Brain function during cognitive flexibility and whitematter integrity in alcohol-dependent patients,problematic drinkers and healthy controls

Jochem M. Jansen1*, Ruth J. van Holst1,2,3*, Wim van den Brink1, Dick J. Veltman1,4,Matthan W. A. Caan5 & Anna E. Goudriaan1,6

Academic Medical Centre, Department of Psychiatry, Amsterdam Institute for Addiction Research, The Netherlands1, Donders Institute for Cognition, Brain andBehaviour, Radboud University, The Netherlands2, Department of Neurology, Radboud University Medical Centre, The Netherlands3, VU University MedicalCenter, The Netherlands4, Academic Medical Centre, Department of Radiology, University of Amsterdam, The Netherlands5 and Arkin Mental Health,The Netherlands6

ABSTRACT

Cognitive flexibility has been associated with prefrontal white matter (WM) integrity in healthy controls (HCs),showing that lower WM integrity is associated with worse performance. Although both cognitive flexibility and WMintegrity have been found to be aberrant in alcohol-dependent (AD) patients, the relationship between the two hasnever been tested. In this study, we investigated the association between WM tract density and cognitive flexibility inpatients with AD (n = 26) and HCs (n = 22). In order to assess the influence of AD severity, we also included a groupof problematic drinkers (PrDs; n = 23) who did not meet the AD criteria. Behavioral responses and brain activityduring a cognitive flexibility task were measured during functional magnetic resonance imaging. Probabilistic fibertracking was performed between the dorsolateral prefrontal cortex and the basal ganglia; two crucial regions for taskswitching. Finally, the task-related functional connectivity between these areas was assessed. There were no signifi-cant group differences in the task performance. However, compared with HCs, AD patients and PrDs showeddecreased WM integrity and increased prefrontal brain activation during task switching. Evidence is presented for acompensatory mechanism, involving recruitment of additional prefrontal resources in order to compensate forWM and neural function impairments in AD patients and PrDs. Although present in both alcohol groups, the PrDswere more successful in invoking this compensatory mechanism when compared to the AD patients. We proposethat this may therefore serve as a protective factor, precluding transition from problematic drinking into alcoholdependence.

Keywords: Alcohol dependence, cognitive flexibility, fiber tracking, fMRI.

Correspondence to: Jochem M. Jansen, Academic Medical Centre, Department of Psychiatry, University of Amsterdam, Room PA3-220, 1100 DDAmsterdam, The Netherlands. E-mail: j.m.jansen@amc.uva.nl or jochemjansen89@gmail.com

INTRODUCTION

Alcohol and drug dependence are chronic relapsingbrain disorders characterized by altered reward circuitryand weakened cognitive control systems that often leadto impairments in cognitive functioning, includingimpaired cognitive flexibility (Baler & Volkow 2006;Koob & Volkow 2010). Cognitive flexibility is a prerequi-site for goal-directed behavior and involves executivedecision making preceding switching behavior (Lezak2004). It has been defined as the ability to adjust one’s

thinking from old to new situations (Moore &Malinowski 2009). Previous studies have shownimpaired cognitive flexibility (Vanes et al. 2014) inalmost all substance use disorders (for a review, seeVolkow et al. 2002) and in problematic gamblers (for areview, see van Holst et al. 2010). These impairments incognitive flexibility are related to the compulsive use ofdrugs and/or alcohol, which is one of the key elementsin the development and persistence of substancedependence (Everitt & Robbins 2005; Stalnaker et al.2009).

*These authors contributed equally to this paper and should be referred to as joint first authors.

ORIGINAL ARTICLE

bs_bs_bannerAddiction Biologydoi:10.1111/adb.12199

© 2014 Society for the Study of Addiction Addiction Biology

Lesion and neuro-imaging studies have shown thatthe dorsolateral prefrontal cortex (dlPFC) is essential forcognitive flexibility (Badre & Wagner 2006; Nyhus &Barcelo 2009). Neuro-imaging studies on cognitiveflexibility in healthy subjects using cognitive switchingparadigms observed increased activation in dlPFC,ventrolateral prefrontal cortex (vlPFC), including theinferior frontal gyrus (IFG), anterior cingulate cortex(ACC), posterior parietal cortex and basal gangliaduring correct task switching (Sohn et al. 2000; Smithet al. 2004; Ravizza & Carter 2008). In addition, manystudies have reported on impaired prefrontal functioningduring cognitive tasks in substance-dependent patients(Baler & Volkow 2006; Goldstein & Volkow 2011).However, neural processes underlying aberrant cognitiveswitching in alcohol use disorders are still not wellexamined.

Previous studies have also shown that prefrontalwhite matter (WM) integrity is related to cognitive func-tioning and cognitive flexibility (Bennett et al. 2011; vanSchouwenburg et al. 2014). Alcohol is neurotoxic andinflicts damage to the brain, and WM in the prefrontalcortex in particular (Kril et al. 1997; Harper 2007; Yehet al. 2009). A decline in WM integrity in alcohol-dependent (AD) patients has previously been linked tocognitive deficits, including (working) memory, attentionand responsiveness to reward (Harris et al. 2008;Chanraud et al. 2009; Pfefferbaum et al. 2010; Schulteet al. 2010, 2012). However, there are no studies thathave investigated the relationship between prefrontalWM integrity and cognitive flexibility in a sample of ADsubjects and problematic drinkers (PrDs). As previousstudies strongly associate the dlPFC and basal gangliawith cognitive flexibility (van Schouwenburg et al. 2012),it is hypothesized that impaired switching behaviorin substance dependence is related to damaged fiberbundles connecting the prefrontal cortex and basalganglia, including ventral striatum and putamen (vanSchouwenburg et al. 2014). WM integrity can be investi-gated non-invasively using diffusion tensor imaging(DTI), which provides fractional anisotropy (FA) andmean diffusivity (MD) values to describe the macroscopiccharacteristics of fiber bundles (Hagmann et al. 2006).Decreases in FA are thought to reflect a decline in fiberdensity and coherence, whereas increases in MD valuesmirror cell membrane damage (Beaulieu 2002). In ADpatients, decreases in FA and increases in MD values havebeen reported (Pfefferbaum & Sullivan 2004; Harper2007, 2009; Pfefferbaum et al. 2009).

