Citation for published version (APA): Huijser, C. (2011). Neuroimaging studies in pediatric obsessive compulsive disorder

Transcription

1 UvA-DARE (Digital Academic Repository) Neuroimaging studies in pediatric obsessive compulsive disorder Huijser, C. Link to publication Citation for published version (APA): Huijser, C. (2011). Neuroimaging studies in pediatric obsessive compulsive disorder General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: or a letter to: Library of the University of Amsterdam, Secretariat, Singel 42, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. UvA-DARE is a service provided by the library of the University of Amsterdam ( Download date: 14 Oct 2018

2 Chapter Functional-MRI during planning before and after cognitive-behavioral therapy in pediatric Obsessive-Compulsive Disorder Chaim Huyser; Dick J. Veltman; Lidewij H.Wolters; Else de Haan; Frits Boer. Published in : Journal of the American Academy of Child and Adolescent Psychiatry (2010) 49(12):

3 Abstract Objective: Pediatric obsessive compulsive disorder (OCD) has been associated with cognitive abnormalities, in particular executive impairments, and dysfunction of frontalstriatal-thalamic circuitry. The aim of this study is to investigate if planning as an executive function is compromised in pediatric obsessive compulsive disorder and is associated with frontal-striatal-thalamic dysfunction, and if this dysfunction would normalize after successful treatment. Method: 2 medication free pediatric patients (Mean age 13.9 year, SD 2.2, range 9-19 years) with obsessive compulsive disorder and 2 healthy controls, matched on age and gender were scanned twice, employing a self-paced pseudo-randomized event-related fmri version of the Tower of London. Patients were rescanned following 16 sessions of protocolized cognitive behavioral therapy; healthy controls were rescanned after a similar time interval. Results: Patients performed the task significantly slower but with similar accuracy compared to controls. Neuroimaging results showed less recruitment of frontal and parietal regions in OCD patients compared to controls during planning versus control task. With increasing task load patients compared to controls showed more recruitment of ventrolateral and medial PFC, as well as insula and anterior cingulate cortex. After treatment, these differences ceased to be significant, time x group x task load interaction analyses showing a significant decrease in right posterior prefrontal activity in OCD patients relative to healthy controls. Conclusion: Pediatric OCD patients show subtle planning impairments and decreased dorsolateral prefrontal and parietal recruitment which normalized following cognitive behavioural treatment. Planning dysfunction is likely to be a state rather than a trait feature of pediatric OCD. 82

4 Introduction Obsessive compulsive disorder (OCD) is characterized by obsessions and compulsions which cause marked anxiety or distress(). In half of the patients with OCD, the disorder starts in childhood(271). Children with obsessions or compulsions at the age of 11 have a high chance of developing OCD in adulthood(67). Neuropsychological studies in adults and pediatric populations with OCD have revealed cognitive impairments in various domains(106), such as planning and organization, mental flexibility, cognitive speed and response inhibition(7;89;260). Also, neuropsychological deficits in childhood, especially with regard to visuospatial processing and executive function, have been associated with adult OCD in a longitudinal study(90), suggesting that executive dysfunction is a core element of OCD. Pediatric populations provide the opportunity to study patients early during development, when potentially confounding factors such as treatment response use of medication or long-term effects of the disorder itself are limited or absent. In pediatric OCD patients, neuroimaging studies employing structural and functional MRI, as well as SPECT and magnetic resonance spectroscopy (MRS) imaging, have provided evidence for frontalstriatal-thalamic abnormalities, similar to studies in adult OCD populations, although different structures appear to be involved (i.e., anterior cingulate cortex (ACC) and putamen/ globus pallidus) compared to adult OCD patients(102). Several findings point at an aberrant development of the brain in pediatric OCD patients when compared with healthy controls(102;232). Until now, activation studies using fmri in pediatric populations with OCD have been rare, and results have been inconsistent: whereas some authors have found hypoactivation of frontostriatal regions(76;317) others reported hyperactivation of these areas(137) using various paradigms. Moreover, all but one(137) used cross-sectional designs, so that state/trait issues could not be addressed, in contrast to repeated measures designs. Following successful pharmacological treatment, structural MRI studies in pediatric OCD have demonstrated normalization of volume differences in thalamus(80), left amygdala(286), and parietal cortex(136), whereas an MRS study demonstrated normalization of caudate glutamergic concentrations(236). The only fmri study to date with a repeated measurement design in pediatric OCD(137) used an implicit learning paradigm (serial reaction task) and reported normalization of bilateral medial prefrontal hyperactivation, and changes of left insula and putamen activity following successful medication treatment. In contrast, two studies which investigated possible treatment 83

