cannabis use and memory brain function in adolescent boys: a cross-sectional multicenter functional...

NEW RESEARCH

Cannabis Use and Memory Brain Functionin Adolescent Boys: A Cross-Sectional

Multicenter Functional Magnetic ResonanceImaging Study

Gerry Jager, Ph.D., Robert I. Block, Ph.D., Maartje Luijten, M.Sc.,Nick F. Ramsey, Ph.D.

Objective: Early-onset cannabis use has been associated with later use/abuse, mental healthproblems (psychosis, depression), and abnormal development of cognition and brain function.During adolescence, ongoing neurodevelopmental maturation and experience shape the neuralcircuitry underlying complex cognitive functions such as memory and executive control. Prefron-tal and temporal regions are critically involved in these functions. Maturational processes leavethese brain areas prone to the potentially harmful effects of cannabis use. Method: Weperformed a two-site (United States and the Netherlands; pooled data) functional magneticresonance imaging (MRI) study with a cross-sectional design, investigating the effects of adoles-cent cannabis use on working memory (WM) and associative memory (AM) brain function in 21abstinent but frequent cannabis–using boys (13–19) years of age and compared them with 24nonusing peers. Brain activity during WM was assessed before and after rule-based learning(automatization). AM was assessed using a pictorial hippocampal-dependent memory task. Re-sults: Cannabis users performed normally on both memory tasks. During WM assessment,cannabis users showed excessive activity in prefrontal regions when a task was novel, whereasautomatization of the task reduced activity to the same level in users and controls. No effectof cannabis use on AM-related brain function was found. Conclusions: In adolescentcannabis users, the WM system was overactive during a novel task, suggesting functionalcompensation. Inefficient WM recruitment was not related to a failure in automatization butbecame evident when processing continuously changing information. The results seem toconfirm the vulnerability of still developing frontal lobe functioning for early-onset cannabisuse. J. Am. Acad. Child Adolesc. Psychiatry, 2010;49(6):561–572. Key Words: cannabis,adolescence, early-onset, fMRI, memory

abi

cl(elsnfs

E arly initiation of cannabis use increases therisk of later use/abuse of other drugs anddrug dependence, and is associated with

mental health problems such as psychosis anddepression. The strength of this association appearsto be dependent on the age when cannabis usebegins.1 A major concern that has only recentlygained attention is the effect of early-onset cannabisuse on adolescent brain function and neurodevel-opment.

Supplemental material cited in this article is available online.

JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY

VOLUME 49 NUMBER 6 JUNE 2010

The still-developing adolescent brain differsnatomically and neurochemically from the adultrain2,3 and is likely more susceptible to drug-

nduced adaptive neuronal plasticity.Animal studies on the neural consequences of

hronic cannabis exposure during the peri-ado-escent period report changes in brain structurepredominantly limbic brain regions) and alteredmotional and cognitive performance in laterife.4 However, these effects were mostly ob-erved at relatively high doses of synthetic can-abinoids (Win 55,212-2; CP 55,940) and there-ore may not be comparable to the humanituation.

Studies in cannabis-using human adolescents

561www.jaacap.org

MTMUt

PIiwowimcal2pctgppbddSc

PSbe

T

JAGER et al.

are still limited, but a number of functionalmagnetic resonance imaging (fMRI) studies havelinked adolescent cannabis use to increased pari-etal activation along with diminished prefrontalactivation during spatial working memory, andto increased parietal and prefrontal activationduring inhibition, indicating reorganization ofneural networks and recruitment of additionalneural resources,5,6 as previously reviewed.7-9

Throughout early and late adolescence, matura-tional processes occur at different rates in variousbrain regions. Maturation is slightly delayed inthe temporal and prefrontal cortex, regions criti-cally involved in memory and learning and cog-nitive control.2,10 The aim of the present studywas to investigate the effects of adolescent regu-lar cannabis use on memory-related brain func-tion, with a focus on temporal and prefrontalregions. As there is evidence for gender differ-ences in the rate and timing of neurodevelop-ment11 and neurodevelopmental responses tocannabinoid exposure,12 only boys were in-cluded in this study. We performed Blood-Oxy-gen-Level-Dependent (BOLD) fMRI using twotasks that reliably engage prefrontal and/or tem-poral brain regions, i.e., a verbal working mem-ory (WM) task and a pictorial associative mem-ory (AM) task, and compared regularlycannabis-using boys with age-matched nonusingpeers. Previous fMRI studies from our laboratoryin adult cannabis users and nonusing controlsapplied the same task paradigms.13,14 These stud-ies yielded no effects on task performance. How-ever, with regard to brain activity, adult cannabisusers displayed subtle alterations in superiorparietal brain activity patterns during WM13 andoverall hypoactivity in a network of prefrontaland parahippocampal areas during AM.14 In thepresent study, therefore, we would not expect aperformance deficit, as the adults (having usedfor several years, i.e., longer than the currentadolescent users) did not exhibit a deficit. Forbrain activity, we anticipated two possible out-comes. For one, effects of cannabis on brainactivity could mimic those in the adults, i.e.,altered parietal brain activity during WM andreduced activity during AM, suggesting age-independent effects of cannabis use. Alterna-tively, brain activity findings could differ fromthose found in adult users. In this case, wehypothesized hyperactivation to compensate forcannabis-induced dysfunction, most prominently
in parahippocampal and prefrontal brain areas. d
JOURN

562 www.jaacap.org

ETHODhis study was a joint venture of the Universityedical Center Utrecht (the Netherlands) and theniversity of Iowa (United States). Data were pooled

o increase statistical power.

articipantsn total, 47 boys, aged 13 through 19 years, were includedn the study; 23 regular cannabis users (11 Dutch, 12 US)

ith at least 200 lifetime episodes of cannabis use, thethers nonusing controls (12 Dutch, 12 US). Eligibilityas ascertained through a screening procedure involv-

ng questionnaires on drug use, medical history, andental health. Table 1 reports inclusion and exclusion

riteria. Written consent was obtained from adolescentsnd their parent/legal guardian in accordance with theocal IRB and conformed to the Helsinki Declaration of004. Adolescents were informed in advance that aarent/legal guardian had to provide signed informedonsent and would be informed by the researchers abouthe group the adolescent was in (e.g., user/nonuserroup) without revealing any details on history and/orattern of cannabis use or other substances. Hence,arents/guardians knew in which group their son was,ut we did not share information with them on theetails of cannabis use, to promote the integrity of therug use information we obtained from the youngsters.ubjects were reimbursed for participation with giftertificates.

rocedureubjects participated in two separate sessions separatedy 1 week. The first involved screening for inclusion andxclusion criteria using self-report questionnaires on

ABLE 1 Study Inclusion and Exclusion Criteria

Inclusion Criteria Exclusion Criteria

Male sex Medical or neurological problemsAge 12–19

yearsRegular use of illegal drugs other

than cannabis (�10 episodeslifetime), except alcohol ornicotine

Right-handedness Axis I psychiatric diagnosis, exceptfor conduct disorder, which is acommon diagnosis in cannabis-using boys

IQ scores �80Use of psychotropic medicationContraindications for MRI

(claustrophobia, metal objects,full dental braces)

Note: MRI � magnetic resonance imaging.

rug use history and a semistructured psychiatric inter-

AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY


tAct(2o

i(mmbcpttpm(t

fI(tdsanahsfl

Fmt“cp

TEENAGE CANNABIS USE AND BRAIN FUNCTION

view (NIMH Diagnostic Interview Schedule for Chil-dren; C-DISC–version IV). To estimate cannabis useparameters, a semi-structured self-report questionnairewas used, containing questions on the number of timessubjects used cannabis during the last 3 months, the lastyear, and (if relevant) previous years, as well as thenumber of joints/blunts smoked per occasion. Lifetimenumber of cannabis use episodes was estimated throughextrapolation. In addition, we estimated the number ofjoints smoked the previous year (number of episodes �average number of joints used per occasion) and numberof joints lifetime, where we took into account periods ofreported more or less frequent use.

