|Home | About | Journals | Submit | Contact Us | Français|
Current pharmacotherapies for addiction represent opportunities for facilitating treatment and are forming a foundation for evaluating new medications. Furthermore, validated animal models of addiction and a surge in understanding of neurocircuitry and neuropharmacological mechanisms involved in the development and maintenance of addiction — such as the neuroadaptive changes that account for the transition to dependence and the vulnerability to relapse — have provided numerous potential therapeutic targets. Here, we emphasize a ‘Rosetta Stone approach’, whereby existing pharmacotherapies for addiction are used to validate and improve animal and human laboratory models to identify viable new treatment candidates. This approach will promote translational research and provide a heuristic framework for developing efficient and effective pharmacotherapies for addiction.
Drug addiction is a chronically relapsing disorder characterized by a compulsion to seek and take a drug, loss of control in limiting intake and emergence of a negative emotional state (for example, dysphoria, anxiety and irritability) when access to the drug is prevented1. An important goal of current neurobiological research is to understand the molecular, neuropharmacological and neurocircuitry changes that mediate the transition from occasional, controlled drug use to the loss of behavioural control over drug seeking and drug taking that defines chronic addiction. In this Review, we suggest that a combination of validated animal models for addiction, neurobiological targets derived from such models, and translation to and from the clinical domain provides a heuristic framework for the development of pharmacotherapies for addiction. Moreover, the application of known treatments for addiction to existing animal and human laboratory models can provide an evolving ‘Rosetta Stone approach’ for accelerating the translation of newly identified targets to pharmacotherapies for addiction (BOX 1; FIG. 1).
Addiction is a brain disease, and is defined as a chronically relapsing disorder of compulsive drug use. Advances in our understanding of the neurobiology of addiction have given substantial support to the disease basis for addiction. Changes in specific neuronal and neurochemical circuits have been identified that correspond to different components of the addiction cycle. Perhaps more importantly, these changes are long lasting and in some cases can be permanent. One goal of medications development for addiction is to reverse or compensate for such pathological effects.
The concept of addiction as a disease is also supported by overwhelming evidence that addiction leads to brain pathology from a functional perspective, and this pathology is manifested by reversible, and possibly some irreversible, brain changes. In the United States alone, illicit-drug abuse and addiction costs society US$180.9 billion per year159; in addition, alcoholism costs $180 billion160 and tobacco addiction costs $167 billion161.
From the perspective of treatment, relapse rates for addiction with abstinence as a goal are high: generally 90% without treatment after 1 year. However, such relapse rates for addiction are similar to those for other chronic relapsing disorders, such as diabetes, hypertension and asthma162. Treatments for addiction have limited success, often only doubling the number of individuals that do not relapse after 1 year. However, even capturing 10% of subjects per year could afford considerable savings in human suffering and societal cost. Appropriately monitored replacement treatments, such as methadone and buprenorphine, have a relatively high success rate in terms of reducing or eliminating illicit use of opioids. Furthermore, recent successes with naltrexone, acamprosate, buprenorphine and varenicline hold promise for future medications development for addiction.
A key element of this approach will be to prevent predictions from animal models being limited by the constructs of the models themselves. Several aspects of the Rosetta Stone approach address this issue. First, because no single animal or human laboratory model exists for all of the aspects of addiction, various models are used to emulate different components of addiction, some of which are still evolving2. Second, there is a dynamic approach to the process of validation, with changes being discovered and implemented at both ends. New targets from neurobiology will feed forwards through the system and new medications will feed backwards, regardless of whether these medications are derived from the feed-forward process, clinical experience or serendipity. Third, symptoms or components of the addiction cycle provide a face-valid model that facilitates the creation of new animal and human laboratory models (for example, the animal model of compulsivity and drug seeking in the context of aversive consequences, or the human model of brain imaging in the context of cue-induced imagery3-5).
A useful psychiatry-based motivational framework that integrates well with animal models of addiction is the concept that drug addiction has aspects of both impulse control disorders (such as kleptomania) and compulsive disorders (such as obsessive–compulsive disorder). It has been suggested that, as an individual moves from an impulsive disorder to a compulsive disorder, a shift from positive reinforcement to negative reinforcement drives the motivated behaviour1 within a cycle comprising three stages: binge–intoxication, withdrawal–negative affect, and preoccupation–anticipation.
Animal models of addiction have outstanding face validity (for example, for intravenous self-administration) and recapitulate aspects of the condition in humans. They also have substantial construct validity (for example, deregulated stress responsivity during drug withdrawal), that is, explanatory power or functional equivalence for the condition in humans. Although no animal model fully reproduces addiction in humans, such models do permit investigation of elements of the drug addiction process that can be defined by models of addiction symptoms within the three stages of the addiction cycle. Different animal models for the study of the neurobiology of addiction can be superimposed on these three stages, collectively reproducing the pathological state known as addiction1 (TABLE 1).
Animal models for the binge–intoxication stage of the addiction cycle incorporate drug reinforcement and include drug and alcohol self-administration. For the withdrawal–negative affect stage, animal models exist for the somatic signs of withdrawal for almost all drugs of abuse. However, more relevant to addiction are the animal models of components of the motivational signs of withdrawal and the negative reinforcing effects of dependence, which are beginning to be used to explore how the nervous system is involved in motivation and adapts to drug use. These include anxiety-like responses, conditioned place aversion (a form of place conditioning), elevated reward thresholds and withdrawal-induced increases in drug self-administration. For the preoccupation–anticipation (craving) stage, models include drug-, cue- and stress-induced reinstatement of drug-seeking behaviour. Animal models of craving can also include the conditioned rewarding effects of drugs of abuse, measures of the conditioned aversive effects of withdrawal, and signs and symptoms of protracted abstinence6,7.