5 effects of cognitive behavioural therapy (CBT) employing structural MRI(230) and MRS(21) yielded negative findings. Therefore, it is not yet clear whether the treatment effects found in the above-reviewed studies are due to direct pharmacological action or symptom improvement, as has been observed in adult OCD following CBT(20;193). A key executive function is planning, which has been frequently studied using the Tower of London task (ToL), modified from the Tower of Hanoi(29). The ToL has been adapted for use in the scanner, and several positron emission tomography (PET)(1), single-photon emission computed tomography (SPECT)(191) and fmri(107;138;299) studies have been performed. These studies have consistently shown activation of dorsolateral prefrontal and parietal-occipital cortex during task performance, whereas a number of studies have also reported ACC(1;299), insular cortex(1;299), striatum(299) and frontopolar(1;299) activation. In OCD patients, the ToL has been used both outside and inside the scanner. In most, but not all(263), neuropsychological studies(23;166) OCD patients showed a decrease in speed of planning but no differences in accuracy compared to controls. Van den Heuvel et al.(299) used a parametric, self-paced fmri version of the ToL. They demonstrated decreased frontal-striatal responsiveness during planning in 22 adult OCD patients. In addition, OCD patients showed increased involvement of anterior cingulate, ventrolateral prefrontal and parahippocampal cortices(301). Den Braber et al.(27), using a similar protocol in monozygotic twins discordant for OCD symptoms, found decreased activation of the dorsolateral prefrontal cortex (DLPFC), thalamus, and inferior parietal cortex, but similar responsiveness of the striatum. The authors suggested that reduced striatal responsiveness in OCD may relate to genetic rather than environmental factors. The aim of the present study was to investigate dorsal prefrontal-striatal function during a planning task in medication free pediatric OCD patients before and after cognitive behavioral treatment compared to healthy controls. In our study we used the ToL task, similar to van den Heuvel et al(301) and den Braber et al(27). We hypothesized that pediatric OCD patients compared to healthy controls would perform worse during the task (increased error rates and/or reaction times), associated with reduced recruitment of prefrontal-striatal areas. Also, we aimed to investigate whether such abnormalities would normalize after successful treatment. 84

6 Method Subjects Twenty-five OCD patients (age , mean 13.9, SD 2.2; 16 girls) performed the Tower of London task while functional MRI data were collected. Inclusion criteria were age between 8 and 19 years, diagnosis of obsessive compulsive disorder with a CY-BOCS score of at least 16, OCD symptoms existing at least six months. Exclusion criteria were IQ below 80, use of psychotropic medication, recent state of the art cognitive behavioral treatment, major psychiatric illness, presence of metal in or at the body. Patients were recruited from the outpatient department of our specialized center for children and adolescents with OCD. In addition, twenty-five controls, pair-wise matched for age and gender (age , mean 13.70, SD 2.8) were recruited. The present study is part of a larger study which investigates (neuro) psychological and biological mechanisms of change during cognitive behavioral treatment for pediatric OCD. Following baseline measurements (T0), all patients were treated with 16 sessions of protocolized cognitive behavioral therapy(93;316). Treatment sessions were conducted by trained and registered cognitive behavioral psychotherapists, who were supervised by the author of the manual (EdH), and consisted of exposure with response prevention and cognitive therapy suited to the needs of the patients. After 16 sessions of CBT patients were rescanned (T1); healthy controls (HC) were also rescanned after a similar time period. The study was approved by the Ethical Committee of the Academic Medical center in Amsterdam (MEC 06/03# ). All patients and controls and their parents gave written informed consent. Measurements Diagnostic assessments were performed by senior clinicians using a semi-structured interview (Anxiety and Depression Inventory Schedule, Child and Parents version (ADIS C/P)(261)). Symptom severity was investigated with the Child version of the Yale-Brown Obsessive Compulsive Scale (CY-BOCS)(24) by one of the investigators (LHW). Depression and anxiety symptoms were rated using the Childhood Depression Inventory (CDI)(12) and the State Trait Anxiety Inventory for children (STAI-C)(270) for anxiety. CBCL and YSR(2) ratings were obtained to assess overall functioning, and the OCS scale(196) was administered to control for OCD symptoms in healthy controls. Intelligence was assessed with the WISC IV(312) (age <17) or WAIS(311)(age >17) with two subtests: Block patterns 8

