INTRODUCTION

Understanding psychological and neural processes, both leading to and consequent on alcohol-use disorders (AUDs), is important for both public health and addiction neuroscience. Adolescence is a high-risk period for initiating alcohol use and developing problem drinking. Most drinkers begin alcohol use in their teens and the greatest rates of alcohol abuse and dependence are between 18 and 25 years (SAMHSA, 2011; Dager et al, 2013). College students drink more and have higher rates of AUDs than same-aged non-college peers (Borsari, 2007).

Impulsivity is an important component of drug and alcohol addiction susceptibility (Dalley et al, 2011; Lejuez et al, 2005, 2010; Oberlin and Grahame, 2009; Petry, 2001), including susceptibility based on familial risk (Knop, 1985; Petry et al, 2002; Saunders et al, 2008; Sher, 1991; Cloninger, 1987; Ernst et al, 2006). Impulsivity is a multifaceted construct (Meda et al, 2009) widely implicated in the development and maintenance of addictive behaviors (Verdejo-Garcia et al, 2008). Distinct aspects of impulsive behavior have been shown to influence alcohol-use initiation, escalation, and dependence differently (Courtney et al, 2012). Neural mechanisms of impulsivity point to dorsal anterior cingulate involvement with risk taking and correlation with hazardous drinking behavior (Claus and Hutchison, 2012).

Impulsive tendencies can be assessed using self-report questionnaires such as the Barratt Impulsiveness Scale (BIS-11) (Patton et al, 1995); such measures detect elevated impulsivity in regular users of multiple substances and those with familial histories of alcoholism (Verdejo-Garcia et al, 2008). Two broad impulsivity subtypes are choice impulsivity (involving rapid temporal discounting) and rapid response impulsivity (involving inhibition of prepotent responses); both have important implications for AUDs (Potenza and De Wit, 2010). A genetic susceptibility (predisposition) to addiction associated with increased impulsivity may manifest behaviorally as impaired response inhibition that may result from abnormal inhibitory control (Goldstein and Volkow, 2002; Kreek et al, 2005). Deficient inhibitory control is reported in multiple neuropsychiatric conditions, including alcohol/substance-use disorders (Bauer, 2001; Kaufman et al, 2003; Kouri et al, 1996) and externalizing disorders frequently linked to addiction risk, including antisocial/conduct-related disorders and attention-deficit hyperactivity disorder (Brandeis et al, 2002; Frank et al, 1998; Pliszka et al, 2000; Rubia et al, 1998). It is possible that pre-existing deficiencies in impulsive control might lead to AUD. However, the question of impulsivity–AUD causality remains a circular one, the answer to which is still largely undetermined.

A major response-inhibition task is the Go/No-Go, where a prepotent bias towards fast responding to ‘Go’ stimuli increases the difficulty of withholding responses to ‘No-Go’ stimuli. Groups characterized by clinically relevant impulsivity (eg, AUDs) show diminished inhibition of responses to No-Go stimuli and thus make more errors of commission (ie, false-alarm responses on No-Go trials, when responses should be withheld/suppressed) (Kaufman et al, 2003; Chamberlain and Sahakian, 2007; Fillmore and Rush, 2002). Prior reports also show binge/heavy drinkers to perform significantly slower compared with their lighter-drinking counterparts, in response inhibition and directed attention tasks (Marczinski et al, 2007; Cox et al, 1999). Measures associated with effortful suppression are regarded as correlates of response inhibition (Pandey et al, 2012), and slow relative to fast stoppers may have weaker inhibition processes and abnormal error processing (Chamberlain and Sahakian, 2007; Lansbergen et al, 2007a, 2007b). Functional MRI versions of the Go/No-Go task consistently show activation of multiregion neural networks (Rubia et al, 2001, Stevens et al, 2007, 2009, Posner and DiGirolamo, 1998; Smith and Jonides, 1999; Barkley, 1997; Weisbrod et al, 2000; Kaiser et al, 2003). Various regions within these networks are likely associated with particular aspects of the task, for example, anterior cingulate cortex (ACC) with choice and error/conflict monitoring, and dorsolateral prefrontal cortex (DLPFC) with higher-order cognitive control over behavior. Moreover, broad networks likely form subcircuits, where groups of network nodes that are engaged by overall task performance might more directly correspond to specific cognitive operations. For example, Stevens et al (2007) showed that correct stops (No-Go’s) engaged a subcircuit comprising left lateral prefrontal cortex, left postcentral gyrus/inferior parietal lobule, striatum, motor/premotor areas, and left cerebellum. Error commission engaged an another subcircuit that was not integrated with activity in regions engaged for higher-order cognitive control over behavior (eg, DLPFC). A third medial/dorsolateral prefrontal–parietal circuit responded to all No-Go stimuli, but with greater BOLD activity to errors.