A crucial issue for the development of treatments for addiction is that selection of relevant targets should be informed by an empirical understanding of the neurobiology of addiction7. Three neurobiological circuits have been identified that have heuristic value for the study of the neurobiological changes associated with the development and persistence of drug dependence (FIG. 2).
The circuitry related to the origin and terminal regions of the mesocorticolimbic dopamine system, which includes signalling by dopamine and opioid peptides, is a crucial mediator of the positive reinforcing effects of drugs associated with the binge–intoxication stage of the addiction cycle8. The preoccupation–anticipation (craving) stage involves key glutamatergic projections to the extended amygdala and nucleus accumbens from the prefrontal cortex (for drug-induced reinstatement of drug seeking) and from the basolateral amygdala (for cue-induced reinstatement of drug seeking)9. Compulsive drug-seeking behaviour is thought to engage ventral striatal–ventral pallidal–thalamic–cortical loops that could subsequently engage dorsal striatal–dorsal pallidal–thalamic–cortical loops10, both of which are exaggerated by concomitant decreased activity in reward circuits11,12. The neural substrates and neuropharmacological mechanisms for the negative motivational effects of the withdrawal–negative affect stage of the addiction cycle may involve not only disruption of the neural systems implicated in the positive reinforcing effects of drugs, but also recruitment of brain stress systems13. Common responses during acute withdrawal from the main drugs of abuse include decreased dopaminergic activity, an activated pituitary–adrenal stress response and an activated brain stress response with activated extrahypothalamic corticotropin-releasing factor (CRF) systems in the amygdala. However, repeated cycles of addiction lead to a blunted pituitary–adrenal response and a sensitized extrahypothalamic CRF stress system response in the amygdala13.
Molecular changes at the signal transduction, gene transcription or gene level are thought to provide insights into how the circuits described above become deregulated and maintain such deregulation, and provide several contributions to medications development. Drugs of abuse perturb intracellular signal transduction, leading to changes in nuclear function and rates of transcription of certain genes14. This leads to altered activity of the neurons in which such changes occur and ultimately to changes in the function of the associated neural circuits. To date, no medication targets have been identified from molecular targets but, in the future, molecular studies may provide the basis for future targets for pharmacotherapeutic approaches and a key to understanding the vulnerability to addiction.
Several medications are currently on the market for the treatment of addiction (FIG. 2; TABLE 2). The validation procedure termed the Rosetta Stone or reverse validity approach uses drugs that are known to be effective in human clinical studies to validate animal models and human laboratory models, and can provide a means of refining such models. For more detailed information on the effects of drugs approved for the treatment of addiction on established animal models of addiction, see Supplementary information S1 (box).
The premise of this Review is that different components of the addiction cycle can be targeted by different medications, selected on the basis of information from three key sources. These comprise research into basic neurobiological mechanisms for the different stages of the addiction cycle, the effects of medications approved for the treatment of addiction on animal models of the different stages of the addiction cycle (Supplementary information S1 (box)) and clinical studies of medications approved for other indications that overlap with specific components of addiction (discussed below) (FIG. 2; TABLE 3).
The effects of selected compounds acting on specific neurobiological targets on models of the motivational components of the addiction cycle that are relevant for pharmacotherapies, using alcoholism as an example, are shown in TABLE 4. It compares novel approaches with two medications currently on the market (naltrexone, and acamprosate (Campral/Aotal; Merck–Serono/Forest Laboratories)). Different patterns of action on the animal models are thought to reflect actions in different stages of the addiction cycle. The Rosetta Stone approach places emphasis on elements of the withdrawal–negative affect stage (the ‘dark side’) of addiction. We propose that this framework is crucial for the neuroadaptations that lead to changes in motivation and thereby drive addiction to maintain an allostatic state12. There is compelling evidence that direct antagonism of the reinforcing effects of drugs of abuse, representing the binge–intoxication stage, produces compensatory increases in drug taking, motivational side effects that limit compliance, or a combination of both. Other reviews have focused on animal models of the preoccupation– anticipation (craving) state15. Here, we explore four neurotransmitter systems: dopamine, γ-aminobutyric acid (GABA), CRF and glutamate, all of which have targets that can restore deregulated reward systems, as in the withdrawal–negative affect stage, and in some cases affect the binge–intoxication stage (dopamine) and the preoccupation–anticipation stage (glutamate). Although GABA, glutamate and CRF are not new targets in the treatment of addiction2, targeting these and the dopamine system using the framework of validation described here has the potential to provide new medications.
The mesolimbic dopamine system projects from the ventral tegmental area to basal forebrain sites, the nucleus accumbens and the central nucleus of the amygdala, and has a key role in motivation. Activation of the mesolimbic dopamine system is thought to be important for directing behaviour towards salient rewarding stimuli16, but may not be necessary for hedonic experience17. Therefore, mesolimbic dopamine activity seems to be crucial for the reinforcing actions of indirect sympathomimetics, such as cocaine and amphetamines, and is involved in, but not essential for, the reinforcing actions of other drugs of abuse, such as opioids and alcohol.
More importantly for our dark side view, dopaminergic function is compromised during acute withdrawal from all major drugs of abuse18: levels of extracellular dopamine decrease in the nucleus accumbens following a binge of self-administered cocaine19; animals kept on a diet of alcohol show a decrease in extracellular levels of dopamine in the nucleus accumbens during withdrawal20. Withdrawal from most major drugs of abuse is also associated with decreased firing of dopaminergic neurons in the ventral tegmental area21.
Given the role of dopamine in the acute reinforcing effects of drugs and its deregulation during drug withdrawal, a dopamine partial agonist may be an effective treatment in different stages of the addiction cycle. A dopamine receptor partial agonist has antagonist properties in situations of high intrinsic activity and agonist properties in situations of low intrinsic activity. Because of its intermediate efficacy, a dopamine partial agonist acts as an agonist in the absence of dopamine and can act as an antagonist in the presence of dopamine23-25, and would hypothetically have less severe or fewer side effects than full agonists or antagonists22. Indeed, partial agonists of dopamine D2 receptors dose-dependently decrease the reinforcing effects of intravenous cocaine and amphetamine self-administration and oral alcohol self-administration in non-dependent rats26-29.
