Mapping Structural Brain Alterations in Obsessive-Compulsive Disorder FREE

Results from a variety of research sources have made it possible toconsider the pathogenesis of obsessive-compulsive disorder (OCD) in termsof brain anatomy.^1- 2 Neurologicalpatients showing obsessions and compulsions generally have diseases involvingthe basal ganglia or the inferior frontal cortex.^1- 4 Neuropsychologicalalterations in patients with OCD include deficits in the executive functionsof the frontal lobes.⁵ Neurosurgical proceduresaimed at treating OCD symptoms directly or indirectly disconnect frontosubcorticalloops.⁶ Functional brain imaging reveals orbitofrontaland basal ganglia hypermetabolism in patients with OCD at rest and duringsymptom provocation that normalizes with response to treatment.^7- 9

Available data, therefore, are consistent in identifying functionalalterations involving frontosubcortical circuits. It is not clear, however,to what extent the altered systems show detectable changes in their anatomy.Structural neuroimaging has indeed provided some evidence of anatomical alterationsin OCD, but the reported findings have been notably heterogeneous, probablyas a consequence of different patient selection and the diversity of methodsadopted to delineate regions of interest. Although volume reduction^10- 12 and enlargement^13- 15 were observed fordifferent components of the basal ganglia and thalamus, other researchers^16- 19 reportedno differences between patients and control subjects. Reduced orbitofrontaland amygdala volumes were also detected,²⁰ andalterations in distant structures, such as the cerebellum, were identifiedusing 3-dimensional magnetic resonance imaging (MRI).^21- 22

The diversity of the described neuroanatomical findings suggests thatimaging procedures that allow analysis of the entire cerebral parenchyma couldfurther contribute to delimiting the distribution of prevailing structuralalterations in the OCD brain, particularly if a large number of patients areassessed. Recent technical developments have made it feasible to measure directsize variations of anatomical elements on a voxel-by-voxel basis. In thisstudy, we use a voxelwise approach to identify possible structural brain alterationsin the MRIs of 72 patients with OCD. We report statistical parametric maps(SPMs) showing substantial brain tissue volume differences between patientswith OCD and a comparable group of 72 controls and describe the correlationsof anatomical changes with relevant clinical variables.

Outpatients were consecutively recruited according to DSM-IV criteria for OCD and the absence of relevant medical, neurological,and other major psychiatric diseases. Specifically, no patient in this studymet the criteria for Tourette syndrome or showed psychoactive substance abuse.Comorbidity with anxiety and depression symptoms was not considered an exclusioncriterion provided that OCD was the primary clinical process. Diagnosis wasindependently assigned by 2 psychiatrists (P.A. and J.M.M.) with extensiveclinical experience in OCD who separately interviewed patients using the Structured Clinical Interview for DSM-IV Axis I Disorders–Clinician Version.²³ Patientswere eligible for the study when both research examiners agreed on all criteria.

Seventy-two patients composed the study group (32 women and 40 men;mean ± SD age, 29.8 ± 10.5 years; range, 18-60 years). All but11 patients were right-handed according to the Edinburgh Inventory.²⁴ The mean ± SD level of education attainedwas 13.2 ± 3.6 years. Table 1 reportsthe clinical characteristics of the patient sample. The Yale-Brown Obsessive-CompulsiveScale²⁶ and a clinician-rated Yale-Brown Obsessive-CompulsiveScale symptom checklist²⁶ were used to assessseverity and to characterize the clinical expression of the OCD. As in otherstudies,^25,27- 30 symptomseverity was recorded for each checklist element. Each participant was assigneda score of 0 (absent), 1 (mild), or 2 (prominent) for each of the 5 clinicaldimensions defined by Mataix-Cols et al²⁵ (Table 1). The score provided for each givendimension reflected the highest score for any of the checklist elements composingthat dimension. Comorbid processes were assessed in the clinical interviewsusing DSM-IV criteria. Hamilton scales^31- 32 wereused to rate depression and anxiety at inclusion. Ten patients had receivedexperimental treatment with transcranial magnetic stimulation on the frontallobe 12 months before inclusion in this study.³³

Seventy-two comparative control subjects from the same sociodemographicenvironment were matched to the patients by age, sex, and handedness. A detailedmedical history was recorded and a psychiatric interview was administeredbefore inclusion to exclude psychiatric disorders, adopting the guidelinesestablished by Shtasel et al.³⁴ The selectedvolunteers, of whom 11 were left-handed, had a mean ± SD age of 30.1± 10.2 years (range, 18-57 years) and an identical sex distributionas the patient group (32 women and 40 men). The mean level of education inthe control group was 14.0 ± 3.1 years (t₁₄₂ = −1.5; P = .14).