Similar to findings in substance dependence, there is adecrease in cognitive functioning and WM integrity inhealthy aging of the brain (for a review, see Goh & Park2009; Park & Reuter-Lorenz 2009). Neuro-imagingstudies have reliably reported increased prefrontal activa-

tion during cognitive tasks in elderly subjects (Park &Reuter-Lorenz 2009; Cooper, Worthy & Maddox 2013).High-cognitive functioning elderly exhibit higher activa-tions when compared to low-cognitive functioning elderly(Park & Reuter-Lorenz 2009). The scaffolding theory ofaging and cognition (STAC) states that increased activa-tion in cognitively demanding tasks in healthy aging is acompensatory mechanism for decreased WM integrityand neural functioning (Park & Reuter-Lorenz 2009).Increased prefrontal activations are sometimes accompa-nied by decreased activity in more posterior parts of thebrain (Davis et al. 2008). Higher activity during cognitiveand emotional tasks has previously been reported andexplained as a compensatory mechanism in substance-dependent populations (Yucel & Lubman 2007; Goldstein& Volkow 2011; Charlet et al. 2014). For example, higheractivations during high-working memory load and affec-tive face recognition were associated with lower relapserisk (Charlet et al. 2014). These higher activations arelikely to be associated with alcohol-induced WM damageand support the explanation that higher activations incognitively demanding situations reflect a compensatorymechanism in patients with AD.

The aim of this study is to investigate cognitive flex-ibility and underlying neuronal mechanisms in a sampleof AD patients, PrDs and healthy controls (HCs). Theinclusion of PrDs enables the assessment of WM integrityand cognitive flexibility across a broad range of alcohol-related problems. On a behavioral level, we hypothesizethat AD patients and PrDs perform worse compared toHC, as indicated by fewer correct switch trials and higherswitch costs. In line with the STAC, we hypothesize thatboth the AD and the PrD groups show a decline in WMintegrity together with an increase in prefrontal neuralactivity during cognitive switching when compared toHCs. In analogy with high and low functioning elderly, itis expected that PrDs show higher compensatory activa-tion of the prefrontal cortex than AD patients. To test ourhypothesis, we applied a cognitive switching task, whilemeasuring the BOLD response in the brain, and per-formed diffusion tensor fiber tracking between the dlPFCand the basal ganglia.

MATERIALS AND METHODS

Participants

AD subjects were recruited from Dutch addiction treat-ment centers where they received cognitive behavioraltherapy. HCs and PrDs were recruited through advertise-ments in local newspapers. All participants were male;HCs did not drink more than 21 standard units (10 g) ofalcohol per week. The Ethical Review Board of the Aca-demic Medical Centre approved the study and all subjects

2 Jochem M. Jansen et al.

© 2014 Society for the Study of Addiction Addiction Biology

provided written informed consent and were remuner-ated for their participation. The number of subjectsvaried per analysis due to missing data in either one ormore modalities; the exact numbers per analysis can befound in Table 2 and are reported throughout the Resultssection.

AD subjects were diagnosed according to DSM-IV-TRalcohol dependence criteria with Section J of the Dutchversion of the Clinical International Diagnostic Inventory(World Health Organization 1997). A measure of alcoholproblem severity was obtained for all participants byadministering the alcohol use disorder identification test(AUDIT; Bush et al. 1998). Furthermore, to ensure thatall participants were detoxified from alcohol, AD patientshad to be abstinent for at least 2 weeks (ranging from 2weeks to 12 months) to be included in the study. The PrDswere abstinent for at least 24 hours, which was con-firmed with urine and breathing tests. For a more elabo-rative description of the PrD group, see Starcke et al.(2013). Urine samples were analyzed by a clinical toxicol-ogy laboratory (ATAL Medical Diagnostics, Amsterdam,the Netherlands) and none of the tests were positive foralcohol.

Exclusion criteria for all groups included lifetime diag-nosis of schizophrenia or psychotic episodes; 12-monthdiagnosis of manic disorder, obsessive compulsive disor-der and post-traumatic stress disorder; other substanceuse disorders than alcohol dependence (except for nico-tine); treatment for mental disorders in the past 12months (except for alcohol dependence); use of psycho-tropic medication; difficulty reading Dutch, age under 18years; IQ below 80 (measured by the Dutch AdultReading Test; Schmand et al. 1991); positive urine screenfor alcohol, amphetamines, benzodiazepines, opioids orcocaine; history or current treatment for neurologicaldisorders; major internal disorders; brain trauma; orexposure to other neurotoxic factors than alcohol.

Assessments and statistical analyses

Functional magnetic resonance imaging (fMRI) paradigm

In order to test cognitive flexibility, we applied an fMRIcompatible self-paced switch task adapted from Sohnet al. (2000). During each trial, a stimulus consisting ofone letter and one digit was shown, the color of whichindicated the task to be performed: vowel/consonant(blue) or odd/even (red) judgments. Letters were takenfrom the set [a, e, i, u, b, c, d, f] and digits were taken fromthe set [2, 4, 6, 8, 3, 5, 7, 9]. Two consecutive trials nevercontained the same letter or digit. Color–task andstimulus–response associations were counterbalancedacross subjects. Task switching (color change) occurredrandomly after four to six trials to avoid predictability.First trials immediately after a color switch were defined

as ‘switch trials’, all other trials were defined as ‘repeattrials’. For example, on a switch trial, the subject had toswitch from making a number judgment to a letter judg-ment or vice versa. If the target was blue, the criteriondimension was letter, and the criterion response was a leftkey press for a vowel and a right key press for a conso-nant. The task ended after 32 task switches and after 160repeat trials, in a single run that took around 21 minutesto be completed. The task was self-paced with amaximum response window of 4000 milliseconds pertrial and an interstimulus interval of 500 milliseconds. Ifthe subject did not respond within 4000 milliseconds,the trial was considered an error and the next trialappeared. In addition, the paradigm included six30-second baseline blocks, during which a fixation crosswas shown and in which the subject had to remain stilland wait until the task resumed, i.e. a passive baselinecondition. Subjects performed a training session outsidethe scanner. There was no provision of feedback duringthe task. Behavioral variables of interest included (1) per-centage correct task switches; (2) percentage of correcttask repeats; and (3) switch cost [reaction times (RTs) ofswitch trials minus RTs of repeat trials].

