research paper-College student substance abuse

F

C c

P S

a

A R R A A

K A A C A I I

1

n a a 2 h t b i

t u s S i s

0 d

Drug and Alcohol Dependence 115 (2011) 94–100

Contents lists available at ScienceDirect

Drug and Alcohol Dependence

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / d r u g a l c d e p

ull length article

ognitive predictors of problem drinking and AUDIT scores among ollege students

hilip Murphy, Hugh Garavan ∗

chool of Psychology and Trinity College Institute of Neuroscience, Trinity College Dublin, College Green, Dublin 2, Ireland

r t i c l e i n f o

rticle history: eceived 19 April 2010 eceived in revised form 16 October 2010 ccepted 29 October 2010 vailable online 9 December 2010

eywords: lcohol UDIT ognitive processes ttentional bias

mpulsivity nhibitory control

a b s t r a c t

Evidence from a number of substance abuse populations suggests that substance abuse is associated with a cluster of differences in cognitive processes. However, investigations of this kind in non-clinical samples are relatively few. The present study examined the ability of alcohol-attentional bias (an alcohol Stroop task), impulsive decision-making (a delay discounting task), and impaired inhibitory control (a GO–NOGO task) to: (a) discriminate problem from non-problem drinkers among a sample of college students; (b) predict scores on the Alcohol Use Disorders Identification Test (AUDIT; a measure of alcohol consump- tion, drinking behaviour, and alcohol-related problems) across all of the student drinkers; (c) predict AUDIT scores within the subgroups of problem and non-problem student drinkers. In logistic regression controlling for gender and age, student drinkers with elevated alcohol-attentional bias and impulsive decision-making were over twice as likely to be a problem than a non-problem drinker. Multiple regres- sion analysis of the entire sample revealed that all three cognitive measures were significant predictors of AUDIT scores after gender and age had been controlled; the cognitive variables together accounted for

48% of the variance. Moreover, subsequent multiple regressions revealed that impaired inhibitory control was the only significant predictor of AUDIT scores for the group of non-problem drinkers, and alcohol- attentional bias and impulsive decision-making were the only significant predictors of AUDIT scores for the group of problem drinkers. Finally, both impulsive decision-making and impaired inhibitory control were significantly correlated with alcohol-attentional bias across the whole sample. Implications are discussed relating to the development of problematic drinking.

. Introduction

Human substance abuse is a complex and multifaceted phe- omenon. The abuse of substances has long been linked to hedonic nd reward-related processes. Indeed, all major substances of buse appear to activate brain reward systems directly (Kenny, 006). In more recent times, however, accumulating evidence as shown that substance abuse is also associated with a clus- er of differences in cognitive processes, including attentional ias for substance-related stimuli, impulsive decision-making, and

mpaired inhibitory control over drives and behaviour. Attentional bias for substance-related stimuli can be seen in how

hese stimuli tend to grab the attention of experienced substance sers. Among paradigms developed to measure attentional bias for ubstance-related stimuli, the most widely used is the addiction

troop task (Cox et al., 2006). This task requires participants to dentify and respond to the colour of a stimulus while ignoring its emantic content. When the semantic content of a stimulus relates

∗ Corresponding author. Tel.: +353 1 896 8448; fax: +353 1 896 3183. E-mail address: [email protected] (H. Garavan).

376-8716/$ – see front matter © 2010 Elsevier Ireland Ltd. All rights reserved. oi:10.1016/j.drugalcdep.2010.10.011

© 2010 Elsevier Ireland Ltd. All rights reserved.

to an individual’s addictive substance, colour naming is found to be slower than when the semantic content of a stimulus is unrelated to substance use.

Previous research using addiction Stroop tasks have shown that abusers of opiates (Franken et al., 2000; Lubman et al., 2000), cocaine (Hester et al., 2006), nicotine (Waters and Feyerabend, 2000), and alcohol (Cox et al., 2000; Stormark et al., 2000) are slower to name the colour of substance-related stimuli, relative to control stimuli and to non-abusers. A meta-analysis of alco- hol abuse and smoking found that the addiction Stroop task has consistently distinguished abusers from non-abusers, produces measurable physiological reactions, and can predict short-term cessation outcomes (Cox et al., 2006).

Although the concept of impulsivity has been defined vari- ously, several authors (de Wit and Richards, 2004; Olmstead, 2006; Reynolds et al., 2006) have suggested that it may involve two relatively independent processes: impulsive decision-making (or cognitive impulsivity), such that individuals tend to choose imme-

diate rewards despite negative consequences of those choices in the future, and impaired inhibitory control (or motor impulsivity), such that individuals find it difficult to withhold or countermand behavioural responses.

Alcoho

r i p i t p p a i k r s

o c u ( a w a d r a a

t r N c N t i 2

w 2 V ( G

d r n a h v a l e d e 1 t a p t b f s f O

h e a

P. Murphy, H. Garavan / Drug and

One approach to understanding impulsive decisions that has eceived considerable research interest in the substance abuse field s the analysis of delay discounting. Delay discounting tasks require articipants to choose between a larger, later reward and a smaller,

mmediate reward, with the magnitude of the latter adjusted until he subjective value of the two rewards is approximately equal. This oint of equivalence is interpreted as the indifference point for that articular delay interval. When indifference points are obtained for variety of delays, one can calculate the free parameter k, which

ndicates the rate at which delayed rewards are discounted. Higher values indicate greater and more rapid discounting of delayed

ewards, which can thus be interpreted as an exacerbation of impul- ive decision-making (Evenden, 1999).