Acute alcohol consumption influences both error processing and associated fMRI responses underlying No-Go false alarms (Anderson et al, 2011). However, abused substances may cause enduring changes in brain function that persist after cessation of use, leading to a ‘hijacking’ of brain reward, motivation, memory, and control circuits (Volkow et al, 1992) and damage to brain regions within neural networks. Specific brain regions (eg, DLPFC and hippocampus) and white matter tracts may be particularly vulnerable to such effects and considerably more so in developing brains of adolescents and young adults (Spear, 2000; Tomlinson et al, 2004). In the context of AUD, recent evidence in an adolescent sample supports the hypothesis that diminished neural activity during response inhibition predicts future involvement with problematic behaviors such as alcohol and substance abuse (Norman et al, 2011). Neural correlates of AUD during response inhibitions tasks show higher measures of AUD associated with lower functioning in regions including insula, inferior frontal gyrus, inferior parietal lobule, and anterior cingulate, and also greater engagement of motor response circuits preinhibition (Claus et al, 2013).

Behavioral measures such as number of blackouts may be indicators of rapid alcohol consumption (ie, gulping rather than drinking more steadily and slowly), which may in part reflect poor impulse control (Goodwin, 1995; Perry et al, 2006). Previous work has also demonstrated increased frontocerebellar response to a Go/No-Go task in substance-naive youth who later experience alcohol-related blackouts, suggesting pre-existing abnormalities in inhibitory processing that underlie blackout propensity (Wetherill et al, 2013). Twin studies indicate that drinking measures such as maximum drinks are highly heritable (Slutske et al, 1999) and are emphasized as an important endophenotype in large-scale alcohol dependence studies such as the Collaborative Study on the Genetics of Alcoholism (COGA) (Saccone et al, 2000).

On the basis of impulsivity-related findings and prior reports of altered fMRI and behavioral responses on Go/No-Go paradigms in impulsive or alcoholic individuals, we hypothesized that collegiate heavy drinkers would show: (a) higher scores on impulsivity-related measures; (b) impaired Go/No-Go response inhibition gauged by RTs and/or error rates (eg, slower RTs on Go correct-hit trials); and (c) diminished fMRI BOLD response patterns involving ACC, frontal cortical, motor cortical, striatal, and hippocampal responses during response inhibition. In line with the blunted response hypothesis, we also predicted that Go/No-Go neural responses would correlate negatively with out-of-magnet alcohol- and impulsivity-related measures.

MATERIALS AND METHODS

Subjects

Participants were 92 young adults aged 18–20 years, recruited during their freshman year as part of an ongoing longitudinal study of alcohol and neurocognitive function (the Brain and Alcohol Research in College Students (BARCS) study; Dager, 2013). All subjects reported any alcohol consumption as well as any related consequences (eg, arrests, memory blackouts, missed classes) monthly, for the preceding month, in considerable detail on a secure website for 24 months. A subset of BARCS subjects participated in a neuroimaging study. All participants provided written informed consent, approved by the institutional review boards at Yale University, Hartford Hospital, Trinity College, and Central Connecticut State University. Exclusion criteria included current/past psychotic or bipolar disorders based on the Mini-International Neuropsychiatric Interview (MINI) (Sheehan et al, 1998), lifetime head injury that resulted in unconsciousness that lasted more than 5 min, positive urine pregnancy test in women, or toxicology screen for abused substances. Eligible participants were evaluated for family-history-of-alcoholism genograms and personal drinking histories. Heavy drinking was defined using a combination of AUD diagnosis and quantity/frequency of current alcohol consumption (Cahalan et al, 1969; Squeglia et al, 2009; Dager et al, 2013). Participants were considered light drinkers if they: (1) did not meet current or past criteria for an AUD; and (2) drank during fewer than half of the weeks during the preceding 6 months. Participants were considered heavy drinkers if they either (1) met the criteria for current AUD or (2) drank more than half of the weeks in the preceding 6 months and reported that they typically binge-drank when drinking (⩾4 drinks per occasion for females or ⩾5 drinks per occasion for males; eg, Dager et al, 2013; Courtney and Polich, 2009; Schweinsburg et al, 2010). The final sample of 56 light and 36 heavy drinkers did not differ on sex, ethnicity, race, family history (FH) of alcoholism and tobacco use (see Table 1 for demographics).

