Addiction is a complex disease process of the brain that results from recurring drug intoxication and is modulated by genetic, developmental, experiential, and environmental factors. The neurobiological changes that accompany drug addiction are not well understood. While until recently it was believed that addiction predominantly involved reward processes mediated by limbic circuits (see, for example, the reward deficiency syndrome ), results from recent neuroimaging studies have implicated additional brain areas, especially the frontal cortex. Here we summarize the findings from these neuroimaging studies and incorporate them with pertinent results from preclinical studies in suggesting a basis for an integrated model of drug addiction.
Most imaging studies have concentrated on the involvement of dopamine in the process of drug addiction because the ability of drugs of abuse to increase brain dopamine concentration in limbic brain regions is considered crucial for their reinforcing effects (2, 3). However, the increase in dopamine per se is insufficient to account for the process of addiction, since drugs of abuse increase dopamine in naive as well as in addicted subjects. In fact, in the case of cocaine addiction, the magnitude of drug-induced dopamine increases, and the intensity of self-reports of the drug’s reinforcing properties (the "high"), appears to be smaller in addicted than in naive subjects (4). This implies that dopamine involvement in drug addiction is likely to be mediated by means of functional and structural changes in the circuits that are modulated by dopamine, including the frontal cortex. In support of this suggestion are the findings of several recent structural/volumetric magnetic resonance imaging studies documenting morphological changes in the frontal lobe in various forms of drug addiction. For example, frontal lobe volume losses have been identified in cocaine-dependent subjects (5, 6), alcoholic subjects (7–9), and heroin-dependent subjects (5). The latter study noted negative correlations between normalized prefrontal volumes and years of either cocaine or heroin use, implying a cumulative effect of substance abuse on frontal volumes. Additional support is provided by studies done in rats. For example, it has been recently shown that self-administration of cocaine, but not food, results in morphological changes in dendrites and dendritic spines in the prefrontal cortex and nucleus accumbens (10). Dopamine activation, as seen during amphetamine administration, also suppresses the inhibition of the amygdala by the medial prefrontal cortex, possibly leading to a disinhibition of sensory-driven affective responses (11). A similar process may be occurring in human drug addiction, in which prefrontal top-down processes (see reference 12) are reduced, releasing behaviors that are normally kept under close monitoring and simulating stress-like reactions in which inhibitory control is suspended and stimulus-driven behavior is accentuated.
If the frontal cortex and its supervisory functions are indeed down-regulated in human drug addiction, the relevance of motivational, higher cognitive, and self-monitoring processes to this affliction cannot be overstated. Specifically, we propose that the behaviors and associated motivational states that are at the core of drug addiction are distinctly the processes of loss of self-directed/willed behaviors to automatic sensory-driven formulas and attribution of primary salience to the drug of abuse at the expense of other available rewarding stimuli. We hypothesize that these states are first evoked in the presence of the drug of abuse or cues conditioned to the drug but then become chronic action tendencies, contributing to relapse/bingeing (behavioral compulsion) and withdrawal/craving (mental compulsion, i.e., obsessiveness), respectively. We thus conceptualize drug addiction as a syndrome of impaired response inhibition and salience attribution and name it the "I-RISA" syndrome of drug addiction.
The I-RISA syndrome encompasses four clusters of behaviors that are interconnected in a positive feedback loop (F1) and depend on the functioning of the prefrontal circuits vis-à-vis the subcortical reward pathway.
The process of short-term drug administration is one that has been traditionally associated with higher extracellular dopamine concentrations in limbic brain regions, in particular, the nucleus accumbens (13, 14). However, there is also evidence of increased concentration of dopamine in frontal regions (15).
Craving is associated with the learned response that links the drug and its environment to a pleasurable or an intensely overpowering experience. The neuroanatomical substrates for consolidation of this memory are likely to involve the amygdala (16, 17) and the hippocampus (18), but activation of the thalamo-orbitofrontal circuit and the anterior cingulate might be a defining element in the actual experience of craving (19).
Compulsive Drug Administration
Compulsive drug self-administration in addicted individuals occurs even when the drug is no longer perceived of as pleasurable and in the presence of adverse physical reactions to the drug (20). This process of loss of control and drug bingeing is associated with dopaminergic, serotonergic, and glutamatergic circuits (21, 22) and probably involves the activation of the thalamo-orbitofrontal circuit and the anterior cingulate gyrus.
