treatment of BN and sibutramine approved for the treatment of obesity (213). Some research suggests that SSRIs are effective in targeting binge eating, psychiatric, and weight symptoms (214), although the effectiveness and duration of these medications remain under debate. In an open-label study, Leombruni et al. (215,216) reported that BED patients who display decreased binge eating and weight loss following initial SSRI administration maintained these beneficial effects for 6 months with continuation of SSRI treatment. In a randomized, double-blind, 12-week study of escitalopram versus placebo for the treatment of individuals with comorbid BED and obesity, individuals receiving high-dose escitalopram treatment had significantly greater reductions in weight, BMI, and total global illness severity (217). However, no significant between-group differences were found for the variables reduction of binge episodes, reduction of days with a binge episode, and reduction of obsessive–compulsive features of BED. In a 2-year follow-up of fluoxetine-treated individuals with BED, no significant improvements in BED symptoms were reported, despite significant improvements in depressive symptoms (218).


Binge eating and comorbid obesity are increasingly common phenomena with wide-ranging public health implications. Recent neurobiologic findings, such as the involvement of the adipose-derived hormone leptin in the dopaminergic reward system, suggest that BED is a brain-based disorder that may share many of the same neurobiologic features as SUDs. Such findings have important treatment implications, and further investigation is required in order to optimize treatment interventions.


CSB is characterized by excessive engagement in normative sexual behaviors. Not included in the DSM-IV, CSB is often referred to as sexual addiction, hypersexual disorder (the proposed diagnostic entity for DSM-5), or sexual impulsivity.

Clinically relevant sexual behaviors may be divided into paraphilic and nonparaphilic behaviors. In paraphilic sexual behaviors, there is a disturbance in the object selection (e.g., an animal, unwilling person, inanimate object). In nonparaphilic sexual behaviors, the individual engages in socially normative sexual behaviors in an excessive, obsessive, or compulsive manner, without a disturbance of object choice. Paraphilic disorders are a distinct category of disorders, already included in the DSM-IV-TR, and are outside the scope of this chapter.


Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Dec 12, 2016 | Posted by in GENERAL & FAMILY MEDICINE | Comments Off on Understanding “Behavioral Addictions”: Insights from Research

Full access? Get Clinical Tree

Get Clinical Tree app for offline access