Table 1 Demographic and Alcohol Use Characteristics of Study Participants
Full size table

Data Collection and Measures

Detailed alcohol use history (memory blackouts, maximum drinks, frequency of drinking) was obtained using the alcohol-use module of the SCID (First et al, 2002). Current and past DSM-IV axis I diagnoses, including substance-use disorders, were ascertained using the MINI and information on daily tobacco use obtained using a Health Screening Questionnaire. At the time of scanning, participants were free of alcohol and illicit substances, as verified by breathalyzer and urine toxicology. Smokers could use tobacco up to 30 min before scan sessions.

Self-reported impulsivity was assessed with the BIS-11 (Patton et al, 1995), Behavioral Inhibition System/Behavioral Activation System (BIS/BAS) scale (Carver and White, 1994) and Zuckerman Sensation Seeking Scale (Zuckerman et al, 1979), with psychometric properties described in the referenced articles. The BIS-11 has been found to factor into three subscales assessing motor, non-planning, and attentional impulsivity (Patton et al, 1995). The Balloon Analog Risk Task (BART; Lejuez et al, 2002) was used to assess risk-taking, with higher values on the adjusted average of the total number of pumps on unexploded balloons indicative of greater risk-taking propensity (Bornovalova et al, 2005; Lejuez et al, 2002; Wallsten et al, 2005).

Go/No-Go Task

We used an event-related fMRI task design described previously (Anderson et al, 2011; Kiehl et al, 2000a, 2000b; Stevens et al, 2007). Participants were instructed to respond by pressing a button with their right index finger as accurately and quickly as possible to Go stimuli (‘X’, 85% probability) and to withhold a response to No-Go stimuli (‘K’, 15% probability). Go and No-Go stimuli were presented for 50 ms with an interstimulus interval of 750, 1750, or 2750 ms. The presentation of Go and No-Go stimuli was pseudorandomized with intervals of 10–15 s between No-Go stimuli. Trials were presented in 2 runs of 246 trials each lasting 7 min 21 s, accounting for a total of 492 total trials per subject. A break of approximately 1 min was provided between runs. Given the high frequency of Go stimuli, the estimated response function was saturated and it was not possible to extract a meaningful result to Go correct-hit stimuli. Thus, imaging results are presented only for No-Go correct rejections>baseline (successful inhibitions) and No-Go false alarms>baseline (unsuccessful inhibitions) only.

Before scanning, participants completed 10 practice trials to ensure that instructions were understood. RT and accuracy were equally emphasized in task instructions. Participants were not given precise instructions about what number of errors was typical, nor provided with feedback during the task on the number of errors or speed of responses, information that could be used to adjust behavior. Participants were encouraged to perform at this level during both sessions to ensure that within-subject performance differences were meaningful.

Data Analysis

Go/No-Go behavioral analysis

The main behavioral-dependent variable was response inhibition, indexed by RT and error proportion. RT data (in-scanner behavior) were presented separately for Go correct-hit and No-Go false-alarm trials. Although many Go/No-Go studies examine only Go correct-hit RTs, we also examined No-Go false-alarm RTs, as No-Go false alarms are believed to be primarily due to the failure to inhibit an impulsive response. RTs for Go correct hits and No-Go false alarms were analyzed for two groups (heavy and light drinkers). Error proportion (ie, false alarms) for No-Go trials was calculated as the number of errors on No-Go trials divided by the total number of No-Go trials. We analyzed both between-group differences and correlations against fMRI response patterns for impulsivity-related data including BIS/BAS (Carver and White, 1994), Zuckerman Sensation Seeking Scale (Zuckerman et al, 1979), BIS-11 (Patton et al, 1995), and BART (Lejuez et al, 2002). We also analyzed correlations between self-reported drinking behavior scores including maximum drinks (the largest number of alcoholic drinks consumed in 24 h (both for the past 6 months and lifetime based on Wechsler et al, 1999)), and alcohol-related memory blackouts (derived from a brief Young-Adult Alcohol Consequences Questionnaire (Kahler et al, 2005)).