To estimate IQ, subjects performed four subtests (sim-ilarities, block design, vocabulary and matrix reasoning)of the Wechsler Intelligence Scale for Children–4th edi-tion (WISC-IV). The second session, 1 week after the first,included neuropsychological testing (results to be re-ported elsewhere) and fMRI scanning. Participants ab-stained from cannabis, alcohol, and other substances forat least 24 hours before the first session and remainedabstinent until completion of the second session. Smok-ing was allowed until 2 hours before the scanning session(to avoid nicotine withdrawal). On both sessions, urinesamples were collected for drug screening (enzyme-multiplied immunoassay for cannabis, alcohol, amphet-amines, ecstasy, opiates, cocaine, and benzodiazepines).Exclusion followed on positive testing on any psychoac-tive substance on the second test day, except for canna-bis, which can linger in the body for several weeksbecause of its lipophilic properties and thus can induce apositive test result even after 1 week of abstinence.15

Instead, cannabis-using subjects were excluded onlywhen their urine toxicology test failed to show a decreasein quantified levels (�g/L) of cannabinoid metabolites(THCCOOH) between the first and second urine sampletaken. Based on the laboratory results, two subjects hadto be excluded from analyses. Two other subjects had apositive urine test on cannabinoids during the secondsession, but THCOOH-levels were decreased comparedwith the first measurement 1 week earlier, which wasconsistent with self-reported abstinence. All other sub-jects had negative urine tests on the day of testing.

Assessment of WM and AMTwo fMRI tasks were administered: a verbal WM task(Figure 1) based on Sternberg’s item-recognition para-digm (denoted STERN), and a pictorial AM task (de-noted PMT; Figure 2). Both tasks have previously beendescribed in detail.13,14 STERN assesses the WM systembefore and after practice (automatization). Subjects wereinstructed to memorize a set of five letters (memory set)and subsequently to respond to single letters (probes) bypressing a button if the probe was in the memory set(target). A novel (NT) and a practiced task (PT) wereadministered. In PT, a fixed memory set was usedrepeatedly, on which subjects were trained before scan-
ning to induce automatization. In the NT, the composi- a


ion of the memory set was changed after every epoch.n additional reaction time control task (CT) was in-

luded during which subjects made a button press whenhe symbol “� �” appeared. In the scanner, each taskCT, PT, and NT) was presented in six epochs (duration,9 seconds) of 10 stimuli each, as well as six rest periodsf equal epoch duration.

PMT assesses (para)hippocampal-dependent AM andnvolves three tasks. First, an associative learning taskAL) is performed that requires subjects to establish a

eaningful connection between two pictures and toemorize the combination. Next, single pictures have to

e classified (SC), which serves as a control task; i.e.,ompared with AL it requires the same amount oferceptual processing and a motor response, but it lacks

he associative learning component. Finally, the retrievalask (RE) asks subjects to recognize specific combinationsreviously presented during AL, and provides a perfor-ance measure. Each task was presented in three epochs

duration, 65 seconds) of eight stimuli each, as well ashree rest periods of equal duration.

MRI acquisitionmaging was performed using a clinical 3 T MRI scannerPhilips Achieva [the Netherlands] and Siemens Magne-on Trio [US]), both with an eight-channel head coil. Pilotata on various T1 and EPI scan sequences from threeubjects (i.e., researchers G.J. and N.F.R. and a researchssistant, J.Z.) were tested for homogeneity of signal-to-oise (SNR) and temporal signal-to-noise (tSNR) ratioscross sites and vendors. The scan parameters yieldingighest similarity between SNR and tSNR maps acrosscanners were used for both tasks: TE/TR 35/2000 ms,ip angle 70°, field of view (FOV) 256 � 256 mm,

IGURE 1 Temporal sequence of events for the workingemory task. Note: Each epoch starts with presentation of

he memory set (a set of five consonants, for example,FGMPT”), and is followed by 10 trials showing a singleonsonant. Subjects have to press a button as fast asossible if the letter belongs to the memory set.

cquisition matrix 64 � 64, slice thickness 3.6 (plus a

563www.jaacap.org

emTtmwaavnfcfwrr8mtPstfaawoi

JAGER et al.

0.4-mm gap), voxel size 4.0-mm isotropic, 26 slices, scanorientation transaxial for STERN and parallel to the longaxis of the hippocampus for PMT. Details on the issue ofmultisite studies and scanner compatibility are presentedin Supplement 1 and Figure S1 (available online). ForSTERN, a single run of 312 scans was acquired over aperiod of 10.4 minutes. For PMT, a single run of 324 scanslasted 10.8 minutes. In addition, a volumetric T1-weighted MR anatomical scan was acquired for spatiallocalization (TR 25 ms, flip angle 30°, FOV read 256 mm,voxel size 1.0 mm isotropic, 176 slices, scan duration 7.8minutes).

Statistical AnalysisSample Characteristics and Drug Use. Cannabis pa-rameters were as follows: age of onset, cumulative uselifetime (estimated number of joints), current or recentuse (number of joints last year), frequency and durationof use, and abstinence (weeks since last use). Additionaldrug use data included last year use of alcohol (averagenumber of drinks/week), tobacco (average number ofcigarettes/week), and lifetime use of other illegal sub-stances (number of occasions lifetime). Because distribu-tion of the drug variables was skewed, scores werelog-transformed, except for age of onset of cannabis,which was normally distributed. Other variables in-cluded age, estimated IQ, country (US versus NL), and a

FIGURE 2 Temporal sequence of events for the associainstruction slide (5 seconds) followed by a fixation cross (each (picture pair 5 seconds, fixation cross 2.5 seconds).to the instruction in each task condition.