All patients and controls gave written informed consent after receivinga complete description of the study, which was approved by the institutionalreview board (Hospital of Bellvitge).

All imaging studies were acquired using a 1.5-T magnet (Signa; GE MedicalSystems, Milwaukee, Wis). A 60-slice, 3-dimensional, spoiled gradient–recalledacquisition sequence was obtained in the sagittal plane. Acquisition parameterswere as follows: repetition time, 40 milliseconds; echo time, 4 milliseconds;pulse angle, 30°; field of view, 26 cm; and matrix size, 256 × 192pixels. Section thickness varied according to brain size, ranging from 2.4to 2.6 mm and covering the brain using a fixed number of 60 slices in eachcase. Acquisition time was 8 minutes 13 seconds. These series parameters produceanisotropic voxels with moderate spatial resolution (voxel dimensions, typically1.0 × 1.3 × 2.5 mm) and an optimal signal-to-noise ratio, allowingreliable brain tissue segmentation.^35- 36

Imaging data were processed on an auxiliary workstation (Sun Ultra 5;Sun Microsystems Inc, Santa Clara, Calif) using a technical computing softwareprogram (MATLAB 5.3; The MathWorks Inc, Natick, Mass) and SPM software (SPM99;The Wellcome Department of Imaging Neuroscience, London, England).

All images were checked for artifacts, and the origin was placed onthe anterior commissure before preparing MRIs for voxel-by-voxel analyses.In short, image preprocessing involved several automated procedures aimedat (1) optimally normalizing gray matter, white matter, and cerebrospinalfluid (CSF) segments using study and tissue-specific templates; (2) modulatingvoxel values from spatial normalization data to preserve volumetric information;and (3) averaging neighboring voxel values (smoothing) to load each elementwith region information. Each image transformation step is described in detailin the following subsections.

A template image was generated for each cranial compartment (gray matter,white matter, and CSF). Following the optimized voxel-based morphometry methodproposed by Good et al,³⁷ each 3-dimensionalMRI was firstly transformed to a standard stereotactic space by normalizingthe imaging data to the SPM99 T₁ template. Next, each image wassmoothed using an 8-mm full-width at half-maximum isotropic Gaussian kernel,and a mean image was created to serve as a whole-brain template in this study.All the structural images were then normalized to this study-specific template.Spatial normalization used the residual sum of squared differences as thematching criterion and involved a first step consisting of a 12-parameteraffine transformation³⁸ and a second step consistingof nonlinear iterations using 7 × 8 × 7 basis functions accountingfor global nonlinear shape differences.³⁹

Normalized images were then segmented into gray matter, white matter,and CSF partitions. As is characteristic of SPM99 segmentations, each voxelwas assigned a probability of belonging to a particular partition based onits signal intensity value and information from previous SPM99 probabilitymaps describing the relative distribution of tissue types. In this step, animage intensity nonuniformity correction was performed.⁴⁰ Thesenormalized and segmented images from the 144 study participants were finallysmoothed using an 8-mm full-width at half-maximum isotropic Gaussian kerneland averaged to create gray matter, white matter, and CSF templates. Thesetissue-specific templates served to obtain optimal normalization parametersfor each tissue type.

Structural images in native space (not normalized) were segmented intogray matter, white matter, and CSF compartments. During this process, a fullyautomated procedure³⁷ was applied to extracttissue of interest from nonbrain voxels (scalp, skull, or dural venous sinuses).Extracted images for each tissue type were normalized to their tissue-specifictemplates. The optimized normalization parameters obtained for each tissuecompartment in this step were applied to the original whole-brain images.The resulting 3 models were resliced to a final voxel size of 1.5 mm³. Each optimally normalized tissue-specific whole-brain model was thensegmented to isolate the corresponding tissue compartment. In this latterprocess, automated extraction of residual nonbrain voxels was again applied.