Behavioral analysis. Demographic and behavioral datawere analyzed using univariate analysis of variance(ANOVA and ANCOVAs) and Tukey’s post hoc tests inSPSS 20.0 (IBM Corp., Armonk, NY, USA). Non-normallydistributed data (i.e. age, AUDIT scores and number ofdrinks per week) were analyzed using Kruskal–Wallistests, and post hoc tests were conducted with Mann–Whitney tests. A paired-sample t-test was used to investi-gate differences between switch and repeat conditionsacross groups. Furthermore, an ANOVA test was used toassess group differences in the task performance. Finally,the effect of abstinence duration was assessed in a corre-lation analysis with task performance, brain activity andWM integrity only in AD patients because PrDs weremostly abstinent for less than 2 weeks. All analyses wereperformed two-tailed with α set at 0.05.

fMRI analysis. Imaging data were obtained using a 3.0 TIntera whole-body fMRI scanner (Philips MedicalSystems, Best, the Netherlands) with a phased arraySENSE RF eight-channel receiver head coil. Task stimuliwere projected on a screen behind the subject’s head atthe end of the scanner table. The screen was visible for thesubject through a mirror mounted above the subject’shead. Two magnet-compatible response boxes were usedto record the subject’s responses. A total of 35 axial slices(voxel size 2.29 × 2.29 × 3 mm) T2*-weighted echoplanar images (EPIs) were obtained, without interslicegap, a matrix size of 96 × 96 mm2, repetition time/echotime = 2.3 seconds/30 milliseconds and a bandwidth of

Brain function and WM in AUD 3

© 2014 Society for the Study of Addiction Addiction Biology

90 kHz. These images are sensitive to blood oxygenationlevel-dependent (BOLD) contrast and covered the entirebrain, except for the inferior regions of the cerebellum.

A T1-weighted structural scan was made forco-registration with the fMRI data (voxel size1 × 1 × 1 mm; 170 slices). Functional imaging analyseswere performed with SPM5 (Statistical ParametricMapping, Wellcome Trust Centre for Neuroimaging,London, UK). Images were manually reoriented and sub-sequently slice-timed, and realigned and unwarped. Next,images were warped to MNI space using each subject’sco-registered T1 image, and spatially smoothed using an8-mm full width at half maximum Gaussian kernel.

fMRI data were analyzed in the context of the generallinear model using delta functions convolved with acanonical hemodynamic response function to modelresponses to each type of stimulus (switch trials andrepeat trials). Contrast images containing parameter esti-mates were entered into a second-level (random effects)analysis. Main effects across groups for each contrastwere analyzed using one-sample t-test implemented inSPM5 and are reported at P < 0.05 corrected for multiplecomparisons at the voxel level across the whole brain,according to the family-wise error (FWE) method. Peaksignificant voxels per group were extracted with Marsbarv0.43 in order to assess the relationship between activitylevels, task performance and DTI measurements (Brettet al. 2002).

Group interactions were investigated using specific apriori regions of interest (ROIs) with a threshold set atP < 0.05, FWE corrected for multiple comparisons(Friston et al. 1996). The dlPFC, vlPFC, ACC, posteriorparietal cortex and basal ganglia were defined as a prioriROIs, given their role in cognitive switching (Sohn et al.2000; Smith et al. 2004; Ravizza & Carter 2008). Thebasal ganglia ROI, including the putamen, was definedusing the WFU PickAtlas Tool v2.4 (Maldjian et al. 2003)that incorporates the automatic anatomical labeling atlas(Tzourio-Mazoyer et al. 2002). Activity in the dlPFC, pos-terior parietal cortex and vlPFC was detected by usingROI templates of the dlPFC, parietal cortex, and themiddle orbitofrontal cortex and inferior orbitofrontalcortex, respectively. These ROIs were defined using theonline Voxel of Interest Database (Fox & Lancaster 1994).Second, for exploratory purposes, we reported brainregions that were not predicted a priori but which met athreshold of P < 0.001, uncorrected with a cluster size of10 contiguous voxels (Supporting Information Appen-dix S1).

Functional connectivity. Functional connectivity analyseswere conducted with the Conn toolbox in Matlab(Whitfield-Gabrieli & Nieto-Castanon 2012). Input forthe functional connectivity analyses were the first level

switch versus repeat contrasts for each subject. Persubject, the correlation coefficient between the dlPFC andbasal ganglia ROIs was computed per hemisphere andanalyzed in SPSS.

Diffusion tensor imaging: acquisition, pre-processing andfiber tracking

Diffusion weighted EPIs were acquired along 32 direc-tions with a b-value of 1000 seconds/mm2 and oneacquisition without diffusion weighting (b = 0). All acqui-sitions were made with the following parameters:TR = 4.862 milliseconds, TE = 94 milliseconds, 38 axialinterleaved slices with a 3-mm slice thickness with nogap, with a 112 × 110 mm2 matrix (0.898 × 0.898 mmin-plane resolution).

The pre-processing of the DTI data was performedusing in-house developed software, written in Matlab(The MathWorks Inc., Natick, MA, USA) and wasexecuted on the Dutch Grid using a web interface to thee-Bioinfra gateway (Olabarriaga, Glatard & de Boer2010; Shahand et al. 2011). We corrected for headmotion and deformations induced by eddy currents byan affine registration of the diffusion weighted images(DWIs) to the non-diffusion weighted image. The gradientdirections were corrected by the rotation componentof the transformation. The DWIs were resampledisotropically. Rician noise in the DWIs was reduced by anadaptive noise filtering method (Caan et al. 2010). Diffu-sion tensors were estimated with a non-linear leastsquares procedure. From the resulting tensors, FA andMD maps were computed.

Additional processing was required to model crossingfibers (BEDPOSTX, Bayesian Estimation of DiffusionParameters Obtained using Sampling Techniques).Probabilistic fiber tracking was performed using [FMRIB’sSoftware Library (FSL) software (Behrens et al. 2007)].Seed regions for the fiber tracking were based on the fMRIresults by saving suprathreshold voxels in the dlPFC andbasal ganglia masks as binary masks, reoriented andtransformed back to single subject space with SPM8.These ROI-masks were resliced with the BEDPOSTXoutput file and subsequently used as seed and waypointmasks in fiber tracking. Fiber tracking was performedbilaterally.

To improve intersubject correspondence along theindividually tracked fibers, we non-rigidly registered(Ashburner 2007) the non-diffusion weighted images(b0) and warped the tracked fibers accordingly, within-house written software in MATLAB (7.10.0, TheMathWorks Inc.). The fiber with minimal distance to allfibers of all subjects was chosen as the reference fiber. Thenormal plane along this fiber was used to define corre-spondence to the fibers of all subjects. The FA profileswith 2-mm distance sampling were generated on the

4 Jochem M. Jansen et al.

© 2014 Society for the Study of Addiction Addiction Biology

bilateral dlPFC-basal ganglia tracts and resulted in 10mean FA and MD points along the tracts. Subse-quent analyses were performed in SPSS using a repeatedmeasures ANCOVA test per tract to check for group dif-ferences in mean FA and MD values, and partial correla-tion analysis to assess the relationship with taskperformance.