7 and vocabulary. Healthy controls were screened for psychopathology with the CBCL, YSR, OCS scale, STAI-C and CDI. At rescanning (T1), the STAI-C state and CDI were again administered to all subjects, and the CY-BOCS to patients only. Task We used a pseudo-randomized self-paced version of the ToL adapted from van den Heuvel et al(299). This version consists of 6 conditions, a control task condition and planning conditions consisting of 1 to moves. Stimuli were presented on a screen and consisted of two configurations, each having three colored beads in 3 different colors, on three vertical rods which can accommodate 3, 2 and 1 beads each. Subjects were asked to indicate the minimum number of steps to reach the target configuration using a response pad. During control task trials, subjects were asked to count the total number of yellow and blue beads. The maximum time for response was 30 seconds for each trial, and the total task lasted 1,000 seconds (17 minutes). Because of the self-paced nature of the task, the number of trials differed across subjects (mean 14, SD 39.9, range , no significant difference between groups). Subjects were asked to rate their anxiety from 0-10, immediately before starting the task. Data acquisition Imaging was performed on a 3.0T Intera MR system (Philips Medical Systems, Best, The Netherlands) with a six-channel SENSE head coil. Head immobilization was established using foam pads inside the coil. Stimuli were generated on a personal computer with E-prime software and projected on a screen at the end of the scanner table which the subject could see through a mirror mounted above the coil. Two response boxes were used to record the subject s responses. Anatomic imaging included a coronal 3-dimensional gradient-echo T1- weighted sequence (Flip angle=8, repetition time (TR) = 9.69ms; echo time (TE) = 4.60ms, 182 slices, 26x26 pixels, voxel size 1x1x1.2mm). For fmri, an echo planar imaging sequence (TR=2.3sec, TE=30ms, 96x96 pixels) was used, creating whole brain acquisitions (40 axial slices, 2.29mmx2.29mm in-plane resolution, 3.0mm slice thickness). In total, 440 volumes per subjects were scanned. The entire scanning procedure included two other task fmri paradigms, following the Tol task, and a DTI scan, so that each subject was in the scanner for about one hour. 86

8 Data analysis Demographic and behavioral data were analyzed with SPSS software (version 17.0; SPSS Inc, Chicago, Ill), using one-way and repeated measures ANOVA. Incorrect trials were excluded from RT analyses. Imaging data were analyzed using SPM (Wellcome Trust Centre for Neuroimaging, London, UK) running in Matlab version (Mathworks, Sherborn, MA). Preprocessing was done by correcting the time series for slice time acquisition and spatial realignment. Following co registration of the mean EPI to the T1-weighted structural scan, images were normalized to MNI space (Montreal Neurological Institute). We used the standard template implemented in SPM for reasons of comparability across studies, and because there is evidence that in the adolescent age range, functional differences relative to young adults are no significant(112). Data were smoothed using an 8mm Gaussian kernel. Next, imaging data were analyzed voxel-wise in the context of the General Linear Model (GLM) to calculate statistical parametric maps of t-statistics for condition specific effects. Control and task trials were modeled using delta functions convolved with a canonical hemodynamic response function and modulated using RTs. Error trials were modeled separately as a regressor of no interest. For each subject, weighted contrasts were calculated for main effects of all planning tasks versus control task, and for task load. Contrast images of the subjects were entered into second level (random effects) analyses using independent samples T-tests to compare patients and controls at T0 and T1, with additional ANOVAs to assess group x time interaction effects. Main effects of planning task vs. control task and task load for each group are reported at p<0.0 corrected for multiple comparison using the False Discovery Rate method with minimum cluster extent of voxels, whereas interaction effects (masked with the group main effect in order to constrain the search for interaction effects to voxels also showing a main effect of task) are reported at p uncorrected (Z>3.09). Post hoc regression analyses were performed between changes in CYBOCS scores and BOLD activation differences before and after treatment. 87