Greater discounting of delayed hypothetical rewards has been bserved in opiate-dependent patients compared with non-abuser ontrols (Madden et al., 1997, 1999), in opiate-dependent individ- als who report sharing needles compared with those who do not Odum et al., 2000), in current smokers compared with both never nd ex-smokers of cigarettes (Bickel et al., 1999), and in individuals ith early-onset alcoholism compared with those with late-onset

lcoholism (Dom et al., 2006). Taken together, these studies have emonstrated that discounting of delayed hypothetical rewards eliably discriminates between substance abusers and non-abusers s well as between individuals with differing severities of substance buse problems.

Inhibitory control deficits have been measured with GO–NOGO asks. These tasks require participants to make a rapid manual esponse to a GO stimulus, but to withhold a response when a OGO stimulus is presented. The NOGO/GO ratio is low, thereby reating a response prepotency that makes it difficult to inhibit on OGO stimuli. A high number of errors of commission (responding

o an incorrect NOGO stimulus) are thought to indicate impaired nhibitory control (Colder and O’Connor, 2002; Reynolds et al., 006).

In addiction research, it has been shown that, compared ith non-abuser controls, abusers of opiates (Forman et al.,

004), cocaine (Hester and Garavan, 2004; Kaufman et al., 2003; erdejo-García et al., 2007), nicotine (Mitchell, 2004), and alcohol

Kamarajan et al., 2005) have increased commission error rates on O–NOGO tasks.

The majority of the research described above has been con- ucted on inpatients and outpatients from substance abuse ehabilitation centres. However, evidence has also emerged of cog- itive processing impairments in non-clinical groups such as heavy nd problem drinkers from high school and college. Researchers ave, for example, used an alcohol Stroop task to demonstrate ele- ated attentional bias for alcohol-related stimuli (hereafter called lcohol-attentional bias) in heavy compared to light drinking col- ege students (Sharma et al., 2001). Additionally, investigators have mployed a delay discounting task to show that heavy and problem rinking college students, but not light drinking students, display xacerbated impulsive decision-making (Vuchinich and Simpson, 998). Consistent with these results, Field et al. (2007) showed hat alcohol-attentional bias and impulsive decision-making (as ssessed by alcohol Stroop and delay discounting tasks) were ositively correlated with scores on the Alcohol Use Disorders Iden- ification Test (AUDIT; a measure of alcohol consumption, drinking ehaviour, and alcohol-related problems), with participants drawn rom high school (Field et al., 2007). In another study, among college tudents, heavy alcohol consumption was associated with more ailures of inhibitory control during a GO–NOGO task (Colder and ’Connor, 2002).

Despite evidence from non-clinical samples suggesting that eavy or problem drinking is associated with individual differ- nces in cognitive processes, studies examining these relationships re relatively few. Thus, the first aim of the present study

l Dependence 115 (2011) 94–100 95

was to use a logistic regression analysis to test the hypothesis that alcohol-attentional bias (an alcohol Stroop task), impul- sive decision-making (a delay discounting task), and impaired inhibitory control (a GO–NOGO task) would discriminate prob- lem from non-problem drinkers (as defined by a cut-off score on the AUDIT) among a sample of college students, after the effects of gender and age had been controlled. It was important to con- trol for gender and age because there is prior evidence that male college students drink more than female students and younger stu- dents drink more than older ones (Gill, 2002). If this hypothesis was supported, then the resulting information would describe a cogni- tive profile linking these various cognitive measures to problematic drinking.

Given that statistical power can be optimised when the outcome measure is continuous, the second aim of the present study was to perform a multiple regression analysis to test the hypothesis that alcohol-attentional bias (an alcohol Stroop task), impul- sive decision-making (a delay discounting task), and impaired inhibitory control (a GO–NOGO task) would predict AUDIT scores across all of the student drinkers, after controlling for the influence of gender and age.

Finally, because relationships between cognition and alco- hol use may vary depending on the severity of that use, the third aim of the present study was to perform separate mul- tiple regression analyses to explore the differential predictive ability of alcohol-attentional bias (an alcohol Stroop task), impul- sive decision-making (a delay discounting task), and impaired inhibitory control (a GO–NOGO task) on AUDIT scores within the subgroups of problem and non-problem student drinkers, after controlling for the effects of gender and age.

2. Methods

2.1. Participants

Eighty-nine students (47 males, 42 females) were recruited from Trinity College Dublin (TCD), via poster advertising in the local campus area. To be eligible, partic- ipants were required to report drinking alcohol at least once a week and to be aged between 18 and 30 years. In addition, participants had to have normal (or corrected to normal) colour vision and to be fluent English speakers. Participants received either course credit or a monetary remuneration for their participation. Two males and three females reported more than occasional use (i.e., one day in the past 30 days) of other illegal drugs (e.g., cannabis, ecstasy) so they were excluded from the analyses, leaving a final total of 84 (45 males; mean age of 20.8 ± 3.0 years; range of 18–30 years). The study received ethical approval from the Ethics Committee in the School of Psychology at TCD.

2.2. Measures

2.2.1. Alcohol Stroop task. A computerised version of the alcohol-related addiction Stroop task, as described by Cox et al. (1999), was adapted for the current study. Four categories of stimuli were used: alcohol-related words (e.g., beer, vodka), music- related words (e.g., violin, banjo), neutral words (e.g., chain, boots), and coloured ‘XXXX’s (see Table 1 for full details). The latter was administered to provide a base- line measure of response time. The different categories provided comparators for the alcohol-related words, controlling for the distractedness of alcohol-related words due to semantic relatedness (music-related words), or words that were neither related with each other or with alcohol (neutral words).