Spatial normalization inherently reduces size variations between differentbrains. To restore original volume information within each voxel, voxel valuesin the segmented images were modulated (multiplied) by the Jacobian determinantsderived from the spatial normalization step. The Jacobian determinant of avoxel is a numerical factor reflecting volume changes (expansions or contractions)occurring when an image from a subject is warped to match the template. If,for example, a brain structure has half the volume of that of the template,then its volume will be doubled during spatial normalization. In this case,modulation will restore original volume information by reducing voxel valuesto half. The analysis of modulated data, therefore, allows direct testingfor regional differences in the absolute amount (volume) of each tissue type.³⁷

Finally, the images were smoothed using a 12-mm full-width at half-maximumisotropic Gaussian kernel. This operation conditions the data to conform moreclosely to the Gaussian field model underlying the statistical proceduresused in regional analyses.⁴¹ After smoothing,the intensity in each voxel is a locally weighted average of tissue volumefrom a region of voxels defined by the size of the smoothing kernel.⁴⁰

The result of the whole process is that global brain shape variabilityis removed by spatial normalization, and different brains, adjusted to thesame stereotactic space, can be compared voxel by voxel. The comparison issensitive to volumetric differences in brain structures, as each voxel valueis adjusted for volume changes occurring in the normalization step (modulation).Therefore, a voxel value difference between 2 subject groups at a specificbrain coordinate means that the 2 populations differ in brain volume at thislocation (and at its surrounding tissue, as the data are finally smoothed).The accuracy in measuring volume changes is determined by precision of spatialnormalization that is typically less optimal for smaller structures in voxel-basedmorphometry. Spatial smoothing helps compensate for inexact normalization.A detailed explanation of what modulated voxel-based morphometry measuresand its inherent limitations is reported elsewhere.⁴²

Global brain volume measurements (obtained from the nonnormalized images)were compared in patients and control subjects using the independent samples t test.

The SPM99 tools were used to map regionally specific volume differencesthroughout the brain on a voxel-by-voxel basis. Comparisons between groupswere conducted separately for each tissue type. Data expressing absolute voxelvalues and data normalized for global differences in voxel intensity acrossscans were analyzed. A proportional scaling method was used to scale eachvoxel value to a grand mean of 100. Each group comparison generated 2 t statistic maps (SPM{t}) correspondingto 2 opposite contrasts: volume decrease and increase, displayed at a thresholdof P<.001. Regional differences were reportedto be significant at P<.05 after correction formultiple comparisons. Age was introduced as a regressor (covariate) in thecomparisons between patients and control subjects. Pearson correlations andpartial correlations were used in the regional and clinical analyses. A possibleinteraction between age and group was specifically investigated by comparingage–tissue volume correlations between patients and control subjects.Spatial coordinates from all the obtained maps were converted to standardTalairach coordinates⁴³ using a nonlinear transformof SPM99 standard space to Talairach space.⁴⁴

Mean ± SD intracranial volume was similar in patients with OCDand control subjects (1407 ± 167 and 1412 ± 145 mL, respectively; t₁₄₂ = −0.2; P =.86). Patients showed a tendency toward smaller gray matter volumes, witha mean difference of 24 mL (mean ± SD, 739 ± 82 mL in patientsand 763 ± 78 mL in controls; t₁₄₂ =−1.8; P = .07).

Compared with control subjects, patients with OCD showed significantabsolute decreases in gray matter volume in the medial frontal gyrus (involvingthe anterior paracingulate cortex), the medial aspect of the orbitofrontalcortex, and the left insulo-opercular region (Figure 1). The posterior paracingulate cortex (at the precuneus-cuneusjunction) and the insulo-opercular region of the right hemisphere showed atendency for gray matter reduction, but differences in these regions werenot statistically significant after correcting for multiple comparisons.

No regions in patients with OCD showed absolute increases in gray mattervolume. Statistically significant increases, however, were observed afternormalizing the data to global gray matter content. We observed relative increasesin gray matter volume bilaterally in the ventral part of the striatum, includingthe ventral aspect of the putamen and the area of the ventral striatum proper(nucleus accumbens and olfactory tubercle), and in the anterior cerebellum(anterior lobule of the cerebellar vermis and part of the left anterior cerebellarhemisphere) (Figure 2).

Figure 3 shows gray mattervolume decreases and increases in 11 representative sagittal views. Theseimages summarize and expand the description of our main results by mappingthe distribution of the observed OCD brain changes more comprehensively.

No region was found to show statistically significant white matter decreasesor increases in patients with OCD compared with controls.

No region in the CSF spaces showed statistically significant volumereductions. Patients, however, did show volume increases involving the medialorbitofrontal and left perisylvian CSF spaces. This CSF pattern broadly matchedthe gray matter findings, as CSF increases were adjacent to the regions showingstatistically significant gray matter volume decreases. For left perisylvianCSF spaces, however, the observed changes were somewhat more anteriorly located(Table 2).

