RiLa

Bgaw(

Mpb

RriCs

Cadf

Ks

Bw2caeIspcodoswhD

F

A

R

0d

egional Cerebral Glucose Metabolic Abnormalitiesn Bipolar II Depressioninda Mah, Carlos A. Zarate Jr, Jaskaran Singh, Yu-Fei Duan, David A. Luckenbaugh, Husseini K. Manji,nd Wayne C. Drevets

ackground: Functional neuroimaging studies of bipolar disorder (BD) performed in conjunction with antidepressant treatment trialsenerally require that patients remain on mood stabilizers to reduce the risk of inducing mania; yet, it is unknown whether the metabolicbnormalities evident in unmedicated BD depressives remain detectable in patients receiving mood stabilizers. This study investigatedhether cerebral metabolic abnormalities previously reported in unmedicated BD subjects are evident in depressed bipolar disorder type II

BD II) subjects receiving lithium or divalproex.

ethods: Using [18F]-fluorodeoxyglucose-positron-emission tomography, cerebral glucose metabolism was compared between 13 de-ressed BD II subjects on therapeutic doses of lithium or divalproex and 18 healthy control subjects. Regional metabolism was comparedetween groups in predefined regions of interest.

esults: Metabolism was increased in the bilateral amygdala, accumbens area, and anteroventral putamen, left orbitofrontal cortex andight pregenual anterior cingulate cortex in depressives versus control subjects. Post hoc exploratory analysis additionally revealedncreased metabolism in left parahippocampal, posterior cingulate, and right anterior insular cortices in depressives versus control subjects.orrelational analyses showed multiple limbic-cortical-striatal interactions in the BD sample not evident in the control sample, permittingensitive and specific classification of subjects by discriminant analysis.

onclusions: These results confirm previous reports that bipolar depression is associated with abnormally increased metabolism in themygdala, ventral striatum, orbitofrontal cortex, anterior cingulate, and anterior insula, and extend these results to bipolar disorder type IIepressives on lithium or divalproex. They also implicate an extended functional anatomical network known to modulate visceromotor

unction in the pathophysiology of BD II depression.

ey Words: Cerebral metabolism, bipolar disorder, depression,triatum, prefrontal cortex, amygdala, mood stabilizers

ipolar disorder (BD) is a chronic, disabling condition witha reported 12-month prevalence of 1% to 3% (Kessler et al2005). Much of the morbidity and mortality associated

ith BD is attributable to the depressive phase (Calabrese et al003), which manifests a substantially more chronic and severeourse and higher suicide rate in bipolar disorder type II (BD II),s compared with bipolar disorder type I (BD I) depression (Juddt al 2003). Yet, little is known about the pathophysiology of BDI. Functional neuroimaging studies of mood disorders havehown that physiological activity is abnormal in limbic andaralimbic structures such as the amygdala, ventral anterioringulate cortex (ACC), and anatomically related areas of therbitofrontal cortex (OFC), striatum, and thalamus during majorepressive episodes, but the majority of these studies focusedn unipolar depression. Few neuroimaging studies have as-essed neurophysiological activity in unmedicated subjectsith bipolar depression, and none of the studies specificallyave limited the study sample to BD II (reviewed in Ketter andrevets 2002).

rom the Section on Neuroimaging in Mood and Anxiety Disorders (LM,Y-FD, WCD), Molecular Imaging Branch, and Mood and Anxiety Disor-ders Program (CAZ, JS, DAL, HKM), National Institute of Mental Health,National Institutes of Health, Bethesda, Maryland.

ddress reprint requests to Linda Mah, M.D., Kunin-Lunenfeld Applied Re-search Unit, Posluns Building, Room 738, Baycrest Centre for GeriatricCare, 3560 Bathurst St., Toronto, Ontario, Canada M6A 2E1. E-mail:mahl@intra.nimh.nih.gov.

eceived January 3, 2006; revised June 9, 2006; accepted June 10, 2006.

006-3223/07/$32.00oi:10.1016/j.biopsych.2006.06.009

Previous positron-emission tomography (PET) studies of un-medicated bipolar depressives that combined BD I and BD IIcases demonstrated abnormally elevated metabolic activity in theamygdala, ventral striatum, and anatomically related limbic andparalimbic cortices. Drevets et al (1995a, 2002) found signifi-cantly increased normalized (regional/global) metabolism in theleft amygdala, with a trend toward hypermetabolism in the rightamygdala and the right ventral striatum in unmedicated bipolardepressives (n � 7, 4 BD II) relative to healthy control subjects.The same BD sample had decreased flow and metabolism in thesubgenual ACC, an area where decreased perfusion also hadbeen identified in an independent sample of bipolar depressives(Drevets et al 1997).

In a larger sample of unmedicated, treatment-resistant, bipo-lar depressives (n � 17, 13 BD II) who were moderately toseverely depressed, Ketter et al (2001) also found increasednormalized metabolism in the right amygdala, bilateral ventralstriatum, medial thalamus, and the medial cerebellum relative tohealthy control subjects. Ketter et al (2001) also reported that inmildly depressed BD cases (n � 16, 9 BD II), normalizedmetabolism was increased in the posterior cingulate cortex, leftventrolateral prefrontal cortex (PFC), left middle and superiorfrontal gyri, left insula, hippocampus, left postcentral gyrus, lefttransverse temporal gyrus, and cerebellum, and decreased in theright inferior and middle temporal gyri. This study also reportedregional differences in absolute metabolism, but the interpreta-tion of these data was confounded by a significant reduction inwhole brain metabolism in BD subjects relative to control subjects.Moreover, it was noteworthy that studies of unmedicated bipolardepressives identified abnormally increased normalized metabolismin many of the same regions where metabolism was elevated inunmedicated unipolar depressives, including the amygdala, ventral

striatum, medial thalamus, ventrolateral PFC, OFC, pregenual ACC,

BIOL PSYCHIATRY 2007;61:765–775© 2007 Society of Biological Psychiatry

a(

tbttcrrdstmsdhi

(dslMBtrAlitba

tpduiDleaamrw

M

S

BcpbtasSaRo

766 BIOL PSYCHIATRY 2007;61:765–775 L. Mah et al

w

nterior insula, posterior cingulate cortex, and medial cerebellumreviewed in Drevets 2000).

Insofar as these abnormalities constitute biologically relevantargets for treatment, the ability to assess how they are affectedy antidepressant treatments is an important goal in investiga-ions aimed at elucidating therapeutic mechanisms. However,he risk that such treatments would trigger hypomanic or maniconversion in BD cases generally requires that such subjectseceive mood stabilizers during antidepressant treatment trials. Itemains unclear, however, whether the metabolic abnormalitiesescribed above would be evident in subjects receiving mood-tabilizing agents. Antidepressant, antipsychotic, and antianxietyreatments have been reported to reduce cerebral blood flow andetabolism in some frontal, parietal, and temporal lobe regions,

uch that many studies of patients on these medications have notetected areas of abnormally elevated metabolism or insteadave reported areas of reduced flow or metabolism not evidentn studies of unmedicated samples (reviewed in Drevets 2000).

Preliminary data in the study conducted by Drevets et al2002) suggested that mood stabilizer treatment reduced amyg-alar metabolic activity toward normative levels in a smallample of clinically remitted subjects with BD but mood stabi-izer treated subjects who remained depressed were not studied.ore recently, Bauer et al (2005) found that currently depressedD patients (n � 10, 1 BD II) being medicated with a combina-

ion of antidepressants and mood stabilizers exhibited increasedelative metabolism versus control subjects in the subgenualCC, right amygdala, right hippocampus, right ventral striatum,

eft thalamus, and cerebellar vermis, and decreased metabolismn the middle frontal gyri bilaterally. However, the variability inype, combination, and dosage of medications and the lack oflood drug concentrations (to ensure that therapeutic levels werechieved) limited the interpretability of these data.