diagnosis of conduct disorder (yes/no). Group differ- u

JOURN

564 www.jaacap.org

nces were tested using t tests and nonparametric Kol-ogorow-Smirnov Z tests.ask Performance. Outcome measures included reac-

ion times (STERN only) and accuracy. General linearodel (GLM) repeated measures analysis was applied,ith task condition (CT, PT, and NT for STERN; AL, SC,

nd RE for PMT) and outcome measure (reaction timend accuracy) as within-subject factor, and group (userersus control) as between-subject factor. To adjust forormal developmental effects, i.e., younger boys per-orming relatively worse than older ones, age was in-luded as a covariate.MRI. Imaging data were analyzed using SPM5 (http://

ww.fil.ion.ucl.ac.uk/spm). Pre-processing includedealignment (motion correction) and unwarping, co-egistration, normalization, and smoothing with an-mm (FWHM) Gaussian kernel. First, statistical activityaps were generated for each subject for all task condi-

ions (CT, PT, and NT for STERN; AL, SC, and RE forMT) by analyzing time series data with multiple regres-ion analyses using a vector representing the design of aask and including cosine basis functions to remove lowrequency drifts in the signal (details on preprocessingnd first-level statistical analysis in Supplement 1, avail-ble online). Next, for STERN, individual contrast mapsere created for NT and PT, corrected for individual

ffset activity levels (i.e., subtracting activity levels dur-ng the control task [CT], representing baseline activity

emory task. Note: Each epoch starts with aneconds). This is followed by eight trials of 7.5 secondsects respond by pressing one of two buttons, according

tive m2.5 sSubj

nrelated to WM processing, from activity levels during



http://www.fil.ion.ucl.ac.uk/spm


vaagbaPmmcUcsdCsArftlealtsrymwob

RTd

uperio


NT and PT, respectively). Similarly, for PMT, we createdindividual activity maps for AL and RE, corrected forindividual offset activity levels during the control condi-tion, contrasting both AL and RE with SC. Contrast mapswere then used in a second-level whole-brain analysis totest for effects of cannabis use on brain activity. Age anda dichotomized variable for country (0 for the Nether-lands; 1 for the United States) were added as covariates,to take into account potential systematic effects of ageand/or differences in MRI scanners across sites. In addi-tion to whole-brain analysis, region-of-interest (ROI)analyses were performed. To check whether users andnonusers activated similar networks of brain regionsduring WM (as expected, as there is no a priori reason toassume that cannabis users show significant functionalreorganization of the brain), we defined ROIs for bothgroups separately (group contrast maps; NT-CT forSTERN and AL-SC for PMT, p � .05, FWE corrected).Visual inspection of the group activation maps indicatedno differential activation patterns between users andcontrols. Therefore, ROIs used for further analyses (Table2) were derived from group contrast maps (NT-CT forSTERN and AL-SC for PMT) of all subjects combined(p � .05, FWE corrected). For STERN (Figure 3), thisyielded activated regions in the left superior parietalcortex (l-SPC), left inferior frontal gyrus (l-IFG), leftprecentral and dorsolateral prefrontal cortex (l-PCC/DLPFC), and anterior cingulate cortex (ACC). For PMT,activated regions included bilateral regions in the para-hippocampal gyrus, middle occipital gyrus, and prefron-tal areas (Figure 4). ROIs for both tasks involved regionsknown from previous studies from our laboratory, usingthe same task paradigms,13,14,16 and hence, met a prioriexpectations. ROIs were marked and activity values forall subjects were obtained per ROI by averaging �-valuesacross all contained voxels for NT and PT, and for ALand RE (corrected for individual offset activity levels),

TABLE 2 Regions of Interest during Working Memory an

Task Region Brodmann area N

STERN ACC 6/24l-PCC/DLPFC 9/46l-IFG 47l-SPC 39

PMT r-PHG/MOG 36/37/19l-PHG 36/37r-DLPFC 9/46ACC 6/24l-DLPFC 46l-MOG 19

Note: Montréal Neurological Institute (MNI) coordinates for the regions oand associative memory (pictorial memory task [PMT]). The coordinatesmap. ACC � anterior cingulate cortex; IFG � inferior frontal gyrus; l �

prefrontal cortex; PHG � parahippocampal gyrus; r � right; SPC � s

using the Marsbar toolbox in SPM5.17 Finally, these f



ariables were entered into GLM repeated-measuresnalyses with task conditions (NT and PT for STERN, ALnd RE for PMT) and ROIs as within-subject factors, androup as between-subject factor. Similar to the whole-rain analyses, country and age were entered as covari-tes in all ROI analyses.otential Confounders. Nine of 12 US cannabis userset criteria for a diagnosis of conduct disorder, whichight act as a confounding factor. However, as the

o-occurrence of conduct disorder was restricted to theS users, it was strongly correlated with the factor

ountry (which was included as a covariate in all analy-es, together with age), and the effects of conduct disor-er could not be disentangled from those of country.annabis use in both US and Dutch users correlated

ignificantly with lower IQ, alcohol use, and tobacco use.ny significant group difference found on output pa-

ameters (task performance, brain activity) resultingrom the main analyses (see previous paragraph) washerefore considered rather an effect of a “cannabis-usingifestyle” than an effect of cannabis use alone. Yet, toxplore the relative strength of the effect of cannabislone, output parameters were recomputed after control-ing for confounding effects of IQ, use of alcohol andobacco by means of multiple regressions, i.e., saving thetandardized residuals. In a subsequent analysis, we thene-entered these standardized residuals into group anal-ses (analysis of variance [ANOVA], GLM repeatedeasures). This constitutes a conservative approachhereby interactions between effects of cannabis and

ther effects are prevented from leaking into the “canna-is” effect.

ESULTSwo users were excluded based on positive urinerug test results, and STERN fMRI data were lost

sociative Memory Tasks

voxels X Y Z T-max

77 �4 12 48 9.5548 �40 0 36 6.2135 �36 24 0 6.13

6 �28 �60 36 4.7310 36 �52 �20 13.4939 �36 �48 �20 12.8043 48 28 20 8.7420 4 20 44 7.6817 �44 20 20 7.1817 �24 �92 4 7.60

st involved in working memory (Sternberg item-recognition task [STERN])and Z represent location of the voxels with the highest t value in the groupMOG � middle occipital gyrus; PCC/DLPFC � precentral/dorsolateralr parietal cortex.

d As

o. of

31

f intereX, Y,left;

or one control because of technical malfunction.

565www.jaacap.org

o[cspdcajD1tw

JAGER et al.

Results are reported for 45 participants (21 users, 24controls) for sample characteristics, drug use, andPMT data, and for 44 participants (21 users, 23controls) for STERN.

Sample Characteristics and Drug UseTable 3 summarizes sample characteristics anddrug use for users (US and Dutch) and nonusers(US and Dutch). Users were, on average, absti-nent for 5.1 weeks (�4.2). Groups did not differin age, but users had significantly lower IQ-scores than controls (p � .01), and reportedsignificantly higher previous-year consumption

FIGURE 3 Regions of interest for the working memorydorsolateral prefrontal cortex. (C) Left superior parietal cobased on the contrast between novel and control task (p �Montreal Neurological Institute system. Slices are in neuro

FIGURE 4 Regions of interest (ROI) for the associative(para)hippocampal gyrus (PHG). (C) Left middle occipitalcortex. (F) Anterior cingulate cortex. ROIs are based on c.001). Numbers above slices indicate z coordinates of theneurological orientation (i.e., left side is left hemisphere).