Image acquisition

All images were acquired with a Siemens (Erlangen, Germany) Allegra 3 T high-performance head-dedicated system with 40 mT/m gradients and a standard quadrature head coil at the Olin Neuropsychiatry Research Center. Functional images were acquired axially using an echo planar image (EPI) gradient-echo pulse sequence covering the whole brain (TR/TE=1500/28 ms, flip angle=65°, FOV=24 cm × 24 cm, 64 × 64 matrix, 3.4 mm × 3.4 mm in-plane resolution, 5 mm effective slice thickness, 30 slices).

Functional image analysis

Functional images from Go/No-Go were preprocessed using SPM5 (Department of Imaging Neuroscience, UK; http://www.fil.ion.ucl.ac.uk/spm/software/spm5/). Differences in EPI slice acquisition timing were corrected using the central slice as a reference. Motion was corrected using INRIAlign (Freire et al 2002), and images were then spatially normalized into Montreal Neurological Institute (MNI) space. Normalized EPIs were then smoothed with a 9-mm FWHM Gaussian kernel. Realignment parameters were examined for excessive motion, and participants with movement >4.5 mm or >3° were not included in analyses. Before first-level analysis, events for each participant were categorized as correct hits to Go stimuli, correct rejects to No-Go stimuli, and false alarms to No-Go stimuli. For first-level analyses, a canonical hemodynamic response function and its temporal derivative (Josephs et al, 1997) were fitted to the onset of these three stimulus types for each session separately. Realignment parameters were included in the model as covariates of no interest.

Whole-brain group analysis

Group random-effects analyses were conducted to examine differences in activity between groups in each contrast, by entering contrast images of correct rejects>baseline (successful inhibition) and false alarms>baseline (unsuccessful inhibition) for each group into a post hoc t-test model. This explored between-group activity differences across the whole brain, using the unmodeled data as an implicit baseline. Post hoc t-test results were thresholded using a combination of voxel-wise p=0.025, false discovery rate-corrected (FDR-corrected) and a cluster extent of k=25 voxels.

For better interpretation of functional differences in each contrast, we also conducted a within-group results for each condition using a one-sample t-test, thresholded using the same parameters as above.

Region of interest analysis

We explored if impulsivity measures (BAS, BIS, BIS-11, Zuckerman, and BART) and alcohol-related measures (maximum alcohol drinks, number of blackouts) correlated with primary clusters that activated as main effects of group differences in each contrast using SPSS statistics IBM version 19. We examined relationships between functional changes and the above-detailed measures by correlating the mean weighted β-estimates extracted from primary clusters that showed group difference in each contrast using MARSBAR (http://marsbar.sourceforge.net/).

RESULTS

Demographics

Demographic and alcohol-use characteristics of study participants were compared among groups using χ2 and t-tests (see Table 1). There were no between-group differences on sex, race, age, tobacco use, or alcoholism FH. Approximately 15% of participants (10% of light and 22% of heavy drinkers) met the criteria for a past depressive or anxiety disorder. Compared with light drinkers, heavy drinkers reported greater levels of lifetime alcohol consumption, more memory blackouts, greater number of drinks per week, greater amounts of drinking during the past 6 months, and higher BAS, BIS-11, and sensation-seeking scores. BIS and BART scores did not differ between groups.

Behavioral Results

Behavioral results are summarized in Table 2.

Table 2 Go/No-Go Task Responses
Full size table

Go correct hits

The groups differed on Go correct-hit RTs (t(89)=−2.95, p=0.004; Figure 1a). The RT slowing was marginally larger in heavy drinkers than in light drinkers. RTs for Go-correct hits also correlated positively with maximum alcohol-consumption scores (the largest number of alcoholic drinks that subjects consumed in 24 h (both lifetime and past 6-month measures) (rlifetime(89)=0.242, p<0.02; r6 months(89)=0.245, p<0.02)).