The four categories of stimuli were presented in blocks of 20 trials, with the order of blocks and the order of stimuli within each block randomised across participants. Each category was presented once resulting in a total of 80 trials. A single trial presented the stimulus (in Arial font – size 40 point) on a black background where it remained until the participant responded, following which a 250 ms blank screen and a 500 ms inter-stimulus fixation cross was presented prior to the next stimulus. Each stimulus was randomly presented in one of four font colours (green, yellow, blue, red), and was centred vertically and horizontally on screen.

Prior to the main task, participants were provided with 60 practice trials com- prising word numbers (e.g., three, seventy), each of which was presented in one of

the four font colours. Participants were instructed to identify the font colour of each word and push a corresponding key on the keyboard as quickly as possible. Like- wise, before the experimental trials began, participants were instructed to press, as quickly and accurately as possible, the colour key that corresponded to the colour of a word or an ‘XXXX’ on the screen, and to ignore the meaning of each stimulus.

96 P. Murphy, H. Garavan / Drug and Alcoho

Table 1 Word stimuli for the alcohol Stroop task.

Alcohol words Music words Neutral words

Beer Guitar Box Alcopops Violin Cue Vodka Cello Telephone Shorts Bassoon Boots Bar Oboe Key Gin Bagpipes Read Alcohol Banjo Shoe Off-licence Clarinet Window Pub Piano Windshield Whiskey Viola Cape Stout Trombone Card Brandy Bass Chain Lager Bongos Lamp Sherry Drums Mirror Spirits Maracas Mouse Pint Trumpet Building Booze Flute Carpet Bitter Keyboard Invitation Drink Recorder Watch

2 t a e a b r t n s q

2 n p n i X a s b E

2 1 t t d 2

‘XXXX’s (p < .001). The difference between the mean RT for music-

T D v

N c

Wine Panpipes Floor

.2.2. Delay discounting task. A paper and pencil version of the delay discounting ask, using hypothetical monetary rewards, as described by Kirby et al. (1999), was dapted for the current study. Participants were instructed to indicate their pref- rences for a number of hypothetical monetary rewards, and make their decisions s if the rewards were real. The questionnaire consisted of a fixed set of 27 choices etween smaller immediate rewards and larger delayed rewards. The immediate ewards ranged from D 11 to D 80, whereas the delayed rewards ranged from D 25 oD 85. The range of delays was 7 to 186 days. The delayed rewards fell into groups of ine small (D 25–D 35), nine medium (D 50–D 60), and nine large (D 75–D 85) reward izes. The nine delayed rewards of each size were spread evenly throughout the uestionnaire.

.2.3. GO–NOGO task. Participants were presented with a serial stream of alter- ating X’s and Y’s (GO stimuli), all of which required participants to respond by ressing a key on the keyboard. NOGO stimuli, in which the target X’s and Y’s did ot alternate, required participants to withhold from pressing this key (e.g., partic-

pants would respond to all stimuli, except the fifth, in the following stream: X, Y, , Y, Y, X, Y). Participants were instructed to recommence responding to GO stimuli fter the NOGO stimulus. There were 225 GO and 25 randomly distributed NOGO timuli presented over one run. A single stimulus was presented in white on a black ackground for 600 ms followed by a 400 ms blank screen prior to the next stimulus. ach stimulus was centred vertically and horizontally on screen.

.2.4. Alcohol Use Disorders Identification Test (AUDIT). The AUDIT (Saunders et al., 993) comprises 10 items that measure frequency and quantity of alcohol consump-

ion, drinking behaviour, and alcohol-related problems. A cut-off score of 11 on he test has previously been employed to discriminate problem from non-problem rinkers (Claussen and Aasland, 1993; Holmila, 1995; Fleming et al., 1991; Tsai et al., 005).

able 2 emographic data, AUDIT scores, and cognitive performance measures for the entire sam alues are mean ± SD.

Variables Entire sample of participants (N = 84)

Gender ratio (M:F) 45:39 Research reward ratio (MR:CC) 55:29 Age 20.8 ± 3.0 AUDIT 13.8 ± 7.6 Alcohol-related words 748.3 ± 148.5 Music-related words 679.7 ± 102.3 Neutral words 679.6 ± 100.3 Coloured ‘XXXX’s 649.8 ± 95.4 Alcohol interference 68.6 ± 125.2 Discounting rate .020 ± .017 Commission error 45.1 ± 23.3 Omission error 3.5 ± 3.9 ote. Only 82 participants were included in the analyses for the GO–NOGO task (41 problem redit.

l Dependence 115 (2011) 94–100

2.3. Apparatus

The two computer-based tasks were performed on a PC computer with a 17- in. screen, and were operated using E-prime (version 1.2) experiment generated software.

2.4. Procedure

All participants were tested individually in a quiet, well-lit room. Upon entering the testing room the experimenter explained to the participants that they were to participate in a study on alcohol drinking involving the completion of ques- tionnaires and computer tasks, following which participants provided informed consent. Participants then provided information regarding their gender, age and illegal drug use, before completing the AUDIT. Participants then performed alco- hol Stroop and GO–NOGO tasks, both of which were counterbalanced across all participants. In between these two computer-based tasks, participants completed the paper and pencil version of the delay discounting task. Upon completion of the experiment, participants were debriefed fully by the experimenter and were rewarded with either course credit or a monetary remuneration depending on their preference.