JOURN

566 www.jaacap.org

f alcohol (mean alcoholic drinks per week 13.3� 13.6] for users compared to 3.4 [� 5.8] forontrols) and tobacco (mean number of cigarettesmoked per week 63.8 [�53.4] for users com-ared with 6.1 [�14.9] for controls). In addition,emographic and cannabis use parameters wereompared for the user and nonuser groups sep-rately, comparing US subjects with Dutch sub-

ects, to identify potentially site related biases.utch users were older than US users (mean age

7.9 years [�� 0.9] and 16.7 [�� 0.8] respec-ively, p � .05), but no significant differences

ere found on IQ scores or cannabis use param-

Note: (A) Left inferior frontal gyrus. (B) Left precentral/(D) Anterior cingulate cortex. Regions of interest are1). Numbers above slices indicate z coordinates of theal orientation (i.e., left side is left hemisphere).

ory task. Note: (A and B) Left and rights. (D and E) Right and left dorsolateral prefrontalst-associative learning versus simple classification (p �ntreal Neurological Institute (MNI) system. Slices are in

task.rtex.

.00logic

memgyruontraMo



pmbtuvsdodsWssiormRtDat.sa


eters. Nine US users had a C-DISC diagnosis ofconduct disorder, whereas none of the Dutchusers or the controls (both US and Dutch) hadsuch a diagnosis.

Task PerformanceFigure 5 shows performance data for the studyparticipants.WM (STERN). Users did not differ from controlson reaction times and accuracy during the STERNtask. On average, reaction times and accuracy weresimilar to adult performance levels as observed inour previous study,13 indicating normal behavioralWM capacity and efficiency. A main effect of agewas observed (F(1,40) � 11.91, p � .001), indicatingthat older boys performed faster and more accu-rately than younger boys, but this was independentof cannabis use.AM (PMT). Users performed equally accurate ascontrols, and accuracy levels were comparable tothose observed in adults,14 indicating that AMperformance was unaffected by cannabis use. Noeffects of age were found.

fMRI dataScanner compatibility (details in Supplement 1,

TABLE 3 Demographic Characteristics and Drug Use (N

Characteristic

Age 17.2IQ 101Cannabis useLifetime (no. of joints) 4,006Last year (no. of joints) 741Age of onset (years) 13.2Abstinence (weeks since last use) 5.1Other substances (use last year)Alcohol (drinks/week last year) 13.3Tobacco (cigarettes/week last year) 63.8Ecstasy (episodes lifetime) 0.3Amphetamines (episodes lifetime) 1.3Cocaine (episodes lifetime) 0.4Psilocybin (magic mushrooms) (episodes lifetime) 0.4LSD (episodes lifetime) 0.1Laughing gas (episodes lifetime) 0.7Benzodiazepines (episodes lifetime) 2.86C-DISCConduct disorder (yes/no) n �

Note: Data are mean with standard deviations and ranges in parenthesdiethylamide; NS � not significant.

aSignificance of differences calculated using independent-samples t tests (Smirnow Z tests, two-tailed.

available online) was quantified by a direct com- (



arison of temporal signal-to-noise maps (tSNR)aps for both STERN and PMT imaging data

etween the two sites, using nonparametric statis-ics in SPM5 (SnPM5-toolbox; see also www.sph.mich.edu/�nichols/SnPM/). The results re-ealed some areas with differential tSNR betweenites, predominantly in the orbito- and ventrome-ial prefrontal regions. However, there was noverlap between regions showing scanner-relatedifferences in tSNR and regions of interest (ROI;ee below) for the WM or AM task.

M (STERN). A second-level whole-brain analy-is in SPM5 (p � .05, FWE corrected) revealed noignificant differences between users and controlsn working-memory–related brain activity beforer after practice (NT-CT and PT-CT contrastespectively). A subsequent ROI- GLM repeated-easures analysis (Table 2 and Figure 3 forOIs), with group as between-subjects factor,

ask (NT, PT) and ROI (l-IFG, l-SPC, l-PCC/LPFC, ACC) as within-subjects factors, and age

nd country as covariates, yielded a significantask � group interaction (F(1,40) � 12.85, p �001). Separate GLM analyses for NT and PThowed that the WM system tended to be over-ctive before automatization (NT) in users

5)

rs (N � 21) Controls (N � 24) pa

, 15–19) 16.8 (1.3, 13–19) NS.7, 82–116) 111 (11.6, 94–138) �.01

55, 224–32,850) 1.8 (4.0, 0–15) �.0012, 208–3528) 1.0 (2.9, 0–12) �.001, 8–16) 15.0 (1.6, 12–17) NS, 1–16)

.6, 0–46) 3.4 (5.8, 0–24) �.01

.4, 0–144) 6.1 (14.9, 0–53) �.001, 0–4) 0.1 (0.4, 0–2) NS, 0–7) –, 0–4) –, 0–2) 0.1 (0.4, 0–2) NS, 0–1) –, 0–9) 0.1 (0.4, 0–2) NS, 0–40) 0.2 (0.8, 0–4) NS

es), n � 12 (no) n � 24 (no) �.001

DISC � Children’s Diagnostic Interview Schedule; LSD � Lysergic acid

ntelligence Quotient [IQ], age of onset) and nonparametric Kolmogorov-

� 4

Use

(1.0(10

(7,5(77(2.3(4.2

(13(53(0.9(2.4(0.9(0.6(0.2(2.1(8.8

9 (y

es. C-

age, I

F(1,40) � 2.77, p � .10; Figure 6). They showed

567www.jaacap.org

http://www.sph.umich.edu/~nichols/SnPM/


aRsurdet(aattDpn

u

JAGER et al.

significantly larger differences in activity beforeand after practice, e.g., NT-PT contrast values inthe l-IFG (F(1,40) � 12.04, p � .001), the l-PCC/DLPFC (F(1,40) � 22.37, p � .001), and ACC(F(1,40) � 5.42, p � .03). As learning (PT) re-duced activity in the WM system to the samelevel in both groups in most areas, this indicatesoverall excessive effort in users to achieve normalperformance when a task is novel (Figure 7).AM (PMT). A second-level whole-brain analysisin SPM5 (p � .05, FWE corrected) yielded nosignificant group differences in associative learningor recognition-related brain activity (AL-SC andRE-SC contrasts, respectively). Neither did a sub-sequent ROI analysis (all p values �.20; Table 3 andFigure 4 for details on ROIs).Potential Confounders. To explore the relativestrength of the effect of cannabis alone on ROI

FIGURE 5 Behavioral data. Note: (A) Working memorof correct responses on targets for both groups and meanthe control task (CT), after (PT), and before practice (NT).(percentage correct responses) classification (SC) and recusers; DN � drug-naive controls.