In order to reduce the number of comparisons, thebehavioral measurements of task switching were onlycorrelated with those segments that differed most in FAand MD in the combined AD and PrD group, comparedwith the HC group. Greater differences are thought toreflect a higher degree of compromised WM integrity andthese segments are therefore of interest. The differencebetween the combined alcohol group compared with theHC group was greatest in the most prefrontal segments(corresponding to number 10) of the left (for MD) andright hemisphere (for FA), indicating that WM integritywas mostly compromised in these areas. Partial correla-tion analysis between these areas and the task perfor-mance measurements were carried out. All DTI analyseswere corrected for age because age has a well-knownnegative effect on WM integrity (Bennett et al. 2011).

RESULTS

Demographics

Table 1 summarizes the demographic and clinical char-acteristics for the three groups: AD, PrDs and HCs. Therewere no significant differences in age or IQ, but all groupssignificantly differed in their AUDIT scores and number ofdrinks per week. Post hoc comparisons revealed that theAD group had significantly higher scores for the AUDITand on the number of standard drinks per week com-

pared with the PrD and the HC groups. Furthermore, thePrD group had higher AUDIT scores and more standarddrinks per week compared with the HC group(AD > PrD > HC). Analysis of the effect of abstinenceduration in AD patients revealed no significant effects onthe ask performance, brain activity or WM integrity.

Behavioral performance on the switch task

Across groups, we found that subjects respondedsignificantly faster during repeat trials (M = 1118.7,SD = 26.6) than to switch trials [M = 1551.0, SD =316.9, t(66) = 17.25, P < 0.001]. In addition, acrossgroups, subjects made more correct responses duringrepeat trials (M = 85.0 percent, SD = 13.4) than duringswitch trials [M = 76.2 percent, SD = 16.7, t(68) =−10.12, P < 0.001]. There were no significant differencesbetween groups in the percentage of correct taskswitches, percentage of erroneous task repeats or switchcosts (see Table 2).

fMRI results

Main effect task switch versus task repeat

Increased activity during switch versus repeat trials in allgroups was found in the whole brain analysis for thebilateral basal ganglia, ventrolateral and dorsolateralPFC, right middle temporal cortex, left anterior cingulatecortex and right premotor cortex (see Supporting Infor-mation Appendix S1).

Group interaction task switch versus repeats

Group interactions were analyzed with ROIs and theactivity differed significantly between groups on theirswitch versus repeat brain activity, mainly in the dlPFC.

Table 1 Sample characteristics. Values for age, IQ, AUDIT, drinks per week, DSM-IV criteria for alcohol abuse and dependence aredenoted as mean (standard deviation) and [min : max].

Healthy controls(n = 22)

Problem drinkers(n = 21)

Alcohol dependent(n = 26) F (d.f.) Significance

Age 38.5 (10.3) 42.1 (13.8) 44,3 (9.2) 3.08 (2) ns[22:53] [22:61] [21:59]

IQ 104.5 (12.5) 106.7 (14.3) 101.0 (15.1) 2.12 (2) ns[80:123] [73:128] [71:125]

AUDIT 5.1 (4.4) 18.5 (5.3) 27.9 (5.1) 51.83 (2) P < 0.001[0:15] [11:33] [11:36]

Number of standard drinks/week 7.3 (7.2) 46.7 (31.5) 96.7 (61.2) 50.90 (2) P < 0.001[0:25] [0:144] [0.5:264]

DSM-IV criteria alcohol abuse 0.14 (0.36) 0.8 (0.95) 1.9 (0.85) 30.32 (2) P < 0.001[0.0:1.0] [0.0:3.0] [1.0:3.0]

DSM-criteria alcohol dependence 0.05 (0.22) 1.3 (1.46) 6.65 (1.56) 100.72 (2) P < 0.001[0.0:1.0] [0.0:6.0] [3.0:9.0]

AUDIT = alcohol use disorder identification test; d.f. = degrees of freedom; ns, non-significant.

Brain function and WM in AUD 5

© 2014 Society for the Study of Addiction Addiction Biology

Specifically, the PrD group particularly showed moreactivity in the left dlPFC than the HC group, and moreactivity in the left dlPFC than the AD group at a (trend)significant level (see Table 3). Task performance corre-lated with task performance only in HCs. The activity inthe left dlPFC peak voxel was positively correlated withpercentage correct switch trials (r = 0.48, n = 18,P < 0.05) and negatively correlated with erroneousrepeats (r = −0.49, n = 17, P < 0.05).

Functional connectivity

fMRI time courses were extracted from the bilateraldlPFC and basal ganglia masks and tested for correla-tional changes between the regions during switchand repeat trials. Both left [t(45) = 4.27, P < 0.001]and right [t(45) = 4.85, P < 0.001] hemisphereconnectivity between these ROIs were significantlyhigher during switch trials compared with repeat trials(left; r = 0.15 P < 0.001, right; r = 0.15, P < 0.001).Functional connectivity did not differ between groupsand did not correlate with AUDIT scores or switch taskperformance.

Fiber tracking

The repeated measures ANCOVA revealed no significantinteraction between group and segment for MD and FA

values in either hemisphere. The FA and MD values in theleft and right hemisphere all revealed significant effectsfor segment, except for left MD values. This indicates thatMD and FA values differ along the fiber tract. Main effectof group was only significant for the left and right MDvalues (see Table 4). Post hoc analyses for these significantgroup effects in MD revealed that HCs had significantlylower MD values than AD patients (left; P < 0.05, right;P < 0.01) and PrD (left; P < 0.05, right; P < 0.01),whereas AD patients and PrD did not significantly differin MD values (see Fig. 1: HC < PrD = AD). Mean FA andMD values are reported in Supporting InformationAppendix S2.

Relationship between FA values and task performance

The results of the correlation analyses revealed that forthe overall group (HCs + PrDs + ADs), prefrontal WM FAvalues positively correlated with the percentage correctswitches and negatively correlated with erroneousrepeats (see Table 5), whereas MD values were not corre-lated with task performance. The results of the analysisper group indicate that the relationships between FA andtask performance were restricted to the HC group (seeTable 6). Moreover, in PrD, MD values negatively corre-lated with percentage correct switches. FA or MD valueswere unrelated to functional connectivity. The MD and FA

Table 2 Behavioral measurements of task switching. This table shows the results of the ANCOVA test with age as covariate.

