Location, location: using functional magnetic resonance imaging to pinpoint brain differences relevant to stimulant use
Martin P. Paulus MD, Department of Psychiatry, UCSD, 8950 Villa La Jolla Drive, Suite C-213, La Jolla CA 92037–0985, USA. E-mail: firstname.lastname@example.org
Aims The purpose of this review is to summarize the neural substrate dysfunctions and disrupted cognitive, affective and experiential processes observed in methamphetamine and cocaine-dependent individuals.
Methods We reviewed all publications in PubMed that conducted comparison studies between healthy volunteers and cocaine-, amphetamine- or methamphetamine-dependent individuals using functional magnetic resonance imaging.
Results Stimulant dependence is characterized by a distributed alteration of functional activation to a number of experimental paradigms. Attenuated anterior and posterior cingulate activation, reduced inferior frontal and dorsolateral prefrontal cortex activation and altered posterior parietal activation point towards an inadequate demand-specific processing of information. Processes reported most consistently to be deficient in these functional neuroimaging studies include inhibitory control, executive functioning and decision-making.
Conclusion One emerging theme is that stimulant-dependent individuals show specific, rather than generic, brain activation differences, i.e. instead of showing more or less brain activation regardless of task, they exhibit process-related brain activation differences that are consistent with a shift from context-specific, effortful processing to more stereotyped, habitual response generation.
Methamphetamine and cocaine dependence are brain disorders characterized by repeated substance use and loss of control, despite the presence of negative consequences. In particular, methamphetamine abuse is an important and growing problem in the United States. Although the prevalence of methamphetamine use among people aged 12 years and older did not change significantly between 2002 and 2004 [1,2], the percentage of frequent users (individuals who had used methamphetamine within the past month) meeting criteria for stimulant dependence increased dramatically from 10.6 to 22.3%. This study also showed that in 2004, more than 1.6 million people (0.6% of the US population) met criteria for cocaine dependence . Tragically, there are no pharmacological treatments available for this devastating disease. A vital first step towards achieving treatment goals is discovering the neurobiological underpinnings of the disease.
Cocaine and amphetamine-like stimulants affect the central nervous system at the synaptic level. Cocaine's primary mechanism of action is thought to be the blockade of the dopamine transporter, whereas other psychostimulants predominantly promote the release of monoamines such as dopamine via reverse transport . Methamphetamine, a compound structurally similar to amphetamine, possesses an additional methyl group which enhances its lipid solubility, so that it crosses the blood–brain barrier more readily than amphetamine. More importantly, the low boiling-point of methamphetamine allows for its smoking and subsequent inhalation, thereby accelerating entry of the active drug into the blood and central nervous system. Apart from its acuteeffects, methamphetamine has been shown to down-regulate striatal D2 dopamine receptors  and dopamine transporters in the striatum , orbitofrontal cortex and dorsolateral prefrontal cortex . Importantly, both cocaine and amphetamines reduce the availability of D2 receptors, which is thought to alter the individual's ability to process reward-related behaviors.
Although the basic pharmacology of these drugs is well understood, how these effects translate to neural system dysfunctions and the development of dependence is much less clear. Nevertheless, there is ample evidence that individuals with stimulant dependence show dysfunctions in cognitive, affective and experiential processes. Specifically, several neuropsychological studies have found that chronic stimulant-dependent subjects have problems with memory [7–11], attention [12,13] and executive functioning . In particular, methamphetamine users show impairments in explicit memory , attention  and selective inhibition , which some investigators have compared to dysfunctions in individuals with fronto-striatal damage . Other studies have shown that use of either methamphetamine or cocaine worsens performance on the Stroop task [16,17], which is consistent with the general hypothesis that stimulant use alter the individual's ability to attend selectively to stimuli or to inhibit pre-potent responses.