Figure 1
figure 1

Mean (standard error) reaction time (RT) (in s) for No-Go false alarms (a) and Go correct hits (b) in the heavy and light drinker groups. The heavy drinkers exhibited significantly greater RTs for both conditions.

PowerPoint slide

Full size image

No-Go false alarms

Although groups did not differ in error proportions, they differed on No-Go false-alarm RTs (t(89)=−2.94, p=0.004; Figure 1b) with increased RT in heavy compared with light drinkers. RTs for No-Go false-alarm trials correlated with the number of reported memory blackouts (r(89)=0.267, p<0.01) and maximum alcohol consumption (lifetime and past-6-month measures) (rlifetime(89)=0.288, p<0.006; r6 months(89)=0.323, p<0.003).

Functional Results

No-Go correct rejects

During No-Go correct rejects, heavy drinkers relative to light drinkers showed decreased activity in the left SMA and bilateral ACC, bilateral parietal lobules, thalamus, putamen, right parahippocampal gyrus/hippocampus, bilateral middle frontal gyrus, and left superior temporal gyrus (Figure 2a, Table 3 (primary clusters), and Supplementary Table 1 (subclusters)). Some group main-effect region of interest contrast values correlated significantly with alcohol- and impulsivity-related measures (Tables 4a and Tables 4b). Maximum alcohol-consumption scores correlated negatively with regional BOLD signal changes in anterior cingulate gyrus, left postcentral, left thalamus, right middle frontal, and right putamen during No-Go correct rejects (vs implicit baseline) (Figure 2 and Table 4a). BOLD signal change in the ACC and left superior temporal gyrus correlated negatively with the number of blackouts (Figure 2 and Table 4a). Negative correlations were observed between activity in multiple main-effect ROIs (primary clusters) and impulsivity scores as noted in Table 4b and Figure 2. All correlations are reported at an uncorrected p-value (see table footnote for details) and thereby should be interpreted with caution.

Figure 2
figure 2

(Top) Significant functional magnetic resonance imaging (fMRI) between-group differences during No-Go correct rejections (p<0.025, false discovery rate (FDR); k=25). Light drinkers exhibited greater blood oxygen level-dependent (BOLD) responses than did heavy drinkers, notably in anterior cingulate gyrus (Brodmann area 24), left supplementary motor area, right parietal lobule, right hippocampus, and left superior temporal gyrus. (Below) Significant negative correlations were seen between activation in the left anterior cingulate cortex (ACC) during No-Go correct rejects and self-reported alcohol-consumption-related measures (number of blackouts, maximum number of drinks) and behavioral-activation-system scores. Individuals with heavier alcohol consumption and higher Behavioral Activation scores showed less BOLD response in right ACC than individuals with lighter consumption and lower behavioral activation.

PowerPoint slide

Full size image
Table 3 Regions Showing Significant Effects for Heavy Drinking on BOLD Response to Correct Rejection vs Baseline (Clusters >25 voxels, pFDR<0.025)
Full size table
Table 4a Correlations Between-Group Main Effect Primary ROI clusters and Alcohol Consumption Measurements
Full size table
Table 4b Correlations Between-Group Main Effect Primary Region-of-interest (ROI) Clusters and Impulsivity Scores
Full size table

No-Go false alarms

No region survived after correction for multiple comparisons in this between-group contrast.

Within-group results showed bilateral activation patterns in both No-Go conditions and revealed largely similar patterns as observed in previously published papers from our group using the same paradigm (Kiehl et al, 2000a). fMRI patterns had substantial overlap in several regions between the two groups. Results are provided in Supplementary Figure 1.

DISCUSSION

This study evaluated neural responses to a Go/No-Go task and their relationships with alcohol- and impulsivity-related measures in heavy- vs light-alcohol-drinking college students. The major finding was that heavy drinkers had less fMRI BOLD response mainly in left SMA, ACC (Brodmann area (BA) 24), parietal lobule, thalamus, putamen, right parahippocampal gyrus, right hippocampus, right middle frontal gyrus, and left superior temporal gyrus during response inhibition. In addition, heavy drinkers had slower RTs during correct and incorrect hits compared to light drinkers, along with higher numbers of memory blackouts and higher impulsivity-related scores. Between-group differences in brain activations in a subset of the above-reported regions correlated with alcohol- and impulsivity-related measures.