between-group differences in WM brain activity, p

JOURN

568 www.jaacap.org

verage � values across all contained voxels perOI were recomputed by running multiple regres-ion analyses, including IQ, alcohol, and tobaccose as regressors, and saving the standardizedesiduals. Standardized residuals, which were nowevoid of effects of IQ, alcohol, and tobacco, werentered into GLM repeated-measures analyses. Theask � group interaction remained significantF(1,40) � 6.81, p � .01). Post hoc ANCOVA withge and country as covariates revealed that evenfter adjustment for effects of IQ or alcohol andobacco use, the group differences in NT-PT con-rast remained significant in l-IFG and l-PCC/LPFC (F(1,40) � 5.99, p � .02 and F(1,40) � 12.05,� .001, respectively) but became marginally sig-

ificant in ACC (F(1,40) � 2.56, p � .10).In users (N � 21), we assessed whether cannabis

se parameters (number of joints, age of onset)

k: mean reaction time � standard error of mean (SEM)centage of errors as percent of all trials (�SEM) duringssociative memory task: accuracy during simple

ion (RE) (� SEM) for both groups. CAN � cannabis

y tasper(B) A

ognit

redicted ROI-specific difference in activity before



.wit


and after automatization (NT-PT contrast). In thel-IFG, number of joints last year was significantlycorrelated to the difference in activity before andafter automatization (Spearman’s rho � 0.52, p �

FIGURE 6 Activity levels per region of interest for working m(NT) for both groups. Note: Activity levels were marginally high� .10), but no significant group effects were found during PT.l-PCC/DLPFC � left precentral/dorsolateral prefrontal cortex; l-

FIGURE 7 Contrast values per region of interest for thedifference in activity before and after practice. Note: ACCgyrus; l-PCC/DLPFC � left precentral/dorsolateral prefro



05), whereas for number of joints lifetime thereas a trend (rho � 0.42, p � .07), tentatively

ndicating that part of the effects found was selec-ively associated with cannabis use.

ry task (in arbitrary units [AU]) after practice (PT) and beforeusers compared with controls during NT (F(1,40) � 2.77, p� anterior cingulate cortex; l-IFG � left inferior frontal gyrus;� left superior parietal cortex.

king memory task (in arbitrary units [AU]), i.e.,anterior cingulate cortex; l-IFG � left inferior frontalortex; l-SPC � left superior parietal cortex.

emoer in

ACCSPC

wor�

ntal c

569www.jaacap.org

odd

fdatacsdpt“fadbgumtsltistbcaisiptipwpbtipconeejo

JAGER et al.

DISCUSSIONThis study examined the effects of regular cannabisuse on adolescent WM and AM brain function. Noevidence was found for effects of cannabis use onAM, at either the behavioral or at the neurophysi-ological level, but the WM system was overactivein users during a novel task, whereas automatiza-tion reduced overall activity to the same level inusers and controls.

Our findings extend those of previous studies inadolescents with cannabis and alcohol use disor-ders reporting increased dorsolateral prefrontal ac-tivation during a spatial WM task.18 Increasedactivity levels in hippocampal and parietal regionshave also been observed in abstinent cannabisusing teens during verbal WM,19 all with normalperformance levels.

We have previously observed subtle alterationsduring WM in the superior parietal cortex in adultcannabis users, but not in prefrontal regions.13 Thepresent results, therefore, support our hypothesisof age-specific effects of adolescent cannabis use onstill-developing brain function and seem to confirmthe vulnerability of developing frontal lobe func-tioning.

Animal studies on the neural consequences ofcannabis exposure during adolescence suggestgreater, more persistent memory deficits and hip-pocampal abnormalities in adolescent than in adultanimals.20 Contrary to what we expected, our re-sults yielded no proof of impaired AM perfor-mance or temporal lobe dysfunction. Our findingsare, however, in line with neuropsychological datain cannabis-using teenagers indicating significantimpact of cannabis on spatial WM and verballearning but not on associative learning.21 In adultcannabis users, we have found hypoactivity com-pared with that in controls in (para)hippocampalregions and the right dorsolateral prefrontal cortexduring AM.14 A recent study by Nestor et al.,22

however, reports parahippocampal hyperactivityand frontocortical hypoactivity in adult users dur-ing an associative face–name learning task. Thecurrent study fails to replicate either of these re-sults. One possible explanation for the discrepan-cies between effects of frequent cannabis use onAM brain function between adolescents and adultsmay be different abstinence periods, as they varybetween short intervals (mean 15 hours; range2–45),22 at least 1 week, but on average not muchlonger,14 and on average 5 weeks (range 1 week to4 months) in the present study. It has been argued
repeatedly that the nonacute effects of cannabis use
JOURN

570 www.jaacap.org

n brain functioning might vary according to theuration of abstinence23 and that this may even beifferent in adolescents.8

This study has several limitations. Groups dif-ered on a number of key variables, with usersisplaying lower estimated IQ scores and greaterlcohol use and tobacco-smoking histories. Al-hough the main findings remained unchangedfter controlling for these factors, and althoughannabis use parameters were linked to the exces-ive activation in the left inferior frontal gyrusuring the novel task, use of alcohol and tobaccoose the possibility of synergistic effects. Together,

hese factors may be considered to constitute acannabis-using lifestyle” that may be predictiveor detrimental effects on development of cognitionnd brain function. The high prevalence of conductisorder in the US users (nine of 12 subjects) maye related to site differences in recruitment strate-ies. In the Netherlands, subjects were recruitedsing a variety of strategies, including advertise-ents on the Internet, and with help of schools. In

he United States, however, recruitment was re-tricted to local substance abuse councils and ado-escent health and resource centers offering educa-ion and treatment programs to minors who gotnvolved with the legal system because of posses-ion of cannabis or other drugs. Boys ending up inhese programs may display more externalizingehavior problems and more often meet criteria foronduct disorder. Banich et al.24 compared brainctivation patterns during a color–word Stroopnterference task between adolescents with severeubstance and conduct problems and controls. Sim-lar to our results, these investigators found thatatients needed to engage prefrontal brain regions

o a greater extent than did controls during thenterference condition to obtain the same level oferformance. However, the question remainshether these differences in brain activation are aredisposing factor in patients with severe, comor-id conduct and substance problems, or whetherhe differences result from the prior ingestion ofllicit substances. The design of our study did notermit conclusions on the nature of the potentialonfounding effect of conduct disorder, as presencef conduct disorder was country-specific and couldot be disentangled from other site-related differ-nces. Still, as any effects related to country differ-nces were regressed out, we believe that we areustified in arguing that the confounding influencef conduct disorder on the main findings is limited.

