New dimensions in human laboratory models of addiction

Drug addiction has been conceptualized as a disorder that involves elements of both impulsivity and compulsivity. Impulsivity can be defined behaviorally as a tendency toward rapid, unintended reactions to internal and external stimuli without regard for the negative consequences of these reactions. Compulsivity can be defined as an element of behavior that results in perseveration in responding in the face of adverse consequences or perseveration in the face of incorrect responses in choice situations and is associated with exaggerated response inhibition. From a psychiatric perspective, impulsivity can be characterized by an increasing sense of tension or arousal before committing an impulsive act, and pleasure, gratification or relief at the time of committing the act. Compulsivity can be characterized by anxiety and stress before committing a compulsive repetitive behavior and relief from the stress by performing the compulsive behavior (American Psychiatric Association 1994). Collapsing the cycles of impulsivity and compulsivity yields a composite addiction cycle composed of three stages: preoccupation/anticipation, binge/intoxication and withdrawal/negative affect (Table 1). Impulsivity often dominates at the early stages and compulsivity dominates at terminal stages. As an individual moves from impulsivity to compulsivity, a shift occurs, from positive reinforcement driving the motivated behavior to negative reinforcement driving the motivated behavior (Koob 2004). These three stages are conceptualized as interacting with each other, with the negative reinforcement component becoming more intense, and ultimately leading to the pathological state known as addiction (Koob & Le Moal 1997). The addiction cycle construct provides an excellent framework by which to interpret animal models for studying the etiology of the addiction process, neurobiology of addiction and treatment of addiction. Importantly, human laboratory studies—the subject of this special issue of Addiction Biology—can provide a powerful means of exploring treatment targets for specific components of the addiction cycle independent of expensive double-blind, placebo-controlled trials.

Animal models for addiction are some of the most robust in neurobiology and have provided a heuristic framework not only for understanding the basic neurobiology of the actions of drugs of abuse, but also for providing a pipeline of potential targets for addiction treatment. While no animal model of addiction fully emulates the human condition, the models do permit investigation of the specific elements of the addiction process. Such elements can be defined by models of actual symptoms of addiction within different stages of the addiction cycle. Different animal models for the study of the neurobiology of addiction can be superimposed on the same three stages of the addiction cycle described above.

Animal models for the binge/intoxication stage of the addiction cycle incorporate the construct of immediate, positive drug reinforcement. Animal models of drugs as reinforcers are extensive and well-validated. Drugs of abuse have powerful reinforcing properties, and animals and humans readily self-administer drugs in the non-dependent state. Animals will perform many different tasks and procedures to obtain the drugs, even in the non-dependent state. Drugs that are self-administered by animals correspond well with those that have high abuse potential in humans, and drug self-administration is considered an animal model that is predictive of abuse potential (Collins et al. 1984). Two other animal models have been used extensively to measure drug reward indirectly: conditioned place preference and brain stimulation reward (Table 1). Conditioned place preference is a non-operant procedure for assessing the reinforcing efficacy of drugs using a classical or Pavlovian conditioning procedure. Animals typically exhibit a conditioned place preference for an environment associated with drugs that are self-administered by humans and avoid environments that induce aversive states (i.e. conditioned place aversion). Lowering of brain stimulation reward thresholds also is a reliable measure of drug reward. Drugs of abuse decrease thresholds for brain stimulation reward, and good correspondence exists between the ability of drugs to decrease brain reward thresholds and their abuse potential (Kornetsky & Bain 1990).

Animal models are specifically designed to address compulsivity-like elements of the binge/intoxication stage and the increased motivation to seek drugs. Such measures include binge drug-taking (Tornatzky & Miczek 2000), increased drug intake produced by extended access to drugs (Ahmed & Koob 1998) and drug-seeking despite negative or aversive consequences (Deroche-Gamonet, Belin & Piazza 2004; Vanderschuren & Everitt 2004) (Table 1). There are compelling studies showing drug-taking in the presence of aversive consequences in animals with extended access to the drug. Rats with extended access to cocaine do not suppress drug-seeking in the presence of an aversive conditioned stimulus, which has face validity for the DSM-IV criteria of ‘continued substance use despite knowledge of having a persistent physical or psychological problem’ (Vanderschuren & Everitt 2004).