Healthy controls(n = 22)

Problem drinkers(n = 21)

Alcohol dependent(n = 26) F value (d.f.) Significance

Percentage correct switches 78.1 (19.6) 79.0 (17.0) 74.2 (16.8) 0.55 (2,66) nsSwitch cost (RT)a 136.7 (20.8) 128.1 (27.2) 124 (29.3) 1.16 (2,61) nsPercentage erroneous repeats 2.0 (1.2) 2.1 (1.3) 2.5 (0.8) 1.81 (2,66) ns

aThe number of participants is smaller for the switch cost analysis due to missing data (healthy controls = 19, problem drinkers = 20, alcohol depend-ent = 25). d.f. = degrees of freedom; ns = non-significant; RT = reaction time.

Table 3 Between group comparisons switch versus repeat functional magnetic resonance imaging. This table reports the differentialactivations between the groups for the pre-selected regions of interest. Significance levels are FWE-corrected, reported results with aP-value smaller than 0.05 are considered significant, whereas a P-value of P = 0.06 is considered trend significant.

Group comparison Brain region Hemisphere Z-value Number of voxels x y zP(FWE-corr)

HC > AD ns – – – – – – –AD > HC dlPFC Left 3.51 4 −36 30 18 0.06AD > PrD ns – – – – – – –PrD > HC dlPFC Left 3.63 28 −33 36 27 0.04– – – 3.64 – −36 33 30 0.04PrD > AD dlPFC Left 3.52 27 −45 30 30 0.06– vlPFC Left 2.90 3 −45 15 −6 0.01

AD = alcohol-dependent patient; dlPFC = dorsolateral prefrontal cortex; few-corr = family-wise error correction; HC = healthy control; ns = non-signifi-cant; PrD = problematic drinker; vlPFC = ventrolateral prefrontal cortex.

6 Jochem M. Jansen et al.

© 2014 Society for the Study of Addiction Addiction Biology

values did not correlate with activity levels in any of thegroups.

DISCUSSION

The current study reports on the underlying functionaland structural brain differences related to cognitive flex-ibility in HC, PrDs and AD patients. Unexpectedly, nogroup differences in cognitive flexibility were found, butthe results of this study confirmed the previouslyreported role of the dlPFC, vlPFC, ACC, posterior parietalcortex and basal ganglia in switching behavior (Sohnet al. 2000; Smith et al. 2004; Ravizza & Carter 2008).Consistent with our hypotheses, PrD and AD groupsrevealed compromised WM integrity, especially MD, andthis was associated with increased dlPFC activity duringtask switching. These findings are in line with the STACtheory, which states that higher prefrontal activation incognitive demanding tasks in the presence of neuraldamage reflects a compensatory mechanism. Our find-ings resonate with this compensatory hypothesis: PrDsshowed higher activations compared with the AD and HCgroups (PrD > AD > HC), whereas both AD patients andPrDs had significantly compromised WM integrity whencompared to controls (Goldin et al. 2008). In healthyaging, some studies have found that increased prefrontalactivation is accompanied by decreased activation inmore posterior parts of the brain; these results were notfound in this study (Davis et al. 2008). Possibly, the

difference in results reflects differences between damagerelated to both healthy aging and the neurotoxic effects ofalcohol. Based on the AUDIT scores, PrDs experiencedless problems in daily life compared with AD patients, andPrDs and AD patients may therefore be considered to rep-resent high and low functioning phenotypes, respectively.The current results suggest a compensatory mechanismin alcohol use disorders, in which additional prefrontalresources are recruited in order to compensate forimpaired WM integrity and neural functioning.

WM integrity, especially MD values, differed betweengroups and indicates a loss of WM integrity in PrDs andAD patients. FA values did not differ significantly betweengroups but were only related to task performance in HC.The absence of decreased FA values in AD compared withHCs is not in line with the previous research, which ishowever inconsistent, as both decrease (Pfefferbaum &Sullivan 2004; Pfefferbaum et al. 2009, 2010) andincrease (De Bellis et al. 2008; Cardenas et al. 2013) in FAvalues compared with HCs have been reported. The asso-ciation between WM integrity and cognitive flexibility is inline with a previous fiber tracking study on cognitive flex-ibility in HCs (van Schouwenburg et al. 2014).

The MD values negatively correlated with task perfor-mance only in PrDs, whereas MD values were not corre-lated with task performance in the AD group.The negativeassociation between lower MD values and higher taskperformance indicates that those PrDs with less damage toprefrontal WM show better switching performance. In ADpatients, no such relationship was apparent as task perfor-mance did not correlate with any of the DTI measure-ments. This might indicate that WM damage in ADpatients is too high to show an association between DTIand functional performance; however, a similar pattern ofWM damage and higher functional activity during theswitch task was found. These results are likely to present acompensatory mechanism in AD patients and PrDs wheredecreased WM integrity and related signal transfer isaccompanied by increased neural recruitment. However,as the current study only assessed one specific fiber tract, itis not possible to make inferences on other WM tracts.Future studies should assess whether neural scaffolding isrelated to additional brain areas and other WM tracts.

Table 4 Results for fiber tracking analysis for both left and right hemisphere. This table reports the results of the one-way repeatedmeasures ANCOVAs that were conducted for FA and MD values in both hemispheres. Analyses are corrected for age.

Left hemisphere (n = 55) Right hemisphere (n = 55)

Segment (repeated) Group (between subjects) Segment (repeated) Group (between subjects)

FA F(9) = 10.94, P < 0.001 F(2) = 0.96, P = ns F(9) = 4.17, P < 0.05 F(2) = 0.23, P = nsMD F(9) = 0.80, P = ns F(2) = 5.87, P < 0.01 F(9) = 6.79, P < 0.01 F(2) = 7.68, P < 0.001

FA = fractional anisotropy; MD = mean diffusivity; ns = non-significant.

Table 5 The results of the correlation analyses of the wholegroup. Results indicate that FA values are correlated with taskperformance. Significant results are displayed in bold.