3. Results

3.1. Participants’ characteristics

Table 2 summarizes demographic data and AUDIT scores for the entire sample of participants. Participants were categorised as problem (N = 42) and non-problem (N = 42) drinkers by using the cut-off score of 11 on the AUDIT. Groups did not differ signifi- cantly in age (t(82) = 1.36, NS), gender ratio (X

2 = 0.00, NS), or in their preference for course credits or monetary remuneration for their participation (X2 = 0.84, NS).

3.2. Alcohol Stroop task

Mean accuracy performance was over 97% for all stimuli cat- egories. All trials with errors were removed from the response time (RT) analyses. Table 2 provides mean RT to each category of stimulus and alcohol interference scores; the latter were cal- culated as participants’ mean RT to the alcohol-related words minus their mean RT to music-related words. A one-way repeated measures ANOVA, with stimulus category as the within subject’s factor, yielded a significant main effect, F (3, 81) = 18.22, p < .001. Post-hoc pairwise comparisons indicated that the mean RT for alcohol-related words was significantly slower than for music- related words (p < .001), neutral words (p < .001), and coloured

related words and neutral words was not significant, but the mean RT for each of these stimulus categories was significantly slower than that of coloured ‘XXXX’s (p < .001).

ple of participants as well as the subgroups of problem and non-problem drinkers;

Non-problem drinkers (N = 42) Problem drinkers (N = 42)

22:20 23:19 25:17 30:12 20.4 ± 2.9 21.2 ± 3.1 7.6 ± 2.5 19.9 ± 5.8 703.6 ± 115.4 793.0 ± 165.0 680.2 ± 119.5 679.2 ± 83.0 672.6 ± 113.7 686.7 ± 85.7 649.8 ± 117.1 649.9 ± 68.7 23.4 ± 73.1 113.8 ± 149.0 .013 ± .013 .027 ± .019 38.4 ± 21.5 51.7 ± 23.4 2.7 ± 3.3 4.2 ± 4.3

drinkers and 41 non-problem drinkers). MR = monetary remuneration; CC = course

P. Murphy, H. Garavan / Drug and Alcohol Dependence 115 (2011) 94–100 97

Table 3 Results of Pearson correlation analysis for all of the predictor variables across the entire sample of participants.

Variables Gender Age Alcohol interference Discounting rate Commission error

Gender −.15 −.14 .03 .13 Age .17 .22* .13 Alcohol interference .28** .33**

Discounting rate .12 Commission error

N

3

e p d g g

3

t w t d e a

3

a p o A d n w r

3

c m ( s d p

3 o s c s d t t e i h y .

trolling for the influence of gender and age. When entered together in Step 1, neither gender nor age significantly predicted AUDIT scores; the two variables accounted for only 1% of the variance. Alcohol interference scores, discounting rates, and commission

Table 4 Results of hierarchical logistic regression analysis of variables discriminating prob- lem from non-problem drinkers.

Variables B SE B OR CI (95%)

Step 1 Gender .05 .45 1.05 .44–2.54 Age .24 .24 1.27 .79–2.05

Step 2 Alcohol interference .78* .36 2.18 1.08–4.43 Discounting rate .91** .32 2.49 1.33–4.68 Commission error .56 .30 1.75 .97–3.14

ote. Gender was coded as male = 0, female = 1. * p < .05 (one-tailed test).

** p < .01 (one-tailed test).

.3. Delay discounting task

Table 2 reports mean discounting rates across all reward cat- gories. These discounting rates were the hyperbolic discount arameter (k) calculated from participants’ choices between imme- iate and delayed rewards, as described by Kirby et al. (1999). The eometric mean of discounting rates for the three reward cate- ories were calculated for each participant.

.4. GO–NOGO task

Two participants were beyond three standard deviations above he mean percentage of omission errors (withholding a response hen a GO stimulus is presented) so they were excluded from fur-

her analyses for this task, leaving a final total of 82 (41 problem rinkers and 41 non-problem drinkers). The mean percentages of rrors of commission (responding to an incorrect NOGO stimulus) nd omission are shown in Table 2.

.5. Correlation analysis

Table 3 shows the results of the Pearson correlation analysis for ll of the predictor variables across the entire sample of partici- ants. Preliminary analyses were performed to ensure no violation f the assumptions of normality, linearity and homoscedasticity. lcohol interference scores were positively correlated with both iscounting rates and commission errors, but discounting rates did ot correlate significantly with commission errors. In addition, age as positively correlated with discounting rates. No significant cor-

elations were found between any of the other variables.

.6. Regression analyses

The adequacies of the regression analyses were ensured by hecking that there were no violations of the assumptions of the odels (see Field, 2009). In addition, the four continuous variables

age, alcohol interference scores, discounting rates, and commis- ion errors) were standardised to have a mean of 0 and standard eviation of 1, so the effect-size estimates would be directly com- arable.

.6.1. Logistic regression analysis. Table 4 summarizes the results f the hierarchical logistic regression analysis for cognitive mea- ures discriminating problem from non-problem drinkers, after ontrolling for the influence of gender and age. When entered imultaneously in Step 1, neither gender nor age significantly iscriminated problem from non-problem drinkers; the variables ogether accounted for only 2% of the variance in drinking sta- us. Alcohol interference scores, discounting rates, and commission rrors were entered at Step 2, and accounted for a 34% increase

n explained variance, p < .001. This step was subjected to post oc effect-size and power analysis (Erdfelder et al., 1996), which ielded f2 = .53. According to Cohen (1992), effect sizes (f2) of .02,

15, and .35 are small, medium, and large, respectively. Exami-

nation of the three cognitive variables within this block revealed that both alcohol interference scores and discounting rates, but not commission errors, made significant and unique contributions. The strongest discriminator of problem from non-problem drinkers was discounting rate, recording an odds ratio of 2.49. This indicated that participants with exacerbated impulsive decision-making were almost two and a half times more likely to be a problem than a non-problem drinker, controlling for all other factors in the model. The odds ratio of 2.19 for alcohol interference scores indicated that participants with elevated alcohol-attentional bias were over twice as likely to be a problem than a non-problem drinker, controlling for all other factors in the model.