Recurrent drug administration and subsequent drug withdrawal result in disruption of behavioral circuits that culminate in dysphoria, anhedonia, and irritability (23), possibly contributing to relapse (24, 25). These changes are likely to involve disruption of frontal cortical circuits and neurotransmitters that include dopamine, serotonin, and corticotropin-releasing factor (26).
Here we review the results from neuroimaging studies that have assessed the effects of drug administration on functional measures, such as glucose metabolism and cerebral blood flow (CBF). Few studies have measured regional brain activity during drug intoxication, and most of these studies have employed a single drug exposure. Such studies have shown lower glucose metabolism throughout the brain, including the frontal cortex, during cocaine, morphine, or alcohol intoxication (27–30). In contrast, marijuana intoxication is associated with higher levels of glucose metabolism in the prefrontal cortex, orbitofrontal cortex, and striatum in marijuana abusers but not in nonabusers (31). Similarly, faster metabolism in the prefrontal cortex, anterior cingulate, orbitofrontal cortex, and striatum has been reported in cocaine abusers after sequential administration of intravenous methylphenidate, which cocaine abusers report to be similar to intravenous cocaine (19). It should be noted that activation in the orbitofrontal cortex was only observed in the subjects in whom methylphenidate induced intense craving and in the prefrontal cortex in the subjects in whom it enhanced mood.
Studies measuring the effects of short-term drug administration on CBF have consistently reported higher levels of prefrontal CBF during intoxication with nicotine (32), marijuana (33), and alcohol (34–36). Moreover, the activation of the right prefrontal cortex during alcohol intoxication was associated with euphoria (35) and during marijuana intoxication with the subjective sense of intoxication (33). In contrast, cocaine lowered CBF throughout the brain, including the frontal cortex, an effect that could be attributed to cocaine’s vasoconstricting effects (37).
Mapping studies during drug intoxication with functional magnetic resonance imaging (fMRI) to measure the blood-oxygenation-level-dependent (BOLD) response have reported activation of the prefrontal cortex and anterior cingulate gyrus during cocaine intoxication, an effect that has been strongly correlated with drug reinforcement properties (38). Studies of nicotine administration have also shown activation in the frontal cortex and anterior cingulate gyrus coinciding in time with the subjective experiences of "rush" and "high" (39).
The discrepancies in activation patterns reported could reflect vasoactive and drug-specific effects or the difference in the temporal course of the processes measured (metabolism studies, 30 minutes; CBF-water studies, 60 seconds; BOLD studies, 3–5 seconds). Since the metabolism and CBF-water studies are limited by their poor time resolution, the BOLD method may be better suited for assessing the relationship between regional changes and drug-induced fast behavioral effects, such as "rush" and "high." On the other hand, the BOLD method is limited by its sensitivity to vasoactive changes that may occur during drug administration.
To summarize, most of the studies show activation in the prefrontal cortex and anterior cingulate gyrus during drug intoxication when using either the CBF or BOLD methods. Also, prefrontal activation appears to be associated with the subjective perception of intoxication, the reinforcing effects of the drug, or enhanced mood. It is also intriguing that in the case of marijuana or methylphenidate, the activation of the frontal regions was predominantly observed in the abusers but not in the nonabusing subjects. This suggests that prefrontal regions and the anterior cingulate are involved in the intoxication process and that their response to drugs is in part related to previous drug experiences.
Acute drug administration is not necessary for the activation of the frontal cortex in individuals previously exposed to the drug of choice, in whom, because of prior exposure, craving alone is possibly sufficient to activate frontolimbic circuits. Thus, higher levels of brain activation (CBF, glucose metabolism, or BOLD) in frontolimbic areas, primarily in the prefrontal cortex and anterior cingulate, has been demonstrated in cocaine abusers exposed to videotapes depicting drug-related stimuli (40–44). Self-reports of craving significantly correlated with glucose metabolism changes in the dorsolateral prefrontal cortex in one study (42) and with the spatial extent of activation in the dorsolateral prefrontal cortex and anterior cingulate in another (43). In all five studies, the drug-related stimuli elicited craving only in the cocaine abusers and not in the comparison subjects, again pointing to the importance of prior drug experience.