In a series of studies, D2 partial agonists have been shown to reverse psychostimulant withdrawal and block the increase in self-administration associated with extended access to the psychostimulant. A D2 partial agonist reversed the motivational deficit that occurs during amphetamine and methamphetamine withdrawal30,31. Animals with extended access to methamphetamine through intravenous self-administration show an increased intake of methamphetamine (escalation in intake)32. A notable effect of the D2 partial agonist aripiprazole on methamphetamine self-administration was a shift to the right of the dose–response function, with a greater effect in rats with higher methamphetamine intake associated with extended access. These data again suggest an increased sensitivity to the effects of the D2 partial agonists in dependent rats32.
Antagonists of the dopamine D3 receptor do not affect baseline cocaine self-administration, but block responses in a progressive-ratio schedule, a measure that is thought to reflect a compulsivity component of cocaine seeking. Additionally, D3 antagonists block cue-induced rein-statement of cocaine and alcohol self-administration33. Similarly, D3 partial agonists do not block baseline cocaine self-administration but block cue-induced reinstatement of self-administration34 and cue-induced drug seeking in a second-order schedule of reinforcement35. A D3 partial agonist also blocked amphetamine-induced conditioned place preference36.
D1 antagonists competitively block cocaine self-administration in rats37, but little work has been done on D1 partial agonists. One study has shown strain-dependent agonist and antagonist effects on cocaine self-administration with a D1 partial agonist in rats38. Together, these results suggest that deregulation of dopamine signalling contributes to the motivational effects of drug withdrawal and reinstatement, and dopamine partial agonists with the appropriate neuropharmacological and pharmacokinetic profile may be effective in treating certain aspects of addiction39.
GABAA receptor antagonists and inverse agonists decrease alcohol self-administration40,41. However, their therapeutic actions are limited by potential side effects involving central nervous system hyperexcitability. By contrast, GABA receptor agonists or modulators can block drug-seeking behaviour through their actions on reward, dependence or both. GABA receptor modulators that increase GABAergic activity directly or indirectly decrease self-administration of cocaine, heroin, nicotine and alcohol in non-dependent rats42-45. GABA receptor agonists also block alcohol withdrawal in animals46 and humans, and decrease drinking and certain components of craving in humans with alcoholism47,48 (TABLE 4). GABAB receptor agonists also block the increased alcohol self-administration observed during acute withdrawal in dependent rats at lower doses than those that block alcohol self-administration in non-dependent rats, suggesting an increased sensitivity of this system during the development of dependence49. The GABAB agonist baclofen has been reported to reduce alcohol craving and intake in a preliminary double-blind, placebo-controlled trial47. However, GABAB agonists currently in therapeutic use for the relief of flexor spasms in multiple sclerosis have substantial sedative effects at therapeutic doses50. Thus, another approach is to explore the role of GABA receptor modulators that indirectly facilitate GABA release.
Gabapentin (Neurontin; Pfizer), an amino acid designed as a structural analogue of GABA51, is a novel anticonvulsant drug that is also used in the treatment of neuropathic pain. Gabapentin increases the concentration of GABA in the brain52 and GABA release from rat brain slices in vitro53. It also decreases synaptic transmission in the brain by selectively inhibiting Ca2+ influx through voltage-operated Ca2+ channels54 and may be an agonist of GABAB receptors55.
In animal models of alcohol dependence, gabapentin has strikingly different effects in non-dependent and alcohol-dependent rats, both cellularly and pharmacologically56. In non-dependent rats, gabapentin facilitated GABAergic transmission in the central nucleus of the amygdala but did not affect alcohol intake. However, in dependent rats, gabapentin decreased GABAergic transmission in the central nucleus of the amygdala and reduced excessive alcohol intake. Furthermore, gabapentin suppressed the anxiogenic-like effects of withdrawal from an acute alcohol injection. A possible explanation for these results is that, during the development of alcohol dependence, neuroadaptive changes occur in the GABAergic system, including a reduced sensitivity and/or a downregulation of presynaptic GABAB receptors56. Gabapentin has proved effective in human laboratory studies in decreasing craving, reversing physiological measures of protracted abstinence and reversing sleep deficits in protracted abstinence57, suggesting a key translation from animals to humans and supporting the translational approach suggested here.
Alcohol is a powerful activator of ‘stress systems’, which could have important implications for our understanding of the neurobiology of dependence and relapse. Alcohol-mediated activation of stress systems occurs through the hypothalamic–pituitary–adrenal axis and extensive extrahypothalamic, extraneuroendocrine CRF systems implicated in behavioural responses to stress58. Central administration of CRF mimics the hormonal, autonomic and behavioural response associated with activation and stress in rodents; competitive CRF receptor antagonists generally have the opposite effects13.
In animal models, a common response to acute withdrawal and protracted abstinence from all major drugs of abuse is the manifestation of anxiety-like responses. Withdrawal from repeated administration of cocaine, alcohol, nicotine and cannabinoids produces an anxiogenic-like response in the elevated plus maze and defensive burying test, and one or both of these effects are reversed by administration of either selective CRF receptor 1 (CRF1) antagonists or mixed CRF1–CRF2 antagonists60-66.
During alcohol withdrawal, extrahypothalamic CRF systems become hyperactive and there is an increase in extracellular CRF within the central nucleus of the amygdala and the bed nucleus of the stria terminalis in dependent rats67-69. Extracellular CRF is increased in the amygdala during withdrawal from chronic administration of nicotine64, cocaine70, opioids71 and cannabinoids66 in rats.