3.6.2. Multiple regression analyses. Table 5 displays the results of the first hierarchical multiple regression analysis (Model 1) for cognitive measures predicting AUDIT scores among the entire sam- ple of participants, after the effects of gender and age had been controlled. The variables included in Step 1 (gender and age) did not significantly predict AUDIT scores; the variables together accounted for only 2% of the variance. In Step 2, alcohol interfer- ence scores, discounting rates, and commission errors accounted for a 48% increase in explained variance, p < .001. Post hoc effect- size and power analysis (Erdfelder et al., 1996) yielded f2 = .95 for this step, which, according to Cohen (1992), is large in magnitude. All three cognitive variables made significant and unique contribu- tions to explaining the variance in AUDIT scores: examination of the standardized coefficients indicated that alcohol interference scores (ˇ = .36, p < .001), discounting rates (ˇ = .35, p < .001), and commis- sion errors (ˇ = .31, p < .01) had a comparable degree of importance in the model.

Table 5 reports the results of the second hierarchical multiple regression analysis (Model 2) for cognitive measures predicting AUDIT scores among the group of non-problem drinkers, after con-

Note. Model: Nagelkerke R2 for Step 1 = 2%, NS; � Nagelkerke R2 for Step 2 = 34% (p < .001). Gender was coded as male = 0, female = 1.

* p < .05. ** p < .01.

*** p < .001.

98 P. Murphy, H. Garavan / Drug and Alcoho

Table 5 Results of hierarchical multiple regression analyses of variables predicting AUDIT scores for the entire sample of participants as well as the subgroups of problem and non-problem drinkers.

Models Variables B SE B ˇ

Model 1 Entire sample of participants

Step 1 Gender −.46 1.71 −.03 Age 1.04 .86 .14

Step 2 Alcohol interference 2.73*** .69 .36 Discounting rate 2.65*** .66 .35 Commission error 3.33** .67 .31

Model 2 Non-problem drinkers

Step 1 Gender −.32 .81 −.07 Age −.19 .41 −.08

Step 2 Alcohol interference .38 .37 .16 Discounting rate .28 .36 .11 Commission error 1.32** .37 .54

Model 3 Problem drinkers

Step 1 Gender −.68 1.88 −.06 Age .45 .95 .08

Step 2 Alcohol interference 2.36** .86 .40 Discounting rate 1.70* .82 .29 Commission error 1.35 .85 .23

Note. Model 1: R2 = 2% for Step 1, NS; � R2 = 48% for Step 2 (p < .001). Model 2: R2 = 1% for Step 1, NS; � R2 = 28% for Step 2 (p < .05). Model 3: R2 = 1% for Step 1, NS; � R2 = 40% for Step 2 (p < .01). Gender was coded as male = 0, female = 1.

e 2 a a n u h

r A e i s a c a e ( a v d p

4

o s i c c l ( p p i

* p < .05. ** p < .01.

*** p < .001.

rrors were entered at Step 2, and accounted for an additional 8% of explained variance, p < .05. Post hoc effect-size and power nalysis (Erdfelder et al., 1996) yielded f2 = .39 for this step, which, ccording to Cohen (1992), is large in magnitude. Of the three cog- itive variables, only commission errors made a significant and nique contribution to explaining the variance and recorded the ighest standardized coefficient (ˇ = .54, p < .01).

Table 5 shows the results of the third hierarchical multiple egression analysis (Model 3) for cognitive measures predicting UDIT scores among the group of problem drinkers, after the influ- nce of gender and age had been controlled. The variables included n Step 1 (gender and age) did not significantly predict AUDIT cores; the variables together accounted for only 1% of the vari- nce. In Step 2, alcohol interference scores, discounting rates, and ommission errors accounted for a 40% increase in explained vari- nce, p < .01. Post-hoc effect-size and power analysis (Erdfelder t al., 1996) yielded f2 = .69 for this step, which, according to Cohen 1992), is large in magnitude. Two of the three cognitive vari- bles made significant and unique contributions to explaining the ariance, with alcohol interference scores recording a higher stan- ardized coefficient (ˇ = .40, p < .01) than discounting rates (ˇ = .29, < .05).

. Discussion

The present study demonstrates, for the first time, the ability f alcohol-attentional bias (an alcohol Stroop task), impul- ive decision-making (a delay discounting task), and impaired nhibitory control (a GO–NOGO task) to simultaneously: (a) dis- riminate problem from non-problem drinkers (as defined by a ut-off score on the AUDIT) among a non-clinical sample of col- ege students; (b) predict AUDIT scores across student drinkers;

c) predict AUDIT scores within subgroups of problem and non- roblem student drinkers. The results of the logistic regression artially support the first hypothesis: alcohol-attentional bias and

mpulsive decision-making, but not impaired inhibitory control,

l Dependence 115 (2011) 94–100

discriminate problem from non-problem drinkers, after controlling for the effects of gender and age. The multiple regression analysis of the entire sample supports the second hypothesis that all three cognitive measures are related to AUDIT scores when the influence of gender and age are controlled. Finally, the results of the subse- quent multiple regressions reveal that impaired inhibitory control is the only predictor of AUDIT scores for the group of non-problem drinkers, and alcohol-attentional bias and impulsive-decision mak- ing are the only predictors of AUDIT scores for the group of problem drinkers.

