goodkind 2015

Copyright 2015 American Medical Association. All rights reserved.

Identification of a CommonNeurobiological SubstrateforMental IllnessMadeleine Goodkind, PhD; Simon B. Eickhoff, DrMed; Desmond J. Oathes, PhD; Ying Jiang, MD; Andrew Chang, BS;Laura B. Jones-Hagata, MA; Brissa N. Ortega, BS; Yevgeniya V. Zaiko, BA; Erika L. Roach, BA; Mayuresh S. Korgaonkar, PhD;Stuart M. Grieve, DPhil; Isaac Galatzer-Levy, PhD; Peter T. Fox, MD; Amit Etkin, MD, PhD

IMPORTANCE Psychiatric diagnoses are currently distinguished based on sets of specificsymptoms. However, genetic and clinical analyses find similarities across a wide variety ofdiagnoses, suggesting that a common neurobiological substratemay exist acrossmentalillness.

OBJECTIVE To conduct a meta-analysis of structural neuroimaging studies across multiplepsychiatric diagnoses, followed by parallel analyses of 3 large-scale healthy participant datasets to help interpret structural findings in themeta-analysis.

DATA SOURCES PubMedwas searched to identify voxel-basedmorphometry studies throughJuly 2012 comparing psychiatric patients to healthy control individuals for themeta-analysis.The 3 parallel healthy participant data sets included resting-state functional magneticresonance imaging, a database of activation foci across thousands of neuroimagingexperiments, and a data set with structural imaging and cognitive task performance data.

DATA EXTRACTION AND SYNTHESIS Studies were included in themeta-analysis if theyreported voxel-basedmorphometry differences between patients with an Axis I diagnosisand control individuals in stereotactic coordinates across the whole brain, did not presentpredominantly in childhood, and had at least 10 studies contributing to that diagnosis (oracross closely related diagnoses). Themeta-analysis was conducted on peak voxelcoordinates using an activation likelihood estimation approach.

MAIN OUTCOMES ANDMEASURES We tested for areas of common graymatter volume increaseor decrease across Axis I diagnoses, as well as areas differing between diagnoses. Follow-upanalyses on other healthy participant data sets tested connectivity related to regions arisingfrom themeta-analysis and the relationship of graymatter volume to cognition.

RESULTS Based on the voxel-basedmorphometry meta-analysis of 193 studies comprising15 892 individuals across 6 diverse diagnostic groups (schizophrenia, bipolar disorder,depression, addiction, obsessive-compulsive disorder, and anxiety), we found that graymatter loss converged across diagnoses in 3 regions: the dorsal anterior cingulate, rightinsula, and left insula. By contrast, there were few diagnosis-specific effects, distinguishingonly schizophrenia and depression from other diagnoses. In the parallel follow-up analyses ofthe 3 independent healthy participant data sets, we found that the common gray matter lossregions formed a tightly interconnected network during tasks and at resting and that lowergray matter in this network was associated with poor executive functioning.

CONCLUSIONS AND REVELANCE We identified a concordance across psychiatric diagnoses interms of integrity of an anterior insula/dorsal anterior cingulatebased network, whichmayrelate to executive function deficits observed across diagnoses. This concordance providesan organizing model that emphasizes the importance of shared neural substrates acrosspsychopathology, despite likely diverse etiologies, which is currently not an explicitcomponent of psychiatric nosology.

JAMA Psychiatry. doi:10.1001/jamapsychiatry.2014.2206Published online February 4, 2015.

Supplemental content atjamapsychiatry.com

Author Affiliations:Authoraffiliations are listed at the end of thisarticle.

Corresponding Author: Amit Etkin,MD, PhD, Department of Psychiatryand Behavioral Sciences, StanfordUniversity 401 Quarry Rd, MC 5797,Stanford, CA 94305-5797 ([email protected]).

Research

Original Investigation | META-ANALYSIS

(Reprinted) E1


Downloaded From: http://archpsyc.jamanetwork.com/ on 02/23/2015


D uring the past several decades, psychiatry has fo-cused on establishing diagnostic categories based onclinical symptoms.1,2 Accordingly, most neuroimag-ing studies have compared brain structure or function in pa-tients with a single, specific diagnosis with healthy partici-pants. In turn, even closely related diagnostic categories arerarely compared with each other. Nonetheless, neuroimag-ing research is suggestive of common neurobiological abnor-malities across phenotypically related diagnoses (eg, schizo-phrenia [SCZ] and bipolar disorder [BPD] or anxiety anddepression).3-6Thesedatahavealsoconvergedon the idea thatpsychiatric illnesses affect the operation of commonly ob-serveddistributedneuralcircuits.7,8Additionally,geneticanaly-ses have identified common polymorphisms associated witha large range of psychiatric diagnoses,9 and comorbidity be-tween diagnoses is considerably higher than expected bychance.10 Thus, there is a disconnectionbetween current psy-chiatric nosology and rapidly emerging biological findings,which emphasizes the need to look for neurobiological sub-strates shared across diagnoses.

To search for a potential transdiagnostic neural signa-ture, we first conducted a meta-analysis of psychiatric neuro-imaging studies that used voxel-based morphometry (VBM)to assess patient/control differences in regional brain volumefrom structural neuroimaging data. Voxel-based morphom-etry analysis of brain structure holds several advantages:(1) VBM analyses use standardized methods, which allow forpooling across studies; (2) they assess the entire brain, thusalleviating the need for a priori assumptions on which neuralcircuits are thought to be affected; (3) the major psychiatricdiagnoses have been examined across numerous indepen-dent VBM studies; and (4) structural markers are relativelystable across time and may provide trait measures of brainabnormalities.11 While there have been many advances instructural brain imaging methods beyond VBM, our goal herewas to quantitatively summarize the very large body of exist-ing (VBM) findings so that this knowledge can be used toguide new studies with improved methods. The use of meta-analytic methods furthermore allows for the summation oflarge amounts of structural imaging data in a spatially unbi-ased manner, thereby reflecting conclusions based on muchof the structural imaging work in psychiatric disorders duringthe past 15 years. Doing so further allows for direct compari-sons of diagnoses that were never compared with each otherin the original data sets.4,12 To more fully contextualizepotential transdiagnostic brain abnormalities in patients, wealso contrasted specific diagnoses or diagnosis groupings, assupported by prior studies of symptom or genetic covariationacross large and diverse patient cohorts.13-17 We used thesegroupings to guide our meta-analytic contrasts because theyare data-driven parcellations that best reflect the currentunderstanding of relationships between diagnoses and thusdo not require an assumption that individual diagnoses rep-resent independent entities.

Despite the advantages of VBM, measures of brain struc-ture are ambiguouswith respect to the connectivity betweenthe identifiedbrain regions, aswell aswith respect to their be-havioral relevance. Thus, we also conducted parallel analy-

ses of 3 data sets of healthy participants to test whether pat-terns of perturbed brain structure reflects a disruption in afunctionally interrelated neural circuit andwhether there arebehavioral correlatesof alteredstructural integrityevenwithina healthy range of functioning. Although these data are de-rived from healthy participants, they can nonetheless aid inthe interpretationofpatient/control differences inbrain struc-ture and help constrain hypotheses formore targeted studiesof potential transdiagnostic neural markers.