Table 2 reports significantpeak differences in voxel volumes observed between patients with OCD and controlsubjects. Each peak corresponds to the voxel showing the most significantdifference, although the changes do not refer exclusively to the reportedcoordinates. Each voxel receives a weighted effect of a voxel region showinga 12-mm diameter (at full-width at half-maximum).

All comparisons were repeated with slice thickness in the regressionas a covariate to exclude a possible confounding effect of slice thicknessvariations across participants. The reported results did not change afterthis analysis.

We investigated whether volume variations in the regions differing betweenpatients with OCD and controls were related to each other. A correlation analysiswas performed using the voxel values corresponding to the 6 peak coordinatesidentified in the gray matter analysis (Table 3). A definite pattern of correlations was observed in patientsshowing a significant inverse correlation between the subcortical (right andleft striatal regions) and cortical (medial frontal, orbitofrontal, and insulo-opercular)structures. In addition, positive correlations were found between corticalregions and between right and left striatal areas. In the control group, thispattern was not found, and the only significant finding corresponded to apositive correlation between the striatal areas.

We also investigated whether volume variations in the regions of interestcorrelated with other parts of the OCD brain. Separate voxel-based brain mapswere performed for the 6 gray matter regions using the peak coordinate valueof each region as a predictor variable. Few findings were obtained in thisanalysis. The right striatal region positively correlated with a bilateralarea involving the posterior cingulate cortex (peak at Talairach x, y, z:7, −44, 15 mm; r = 0.53, t₆₉ = 5.2; corrected P = .01) andinversely correlated with a diencephalic area that included part of the medialand anterior thalamus bilaterally (peak at Talairach x, y, z: 0, −3,3 mm; r = −0.54, t₆₉ = 5.4; corrected P = .008).

In a post hoc analysis, we tested whether volume increases or decreasesoccurred in the OCD posterior cingulate and thalamus at lower significancethresholds. We did not find any difference between patients and control subjectsin this analysis.

We observed a different effect of age on tissue volume for both studygroups (Table 4 and Figure 4). Patients with OCD showed a preservation (absence of age-relatedvolume decrease) of the ventral striatum areas bilaterally, although the effectwas particularly relevant for the right side (age × group interactionwas significant at Talairach x, y, z: 21, 6, –4 mm; t₁₄₀ = 5.41; corrected P = .001).In contrast, the effect of age on the cortical regions was similar for patientsand control subjects. In both groups, gray matter volume decreased with ageat a comparable rate. Age-related reduction of gray matter was a general effectin the brain. The correlation of age with the total gray matter volume was r = −0.41 in patients with OCD and r = −0.48 in control subjects (P<.001for both).

Illness duration effects were comparable to those observed for age.We found, for example, that this variable correlated inversely with totalgray matter volume (r = −0.48; P<.001) and with normalized right striatal area (r = 0.50; P<.001). Age and illness durationwere strongly related in this OCD sample (r = 0.84; P<.001). No significant effect was observed when theage of patients at the onset of OCD was analyzed.

Separate voxel-based analyses were conducted for sex, disease severity(Yale-Brown Obsessive-Compulsive Scale scores), comorbid anxiety and depressionsymptoms (Hamilton scores), treatment type (including medication and behavioraltherapy, current and past history), number of treatments received, and totaltreatment duration. In no case did these variables significantly predict anatomicalchanges in the OCD brain.

Scores on the 5 defined symptom dimensions were used as predictors infurther analyses attempting to map brain changes associated with a given factor.We found no significant correlation between these scores and the values ofvoxels included in the abnormal regions. Similarly, we found no significantdifference for these regions when comparing patients with symptoms for a givendimension with the rest of the sample.

We made one relevant finding when the voxel-based maps obtained in thisanalysis were checked to identify possible changes in other brain areas. Patientswith "prominent" aggressive obsessions and checking compulsions (factor 4)showed a relative decrease in gray matter volume in the right amygdala regioncompared with the rest of the OCD sample (Figure 5). Differences at peak coordinates (Talairach x, y, z: 18,−9, −13 mm) were significant (t₆₀ = 5.3; corrected P = .01), adjusted to brainvolume, age, and sex. These 30 patients also showed a smaller right amygdalaregion than their 30 matched control subjects in relative and absolute terms(t₁₂₀ = 5.8; corrected P<.001, adjusted to brain volume; t₁₂₆ = 4.5; corrected P = .04, for nonadjustedcomparison). The rest of the sample, however, did not differ from their controlcounterparts.