The results of the logistic regression model are consistent with previous studies that showed slower colour naming of alcohol- related stimuli (relative to control stimuli) and greater discounting of delayed hypothetical rewards among heavy or problem drink- ing college students than light drinking students (Sharma et al., 2001; Vuchinich and Simpson, 1998). The present results extend these previous findings by showing that student drinkers with elevated alcohol-attentional bias and impulsive decision- making are more than twice as likely to be a problem than a non-problem drinker. In contrast, impaired inhibitory control is not a significant discriminator of problem from non-problem student drinkers, despite previous research demonstrating that heavy alcohol consumption was associated with more failures of inhibitory control in a college sample (Colder and O’Connor, 2002). However, the outcome measure is dichotomous (i.e., prob- lem or non-problem drinking) in the logistic regression model reported here whereas the dependent variable was continuous (i.e., alcohol consumption) in the regression model reported by Colder and O’Connor (2002). Thus, it is possible that splitting the sample of student drinkers into subgroups of problem and non-problem drinkers resulted in the loss of statistical power. Indeed, as discussed below, the results of the present paper’s mul- tiple regression model of the entire sample lend further support to this view. Together, these results describe a cognitive profile linking differences in cognitive processes to problematic drink- ing.

The multiple regression model of the entire sample is con- sistent with previous research that demonstrated measures of alcohol-attentional bias, impulsive decision-making, and impaired inhibitory control are associated with AUDIT or other alcohol use scores, with participants drawn from high school or college (Colder and O’Connor, 2002; Field et al., 2007). The present results extend these previous findings by showing that each of the three cog- nitive measures explains unique variance in AUDIT scores, and contributes similarly to the prediction of the outcome measure. The finding that impaired inhibitory control is a significant predic- tor in this model, but not in the logistic regression model, suggests that statistical power can be optimised by choosing a continu- ous outcome measure (i.e., AUDIT scores) over a dichotomous and somewhat arbitrary one (i.e., problem or non-problem drinking based on a specific AUDIT score). Moreover, the results of this model revealed a cognitive profile that accounts for 48% of the variance in AUDIT scores. In contrast, the cognitive variables in the logis- tic regression model explain only 34% of the variance in drinking status. A further consideration is that the underlying measure of alcohol use (AUDIT) varies in a continuous manner across partici- pants and treating it as a continuous variable in the analyses may better describe the sample than creating a somewhat arbitrary cat- egorisation of participants into problem and non-problem drinkers.

Although the three cognitive measures had roughly equiva- lent predictive power for the entire sample’s AUDIT scores, the subsequent multiple regressions show differences in the pat-

tern of relationships for subgroups of the sample: as described above, impaired inhibitory control is the only significant predic- tor of AUDIT scores for the group of non-problem drinkers, and alcohol-attentional bias and impulsive decision-making are the

Alcoho

o d p fi m e m b fi o c c f a s r

a e i o e t a e a d m m B c i

t o c e s s i 1 c u s o s 9 a t a a t t u

e a b s f a I n a s t

for cocaine-related material and drug-seeking behaviour in active cocaine users. Drug Alcohol Depend. 81, 251–257.

P. Murphy, H. Garavan / Drug and

nly significant predictors of AUDIT scores for the group of problem rinkers. Indeed, these results reveal somewhat distinct cognitive rofiles for each group. Moreover, in light of previous research ndings demonstrating that deficient inhibitory control is a risk arker for progressing towards compulsive substance use (Belin

t al., 2008), and alcohol-attentional bias and impulsive decision- aking are vulnerability markers for maintaining addiction-like

ehaviours (Cox et al., 2002; Yoon et al., 2007), these cognitive pro- les may help to establish cognitive risk profiles at different stages f the development of problematic drinking. The early stages of ollege student drinking may be mediated by deficits in inhibitory ontrol that influence whether or not the individual will transition rom a non-problem to a problem drinker. Once the individual is

problem drinker, however, alcohol-attentional bias and impul- ive decision-making may serve as the strongest risk markers for emaining a problem drinker.

Across the whole sample, measures of alcohol-attentional bias nd impulsive decision-making were significantly correlated with ach other (Field et al., 2007). A novel finding from the present study s the demonstration of a significant correlation between measures f alcohol-attentional bias and impaired inhibitory control. These ffects reveal the complement of cognitive processes that may con- ribute to alcohol use. An alcohol-attentional bias may be related to lcohol-urges and alcohol-seeking behaviours as suggested by the vidence that such biases can predict relapse in alcohol abusers ttempting to abstain (Cox et al., 2002). The correlated impulsive ecision-making and deficient inhibitory control suggest compro- ise in the top-down, prefrontally mediated control that drinkers ight rely upon in order to control or reduce their use; for example,

rewer et al. (2008) identified prefrontal regions linked to cognitive ontrol as being among the best predictors of treatment outcome n a treatment-receiving sample of cocaine users.