The current study used [18F]-fluorodeoxyglucose (18FDG) PETo assess cerebral metabolic rates for glucose (CMRglu) in BD IIatients who remained depressed while receiving therapeuticoses of mood stabilizers. Based on previous PET studies ofnmedicated depressed patients and preclinical studies of lith-um and divalproex sodium (Bauer et al 2005; Drevets et al 2002;u et al 2003; Ketter et al 2001), BD II depressives on therapeutic

evels of mood stabilizers in the current study were expected toxhibit abnormally elevated metabolic activity in similar regionss found in unmedicated depressed samples, including themygdala and anatomically related areas of the orbital andedial PFC and the ventral striatum. Associations between

egional glucose metabolism and depression and anxiety ratingsere examined post hoc.

ethods and Materials

ubjectsPatients (n � 13; 11 female subjects) met DSM-IV criteria for

D type II and for a current major depressive episode. Healthyontrol subjects (n � 18, 15 female subjects) with no history ofsychiatric illness also participated. Diagnoses were establishedy an unstructured interview with a psychiatrist and the Struc-ured Clinical Interview for DSM-IV-Patient Edition (SCID-P) withsecond clinician. Patients were included if they had an initial

core of �20 on the Montgomery-Asberg Depression Ratingcale (MADRS) (Montgomery and Asberg 1979). Severity ofnxiety symptoms also was rated using the Hamilton Anxietyating Scale (HAM-A) (Hamilton 1959). All ratings were obtained

n the day of the PET scan. Exclusion criteria included current

ww.sobp.org/journal

psychotic features, serious suicidal risk, substance abuse within90 days, substance dependence within 5 years, rapid cyclingcourse within 12 months, major medical or neurological disor-ders, pregnancy, and lactating female subjects. Patients wererecruited as part of a clinical trial studying the effect of adding anadjunct antidepressant medication to mood stabilizer treatmentand were required to take a stable dose of either divalproexsodium or lithium for at least 4 weeks, during which two weeklyblood drug levels fell within the therapeutic range before imag-ing (plasma valproic acid concentrations of 50 to 125 ug/mL;serum lithium concentrations of .6–1.2 mEq/L). Other psycho-tropic medications and all medications likely to influence cere-bral physiology, perfusion, or metabolism were not permittedwithin 2 weeks of scanning. Subjects provided written informedconsent, as approved by the National Institute of Mental HealthInstitutional Review Board (NIMH-IRB).

Image AcquisitionQuantitative whole brain and regional metabolic measures

were obtained using a technique that combined left cardiacventricular chamber time-activity curve data with venous bloodsampling to yield the input function for calculating CMRGlu. Thismethod previously was validated against methods that employedarterial blood sampling to generate the 18FDG input function(Moore et al 2003). Positron-emission tomography images wereacquired using a GE Advance PET scanner (GE Medical Systems,Waukesha, Wisconsin) (35 contiguous slices 4.25 mm thick; axialresolution � 4.9 and 5.3 mm full-width at half maximum [FWHM]in two-dimensional [2-D] and three-dimensional [3-D] modes,respectively). Subjects received 4.5 mCi of 18FDG following afasting period of at least 6 hours. Following an initial transmis-sion scan of the chest to permit measured attenuation correctionof the cardiac emission scan, a 35-minute dynamic emission scanof the heart was acquired in 2-D mode (10 30-second frames and10 3-minute frames), with concurrent serial venous blood sam-pling beginning 15 minutes posttracer injection. At 45 minutesposttracer injection, a 10-minute static emission scan of the brainwas acquired in 3-D mode, which immediately was followed byan 8-minute transmission scan of the head to perform measuredattenuation-correction of the emission scan.

An anatomical magnetic resonance image (MRI) was obtainedfor each subject using a 3.0 Tesla GE Signa Scanner (GE MedicalSystems, Waukesha, WI) and a 3-D MPRAGE (echo time [TE] �2.982 milliseconds, repetition time [TR] � 7.5 milliseconds,inversion time � 725 milliseconds, voxel size � .9 � .9 � 1.2mm) for co-registration of PET images.

Image AnalysisTo quantify whole brain and regional CMRGlu from 18FDG

emission images, the cardiac input function was derived bycombining left cardiac ventricular chamber time-activity curvedata with venous blood sampling, as detailed in Moore et al(2003). Briefly, cardiac slices were reconstructed and five leftventricular slices were identified for region-of-interest (ROI)placement. The cardiac image frames acquired from 0 to 5minutes initially were averaged to allow localization of the leftventricular blood pool, while the frames acquired between 25and 35 minutes permitted identification of myocardial wall18FDG uptake. Circular ROIs of 2 cm diameter were positionedover the left ventricular chamber on difference images obtainedby subtracting the left ventricular myocardial 18FDG uptake fromthe blood pool image to minimize spillover of radioactivity from

the myocardium. An average left ventricular time-activity curve

wRt3vrtco(de(itsnmu(

as(NiMGg

ntApdDTabatb(poslrravatarcodc

app

L. Mah et al BIOL PSYCHIATRY 2007;61:765–775 767

as computed from the time-activity curves obtained from theOI in each of the five slices. The ratio of the average radioac-ivity concentration in venous blood samples taken at 25, 30, and5 minutes to the average radioactivity concentration in the leftentricular blood pool from 25 to 35 minutes was calculated. Thisatio was used to scale the 50-minute venous sample concentra-ion, which was then appended to the left ventricular curve toomplete the input function used to generate parametric imagesf regional cerebral metabolic rates for glucose (rCMRGlu)Brooks 1982). An image of quantitative glucose metabolism waserived from the regional tissue radioactivity in the cerebralmission scan and the input function as described in Moore et al2003). This image was aligned to the magnetic resonance (MR)mage, which was resliced to yield images with the same slicehickness and voxel size as the PET images using AIR (Ahman-on-Lovelace Brain Mapping Center, UCLA, Los Angeles, Califor-ia; Woods et al 1993). The whole brain metabolism waseasured as the average CMRGlu within a template definedsing the co-registered MRI image to define the brain edgeDrevets et al 2002).

For analysis of regional metabolism, each subject’s MRI imagend co-registered PET image were spatially normalized to thetandard stereotaxic space of the Montreal Neurological InstituteMNI) template using SPM99 (Wellcome Department of Imagingeuroscience, London, United Kingdom). Spatially normalized

mages were resampled to a voxel size of 2 � 2 � 2 mm.etabolic images were filtered using a 12 � 12 � 12 mmaussian smoothing kernel within SPM99. Nonspecific effects oflobal activity were removed by proportional scaling.