9 Results Demographic and Behavioral data The groups did not differ with respect to handedness and state anxiety, as listed in Table 1. Intelligence scores on vocabulary and block patterns were significantly higher for controls although all scores were within the normal range. Ratings on the OCS scale of the CBCL, on the STAI and the CDI were significantly higher in patients than in controls. No controls were within the clinical range. OCD patients had a considerable overlap of OCD symptom categories based on the CY-BOCS symptom checklist and had high rates of co morbid disorders. Five patients had a medication history. Table 1. Demographic and Clinical data Measurements Patients N=2 Controls N=2 Age Mean 13.9 SD 2.2 Range Mean SD 2.8 Range Gender 16 girls-9 boys 16 girls-9 boys Handedness 22 right, 3 left 23 right, 2 left Intelligence (WISC IV) Psychopatology (CBCL) Depression (CDI) T0 T1 Anxiety (STAI-C) T0 T1 Anxiety in the scanner (0-10) Time between scans Diagnosis (ADIS) (Patients only) Block pattern= 9.4 SD 3.1 Vocabulary= 10. SD 1.9 Internalizing 20.8 SD 8.1 Externalizing 13.8 SD 7.8 Total score 61.0 SD 20.7 OCS scale 9.7 SD SD 6.2 (2 cases >19) 9.42 SD 6.7 Trait 34.6 SD 6.4 State 33.9 SD 6.1 State 31.7 SD SD SD 1.8 Block Pattern= 11.2 SD 2.7 Vocabulary= 12.0 SD 1.49 Internalizing 2.7 SD 3.8 Externalizing 2.3 SD 2.8 Total score 9.2 SD 8.7 OCS scale 0.67 SD SD 3. Trait 27.2 SD 4.4 State 29.8 SD 3.3 State 28. SD SD 1.89 NS 2.43 SD 1.2 NS. month (1.3).7 month (1.76) NS OCD= 2 Other anxiety disorder 48%: SAD 4%, SoPh 16 %, SpPh 16%, GAD 8 %, PTSD 4%. Affective disorders 12% :8 % depression, 4% dysthimic disorder Externalizing disorders 12%: 8 % ADHD, 4% ODD. 88

10 OCD severity (CY- BOCS) (Patients only) Symptom dimensions on CYBOCS symptom checklist Medication history Tics 8%. Obsessions T0: SD 2.71; T1: 6.63 SD 4.83 Compulsions T0: SD 2.48 T1: 6.38 SD 4.8 Total T0: SD.08 T1: SD 9.62 (Range 16-3) (Range 0-33) Washing/contamination, 7 %; Checking/aggressive, 94%; Ordering/symmetry, 8 %; Hoarding, 33 % patients: - 2 fluoxetine (discontinued 6 months, 1 year); - 2 risperidone (discontinued 3month and 14 days before scanning) - methylphenidate, dexamphetamine and atomoxetine (discontinued 1 year before scanning). p<0.001, p<0.01, p<0.0 patients versus controls None Table 1. Demographic and Clinical data Abbreviations: WISC IV= Wechsler Intelligence Scale for Children, CBCL=Child Behavior Checklist, OCS scale=obsessive compulsive symptom scale, CDI=Child depression inventory, STAI=State and trait anxiety inventory, ADIS= Anxiety and Depression Interview Schedule, OCD=Obsessive compulsive disorder, SAD= Separation anxiety Disorder, SoPh= Social Phobia, SpPh= Specific Phobia, GAD= Generalized Anxiety Disorder, PTSD= Posttraumatic stress disorder, ADHD= Attention Deficit Hyperactivity Disorder, ODD= Oppositional Defiant Disorder, CY-BOCS= Children s Yale-Brown Obsessive Compulsive Scale. Following treatment with 16 sessions CBT, CY-BOCS scores showed a significant decrease. Clinical response (i.e., reduction >30 %) was established in 18 out of 24 patients, and 16 out of 24 patients reached a CY-BOCS below 16. At T0, overall reaction times were significantly (p=0.03) longer in patients (8.01s (SD 2.63)) than controls (6.7s (SD 1.29) (Supplement table 1). Planning accuracy was similar in both groups (patients 90.9 % (SD 4.6), HC 89. % (SD 6.6)). We did not find a significant association of gender, age or intelligence scores with regard to accuracy or overall mean reaction time. Following treatment, we found a significant time x group effect for overall mean reaction time (F=7.034, p=0.01), indicating a greater decrease in patients (T1:.96s (SD 1.99)) than control subjects (T1:.76s (SD 1.22)), the difference in RT at T1 between patients and controls was no longer significant (p=0.68) (see Table 2 and Figure 1). 89