For the withdrawal/negative affect stage, animal models exist for the somatic signs of withdrawal for virtually all drugs of abuse. However, more relevant for addiction are the animal models of components of the negative reinforcing effects of dependence that are beginning to be used to explore how the nervous system that is involved in motivation adapts to drug use. These include conditioned place aversion, elevations in reward thresholds and decreases in responding for non-drug rewards on a progressive-ratio schedule of reinforcement. Animal models of the negative reinforcing effects of dependence include the same models used for the rewarding effects of drugs of abuse (Table 1), but with changes in valence of the reward. Animals will show a conditioned place aversion to precipitated withdrawal from chronic administration of a drug. For example, in a conditioned place preference procedure, rodents avoid the compartment paired with an acute withdrawal syndrome precipitated by administration of an antagonist to the neurochemical system under study. Thus, rats will show a conditioned place aversion to precipitated opioid, cannabinoid and nicotine withdrawal; each precipitated by opioid, cannabinoid and nicotinic antagonists (Tzschentke 1998). Similarly, precipitated or spontaneous withdrawal from all drugs of abuse raise, rather than lower, brain reward thresholds (Markou, Kosten & Koob 1998). Increased self-administration in dependent animals presumably reflects both positive and negative reinforcement mechanisms and is a model that overlaps with the binge/intoxication stage. Dependence-induced increases in self-administration have now been observed with cocaine, methamphetamine, nicotine, heroin and alcohol (Ahmed & Koob 1998; Ahmed, Walker & Koob 2000; Roberts et al. 2000; Rimondini et al. 2002; Kitamura et al. 2006; O'Dell & Koob 2007).

Models for the preoccupation/anticipation stage involve drug-, cue- and stress-induced reinstatement of drug-seeking behavior (Table 1). Animal models of craving involve both the conditioned rewarding effects of drugs of abuse and measures of the conditioned aversive effects of dependence (Shippenberg & Koob 2002). Resistance to extinction has been used to measure the residual motivational properties of drugs by assessing the persistence of drug-seeking behavior in the absence of response-contingent drug availability. Other measures of craving have examined the motivational properties of the drugs themselves or cues paired with the drugs after extinction. Drug-induced reinstatement first involves extinction then presentation of a priming injection of a drug. Latency to reinitiate responding or the number of responses on the previously extinguished lever is hypothesized to reflect the motivation for drug-seeking. Similarly, drug-paired or drug-associated stimuli can reinitiate drug-seeking (cue-induced reinstatement). Other measures of craving include second-order schedules of reinforcement, in which animals can be trained to work for a previously neutral stimulus that predicts drug availability.

From the aversive perspective, acute stressors also can reinitiate drug-seeking behavior in animals that have been extinguished, and this is called stress-induced reinstatement. Increased brain reward thresholds and increases in anxiety-like behavior have been shown to persist after acute withdrawal in animals studies and have been hypothesized to reflect protracted abstinence, perhaps most comprehensively demonstrated for alcohol dependence (Roberts, Cole & Koob 1996; Roberts et al. 2000; Rimondini et al. 2002; O'Dell et al. 2004; Heilig & Koob 2007). Finally, with opioids, conditioned withdrawal has been observed in which previously neutral stimuli paired with precipitated opioid withdrawal have been shown to produce place aversions and to have motivational properties in increasing self-administration of opioids (Kenny et al. 2006).

Using the structure of the addiction cycle outlined above, human laboratory studies also can provide measures for each addiction stage and are hypothesized to have heuristic value for predicting potential treatment efficacy in these domains (Table 1). Although the predictive validity of human laboratory models remains to be determined, ongoing studies with established medications used for addiction treatment provide the ‘Rosetta Stone’ by which to evaluate the validity of the laboratory model and then to use it as a springboard for novel medication development. This issue of Addiction Biology builds on the initial pioneering work in human laboratory models and extends these models to real-world constructs such as vulnerability to addiction, impulsivity, craving and resistance to relapse. As such, human laboratory models are well on their way to providing elements of construct validity, the Holy Grail of validation in biology.