Correlations

Percentagecorrect switches(n = 51)

RT switchcost(n = 47)

Percentageerror repeata

(n = 50)

Right FA segment 10 0.35* 0.17 −0.37*Left MD segment 10 0.01 −0.16 0.11

*Significant at P < 0.01, tested for two-tailed significance. aLog-transformed. FA = fractional anisotropy; MD = mean diffusivity;RT = reaction time.

Brain function and WM in AUD 7

© 2014 Society for the Study of Addiction Addiction Biology

These results should be viewed in light of some limi-tations. The non-significant difference between thegroups on task performance might be related to prac-tice effects, task design and/or sample sizes. First, all sub-jects practiced outside the scanner, which may haveobscured performance differences. Second, as the focusof our experiment was to assess the differential neural

involvement in cognitive flexibility, the current version ofthe task was adapted for efficiently measuring the BOLDresponse while at the same time minimizing the amountof trials to keep the paradigm as short as possible. Thesmall amount of switch trials may therefore have beenless suited to assess behavioral differences. Based onprevious studies on diminished cognitive flexibility in

Figure 1 Fractional anisotropy (FA) and mean diffusivity (MD) profiles for prefrontal white matter.The fiber tract is divided in ten segments,segment 1 is the most subcortical segment and segment 10 is the most prefrontal segment, as is depicted at the top of the figure. Differentcolors within the brain correspond to different segments. For the graphs, differently colored lines represent the different groups. MD valuesare in mm2/s. Error bars denote standard error of the mean

8 Jochem M. Jansen et al.

© 2014 Society for the Study of Addiction Addiction Biology

substance dependence (Hildebrandt et al. 2006; Pitelet al. 2009; van der Plas et al. 2009) and combined withour reported differences regarding neural involvementduring cognitive switching, we argue that these neuraldifferences are indicative of subtle impairments in cog-nitive flexibility in both AD subjects and PrDs. However,future studies should consider using a different switchparadigm [e.g. as was used in the work of de Ruiter et al.(2009)], which is powerful enough to detect behavioraldifferences to assess cognitive flexibility, and relate thoseresults to WM integrity. Furthermore, there was no rela-tionship between neural activity levels and DTI measure-ments. This would have underscored the results of thisstudy, but this non-significant result does not dismiss theproposed compensatory mechanism in AD patients andPrDs. Although the non-significant result could be anindication that activity is unrelated to WM integrity, itmay also reflect some methodological issue. The fibertracking was based on the whole dlPFC ROI, whereas wecorrelated the DTI measurements with the peak activa-tion voxel and the location of the peak voxel was slightlydifferent for the three groups. As mentioned in theMaterials and Methods section, the PrDs and AD patientsdiffered in their period of abstinence based on the inclu-sion criteria of the study. We could therefore onlyperform correlation analyses for the AD group.Although abstinence did not have any relationship withtask performance, fMRI or DTI measurements, the avail-able data were not very specific and future studies shoulduse a more precise way of determining abstinence dura-tion. Furthermore, the AD patients all received cognitivebehavioral therapy treatment, which may have had aneffect on the brain function (Dichter, Felder & Smoski2010; Zatorre, Fields & Johansen-Berg 2012). This mayhave influenced the results of the study, but unfortu-nately, data on the exact period of treatment were notcollected. Finally, sample sizes, although adequate fordetecting BOLD differences, may have limited the possi-bility to detect significant differences in behavioralperformance.

CONCLUSION

The current study reports on the underlying functionaland structural brain differences associated with cogni-tive flexibility between HCs, PrDs and AD patients. Evi-dence is reported for preserved cognitive flexibility inPrDs and AD patients by recruiting more prefrontalresources in order to compensate for reduced WMintegrity. Furthermore, PrDs showed more recruitmentof compensatory prefrontal resources compared withAD patients. This compensatory mechanism couldindicate a protective mechanism against a decline incognitive functioning and/or transition into compulsivedrug use, but future research should confirm thesehypotheses.

Acknowledgement

This research was funded by The European Foundationfor Alcohol Research (ERAB), grant no. EA 10 27“Changing the vulnerable brain: a neuromodulationstudy in alcohol dependence” to AEG, WB and DJ.

Authors Contribution

Authors RJvH and JMJ have contributed equally to thispaper and should be referred to as joint first authors. RvJ,AG and WvdB designed the study. RvJ was responsible forthe data collection. RvJ and JJ were responsible for thedata analysis. DV provided assistance for the fMRI analy-sis, whereas MC did so for the DTI data analysis. Then JJinterpreted the results and wrote the paper. All authorshave critically reviewed content and approved finalversion submitted for publication.

References

Ashburner J (2007) A fast diffeomorphic image registrationalgorithm. Neuroimage 38:95–113.

Badre D, Wagner AD (2006) Computational and neurobiologicalmechanisms underlying cognitive flexibility. Proc Natl AcadSci U S A 103:7186–7191.

Table 6 The results of the correlation analysis per group. Results indicate that the FA values correlate with task performance only inthe control group, whereas MD values only correlate with task performance in the problematic drinkers. Significant results aredisplayed in bold.

Correlations

Percentage correct switches RT switch cost Percentage error repeata

HC(n = 15)

PrD(n = 13)

AD(n = 17)

HC(n = 13)

PrD(n = 12)

AD(n = 16)

HC(n = 14)

PrD(n = 12)

AD(n = 17)

Right FA segment 10 0.65** 0.42 0.20 −0.32 0.22 0.14 −0.66** −0.36 −0.22Left MD segment 10 0.37 −0.53* −0.03 −0.03 −0.26 0.09 −0.22 0.37 0.18

*Significant at P < 0.05, tested for two-tailed significance. **Significant at P < 0.01, tested for two-tailed significance. aLog-transformed. AD = alcohol-dependent patient; FA = fractional anisotropy; HC = healthy control; MD = mean diffusivity; PrD = problematic drinker; RT = reaction time.

Brain function and WM in AUD 9

© 2014 Society for the Study of Addiction Addiction Biology

Baler RD, Volkow ND (2006) Drug addiction: the neurobiology ifdisrupted self-control. Trends Mol Med 12:559–566.

Beaulieu C (2002) The basis of anisotropic water diffusion in thenervous system—a technical review. NMR Biomed 15:435–455.

Behrens TEJ, Berg HJ, Jbabdi S, Rushworth MFS, WoolrichMW (2007) Probabilistic diffusion tractography with multiplefibre orientations: what can we gain? Neuroimage 34:144–155.

Bennett IJ, Madden DJ, Vaidya CJ, Howard JH, Jr., Howard DV(2011) White matter integrity correlates of implicit sequencelearning in healthy aging. Neurobiol Aging 32:2317.e1–2317.e12.