11 Table 2. Behavioral data Tower of London task Patients N=2 (T1= 24) Controls N=2 (T1=22) Condition RT T0 (SD) RT T1 ACC T0 ACC T1 RT T0 RT T1 ACC T0 ACC T1 Baseline.14 (1.98) 1 move 6.06 (2.06) 2 moves 8.16 (3.23) 3 moves (3.23) 4 moves (6.10) moves 18,2 (8.34) All moves 8.01 (2.63) 3.81 (1.64) 4.86 (1.67) 6.03 (1.88) 7.74 (2.93) 9.0 (3.30) 14.1 (7.01).96 (1.99) 94.9 (3.4) 9.2 (6.9) 89.8 (13.1) 88.1 (10.4) 78.8 (12.3) 79.7 (23.7) 90.9 (4.6) 93.2 (3.7) 94.8 (6.2) 86.6 (13.3) 89.8 (1.1) 76.7 (1.7) 72.9 (27.2) 89.3 (8.0) 4.22 (0,84).07 (0.99) 6.89 (1.7) 8.9 (2.38) (2.1) 1.31 (4.2) 6.7 (1.29) 3.46 (.74) 4.0 (.99).90 (1.68) 7.9 (1.80) 9.4 (2.40) (3,81).76 (1.22) 92.8 (2.9) 94.2 (1.9) (10.9) (3.7) (14.6) (7.6) (14.07) (10.0) 80.7 (11.1) 77.9 (16.4) 79.3 (13.9) 77.2 (16.3) 89. (6.6) 91. (4.) p<0.0 patient versus control Table 2: Reaction times (RT) and accuracy (ACC) on Tower of London task before (T0) and after (T1) treatment. Reaction times are in seconds, accuracy is percentage of accurate responses. Figure 1. OCD Controls Figure 1. General Linear Model repeated measure Reaction Time at T0 Mean all moves versus Reaction Time at T1 Mean all moves for patients and controls show significant (p<0.01) time x group effect for overall mean reaction time (F= 7,034, p 0,01), indicating a greater decrease in patients (T1:.96 s (SD 1.99) than control subjects (T1:.76 s (SD 1.22)). 90

12 Imaging data Effects of planning task versus control task We did not observe significant group x time x task interaction effects for planning versus control task (counting). However, group-by-task interaction analysis at T0 showed greater BOLD signal in healthy controls (HC) compared to OCD patients in left posterior DLPFC/left premotor cortex (MNI , Z=3.3) and the right parietal cortex (MNI , Z=3.21) (Figure 2A-D), OCD patients revealed no areas of greater BOLD signal compared to HC at T0. At T1 no significant differences in the group-by-task interaction analysis were observed between HC and OCD patients. Also, changes between T0 and T1 in both groups were not significant at our a priori threshold. Figure 2 2A 2B 2C 2D Figure 2. Brain regions showing significant Blood Oxygenation Level-Dependent signal increase during planning compared with control task in healthy controls versus OCD patients at P<0.001 uncorrected before treatment. A(above): Frontal lobe, Posterior DLPFC/premotor region; B) BOLD signal before and after treatment in OCD (black, T0=first column, T1=second column) and HC (grey, T0= third column, T1= forth column) in DLPFC (MNI ); C(under): Parietal lobe, post central gyrus; D) BOLD signal before and after treatment in OCD and HC in parietal lobe (MNI 4, -27, 4) 91