The a priori hypotheses were tested by comparing theormalized regional metabolism between depressives and con-rol subjects within 12 predefined ROIs: the amygdala, OFC,CC, and three striatal regions: the accumbens area, caudate andutamen in the left and right hemispheres. These regions wereelimited using standard anatomical criteria (Breiter et al 1997;revets et al 2001; Duvernoy 1999; Szeszko et al 1999, 2002;alairach and Tournoux 1988) on a single MNI template andpplied to all normalized brains at the group level contrasts. Theoundaries of the amygdala were defined anteriorly, laterally,nd ventrally by the temporal lobe white matter, posteriorly byhe temporal horn of the lateral ventricle and the alveus, dorsallyy the basal ganglia, and medially by the cerebrospinal fluidCSF) space. The boundaries of the ACC ROI were (anterior,osterior, ventral, dorsal) tip of the cingulate sulcus, connectionf the superior and precentral sulci, callosal sulcus, and cingulateulcus. The boundaries of the OFC ROI were (anterior, posterior,ateral, and medial) last appearance of the anterior horizontalamus, last appearance of the olfactory sulcus, anterior horizontalamus or circular sulcus of insula, and the olfactory sulcus. Theccumbens area encompassed the gray matter ventral to theentral tip of the internal capsule and was delimited dorsally byline connecting the ventral aspect of the lateral ventricle and

he ventral tip of the internal capsule and posteriorly by thenterior commissure. The caudate ROI extended from the ante-ior aspect of the caudate head posteriorly to the coronal planeontaining the lateral geniculate nucleus. The putamen wasutlined from the posterior boundary of the accumbens ROIorsally and posteriorly as delimited laterally by the externalapsule, medially by the internal capsule and the globus pallidus.

Within each ROI, the mean normalized metabolism for the BDnd control groups was compared voxel-by-voxel using un-aired t tests to identify voxels with t-values corresponding to

� .05. Correction for multiple comparisons was applied for

each ROI using the cluster test, which tests the probability of typeI error based on the spatial extent of clusters of adjacent voxelsfor which the voxel t-values correspond to p � .05 (Poline et al1997). The number of voxels used to compute the correctedp-values in the cluster-level comparisons was limited using theSmall Volume Correction option within SPM99. To adjust for thenumber of ROIs searched, the resulting p-values also weredivided by 12 (the number of ROIs), following the methoddescribed by Bauer et al (2005) for applying Bonferroni correc-tion to multiple ROIs assessed by the Small Volume Correctionapproach (but see Discussion).

To address the probability of type II error, the whole brainwas searched post hoc using an exploratory voxelwise analysisto identify clusters of voxels with t-values corresponding topuncorrected � .001 that also exceeded the “minimum expectedcluster size” (as calculated by SPM99 based on Gaussian randomfield theory and the smoothness of the image data and listed inthe SPM99 output file) (Friston et al 1994). The significancethreshold was set at p � .001 to reduce the likelihood of type Ierror because of the larger number of statistical comparisonsperformed for the whole brain analysis.

Another post hoc analysis was performed to assess thecovariance of regional metabolism across the areas where meanmetabolism differed between groups within the predefined ROI.The normalized metabolism was extracted for each subject fromthe voxel containing the peak t-value within each region (thefinal spatial resolution of such voxel measurements after recon-struction and filtering was much larger than that of individualvoxels), and the covariance between metabolic activity acrossregions was assessed by computing Pearson’s product momentcorrelation coefficients using SPSS Inc. (Chicago, Illinois). To testthe significance of differences in intercorrelations betweenregional metabolism in patients versus control subjects, z testson correlations transformed using Fisher’s z transformationwere performed (Rosenthal 1991). Because these analyses wereconsidered exploratory, the significance threshold for correla-tions performed within groups and the z tests comparing corre-lations across groups was set at p � .05.

Table 1. Demographic and Clinical Characteristics of Depressed andControl Samples

Bipolar IIDepressed

ControlSubjects

Number 13 18Gender (M/F) 2/11 5/13Agea (Mean Years � SD) 43 (� 8.4) 39 (� 8.0)IQb 117 (� 6.6) 116 (� 9.1)MADRS (Mean � SD) 32.2 � 4

Range (26–39)HAM-A (Mean � SD) 17.2 (� 3.9)

Range (10–24)Age of Onset (Mean Years � SD) 20 (� 10.5)Length of Illness (Mean Years � SD) 22.9 (� 12.0)Length of Current Episode

(Mean Months � SD) 4.2 (� 2.0)

IQ, intelligence quotient; MADRS, Montgomery-Asberg DepressionRating Scale; HAM-A, Hamilton Anxiety Rating Scale.

aNo significant difference in age between patients and control subjects[t (29) � 1.1, p � .28].

bNo significant difference in IQ between patients and control subjects[t (22) � .26, p � .80]. Data unavailable for 3/18 control subjects and 4/13

patients.

www.sobp.org/journal

mtaStvoolnrvoewm

sultw

768 BIOL PSYCHIATRY 2007;61:765–775 L. Mah et al

w

To optimize the pattern of regions where the covarianceatrix of metabolic activities provided the greatest discrimina-

ion between individual depressed cases versus control subjects,stepwise discriminant function analysis was performed using

PSS. Normality was examined using the Kolmorgorov-Smirnovest and visual review of histograms and boxplots. Outlyingalues with potential for having undue influence on the outcomef any statistics were adjusted by reducing the values to withinne unit above the next highest value or one unit below the nextowest unit, depending on the end of the scale concerned (Tabach-ick and Fidell 1989). Stepwise entry criteria were set at p � .05 withemoval at p � .10 using Wilks lambda. A leave-one-out cross-alidation method was used to verify the sensitivity and specificityf the model. Following the stepwise procedure, the variables notntering the model were added individually to determinehether they would improve the sensitivity and specificity of theodel.To identify areas where metabolism correlated with illness

everity, linear regression analyses were performed post hocsing SPM99 to assess relationships between regional metabo-ism and depression or anxiety ratings. The statistical significancehreshold was set at a voxelwise probability level of p � .01ithin the predefined ROIs and p � .005 for other regions.

Table 2. Comparison of Regional Metabolism in Bipolaof Interest Defined A Priori

Region of Interest

Stereotaxic Coordinates

x y

AmygdalaLeft �24 �1 �Right 22 �1 �

Orbitofrontal CortexLeft �18 17 �

�30 30 ��22 34 �

Right 55 35 �16 34 �

Anterior CingulateLeft �6 48 �

�4 18Right 12 47

Accumbens AreaLeft �14 11 �Right 16 11 �

CaudateLeft �10 10Right 14 10

PutamenLeft �22 8 �Right 26 8 �

30 418 8

aStereotaxic coordinates locate peak voxel t-valuespositive x and y indicating right and anterior, respectiveland posterior commissures. Voxel coordinates were traspace using the Talairach Daemon database (http://ric.u

bThe voxel-level t-value for the voxel containing thmetabolism in bipolar II subjects relative to control subjbipolar II subjects relative to control subjects.

cThe cluster-level corrected p-value represents the plarge or larger (number of voxels) within the region of incomparisons (Poline et al 1997).

dSignificant after correction for multiple comparisons usin

ww.sobp.org/journal

Results

Demographic and clinical characteristics of the subjects ap-pear in Table 1. The depressed and control samples were similarin gender composition, mean age, and intelligence quotient (IQ).

With respect to comorbid conditions in the BD sample, twosubjects met DSM-IV criteria for past alcohol dependence (inremission � 5 years), one for past polysubstance abuse (inremission � 3 months), one for obsessive-compulsive disorder,and one for eating disorder not otherwise specified. Threedepressives had mild medical conditions including chronic si-nusitis (n � 1), chronic low back pain (n � 1), and irritable bowelsyndrome (n � 1). The control subjects were medically healthy andhad no personal or family history of psychiatric illness.