The ability of CRF antagonists to block the anxiogenic-like and aversive-like effects of drug withdrawal would predict the motivational effects of CRF antagonists in animal models of dependence40,62,63,72. A CRF1–CRF2 peptide antagonist that has no effect on alcohol self-administration in non-dependent rats eliminates excessive drinking in dependent rats during acute withdrawal and protracted abstinence73. Direct administration of a CRF1–CRF2 peptide antagonist into the central nucleus of the amygdala also eliminated excessive drinking during acute withdrawal67. Systemic injections of small-molecule CRF1 antagonists also blocked the increased alcohol intake during acute withdrawal and protracted abstinence in alcohol-dependent rats, but not in non-dependent rats74,75 (TABLE 4). These data suggest an important role for CRF, primarily in the central nucleus of the amygdala, in mediating the increased self-administration associated with alcohol dependence. CRF1 antagonists also selectively blocked the increased self-administration associated with extended access to cocaine76, nicotine64 and heroin77.
Several clinical trials that have been completed explored the effects of CRF1 antagonists on anxiety and depression, but no human laboratory studies or clinical trials have yet been initiated to investigate the effects of such drugs on alcohol dependence. In a trial terminated in Phase II because of increased levels of liver transaminases, the CRF1 antagonist R121919 caused a significant reduction in scores on the Hamilton Depression Inventory during the 30 day treatment period of the open trial78. A randomized, double-blind, placebo-controlled trial of another CRF1 antagonist, CP-316311, had no effect compared with the placebo, whereas the selective serotonin reuptake inhibitor sertraline produced a positive result79. CP-316311 had no adverse effects and in both trials the CRF1 antagonists were well tolerated, suggesting that CRF1 antagonists could be viable therapeutics for other addiction-related disorders.
Preclinical data are emerging which suggest that other neurotransmitter systems and neuromodulators within the extended amygdala can be deregulated during the development of dependence on drugs of abuse13. Evidence for deregulation of noradrenaline in alcohol, cocaine and opioid dependence; dynorphin in cocaine and alcohol dependence; vasopressin in opioid and alcohol dependence; and orexin and substance P in cocaine, opioid and alcohol dependence are pertinent examples13. Administration of the noradrenergic α1-receptor antagonist prazosin decreases self-administration in rats that are dependent on alcohol, cocaine and opioids80-82 (TABLE 4). It has been proposed that noradrenergic–CRF interactions contribute to the brain stress activation associated with withdrawal from drugs of abuse13. In animal models of alcohol self-administration, the dramatic motivational effects of CRF in dependence can be observed in dependent animals.
Activation of the dynorphin–κ-opioid receptor system produces effects that are similar to those of most other opioid systems, but often opposite to those of μ-opioid receptors in the motivational domain83.κ-Opioid receptor agonists produce conditioned place aversion84, and depression and dysphoria in humans85. Substantial evidence suggests that expression of the gene encoding the dynorphin peptide and κ-opioid receptor activation are increased in the striatum and amygdala during acute and chronic administration of cocaine in rats86,87 and in humans88,89. The activation of the dynorphin system in the nucleus accumbens decreases activity in the dopamine system. It is therefore possible that the activation of the dynorphin system could contribute to the dysphoric syndrome associated with cocaine dependence90.
Dynorphin activity is also increased by stress91, suggesting a potential interaction with the CRF system. In mice, forced-swim stress and inescapable foot shock produced place aversions that were blocked by a κ-opioid receptor antagonist and dynorphin knockout92. Blockade of dynorphin activity, either by κ-opioid receptor antagonism or by disruption of the prodynorphin gene, blocked stress-induced reinstatement of cocaine-induced place preference in mice93 and blocked stress-induced reinstatement of cocaine-seeking behaviour in rats94, and CRF is thought to produce its aversive effect through activation of the dynorphin system92. There is also evidence that the link between reinstatement of drug-seeking behaviour and activation of κ-opioid receptors is mediated by CRF95. Thus, the dynorphin–κ-opioid system mimics stressor administration in animals in producing aversive effects and inducing drug-seeking behaviour. This aversive response could involve reciprocal interactions with dopamine in the nucleus accumbens and the extrahypothalamic brain CRF system.
Administration of a κ-opioid receptor antagonist had no effect on baseline self-administration of limited-access cocaine or heroin in primates96, but blunted the increased self-administration of cocaine in rats with extended access to the drug (S. Wee et al., in the press). A κ-opioid receptor antagonist also selectively blocked the increase in ethanol self-administration associated with withdrawal in alcohol-dependent rats97.
Neuromodulatory systems that oppose the actions of CRF in modulating stress and emotional behaviour could also be future targets for addiction treatment. These include neuropeptide Y and nociceptin systems, in which agonists reduce the excessive drinking associated with alcohol dependence98,99. Targeting the receptor system for substance P, known as the neurokinin 1 receptor (NK1R; also known as TACR1) system, which modulates emotional states, provides an example of a translational approach100. NK1R-knockout mice backcrossed onto a high-drinking strain of mice showed a major decrease in voluntary alcohol consumption, and recently detoxified subjects with alcohol dependence and treated with an NK1R antagonist showed decreased craving, blunted cortisol responses and decreased functional magnetic resonance imaging (fMRI) responses to affective stimuli. This suggests that NK1R may be a viable therapeutic target for the treatment of the dark side of addiction.
Glutamate is thought to have several roles in the neurobiology of addiction, many of which provide potential targets for medications development. At low doses, ethanol may act as an NMDA (N-methyl-d-aspartate) receptor antagonist, which could contribute to the acute rewarding effects of ethanol102. Acute and protracted abstinence from ethanol described in alcohol dependence models seems to involve overactive glutamatergic systems, which are thought to be a target for the therapeutic effects of acamprosate39,102,103 (TABLE 4).