13 Main effects of the planning task compared with our control task at T0 revealed an increase in blood oxygenation level dependent signal (BOLD) in healthy controls in bilateral dorsolateral prefrontal regions, including bilateral premotor regions, left ventrolateral prefrontal cortex, bilateral parietal cortex, bilateral insular cortex and left caudate nucleus. OCD patients showed similar (although somewhat less robust) activation patterns in bilateral premotor, parietal, and insular cortex and left caudate nucleus, as well as in left occipital cortex (Figure 3 A-B, and Table 3). Figure 3A 3B Figure 3. Brain regions showing significant Blood Oxygenation Level-Dependent signal increase during planning task compared with control task in OCD patients (left, 1A) and controls (right, 1B) before treatment. P<0.0 corrected (FDR). Table 3 Main effect of task: Planning versus control task (counting) OCD patients at p<0.0 corrected (FDR) Controls at p<0.0 corrected (FDR) Voxels in MNI space Voxels in MNI-space Cluster p(fdr) Z x,y,z (mm) Cluster p(fdr) Z x,y,z (mm) Frontal lobe, Premotor area, left (BA 6, 8): Frontal lobe, Premotor area, right (BA 6,8) : Frontal lobe, Middle frontal gyrus, left (BA 9, 10)

14 Frontal lobe, Middle frontal gyrus, right (BA 9,10): Frontal lobe, inferior frontal gyrus, left Frontal lobe, inferior frontal gyrus left(ba 44) Parietal lobe, precuneus, left (BA 7,40) : Parietal lobe, precuneus right (BA 7, 40,19): Insular cortex, left: Insular cortex, right: Caudate nucleus, left Occipital lobe, superior occipital gyrus, left (BA 19) Table 3. Brain regions showing significant Blood oxygenation Level-Dependent Signal Increase during the planning task compared with control task (counting) in OCD patients and controls before treatment. P<0.0 corrected (FDR). Abbreviations: BA = Brodmann area Figure 4. 4A. 4B. Figures 4.Changes in CYBOCS scores before and after treatment (DeltaCybocs) were significantly correlated with BOLD-changes before and after treatment in OCD patients during planning versus control task within the left parietal cortex (MNI -60, -6, 18, Z=3.48, r=0.66, p 0.001). 93

15 Post hoc analyses revealed that differences in CYBOCS scores were significantly correlated with changes in BOLD activation before and after treatment within the left DLPFC (MNI , Z= 3.68, r =.683, p = 0.000) and left parietal cortex (MNI -60, -6, 18, Z=3.48, r =.66, p = 0.001) (Figure 4A-B). Effects of task load Time x group x task interaction analyses with increasing task load revealed a significant decrease of BOLD signal in OCD patients relative to HC over time in the right inferior frontal gyrus, extending into frontal operculum (MNI , Z= 3.16). Group-by-task interaction analyses at T0 (Figure A and B, and Table 4) showed that increasing task load was associated with a greater BOLD signal in OCD patients compared to healthy controls in left DLPFC, left dorsal ACC, and right dorsomedial PFC, as well as left insular cortex. These differences in BOLD signal before treatment (T0) between OCD and HCs were no longer present at T1. Time x load analyses did not reveal significant changes between T0 and T1 in HC, whereas in our OCD subjects the right inferior frontal region identified in our time x group x load analysis was observed only at a slightly lower threshold (MNI , Z=2.9). Figure A B Figure. A) Increased BOLD signal at increased task load in insular cortex (cross-hairs), medial dorsal prefrontal lobe (mdpfc) and left dorsal anterior cingulate cortex (dacc) in OCD patients compared to healthy controls before treatment. B) BOLD signal in left insula (MNI -42, 12, -12) before and after treatment for OCD (black) and HC (grey). 94

16 Table 4 Region side MNI-space Coord. Z p Frontal lobe, Inferior gyrus left Insular Cortex left Frontal Lobe, Middle Frontal Gyrus left Frontal Lobe, Medial Frontal Gyrus right Limbic Lobe, Cingulate Gyrus left Brain regions showing significant Blood oxygenation Level-Dependent Signal Increase during increased task load. OCD patients versus healthy controls at P uncorrected, Z>3.09, before treatment. Main effect of increased task load at T0 (Figure 6A-B, and Table ) was found in HC in bilateral DLPFC and precuneus, right occipital lobe cortex and bilateral striatal activation (left caudate nucleus and right putamen). Again, OCD patients showed similar activations, also including bilateral parietal cortex and right cerebellum. Post hoc analyses revealed that changes in CYBOCS scores were significantly correlated with BOLD-changes during increased task load within the left parietal cortex (MNI -4, -1, 27, Z=3.46, r=.64, p=0.001). Figure 6 6a 6B Figure 6.Brain regions showing significant Blood Oxygenation Level-Dependent signal increase during increased task load in OCD patients (left, 4A) and controls (right, 4B) before treatment. P<0.0 corrected (FDR). 9