Patients were receiving lithium or divalproex only. Serumlithium levels (n � 8, mean � .73 mEq/L, range .6–.9 mEq/L) andplasma divalproex levels (n � 5, mean � 75 ug/mL, range 54–90ug/mL) were within the therapeutic range. Control subjects wereunmedicated.

Whole Brain MetabolismThe mean whole-brain CMRGlu did not differ between groups

[BD II: .0560 � .015 mg/min/mL; control subjects: .0558 � .0068

pressed Subjects Versus Control Subjects the Regions

Tb

Voxel-LevelUncorrected

p-Value

Cluster-LevelCorrectedp-Valuec

2.12 .02 .041d

1.78 .04 .032d

2.93 .003 .003d

2.59 .0072.40 .012.12 .02 .322.11 .02 .31

2.23 .03 .096�2.70 .006 .14

2.62 .007 .009d

3.24 .001 �.001d

3.12 .002 �.001d

3.26 .001 .037d

2.97 .003 .061

4.35 �.001 �.001d

4.18 �.001 �.001d

4.09 �.0013.45 �.001

eir distance in mm from the anterior commissure, withpositive z dorsal to a plane containing both the anteriormed from Montreal Neurological Institute to Talairacha.edu).k difference. Positive t-values indicate greater regionalegative t-values indicate lower regional metabolism in

bility of obtaining any cluster with a spatial extent thist search volume after applying corrections for multiple

r II De

a

z

2018

141317

318

940

5

76

15

2205

by thy, andnsforthscs

e peaects. N

robateres

g the cluster test.

mrwn

Fsddpcbbhcptcr

L. Mah et al BIOL PSYCHIATRY 2007;61:765–775 769

g/min/mL, t (29) � �.063, p � .95]. Consequently, for theegional analyses, local metabolic values were normalized tohole brain metabolism to reduce the variability associated with

igure 1. Image sections from the statistical parametric mapping (SPM) analy-is showing regions with significantly greater metabolic activity in the bipolarisorder type II depressives versus the healthy control subjects. The SPM imagesemonstrate voxel t-values corresponding to a significance threshold set atuncorrected �.001. The horizontal sections run parallel to the horizontal planeontaining both the anterior and posterior commissures and are located at theicommissural plane (upper panel) and 16 mm (lower panel) ventral to thisicommissural plane. The coronal and sagittal planes run orthogonal to theorizontal planes and are located at 8 and 44 mm anterior to the anteriorommissure and 22 and 24 mm to the left of the midline in the upper and loweranels, respectively. This post hoc whole brain analysis showed elevated me-

abolism in left ventral striatum (A), right ventral striatum (B), left posterioringulate (C), left orbitofrontal cortex (D), left parahippocampal cortex (E), andight insula/claustrum (see Table 2 for coordinates and significance levels).

onspecific global fluctuations across subjects.

Regions-of-Interest AnalysisTable 2 summarizes the differences in regional metabolism

between groups within the predefined ROI. Metabolism wasincreased (pcorrected � .05 after application of the cluster test) inthe depressives versus the control subjects in areas within the leftand right amygdala, left OFC, right ACC, left and right accumbensarea, left and right putamen, and the left caudate. In other areas,metabolism appeared increased in the right OFC, left ventralACC, and right caudate, and decreased in the left dorsal ACC, ata voxel-level significance threshold of puncorrected � .05, butthese differences were not significant after applying the clustertest to correct for multiple comparisons within the ROI. Afteradditionally applying Bonferroni adjustment to correct for mul-tiple comparisons across the 12 ROI after Bauer et al (1995), onlythe differences in the left OFC, bilateral accumbens area, andbilateral putamen would have remained significant (but seeDiscussion).

Whole-Brain Exploratory AnalysesFigure 1 and Table 3 present the results of the post hoc,

voxelwise analysis comparing metabolism between groupsacross the entire brain. The depressives had greater metabolismthan control subjects in the anteroventral portion of the putamenbilaterally, right caudate, left posterior cingulate cortex/lingualgyrus, left OFC (anterior to the predefined OFC-ROI), rightanterior insula, and left parahippocampal gyrus (Table 3). Noareas were identified where metabolism was decreased signifi-cantly in the depressives versus the control subjects.

Exploratory Analyses of Relationships between RegionalMetabolism and Clinical Ratings

Results from the exploratory, voxelwise regression analysesare summarized in Table 4. Depression severity ratings (MADRSscores) correlated positively with metabolism in the left lateral

Table 3. Post Hoc Voxelwise Analysis of Regional Metabolism BetweenBipolar II Depressed Subjects and Control Subjects

Regiona

StereotaxicCoordinatesb

Tcx Y z

Depressives � Control SubjectsLeft Anteroventral Putamen �22 8 �2 4.35Right Anteroventral Putamen 26 8 �2 4.18Right Caudate Head 16 10 3 3.70Left Posterior Cingulate Cortex �20 �72 0 4.20

�22 �64 1 4.14�12d �77 7 3.48

Left Anterior Orbitofrontal Cortex �24 44 �16 3.95Left Parahippocampal Gyrus �22 �35 �7 3.72Right Anterior Insula 34 2 2 3.99

Control Subjects � DepressivesNo Suprathreshold Clusters

Areas where mean metabolism was increased in bipolar II subjects rela-tive to control subjects indicated by “Depressives � Control Subjects.” Noareas were identified where metabolism was decreased at p � .001 in thedepressives versus the control subjects.

aStereotaxic coordinates locate peak voxel t-values corresponding top � .001 that are part of clusters of voxels that exceeded the minimumexpected cluster size of 36 voxels per cluster (defined by SPM99).

bCoordinates interpreted as in Table 2.cThe voxel-level t-value for the voxel containing the peak difference.dLeft posterior cingulate cortex region lying along the calcarine and

parieto-occipital sulci.

www.sobp.org/journal

appiptmlechp

EA

chsobAtdama

T

B

R

R

p

770 BIOL PSYCHIATRY 2007;61:765–775 L. Mah et al

w

spect of the superior frontal gyrus (dorsal anterolateral PFC), leftrecentral gyrus, right superior temporal gyrus, and right inferiorarietal cortex but negatively with metabolism in the left anterior

nsula and left lateral OFC. Anxiety ratings (HAM-A) correlatedositively with metabolism in the left medial OFC but nega-ively with metabolism in the left lateral OFC, bilateral puta-en, left caudate, left ACC, left anterior and posterior insula,

eft frontal polar cortex, left postcentral gyrus, superior pari-tal lobule, left superior temporal cortex, and right occipitalortex (Table 4B). None of these clinical correlations wouldave remained significant after correction for multiple com-arisons.

xploratory Correlational Analyses of the Metabolic Valuescross Regions

Within the striatal ROI, the metabolism in each accumbens,audate, and putamen ROI correlated positively with that in theomologous ROI in the contralateral hemisphere in both depres-ives and control subjects (Table 5). The left accumbens metab-lism correlated positively with metabolism in the left caudate inoth groups but correlated (positively) with activity in the leftCC, left OFC, bilateral amygdala, and bilateral putamen only in

he depressives. The strength of the correlation coefficientsiffered across groups only for the association between leftccumbens and left ACC (p � .01). The right accumbensetabolism correlated positively with metabolism in the bilateral

able 4. Regions Where Metabolism Correlated with Depression and Anxie

Positive Correlation

rain Regiona

StereotaxicCoordinatesa

Tbx y z

egions Where Metabolism Correlated withDepression Severity

Left Superior Frontal Gyrus �18 35 39 3.84Left Precentral Gyrus �44 4 35 3.77

�26 �24 64 3.45Right Superior Temporal Cortex 65 0 0 3.69Right Inferior Parietal Lobule 44 �42 44 2.85

egions Where Metabolism Correlated withAnxiety Severity

Left OFC �20 40 �24 2.67

Depression severity rated by the Montgomery-Asberg Rating Scale; anxOFC, orbitofrontal cortex; ACC, anterior cingulate cortex.aStereotaxic coordinates, interpreted as in Table 2, locate peak voxel t-va