For the binge/intoxication stage of the addiction cycle, self-administration procedures for cocaine, heroin and marijuana have been established using largely operant responding in which participants who are dependent on the drug make a behavioral response such as pressing a key on a computer to receive a drug (Haney 2008). The procedures involve administering the drug in the pattern by which the drug is abused. Thus, smoked cocaine may be available every 15 minutes; in contrast, heroin may be available every 6 hours. Different schedules of reinforcement are used, such as fixed-ratio or progressive-ratio, or choice procedures are used that are similar to the animal models (Kelly et al. 1997; Haney, Foltin & Fischman 1998; Haney 2008). From a validation perspective, similar to animal models, heroin self-administration is reduced by all three medications approved by the Food and Drug Administration to treat opioid dependence: methadone, naltrexone and buprenorphine (Mello et al. 1981; Comer, Collins & Fischman 1997). However, a range of medications have been shown to reduce the subjective effects and craving associated with cocaine but do not decrease cocaine self-administration (Haney & Spealman 2008). These results are consistent with clinical data showing that of over 60 medications tested, none have proven reliably effective in clinical trials (Vocci & Elkashef 2005). To date, little or no work has been done with treatments for marijuana self-administration or clinical trials.

Endophenotypes of the binge/intoxication stage with potential for human laboratory study include the construct of impulsivity. Impulsivity can likely contribute to increasing the probability of engaging in initial drug intake, and the subsequent drug effects on impulsivity may increase impulsive behaviors that in turn facilitate further drug use, prolonging the binge or even provoking relapse (de Wit 2008). A variety of tasks are used to assess impulsivity (de Wit 2008). Commonly used behavioral measures of impulsivity include the task of delayed discounting (in which there is a relative preference for smaller, more immediate rewards over larger, more delayed rewards; Rachlin & Green 1972), tasks of behavioral inhibition (e.g. Stop Task; Logan 1994) and attentional measures (in which subjects show increased variability in reaction times on a simple reaction time task reflecting ‘lapses in attention’; Leth-Steensen, Elbaz & Douglas 2000). Given that longitudinal studies show that behavioral disinhibition predicts substance use disorders (Sher, Bartholow & Wood 2000), a reasonable hypothesis is that drug users may differ on measures of impulsivity (Sher & Trull 1994). Also relevant to the use of measures of impulsivity in human laboratory studies, acute and chronic administration of drugs of abuse produces effects on delayed discounting, behavioral inhibition and lapses in attention that impair impulse control even further (de Wit 2008). Acute administration of drugs of abuse has little effect on delayed discounting, but interestingly, acute drug withdrawal such as with heroin increases discounting (Giordano et al. 2002). Acute administration of drugs such as alcohol impaired performance in the Stop Task; in contrast, psychostimulants improved performance. D-amphetamine similarly improved performance on lapses of attention in a reaction time task (de Wit 2008). Thus, measures of impulsivity may provide unique insights into factors guiding initiation and early drug use; they may also prove useful for predicting impairments in relapse in acute withdrawal and protracted abstinence.

For the withdrawal/negative affect stage, negative reinforcement mechanisms are in effect, rather than positive reinforcement mechanisms. With negative reinforcement, individuals can take drugs to relieve the emotional withdrawal state such as with opioids and sometimes alcohol. However, a more likely scenario is that individuals administer the drugs to self-medicate the malaise or negative affect associated with protracted abstinence. Numerous human laboratory measures of acute withdrawal are available. Chronic treatment of nicotine withdrawal with nicotine substitution (Hjalmarson 1984) improves outcome, and methadone maintenance at appropriate levels of drug will block opioid-seeking behavior (Dole 1988). For marijuana, medications also have been tested on non-treatment-seeking volunteers who smoke marijuana repeatedly each day (Haney 2005). In the laboratory, marijuana withdrawal is alleviated by the resumption of marijuana smoking or by administration of oral Δ⁹-THC (Haney et al. 2004). Using this approach, the medication regimen that was most effective was the frequent administration of oral Δ⁹-THC (dronabinol) (Haney 2005). Whether such effective withdrawal treatments will translate to treatment of relapse in abstinent marijuana smokers remains a challenge for future research (Haney 2008).