17 Table. Main effect of increased task load OCD patients Healthy Controls Voxels in MNI-space Voxels in MNI-space Cluster p (FDR) Z x,y,z (mm) Cluster p(fdr) Z x,y,z (mm) Frontal Lobe, Superior Frontal Gyrus, left, (BA10) (BA6) (BA 6) (BA6) (BA 8) Frontal Lobe, Superior Frontal Gyrus, right, BA Frontal Lobe, Inferior Frontal Gyrus, left, BA Frontal Lobe, Precentral Gyrus, left, BA Parietal Lobe, Precuneus, left, BA (BA 19) Parietal Lobe, Precuneus, right, BA Parietal Lobe, Inferior Parietal Lobule, left BA Parietal Lobe, Inferior Parietal Lobule, right BA Occipital Lobe, Superior Occipital Gyrus, right, BA Nucleus Caudate, left Putamen, right Globus Pallidus, right Anterior Lobe Cerebellar Lingual, right Table. Brain regions showing significant Blood oxygenation Level-Dependent Signal Increase during increasing task load in OCD patients and controls before treatment. P<0.0 corrected (FDR). 96

18 Discussion The present study is, to our knowledge, the first to report on executive dysfunction and its neural correlates in unmedicated pediatric OCD patients, and to demonstrate normalisation of these abnormalities following cognitive behavioural therapy. In our study we investigated behavioral performance and associated brain activation during a Tower of London planning task, in medication-free pediatric OCD patients before and after cognitive behavioural therapy compared to healthy controls (HC) who were also scanned twice. Our results showed that at baseline, pediatric OCD patients compared to healthy controls had longer mean reaction times but similar accuracy during planning, associated with decreased recruitment of frontal-parietal areas. With increasing task load, pediatric OCD patients were found to activate additional brain regions, in particular dorsomedial prefrontal cortex and dorsal anterior cingulate cortex, and insula, compared with control subjects. Following cognitive behavioral therapy, these differences between OCD and HC ceased to be significant, time x group x task load interaction analyses showing a significant decrease in right posterior prefrontal activity in OCD patients relative to healthy controls, indicating a decrease in task load associated activity between T0 and T1 present in OCD patients but not in controls. In addition, changes in symptom severity were correlated with changes in BOLD activation patterns before and after treatment in DLPFC and parietal cortex. Time x group x planning versus control task were not significant at our a priori threshold, which is somewhat surprising given that task vs. control condition contrasts tend to be more robust, albeit less specific, than parametric load contrasts(66). A possible explanation is that time x task vs. control comparisons may be confounded by differences in baseline activity, as was also suggested by van den Heuvel et al. in their ToL study in adult OCD. However, although previous resting-state studies employing 18F-deoxyglucose PET have shown a decrease in frontal-striatal metabolism following successful CBT (20), such an explanation must be considered speculative since we did not measure resting-state perfusion. Therefore, the alternative explanation of differential novelty and/or learning effects between our groups cannot be ruled out. Our behavioral findings of increased mean reaction times coupled with normal accuracy are in line with several ToL studies in adult OCD patients (23;166). In our study, both groups showed increased performance speed at retest, but this effect was larger in OCD patients, likely reflecting clinical improvement as well as learning processes. The first fmri study with the ToL of van den Heuvel et al (301) reported not only slower performance but also 97