It is important to emphasize that stimulant dependence is not a simple, single-location brain area disease. Rather, brain imaging tools and systems neuroscience approaches are beginning to delineate networks affected by chronic stimulant use. Functional magnetic resonance imaging (fMRI) has emerged as an invaluable method for correlating altered neural substrate activation with specific cognitive or affective dysfunctions due to stimulant use. Although magnetic resonance imaging lacks the chemical specificity of single photon emission computerized tomography (SPECT) and positron emission tomography (PET), it possesses several advantages over these imaging methods. First, functional MRI is non-invasive and does not require radiotracer compounds, thus ionizing radiation risks are not a concern. Secondly, it provides superior spatial resolution to allow distinction of small cortical regions . Thirdly, because fMRI reflects brain activation via blood oxygen level-dependent (BOLD) responses that are indicative of blood flow changes, its temporal resolution, while lagging behind that of electrophysiological methods, exceeds that of PET and SPECT and is sufficient for event-related paradigms. All features considered, fMRI has excellent ‘functional resolution’ for examining substrates which may be affected by stimulant use. When comparing drug users to controls, investigators may begin by evaluating groups for differential percentage change in BOLD activation in relevant brain areas. Altered activation patterns, i.e. either more and less activation of a particular brain area in a target population group, can be due to several mechanisms. The brain region could be (1) more or less engaged in the active task; (2) more or less engaged in the comparative (control) task, resulting in an altered activation difference; or (3) similarly engaged but exhibiting different levels of performance. Moreover, anatomical differences resulting from long-standing drug use may also affect BOLD patterns. Therefore, one has to examine carefully task-related activation, level of performance on both target and control task and characteristics of the control task in order to interpret the brain imaging differences across groups. Additionally, although fMRI may be used to demonstrate differences between groups, the sole comparison of group differences is insufficient to elucidate the cause and effect relationships behind these differences. For example, whether altered BOLD patterns observed in stimulant users are due to pre-existing brain processing differences or sequelae of long-term substance use has yet to be resolved. Currently, longitudinal studies with at-risk populations are under way to address this question.
This review aims to summarize the current knowledge obtained by fMRI studies of the brain systems affected in subjects with cocaine or methamphetamine dependence. In this assessment, we report solely on studies conducted using fMRI and behavioral tasks, and exclude other imaging methods [structural MRI, PET, SPECT, magnetic resonance spectroscopy (MRS)]. The selected publications are studies examining brain activation in individuals with cocaine or methamphetamine dependence relative to healthy comparison subjects using different behavioral paradigms. More recently, several research groups have begun to utilize fMRI and other neuroimaging techniques to assess brain activity differences between stimulant-dependent subjects who relapse and those who maintain abstinence. Results from these studies may ultimately provide clinicians with a practical tool to quantify risk for relapse and develop risk-specific treatment interventions. In addition, we refer to, but do not review formally, studies of acute cocaine administration in dependent subjects to highlight brain structures that are important for the processing of drug expectancy and craving. Finally, studies of pre-natal stimulant exposure and adolescent populations are also outside the scope of this review.
This review is organized using a matrix approach of brain areas and cognitive processes. We consider the changes in brain activation as columns of a matrix and the different tasks used as rows of this matrix. Emerging differences across tasks support the idea that a particular brain area is generically dysfunctional in stimulant-dependent subjects. Alternatively, differences across brain areas within the same cognitive task highlight the fact that the brain acts as a system and that methamphetamine dependence is not simply a brain lesion disorder. Thus, we discuss the differences both from a process point of view as well as from a neural substrate perspective. By cross-examining the topic in this manner, we are able to integrate recent findings more effectively into the existing stimulant and neuroimaging literature.
SUMMARY OF FINDINGS ACCORDING TO NEURAL SUBSTRATES
Frontal brain areas comprise three distinct regions, i.e. the dorsolateral prefrontal cortex, the orbitofrontal cortex and the ventromedial cortex. The dorsolateral prefrontal cortex is involved primarily in executive functions and higher-order cognition, including memory, language, impulse control and problem-solving. The ventrolateral regions [Brodmann areas (BA) 45, 47] have been shown to mediate the selection of information for short- and long-term memory [19–21], while mid-dorsal regions (BA 9, 46) monitor information in working memory and play a role in sustained attention. In comparison, the orbitofrontal cortex is associated with processing the value of food, odor, liquid and other primary reinforcers. This area has also been implicated in higher-order reward and reinforcement processing, such as the assessment of short- versus long-term gains or losses . The inferior prefrontal cortex, which encompasses the lateral lower surface of the brain, has been implicated in a number of processes such as inhibition , detection of irregularity  and processing of decision outcome . Frontal and prefrontal areas mediate executive functions relevant in the study of drug abuse and relapse: attending to and assessing situations, recognizing potential consequences, making decisions, planning and coordinating behavior and inhibiting inappropriate behavior. A study of methamphetamine users revealed that, while performing a prediction task, methamphetamine-dependent subjects failed to activate ventromedial cortex corresponding to BA 10 and 11 and activated less dorsolateral prefrontal cortex (BA 9). These frontal activation differences were accompanied by an increased susceptibility to the influence of immediately preceding trial outcomes . Thus functional changes in dorsolateral, orbitofrontal and ventromedial cortices may impact the executive system, as evidenced by the atypical patterns of expectancy, compulsion and decision-making observed in users of cocaine and methamphetamine [25,26].