MethodsVBMMeta-analysis: Study SelectionWe searched PubMed for all studies through July 2012 usingVBM (Figure 1). Of these, we selected those studies thatinvestigated Axis I psychiatric diagnoses. Studies includingpatients with SCZ, schizophreniform, or schizoaffective dis-order were included in a psychotic diagnoses category. Thisincluded both chronic and first-onset psychosis. There wereno studies of transient psychotic disorders. Also, althoughindividuals with depression or bipolar disorder may presentwith psychotic features, studies of these patients wereincluded in separate categories; in these cases, the mood dis-order was the primary diagnosis and, in general, thesepatients did not have psychosis. One study of patients withaffective psychosis was excluded because it was unclear how

Figure 1. FlowDiagram of Study Selection

1910 Records identified throughdatabase search

1920 Records after duplicatesremoved

1485 Records excludedbecause they did not include patientswith Axis I diagnoses

242 Full-text articles excluded 28 No healthy control

group or patient/control comparisons

34 No patient/controldifferences

48 No patient groupwith Axis I diagnoses

44 No whole-brainanalysis or differentmethod used

14 No coordinates listedin stereotaxic space

32 Meta-analyses/reviews

12 Articles in a differentlanguage

9 Too few studies withthat diagnosis

21 Other (eg, patientswith medicalcomorbidity)

24 Additional records identifiedthrough other sources

1920 Records screened

435 Full-text articles assessed foreligibility

193 Studies includedin the quantitativesynthesis(meta-analysis)

193 Studies includedin qualitativesynthesis

Research Original Investigation Neurobiological Substrate for Mental Illness

E2 JAMAPsychiatry Published online February 4, 2015 (Reprinted) jamapsychiatry.com




to categorize it. We excluded neurological disorders, diagno-ses presenting predominantly in childhood (eg, attention-deficit/hyperactivity disorder), personality disorders, andany diagnosis that had fewer than 10 studies and could notbe readily grouped with another diagnosis (eg, affectivepsychosis).4 Additionally, while autism spectrum disorderspresent symptoms throughout the lifespan, they typicallyfirst present very early in life and are associated with altereddevelopmental trajectories of brain structure,18-20 which mayfurthermore interact with key features of the diagnosis (eg, asindicated by IQ).18,19 In light of this, and the fact that the 10VBM studies we identified of adults with autism spectrumdisorders largely did not address this heterogeneity, we optedto exclude autism to maximize interpretability of our results.Studies in which some patients were younger than age 18years were included so that for the diagnoses selected allavailable studies were included.

Studies were selected if they (1) used VBM to analyze graymatter in patients with a psychiatric diagnosis, (2) included acomparison between these patients and matched healthycontrol participants, (3) performed a whole-brain analysis,and (4) reported coordinates in a defined stereotaxic space(eg, Talaraich space or Montreal Neurological Institute space).This meant that any studies that only reported results aftersmall-volume correction within a region of interest wereexcluded. Some studies included multiple patient groups andwe included as separate contrasts each patient/healthy con-trol participant comparison. While it is not possible to deter-mine whether some participants in a prior publication wereincluded in subsequent publications, this confound wasmini-mized because we included a large number of studies thatrepresented a diversity of authors, scanners, and institutions.It is also unlikely that this type of error would explain theexistence of a common gray matter change pattern acrossdiagnoses. Coordinates reported in Talairach space were con-verted into Montreal Neurological Institute space for themeta-analysis.21

Activation Likelihood EstimationMeta-analysisWeused the revised activation likelihoodestimation (ALE) al-gorithm to identify consistent patterns of graymatter changeacross studies.22,23 This algorithmaims to identify areas show-ing a convergence of reported coordinates across experi-ments, which is higher than expected under a random spatialassociation. The key idea behind ALE is to treat the reportedfoci not as single points, but rather as centers for 3-dimen-sional gaussian probability distributions capturing the spa-tial uncertainty associated with each focus. Then, for eachvoxel, the probabilities of all foci of a given experiment wereaggregated, yielding a modeled activation map.24 The unionof allmodeledactivationmaps then resulted invoxelwiseALEscores, which reflect the convergence of results at each par-ticular location of the brain. Significant convergence was as-sessedbycomparisonofALEscoreswithanempirical null dis-tribution that reflects a random spatial association betweenexperiments with a fixed within-experiment distribution offoci.22 Hereby, random-effects inference was applied, whichdoes not cluster foci within a particular study but rather as-

sesses above-chance convergence between experiments. Theobserved ALE scores were tested against the expectation onthe ALE scores under the null distribution of random spatialassociation across experiments. The resulting nonparametricP values were then thresholded at a cluster-level familywiseerrorcorrected threshold of P < .05 (cluster-forming thresh-old at voxel-level P < .005) and transformed into z scores fordisplay. To identify the regions showing commongraymatterchanges in psychotic and nonpsychotic diagnosis groups, weconducted a conjunction analysis. That is, by using themini-mumstatistics under the conjunctionnull hypothesis25,26 andcomputing the intersection of the thresholdedmeta-analyticmaps derived from both approaches, we aimed to delineateconsistentpatternsofgraymatterchangeacrosspatientgroups.The conjunction null hypothesis tests whether all effects aredifferent from null rather than whether the combined effectis null (ie, the global null hypothesis).

Differences in activation likelihood between were testedby performing ALE separately on the experiments associatedwith either groupandcomputing thevoxelwisedifferencebe-tween the ensuingALEmaps.All experiments contributing toeither analysiswere thenpooled and randomly divided into 2groups of the same size as the 2 original sets of experiments(eg, findings inpsychoticandnonpsychoticdiagnosisgroups).27

Activation likelihood estimation scores for these 2 randomlyassembled groups, reflecting the null hypothesis of label ex-changeability, were calculated and the difference betweentheseALE scoreswas recorded for each voxel in the brain. Re-peating thisprocess 10 000 times thenyieldedavoxelwisenulldistribution on the differences in ALE scores between the 2(sub) analyses. The true differences in ALE scores were thentested against this null distribution, yielding a P value for thedifference at each voxel based on the proportion of equal orhigher differences under label exchangeability. The resultingP values were thresholded at P > .95 (95% chance of true dif-ference), transformed into z scores, and inclusivelymaskedbythe respectivemaineffects (ie, thesignificanteffects in theALEfor a particular group).

Functional Significance of the Identified GrayMatter RegionsTo test whether the identified regions of common gray mat-ter loss identified in thepatientVBMmeta-analysis formedaninterconnected network and to understand its functional sig-nificance,weanalyzed 3 large complimentaryhealthypartici-pant data sets.