� .005 for other regions.bThe voxel-level t-value for the voxel containing the peak t-value.

mygdala, bilateral putamen, right caudate, left OFC, and left

ww.sobp.org/journal

ACC only in the depressives, but the strength of these correla-tions differed between groups only for the left OFC and bilateralputamen (p � .05). Metabolism in the putamen also correlatedpositively with metabolism in the amygdala bilaterally in thedepressives, but the strength of these correlations did not differsignificantly between groups. The left and right amygdala meta-bolic values were positively correlated in the depressives, andthe strength of this correlation differed between groups (p � .05).In the depressives, right amygdala metabolism correlated withmetabolism in the right ACC and OFC bilaterally, but thestrength of these correlations differed between groups onlyfor the right OFC (p � .05). Furthermore, the depressivesshowed correlations between left OFC activity and activity inthe left putamen and ACC, and the difference in the correla-tion coefficients between groups reached significance for theputamen. Finally, in the depressives, the left ACC metabolismcorrelated with left putamen metabolism and the right ACCmetabolism correlated with right OFC metabolism, but onlythe intercorrelation between left ACC and putamen differedbetween groups (p � .05).

Discriminant AnalysisThe stepwise discriminant analysis allowed the right putamen

and right OFC to enter the model to differentiate patients andcontrol subjects (Wald �2 � 18.42, df � 2, p � .0001). Table 6shows the sensitivity and specificity of this model together with

verity in Bipolar Subjects

Negative Correlation

Brain Regiona

StereotaxicCoordinatesa

Tb pax y z

1 Left Anterior Insula �34 24 14 3.75 .0022 Left Lateral OFC �34 34 �10 2.66 .01323

Right Putamen 24 �7 15 5.44 �.00118 1 13 4.56 �.00120 10 4 3.34 .003

Left Putamen �20 4 3 2.95 .007Left Caudate �12 8 5 2.93 .007Right Medial Frontal Gyrus 8 �23 53 4.48 �.001Left ACC �20 40 24 3.21 .004Left Frontopolar Cortex �16 57 5 3.96 .001Left OFC �24 34 �12 3.77 .002Left Anterior Insula �32 14 14 3.58 .002Left Posterior Insula �38 �28 14 3.07 .005Left Superior Parietal Lobule �22 �57 62 4.57 �.001Left Postcentral Gyrus �8 �32 64 3.06 .005Left Superior Temporal Gyrus �55 �40 18 3.18 .004Right Occipital Cortex 48 �70 5 3.40 .003

everity rated by the Hamilton Anxiety Rating Scale.

orresponding to puncorrected � .01 within predefined regions of interest and

ty Se

pa

.00

.00

.00

.00

.00

.01

iety s

lues c

the cross-validated values. The right ACC and left OFC were

Table 5. Intercorrelations Between Regional Metabolic Values for the Bipolar II Depressed and Control Subjects

L Amygdala R Amygdala L OFC R OFC L Cingulate R Cingulate L Accumbens R Accumbens L Caudate R Caudate L Putamen R Putamen

Bipolar II DepressedSubjects (n � 13)

L Amygdala .82a,f .53c .34 .50d .31 .71b .68b,g .55d .35 .60a .62c,g

R Amygdala .52c .58c,f .55d .60c,g .66c .72b,g .39 .38f .52d .57c

L Orbital Cortex .23 .68b .35 .73b .67c,f .34 .40f .69b,f .43R Orbital Cortex .02 .70b,g .29 .50d .41 .29 .32 .44f

L Cingulate .45 .69b .56c,g .19 .32 .64c,f .53d,g

R Cingulate .46 .33 .41 .14 .45g .47f

L Accumbens .77b .55d .45 .77b,g .80a,g

R Accumbens .45 .63c .78b,f .83b,f

L Caudate .66c .71b .67c

R Caudate .66c .60c

L Putamen .84a

R PutamenHealthy Control Subjects

(n � 18)L Amygdala .31 .24 �.05 .08 .06 .25 .06 .21 .08 .16 �.06R Amygdala .59c �.22 .33 �.03 .36 .21 �.24 �.40d �.07 .09L Orbital Cortex .02 .38 .20 .41d .01 �.08 �.37 �.04 �.10R Orbital Cortex �.20 .14 .16 .00 .22 �.08 �.31 �.38L Cingulate .31 �.24 �.16 �.24 �.04 �.05 �.18R Cingulate �.17 .05 �.01 .10 �.21 �.32L Accumbens .64b .47c �.11 .26 .38R Accumbens .39 .14 .03 .17L Caudate .74a .47c .23R Caudate .40d .10L Putamen .80a

R Putamen

OFC, orbitofrontal cortex.aCorrelation is significant at the .001 level (two-tailed).bCorrelation is significant at the .01 level (two-tailed).cCorrelation is significant at the .05 level (two-tailed).dTrend, p � .1 (two-tailed).eDifference in strength of correlation coefficient between depressives and control subjects, p � .01.fDifference in strength of correlation coefficient between depressives and control subjects, p � .05.gDifference in strength of correlation coefficient between depressives and control subjects, .05 � p � .1.

L.Mah

etalBIO

LPSY

CH

IATRY

2007;61:765–775771

ww

w.so

bp

.org

/jou

rnal

atEmtapn

D

ipaeaasavpMBvvu

mdsoircBtbittcsa(itnp

T

O

C

772 BIOL PSYCHIATRY 2007;61:765–775 L. Mah et al

w

dded successively to the stepwise model to determine whetherhey would improve the discriminative capability of the model.ach region improved the specificity and sensitivity for theodel based on the observed values. However, the sensitivity for

he model using the cross-validated values decreased with theddition of the left OFC. Thus, the model including rightutamen, right OFC, and right ACC provided the best discrimi-ative ability overall.

iscussion

This study demonstrates for the first time that the abnormalncreases in metabolic activity in limbic and paralimbic regionsreviously reported in unmedicated, depressed BD samples alsore evident in depressed BD II subjects receiving mood stabiliz-rs. In regions defined a priori, metabolism was elevated in themygdala, accumbens area and anteroventral putamen bilater-lly, left OFC, right pregenual ACC, and the left caudate in the BDubjects versus the control subjects (Table 2). Post hoc analysesdditionally revealed increases in metabolism in depressivesersus control subjects in the left parahippocampal cortex,osterior cingulate cortex, and right anterior insula (Table 3).ean whole-brain CMRglu did not differ significantly betweenD and control samples, justifying the comparison of regionalalues normalized to global metabolism to reduce nonspecificariability and ensuring that the regional results were not attrib-table to widespread increases in metabolism in BD.