Besides the advantage of increased power, the



ecnattdceamwirdscbtttsolwdtTentcnniiopccuTpanntc

tqp


multicenter design also poses a challenge in termsof scanner compatibility. We assessed scanner com-patibility in the preparation stage of the study,opted for scan sequences that showed the greatestsimilarities in scans across sites, and, after studycompletion, used several statistical methods to fur-ther minimize and/or quantify systematic effectsdue to site-related differences in scanner equip-ment. Nonetheless, we cannot completely rule outthe influence of scanner differences on between-subjects variability, predominantly in the orbito-and ventromedial regions. These areas are involvedin inhibitory processes and decision making, andaltered brain function in these areas has beenreported in relation to chronic cannabis use/abuseand/or use of other drugs.25,26 We acknowledgelimited sensitivity to detect cannabis-related effectsin orbito- and ventromedial regions in the presentstudy. However, the experimental paradigms ap-plied in the present study are unlikely to elicitactivations in those areas. Hence, we feel confidentthat the results of our ROI analyses have not beencompromised by the multicenter approach.

A third limitation is the cross-sectional design,and hence, the possibility that the observed differ-ences in WM related brain function predated theonset of regular cannabis use. There is ample liter-ature on genetic, neurobehavioral, and personalityprofiles increasing the liability to substance use andabuse, as reviewed elsewhere.27,28 Such pre-exist-ing factors may both increase the tendency ofadolescents to get involved in risky behaviors suchas drug use/abuse, and may affect neurocognitivedevelopment. Finally, the inclusion of cannabis-usingboys, only, excludes the possibility to explore gender-specific differences in the impact of adolescent canna-bis use on cognition and brain function, for whichthere is tentative evidence from animal studies.13

Future neuroimaging studies should attempt toelucidate the structural and neurochemical corre-lates that underlie the observed alterations in brainactivation in cannabis users during a cognitivechallenge, because the clinical significance of thesealterations is far from clear.23 Findings of increasedbrain activation combined with normal perfor-mance are commonly interpreted as functionalcompensation to maintain normal task perfor-mance. Yet, a warning against simplistic or mech-anistic interpretation of increased versus decreasedbrain activation in the absence of differences in taskperformance seems appropriate, especially withregard to fMRI studies in adolescents. During ad-
olescence, both brain maturation and learning and t


xperience shape the neural circuitry that underliesognitive brain function, resulting in highly dy-amic changes in brain function over time.2 Hence,lterations in brain activity patterns as observed inhe present study may signify persistent dysfunc-ion but, alternatively, may also reflect a shift orelay in normal neurodevelopmental changes inognitive brain function. In this context, it is inter-sting to note that our current findings of excessivectivation in several prefrontal areas during theost demanding task condition show similaritiesith a study on automatization and WM capacity

n schizophrenia by Van Raalten et al.16 This studyeports that patients with schizophrenia, who oftenisplay deficits in executive functioning, displayedimilar levels of brain activity after automatizationompared with controls; however higher levels ofrain activity during the novel, more demandingask indicated inefficient WM function and a failureo properly engage the WM brain system whenask demands increase.16 The present results re-emble these findings. It has been suggested, basedn overlap in cognitive dysfunction associated with

ong-term cannabis use and schizophrenia, thatith regard to the neurobiology underlying theysregulation of higher-order cognitive processes,

he endocannabinoid system may be implicated.29

he endocannabinoid system could be involved,ither directly or through its interactions with othereurotransmitter systems, notably dopamine, in

he development of similar cognitive deficits asso-iated with both cannabis abuse and schizophre-ia.30 This notion is consistent with the increasingumber of studies detecting cognitive dysfunction

n adolescent cannabis users7-9 as well as a greaterncidence of juvenile psychotic symptoms andther mental health problems.31 During criticaleriods of neurodevelopment, in particular, adoles-ence, cannabis use may have more impact, espe-ially in otherwise genetically predisposed individ-als, and may precipitate the onset of psychosis.32

his is not to say that cannabis use itself causessychosis in young people, but that a complexssociation between cannabis use and schizophre-ia may be due to dysfunction of the endocannabi-oid system,29 which may be reflected in similari-

ies in abnormal neurophysiology underlyingognitive brain functions.

In conclusion, teenage cannabis use may reducehe ability to process information requiring fre-uent updating. The present study indicates thatredominantly prefrontal brain regions are prone

o adverse consequences of cannabis use during

571www.jaacap.org

JAGER et al.

this stage of life. Whether the effects of adolescentcannabis use on working memory–related brainactivation persist over longer periods of abstinence,as well as their clinical relevance in terms cognitivedysfunction, remains to be determined. &

Accepted February 3, 2010.

Dr. Jager and Ramsey are with the Rudolf Magnus Institute of Neuro-science, University Medical Center Utrecht; Drs. Luijten is with theInstitute of Psychology, Erasmus University Rotterdam; Dr. Block is withthe University of Iowa.

This work was supported by grants of the Netherlands Organizationfor Health Research and Development (ZonMW 31100003) and TheNational Institute of Drug Abuse (NIDA 5R01DA019338) as part of
their bi-national Addiction Program.
analysis using an SPM toolbox. Paper presented at the 8th

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

JOURN

572 www.jaacap.org

The authors gratefully acknowledge Jordan Zuccarelli, MeganBecker, and Catherine Fruehling-Wall for US data acquisition;Vincent Magnotta and Daniel O’Leary for support at the IowaImaging Center and the Iowa Imaging Processing Laboratory; andall Dutch and US institutions that assisted with recruitment of theparticipants.

Disclosure: Drs. Jager, Block, Luijten, and Ramsey report no biomedi-cal financial interests or potential conflicts of interest.

Correspondence to Dr. Gerry Jager, Rudolf Magnus Institute ofNeuroscience, Department of Neurology and Neurosurgery, HPG 03.124, University Medical Center Utrecht, Heidelberglaan100, 3584 CX Utrecht, the Netherlands; e-mail: [email protected]

0890-8567/$36.00/©2010 American Academy of Child andAdolescent Psychiatry

DOI: 10.1016/j.jaac.2010.02.001

REFERENCES1. Sundram S. Cannabis and neurodevelopment: implications for

psychiatric disorders. Hum Psychopharmacol. 2006;21:245-254.2. Bunge SA, Wright SB. Neurodevelopmental changes in WM and

cognitive control. Curr Opin Neurobiol. 2007;17:243-250.3. Giedd JN, Blumenthal J, Jeffries NO, et al. Brain development

during childhood and adolescence: a longitudinal MRI study. NatNeurosci. 1999;2:861-863.

4. Adriani W, Laviola G. Windows of vulnerability to psychopathol-ogy and therapeutic strategy in the adolescent rodent model.Behav Pharmacol. 2004;15:341-352.

5. Schweinsburg AD, Nagel BJ, Schweinsburg BC, et al. Abstinentadolescent marijuana users show altered fMRI response duringspatial WM. Psychiatry Res. 2008;163:40-51.

6. Tapert SF, Schweinsburg AD, Drummond SP, et al. FunctionalMRI of inhibitory processing in abstinent adolescent marijuanausers. Psychopharmacology (Berl). 2007;194:173-183.

7. Jacobus J, Bava S, Cohen-Zion M, Mahmood O, Tapert SF.Functional consequences of marijuana use in adolescents. Phar-macol Biochem Behav. 2009;92:559-565.