Self-medication also overlaps with protracted abstinence. Indeed, nicotine may reduce negative affect while increasing cognitive efficiency (Kassel & Unrod 2000). In human laboratory studies, nicotine can have direct facilitative effects on cognitive processing in both non-abstaining smokers and non-smokers (Rusted et al. 2000; Lawrence, Ross & Stein 2002; Giessing et al. 2006). More sophisticated cognitive measures, neuro-imaging studies and more fine-tuned analyses such as task difficulty and cognitive load are suggested to show more robust improvements in cognitive function with nicotine (Evans & Drobes 2008).

Stress responses and changes in activity of the hypothalamic pituitary adrenal axis impact on all phases of the addiction cycle, but may be of particular relevance to the withdrawal negative affect stage and preoccupation/anticipation stage (Koob & Kreek 2007). Glucocorticoids facilitate dopaminergic activity, and high levels of glucocorticoids increase the vulnerability to drug-seeking in animal models (Piazza et al. 1989; Goeders 2002; Uhart & Wand 2008). Adverse childhood experiences such as emotional abuse are associated with increased initiation of substance abuse (Widom, Weiler & Cottler 1999). In dependent humans, active drug-taking and acute withdrawal are accompanied by increased cortisol levels; however, with chronic alcohol and drug use, there is impairment in the cortisol response that can extend into protracted abstinence. Subjects confronted with stressful situations following treatment are more likely to relapse than those not confronted with stressful experiences (Noone, Dua & Markham 1999; Uhart & Wand 2008). Thus, both stress and drugs of abuse activate the hypothalamic pituitary adrenal axis, but the glucocorticoid response becomes hypoactive with chronic high dose drug use. Concomitantly, there is activation of brain stress systems in the extended amygdala that can contribute to negative reinforcement mechanisms associated with withdrawal and protracted abstinence (Imaki et al. 1991), and the activation of the hypothalamic pituitary adrenal axis may be a key part of this transition to dependence. Indeed, it is well documented that high glucocorticoids can drive the brain stress systems in the amygdala (Koob & Kreek 2007). Thus, there may be a cascade of stress hormone interactions with drugs of abuse from facilitation of the binge/intoxication stage, to exaggeration of the withdrawal/negative affect stage to hypersensitivity of brain stress systems that contribute to maintaining the withdrawal-affect stage and sensitizing stress-induced relapse. All of these changes may be amenable to study in the human laboratory setting.

For the preoccupation/anticipation (craving) stage, three major factors are hypothesized to contribute to relapse: priming doses of drug, drug-associated cues and stressor exposure. Several human laboratory procedures have been developed to reflect these aspects of the preoccupation/anticipation stage. Drug reinstatement has been developed in human laboratory models. With smoking, following abstinence, participants smoked five cigarettes in their natural environment, and all reexposed individuals relapsed within 2 days (Chornock et al. 1992). Naltrexone given during the relapse period in dependent smokers reduced craving and reduced the number of cigarettes smoked during the free access period (King & Meyer 2000).

One well-studied model of the drug-associated cue component of craving is the cue-reactivity paradigm. Here, psychological and physiological reactivity to stimuli associated with drug-taking behavior are measured. Developed initially with smoking, craving states have been measured that are associated with presentation of cues for smoking, alcohol and cocaine (Carter & Tiffany 1999). Exposure to stimuli or cues associated with drug consumption produces urges to take the drug, conditioned appetitive responses and changes in autonomic responses.