1. Task-Based Connectivity: Meta-analytic ConnectivityModelingIn the first healthy participant data set, we assessed the nor-mal pattern of task-based coactivation throughout the brainfor each of the VBM common gray matter loss regions. Meta-analytic connectivity modeling (MACM) assesses connectiv-ity by determining brain areas that coactivate with a seed re-gion acrossmany neuroimaging experiments at a level abovechance.28 This analysis was conducted using the BrainMapdatabase.29-31We constrained our analysis to functionalmag-netic resonance imaging (MRI) and positron emission tomo-graphicexperiments fromnormalmappingneuroimagingstud-ies (no interventions and no group comparisons) in healthy

Neurobiological Substrate for Mental Illness Original Investigation Research

jamapsychiatry.com (Reprinted) JAMAPsychiatry Published online February 4, 2015 E3




participants, which report results as coordinates in stereo-taxic space. These inclusion criteria yielded approximately7500 eligible experiments at the time of analysis.

The first step in anMACM analysis is to identify all experi-ments in a database that activate the seed region (ie, thatreported at least 1 focus within the seed volume). Subse-quently, quantitative meta-analysis is used to test for conver-gence across the foci reported in these experiments.28,32 Sig-nificant convergence outside the seed, as computed using theALE methods previously described and tested by a cluster-level familywise errorcorrected threshold of P < .05 (cluster-forming threshold of P < .005), indicates consistent coactiva-tion. Identification of overlapping regions of meta-analyticcoactivation with each of the common gray matter loss regionseeds was performed using the minimum statistics justdescribed. Significance in this conjunction indicated that aregion was coactivated separately with each of the seedregions (and combination) and is not sensitive to the fact thatthere may be an overlap between coactivated regions in onemap and a seed in another map because it evaluates the con-junction null hypothesis.25,26

2. Task-Independent Connectivity: Resting-State FunctionalMRI ConnectivityIn the second healthy participant data set, our goal was todetermine patterns of task-independent functional connec-tivity patterns across the whole brain for each of the VBMmeta-analysis common gray matter loss regions as seeds in atask-free resting state (RS). In total, the processed sampleconsisted of 99 healthy individuals between 21 and 60 years(mean [SD] age, 36.3 [11.1] years; 63 men and 36 women) with260 echoplanar images per individual. We limited our sampleto nongeriatric adults to diminish the influence of develop-ment or aging on our results. See the eAppendix in theSupplement for a description of scan parameters and pro-cessing methods for functional connectivity analyses.

Themaineffect of connectivity for individual clusters andconjunctionsacross thosewere testedusing thestandardSPM8implementations with the appropriate nonsphericity correc-tion.Theresultsof these random-effectsanalyseswerecluster-level thresholdedatP < .05 (cluster-forming thresholdatvoxellevel:P < .005),analogoustotheMACManalysis.Similarly, con-junctionmapswere createdusing theminimumstatistics justdescribed to identify overlapping patterns of functional MRIconnectivity across each of the common gray matter loss re-gion seeds. Finally, a minimum statistics conjunction analy-sis between MACM and RS results was performed to detectareas showing both task-dependent and task-independentfunctional connectivity across all of the common graymatterloss seed regions. Thus, regions significant across these con-junctionswere connected separatelywith eachof the seed re-gions during both tasks and rest.

3. Behavioral Task Performance and VBMGrayMatter VolumesIn the third healthy participant data set, our goal was to de-termine whether gray matter decreases in the common graymatter loss regionspredictedperformanceonbehavioral testsof a range of cognitive functions. In total, 163 healthy indi-

viduals between 21 and 60 years (mean [SD] age, 38.2 [12.7]years; 72men and 91women)were drawn from theBRAINnetFoundation Database.33,34 Exclusion criteriawere any knownneurological disorder, previous head injury, mental retarda-tion, DSM-IV Axis I diagnosis, and history of drug depen-dence. See the eAppendix in the Supplement for MRI acqui-sition parameters and processing methods for VBM andbehavioral data.

The standardized computerized behavioral performancebattery included 10 tasks, probing a range of cognitivedomains35: digit span task (working memory), span of visualmemorytask (workingmemory), trail-makingtask (taskswitch-ing), color-wordStrooptask (interferenceresolution),mazetask(visuospatial navigation), verbal fluency task, continuousper-formance task (sustained attention), go/no go task (responseinhibition), choice reaction time task (information processingspeed), and a finger tapping task (motor speed). Performancemetrics are listed in eTable 1 in the Supplement for each task.

The relationshipbetweenregional graymattervolumeandbehavioral performance was examined by regressing the av-erage of z scorenormalized volumes of each of the commongraymatter loss regions of interest against each of the behav-ioral performanceprincipal components,while controlling forage, education, sex, and their interactions because these de-mographics predicted either behavior or graymatter volume.

ResultsVBMMeta-analysis Across Psychiatric DisordersIncludedinthemeta-analysiswerepeakvoxelcoordinates frompublished studies that compared a psychiatric group withhealthyparticipants,which thus represented indicators of re-gional gray matter volume change associated with that diag-nosis (see theMethods section for studyselectioncriteria).Ourfinal sample included 212 comparisons between patients andcontrol individuals from193peer-reviewedarticles, represent-ing a total of 7381 patients and 8511 matched healthy controlindividuals (eTable 2 in the Supplement). Included diagnos-tic groups were SCZ (including schizoaffective and schizo-phreniformdiagnoses),BPD,majordepressivedisorder (MDD),substanceusedisorder, obsessive-compulsivedisorder (OCD),and a group of several anxiety disorders combined for ad-equate sample size (ANX). Thus, the meta-analysis includedahighlydiverse sampleofdiagnoses fromacross themajor cat-egories of adult Axis I psychopathology.

Meta-analyses were conducted using the revised ALEmethod,22witha familywiseerror correction formultiple com-parisons. Across all studies, the clear majority (85%) of peakvoxels represented decreased gray matter in patients com-pared with control individuals. Consistent gray matter de-creases in patientswere found in the bilateral anterior insula,dorsal anterior cingulate (dACC), dorsomedial prefrontal cor-tex, ventromedial prefrontal cortex, thalamus, amygdala, hip-pocampus, superior temporal gyrus, and parietal operculum(Figure 2A and eTable 3 in the Supplement). By contrast, graymatter increases inpatientswere foundexclusively in thestria-tum (eFigure 1A and eTable 3 in the Supplement).






Because one of the most fundamental diagnostic divi-sions in psychopathology is between primarily psychotic di-agnoses (ie, SCZ) and primarily nonpsychotic diagnoses,15,16

wenextassessed for commongraymatter changesacross thesefundamental categories. In doing so,we also ensured that our

findings were not driven by studies of SCZ, which repre-sentednearlyhalfof the includedstudies.Aconservativemini-mum statistic conjunction across the psychotic and nonpsy-chotic diagnosis groups revealed significant gray matter lossin the bilateral anterior insula and dACC of both groups(Figure 2B and C; eTables 4-6 in the Supplement). Further-more, the presence of gray matter loss in one region pre-dicted a higher than chance probability of gray matter loss inthe other regions (21 > 4.3;P < .05 for all),36,37 suggesting thatthese gray matter losses may occur in a coordinated fashionacross a structural network including these regions. By con-trast, thegraymatter increases in the striatumofpatientswereevident only in the psychotic diagnosis group (eFigure 1B andeTable 4 in the Supplement).