As part of the medial wall of the frontal cortex, the anterior cingulate cortex (ACC) rostrally surrounds the corpus callosum and has been implicated in processes that are important for cognitive control . The term ‘cognitive control’ refers to a set of functions which enables the cognitive system to perform adaptively specific tasks in the presence of stimuli and other behavioral options competing for attentional resources and enables one to redirect processing strategies when goal-states change . Therefore, the cingulate cortex is involved actively in tasks that involve effortful orchestration of emotion, sensory, motor and cognitive functions. In particular, anterior cingulate function has been associated with the detection or processing of errors as well as the processing of responses within a context of stimulus conflict [29,30]. The anterior cingulate has also been described as an initiator of appropriate responses and a suppressor of inappropriate behaviors . Therefore, dysfunction in ACC has been observed in a number of different psychiatric conditions, including substance use disorders. Specifically, reduced activity in the ACC has been demonstrated in marijuana users , opiate addicts , schizophrenia patients  and grieving subjects . Most notably, stimulant-dependent subjects relative to comparison subjects showed lower activation in this area during tasks that require an individual to inhibit pre-potent responses [36,37]. The pattern of attenuated anterior cingulate activity in cocaine and methamphetamine users also occurred in paradigms involving executive functions such as verbal working memory , decision-making  and stress imagery . In contrast, cocaine-dependent individuals activated the anterior cingulate more so than controls when exposed to audiovisual drug-related cues [41,42]. This activation preceded the experience of craving , which was shown to correlate with the level of ACC activation [40–42]. Taken together, these studies suggest that the anterior cingulate in stimulant-dependent individuals is less able to redirect cognitive control to appropriate and adaptive behavioral patterns in general, but is overly sensitive to redirect behavior in response to drug-related cues.
In comparison to the ACC, the posterior cingulate cortex (PCC) is located below the parietal cortex and receives afferents primarily from the anterior thalamic nuclei and from extensive cortical areas in the frontal, parietal and temporal lobes . This structure has been described as the ‘evaluative region’ of the cingulate  and is frequently involved in the degree to which actions lead to predictable (versus unpredictable) outcomes  and controlled (versus uncontrolled) movements . Similarly, parietal areas and the PCC participate in perspective shifting and judgment of self [46–49]. Neuronal activity in the posterior cingulate cortex has also been linked to making risky choices , which could contribute to both the initiation of drug use and its continuation. Studies in cocaine-dependent populations demonstrated that a network of structures including the cingulate gyrus activated selectively during cocaine-induced craving [51,52]. Moreover, change of activity in left anterior cingulate and right posterior cingulate cortices correlated inversely with the change of cocaine craving rating during stress imagery . Greater BOLD activation in posterior cingulate cortices was also observed in an investigation of relapse in cocaine-dependent patients, in which the increased activation correlated with poorer treatment effectiveness scores . Studies in control and drug-using populations show that the PCC may play a crucial role in judgment and decision-making as they relate to the processing of self-relevant stimuli and the ability to control self-generated action outcomes. The degree of activation in this area may be useful to quantify the degree of intact decision-making as it relates to stressful or challenging situations, such as cueing environments which present temptation to relapse.
The parietal cortex has been viewed traditionally as important for attentional processing, with additional roles proposed in episodic and working memory, skills learning and spatial perception and imagery . As part of the executive system, this structure maintains close cortico–cortical connections to all areas of the frontal cortex.