Glutamate has also long been associated with the neuroplasticity that is important for behavioural sensitization to drugs of abuse, particularly psychostimulants. AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid) receptor antagonists and NMDA receptor antagonists block the development of locomotor sensitization to psychomotor stimulants104, and NMDA receptor antagonists also block the long-term potentiation and long-term depression associated with repeated administration of psychostimulants105. Some of these effects have been linked to increases in AMPA receptors that lack the subunit known as glutamate receptor 2 (GluR2)106.
One prominent hypothesis is that repeated self-administration of psychostimulants decreases basal release of glutamate in key brain circuits associated with the preoccupation–anticipation (craving) stage of the addiction cycle, but an exaggerated response of glutamate to activity in these circuits could confer susceptibility to relapse107. For example, drug-induced reinstatement of drug seeking seems to be mediated by a glutamatergic projection from the prefrontal cortex to the nucleus accumbens9. Cue-induced reinstatement of drug seeking involves a glutamatergic projection to the nucleus accumbens from both the basolateral amygdala and the ventral subiculum108,109. In addition, in cocaine-treated mice, there was less strengthening of synaptic transmission in ventral tegmental area slices from mice that lacked mGluR1 or the NMDA receptor subunit NR1 in dopamine neurons, suggesting a direct dopamine neuron target produced by glutamatergic plasticity110.
Additional evidence that AMPA or kainate receptors mediate glutamate modulation in addiction comes from a ‘top–down’ perspective. Clinical trials of topiramate, an anticonvulsant with some glutamate antagonist activity, have reported decreases in drinking behaviour in alcohol dependence and improvements in quality of life, but with significant adverse effects on memory and concentration111,112. In preclinical studies, topiramate decreased alcohol consumption and preference113 and decreased stress-induced increases in alcohol consumption and preference in mice114. However, topiramate also decreased saccharin preference, increased water intake and failed to block conditioned place preference to alcohol113,115. Together, these results suggest that topiramate can interact with both withdrawal-induced negative affective states and the rewarding effects of ethanol.
Pharmacological agents that modulate glutamate function may not only have a role in the basal hypoexcitability or hyperexcitability of glutamatergic systems during protracted abstinence, depending on the drug of abuse, but may also decrease drug-and cue-induced reinstatement of drug self-administration. To this end, antagonists of AMPA receptors, NMDA receptors and metabotropic glutamate receptor 5 (mGluR5), and agonists of mGluR2 and mGluR3, all of which decrease glutamate function, have been shown to block cue-induced reinstatement of drug self-administration116-119.
Given the side effects associated with direct glutamatergic antagonists, drugs that modulate the system may be more promising candidates for the treatment of addiction. Agents that can restore homeostasis to systems associated with both the withdrawal–negative affect (dark side) stage and the preoccupation–anticipation (craving) stage would be optimal. Studying the withdrawal–negative affect stage of the addiction cycle reveals numerous targets for pharmacotherapy development for addiction. A similar case can be made for the preoccupation–anticipation stage using the neurobiological targets identified here7. However, in both domains, a limiting step for clinical development is the progression from identifying preclinical targets to testing in humans.
A crucial step in medications development is the submission of an investigational new drug (IND) application to the US Food and Drug Administration (FDA), or its equivalent in other countries, so that the drug can be tested in humans. See Supplementary information S2 (box) for further information regarding INDs for pharmacotherapies for the treatment of addiction.
Human laboratory studies provide a potentially powerful means of exploring treatment targets for specific components of the addiction cycle without the need for expensive double-blind, placebo-controlled trials. They can potentially predict efficacy measures of potential treatments for each stage of the addiction cycle (TABLES 1,,4).4). Although the predictive validity of human laboratory models remains to be determined, ongoing studies with established medications for addiction can be used in a Rosetta Stone approach to evaluate the validity of animal and human models. Human laboratory models can also then serve as a springboard for the development of new pharmacotherapies.
For the binge–intoxication stage of the addiction cycle, self-administration procedures for cocaine, heroin and marijuana in humans have been established largely using operant responding paradigms, in which participants who are dependent on a drug make a behavioural response such as pressing a key on a computer to receive the drug120. As in animal models, heroin self-administration is reduced by all three medications approved by the FDA to treat opioid dependence — methadone, naltrexone and buprenorphine (Subutex; Schering–Plough) — supporting the predictive validity of the animal models121,122. However, for cocaine, support for the validity of these models is less robust. A range of medications have been shown to reduce the subjective effects and craving associated with cocaine but do not decrease cocaine self-administration itself120. These results are consistent with clinical data showing that, of more than 60 medications tested, none has proved reliably effective in clinical trials123. To date, little or no work has been done on treatments for marijuana dependence.
Other measures, such as impulsivity, could be considered endophenotypes of the binge–intoxication stage and have some potential for predicting which pharmacotherapies could have efficacy for addiction treatment. Impulsivity can increase the probability that an individual will engage in initial drug intake, and the subsequent effects of the drug on impulsivity may increase impulsive behaviours that facilitate further drug use, prolonging the binge or even provoking relapse (discussed below). Various tasks have been used to assess impulsivity, including delayed discounting (that is, whether subjects show a relative preference for smaller, more immediate rewards over larger, more delayed rewards), behavioural inhibition (for example, the Stop Task124) and attentional measures (that is, whether subjects show increased variability in performance in a simple reaction time task that reflects ‘lapses in attention’).
For the withdrawal–negative affect stage, negative reinforcement mechanisms are in operation. Numerous human laboratory measures of acute withdrawal have proved to be sensitive to drug substitution125. In the laboratory, marijuana withdrawal is alleviated by marijuana smoking or by administration of oral Δ9-tetrahydrocannabinol (THC)126. Cognitive measures are sensitive to withdrawal effects of drug dependence during acute and protracted abstinence, and can be considered to be another endophenotype of the addiction process that could be used in medication screening127 (TABLE 1). Nicotine can improve cognitive processing and reduce negative affect128. The cascade of stress hormone interactions with drugs of abuse — which can facilitate the binge–intoxication stage, exaggerate the withdrawal–negative affect stage, and cause sensitization to stress-induced relapse — may all be amenable to human laboratory studies129.