For follow-up analyses, we extracted per-voxel prob-abilities of decreased gray matter in the VBM meta-analysisfor each of the 3 common regions and conducted nonpara-metric Kruskal-Wallis tests to examine the effects of diagno-sis, age, medication, and comorbidity. The values representthe probability of identifying a gray matter abnormality foran average voxel within the region of interest derived fromthe modeled activation maps. We found similar magnitudeeffects for gray matter loss across all nonpsychotic diagno-ses, as shown in Figure 3 (Kruskal-Wallis H test: H[4] < 3.3;P > .51 for all). These effects were significantly larger in thepsychotic than the nonpsychotic diagnosis group (H[1] > 4.9;P < .03 for all; see voxelwise comparison in eFigure 2A andeTable 7 in the Supplement). Gray matter differences werenot related to either age at onset (reported in 106 studies) orduration of illness (reported in 148 studies) using either non-parametric correlations (Spearman rho < 0.13; P > .18 for all)or Kruskal-Wallis tests based on a median split (H[1] < 0.52;P > .47 for all).

Next, we examined the potential role of medications.Within the nonpsychotic diagnosis group, 64% of studies in-cluded medicated patients; however, this did not explain in-sula and dACC gray matter loss (H[1] < 0.9; P > .35 for all).Within the psychotic diagnosis group, 90% of studies in-cluded medicated patients, and here too medication did not

Figure 2. Shared Patterns of Decreased GrayMatter From the Voxel-BasedMorphometryMeta-analysis

A

B

C

All patients

CommonInsula

dACC

L RZ

6

4

2

Psychotic only Nonpsychotic only

Results are from patient vs healthy participant comparisons for studies pooledacross all diagnoses (A), separately by psychotic or nonpsychotic diagnosisstudies (B), and from a conjunction across the psychotic and nondiagnosisdiagnosis groupmaps in panel B (C). Results show common gray matter lossacross diagnoses in the anterior insula and dorsal anterior cingulate (dACC).The z score is for the activation likelihood estimation analysis for gray matterloss. L indicates left; and r, right.

Figure 3. Extracted per-Voxel Probabilities of Decreased GrayMatter in the Voxel-BasedMorphometryMeta-analysis, Separated by Individual Diagnosis and CommonGrayMatter Loss Region(Left and Right Anterior Insula)

0

4

Cont

rol>

Patie

nt V

oxel

wis

e Pr

obab

ility

, x10

4

3

2

1

SCZ BPD MDD SUD OCD ANX

92N= 26 28 21 15 17

Psychotic vs nonpsychotica

Left insula

Right insula

dACC

Values represent the probability ofidentifying a gray matter abnormalityfor an average voxel within the regionof interest, derived from themodeledactivationmaps. ANX indicatesanxiety disorders; BPD, bipolardisorder; dACC, dorsal anteriorcingulate; MDD, major depressivedisorder; OCD, obsessive-compulsivedisorder; SCZ, schizophrenia; andSUD, substance use disorder.aP < .05 for comparison of thepsychotic with the nonpsychoticdisorders.






predict gray matter loss (H[1] < 0.7; P > .40 for all). Antipsy-choticmedicationhas been associatedwith increases in stria-tal graymatter, consistentwith our finding of increased stria-tal gray matter only in the psychotic group (eFigure 1 in theSupplement).38-40 We have not identified reports of de-creased insular volume with antipsychotic medications ineither humans or animals. Reports of decreased frontal vol-umes (determined using a very broad region of interest)41 inhumans appear to reflect volume decreases in the lateral pre-frontal cortex.40,42 In turn, if anything, volume inmidline re-gions locatedwithina fewcentimetersofourdACCclusterwerereported to be increased with antipsychotic medication.40,43

Opposite effectsof typical andatypical antipsychoticsonbrainvolumehavealsobeen reported,44whileothereffectshavenotsurvived a correction for multiple comparisons.45,46 Criti-cally, themedications typically used inpsychotic patients dif-fer highly from those used with nonpsychotic patients, mak-ing it unlikely that medication could explain a commondecrease in the insulaanddACCacrossbothpsychoticandnon-psychotic patients.

Finally, we considered whether the common graymatterloss findingswere owing to thepresence of a commoncomor-bid diagnosis among those studies reporting on comorbidity(reported in 90% of all studies). Overall, rates of comorbidityvariedacrossdiagnosis: substanceusedisorder, 9%;SCZ, 10%;BPD, 18%;MDD, 24%;OCD, 50%; andANX, 58%.However, thepresenceof comorbiditydidnotaccount fordifferences ingraymatter in either the insula or dACC (H[1] < 1.7; P > .20 for all).Wealsoobserveda similar patternof commongraymatter lossin the anterior insula and dACC, as just described, after ex-cludingstudieswithAxis Icomorbidity (eFigure2 intheSupple-ment). In summary, these results suggest that anterior insulaand dACC graymatter loss represent a transdiagnostic neural

abnormality evident across a wide variety of mental ill-nesses, most pronounced in psychosis.

In a contrast of psychotic and nonpsychotic diagnosisgroups, we found that the psychotic group had both greatergraymatter loss in themedial prefrontal cortex, insula, thala-mus, and amygdala (eFigure 3A and eTable 7 in the Supple-ment), aswell as greater increases in striatal graymatter (eFig-ure 3B and eTable 8 in the Supplement). We then furthersubdivided the nonpsychotic diagnosis group into 3 group-ings: internalizing (depression [MDD], OCD, and anxiety dis-orders [ANX]), externalizing (substance use disorders), andBPD. These groupings were based on prior symptom and ge-netic covariation studies of the structural organizationof psy-chiatric disorders.13-15,17 Contrasts between these groups iden-tified clusters in theanteriorhippocampus, extending into theamygdala. Specifically, internalizingdisordershadgreaterhip-pocampal/amygdala gray matter loss than both the external-izing (eFigure 3A and eTable 1 in the Supplement) or BPD(Figure4; eTable8 in theSupplement) diagnostic groups. Thiseffectwasdrivenby theMDDgroup,whereinwe foundgreaterhippocampal/amygdala graymatter loss than in the other in-ternalizing (OCD and ANX; Figure 4; eTable 1 in the Supple-ment), externalizing, or BPDgroups (eFigure 4 andeTable 1 inthe Supplement).