The major goal of this study was to determine whetheretabolic abnormalities previously identified in unmedicated,epressed BD subjects also would be evident in depressed BDubjects currently taking mood stabilizers. Consequently, a pri-rity of the hypothesis testing was to avoid type II error, andntergroup differences that replicated previous findings wereeported in Table 2 if they survived correction for multipleomparisons within an ROI. Nevertheless, even when applyingonferroni adjustment to correct for multiple comparisons across

he 12 ROI after Bauer et al (2005), the differences in left OFC,ilateral accumbens area, and bilateral putamen remained signif-cant. The validity of applying this method for correcting acrosshe 12 regions searched by SPM99 was limited, however, becausehe metabolic measures across many of the regions were highlyorrelated (and thus nonindependent; Table 5). In addition, theizes of the area searched using the small volume concentrationnalyses differed across ROI both herein and in Bauer et al2005). Thus, important evidence that the abnormalities reportedn Tables 2 and 3 did not constitute type I error was provided byheir consistency with previous findings in BD (excepting theovel finding that metabolism was elevated in the parahip-

able 6. Predictive Ability of Models from Discriminant Analysis

Putamen (Right)Orbital Frontal (Right)

� Cingulate(Right)

� Orbital Frontal(Left)

bservedWilk’s Lambda .518 .483 .465Sensitivity 84.6 84.6 84.6Specificity 72.2 88.9 94.4% Correct 77.4 87.1 90.3

ross-ValidationSensitivity 84.6 84.6 69.2Specificity 72.2 83.3 88.9% Correct 77.4 83.9 80.6

ocampal gyrus in the BD II sample).

ww.sobp.org/journal

Although few previous functional imaging studies examinedspecifically the depressed phase of BD and none focusedexclusively on BD II, it is noteworthy that the regional abnor-malities reported herein in mood stabilizer treated, BD II depres-sives generally appear consistent with those reported in studiesof unmedicated depressed samples that combined BD I and IIdepressives, combined BD and major depressive disorder (MDD)depressives, or included only MDD depressives. Specifically, theabnormal elevations of regional metabolism found in our samplereplicated the findings of other studies of unmedicated BDdepressives imaged at rest or during a continuous performancetask in the right amygdala (Bauer et al 2005; Drevets et al 2002;Ketter et al 2001), left amygdala (Drevets et al 2002), ventralstriatum (Ketter et al 2001; Drevets et al 1995a), and right ventralACC (Bauer et al 2005). Our finding of increased metabolism inthe posterior OFC situated on the left inferior frontal gyrus wasconsistent with the findings of Ketter et al (2001) in unmedicated,“mildly depressed” bipolar patients (BD I and BD II combined)and of Cohen et al (1992) in unmedicated, depressed patients(BD II and MDD combined). The mildly depressed BD sample ofKetter et al (2001) also showed increased normalized metabolismin the posterior cingulate and anterior insular cortices, consistentwith the findings of our post hoc whole brain analysis (Table 3).Each of these abnormalities also has been reported in unmedi-cated depressed patients with MDD (reviewed in Drevets 2000).Finally, the area of increased metabolism in the left parahip-pocampal cortex in our BD II depressives (Table 3) is noteworthyin light of a report that metabolism in this region correlatedpositively with depression severity in MDD (Drevets et al 2002).

Similarly, the absence of significant differences in whole brainmetabolism between mood stabilizer medicated BD II patientsand control subjects was consistent with the previous findings ofDrevets et al (1997, 2002) in samples that included both BD I andBD II depressives. In contrast, Ketter et al (2001) reported thatmetabolism was reduced 7.7 % in a combined sample of BD Iand BD II patients with moderate to severe depression. Otherdifferences in the patient samples may have accounted for thisapparent discrepancy across studies, however, as Ketter et al(2001) selected only treatment-resistant patients who in mostcases also were rapid-cycling. Of the subjects studied herein andthe BD subjects previously studied by Drevets et al (1997, 2002),none were rapid cycling and treatment refractoriness was notestablished.

The current results suggest that lithium or divalproex do notproduce the widespread reductions in metabolism that occur inresponse to benzodiazepine and some antipsychotic agents(reviewed in Drevets et al 2004). Moreover, the substantialoverlap between the regional findings of the current study andthose of previous studies conducted in unmedicated samples(reviewed above) suggests that the differences found hereinbetween bipolar depressives and control subjects (Tables 2 and 3)were not attributable to effects of mood stabilizer treatment.Instead, these data suggest that currently depressed BD IIsubjects exhibit elevated neural transmission through the limbic-cortical-striatal circuitry (Drevets et al 1992), which remainsuncorrected by mood stabilizers in cases whose depressivesymptoms persist in spite of mood stabilizer monotherapy. Incontrast, mood stabilizer treatment that successfully alleviates thedepressive syndrome may attenuate the hypermetabolism inthese regions, as Drevets et al (2002) previously reported pre-liminary evidence that amygdala metabolism decreases to thenormative range in BD patients whose depressive symptoms

remitted during mood stabilizer monotherapy.

otstempeaegri2ip(emoibilsipct

mcoltemptitBatahsrpettottasci2

a

L. Mah et al BIOL PSYCHIATRY 2007;61:765–775 773

One noteworthy difference between the metabolic resultsbtained in symptomatic BD patients on mood stabilizers versushose reported in unmedicated BD depressives involved theubgenual ACC. Metabolic activity was abnormally decreased inhis region in unmedicated depressed subjects with BD (Drevetst al 1997), a finding not observed herein. The reduction inetabolism in unmedicated BD samples was associated with arominent reduction in gray matter (Drevets et al 1997; Hirayasut al 1999; Ongur et al 1998), which appeared to account for thepparent reduction in metabolism via partial volume averagingffects (Drevets 2005). Lithium treatment resulted in increasedray matter volume in specific PFC regions where abnormaleductions in gray matter were found in the untreated state,ncluding the subgenual ACC (Moore et al, unpublished data,004), consistent with evidence from preclinical studies indicat-ng that lithium and divalproex exert neurotrophic and neuro-rotective effects in the frontal cortex of experimental animalsDu et al 2003). Potentially consistent with this hypothesis, Bauert al (2005) found that in currently depressed BD I patients takingood stabilizers (predominantly lithium or divalproex), metab-lism was abnormally increased in the subgenual ACC (as well asn the right amygdala, ventral striatum, left thalamus, and cere-ellar vermis). The previous finding of reduced metabolic activityn BD in the subgenual ACC thus may be significantly altered byithium therapy, if the increase in subgenual ACC tissue isufficient to eliminate the partial volume averaging effect in PETmages. Longitudinal imaging studies acquired both pre- andost-treatment with mood stabilizers are needed to adequatelyharacterize these relationships between morphometry and me-abolism.

The significant, positive correlations found between regionaletabolic activity across the regions studied (Table 5) are

onsistent with their known anatomical connections with eachther (Ongur et al 2003) because regional glucose metabolismargely reflects the energy utilization associated with synapticransmission in neuronal terminal fields (Shulman et al 2004). Forxample, in the striatum, the positive correlations between theetabolism of the left and right side accumbens, caudate, andutamen are consistent with the known interhemispheric projec-ions connecting these regions with their homologous structuresn the contralateral hemisphere. Moreover, the specific areas ofhe striatum where metabolism was increased in the depressedD sample, namely the accumbens, ventromedial caudate, andnteroventral putamen (Tables 2 and 3; Figure 1), are the specificargets of extensive afferent anatomical projections from themygdala and orbital and medial PFC regions also implicatederein (Ongur et al 2003). Further, the limbic (amygdala, ventraltriatum, anterior and posterior cingulate cortex) and paralimbicegions (medial and orbital PFC, anterior insula, parahippocam-al gyrus) implicated in the current study participate in anxtensively interconnected, “visceromotor” network of structureshat modulate endocrine, autonomic, behavioral, and experien-ial aspects of emotional behavior (Ongur et al 2003). The patternf intercorrelated, abnormal metabolic increases across the struc-ures within this circuit in the depressed group suggests thathese regions function in concert to mediate, modulate, and/ordapt to the pathophysiology underlying the bipolar depres-ion phenotype. The absence of similar intercorrelations in theontrol group suggests that this circuit may not be physiolog-cally activated in the normative, resting state (Shulman et al004).