Brett M, Anton JL, Valabregue R, Poline JB (2002) Region ofinterest analysis using the MarsBar toolbox for SPM 99.Neuroimage 16:S497.

Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA(1998) The AUDIT alcohol consumption questions (AUDIT-C):an effective brief screening test for problem drinking. Ambu-latory Care Quality Improvement Project (ACQUIP). AlcoholUse Disorders Identification Test. Arch Intern Med 158:1789–1795.

Caan MW, Khedoe G, Poot D, den Dekker A, Olabarriaga S,Grimbergen K, van Vliet L, Vos F (2010) Adaptive noisefiltering for accurate and precise diffusion estimation in fibercrossings. Med Image Comput Comput Assist Interv 13 (Pt1):167–174.

Cardenas VA, Greenstein D, Fouche JP, Ferrett H, Cuzen N, SteinDJ, Fein G (2013) Not lesser but greater fractional anisotropyin adolescents with alcohol use disorders. Neuroimage Clin2:804–809.

Chanraud S, Reynaud M, Wessa M, Penttila J, Kostogianni N,Cachia A, Artiges E, Delain F, Perrin M, Aubin HJ, Cointepas Y,Martelli C, Martinot JL (2009) Diffusion tensor tractography inmesencephalic bundles: relation to mental flexibility in detoxi-fied alcohol-dependent subjects. Neuropsychopharmacology34:1223–1232.

Charlet K, Beck A, Jorde A, Wimmer L, Vollstädt-Klein S, GallinatJ, Walter H, Kiefer F, Heinz A (2014) Increased neural activityduring high working memory load predicts low relapse risk inalcohol dependence. Addict Biol 19:402–414.

Charlet K, Schlagenhauf F, Richter A, Naundorf K, Dornhof L,Weinfurtner CE, König F, Walaszek B, Schubert F, Müller CA,Gutwinski S, Seissinger A, Schmitz L, Walter H, Beck, A,Gallinat J, Kiefer F, Heinz A (2014) Neural activation duringprocessing of aversive faces predicts treatment outcome inalcoholism. Addict Biol 19:439–451.

Cooper JA, Worthy DA, Maddox WT (2013) Scaffolding acrossthe lifespan in history-dependent decision making. PsycholAging 28:505–514.

Davis SW, Dennis NA, Daselaar SM, Fleck MS, Cabeza R (2008)Que PASA? The posterior-anterior shift in aging. Cereb Cortex18:1201–1209.

De Bellis MD, Van Voorhees E, Hooper SR, Gibler N, Nelson L,Hege SG, MacFall J (2008) Diffusion tensor measures of thecorpus callosum in adolescents with adolescent onset alcoholuse disorders. Alcohol Clin Exp Res 32:395–404.

de Ruiter MB, Veltman DJ, Goudriaan AE, Oosterlaan J, Sjoerds Z,van den Brink W (2009). Response perseveration and ventralprefrontal sensitivity to reward and punishment in maleproblem gamblers and smokers. Neuropsychopharmacology34:1027–1038.

Dichter GS, Felder JN, Smoski MJ (2010) The effects of briefbehavioral activation therapy for depression on cognitive

control in affective contexts: an fMRI investigation. J AffectDisord 126:236–244.

Everitt BJ, Robbins TW (2005) Neural systems of reinforcementfor drug addiction: from actions to habits to compulsion. NatNeurosci 8:1481–1489.

Fox PT, Lancaster JL (1994) Neuroscience on the net. Science266:994–996.

Friston KJ, Holmes A, Poline JB, Price CJ, Frith CD (1996) Detect-ing activations in PET and fMRI: levels of inference and power.Neuroimage 4 (3 Pt 1):223–235.

Goh JO, Park DC (2009) Neuroplasticity and cognitive aging: thescaffolding theory of aging and cognition. Restor NeurolNeurosci 27:391–403.

Goldin PR, McRae K, Ramel W, Gross JJ (2008) The neural basesof emotion regulation: reappraisal and suppression of nega-tive emotion. Biol Psychiatry 63:577–586.

Goldstein RZ, Volkow ND (2011) Dysfunction of the prefrontalcortex in addiction: neuroimaging findings and clinical impli-cations. Nat Rev Neurosci 12:652–669.

Hagmann P, Jonasson L, Maeder P, Thiran JP, Wedeen VJ, MeuliR (2006) Understanding diffusion MR imaging techniques:from scalar diffusion-weighted imaging to diffusion tensorimaging and beyond. Radiographics 26 (Suppl. 1):s205–s223.

Harper C (2007) The neurotoxicity of alcohol. Hum Exp Toxicol26:251–257.

Harper C (2009) The neuropathology of alcohol related braindamage. Alcohol Alcohol 44:136–140.

Harris GJ, Jaffin SK, Hodge SM, Kennedy D, Caviness VS,Marinkovic K, Papadimitriou GM, Makris N, Oscar-Berman M(2008) Frontal white matter and cingulum diffusion tensorimaging deficits in alcoholism. Alcohol Clin Exp Res 32:1001–1013.

Hildebrandt H, Brokate B, Hoffmann E, Kröger B, Eling PATM(2006) Conditional responding is impaired in chronic alcohol-ics. J Clin Exp Neuropsychol 28:631–645.

Koob GF, Volkow ND (2010) Neurocircuitry of addiction.Neuropsychopharmacology 35:217–238.

Kril JJ, Halliday GM, Svoboda MD, Cartwright H (1997) Thecerebral cortex is damaged in chronic alcoholics. Neurosci-ence 79:983–998.

Lezak MD (2004) The behavioral geography of the brain. In:Lezak MD, ed. Neuropsychological Assessment, 4th edn, pp.39–85. New York: Oxford University Press.

Maldjian JA, Laurienti PJ, Burdette JB, Kaft RA (2003) An auto-mated method for neuroanatomic and cytoarchitectonicatlas-based interrogation of fMRI data sets. Neuroimage19:1233–1239.

Moore A, Malinowski P (2009) Meditation, mindfulness andcognitive flexibility. Conscious Cogn 18:176–186.

Nyhus E, Barcelo F (2009) The Wisconsin Card Sorting Test andthe cognitive assessment of prefrontal executive functions: acritical update. Brain Cogn 71:437–451.

Olabarriaga SD, Glatard T, de Boer PT. (2010) A virtual labora-tory for medical image analysis. IEEE Trans Inf TechnolBiomed 14:979–985.