Task-Dependent and Task-Independent ConnectivityTo determine whether these common gray matter loss re-gions are normally part of a coherent brain circuit, we exam-ined their patterns of task-dependent coactivation and task-independentRS functional connectivity (FC).We constructedMACMs47 using the BrainMap database that identify above-chance task-based coactivations across thousands of neuro-imaging studies inhealthy individuals, givenactivation inone

Figure 4. Voxel-BasedMorphometryMeta-analysis Contrasts Between Subdivisions of the NonpsychoticDisorder Studies Group, Broken Down by Internalizing, Externalizing, and Bipolar Disorder (BPD) Groupings

Internalizing = more lossthan externalizing

Internalizing = more lossthan BPD

Nonpsychotic disorders

Internalizing (MDD, ANX, and OCD)

MDD = more loss than ANX + OCD

BPD

MDDANX + OCD

Externalizing (SUD)

Graymatterloss: z

3

2

1

Internalizing disorders show greatergray matter loss in the anteriorhippocampus and amygdala whencompared with either externalizing orBPD diagnoses. This effect is drivenby themajor depressive disorder(MDD) group, which shows greatergray matter loss in these regions thanthe other diagnoses within theinternalizing grouping (anxietydisorder [ANX] andobsessive-compulsive disorder[OCD]). The z score is for theactivation likelihood estimationanalysis for gray matter loss. SUDindicates substance use disorder.






of the identified regions of convergent gray matter loss(Figure 5A). A minimum statistic conjunction across the 3MACMs (left/right anterior insula or dACC seeded) revealedoverlappingcoactivation inall 3of these regions.UsingRSfunc-tionalMRIdata from99healthyparticipants (21-60years old),we similarly seeded each of the common gray matter loss re-gions in task-independentFCanalyses (Figure 5B). A conjunc-tion across the 3 RS-FC maps revealed a similar overlappingpattern of connectivity as in theMACManalysis. This overlapwas further verified in a conjunction across the MACMs andRS-FC maps (Figure 5C; eTable 9 in the Supplement), whichdemonstrated coactivation and RS-FCwith the anterior insu-lae and dACC for each of the common gray matter loss re-gions. Thus,we showed in 2 independent healthy participantdata sets that the regions in which gray matter loss was ob-served in a transdiagnostic fashion in the VBMmeta-analysisform a closely interacting functional network across a broadrange of tasks and in the task-independent RS.48

Behavioral Correlates of Decreased Anterior Insulaand dACC Structural IntegrityAmong other functions, activity in the dACC and anterior in-sula is thought to signal events that deviate from expecta-tions, which is then used to drive adaptive behavioralcontrol.48-51 Executivedysfunction is an important transdiag-nosticdomainofdisruptedadaptivecontrol inmental illness.52

Altered structural integrity of the commongraymatter loss re-gions may predict performance on cognitive tests of execu-tive function. To test this hypothesis,weused a separate dataset of 163 healthy participants (21-60 years old)whohad com-pletedacomputerizedneurocognitiveassessmentbattery thatcovered a broad range of basic cognitive and higher-level ex-ecutive functions. Because these tasks assessed overlappingcognitive domains and because performance data are par-tially correlated across tasks, we conducted a data-reductionstep using a principal components analysis. This resulted inthe identification of 3 principal components (eTable 10 in theSupplement), which reflected general executive functioning(task switching, interference, andworkingmemory), the spe-cificdomainof sustainedattention, andcombinedgeneral cog-nitive and performance speed (simple reaction time and fin-ger tapping), consistent with identification of an overridingexecutive function factor in latent variable analyses of cogni-tive performance data.53,54 We then correlated individualbehavioralperformanceoneachof thesecomponentswithpar-ticipant-specific graymatter volumes,measuredusingwhole-brain volume-corrected VBM, while controlling for age, edu-cation, sex, and their interactions. After adjusting for thesecovariates, lower graymatter across the 3 common graymat-ter loss regions still predicted worse performance in terms ofgeneral executive function (standardized = 0.24;P = .007; r2

change over covariate-only model, 0.025; Figure 6A), with a

Figure 5. CommonGrayMatter Loss Regions From the Voxel-BasedMorphometryMeta-analysis Are Part of An Interconnected Brain Network

Coactivated regions during tasksA Connected regions at restB

Conjunction across task coactivationand resting connectivity

C

L insula L insula R insula

Cingulate

Overlap(MACM)

Overlap(RS-FC)

Cingulate

R insula

A, Meta-analytic coactivationmaps (MACMs) showing regions coactivated witheach of the common gray matter loss regions in healthy participant task-basedactivation studies in the BrainMap database, as well as a conjunction across all3 MACMmaps. B, Resting-state (RS) functional connectivity (FC) in healthyindividuals seeded by each of the common gray matter loss regions, as well as a

conjunction across all RS-FCmaps. C, Conjunction across all of the MACMs andRS-FCmap demonstrates that each of the common gray matter loss regionsshows both task-dependent and task-independent FC with the bilateral anteriorinsula and dorsal anterior cingulate (the regions showing consistent gray matterchanges) as well as the thalamus. L indicates left; and R, right.






similar-directiontrendforsustainedattention ( = 0.19;P = .10;Figure 6B), but no effect on general cognitive and perfor-mance speed ( = 0.06; P = .63; Figure 6C). Graymatter vol-ume in theprimary visual cortex, a control region,wasnot re-lated to anyof theperformancemeasures ( < 0.07;P > .43 forall). These results suggest that lower gray matter in the com-mongraymatter loss regions (ie,morepatient-like regionalvol-ume) is associatedwith poorer executive functioning but notgeneral aspects of task performance.

DiscussionIn this study, we identified a transdiagnostic pattern of graymatter loss in the anterior insula and dACC across psychiatricpatients, reflecting volumetric change within an intercon-nected network. Follow-up analyses in healthy participantssuggest that decreased graymatter in these regions is associ-ated with worse executive functioning. In contrast to thissharedneural substrate, diagnosis-specific effectswere foundonly for SCZ anddepression. Secondary analyses also suggestthat these findings are likely not due tomedication effects orthe presence of a common comorbid diagnosis across ourgroups.

Cognitive symptoms are part of the diagnostic criteria formany (butnot all) psychiatricdiagnoses.Our connectionof ex-ecutive functioning to integrityof awell-establishedbrainnet-work that is perturbed across a broad range of psychiatric di-agnoses helps ground a transdiagnostic understanding ofmental illness in a context suggestive of common neuralmechanisms for disease etiology and/or expression. Execu-tive dysfunction also predicts socio-occupational impair-ment, a central problem in the lives ofmanypatientswithpsy-chiatric illness.55,56 In lightof these associations, itmaybe that

the common graymatter loss in the anterior insula and dACCaccounts for this aspect of dysfunction in psychiatric disor-ders and perhaps less so diagnosis-specific symptoms.

While our datadidnot include tests of emotional process-ing, there may also be a relationship between the volumes ofthese regions and emotional perturbations, given the role oftheanterior insula anddACC inemotionalprocessingand theirabnormal activation during affective tasks in at least some ofthe assesseddisorders.4,5,57-59 Additional emotion-related ab-normalities in individuals with decreased anterior insula anddACC gray matter would only further compound their func-tional impairment.Moreover, graph theoretical workwith RSdata suggests that theremay even be 2 adjacent but function-allydistinct insula-dACCnetworks,60 onepotentiallymore in-volved in task control while the other in salience processing.

Convergent data also come fromworkonneurodegenera-tive disorders. Specific dementia syndromes have been re-lated to regionally specific gray matter loss in well-describedbrainnetworks.36Of these, the anterior insula anddACChavebeen implicated inbehavioral variant frontotemporal demen-tia, themost psychiatric like of dementia syndromes,61whichcan be mistaken for a psychiatric disorder early in its course.