Voxel-based mapping of structural brain alterations in patients withOCD revealed statistically significant reductions of gray matter volume inthe medial frontal and orbitofrontal cortices and in the left insulo-opercularregion and a relative volume increase in the ventral part of the striatumand anterior cerebellum. Cortical and subcortical changes were abnormallycoupled in OCD, and age statistically significantly contributed to the relativeenlargement in the ventral striatum area. The clinical expression of OCD wasnot related to these anatomical changes, although patients with aggressiveobsessions and checking compulsions showed a reduced right amygdala volume.

Early imaging studies typically evaluated selected cerebral regionsin few patients and provided notably heterogeneous results.⁸ Morerecently, Jenike et al²¹ used 3-dimensionalMRI and parceled the entire brain in several regions of interest. They suggestedthat structural brain alterations in OCD may be widely distributed. Kim etal²² assessed the OCD brain using voxel-basedmeans and analyzed MRI studies of 25 patients with no correction for multiplecomparisons. They identified brain regions showing altered gray matter "density"(direct volume testing was not possible until recent technical improvements)and found alterations distributed over multiple areas of the 4 cerebral lobesand diencephalon. Their results only marginally coincide with our data, butboth studies may be considered complementary.

In keeping with the increasing interest in the comprehensive assessmentof OCD brain anatomy, we used highly automated voxel-based morphometric proceduresand examined the entire cerebral parenchyma in a large number of patients(the largest structural MRI study to date in OCD). We adopted an improvedmethod for modulating data that allowed direct volume measurements. In ouranalysis, voxel values specifically expressed variations in the absolute amountof brain tissue, in contrast with previous voxel-based studies in which theinterpretation of results was not self-evident, as voxel values expressedvariations in structure shape and tissue composition.

These study characteristics may well improve the probability of capturingthe more prevalent structural alterations in OCD. Moreover, some relevantfeatures of the OCD brain may probably be detected on a voxelwise basis only.Indeed, structural changes in patients do not exactly match specific cerebralnuclei or definite gyri but instead involve divisions of anatomical structures.This finding may in part explain the difficulty in identifying structuralbrain alterations in the studies that used anatomical boundaries to delimitthe regions of interest.⁸ Nevertheless, voxel-basedmorphometry is still limited in that brain normalization and severe smoothingmay cause information loss and degrade anatomical details in small structures.In addition, there is a real need to ascertain to what extent the resultsobtained with volume-modulated voxel-based approaches are comparable to thoseof conventional morphometry. We would, however, point out that the studiesconducted in large samples in other neuropsychiatric populations suggest thatstandard and voxel-based methods produce compatible findings.⁴⁵

The proposed models of OCD symptom mediation generally coincide in thatbrain dysfunction does not involve the entire frontosubcortical system butrather the ventral striatum and functionally related orbitofrontal and cingulatecortices specifically.^{2,7- 8,46- 49} Wedetected alterations in the ventral part of the striatum and the medial (orbitaland paracingulate) aspect of the frontal lobes. Thus, it seems that alterationsin the "limbic" component of the frontosubcortical system may also be observedin the structural domain. It is relevant, however, that we found volume reductionin the cortex and relative expansion of tissue at the subcortical level, whereasin the functional studies,^7- 9 bothlevels seem hyperactive.

Our interregional correlation analysis revealed that cortical and striatalalterations are coupled in OCD. The observed inverse correlation between the2 brain levels is pathologic as it was not present in control subjects. Suchan analysis may indicate that the whole pattern of alterations detected inthe study reflects a global disorder in a large-scale system as opposed toa simple sum of isolated changes. Baxter et al⁴⁸ alsofound abnormal frontosubcortical correlations in OCD in their early positronemission tomographic assessment. The correlation disappeared with symptomimprovement in their treatment-responsive patients.

Prevailing hypotheses for the pathogenesis of OCD propose that a defective,⁴⁸ imbalanced,⁷ hypertonic⁴⁷ striatum originates OCD symptoms by disturbing thefrontosubcortical balance in the striatal loops. Our age analysis may provideevidence for the dynamic nature of the striatal disturbance, as it suggeststhat subcortical alterations progress throughout the age period studied. Inthis context, structural changes in the striatum could be the anatomical expressionof enduring striatal dysfunction. Age did not contribute to the occurrenceof OCD cortical alterations. In both study groups, cortical gray matter volumedecreased with age at a comparable rate. Therefore, early structural changesand changes developing during the illness may coexist in OCD.