Park DC, Reuter-Lorenz P (2009) The adaptive brain: agingand neurocognitive scaffolding. Annu Rev Psychol 60:173–196.

Pfefferbaum A, Sullivan EV (2004) Disruption of brain whitematter microstructure by excessive intracellular andextracellular fluid in alcoholism: evidence from diffusiontensor imaging. Neuropsychopharmacology 30:423–432.

Pfefferbaum A, Rosenbloom M, Rohlfing T, Sullivan EV (2009)Degradation of association and projection white matter

10 Jochem M. Jansen et al.

© 2014 Society for the Study of Addiction Addiction Biology

systems in alcoholism detected with quantitative fiber track-ing. Biol Psychiatry 65:680–690.

Pfefferbaum A, Rosenbloom MJ, Fama R, Sassoon SA, SullivanEV (2010) Transcallosal white matter degradation detectedwith quantitative fiber tracking in alcoholic men and women:selective relations to dissociable functions. Alcohol Clin ExpRes 34:1201–1211.

Pitel AL, Rivier J, Beaunieux H, Vabret F, Desgranges B, EustacheF (2009) Changes in the episodic memory and executive func-tions of abstinent and relapsed alcoholics over a 6 monthperiod. Alcohol Clin Exp Res 33:490–498.

Ravizza SM, Carter CS (2008) Shifting set about task switching:behavioral and neural evidence for distinct forms of cognitiveflexibility. Neuropsychologia 46:2924–2935.

Schmand B, Bakker D, Saan R, Louman J (1991) [The DutchReading Test for Adults: a measure of premorbid intelligencelevel]. Tijdschr Gerontol Geriatr 22:15–19.

Schulte T, Müller-Oehring EM, Pfefferbaum A, Sullivan EV(2010) Neurocircuitry of emotion and cognition in alcohol-ism: contributions from white matter fiber tractography. Dia-logues Clin Neurosci 12:554–560.

Schulte T, Müller-Oehring EM, Sullivan EV, Pfefferbaum A(2012) White matter fiber compromise contributes differen-tially to attention and emotion processing impairment inalcoholism, HIV-infection, and their comorbidity. Neuropsy-chologia 50:2802–2822.

Shahand S, Santcroos M, Mohammed Y, Korkhov V, Luyf ACM,van Kampen A, Olabarriaga SD (2011) Front-ends to biomedicaldata analysis on grids. Proceedings of the HealthGrid Confer-ence. Bristol, UK.

Smith AB, Taylor E, Brammer M, Rubia K (2004) Neural corre-lates of switching set as measured in fast, event-related func-tional magnetic resonance imaging. Hum Brain Mapp21:247–256.

Sohn MH, Ursu S, Anderson JR, Stenger VA, Carter CS (2000)Inaugural article: the role of prefrontal cortex and posteriorparietal cortex in task switching. Proc Natl Acad Sci U S A97:13448–13453.

Stalnaker TA, Takahashi Y, Roesch MR, Schoenbaum G (2009)Neural substrates of cognitive inflexibility after chroniccocaine exposure. Neuropharmacology 56 (Suppl. 1):63–72.

Starcke K, Holst RJ, van den Brink W, Veltman DJ, Goudriaan AE(2013) Physiological and endocrine reactions to psychosocialstress in alcohol use disorders: duration of abstinence matters.Alcohol Clin Exp Res 37:1343–1350.

Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F,Etard O, Delcroix N (2002) Automated anatomical labelingof activations in SPM using a macroscopic anatomicalparcellation of the MNI MRI single-subject brain. Neuroimage15:273–289.

van der Plas EA, Crone EA, van den Wildenberg WP, Tranel D,Bechara A (2009) Executive control deficits in substance-

dependent individuals: a comparison of alcohol, cocaine, andmethamphetamine and of men and women. J Clin ExpNeuropsychol 31:706–719.

van Holst RJ, van den Brink W, Veltman DJ, Goudriaan AE(2010) Why gamblers fail to win: a review of cognitive andneuroimaging findings in pathological gambling. NeurosciBiobehav Rev 34:87–107.

van Schouwenburg MR, O’Shea J, Mars RB, Rushworth MF,Cools R (2012) Controlling human striatal cognitive functionvia the frontal cortex. J Neurosci 32:5631–5637.

van Schouwenburg MR, Onnink AMH, ter Huurne N, Kan CC,Zwiers MP, Hoogman M, Franke B, Buitelaar JK, Cools R(2014) Cognitive flexibility depends on white matter micro-structure of the basal ganglia. Neuropsychologia 53:171–177.

Vanes LD, van Holst RJ, Jansen JM, van den Brink W, OosterlaanJ, Goudriaan AE (2014) Contingency learning in alcoholdependence and pathological gambling: learning andunlearning reward contingencies. Alcohol Clin Exp Res38:1602–1610.

Volkow ND, Fowler JS, Wang GJ, Goldstein RZ (2002) Role ofdopamine, the frontal cortex and memory circuits in drugaddiction: insight from imaging studies. Neurobiol Learn Mem78:610–624.

Whitfield-Gabrieli S, Nieto-Castanon A (2012) Conn: a func-tional connectivity toolbox for correlated and anticorrelatedbrain networks. Brain Connect 2:125–141.

World Health Organization (1997) Composite InternationalDiagnostic Interview—Version 2.1. Geneva: World HealthOrganization.

Yeh PH, Simpson K, Durazzo TC, Gazdzinski S, Meyerhoff DJ(2009) Tract-Based Spatial Statistics (TBSS) of diffusiontensor imaging data in alcohol dependence: abnormalities ofthe motivational neurocircuitry. Psychiatry Res 173:22–30.

Yucel M, Lubman DI (2007) Neurocognitive and neuroimagingevidence of behavioural dysregulation in human drug addic-tion: implications for diagnosis, treatment and prevention.Drug Alcohol Rev 26:33–39.

Zatorre RJ, Fields RD, Johansen-Berg H (2012) Plasticity in grayand white: neuroimaging changes in brain structure duringlearning. Nat Neurosci 15:528–536.

SUPPORTING INFORMATION

Additional Supporting Information may be found in theonline version of this article at the publisher’s web-site:

Appendix S1 The main effect of task switch versus taskrepeat for the whole groupAppendix S2 Mean FA and MD values per group andhemisphere

Brain function and WM in AUD 11

© 2014 Society for the Study of Addiction Addiction Biology