One feature of the present study should be noted. In this study, he cut-off score of 11 on the AUDIT was used to distinguish perationally problem from non-problem drinkers. Although the ut-off score of 11 was originally recommended by the develop- rs of the AUDIT (Saunders and Aasland, 1987), a lower cut-off core of 8 was later adopted, in line with lower recommended afe drinking limits (Babor et al., 1989). However, several stud- es (Claussen and Aasland, 1993; Holmila, 1995; Fleming et al., 991; Tsai et al., 2005) have continued to employ the higher ut-off point. The disadvantage is that these studies have doc- mented a low specificity; for example, Fleming et al. (1991) howed a 29% false positive rate when utilising the cut-off score f 11. On the other hand, these studies have demonstrated good ensitivity; for example, Claussen and Aasland (1993) detected 4% of those with DSM-III alcohol misuse or dependence using cut-off score of 11. Consequently, it should be understood that

he present study’s categorisation of participants into problem nd non-problem drinkers is somewhat arbitrary and, given the bsence of a corroborative clinical interview, does not define wo qualitatively different groups, but instead, serves to opera- ionally define two groups who differ in the excess of their alcohol se.

In conclusion, the present results suggest that individual differ- nces in cognitive processes are associated with problem drinking mong college students. They also suggest that the relationships etween cognition and alcohol use may vary depending on the everity of that use; in this regard, the distinct cognitive profiles or each subgroup may help to establish cognitive risk profiles t different stages of the development of problematic drinking. t will be important to determine the extent to which the cog- itive processes implicated here as contributing to alcohol use nd misuse are inter-dependent, contribute to the use of other

ubstances, and are potentially amenable to therapeutic interven- ion.

l Dependence 115 (2011) 94–100 99

Role of funding source

This study was unfunded.

Contributors

Both authors designed the study and wrote the protocol. The first author conducted the literature review, data collection, sta- tistical analysis, and wrote the first draft of the manuscript. Both authors contributed to and approved the final manuscript.

Conflict of interest

Both authors declare they have no conflicts of interest.

Acknowledgement

The authors would like to thank Jude Cosgrove for her statistical advice.

References

Babor, T.F., De la Fuente, J.R., Saunders, J., Grant, M., 1989. AUDIT: The Alcohol Use Disorder Identification Test: Guidelines for Use in Primary Health Care. WHO, Division of Mental Health, Geneva.

Belin, D., Mar, A.C., Dalley, J.W., Robbins, T.W., Everitt, B.J., 2008. High impulsivity predicts the switch to compulsive cocaine-taking. Science 320, 1352–1355.

Bickel, W.K., Odum, A.L., Madden, G.J., 1999. Impulsivity and cigarette smoking: delay discounting in current, never, and ex-smokers. Psychopharmacology 146, 447–454.

Brewer, J.A., Worhunsky, P.D., Carroll, K.M., Rounsaville, B.J., Potenza, M.N., 2008. Pretreatment brain activation during stroop task is associated with outcomes in cocaine-dependent patients. Biol. Psychiatry 64, 998–1004.

Cohen, J., 1992. A power primer. Psychol. Bull. 112, 115–119. Colder, C.R., O’Connor, R., 2002. Attention biases and disinhibited behavior as pre-

dictors of alcohol use and enhancement reasons for drinking. Psychol. Addict. Behav. 16, 325–332.

Cox, W.M., Blount, J.P., Rozak, A.M., 2000. Alcohol abusers’ and nonabusers’ dis- traction by alcohol and concern-related stimuli. Am. J. Drug Alcohol Abuse 26, 489–495.

Cox, W.M., Fadardi, J.S., Pothos, E.M., 2006. The addiction Stroop test: theoretical considerations and procedural recommendations. Psychol. Bull. 132, 443–476.

Cox, W.M., Hogan, L.M., Kristian, M.R., Race, J.H., 2002. Alcohol attentional bias as a predictor of alcohol abusers’ treatment outcome. Drug Alcohol Depend. 68, 237–243.

Cox, W.M., Yeates, G.N., Regan, C.M., 1999. Effects of alcohol cues on cognitive pro- cessing in heavy and light drinkers. Drug Alcohol Depend. 55, 85–89.

Claussen, B., Aasland, O.G., 1993. The Alcohol Use Disorders Identification Test (AUDIT) in a routine health examination of long-term unemployed. Addiction 88, 363–368.

de Wit, H., Richards, J.B., 2004. Dual determinants of drug use in humans: reward and impulsivity. In: Bevans, R., Bardo, M. (Eds.), Motivational Factors in the Etiology of Drug Abuse: Volume 50 of the Nebraska Symposium on Motivation. University of Nebraska Press, Lincoln, NE, pp. 19–55.

Dom, G., D’haene, P., Hulsyijn, W., Sabbe, B., 2006. Impulsivity in abstinent early- and late-onset alcoholics: differences in self-report measures and a discounting task. Addiction 101, 50–59.

Erdfelder, E., Faul, F., Buchner, A., 1996. G*Power: a general power analysis program. Behav. Res. Methods Instrum. 28, 1–11.

Evenden, J.L., 1999. Varieties of impulsivity. Psychopharmacology 146, 348–361. Field, A., 2009. Discovering Statistics using SPSS. Sage Publications, London. Field, M., Christiansen, P., Cole, J., Goudie, A., 2007. Delay discounting and the alcohol

Stroop in heavy drinking adolescents. Addiction 102, 579–586. Fleming, M.F., Barry, K.L., Mac Donald, R., 1991. The Alcohol Use Disorders Identifi-

cation Test (AUDIT) in a college sample. Int. J. Addict. 26, 1173–1185. Forman, S.D., Dougherty, G.G., Casey, B.J., Siegle, G.J., Braver, T.S., Barch, D.M., Stenger,

V.A., Wick-Hull, C., Pisarov, L.A., Lorensen, E., 2004. Opiate addicts lack error- dependent activation of rostral anterior cingulate. Biol. Psychiatry 55, 531–537.