Available evidence in SCZ andposttraumatic stress disor-der suggests that insula and dACC gray matter loss may re-flect the illness itself rather than a risk state. In SCZ, volumein these regions decreaseswith psychosis onset relative to in-dividuals inahigh-risk state.62,63Decreased insulaordACCvol-ume is seen in individuals with recent-onset posttraumaticstress disorder,64,65 butnot the twinof patientswithposttrau-matic stress disorder, who would carry their genetic risk butnot have been exposed to trauma.66 However, there may becertain risk states in which insula/dACC volumes are re-duced, suchaschildhoodmaltreatment,67,68which is a risk fac-tor for most psychiatric diagnoses.

Figure 6. Relationship Between GrayMatter Volume in the CommonGrayMatter Loss Regions and Performance on a Computerized Batteryof Behavioral Cognitive Tests

2 1

3

1

2

0

1

2

3 3 31 2

Gene

ral S

peed

, z S

core

Gray Matter Volume, z Score0

SpeedC

2 1

3

1

2

0

1

2

3 3 31 2

Sust

aine

d At

tent

ion,

z S

core


AttentionB

2 1

3

1

2

0

1

2

3 3 31 2

Exec

utiv

e Fu

nctio

n, z

Sco

re


Executive functionA

Based on a principal components analysis, cognitive test performance was reduced to 3 components: general executive function, the specific domain of sustainedattention, and general cognitive performance and performance speed. Lower voxel-basedmorphometrymeasured gray matter volume in these regions isassociated with worse executive functioning (A) and a trend for worse sustained attention (B) but not general performance and speed (C).






Structural neuroimaging meta-analyses have been previ-ously reported for a number of psychiatric disorders.3,69-83

However, these have either focused on either single diagnosesor on only 2 phenotypically closely related groups.71,77 As aconsequence, interpretation of findings has often reflecteddiagnosis-specific neural circuit models. Likewise, othermeta-analyses have not provided spatially unbiased informa-tion across the brain (eg, in manual volumetric tracingstudies).84 In meta-analytically summarizing a more com-plete spectrum of psychopathology across the entire brain,our findings emphasized the biological commonalities thatmay have been underappreciated in prior work. Indeed, ouronly diagnosis-specific findings are in the association ofdecreased gray matter volumes with MDD84 and associationof SCZ with a combination of decreased medial prefrontal,medial temporal, and thalamic gray matter and increasedstriatal gray matter. The lack of an effect of comorbid disorder(which may otherwise reflect common neurobiology) may bepartly owing to the fact that many investigators may haverecruited more clinically pure populations. In light of thecommon neurobiological changes observed here, which aregreatest in the disorder that is most disruptive to functioning(ie, SCZ), future work can focus on determining which aspectof the course of mental illnesses insula and dACC gray matterloss best represent. For example, the severity of an illnessmay be indexed by its chronicity, diagnostic comorbidity, andcurrent symptom levels, any or all of which may be reflectedin greater gray matter loss.

A few additional factors are also important to consider ininterpreting our findings. First, we found no effects of cur-rentmedicationuseongraymatter volume, andpriorworkex-

amining these effects havenot implicated reductions in ante-rior insula and dACC gray matter as a result of psychotropicmedication.38-44Nonetheless,wecannot ruleoutamoresubtleeffect of prior medication use. Similarly, nonabuse levels ofsmoking, alcohol, or other druguse cannot be excluded as ex-planatory factors.85 Second, comparisons between indi-vidual diagnoses may fail to yield evidence of distinct defi-cits by virtue of insufficient power due to limited sample sizefor smaller magnitude effects.

ConclusionsOur findings suggest that a generalmapping exists between abroadrangeofsymptomsandthe integrityofananterior insula/dACCbased network across awide variety of neuropsychiat-ric illnesses. These results do not imply that phenotypic dif-ferencesbetweendiagnosesarenegligible.Rather, theyprovideanorganizingmodel that emphasizes the import of shareden-dophenotypes across psychopathology, which is not cur-rently an explicit component of psychiatric nosology. Thistransdiagnosticperspective is consistent,however,withnewerdimensional models such as the National Institute of MentalHealths Research Domain Criteria Project.86 Although thisshared neural substrate suggests common brain structuralchanges at some level, it is likely that these changes reflect adiverse set of etiologies. Nonetheless, the fact that commonstructural changes are seen despite potentially differing eti-ologies raises the possibility that some interventions that tar-get theanterior insula anddACCmayproveofbroaduseacrosspsychopathology.

ARTICLE INFORMATION

Submitted for Publication:March 18, 2014; finalrevision received August 13, 2014; accepted August18, 2014.

Published Online: February 4, 2015.doi:10.1001/jamapsychiatry.2014.2206.

Author Affiliations: Veterans Affairs Palo AltoHealthcare System and the Sierra Pacific MentalIllness, Research, Education, and Clinical Center(MIRECC), Palo Alto, California (Goodkind, Oathes,Jiang, Chang, Jones-Hagata, Ortega, Zaiko, Roach,Etkin); Department of Psychiatry and BehavioralSciences, Stanford University School of Medicine,Stanford, California (Goodkind, Oathes, Jiang,Chang, Jones-Hagata, Ortega, Zaiko, Roach, Etkin);Institute for Neuroscience andMedicine (INM-1),Research Center Jlich, Jlich, Germany (Eickhoff);Institute for Clinical Neuroscience andMedicalPsychology, Heinrich-Heine University Dsseldorf,Dsseldorf, Germany (Eickhoff); Brain DynamicsCentre, WestmeadMillennium Institute and SydneyMedical SchoolWestmead, Sydney, Australia(Korgaonkar, Grieve); Sydney Translational ImagingLaboratory, SydneyMedical School, University ofSydney, Sydney, Australia (Korgaonkar, Grieve);Department of Psychiatry, New York University,New York (Galatzer-Levy); Research ImagingInstitute, University of Texas Health Science Centerat San Antonio (Fox); South Texas Veterans HealthCare System, San Antonio (Fox); School ofHumanities, University of Hong Kong, Hong Kong,

China (Fox); State Key Laboratory for Brain andCognitive Science, University of Hong Kong, HongKong, China (Fox).

Author Contributions:Dr Etkin had full access toall of the data in the study and takes responsibilityfor the integrity of the data and the accuracy of thedata analysis. Drs Goodkind and Eickhoffcontributed equally as first authors.Study concept and design: Goodkind, Eickhoff,Oathes, Chang, Fox, Etkin.Acquisition, analysis, or interpretation of data: Allauthors.Drafting of the manuscript: Goodkind, Eickhoff,Oathes, Jiang, Jones-Hagata, Zaiko, Grieve,Galatzer-Levy, Fox, Etkin.Critical revision of the manuscript for importantintellectual content: Goodkind, Eickhoff, Chang,Ortega, Roach, Korgaonkar, Grieve, Galatzer-Levy,Fox, Etkin.Statistical analysis: Goodkind, Eickhoff, Jiang,Roach, Korgaonkar, Galatzer-Levy, Fox, Etkin.Obtained funding: Grieve, Fox, Etkin.Administrative, technical, or material support:Goodkind, Oathes, Jones-Hagata, Ortega, Grieve,Etkin.Study supervision: Goodkind, Grieve, Etkin.

Conflict of Interest Disclosures:None reported.