The mechanism that underlies craving may entail recall of emotionally laden previous experiences. Indeed, craving correlated with activation of the amygdala in one study (42), and activation of the orbitofrontal cortex was observed when cocaine abusers recalled and described their own method of preparing cocaine but not when they described their family genealogical tree in a study from our laboratory (45) (F2). In addition, craving may involve anticipation of a future drug reward. The role of the dorsolateral prefrontal cortex in the anticipation of immediate drug self-administration was previously suggested (42), although the dissociation of craving from anticipation and of both from the actual drug experience has not been supported. Nevertheless, the actual drug experience may be related with more circumscribed activations than the anticipation phase (42), in line with evidence for a similar dissociation of anticipation from an actual sensory (tactile) experience (46).
Another line of evidence supporting the role of the frontal cortex in craving derives from studies conducted shortly after the subjects’ last drug use. For example, we have demonstrated higher regional brain glucose metabolism, including in the orbitofrontal cortex and striatum, in cocaine abusers tested during early withdrawal (<1 week since last cocaine use) than in normal comparison subjects (47). These higher levels were proportional to the craving intensity, such that the higher the metabolism, the greater the drug craving. A central role for craving in orbitofrontal cortex activation has also been suggested by the results of a study from our laboratory in which methylphenidate increased orbitofrontal cortex and striatal metabolism only in the subjects in whom it enhanced craving (19).
The problem remains that cocaine craving is not a direct measure of compulsive cocaine use, and, in fact, its association with drug use and relapse continues to be challenged (48). Demonstrating the involvement of the orbitofrontal cortex in compulsive drug self-administration would require investigation of the abuser during actual use in which drug supply is unrestricted. Alternatively, a paradigm that simulates compulsive behavior (such as gambling when it is clearly no longer beneficial) might offer invaluable insight into the circuits underlying loss of control in addiction.
Abnormalities in the human cortex associated with withdrawal from cocaine in regular cocaine abusers were documented as early as 1988 in our laboratory (49). We demonstrated that the relative CBF values for the prefrontal cortex and the left lateral frontal cortex were significantly lower in the cocaine users than in the normal comparison subjects. A follow up study in active cocaine abusers demonstrated differences in regional brain glucose metabolism between cocaine abusers tested within 1 week of last cocaine use and cocaine abusers tested 2–4 weeks after last cocaine use (47). Of interest, glucose metabolism was higher in the orbitofrontal cortex and in the striatum in the former group than in the normal comparison subjects. During more protracted withdrawal (1–6 weeks since last use), brain metabolism was found to be lower in cocaine abusers than in normal comparison subjects, an effect that was most accentuated in the frontal cortex (F3) (50).
Studies of alcohol abusers have provided similar evidence. For example, glucose metabolism abnormalities (including in the frontal cortex) were documented in otherwise healthy alcoholic subjects with mean duration of alcohol withdrawal of 11 days (51). Persistent lower striatal metabolism has been shown in regional metabolism studies after more protracted alcohol withdrawal (52). In addition, alcoholic subjects have shown less sensitivity to the lower metabolism induced by lorazepam, a benzodiazepine that facilitates g-aminobutyric acid neurotransmission in the striatal-thalamo-orbitofrontal cortex circuit during early (1–4 weeks) detoxification (53) and in the orbitofrontal cortex during protracted (8–11 weeks) detoxification (54), suggesting long-lasting drug-related adjustments in these brain regions. Persistent abnormalities after alcohol detoxification were also documented for the anterior cingulate (54). Lower activity in the prefrontal cortex in alcoholic subjects during detoxification was also documented in other laboratories using slightly different study groups (Cloninger-type 2 alcoholics) and techniques (single photon emission computed tomography) (55).