The right inferior parietal lobule, a portion of the parietal cortex which maintains bidirectional connections to right dorsolateral prefrontal and anterior insular cortex, has been implicated in a number of decision-making processes. These include sustained, and possibly selective, attention , voluntary attentional control , inhibitory control  and local to global task-relevant switching , as well as distinguishing between task-relevant and task-irrelevant events . Thus, this area may be critical for the extraction and selection of task-relevant information. Moreover, several studies have implicated this structure in autonomic arousal processes , risk-taking decision-making  and guessing . The right inferior parietal lobule, which integrates attentional resources to select actions (i.e. predicting the location of the stimulus) and inhibits previous prediction strategies as they relate to success or failure, may be critical for the assessment process during decision-making.
In the posteromedial portion of the parietal area, the precuneus (BA 7) is engaged in imagery, episodic memory retrieval and consciousness of self (e.g. first-person perspective, personal identity, past personal experiences). It is also ‘strongly interconnected’ with the frontal cortex  and maintains connections with medial and lateral parietal cortex, posterior and anterior cingulate, thalamus, striatum, claustrum and brain stem. The precuneus, together with the posterior cingulate, was shown to be active when reflecting upon one's own personality traits, whereas the anterior cingulate activated during reflection upon one's own physical traits . Other regions implicated in first-person perspective tasks are medial parietal and prefrontal cortices and the right inferior parietal cortex . Methamphetamine-dependent subjects decreasingly activated areas within the post-central gyrus (BA 5) and precuneus (BA 19, 7) while completing a decision-making task . Areas of activation within the precuneus and inferior parietal lobule (BA 40) of these subjects correlated with different set error rates, implying that stimulant users in comparison to controls are influenced differently by experiences of success and failure. The less predictable the outcome, the more these subjects activated these parietal areas, as well as the left medial frontal gyrus (BA 10) and left insula (BA 13). In cocaine-dependent participants, stress was associated with less activation in the post-central gyri , while the precuneus showed less activation during a verbal working memory task . Future investigation into the contribution of emotional and social context to self-processing may reveal more involvement of parietal regions in patterns of drug use and abuse. Research thus far suggests that these areas are less activated in stimulant-dependent individuals, except during decision-making when the outcome of a decision is less certain.
Altered processing of errors and outcomes by the parietal cortex during decision-making in methamphetamine-dependent individuals is consistent with PET imaging findings by Volkow and colleagues , which have shown increased glucose metabolism even after prolonged abstinence in this area. Moreover, animal experiments have demonstrated parietal cortex neuronal degeneration following methamphetamine administration. Fundamentally, altered parietal cortex modulation of attentional and decision-making processes may constitute part of the progressive process of dependence resulting in a less adaptive and more habit-driven stimulus processing and response formation.
Occipito-temporal brain structures, which include the superior temporal gyrus (STG) and the middle temporal gyrus, are involved in lexical–semantic processing . The STG has primarily been implicated in semantic processing [69,70], in particular of spoken words  and object-naming . However, the role of these structures in non-speech-related processing has only recently been appreciated. Specifically, the role of the STG in non-speech-related processing has been described as a correlate of temporal action planning , semanticjudgment , processing complex configurations of symbolic information , hand imitation , cross-modal integration  and artificial grammar learning . In addition, this structure is thought to identify salient events in the sensory environment both within and independent of the current behavioral context . Thus, the STG appears to be critical for providing information that occurs over time in order to identify, process or prepare for actions, which have to be integrated over time. In the context of this investigation, distinct portions of the STG activate differentially as a function of prior response and prior outcome. In cocaine-dependent individuals, poorer treatment effectiveness scores were correlated with activation in the left superior temporal gyrus , a finding that is consistent with the altered activation related to relapse in methamphetamine-dependent subjects  and cerebral metabolic and hypoperfusion anomalies seen in these areas . As the role of these structures beyond speech processing continues to be elucidated, so will their possible contributions to drug use and dependence.