Funding/Support:Drs Goodkind and Etkin werefunded by the Sierra-Pacific Mental Illness Research,Education and Clinical Center (MIRECC) at the PaloAlto Veterans Affairs. Dr Eickhoff was supported by

the Deutsche Forschungsgemeinschaft (EI 816/4-1;SBE and LA 3071/3-1), the National Institute ofMental Health (R01-MH074457), and the EuropeanCommission (Human Brain Project). Drs Korgaonkarand Grieve were funded by the National Health andMedical Research Council of Australia project grantsscheme (project grant 1008080) and the SydneyUniversity Medical Foundation. Comprehensiveaccess to the BrainMap database was authorized bya collaborative-use license agreement. BrainMap issupported by National Institutes of Health/NationalInstitute of Mental Health Award R01MH074457. AllVBM studies reported in this article are now codedand accessible through BrainMap.

Role of the Funder/Sponsor: The funders had norole in the design and conduct of the study;collection, management, analysis, andinterpretation of the data; preparation, review, orapproval of themanuscript; and decision to submitthemanuscript for publication.

Additional Contributions:We acknowledge use ofdata from the BRAINnet Foundation, a 501(c)(3)nonprofit organization, which administers access todata from the Brain Resource Internationaldatabase and the scientific projects that havecontributed to this database. We acknowledge thehelp of Michelle Wang, BPsych, in analysis of theBRAINnet data, and feedback on themanuscriptfromMichael Greicius, MD, Alan Schatzberg, MD,Anett Gyurak, PhD, and Greg Fonzo, PhD (StanfordUniversity School of Medicine), and Lisa McTeague,






PhD (Medical University of South Carolina). Nofunding was received for these contributions.

REFERENCES

1. Spitzer RL, Endicott J, Robins E. Researchdiagnostic criteria: rationale and reliability. Arch GenPsychiatry. 1978;35(6):773-782.

2. Fischer BA. A review of American psychiatrythrough its diagnoses: the history and developmentof the Diagnostic and Statistical Manual of MentalDisorders. J NervMent Dis. 2012;200(12):1022-1030.

3. Ellison-Wright I, Bullmore E. Anatomy of bipolardisorder and schizophrenia: a meta-analysis.Schizophr Res. 2010;117(1):1-12.

4. Etkin A, Wager TD. Functional neuroimaging ofanxiety: a meta-analysis of emotional processing inPTSD, social anxiety disorder, and specific phobia.Am J Psychiatry. 2007;164(10):1476-1488.

5. Hamilton JP, Etkin A, Furman DJ, LemusMG,Johnson RF, Gotlib IH. Functional neuroimaging ofmajor depressive disorder: a meta-analysis and newintegration of base line activation and neuralresponse data.AmJPsychiatry. 2012;169(7):693-703.

6. Baker JT, Holmes AJ, Masters GA, et al.Disruption of cortical association networks inschizophrenia and psychotic bipolar disorder. JAMAPsychiatry. 2013;71(2):109-118.

7. Menon V. Large-scale brain networks andpsychopathology: a unifying triple network model.Trends Cogn Sci. 2011;15(10):483-506.

8. Whitfield-Gabrieli S, Ford JM. Default modenetwork activity and connectivity inpsychopathology. Annu Rev Clin Psychol. 2012;8:49-76.

9. Cross-Disorder Group of the PsychiatricGenomics Consortium. Identification of risk lociwith shared effects on five major psychiatricdisorders: a genome-wide analysis. Lancet. 2013;381(9875):1371-1379.

10. Clark LA, Watson D, Reynolds S. Diagnosis andclassification of psychopathology: challenges to thecurrent system and future directions. Annu RevPsychol. 1995;46:121-153.

11. Jovicich J, Marizzoni M, Sala-Llonch R, et al;PharmaCog Consortium. Brain morphometryreproducibility in multi-center 3TMRI studies:a comparison of cross-sectional and longitudinalsegmentations.Neuroimage. 2013;83:472-484.

12. Kober H, Wager TD. Meta-analysis ofneuroimaging data.Wiley Interdiscip Rev Cogn Sci.2010;1(2):293-300.

13. Cox BJ, Clara IP, Hills AL, Sareen J.Obsessive-compulsive disorder and the underlyingstructure of anxiety disorders in a nationallyrepresentative sample: confirmatory factor analyticfindings from the German Health Survey. J AnxietyDisord. 2010;24(1):30-33.

14. Slade T,Watson D. The structure of commonDSM-IV and ICD-10mental disorders in the Australiangeneral population. PsycholMed. 2006;36(11):1593-1600.

15. Vaidyanathan U, Patrick CJ, IaconoWG.Examining the overlap between bipolar disorder,nonaffective psychosis, and commonmentaldisorders using latent class analysis. Psychopathology.2012;45(6):361-365.

16. Wright AG, Krueger RF, HobbsMJ, Markon KE,Eaton NR, Slade T. The structure of psychopathology:

toward an expanded quantitative empiricalmodel.J AbnormPsychol. 2013;122(1):281-294.

17. Kendler KS, Prescott CA, Myers J, Neale MC.The structure of genetic and environmental riskfactors for common psychiatric and substance usedisorders in men and women. Arch Gen Psychiatry.2003;60(9):929-937.

18. Zielinski BA, PriggeMB, Nielsen JA, et al.Longitudinal changes in cortical thickness in autismand typical development. Brain. 2014;137(pt 6):1799-1812.

19. Ecker C, Shahidiani A, Feng Y, et al. The effect ofage, diagnosis, and their interaction on vertex-basedmeasures of cortical thickness and surface area inautism spectrumdisorder. J Neural Transm. 2014;121(9):1157-1170.

20. Greimel E, Nehrkorn B, Schulte-Rther M, et al.Changes in grey matter development in autismspectrum disorder. Brain Struct Funct. 2013;218(4):929-942.

21. Lancaster JL, Tordesillas-Gutirrez D, MartinezM, et al. Bias betweenMNI and Talairachcoordinates analyzed using the ICBM-152 braintemplate.Hum Brain Mapp. 2007;28(11):1194-1205.

22. Eickhoff SB, Bzdok D, Laird AR, Kurth F, Fox PT.Activation likelihood estimationmeta-analysisrevisited.Neuroimage. 2012;59(3):2349-2361.

23. Eickhoff SB, Laird AR, Grefkes C, Wang LE,Zilles K, Fox PT. Coordinate-based activationlikelihood estimationmeta-analysis ofneuroimaging data: a random-effects approachbased on empirical estimates of spatial uncertainty.Hum Brain Mapp. 2009;30(9):2907-2926.

24. Turkeltaub PE, Eickhoff SB, Laird AR, FoxM,Wiener M, Fox P. Minimizing within-experiment andwithin-group effects in activation likelihoodestimationmeta-analyses.Hum Brain Mapp. 2012;33(1):1-13.

25. Jakobs O, Langner R, Caspers S, et al.Across-study and within-subject functionalconnectivity of a right temporo-parietal junctionsubregion involved in stimulus-context integration.Neuroimage. 2012;60(4):2389-2398.