The significant covariation between the metabolic values

cross these regions in the BD group but not the control group

also was reflected in the results of the discriminant analysis,which allowed relatively specific and sensitive classification ofindividual subjects into correct diagnostic groups. Notably, thestructures identified by this analysis as providing the greatestdiscriminating capability were predominantly lateralized to theright hemisphere. In a previous, similar discriminant analysisinvolving unipolar depressives, left-sided structures within thesame circuit (amygdala, lateral OFC, medial thalamus, andventral ACC) provided the greatest discrimination of unipolardepressives from control subjects, as well as from bipolardepressives (Drevets et al 1995b). These data support hypothesespreviously based on studies of secondary mood disorders asso-ciated with cerebrovascular lesions proposing that frontal-striataldysfunction within the right hemisphere is involved more spe-cifically in the pathophysiology of bipolar mood syndromes,whereas frontal-striatal dysfunction on the left side is morerelevant to the development of unipolar depression (Robinson etal 1988).

Post hoc exploratory analyses provided preliminary evidenceof associations between regional metabolism and symptomratings. Depression severity correlated positively with metabo-lism in the dorsal anterolateral PFC. This region reportedlyshowed reduced metabolism in both unipolar and bipolar de-pression (Baxter et al 1989) and reductions in glial cell countsand neuronal size in MDD (Rajkowska et al 1999). Depressionratings also correlated positively with metabolism in the rightsuperior temporal cortex, a region that forms part of the ex-tended visceromotor network (Ongur et al 2003) and is reducedin volume in depressed BD subjects (Nugent et al 2006). Theinverse correlation found between depression severity and me-tabolism in the left OFC and anterior insula is consistent withstudies of unmedicated unipolar depressives imaged at rest(Drevets et al 1992, 1995b) or during a continuous performancetask (Kimbrell et al 2002). Notably, lesions within the left OFCincrease the risk for developing late-onset depression (Mac-Fall et al 2001). Taken together, these data suggest that leftlateral OFC/anterior insular function exerts a modulatory orinhibitory effect on depressive symptomatology (Drevets et al2004).

A complex relationship was observed between anxiety sever-ity and metabolism in the left medial OFC, such that anxietycorrelated negatively with metabolism in one left medial OFCregion but positively in another that was situated more ventrallyand anteriorly. These observations are notable, given evidencethat two adjacent regions in the medial PFC exert opposingeffects on amygdala activity and fear-related behavior in rats(Vidal-González et al, unpublished data). Anxiety severity alsocorrelated inversely with metabolism in the left frontal polarcortex and left anterior insula, compatible with previous reportsthat metabolism in these regions correlated inversely with de-pressive symptoms rated by the Hamilton Depression RatingScale, which also assesses multiple anxiety symptoms (Drevets etal 1992). In addition, anxiety ratings correlated negatively withmetabolism in the left superior temporal cortex and left posteriorinsula, consistent with the previous findings of Osuch et al (2000)in BD subjects imaged across a variety of mood states, and in thecaudate, similar to reports that right caudate activity correlatednegatively with symptom severity in posttraumatic stressdisorder (Lucey et al 1997). Finally, the observation thatanxiety symptoms correlated inversely with metabolism in theanteroventral putamen was notable, given evidence that inhealthy humans, anxiety symptoms associated with amphet-

amine administration correlated inversely with the magnitude

www.sobp.org/journal

oe

dtpa(mmavi(stacdtce

PItt

McSm

B

B

B

B

C

C

D

D

D

D

D

774 BIOL PSYCHIATRY 2007;61:765–775 L. Mah et al

w

f dopamine release in the anteroventral striatum (Drevetst al 2001).

In summary, during mood stabilizer treatment, persistentlyepressed subjects with BD II show increased metabolism inhe amygdala and other limbic and paralimbic structures. Theresence of elevated activity within this extended functionalnatomical network that modulates visceromotor functionOngur et al 2003) thus may depend primarily on the currentood state when considered together with evidence thatetabolism in these regions decreases following effective

ntidepressant treatment in MDD subjects (reviewed in Dre-ets et al 2004) but increases during tryptophan depletion-nduced depressive relapse in recovered MDD subjectsNeumeister et al 2004). These results confirm the feasibility oftudying the metabolic effects of adjunctive antidepressantherapy in patients receiving mood stabilizers to prophylaxgainst the development of hypomania. In addition, theonvergence of findings between bipolar disorder type IIepression and unipolar depression suggests that the func-ional anatomical correlates of the major depressive syndromeonstitute a neurophysiological phenotype that, to somextent, appears independent of mood disorder subtype.

This research was supported by the Intramural Researchrogram of the Naitonal Institutes of Health (NIH), Nationalnstitute of Mental Health (NIMH) and was presented to fulfillhesis requirements toward a Masters in Health Sciences degreehrough the Duke-NIH Clinical Research Training Program.

We thank examining committee members David C. Steffens,.D., M.HSc., and David M. DeLong, Ph.D., for their helpful

omments, the NIH PET Staff, and Allison C. Nugent, Ph.D., andtephen J. Fromm, Ph.D., for advice regarding scientific develop-ent, image processing, and analysis.

auer M, London ED, Rasgon N, Berman SM, Frye MA, Altshuler LL, et al(2005): Supraphysiological doses of levothyroxine alter regional cerebralmetabolism and improve mood in bipolar depression. Mol Psychiatry10:456 – 469.

axter LR Jr, Schwartz JM, Phelps ME, Mazziotta JC, Guze BH, Selin CE, et al(1989): Reduction of prefrontal cortex glucose metabolism common tothree types of depression. Arch Gen Psychiatry 46:243–250.

reiter HC, Gollub RL, Weisskoff RM, Kennedy DN, Makris N, Berke JD, et al(1997): Acute effects of cocaine on human brain activity and emotion.Neuron 19:591– 611.

rooks RA (1982): Alternative formula for glucose utilization using labeleddeoxyglucose. J Nucl Med 23:538 –539.

alabrese JR, Hirschfeld RM, Reed M, Davies MA, Frye MA, Keck PE, et al(2003): Impact of bipolar disorder on a U.S. community sample. J ClinPsychiatry 64:425– 432.

ohen RM, Gross M, Nordahl TE, Semple WE, Oren DA, Rosenthal N (1992):Preliminary data on the metabolic brain pattern of patients with winterseasonal affective disorder. Arch Gen Psychiatry 49:545–552.

revets WC (2000): Neuroimaging studies of mood disorders. Biol Psychiatry48:813– 829.

revets WC (2005): Brain structural abnormalities in mood disorders. In:Zorumski CF, Rubin EH, editors. Psychopathology in the Genome andNeuroscience Era. Washington, DC: American Psychiatric Publishing Inc.,119 –152.

revets WC, Gadde K, Krishnan R (2004): Neuroimaging studies of depres-sion. In: Charney DS, Nestler EJ, Bunney BJ, editors. The NeurobiologicalFoundation of Mental Illness, 2nd ed. New York: Oxford University Press,461– 490.

revets WC, Gautier C, Price JC, Kupfer DJ, Kinahan PE, Grace AA, et al (2001):Amphetamine-induced dopamine release in human ventral striatumcorrelates with euphoria. Biol Psychiatry 49:81–96.

revets WC, Price JL, Bardgett ME, Reich T, Todd RD, Raichle ME (2002):

Glucose metabolism in the amygdala in depression: Relationship to

ww.sobp.org/journal

diagnostic subtype and plasma cortisol levels. Pharmacol Biochem Behav71:431– 447.