The frontostriatal system serves to modulate behavioral responses andworks together with other brain systems, such as the cerebellum.⁵⁰ Thecerebellar hemispheres affect fine motor and cognitive responses, whereasthe medial rostral cerebellum affects arousal, autonomic behavior, and emotionalresponsiveness.^51- 52 The anteriorvermis and the associated fastigial nucleus are considered to be the "limbiccerebellum," which is directly connected to the ventral tegmental area thatprovides dopamine to the ventral striatum and facilitates neural activityin the septal region (reviewed by Schmahmann⁵²).We found relative gray matter volume increases in the ventral part of thestriatum and in the functionally related rostral cerebellum. This findingmay suggest cerebellar involvement in the pathogenesis of OCD.

At first sight, the changes observed in the left posterior insula andadjacent opercular tissue may seem difficult to harmonize with the other alterations.This region of the brain has traditionally been related to language operations,but recent studies have provided a wealth of data suggesting a key role inthe perception of the body scheme, mediation of pain responses, visceral awareness,and evaluation of the affective components of sensory information.⁵³ We do not know the intrinsic mechanisms that mayfunctionally link the insulo-opercular changes to the alterations occurringin the frontostriatal system in OCD. Anatomically, however, the relationshipis direct, as the frontoparietal operculum and the insula send strong connectionsto the ventral part of the striatum.^54- 55

The operculo-insular system processes sensory inflow before reachingthe amygdala.^53,56 Therefore,in this context, the amygdala (a third party in the genesis of compulsions⁴⁹) is doubly prone to hyperactivation in OCD, fromreduction of the inhibitory tone provided by an altered orbitofrontal cortexas proposed by Rauch et al⁴⁹ and from an excessiveinput of deficiently filtered sensory stimulation as a result of a "weak"operculo-insular system. In tune with this amygdalocentric perspective, wefound that aggressive obsessions and checking compulsions, a dimension involvingOCD behavior openly related to fear, were associated with a smaller rightamygdala area. Szeszko et al²⁰ previously reportedreduced right amygdala size in OCD, although their finding was not ascribedto a particular clinical subtype.

We attempted to link the observed findings with the core of the OCDphenomenon. Nevertheless, we should alternatively consider treatment as apotential contributor to regional brain size variations. Volume increasesin basal ganglia have been reported as a consequence of prolonged neurolepticadministration,⁵⁷ and, furthermore, increasedthalamic volume was observed in medication-free patients^15,22 normalizingwith paroxetine monotherapy.¹⁵ Although wefound no association between the administered treatments and brain changespresent in our patients, medication could be responsible for obtaining negativefindings for the thalamus. It is possible in this study, as in others,²¹ that active treatment at the time of imaging preventedus from obtaining the increased thalamic volume reported for medication-freepatients. Another possible study limitation is that 10 patients had receivedexperimental treatment with transcranial magnetic stimulation and that thelong-term neural consequences of such treatment are poorly understood.

Although we found structural alterations in brain systems metabolicallyaltered in OCD, there is no complete anatomical coincidence between our resultsand functional data. Hypermetabolism in positron emission tomographic studiesis also observed in more lateral aspects of the orbitofrontal cortex and inmore medial aspects of the ventral striatum.⁹ Partof such a discordance could perhaps be explained by technical arguments, butit is also feasible that functional and structural approaches emphasize, inpart, different aspects of the disorder. Our data may complement functionalstudies mainly by further delimiting the anatomy of changes, by showing thateither decreases or increases in tissue volume are compatible with hypermetabolism,and by suggesting that brain alterations in OCD may progressively vary duringthe disease.

Correspondence: Jesús Pujol, MD, Magnetic Resonance Centerof Pedralbes, J Anselm Clavé 100, Esplugues de Llobregat, Barcelona08950, Spain (jpujol@cetir.es).

Submitted for publication June 3, 2003; final revision received January21, 2004; accepted January 29, 2004.

This study was supported in part by grants 00/0226 and PI020102 fromthe Fondo de Investigación Sanitaria, Madrid, Spain; by the FundacióLa Marató TV3, Barcelona; and by grants 1999SGR-328 and 2000XT-43 fromthe Direcció General de Recerca de la Generalitat de Catalunya, Barcelona.

We thank Gerald Fannon, PhD, for revising the manuscript.