Alcoholic subjects show less sensitivity in the striatal-thalamo-orbitofrontal cortex circuit to the serotonin agonist m-chlorophenylpiperazine, which provides evidence for the relevance of serotonin in these abnormalities (56). Studies from our laboratory also point to the relevance of dopamine in withdrawal. First, we documented that in cocaine abusers during early (up to 1 month since last cocaine use) and protracted (up to 4 months since last cocaine use) withdrawal, striatal dopamine response or receptor availability was significantly lower (4, 57, 58) than in normal comparison subjects. We also reported lower striatal dopamine D2 receptor binding in heroin (59) and methamphetamine (60) abusers and in alcoholic subjects (61) (F4). Moreover, the lower levels of striatal D2 receptors were found to be associated with lower metabolism in the orbitofrontal cortex and anterior cingulate gyrus in cocaine addicted subjects (58) (F5) and in the orbitofrontal cortex in methamphetamine abusers (62). Finally, higher metabolism in the anterior cingulate gyrus has been shown in response to methylphenidate, which increases dopamine by blocking the dopamine transporter (19), providing further support for the role of lower dopamine activation in frontal hypometabolism in drug addiction.
For a review of the role of the mesolimbic and mesocortical dopamine system in inhibitory control and motivational processes in the rodent and primate brain and their putative impact on drug seeking, see Jentsch and Taylor (63). Here we will mention the mesolimbic and the mesocortical dopamine systems, which are classically associated with drug reinforcement and addiction. The mesolimbic dopamine circuit, which includes the nucleus accumbens, amygdala, and hippocampus, has been traditionally associated with the acute reinforcing effects of a drug and with the memory and conditioned responses that have been linked to craving. It is also likely to be involved in the emotional and motivational changes seen in drug abusers during withdrawal. The mesocortical dopamine circuit, which includes the prefrontal cortex, orbitofrontal cortex, and anterior cingulate, is likely to be involved in the conscious experience of drug intoxication, drug incentive salience, drug expectation/craving, and compulsive drug administration. Because these circuits operate in parallel and interact with one another, it is likely that a given behavior involves, to a greater or lesser extent, their joint participation. The nature of these interactions affects response to the drug. For example, the activation of memory circuits (the hippocampus and amygdala) in association with a drug-related context activates the orbitofrontal cortex and anterior cingulate in expectation of the reinforcer, which in turn activates the dopamine cells (64), leading to a further increase in the craving sensation and a possible decrease in inhibitory control. Note the circular nature of this interaction: the attribution of salience to a given stimulus, which is a function of the orbitofrontal cortex, depends on the relative value of a reinforcer compared to simultaneously available reinforcers (65), which requires knowledge of the strength of the stimulus as a reinforcer, a function of the hippocampus and amygdala. Consumption of the drug in turn will further activate cortical circuits (the orbitofrontal cortex and anterior cingulate) in proportion to the dopamine stimulation by favoring the target response and decreasing non-target-related background activity (66). The activation of these interacting circuits (F6) may be indispensable for maintaining the compulsive drug administration observed during bingeing and to the vicious circle of drug addition (F7).
A question remains why some people become addicted and others, under similar circumstances, do not. This question is one of the most challenging issues in drug abuse research. The neurobiological mechanisms underlying vulnerability to drug addiction are poorly understood and are likely to involve a balance between factors that confer vulnerability and those that protect against it. On the basis of imaging studies, one cannot rule out the possibility that the changes in frontal activity in drug-addicted subjects may have antedated their drug use and could have enabled the loss of control and compulsive drug intake in these subjects. But it is also possible that the changes in frontal activity are secondary to recurrent drug use and that other variables are responsible for the addiction vulnerability. We and others have hypothesized that the decrease in activity of D2 receptors may put individuals at risk for addictive behaviors as a means of temporarily compensating for the fewer D2-regulated reward circuits (67). Moreover, we have shown that overexpression of D2 receptors in the nucleus accumbens of rats previously trained to self-administer alcohol markedly reduces their alcohol intake (68). Since D2 receptor availability is positively associated with frontal activity in the human brain (reviewed in reference 69), this suggests that one of the mechanisms by which D2 receptors regulate drug self-administration, and possibly the potential for addiction, is by modulating frontal activity.