Exposure to alcohol cues, such as the sight or smell of alcoholic beverages, reliably increases the urge to drink alcohol, increases salivation and increases attention to cues (Monti et al. 1999). Even more compelling, cue-reactivity can predict treatment outcome (Cooney et al. 1997) and has been validated in some cases using medications that successfully treat alcoholism. For example, naltrexone, but not topiramate, blocks cue-reactivity in alcohol-dependent subjects (Monti et al. 1999; Miranda et al. 2008), and nicotine replacement therapy decreases craving associated with smoking cues (Shiffman et al. 2003). Other drugs not currently in use in treatment have shown positive results with cue-induced reactivity paradigms, including carbamazepine for alcohol (Hersh, Bauer & Kranzler 1995) and amantadine for cocaine (Robbins, Ehrman & Childress 1992).

A real-world aspect to cue-reactivity is the association between drug intake, cues in the environment and vulnerability for addiction. With alcohol, a positive family history of alcohol use disorders is a strong predictor of future alcohol abuse and dependence. Although the genetic component is well established (Schuckit 1985), environmental interactions also provide a rich basis for human laboratory study. Alcohol cue-reactivity, such as alcohol-related advertisements, is an important predictor of alcohol use (Snyder et al. 2006). Personal drinking levels in college students predicted positive affective responses to alcohol pictures (i.e. experiencing pleasure while viewing a visual alcohol cue; Pulido et al. 2008).

A novel approach to cue-reactivity that was developed for the study of craving in alcoholism during protracted abstinence is the exploration of the interaction of cue exposure with emotional states during protracted abstinence (Mason et al. 2008). A non-treatment-seeking sample of alcohol-dependent subjects were exposed to affective stimuli that had positive or negative valence and then to a beverage cue. but with no opportunity to self-administer alcohol. Cue-reactivity was measured using subjective measures of craving, measures of emotional reactivity and psychophysiological measures including heart rate, skin conductance and facial electromyogram. Of the subjects who showed a cue-reactivity response that was one standard deviation from the mean, 12 subjects received gabapentin and 21 subjects received placebo. Alcohol exposure and both positive and negative emotional cues had the expected effects on subjective and emotional reactivity but less effects on psychophysiological measures (Mason et al. 2008). These results were similar to what one might expect to observe in the first week of a clinical trial. Gabapentin significantly decreased subjective craving, decreased craving evoked by stimuli, and improved several measures of sleep quality. These results suggest that cue-reactivity, combined with an emotional overlay, may provide a powerful means of evaluating potential medications for addiction treatment.

A similar translation from animals to humans was observed recently in a study of the substance P/ neurokinin-1 receptor system (George et al. 2008). Mice genetically deficient in the neurokinin-1 receptor (knockouts) showed a major decrease in voluntary alcohol consumption. Recently detoxified subjects with alcohol dependence and treated with a neurokinin-1 receptor antagonist showed decreased craving, blunted cortisol responses and decreased functional magnetic resonance imaging responses to affective stimuli.

Stress and stressors have long been associated with relapse and vulnerability to relapse (Marlatt & Gordon 1980; Koob & Kreek 2007). In human alcoholics, numerous symptoms that can be characterized by negative affect persist long after acute physical withdrawal from ethanol (Alling et al. 1982). These symptoms, post-acute withdrawal, tend to be affective in nature, subacute and often precede relapses (Hershon 1977; Annis, Sklar & Moser 1998). A factor analysis of Marlatt's relapse taxonomy found that negative emotion, which includes elements of anger, frustration, sadness, anxiety and guilt, was a key factor in relapse (Zywiak et al. 1996) and the leading precipitant of relapse in a large scale replication of Marlatt's taxonomy (Lowman, Allen & Stout 1996). Exposure to negative affect, stress or withdrawal-related distress also increases drug craving (Childress et al. 1994; Cooney et al. 1997; Sinha, Fuse & Aubin 2000).