Drevets WC, Price JL, Simpson JR Jr, Todd RD, Reich T, Vannier M, et al (1997):Subgenual prefrontal cortex abnormalities in mood disorders. Nature386:824 – 827.

Drevets WC, Price JL, Videen TO, Todd RD, Reich T, Vannier ME (1995a):Metabolic abnormalities in the subgenual prefrontal cortex and ventralstriatum in mood disorders. Astr Soc Neurosci 21:260.

Drevets WC, Spitznagel E, Raichle ME (1995b): Functional anatomical differ-ences between major depressive subtypes. J Cereb Blood Flow Metab15:S93.

Drevets WC, Videen TO, Price JL, Preskorn SH, Carmichael ST, Raichle ME(1992): A functional anatomical study of unipolar depression. J Neurosci12:3628 –3641.

Du J, Gray NA, Falke C, Yuan P, Szabo S, Manji HK (2003): Structurally dissim-ilar antimanic agents modulate synaptic plasticity by regulating AMPAglutamate receptor subunit GluR1 synaptic expression. Ann N Y Acad Sci1003:378 –380.

Duvernoy H (1999): The Human Brain, 2nd ed. New York: Springer-VerlagWien.

Friston KJ, Worsley KJ, Frackowiak RSJ, Mazziotta JC, Evans AC (1994): Assess-ing the significance of focal activations using their spatial extent. HumBrain Mapp 1:214 –220.

Hamilton M (1959): The assessment of anxiety states by rating. Br J MedPsychol 32:50 –55.

Hirayasu Y, Shenton ME, Salisbury DF, Kwon JS, Wible CG, Fischer IA, et al(1999): Subgenual cingulate cortex volume in first-episode psychosis.Am J Psychiatry 156:1091–1093.

Judd LL, Schettler PJ, Akiskal HS, Maser J, Coryell W, Solomon D, et al (2003):Long-term symptomatic status of bipolar I vs. bipolar II disorders. IntJ Neuropsychopharmacol 6:127–137.

Kessler RC, Chiu WT, Demler O, Merikangas KR, Walters EE (2005): Preva-lence, severity, and comorbidity of 12-month DSM-IV disorders in theNational Comorbidity Survey Replication. Arch Gen Psychiatry 62:617–627.

Ketter T, Drevets WC (2002): Neuroimaging studies of bipolar depression:Functional neuropathology, treatment effects, and predictors of clinicalresponse. Clin Neurosci Res 2:182–192.

Ketter TA, Kimbrell TA, George MS, Dunn RT, Speer AM, Benson BE, et al(2001): Effects of mood and subtype on cerebral glucose metabolismin treatment-resistant bipolar disorder. Biol Psychiatry 49:97–109.

Kimbrell TA, Ketter TA, George MS, Little JT, Benson BE, Willis MW, et al(2002): Regional cerebral glucose utilization in patients with a range ofseverities of unipolar depression. Biol Psychiatry 51:237–252.

Lucey JV, Costa DC, Busatto G, Pilowsky LS, Marks IM, Ell PJ, et al (1997):Caudate regional cerebral blood flow in obsessive-compulsive disorder,panic disorder and healthy controls on single photon emission comput-erised tomography. Psychiatry Res 74:25–33.

MacFall JR, Payne ME, Provenzale JE, Krishnan KR (2001): Medial orbitalfrontal lesions in late-onset depression. Biol Psychiatry 49:803– 806.

Montgomery SA, Asberg M (1979): A new depression scale designed to besensitive to change. Br J Psychiatry 134:382–389.

Moore DF, Altarescu G, Barker WC, Patronas NJ, Herscovitch P, Schiffmann R(2003): White matter lesions in Fabry disease occur in ’prior’ selectivelyhypometabolic and hyperperfused brain regions. Brain Res Bull 62:231–240.

Neumeister A, Nugent AC, Waldeck T, Geraci M, Schwarz M, Bonne O, et al(2004): Neural and behavioral responses to tryptophan depletion inunmedicated patients with remitted major depressive disorder and con-trols. Arch Gen Psychiatry 61:765–773.

Nugent AC, Milham MP, Bain EE, Mah L, Cannon DM, Marrett S, et al (2006):Cortical abnormalities in bipolar disorder investigated with MRI andvoxel-based morphometry. Neuroimage 30(2):485– 497.

Ongur D, Drevets WC, Price JL (1998): Glial reduction in the subgenual prefrontalcortex in mood disorders. Proc Natl Acad Sci U S A 95:13290–13295.

Ongur D, Ferry AT, Price JL (2003): Architectonic subdivision of the humanorbital and medial prefrontal cortex. J Comp Neurol 460:425– 449.

Osuch EA, Ketter TA, Kimbrell TA, George MS, Benson BE, Willis MW,et al (2000): Regional cerebral metabolism associated with anxiety symp-toms in affective disorder patients. Biol Psychiatry 48:1020–1023.

Poline JB, Worsley KJ, Evans AC, Friston KJ (1997): Combining spatial extentand peak intensity to test for activations in functional imaging. Neuroim-

age 5:83–96.

R

R

R

S

S

L. Mah et al BIOL PSYCHIATRY 2007;61:765–775 775

ajkowska G, Miguel-Hidalgo JJ, Wei J, Dilley G, Pittman SD, Meltzer HY,et al (1999): Morphometric evidence for neuronal and glial prefrontalcell pathology in major depression. Biol Psychiatry 45:1085–1098.

obinson RG, Boston JD, Starkstein SE, Price TR (1988): Comparison of maniaand depression after brain injury: Causal factors. Am J Psychiatry 145:172–178.

osenthal R (1991): Meta-Analytic Procedures for Social Research, RevisedEdition. Newbury Park, CA: Sage Publications.

hulman RG, Rothman DL, Behar KL, Hyder F (2004): Energetic basis of brainactivity: Implications for neuroimaging. Trends Neurosci 27:489 – 495.

zeszko PR, Robinson D, Alvir JM, Bilder RM, Lencz T, Ashtari M, et al (1999):

Orbital frontal and amygdala volume reductions in obsessive-compul-sive disorder. Arch Gen Psychiatry 56:913–919.

Szeszko PR, Strous RD, Goldman RS, Ashtari M, Knuth KH, Lieberman JA, et al(2002): Neuropsychological correlates of hippocampal volumes in pa-tients experiencing a first episode of schizophrenia. Am J Psychiatry159:217–226.

Tabachnick BG, Fidell LS (1989): Using Multivariate Statistics. New York:Harper & Row.

Talairach J, Tournoux P (1988): Co-Planar Stereotaxic Atlas of the HumanBrain. New York: Thieme Medical Publishers Inc.

Vidal-González I, Vidal-González B, Rauch SL, Quirk GJ (in review): Micro-stimulation reveals opposing influences of prelimbic and infralimbiccortex on the expression of conditioned fear.

Woods RP, Mazziotta JC, Cherry SR (1993): MRI-PET registration with auto-mated algorithm. J Comput Assist Tomogr 17:536 –546.

www.sobp.org/journal