19 increased error rates in their sample of adult OCD patients. Interestingly, Roth et al.(239) showed in a study directly comparing early vs. late onset OCD that the late onset group performed worse on EF tasks than the early onset group, which may explain why we failed to observe accuracy differences in our pediatric OCD group. The present fmri study is also the first to employ a parametric ToL paradigm in adolescents. Imaging results from the present study showed main effects of task in brain regions also found in adult samples during a planning task(1;27;138;191), in particular dorsolateral prefrontal and parietal cortex, and basal ganglia. The DLPFC has been extensively associated with executive function, including planning, set shifting, decision making (203). DLPFC activity during the TOL is likely to include generating, selecting and/or remembering mental moves, whereas relating these moves to the target configuration involves a comparison with goal representations in parietal areas(240). Our pediatric OCD patients showed less recruitment of these regions compared to HC, which is in line with the findings of previous studies. Although fmri studies in pediatric OCD have been rare, our results of decreased frontal activity are partly in agreement with those of Wooley et al.(317), who reported reduced activation of orbitofrontal-striatal and DLPFC regions during a stop signal task. Parietal involvement in OCD has been reported in adult populations across several imaging modalities(283;298). In pediatric OCD, although several structural studies(232;28) did not reveal parietal abnormalities, one study found increased parietal gray matter(284), whereas another(37) found reduced white matter in right parietal cortex. Lazaro et al.(136) found decreased bilateral parietal GM and right-sided WM reductions, which normalized following antidepressant therapy and behavioral counseling. The latter finding is of interest in view of results from the present study, showing decreased parietal activity in patients normalizing following CBT, which was correlated with changes in severity scores of OCD symptoms, and may reflect a similar neuroanatomical substrate, e.g. regional dyspruning as suggested by Rosenberg and Keshavan(232). When comparing BOLD signals associated with increasing task load we found increased activity in DLPFC and precuneus in both groups, as well as bilateral parietal inferior gyrus in pediatric OCD patients, the latter finding being in agreement with earlier ToL studies(27;301). Group by task interaction analyses at baseline showed additional involvement of dorsal ACC, dorsomedial and dorsolateral PFC, and insular cortex in pediatric OCD patients, whereas time x group x task load interaction analyses revealed a significant decrease in right posterior prefrontal activity in OCD patients relative to healthy controls following CBT, 98

20 indicating normalization of prefrontal recruitment with increasing task load. In addition, we suggest that increased dorsal ACC and dmpfc activity at baseline may reflect increased error monitoring in OCD patients when solving complex trials. Insular cortex involvement has been reported previously in OCD during symptom provocation paradigms(207), and may reflect increased arousal in our OCD group, even though state anxiety ratings were not significantly different. Our study supports neurobiological models of OCD characterized by both dysfunction of dorsal frontal-parietal-striatal (cognitive) networks and hyperactivity of medial prefrontallimbic (affective) circuitry(176), although the latter was observed only when task demands increased. We also showed that planning impairments may normalize following CBT, although executive dysfunction is usually not targeted in cognitive behavioral treatment approaches of OCD. Therefore, it is worth investigating whether treatment enhancement in OCD can be achieved by using elements of cognitive remediation therapy, as has been advocated for anorexia nervosa(290). Several potential limitations need to be addressed. First, our groups differed significantly with regard to intelligence as measured using vocabulary and block pattern subtests. However, neither score was correlated with overall mean reaction time or accuracy. In addition, post hoc analyses of our imaging data with covariance analysis controlling for intelligence scores revealed similar group differences, both for our task vs. control task and task load contrasts (data not shown). Second, our groups differed with regard to depression and anxiety ratings as measured using CDI/STAI-C, although anxiety ratings in the scanner were similar. Controlling for trait anxiety by including STAI scores as a covariate revealed highly similar group effects to those reported. However, when adding depression rates (CDI), differences in DLPFC activation for planning vs. control condition ceased to be significant at our a priori threshold, therefore we are not fully able to rule out the effect of co morbid depression on our results. Also, we were unable to differentiate between OCD symptom dimensions within our sample. In adult OCD, several neuroimaging studies have reported differences between subgroups(171;300), although in pediatric OCD findings have been inconsistent (76;284). Notwithstanding these potential limitations, our study is the first to show slower performance and decreased neurophysiologic responsiveness in unmedicated pediatric OCD patients during a planning task, which can be reversed following successful CBT. Future research should focus on studying subgroups with respect to symptom sub dimensions and age, and aim to control for co morbidity by including additional experimental groups, to investigate the usefulness of cognitive fmri paradigms as clinical markers in pediatric OCD. 99