Go/No-Go Behavioral Performance and fMRI Activity

Heavy and light drinkers differed in Go correct-hit and No-Go false-alarm RTs, with heavy drinkers manifesting slower RTs in Go and No-Go conditions. This slower reaction time may represent a conflict during prepotent response inhibition resulting in response delays, consistent with previous reports (Zack et al, 2011). Nonsignificant differences in error proportion between groups further improves our interpretation of fMRI response differences seen in this study, as the data suggest that the two groups are matched on task performance at a behavioral level.

A critical interpretive issue is that heavy drinkers were slower to respond overall, in doing so they might have ironically also benefitted from committing fewer errors. To address this issue empirically, we correlated RT during error commission to proportion of errors and saw a significant negative correlation in both groups, suggesting that when subjects took more time to respond they committed fewer errors. However, importantly, when we compared the correlation coefficients between groups using a Fisher’s r–Z transformation method, we found no significant between-group differences, indicating that this pattern of general slowness was observed in both groups equally, thereby not causing an interpretation bias in the reported results.

Aspects of our fMRI results are consistent with our initial hypothesis of differential fMRI activity in task-relevant attention and response inhibition areas. One such example is reduced fMRI activity in left SMA and ACC, parietal lobule, right parahippocampal gyrus, right hippocampus, right middle frontal gyrus, and left superior temporal gyrus during successful response inhibition (No-Go correct rejection vs baseline) in heavy drinkers, but no significant group differences in fMRI activity for unsuccessful response inhibition (false alarms) after multiple comparisons. Our results are also compatible with previous studies of alcoholics vs healthy controls that showed differential activity associated with No-Go correct rejects in the ACC (Pandey et al, 2012).

The ACC and parietal cortex are part of a distributed network that underlies cognitive control, conflict monitoring, effortful processing, and impulse control (Botvinick et al, 1999, 2001; Luu and Pederson, 2004; Ridderinkhof et al, 2004; Aron et al, 2004; Jentsch and Taylor, 1999; Lyvers, 2000; Everitt and Robbins, 2005; Bekker et al, 2005). Frontal lobe dysfunction may result from alcohol-related brain damage (Oscar-Berman and Ksenija, 2007; Dirksen et al, 2006). Anderson et al (2005) showed that greater BOLD response to inhibition during a Go/No-Go task predicted more expectancies of cognitive and motor impairments from alcohol in adolescents. These results suggest that decreased inhibitory control may contribute to more positive and less negative alcohol expectancies, which could eventually lead to problem drinking (Anderson et al, 2005). Our results regarding deficits in SMA, ACC, frontal lobe, and parietal lobule are broadly compatible with the above reports.

Impulsivity-Related Measures and fMRI Activity

Impulsivity-related constructs contribute importantly to addictions (Andrews et al, 2011; Dalley et al, 2011; Lejuez et al, 2005, 2010; Oberlin and Grahame, 2009; Petry, 2001). Individuals with familial alcoholism show increased impulsivity relative to those without (Cloninger, 1987; Ernst et al, 2006; Knop, 1985; Petry et al, 2002; Saunders et al, 2008; Sher, 1991). Thus, we hypothesized that heavy drinkers would score higher on the impulsivity-related measures used in this study. We found significant group differences on self-reported measures including the Sensation Seeking Scale (Zuckerman et al, 1979), behavioral activation system score of the BIS/BAS scale (Carver and White, 1994), and BIS-11 (Patton et al, 1995), consistent with previous results showing higher behavioral activation, impulsivity and sensation-seeking in relation to hazardous drinking, alcohol abuse/dependence, and FH of alcoholism (Verdejo-Garcia et al, 2008; Knop, 1985; Petry et al, 2002; Saunders et al, 2001; Sher, 1991; Cloninger, 1987; Hamilton et al, 2012; Ernst et al, 2006). Interestingly, between-group differences were not observed on the behavioral task (the BART) assessing risk-taking propensities. BART scores have previously been found to factor independently from self-reported impulsivity-related measures used in this study (Meda et al, 2009), suggesting that behavioral risk-taking may be a dissociable construct from perceived assessments of impulsivity-related tendencies. As self-reported and not behavioral measures of impulsivity were found recently to mediate the relationship between stress and hazardous drinking in a community sample of adults (Hamilton et al, 2013), future studies should consider how stress exposure might interact with individual differences in impulsivity-related tendencies in the propensities of college students to consume alcohol. Given data linking both self-reported sensation-seeking and behavioral risk-taking in early adolescence (MacPherson et al, 2010) to prospective alcohol-use behaviors, the extent to which these relationships might change over time and affect college students warrants additional investigation.