8. Schweinsburg AD, Brown SA, Tapert SF. The influence of mari-juana use on neurocognitive functioning in adolescents. CurrDrug Abuse Rev. 2008;1:99-111.

9. Jager G, Ramsey NF. Long-term consequences of adolescentcannabis exposure on the development of cognition, brain struc-ture and function: an overview of animal and human research.Curr Drug Abuse Rev. 2008;1:114-123.

10. Paus T. Mapping brain maturation and cognitive developmentduring adolescence. Trends Cogn Sci. 2005;9:60-68.

11. Giedd JN, Clasen LS, Lenroot R, et al. Puberty-related influenceson brain development. Mol Cell Endocrinol. 2006;25:154-162.

12. Rodriguez de FF, Ramos JA, Bonnin A, Fernandez-Ruiz JJ. Pres-ence of cannabinoid binding sites in the brain from early postna-tal ages. Neuroreport. 1993;4:135-138.

13. Jager G, Kahn RS, Van Den Brink W, Van Ree JM, Ramsey NF.Long-term effects of frequent cannabis use on WM and attention:an fMRI study. Psychopharmacology (Berl). 2006;185:358-368.

14. Jager G, Van Hell HH, De Win MM, et al. Effects of frequentcannabis use on hippocampal activity during an associativememory task. Eur Neuropsychopharmacol. 2007;17:289-297.

15. Goodwin RS, Darwin WD, Chiang CN, et al. Urinary eliminationof 11-nor-9-carboxy-delta9-tetrahydrocannnabinol in cannabis us-ers during continuously monitored abstinence. J Anal Toxicol.2008;32:562-569.

16. Van Raalten TR, Ramsey NF, Jansma JM, Jager G, Kahn RS.Automatization and WM capacity in schizophrenia. SchizophrRes. 2008;100:161-171.

17. Brett M, Anton J, Valabregue R, Poline J. Region of interest

International Conference on Functional Mapping of the HumanBrain, Sendai, Japan, June 10–16, 2002.

8. Schweinsburg AD, Schweinsburg BC, Cheung EH, Brown GG,Brown SA, Tapert SF. fMRI response to spatial WM in adolescentswith comorbid marijuana and alcohol use disorders. Drug Alco-hol Depend. 2005;79:201-210.

9. Jacobsen LK, Pugh KR, Constable RT, Westerveld M, Mencl WE.Functional correlates of verbal memory deficits emerging duringnicotine withdrawal in abstinent adolescent cannabis users. BiolPsychiatry. 2007;61:31-40.

0. Quinn HR, Matsumoto I, Callaghan PD, et al. Adolescent rats findrepeated Delta(9)-THC less aversive than adult rats but displaygreater residual cognitive deficits and changes in hippocampalprotein expression following exposure. Neuropsychopharmacol-ogy. 2008;33:1113-1126.

1. Harvey MA, Sellman JD, Porter RJ, Frampton CM. The relation-ship between non-acute adolescent cannabis use and cognition.Drug Alcohol Rev. 2007;26:309-319.

2. Nestor L, Roberts G, Garavan H, Hester R. Deficits in learningand memory: parahippocampal hyperactivity and frontocorticalhypoactivity in cannabis users. Neuroimage. 2008;40:1328-1339.

3. Gonzalez R. Acute and non-acute effects of cannabis on brainfunctioning and neuropsychological performance. NeuropsycholRev. 2007;17:347-361.

4. Banich MT, Crowley TJ, Laetitia L, et al. Brain activation duringthe Stroop task in adolescents with severe substance and conductproblems: a pilot study. Drug Alcohol Depend. 2007;90:175-182.

5. Volkow ND, Fowler JS, Wang GJ, Baler R, Telang F. Imagingdopamine’s role in drug abuse and addiction. Neuropharmacol-ogy. 2009;56:3-8.

6. Bolla KI, Eldreth DA, Matochik JA, Cadet JL. Neural substrates offaulty decision-making in abstinent marijuana users. Neuroim-age. 2005;26:480-492.

7. Ivanov I, Schulz KP, London ED, Newcorn JH. Inhibitory controldeficits in childhood and risk for substance use disorders: areview. Am J Drug Alcohol Abuse. 2008;34:239-258.

8. Schepis TS, Adinoff B, Rao U. Neurobiological processes inadolescent addictive disorders. Am J Addict. 2008;17:6-23.

9. Solowij N, Michie PT. Cannabis and cognitive dysfunction: par-allels with endophenotypes of schizophrenia? J Psychiatry Neu-rosci. 2007;32:30-52.

0. D’Souza DC, bi-Saab WM, Madonick S, et al. Delta-9-tetrahydro-cannabinol effects in schizophrenia: implications for cognition,psychosis, and addiction. Biol Psychiatry. 2005;57:594-608.

1. Rey JM, Martin A, Krabman P. Is the party over? Cannabis andjuvenile psychiatric disorder: the past 10 years. J Am Acad ChildAdolesc Psychiatry. 2004;43:1194-1205.

2. Weiser M, Davidson M, Noy S. Comments on risk for schizophre-
nia. Schizophr Res. 2005;79:15-21.


mailto:[email protected]

mailto:[email protected]

dPtnynswaiap(2niShatfsb(


SUPPLEMENT 1Magnetic resonance imaging (MRI) scanner com-patibility. This study was a joint venture of theUniversity Medical Center Utrecht (the Nether-lands) and the University of Iowa (United States).Imaging data were collected using two clinical3.0-Tesla MRI scanners, both with an eight-channelhead coil but from different vendors (PhilipsAchieva in Utrech and Siemens Magneton Trio inIowa. As data were pooled to increase statisticalpower, the issue of scanner compatibility had tobe dealt with, focusing both on scanner hardwareand software. The following approach to assessscanner compatibility and to overcome problemsresulting from scanner differences was adopted.

First, in the preparation stage of the study,pilot data were collected of three subjects (i.e.,researchers G.J. and N.F.R.) and a research assis-tant (J.Z.) on both scanners, acquiring severaltry-out three-dimensional (3-D) anatomical scansand functional time series (using different TEs,TRs and flip angles). Data were tested for homo-geneity of signal-to-noise (SNR) and temporalsignal-to-noise (tSNR) ratios across sites and ven-

FIGURE S1A Graphic presentation of the results of a n(STERN) contrasting signal-to-noise (SNR) images from thwise error (FWE)–corrected p value � .05 (correspondinthat show significantly higher SNR in US scans comparedcontrast, i.e., significant higher SNR in Dutch scans compinterest (ROI) for STERN, used in the ROI analysis. Image



ors. Three pulse sequences were compared: 3DRESTO, fast 2D-EPI, and slow 2D-EPI. The first

wo scans involved SENSE/GRAPPA (both scan-ers had the same eight-channel head coil) andielded different tSNR maps. Moreover, we wereot able to make the pulse sequences exactly theame for both scanners. The slow 2D-EPI scan,hich uses the eight channels but does not apply

ny SENSE or GRAPPA, turned out to be the bestn terms of the SNR and tSNR (data not reported)nd was selected as the scan with which toroceed. All scan parameters were the same