In the next several paragraphs, we summarize the functional neuroimaging studies conducted in healthy comparison subjects or non-drug-abusing populations that have recently implicated the striatal-thalamo-orbitofrontal cortex circuit in the I-RISA components. Specifically, we target perception of response-reinforcement relations and response inhibition, but we will also discuss the expectation of reward and depression. These four components can be viewed as intricately related to the four dimensions of our drug-addiction model (F1), each potentially predisposing to drug addiction; thus
1. Drug intoxication is associated with the experience of its strong positive and negative reinforcement effects, an association that is strengthened through repeated self-administration and that possibly hinders the formation of similar associations; attribution of primary salience to the drug occurs at the expense of less powerful reinforcers.
2. We conceive impairment in response inhibition as underlying the experience of relapse and bingeing. When response-reinforcement regulation is down because of impaired salience attribution, response disinhibition, or impulsive responding to immediately salient, drug-related rewards is expected.
3. Expectation of the effects of the drug of abuse, whether it is the "high" or a lower negative state, is integral to drug craving.
4. Dysthymia is a core symptom of withdrawal, possibly reflecting adaptation responses to repeated dopamine enhancement by drugs of abuse in the reward circuits that render the latter less responsive to natural reinforcers (70–72). Behaviorally, this lower sensitivity in the reward circuits may represent a generalized impairment in the ability to derive pleasure from non-drug-related stimuli, leading to a state of anhedonia, which puts drug-addicted individuals at greater risk for seeking drug stimulation.
Response-Reinforcement Relations and Salience Attribution
Regarding perception of response-reinforcement relations, several recent fMRI studies have monitored brain hemodynamic changes during performance of game-playing tasks with monetary reinforcers (see reference 73 for review). Responses to monetary gains and losses or to winning and losing game points have been noted in the prefrontal cortex, orbitofrontal cortex, anterior cingulate, and thalamus. Activations of these areas in guessing, compared to reporting (the orbitofrontal cortex was exclusively activated in the guessing task in the more difficult condition, in which probability of being correct was down to 25% from 50%), point to a unique role of the correctness of a response and greater dependence on feedback under conditions of inherent uncertainty (74). Indeed, a recent fMRI study documented that unpredictability of a reward (water or juice) correlated with activity in the orbitofrontal cortex, thalamus, and nucleus accumbens (75). Positron emission tomography (PET) studies have similarly demonstrated striatal-thalamo-orbitofrontal cortex activations (including the anterior cingulate and dorsolateral prefrontal cortex) in association with gambling (76, 77), receiving a salient feedback (78), or receiving a monetary reward (79).
The processing of emotionally salient and behaviorally adaptive information may be at the core of advantageous assessment of response-reinforcement relations. Indeed, the role of the frontal cortex, and specifically the anterior cingulate, in emotional processing has been demonstrated in several PET studies (e.g., references 80–83). Consistent with these studies are the results of a fMRI study (84) in which the inferior frontal gyrus and dorsal anterior cingulate were involved in making a semantic, emotionally laden versus an orthographic decision in a verbal go/no-go task. Other fMRI studies have implicated the orbitofrontal cortex and anterior cingulate in the experience of pleasant sensations (85) and the orbitofrontal cortex in recognizing fearful, angry, and disgusted emotional facial expressions compared to neutral expressions (86). Of interest, in a PET study (87), angry but not sad faces specifically activated the orbitofrontal cortex, proportionally with the increasing intensity of the emotion, while the anterior cingulate cortex was coactivated by both expressions. Taken together, the results of these studies suggest an important integrative role for the orbitofrontal cortex and anterior cingulate in the analysis of the information that carries emotive, evaluative, and, in the long-term, survival significance for an individual, which comprise salience attribution, an integral part of our I-RISA syndrome of drug addiction.