For the preoccupation–anticipation stage, three main external factors (priming doses of drug, drug-associated cues and stressor exposure) and two internal factors (the malaise of protracted abstinence and an associated state of stress that contributes to malaise) are thought to contribute to relapse. 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, notably for alcohol and tobacco addiction. Priming-induced drinking in alcohol-dependent subjects in a bar-like setting was greater than in social drinkers and was selectively decreased in the alcohol-dependent groups by administration of opioid antagonists130. Similar results were observed in individuals who were dependent on alcohol, had a family history of alcohol dependence and received a priming dose of alcohol131, and in cigarette smokers primed with five cigarettes132.
Exposure to alcohol cues, such as the sight or smell of alcoholic beverages using the cue reactivity paradigm, reliably increases the urge to drink alcohol, salivation and attention to cues57,133,134. Furthermore, cue reactivity can predict treatment outcome135 and has been validated in some cases using medications that successfully treat alcohol dependence. For example, naltrexone, but not topiramate, blocked cue reactivity in subjects with alcohol dependence134,136, and nicotine replacement therapy decreased craving associated with smoking cues137. Other drugs not currently in therapeutic use for addiction have shown positive results with cue-induced reactivity paradigms, including carbamazepine (Tegretol; Novartis) for alcohol138 and amantadine for cocaine139.
Stress responses, including changes in the activities of the hypothalamic–pituitary–adrenal axis and extra-hypothalamic brain stress systems, affect all phases of the addiction cycle but may be particularly relevant to both the withdrawal–negative affect stage and the preoccupation–anticipation stage. Stress and stressors have also been associated with relapse and vulnerability to relapse129,140. Negative affect, stress or withdrawal-related distress also increases drug craving57,135,141,142. Both stress and drugs of abuse activate the hypothalamic–pituitary–adrenal axis, but the glucocorticoid response becomes blunted with chronic high-dose drug use. High glucocorticoid tone can, in turn, drive the brain stress systems in the amygdala129. Thus, drugs of abuse can trigger a cascade of stress hormone interactions that can facilitate the binge–intoxication stage and exacerbate the withdrawal–negative affect stage. They may also cause hypersensitivity of brain stress systems that contribute to maintaining the withdrawal–negative affect stage and sensitize the individual to stress-induced relapse. All of these changes may be amenable to study in humans in a laboratory setting.
Stress-related responses and stress-induced craving have been elicited in individuals with an addiction using a new model of stress-induced responsivity with an emotional imagery paradigm142 based on the early work of Lang and colleagues143. In this paradigm, individuals using the higher amounts of cocaine and alcohol and subjects recovering from an alcohol dependence showed greater craving and physiological responses to stressors than control social drinkers144. Perhaps of greatest importance in terms of paradigm validation, stress-induced cocaine craving in the laboratory could be used to accurately predict time to relapse145. Similar results have been observed for subjects dependent on alcohol or nicotine146,147. Preliminary results suggest that an α2-adrenoceptor agonist and an antagonist of sympathetic signalling, but not naltrexone, significantly decreased stress-induced opioid craving in subjects dependent on opioids148. These results support the construct and predictive validity of this laboratory model for stress-induced craving. Future studies crossreferencing pharmacological probes from the animal and human studies should provide an excellent basis for translational advances.
Exploration of the interaction of cue exposure with emotional states during protracted abstinence has provided a novel approach to the study of cue reactivity in alcohol craving57,149. A sample of non-treatment-seeking subjects with alcohol dependence was exposed to affective stimuli that had positive or negative valence (that is, they produced emotional responses that were positive or negative in nature) 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. Alcohol exposure and both positive and negative emotional cues had the expected effects on subjective and emotional reactivity, and related effects on psychophysiological measures57. Gabapentin significantly decreased subjective craving and craving that was affectively evoked, and improved several measures of sleep quality57. These results suggest that affective priming, combined with alcohol cue exposure, could provide a powerful means to evaluate potential pharmacotherapies for addiction treatment.
Another novel approach involves measuring resistance to relapse in humans. The ‘smoking lapse behaviour’ model allows the measurement of two crucial features of relapse: the ability to resist the first cigarette and subsequent smoking behaviour150. Subjects with nicotine dependence 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 measured150. This model remains to be validated with existing anti-craving medications but provides an intriguing extension of cue reactivity studies that could be useful as an intermediary step between preclinical models and clinical trials.
An evolving area in human laboratory relapse models is the measurement of neural correlates of cues for relapse. Increased functional brain activation elicited by drug-associated cues, as measured in brain imaging studies, may correlate with an increased risk of relapse. 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 by combining blood flow measurements with fMRI. Core regions activated in most studies include the anterior cingulate, orbitofrontal cortex, basolateral amygdala, ventral striatum and dorsal striatum5. Strong cue-induced activation of similar regions, including the ventral striatum, dorsal striatum, medial prefrontal cortex and anterior cingulate, has been observed in subjects with alcohol dependence who have repeatedly suffered relapses151,152. Even more intriguingly, reduced functional activation of the ventral striatum in response to cues that signal non-drug rewards was observed in individuals with alcoholism, suggesting a shift in incentive salience to drug-related cues153. Imaging studies may therefore provide unique insights into subjects who exhibit the most dramatic functional activation in response to cues and, by extrapolation, the subjects who are more likely to relapse. Future studies will need to explore pharmacotherapeutic approaches to normalizing such cue-induced responses and whether such measures will predict therapeutic efficacy in treatment100.