Limbic and paralimbic regions
These regions, which encompass the amygdala, hippocampus, parahippocampal gyrus and insula, are important for processing of emotions. A network of the amygdala, insula, anterior cingulate gyrus and medial prefrontal cortex functions to identify the emotional significance of a stimulus, generate an affective response and regulate an affective state . The insula, a structure deep to the frontal, temporal and parietal lobes within the lateral fissure, maintains afferent and efferent connections to the medial and orbitofrontal cortices, anterior cingulate and several nuclei of the amygdala . Although activation of the insula has been frequently associated with disgust , unease and anticipation of adverse events, there is increasing evidence of a broader role for this brain area in emotion processing . Insular activation is thought to be involved in differential positive versus negative emotion processing , in particular fearful face processing , pain perception [87,88] and emotional valence judgements .
There is probably a role for the insula in inhibition and/or anticipation and processing of particular outcomes. During successful inhibitions in a go–no-go task, the right insula was one of two regions activated increasingly in controls than in cocaine-dependent individuals . The left insula was significantly more active in controls relative to users when they failed to inhibit the pre-potent response. Although Paulus et al. observed an error-rate effect in the right insula during a decision-making task, there was no group difference found between methamphetamine-dependent and control participants . In the left insula, however, a group by error-rate effect was found showing an inverse relationship of brain activation to outcome predictability in methamphetamine users. This observation, along with the left insula activity attenuation found by Kaufman et al., suggests that stimulant users are less engaged in processing success versus failure when making decisions or attempting to inhibit their behavior . Overall, the insula is generally less activated in stimulant users; however, task conditions, such as outcome predictability in a decision-making paradigm, show a more complex relationship between BOLD activation and drug use in this limbic region.
The altered function of the insula, which is critical for negative emotion processing , may provide a link to the clinical finding of increased negative emotionality in subjects who are at risk for stimulant dependence. This structure has also been implicated by London and colleagues in PET studies . At this stage, however, it is not clear whether altered insula functioning is a pre-existing condition, increasing the likelihood that individuals engage in drug-taking behavior, or a consequence, resulting from years of use.
SUMMARY OF FINDINGS ORGANIZED BY COGNITIVE/AFFECTIVE PROCESSES
Inhibition and Attention
Inhibition is the process by which the execution of a thought, action or emotion is overridden or reversed. Whether it involves stopping at a red light or not spending beyond one's means, inhibition is an important function for daily life. Several studies with healthy volunteers have identified critical neural substrates of inhibition during go no-go performance, which include the inferior and middle frontal cortices [58,90–96], inferior parietal lobe [90–92,95,97,98], anterior cingulate [99,100], precuneus [91,92,95,98], insula [59,91,97,101] and superior temporal gyrus [91,95].
Kaufman et al. examined the degree of inhibitory control and action monitoring using a go no-go task. Cocaine users perform poorly on this task compared to controls [36,102,103]. During successful inhibition events, cocaine-using subjects activated the anterior cingulate and the right insula less robustly. Areas within the anterior cingulate, right medial frontal gyrus/pre-SMA, left inferior frontal gyrus and left insula were hypoactive during unsuccessful inhibitions. An insufficient anticipatory response from the left insula possibly contributes to the mechanism behind the inability to inhibit. The attenuated responsiveness of the anterior cingulate is not surprising, considering that it appears to play a role in urgent inhibitions .
Hester & Garavan  examined inhibition by administering a go no-go task, which also had a significant working memory component. After being trained to remember a list of letters, participants were shown presentations of single letters and instructed to press a button in response to a letter not from the list (go trials), but to refrain from responding when shown a familiar letter (no-go trials). Go trials outnumbered no-go trials to create a pre-potency for the button press response. Task performance (measured by the number of successful inhibitions) was significantly better in controls than in cocaine-dependent subjects. This advantage increased proportionately to the level of working memory demand (memory list length). Controls activated the ACC, pre-supplementary motor area and right superior frontal gyrus more so than cocaine users. Increased memory load was associated with a pattern of increased ACC activation in the control group and increasingly activated areas within the left cerebellum (specifically, the culmen and pyramis) of cocaine users. In the control subjects, cerebellar areas were decreasingly activated with increasing working memory demands. The authors interpreted the results by proposing that cocaine users have a reduced capacity for activating the ACC and prefrontal regions when presented with a challenging inhibition task. Recruiting the left cerebellum could possibly compensate for insufficient activity in the prefrontal areas or serve to promote more activation in frontal areas, as cerebro–cerebellar loops have been purported [104,105], and activations of the cerebellum and contralateral frontal lobe shown to occur concurrently .