The other component of the proposed I-RISA syndrome is the control of behavior, which is assumed to break down in periods of relapse and drug bingeing. Response inhibition has been relatively well studied in neuroimaging paradigms. For example, the orbitofrontal cortex, anterior cingulate cortex, and striatum were activated in a go/no-go task in two fMRI studies (88, 89). Better response inhibition was associated with greater volume of activation in the orbitofrontal cortex and a smaller magnitude of activation in the anterior cingulate cortex, possibly implicating the orbitofrontal cortex in the effort exerted when inhibiting a response and the anterior cingulate cortex in error detection (88). Further support for the role of the anterior cingulate in response inhibition, including response competition and selection, is provided by other fMRI (84, 90–92) and PET (e.g., reference 93) studies of the go/no-go paradigm. In addition, the role of the anterior cingulate in response inhibition has been established in studies of the suppression of prepotent response tendencies by using the Stroop effect (e.g., reference 94). We have recently provided more direct evidence for the role of the prefrontal circuit in response inhibition in drug addiction (95). We examined the association between Stroop interference and relative glucose metabolism in selected prefrontal brain regions in cocaine-addicted subjects, alcoholic subjects, and comparison subjects. The results revealed that for the cocaine-addicted subjects and alcoholic subjects, higher levels of orbitofrontal cortex metabolism at baseline was associated with lower conflict (higher Stroop interference score), while for the comparison subjects, higher orbitofrontal cortex metabolism was associated with higher conflict (lower Stroop interference score), suggesting a change in the role of the orbitofrontal cortex as a function of addiction.
Supporting the role of the frontal cortex in expectation is an fMRI study that demonstrated distinct brain regions and different response characteristics in anticipation of pain versus the experience of pain, with the former activating more anterior regions (including the anterior medial frontal cortex) than the latter (46). Activation of the orbitofrontal cortex has also been associated with expectation in several PET studies, including expectation in tasks of visual attention (96) and in tasks involving a shock (97, 98).
Finally, an association between depression and prefrontal abnormalities has been demonstrated in neuroimaging studies conducted in depressed patients, with suggested disruptions of frontostriatal (99) and corticolimbic (100) networks. Results of these studies revealed resting abnormalities in the dorsolateral, ventrolateral, and medial aspects of the prefrontal cortex and the anterior cingulate, blunted responses in the anterior cingulate and medial prefrontal cortex to behavioral and pharmacological challenges, and abnormalities localized to the orbitofrontal cortex (100, 101). Lower activity in the striatum of depressed patients in the resting state and in response to a reaction-time task and feedback have also been reported (102, 103).
The imaging studies reviewed here provide evidence for the involvement of the frontal cortex in the various aspects of drug addiction, including reinforcing responses to drugs during intoxication, activation during craving, and deactivation during withdrawal. The involvement of the frontal cortex throughout these cyclical stages of addiction is likely to play an important role in the cognitive behavioral and emotional changes that perpetuate drug self-administration and that are highlighted in the I-RISA syndrome of drug addition.
Future investigations should target the interplay between impaired salience attribution and response inhibition and their possible causal or predisposing effects toward developing drug addition.
Received Dec. 6, 2001; revision received April 18, 2002; accepted April 24, 2002. From Brookhaven National Laboratory. Address reprint requests to Dr. Goldstein, Brookhaven National Laboratory, Medical Research, Bldg. 490, 30 Bell Ave, Upton, NY 11973; firstname.lastname@example.org (e-mail). Supported in part by the U.S. Department of Energy (Office of Health and Environmental Research) under contract DE-ACO2-98CH10886, the National Institute of Drug Abuse (under grant DA-06891), and the Institute of Alcohol Abuse and Alcoholism (under grant AA-09481).
Behavioral Manifestations of the I-RISA (Impaired Response Inhibition and Salience Attribution) Syndrome of Drug Addiction
Orbitofrontal Cortical Activation in Active Cocaine Abusers During a Cocaine Theme Interview and a Neutral Theme Interview, as Measured by FDG PET
Lower Relative Glucose Metabolism in the Prefrontal Cortex and Anterior Cingulate Gyrus of a Cocaine Abuser Than in a Normal Comparison Subject
Lower Striatal Dopamine D2 Receptor Binding in Drug Users During Withdrawal From Cocaine, Methamphetamine, and Alcohol Than in Normal Comparison Subjects
Relation of Striatal Dopamine D2 Receptor Availability and Orbitofrontal Cortical Metabolism in Cocaine Usersa,b
aAdapted from an earlier article (59). Reprinted from Synapse, ©1993 Wiley-Liss, Inc., with permission.
Interactions of the Mesocortical and Mesolimbic Circuits in Drug Addiction
Integrative Model of Brain and Behavior: the I-RISA (Impaired Response Inhibition and Salience Attribution) Syndrome of Drug Addiction