TE/TR 35/2000 ms, flip angle 70°, FOV 256 �56 mm, acquisition matrix 64 � 64, slice thick-ess 3.6 (plus a 0.4-mm gap), voxel size 4.0 mm

sotropic, 26 slices, scan orientation transaxial forTERN and parallel to the long axis of theippocampus for PMT). The only differencecross scanners was an opposite step-direction ofhe phase-encoding gradient. This caused a dif-erence in the orbitofrontal region, where macro-usceptibility causes some signal loss and signalecomes shifted away from the brain midlinedue to rapid signal decay and a concomitant T2*

rametric two-sample t test for the working memory taskscanner (n � 23) and Dutch scanner (n � 21), family-

eshold value of T � 5.11). Note: In green, the areasDutch scans. In blue, the areas of the oppositewith US scans. In red (superimposed), the regions ofin radiological orientation, i.e., left � right.

onpae USg thrwith

areds are

572.e1www.jaacap.org

ttre(tmsrsdtaadtbcda

ani

JAGER et al.

weighting across K-space), resulting in reducedtSNR locally and thus reduces sensitivity forBOLD-signal change. The step direction couldnot be made the same for both scanners, so in thelowest slices SNR (and, as a consequence, tosome degree tSNR) is lower in the right hemi-sphere for the Philips scanner and in the lefthemisphere for the Siemens.

Second, after completion of the study the tSNRmaps were quantitatively compared between thetwo sites to ascertain the expected similarity, bydirectly comparing tSNR maps for both STERN(working memory) and PMT (associative mem-ory) image data. As tSNR cannot be assumed tobe normally distributed (both thermal and phys-iological noise contribute, which have differentdistributions), nonparametric statistics were ap-plied, using the SnPM5b toolbox in SPM5 devel-oped by Andrew Holmes and Tom Nichols (seealso www.sph.umich.edu/�nichols/SnPM/). Foreach subject’s time series, tSNR-images were cre-ated based on the preprocessed functional images(i.e., the unwarped, coregistered, normalized, andsmoothed time series) for each task separately.

FIGURE S1B Graphic presentation of the results of a ntask (PMT) contrasting signal-to-noise (SNR) images fromfamily-wise error (FWE)–corrected p value � .05 (correspareas that show significantly higher SNR in US scans comcontrast, i.e., significant higher SNR in Dutch scans compinterest (ROI) for PMT, used in the ROI analysis. Images a

tSNR-images were entered into a nonparametric (

JOURN

572.e2 www.jaacap.org

wo-sample t test in SnPM, using an approximateest of 1,000 permutations (default procedure toeduce computation time when the exact testxceeds 5,000 permutations), a family-wise errorFWE)–corrected p value of .05, and contrastingSNR-maps obtained on the US scanner (Sie-

ens) with tSNR-maps obtained on the Dutchcanner (Philips) for both STERN and PMT. Theesults are shown in Figure S1, and revealedeveral areas that showed significant site-relatedifferences in tSNR (as was expected based on

he pilot data because of different step directions)nd, hence, in activation effect size. Differencesre most prominent in the orbito- and ventrome-ial regions, where the difference in step direc-

ion between scanners likely added to suscepti-ility in these areas to lower signal-to-noise ratiosaused by artifacts resulting from local fieldistortions in the proximity of the eye socketsnd nasal cavities.

Finally, for data analysis, two procedures weredopted to minimize systematic effects of scan-er differences on the results. First, preprocess-

ng included a smoothing step of 8 mm FWHM

rametric two-sample t test for the associative memoryS scanner (n � 24) and Dutch scanner (n � 21),

ing threshold value of T � 5.18). Note: In green, thed with Dutch scans. In blue, the areas of the oppositewith US scans. In red (superimposed), the regions ofradiological orientation, i.e., left � right.

onpathe Uondparearedre in

i.e., two times the voxel size). This is not uncom-




inmarumuvtnsfidssrmdtisdapdhbm

R1


mon in fMRI analyses because of the assump-tions of Gaussian Random Field Theory neededfor some algorithms, but is important in thecontext of scanner differences in that smoothnessequalization (the procedure of smoothing imagedata from different scanners with scanner-relatedvariability in “raw” smoothness to a constantFWHM) markedly reduces any possible activa-tion effect size differences between scanners.1

Second, country (US/Netherlands) was enteredas a covariate in the group-wise comparisons onbrain activity, as any site-related differences intSNR would result in a systematic difference inthe magnitude of activation (i.e., the contrast-to-noise ratio).

Preprocessing of fMRI dataImaging data were analyzed using SPM5(http://www.fil.ion.ucl.ac.uk/spm). Preprocess-ing included realignment (motion correction)and unwarping, coregistration, normalization,and smoothing with an 8-mm (FWHM) Gaussiankernel. For realignment (motion correction), allimages were aligned to the first functional imageusing a rigid body transformation procedure(default option SPM). EPI images are sensitive todistortions due to magnetic field inhomogene-ities, caused by magnetic susceptibility differ-ences in neighboring tissues within the head,resulting in geometrical distortion and signalloss. SPM5 offers a default procedure (unwarp-ing), using algorithms to calculate the geometricdistortion with a field mapping sequence, andthen compensates for these artifacts by geomet-
rically unwarping the EPI images and by apply-


ng cost-function masking in registrations to ig-ore areas of signal loss. Coregistration involvedoving all source images (the anatomical scan

nd the unwarped realigned EPI images) to aeference image (mean unwarped EPI image)sing interpolation methods and affine transfor-ation. An indirect normalization approach was

sed, coregistering EPI images from each indi-idual subject with his high-resolution T1 ana-omical scan (default settings in SPM5), andormalizing this anatomical scan to stereotaxicpace (i.e., the MNI305-template). Parametersrom this step were used to transfer the EPImages to stereotaxic space. No resampling wasone (i.e., voxel size used was the size of acqui-ition (i.e., 4.0 isotropic). First-level fMRI timeeries model specifications included a vector rep-esenting the block designs of the tasks, andodeling the four task conditions (i.e., onset and

urations (in scans) for CT, PT, NT, and instruc-ion frames for STERN, and AL, SC, RE, andnstruction frames for PMT) along with a basicet of cosine functions that high-pass filtered theata (cutoff value of 249.6 seconds for STERNnd 324 seconds for PMT). Cutoff values for highass filtering were calculated using in-house–eveloped software to determine the optimaligh-pass filter taking into account the R-squareetween the factors included in the model (i.e.,inimizing multicollinearity).

EFERENCE. Friedman L, Glover GH, Krenz D, Magnotta V; FIRST BIRN.

Reducing inter-scanner variability of activation in a multicenter
fMRI study: role of smoothness equalization. Neuroimage. 2006;32:1656-1668.
572.e3www.jaacap.org


cannabis use and memory brain function in adolescent boys: a cross-sectional multicenter functional...

Documents