Impulsivity-related measures also correlated with several regions showing differential brain activity between groups. In the No-Go correct rejection vs baseline contrast, BOLD activity in the left ACC, left thalamus, right lingual gyrus, right middle frontal, right putamen, and left postcentral gyrus correlated inversely with BAS scores. Similarly, BIS-11 and sensation-seeking scores correlated inversely with BOLD signal in the right putamen. Despite correlations with sensation seeking and other impulsive measures, we did not find either between-group differences or correlations with the BART. This finding is consistent with the notion that the BART might index a separate impulsivity-related construct as discussed previously (Meda et al, 2009). Despite being characterized by premorbid impulsiveness, sensation seeking, and higher rates of anxiety, it could be possible that certain internalizing features in the heavy-drinking group biases their actions in a more risk-aversive manner regarding asset forfeiture and cognitive complexity, reflected by a lack of difference on tasks such as BART (Lejuez et al, 2002). Using fMRI, Claus et al (2011) investigated the neural basis of impulsive choice in AUDs, suggesting that these may result from functional anomalies in widely distributed but interconnected brain regions involved in cognitive and emotional control. Our results are compatible with previous observations pointing to dysfunction of the orbitofrontal cortex (Berlin et al, 2004; Winstanley et al, 2004), superior frontal gyrus (Horna et al, 2003), ACC, superior temporal gyrus (Garavan et al, 2002), and hippocampus (Cheung and Cardinal, 2005) as possible substrates of elevated impulsivity and behavioral activation. Acute alcohol use itself influences error processing and associated fMRI response during No-Go false alarms (Anderson et al, 2011). Animal studies of impulsive choice implicate regions including the hippocampus (Cheung and Cardinal, 2005), complementary to our findings on relationships between self-reported impulsivity and hippocampal activation during Go/No-Go performance. Taken together, our findings suggest that the neural correlates of successful response inhibition on the Go/No-Go task involve a distributed network of cortical and subcortical regions and individual differences in the degree of regional brain activation relate to out-of-magnet measures of self-reported behavioral activation, impulsivity, and sensation-seeking. Owing to the ongoing nature of this study, we do not yet have current follow-up drinking/imaging information on our subjects, which might limit the interpretability of the results in the context of brain-behavior prediction of substance abuse. However, our study can be compared with a recently published paper of Norman et al (2011), who performed a longitudinal study using a Go-NoGo fMRI task similar to ours in an adolescent population (ages 12–14 years). Results show that at baseline (ie, before using alcohol), youth who had transitioned to heavy alcohol use had largely reduced activation during response inhibition in a core set of 12 regions that largely overlapped with the reduced regional activity reported in heavy drinkers in this study. Given that a subset of these regions also correlated with higher impulsivity scores in our subjects, it is possible that blunted activation and/or increased impulsitivity related to these regions might indicate delayed/abnormal maturation of inhibitory networks in future substance abusers. Given links between these constructs and alcohol-use patterns, the results suggest specific neural regions/circuits that might represent targets for therapeutic interventions for heavy drinking among college students.