Widespread attempts are being made to identify genetic markers for addiction. However, a more exciting possibility is that certain single nucleotide polymorphisms may predict vulnerability to certain subtypes of excessive drinking syndromes and, of particular relevance to this Review, may predict responsiveness to pharmacotherapies in the treatment of alcoholism. Animal and human studies are beginning to realize this potential opportunity. Genetic association studies have focused on two pathways: one representing the reward side of addiction (the μ-opioid peptide system) and one representing the dark side of addiction (the CRF brain stress system). The human μ-opioid receptor is encoded by the OPRM1 gene and is a primary candidate for causing the pharmacogenetic variability of the clinical effects of opioid drugs and opioid receptor antagonists in the treatment of addiction. Mutations in OPRM1 have been found in the promoter, coding regions and introns of the gene. One mutation that has received considerable attention is the A118G single nucleotide polymorphism, which causes an amino acid substitution of asparagine with aspartate at position 40 of the μ-opioid receptor protein (Asn40Asp)154 (BOX 2).
In a separate series of studies, an association was found between single nucleotide polymorphisms of the gene that encodes CRF1 and binge drinking in adolescents and in adults with an alcohol dependence155. One of these single nucleotide polymorphisms, R1876831, is located in an intron that could potentially influence transcription of the gene that encodes CRF1 (BOX 2).
Double-blind, placebo-controlled trials with random assignment to treatments are the accepted standard for determining pharmacotherapy efficacy. A number of unique features of clinical trials for medications to treat addiction need to be emphasized. These include a lack of consensus about clinically relevant outcome measures, admission criteria and methods for detecting relapse to use of substances between study visits. Additionally, general issues related to medication compliance, placebo response and dropout rates present challenges for the design of clinical trials in addiction. There are also safety and tolerability issues that are specific to addiction, including the potential interaction of alcohol or other substances of abuse with the pharmacotherapy under study.
The choice of the outcome for which a pharmacotherapy is likely to show efficacy may be guided by results from preclinical models of addiction and human laboratory studies. For example, naltrexone is thought to reduce the rewarding effects of drinking such that the individual with an alcohol dependence is no longer motivated to drink heavily. An outcome of reduced heavy drinking, as opposed to complete abstinence, is well characterized for naltrexone in human laboratory models involving alcohol administration and in clinical trials with subjects that are non-abstinent. Conversely, medications such as acamprosate or gabapentin, which are thought to support abstinence from alcohol by normalizing activity in a brain pathway that has become chronically deregulated, have shown efficacy in human laboratory models of cue reactivity but not in alcohol administration models, and have primarily shown efficacy on abstinence outcomes in clinical trials (see below).
Historically, the rate of complete abstinence has been the primary outcome measure of pharmacotherapy efficacy. However, clinically relevant abstinence-oriented outcomes also include latency to first lapse, percentage of abstinence days over the study duration, longest duration of abstinence during the study, abstinence at the end of the study (as indicative of behaviour likely to continue after the study) and statistical modelling of the trajectory of abstinence over the course of the study. Depending on the pharmacokinetic properties of the medication (for example, time to achieve steady-state serum concentrations), it may be necessary to specify a grace period during which a lapse may occur that would not normally be considered in efficacy determinations. Alternatively, to ensure an abstinent starting point across all patients, detoxification and a brief abstinent interval (for example, 2–5 days) may be required before randomization.
Harm reduction and non-hazardous use of substances are alternative end points for patients that do not have abstinence as their treatment goal. Related outcomes may include time to first heavy use, percentage of heavy use days during the study or statistical models of the trajectory of heavy use days over the course of the study.
Admission criteria related to pre-randomization substance use or abstinence vary as a function of the stage of the addiction cycle that a pharmacotherapy is thought to target. To show efficacy, pharmacotherapies that are thought to support abstinence through normalization of a brain pathway deregulated by discontinuation of the addictive substance require a study sample that has achieved a minimal period of abstinence before randomization (usually 2–5 days to avoid the effects of acute withdrawal). Conversely, pharmacotherapies that are predicted to decrease heavy or hazardous use may require a non-abstinent sample with a high pre-randomization rate of heavy use to show efficacy.
Safety considerations specific to a medication also influence admission criteria. For example, subjects with an addiction, especially those dependent on alcohol, frequently present for treatment with pathologically elevated liver function test values. Trials of pharmacotherapies that are associated with hepatotoxicity (for example, naltrexone) typically exclude patients with liver function test results that are more than three times the upper limit of the normal range. However, such necessary exclusion criteria may also limit the dependence severity in the study sample and the extent to which clinically relevant generalizations can be made. Although women with childbearing potential are often omitted from clinical trials, this group needs to be represented in clinical trials of substance dependence because gender can affect drug efficacy. For example, long-acting injectable naltrexone showed efficacy in males but not in females in a multicentre trial of alcohol dependence156.
Measures of substance use are another unique aspect of clinical trials for pharmacotherapy development for addiction. Urine dipsticks and hand-held analysers of alcohol or carbon monoxide in expired breath can provide an immediate indication of drug use, drinking and cigarette smoking, respectively, at the time of a clinical trial research visit. However, most substances of abuse and their active metabolites (if any) have short half-lives, and their use cannot be assessed in breath, urine or plasma during the interval between study visits. One exception is Δ9-THC, the primary metabolite of cannabis, which can be present in urine for 1 month after use, potentially resulting in an underestimate of abstinence if detected. Because of such a long half-life, urinary Δ9-THC levels can be normalized to the urinary creatinine concentration to reduce variability in drug measurement owing to urine dilution157; abstinence is reflected by a reduction in the Δ9-THC/creatinine ratio across visits.
Available biomarkers of substance use between study visits (for example, γ-glutamyl transpeptidase levels, mean corpuscular volume and carbohydrate-deficient transferrin levels in the case of alcohol) have sensitivity and/or specificity limitations that can contribute to inaccurate estimations of use. A standardized interview using a calendar format, typically a variation on the timeline followback interview158, uses prompts such as weekend versus weekdays, paydays or holidays to assess quantity and frequency of substance use between study visits. Periodic interviews with collateral informants such as close friends or relatives of the subject can also be used in conjunction with biological measures to validate the subject’s timeline followback interview self-report measures of substance use. If discrepancies between sources cannot be resolved, the most negative outcome is typically assumed to be accurate.