To investigate executive functioning, Kübler and colleagues  devised three paradigms of attention switching within working memory: a verbal working memory task, a visuospatial task and a combined task. Functional data from the visuospatial and combined tasks were deemed ambiguous, because task performance differed between cocaine users and control subjects. During the verbal task, cocaine-dependent participants displayed less activation in frontal areas (cingulate, medial frontal and middle frontal gyri), the right precuneus and two subcortical areas (left thalamus and left lentiform nucleus). The left cingulate, right precuneus and left culmen of the cerebellum contained areas where there were significant interactions between group and task demands. These results are consistent with behavioral findings, which have shown that cocaine and methamphetamine users perform poorly on the measures of selective attention [16,17]. These behavioral effects could be due to the drug acting upon the orbitofrontal cortex and anterior cingulate, as lesions within these areas have been associated with inattention and poor Stroop task performance [17,107]. Therefore, reduced attention or attending to inappropriate stimuli may be a fundamental cognitive dysfunction either associated with or induced by long-standing use of stimulants.
Paulus and colleagues  reported on dysfunctions in decision-making and associated decreased cortical activation among methamphetamine users. Subjects performed a two-choice prediction task in which they chose between two possible outcomes and choices were reinforced to be ‘correct’ 50% of the time. Relative to control participants and the control task (two-choice response task), methamphetamine users failed to activate, or activated less, regions within the dorsolateral and orbitofrontal cortices (BA 9, 10, 11) while performing the prediction task. Methamphetamine-dependent individuals also exhibited less task-related activation in the right inferior frontal gyrus (BA 44) and left middle frontal gyrus (BA 46). The degree to which responses were affected by the previous outcome also differed between groups; the methamphetamine-dependent participants' choices were more influenced by the outcome of the immediately preceding choice, leaning toward a ‘win–stay/lose–shift’ strategy. This behavioral difference diminished with increased duration of abstinence from stimulants. Stimulant use duration could be predicted by the interaction effect between task and group within the area of the orbitofrontal gyrus (BA 11); subjects with a shorter duration of methamphetamine use activated more within this region than subjects with a longer use history. In a follow-up study , Paulus and colleagues investigated the effect of outcome on decision-making in methamphetamine-dependent individuals by varying the success rate of a two-choice prediction task. Again, the methamphetamine users exhibited a significant preference for the win–stay/lose–shift pattern when compared to control subjects, and showed less task-related activation in bilateral inferior prefrontal cortex, dorsolateral prefrontal cortex, bilateral parietal cortex, left post-central gyrus and left superior temporal gyrus. Both groups were affected by error rate so that they activated more the right insula, right inferior and right middle frontal gyrus in the low error rate condition, and activated the least under the most unpredictable condition (50% error rate). Significant group–error rate interactions were reported in the left insula, inferior frontal gyrus, middle frontal gyrus, precuneus and inferior parietal lobule. Duration of use and days of sobriety were correlated with BOLD activation to explore the relationships between neural activation and long-term use or prolonged withdrawal. The left middle frontal gyrus, anterior cingulate and left precuneus were activated less the longer the subject's use of methamphetamine.Duration of sobriety correlated with more activation in the left medial frontal gyrus.
Decision-making abnormalities in methamphetamine-dependent individuals may include alterations in trend detection, i.e. determining which options are advantageous and which are not, error processing, i.e. adapting decision-making as a consequence of positive versus negative outcomes, and risk-taking, i.e. the degree to which options with probable gains and losses are preferred. It appears that stimulant-dependent individuals show a lack of flexible association of outcomes with advantageous actions, which is consistent with previous studies that found altered inferior prefrontal activation at baseline and altered task-dependent activation patterns in stimulant-dependent subjects. Moreover, our own data have shown altered activation patterns in neural substrates critical for error processing but not error-rate-dependent changes in methamphetamine-dependent subjects. Other investigators have shown altered risk-taking patterns in substance-dependent subjects. Taken together, these deficits may be a consequence of altered dopaminergic functioning, which is critical for signaling the mismatch between what is expected and what is observed.