Franken, I.H., Kroom, L.Y., Wiers, R.W., Jansen, A., 2000. Selective cognitive process- ing of drug cues in heroin dependence. J. Psychopharmacol. 14, 395–400.

Gill, J.S., 2002. Reported levels of alcohol consumption and binge drinking within the UK undergraduate student population over the last 25 years. Alcohol Alcohol. 37, 109–120.

Hester, R., Dixon, V., Garavan, H., 2006. The relationship between attentional bias

Hester, R., Garavan, H., 2004. Executive dysfunction in cocaine addiction: evi- dence for discordant frontal, cingulate, and cerebellar activity. J. Neurosci. 24, 11017–11022.

1 Alcoho

H

K

K

K

K

L

M

M

M

O

O

Waters, A.J., Feyerabend, C., 2000. Determinants and effects of attentional bias in smokers. Psychol. Addict. Behav. 14, 111–120.

Yoon, J.H., Higgins, S.T., Heil, S.H., Sugarbaker, R.J., Thomas, C.S., Badger, G.J., 2007.

00 P. Murphy, H. Garavan / Drug and

olmila, M., 1995. Intoxication and hazardous use of alcohol: results from the 1992 Finnish Drinking Habits Study. Addiction 90, 785–792.

amarajan, C., Porjesz, B., Jones, K.A., Choi, K., Chorlian, D.B., Padmanabhapillai, A., Rangaswamy, M., Stimus, A.T., Begleiter, H., 2005. Alcoholism is a disinhibitory disorder: neurophysiological evidence from a Go/No-Go task. Biol. Psychol. 69, 353–373.

aufman, J.N., Ross, T.J., Stein, E.A., Garavan, H., 2003. Cingulate hypoactivity in cocaine users during a GO/NOGO task as revealed by event-related fMRI. J. Neurosci. 23, 7839–7843.

enny, P.J., 2006. Brain reward systems and compulsive drug use. Trends Pharmacol. Sci. 28, 135–141.

irby, K.N., Petry, N.M., Bickel, W.K., 1999. Heroin addicts have higher discount rates for delayed rewards than non-drug-using controls. J. Exp. Psychol. Gen. 128, 78–87.

ubman, D.I., Peters, L.A., Mogg, K., Bradley, B.P., Deakin, J.F.W., 2000. Attentional bias for drug cues in opiate dependence. Psychol. Med. 30, 169–175.

adden, G.J., Bickel, W.K., Jacobs, E.A., 1999. Discounting of delayed rewards in opioid-dependent outpatients: exponential or hyperbolic discounting func- tions? Exp. Clin. Psychopharmacol. 7, 284–293.

adden, G.J., Petry, N.M., Badger, G.J., Bickel, W.K., 1997. Impulsive and self-control choices in opioid-dependent patients and non-drug using control participants: drug and monetary rewards. Exp. Clin. Psychopharmacol. 5, 256–262.

itchell, S.H., 2004. Measuring impulsivity and modeling its association with cigarette smoking. Behav. Cogn. Neurosci. Rev. 3, 261–275.

dum, A.L., Madden, G.J., Badger, G.J., Bickel, W.K., 2000. Needle sharing in opioid- dependent outpatients: psychological processes underlying risk. Drug Alcohol Depend. 60, 259–266.

lmstead, M.C., 2006. Animal models of drug addiction: where do we go from here? Q. J. Exp. Psychol. 59, 625–653.

l Dependence 115 (2011) 94–100

Reynolds, B., Ortengren, A., Richards, J.B., de Wit, H., 2006. Dimensions of impulsive behavior: personality and behavioral measures. Pers. Indiv. Differ. 40, 305–315.

Saunders, J.B., Aasland, O.G., 1987. WHO Collaborative Project on Identification and Treatment of Persons with Harmful Alcohol Consumption. Report on Phase 1, Development of a Screening Instrument. WHO, Division of Mental Health, Geneva.

Saunders, J.B., Aasland, O.G., Babor, T.F., de la Fuente, J.R., Grant, M., 1993. Develop- ment of the Alcohol Use Disorders Screening Test (AUDIT). WHO collaborative project on early detection of persons with harmful alcohol consumption. II. Addiction 88, 791–804.

Sharma, D., Albery, I.P., Cook, C., 2001. Selective attentional bias to alcohol related stimuli in problem drinkers and non-problem drinkers. Addiction 96, 285–295.

Stormark, K.M., Laberg, J.C., Nordby, H., Hugdahl, K., 2000. Alcoholics’ selective atten- tion to alcohol stimuli: automated processing? J. Stud. Alcohol 61, 18–23.

Tsai, M.C., Tsai, Y.F., Chen, C.Y., Liu, C.Y., 2005. Alcohol Use Disorders Identifica- tion Test (AUDIT): establishment of cut-off scores in a hospitalized Chinese population. Alcohol. Clin. Exp. Res. 29, 53–57.

Verdejo-García, A.J., Perales, J.C., Pérez-Garcia, M., 2007. Cognitive impulsivity in cocaine and heroin polysubstance abusers. Addict. Behav. 32, 950–966.

Vuchinich, R.E., Simpson, C.A., 1998. Hyperbolic temporal discounting in social drinkers and problem drinkers. Exp. Clin. Psychopharmacol. 6, 292–305.

Delay discounting predicts postpartum relapse to cigarette smoking among pregnant women. Exp. Clin. Psychopharmacol. 2, 176–186.