Medication non-compliance is not a common factor that influences outcomes in clinical trials involving addiction. When it occurs, factors such as drug tolerability should be assessed if standard measures to ensure compliance have been implemented. Such measures can include packaging medication in blister cards with day and time of day indicated for each dose, identifying frequently missed doses by reviewing unused medication at every visit or using multiple event monitoring system cap data, linking missed dosing with an activity of daily living, such as meals or brushing teeth, and a suggestion to affirm a commitment to recovery with every dose.
Part of the definition of substance dependence by the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM IV) is an increased risk of relapse, impaired impulse control and psychosocial difficulties. Not surprisingly, therefore, trials of substance dependence may involve higher dropout rates than clinical trials of treatments for other psychiatric disorders. Therefore, incorporating strategies to enhance study completion in order to adequately assess use, safety and efficacy of a pharmacotherapy is important. Such strategies can include a flat rate of monetary compensation to offset transportation expenses at each visit and a lump sum payment for trial completion that is sufficient to be motivating but not coercive, the offer of a ~US$5 value coupon at each visit or systematic acknowledgement of the gains the subject is making (for example, income saved on days of abstinence that was previously spent on substances or normalization of liver function tests). Obtaining comprehensive contact information is also prudent.
The US National Institutes of Health (NIH) has promoted many of the elements required for the development of pharmacotherapies for addiction. To our knowledge, equivalent programmes do not exist in Europe, and should be encouraged. Areas of success include tremendous breakthroughs in the basic neurobiology of addiction, the successful development and validation of behavioural and pharmacological treatments of addiction (such as buprenorphine, supported by the National Institute on Drug Abuse (NIDA), and naltrexone, supported by the National Institute on Alcohol Abuse and Alcoholism (NIAAA)), and improved infrastructure for some aspects of pharmacotherapy development. For example, NIDA has established an extensive clinical trials network (see Further information for a link to the NIDA clinical trials network website). However, tremendous resources have been devoted to the development of pharmacotherapies for cocaine addiction, with little or no success reported to date2. It is hoped that the burgeoning use of human laboratory studies, outlined above, and the Rosetta Stone approach linking human and animal studies promoted here will yield better results.
A relevant issue is therefore how the NIH can do more to help. Several suggestions can be provided on the basis of this Review. These include developing a more balanced portfolio at the preclinical and clinical stage regarding which drugs of abuse should be studied and expanding the development and validation of human laboratory studies for potential medications. A major bottleneck for testing new medications is obtaining IND approvals for new drugs for human laboratory studies and Phase II clinical trials. A core facility like that of the NIDA clinical trials network should be considered for pooling resources for IND development, not only by NIDA and NIAAA, but by all central nervous system-related NIH institutes. Of crucial importance is a scientifically determined and clinically relevant ‘decision tree’ of which drugs should go forward. Such a decision tree could be built on the framework elaborated here, by incorporating a Rosetta Stone-like validation into the strategic planning of the NIH to facilitate translational research.
Another key element is that the pharmaceutical industry must consider that pharmacotherapies to target addiction are potentially profitable. Recent success with acamprosate and varenicline (Chantix/Champix; Pfizer) should provide some indication that further investigations in this area of disease are merited. Increased interaction with NIH in general to pool resources could reduce some of the developmental costs for industry, and some mechanisms are clearly underway in the form of small business innovation research grants and small business technology transfer grants (see Further information for a link to the Small Business Innovation Research (SBIR) and Small Business Technology Transfer (STTR) programmes). However, more efforts could be made by industry to partner with the NIH for the development of medications for addiction treatment. Furthermore, the pharmaceutical industry should consider allowing compounds that have been granted IND status but which they are no longer developing to be used in proof-of-concept human laboratory studies of addiction.
Thus, there is substantial potential for the development of future pharmacotherapies for addiction. Medications currently on the market for this indication have not only provided information on the opportunities for facilitating treatment but are also forming a means to evaluate future medications. A combination of validated animal models of addiction and a surge in understanding the neurocircuits and neuropharmacological mechanisms involved in the development and maintenance of addiction through basic research has provided numerous viable targets for future medications. New neurobiological targets will be derived from this basic research on addiction, with a focus on the neuroadaptive changes that account for the transition to dependence and the vulnerability to relapse, possibly within a genetic context. We propose that the Rosetta Stone framework outlined here could provide a heuristic approach for efficient and effective development of pharmacotherapies for addiction, and promote the flourishing of translational research interactions among independent investigators, the NIH and private industry.
This is manuscript number 19,996 authored from The Scripps Research Institute. The authors thank M. Arends for his assistance with manuscript preparation. Preparation of this work was supported by the Pearson Center for Alcoholism and Addiction Research and National Institutes of Health grants AA12602, AA08459, (R37)AA014028 and AA06420 from the National Institute on Alcohol Abuse and Alcoholism, DA04043, DA04398, (P20)DA024194 and DA10072 from the National Institute on Drug Abuse, DK26741 from the National Institute of Diabetes and Digestive and Kidney Diseases, and 17RT-0095 from the Tobacco-Related Disease Research Program from the State of California. The opinions expressed in this article by G.K.L. are as an individual and not as an employee of Nereus Pharmaceuticals.
Competing interests statement The authors declare competing financial interests: see web version for details.
κ-opioid receptor | μ-opioid receptor | CRF1 | D2 | D3 | NK1R
NIDA clinical trials network: http://www.nida.nih.gov/ctn/
Small Business Innovation Research (SBIR) and Small Business Technology Transfer (STTR) programmes: http://grants.nih.gov/grants/funding/sbirsttr_programs.htm