Stress-related responses and stress-induced craving have been elicited in addicted individuals using a novel model of stress-induced responsivity with an emotional imagery paradigm (Sinha 2008). Here, based on the early work of Lang et al. (1980), both controls and dependent subjects are asked to develop an individualized stress imagery script based on the description of a recent event that a given subject found ‘most stressful’. Similarly, a drug-related script and control neutral-relaxing script were developed for each subject. Subsequent studies used laboratory exposure to the audio-tape scripts as stimuli for guided imagery (Sinha 2008). Using this paradigm, drug craving with mild to moderate levels of physiological arousal and subjective distress was reliably induced in multiple groups of cocaine-, alcohol- and opioid-dependent individuals engaged in treatment (Sinha et al. 2003; Fox et al. 2007). Individuals using higher amounts of cocaine and alcohol per week showed greater craving and physiological responses to stressors (Fox et al. 2005), and recovering alcohol-dependent subjects showed higher craving and physiological responses to stressors than control social drinkers (Sinha 2008). Perhaps most compelling from a validation perspective, stress-induced cocaine craving in the laboratory significantly predicted time to cocaine relapse (Sinha et al. 2006). Similar results have been observed for alcohol- and nicotine-dependent subjects (al'Absi, Hatsukami & Davis 2005; Breese et al. 2005). Preliminary results suggest that an α adrenergic agonist, an anti-sympathetic agent, but not naltrexone, significantly decreased stress-induced opioid craving in opioid-dependent subjects (Sinha et al. 2007). Together, these results provide a compelling laboratory model for stress-induced craving that may have some construct and predictive validity. Future studies cross-referencing pharmacological probes from the animal and human studies should provide an excellent basis for translational advances.

The above models, however, do not take into account resistance to relapse. One could argue that the first occurrence of smoking during a cessation attempt is a critical period and represents an important target for medication treatment. A model termed ‘smoking lapse behavior’ allows the measurement of two critical features of relapse: the ability to resist the first cigarette and subsequent smoking behavior (McKee 2008). Subjects are first exposed to precipitants of smoking relapse, such as alcohol, stress and nicotine deprivation, and then their ability to resist smoking when presented with their preferred brand of cigarettes is measured (McKee 2008). Following the prime, the preferred brand of cigarettes is made available, and the subject has the option of initiating a tobacco self-administration session or delaying initiation for up to 50 minutes in exchange for an incremental monetary reward that increases with every 5 minutes of resistance. Should a subject choose to smoke, a monetary reward can be gained for cigarettes that are not smoked. Increasing levels of nicotine deprivation decreased the ability to resist smoking in smokers (McKee 2008). This model remains to be validated with existing anti-craving medications but provides an intriguing extension of cue-reactivity that may be useful as an intermediary step between pre-clinical models and clinical trials.

Another evolving area in human laboratory relapse models, currently unique to humans, is the measurement of neural correlates of cues for relapse using brain imaging studies. Increased functional brain activation elicited by drug-associated cues may predict increased relapse risk. For alcohol-related cues, both external and internal cues may trigger craving (Verheul, van den Brink & Geerlings 1999). Cue-induced functional activation of the brain can be assessed by measuring changes in cerebral blood flow with positron emission tomography or single photon emission computed tomography, or measuring blood flow combined with functional magnetic resonance imaging. Core regions activated in most studies include the anterior cingulate, orbitofrontal cortex, basolateral amygdala, ventral striatum and dorsal striatum (Heinz et al. 2008). A more key question is whether brain activation during presentation of drug-associated stimuli predicts increased risk for relapse. Stronger cue-induced activation of the ventral striatum, dorsal striatum, medial prefrontal cortex and anterior cingulate has been observed in patients who suffered multiple relapses in alcohol-dependent subjects (Braus et al. 2001; Grusser et al. 2004). Even more intriguing, reduced functional activation of the ventral striatum in response to cues that signal non-drug rewards was observed in alcoholics, suggesting a shift in incentive salience to drug-related cues (Wrase et al. 2007). Thus, imaging studies may provide unique insights into subjects who exhibit the most dramatic functional activation to cues and, by extrapolation, the subjects who are more likely to relapse. Future studies will explore pharmacotherapeutic approaches to normalizing such cue-induced responses and whether such measures will predict therapeutic efficacy in treatment.