Craving is an important psychological phenomenon that has been linked closely to relapse [108,109]. Intensity of craving predicted methamphetamine use significantly in the week immediately following each craving report. Craving remains a highly significant predictor of relapse in multivariate models even after controlling for pharmacological intervention or methamphetamine use during the prior week. Craving scores preceding use were 2.7 times higher than scores preceding abstinence. Risk of subsequent use was 2.5 times greater for scores in the upper half of the scale relative to scores in the lower half . Investigating the neural activity revealed by neuroimaging during craving in stimulant users could also lead to a better understanding of precipitating factors for dependence and preventative strategies for relapse. In turn, treatments that affect craving measurements or functional imaging patterns may be expected to improve treatment outcome.
When Maas and colleagues  exposed subjects to cocaine cues, cocaine users activated the anterior cingulate and left dorsolateral prefrontal cortices more so than controls. These findings corroborate the PET results of Grant et al.  and Childress et al. , showing significant signal change increases in the dorsolateral prefrontal cortex and anterior cingulate, respectively, in addition to increases in the medial temporal lobes and amygdala. Goldstein & Volkow note that the amygdala and hippocampus are probably key structures that mediate the memory consolidation required for craving, whereas the anterior cingulate cortex and thalamo–orbitofrontal circuitry might play critical roles in the craving experience and maintenance of drug addiction through loss of control .
Behavioral and longitudinal studies in cocaine-dependent subjects have shown that stress augments craving, and craving intensity correlates with relapse risk in individuals not receiving psychosocial treatment . Experiments conducted by Sinha et al. revealed that in response to individually tailored stress imagery scripts, patients activated the ACC, left hippocampus/parahippocampal gyrus, right fusiform gyrus and right postcentral gyrus less. The authors suggest that an inability to practice behavioral control under stress may result from the inactivation of anterior cingulate and paralimbic areas. Stress has been shown to correlate with increased anterior cingulate activity ; diminished anterior cingulate function in cocaine users may then disturb normal stress-processing. A weakened response to stress in turn may contribute to detrimental behavioral responses to craving. Given that suppression of inappropriate actions has been ascribed to the ACC , this region is presumably important for avoiding relapse in abstinent individuals.
Several limitations are inherent in these studies. First, stimulants not only affect neuronal brain functioning but also have direct effects on brain vasculature , which may be complex and non-dopamine-related . Therefore, global hemodynamic changes must be considered when comparing functional activation patterns of stimulant users and control subjects, as they may cause altered differences in the BOLD signal. Accordingly, enhanced activation patterns in response to visual stimuli were found in chronic cocaine users, and were attributed to low resting cortical blood flow and/or low oxidative metabolism . Acute cocaine administration has been shown to decrease cortical blood flow, but apparently BOLD imaging can still be conducted reliably . Secondly, retrospective studies such as these are always confounded by the classical chicken or egg problem—are differences pre-existing or resultant of drug use? Studies to correlate fMRI findings with duration and amount of use can help elucidate this issue, while longitudinal studies in at-risk populations are preferable for clarifying the relationship. Thirdly, stimulant use rarely occurs in isolation and is often accompanied by use of other drugs or by psychiatric comorbidity . Therefore, determining the specificity of the observed brain differences is always difficult.
Understanding the neurobiological underpinnings of stimulant dependence has several important clinical implications. First, a better measurement of neural substrate dysfunction induced by repeated use of stimulants may result in tailored treatments that take into account the processing difficulties in these individuals. Secondly, quantifying neural substrate dysfunctions may enable one to predict who is at high risk for relapse, which in turn may enable one to better appropriate resources to reduce relapse risks. Thirdly, identifying which processes are affected by methamphetamine and other stimulants will provide a better understanding of how addiction to stimulants develops. For example, several investigators have begun to relate brain imaging differences to relapse. One emerging theme is that stimulant-dependent individuals show specific, rather than generic, brain activation differences, i.e. instead of showing more or less brain activation independent of the process being investigated, they exhibit process-related brain activation differences that are consistent with a shift from context-specific, effortful processing to more stereotyped, habitual, response generation.
The authors would like to thank the anonymous reviewers for their time and suggestions.