Limitations and future directions

Our study involved several limitations. It would be useful to compare No-Go correct rejects to a Go correct-hits baseline, thereby increasing the specificity of the results to inhibition over execution. However, our design was not optimized to study activity associated with Go correct hits. Secondly, to measure BOLD correlates of response inhibition, we compared No-Go correct rejects to an implicit baseline. Although this offers some advantages over subtractive contrasts in interpretation (ie, changes in activity in subtractive contrasts can be attributed to an increase/decrease in BOLD in one condition compared with another, or a change in both), disadvantages entail attributing results specifically to an inhibitory process. To address this, we conducted a supplementary analysis of No-Go correct rejects>No-Go false alarms (successful vs unsuccessful inhibition). These results support our conclusion that our No-Go correct rejects>baseline contrast indexed inhibition. However, these results are not reported here as findings were weak (p<0.01 uncorrected value) and did not survive corrections for multiple comparisons. This question should be explored in future work using samples providing sufficient statistical power to address this issue conclusively. The current data also do not speak to the issue of cause vs consequence; it is unclear whether different Go/No-Go behavior and brain BOLD patterns represent pre-existing vulnerabilities in response inhibition that are risk factors for later AUDs, consequences of heavy drinking, or both. To improve sample representativeness, we adopted a slightly unconventional (but previously published) method of classifying subjects into light (a mix of both AUD and non-AUD subjects) and heavy drinkers, which might have added some potential noise to the data. It is also unclear if individual differences in motivation might have confounded subjects’ signal detection/perception capability. In addition, the significant differences in occurrence of depressive/anxiety disorder rates among groups might be an additional confounding factor. Also, as fMRI–impulsivity correlations did not survive corrections for multiple comparisons, these findings need to be interpreted with caution.

As alcoholism can be associated with comorbid psychiatric symptoms, larger samples should investigate directly the impact of psychiatric comorbidity as related to the heavy-drinking-related neural correlates of response inhibition. Personality and alcohol expectancies have also been examined as potential risk factors for the initiation and maintenance of alcohol use in adolescents and young adults, and these should be examined further with respect to the heavy drinking related neural correlates of response inhibition. Nonetheless, the findings that heavy drinkers demonstrate evidence of decreased processing associated with regions subserving attention, motivation, and response inhibition during No-Go response withholding and slower RTs in response to an fMRI Go/No-Go task, suggest a neural mechanism that may underlie heavy drinking among college students. Future studies should consider investigating prevention and treatment strategies targeting impulsivity-related constructs, particularly as changes in impulsivity have correlated with changes in other addictive behaviors during treatment (Blanco et al, 2009).

In summary, young heavy drinkers demonstrated altered task performance, greater impulsivity-related ratings, and reduced response-inhibition-task-associated brain activity, most prominently in ACC, portions of frontal lobe, hippocampus, thalamus, and superior temporal regions, brain areas associated with impulsivity, alcoholism, and/or alcohol-related toxicity. Specific alcohol- and impulsivity-related measures were associated with between-group differences in brain activation. Differences in impulsivity-related tendencies, behavior, and brain activation patterns could not be attributed to group differences in alcoholism FH in our sample, as this measure did not differ between groups. If the observed brain activity differences in regions involved with cognitive control, attention, and response inhibition in heavy drinkers results in a reduced capacity to inhibit responses to No-Go stimuli, such individuals might require more effortful inhibitory control to overcome impulsiveness tendencies and yield equal stopping performance, and this might translate into heavier drinking in real-life situations.

FUNDING AND DISCLOSURE

Dr Pearlson has been a consultant to Bristol Myers Squibb for projects unrelated to the current research. Dr Potenza has served as a consultant or advisor to Boehringer Ingelheim, Somaxon, gambling businesses and organizations, law offices, and the federal defender's office in issues regarding impulse control disorders. He has received research support from the National Institutes of Health, Veteran's Administration, Mohegan Sun Casino, the National Center for Responsible Gaming, Psyadon, Forest Laboratories, Ortho-McNeil, Oy-Control/Biotie, and GlaxoSmithKline. He has participated in surveys, mailings, or telephone consultations related to drug addiction, impulse control disorders, or other topics. He has provided clinical care in the Connecticut Department of Mental Health and Addiction Services Problem Gambling Services Program. He has performed grant reviews for the National Institutes of Health and other agencies. He has guest-edited journal sections, has given academic lectures in grand rounds, continuing medical education events, and other clinical and scientific venues, and has generated book or book chapters for publishers of mental health texts. The other authors declare no conflicts of interest.