26. Nichols T, Brett M, Andersson J, Wager T, PolineJB. Valid conjunction inference with theminimumstatistic. Neuroimage. 2005;25(3):653-660.

27. Eickhoff SB, Bzdok D, Laird AR, et al.Co-activation patterns distinguish cortical modules,their connectivity and functional differentiation.Neuroimage. 2011;57(3):938-949.

28. Eickhoff SB, Jbabdi S, Caspers S, et al.Anatomical and functional connectivity ofcytoarchitectonic areas within the human parietaloperculum. J Neurosci. 2010;30(18):6409-6421.

29. Laird AR, Eickhoff SB, Fox PM, et al. TheBrainMap strategy for standardization, sharing, andmeta-analysis of neuroimaging data. BMC Res Notes.2011;4:349.

30. Laird AR, Eickhoff SB, Kurth F, et al. ALEmeta-analysisworkflows via the BrainMapdatabase:progress towards a probabilistic functional brainatlas. Front Neuroinform. 2009;3:23.

31. Fox PT, Lancaster JL. Opinion: mapping contextand content: the BrainMapmodel.Nat Rev Neurosci.2002;3(4):319-321.

32. Robinson JL, Laird AR, Glahn DC, Lovallo WR,Fox PT. Metaanalytic connectivitymodeling:

delineating the functional connectivity of the humanamygdala.HumBrainMapp. 2010;31(2):173-184.

33. Grieve SM, Korgaonkar MS, Clark CR, WilliamsLM. Regional heterogeneity in limbic maturationalchanges: evidence from integrating corticalthickness, volumetric and diffusion tensor imagingmeasures. Neuroimage. 2011;55(3):868-879.

34. Koslow SH,Wang Y, Palmer DM, Gordon E,Williams LM. BRAINnet: a standardized global humanbrain project. Technol Innovation. 2013;15:17-29.

35. Paul RH, Lawrence J, Williams LM, Richard CC,Cooper N, Gordon E. Preliminary validity ofintegneuro: a new computerized battery ofneurocognitive tests. Int J Neurosci. 2005;115(11):1549-1567.

36. SeeleyWW, Crawford RK, Zhou J, Miller BL,Greicius MD. Neurodegenerative diseases targetlarge-scale human brain networks.Neuron. 2009;62(1):42-52.

37. Alexander-Bloch A, Giedd JN, Bullmore E.Imaging structural co-variance between humanbrain regions.Nat Rev Neurosci. 2013;14(5):322-336.

38. Chakos MH, Shirakawa O, Lieberman J, Lee H,Bilder R, Tamminga CA. Striatal enlargement in ratschronically treated with neuroleptic. Biol Psychiatry.1998;44(8):675-684.

39. Chua SE, Deng Y, Chen EY, et al. Early striatalhypertrophy in first-episode psychosis within 3weeks of initiating antipsychotic drug treatment.Psychol Med. 2009;39(5):793-800.

40. Torres US, Portela-Oliveira E, Borgwardt S,Busatto GF. Structural brain changes associatedwith antipsychotic treatment in schizophrenia asrevealed by voxel-basedmorphometric MRI: anactivation likelihood estimationmeta-analysis. BMCPsychiatry. 2013;13:342.

41. Ho BC, Andreasen NC, Ziebell S, Pierson R,Magnotta V. Long-term antipsychotic treatmentand brain volumes: a longitudinal study offirst-episode schizophrenia. Arch Gen Psychiatry.2011;68(2):128-137.

42. Selemon LD, LidowMS, Goldman-Rakic PS.Increased volume and glial density in primateprefrontal cortex associated with chronicantipsychotic drug exposure. Biol Psychiatry. 1999;46(2):161-172.

43. Tomelleri L, Jogia J, Perlini C, et al. Brainstructural changes associated with chronicity andantipsychotic treatment in schizophrenia. EurNeuropsychopharmacol. 2009;19(12):835-840.

44. McCormick L, Decker L, Nopoulos P, Ho BC,Andreasen N. Effects of atypical and typicalneuroleptics on anterior cingulate volume inschizophrenia. Schizophr Res. 2005;80(1):73-84.

45. McClure RK, Carew K, Greeter S, Maushauer E,Steen G, Weinberger DR. Absence of regional brainvolume change in schizophrenia associated withshort-term atypical antipsychotic treatment.Schizophr Res. 2008;98(1-3):29-39.

46. Vernon AC, CrumWR, Lerch JP, et al. Reducedcortical volume and elevated astrocyte density in ratschronically treatedwith antipsychotic drugs-linkingmagnetic resonance imaging findings to cellularpathology.Biol Psychiatry. 2014;75(12):982-990.

47. Laird AR, Eickhoff SB, Rottschy C, Bzdok D, RayKL, Fox PT. Networks of task co-activations.Neuroimage. 2013;80:505-514.






48. Smith SM, Fox PT, Miller KL, et al.Correspondence of the brains functionalarchitecture during activation and rest. Proc NatlAcad Sci U S A. 2009;106(31):13040-13045.

49. Menon V, Uddin LQ. Saliency, switching,attention and control: a network model of insulafunction. Brain Struct Funct. 2010;214(5-6):655-667.

50. Dosenbach NU, Visscher KM, Palmer ED, et al.A core system for the implementation of task sets.Neuron. 2006;50(5):799-812.

51. Sridharan D, Levitin DJ, Menon V. A critical rolefor the right fronto-insular cortex in switchingbetween central-executive and default-modenetworks. Proc Natl Acad Sci U S A. 2008;105(34):12569-12574.

52. Etkin A, Gyurak A, OHara R. A neurobiologicalapproach to the cognitive deficits of psychiatricdisorders.DialoguesClinNeurosci. 2013;15(4):419-429.

53. Miyake A, Friedman NP, EmersonMJ, WitzkiAH, Howerter A, Wager TD. The unity and diversityof executive functions and their contributions tocomplex frontal lobe tasks: a latent variableanalysis. Cogn Psychol. 2000;41(1):49-100.

54. Adrover-Roig D, Ses A, Barcel F, Palmer A.A latent variable approach to executive control inhealthy ageing. Brain Cogn. 2012;78(3):284-299.

55. Depp CA, Mausbach BT, Harmell AL, et al.Meta-analysis of the association between cognitiveabilities and everyday functioning in bipolardisorder. Bipolar Disord. 2012;14(3):217-226.

56. Harvey PD, Strassnig M. Predicting the severityof everyday functional disability in people withschizophrenia: cognitive deficits, functionalcapacity, symptoms, and health status.WorldPsychiatry. 2012;11(2):73-79.

57. Craig AD. How do you feel? interoception: thesense of the physiological condition of the body.Nat Rev Neurosci. 2002;3(8):655-666.

58. Etkin A, Egner T, Kalisch R. Emotionalprocessing in anterior cingulate andmedialprefrontal cortex. Trends Cogn Sci. 2011;15(2):85-93.

59. Mechias ML, Etkin A, Kalisch R. Ameta-analysisof instructed fear studies: implications for consciousappraisal of threat. Neuroimage. 2010;49(2):1760-1768.

60. Power JD, Cohen AL, Nelson SM, et al.Functional network organization of the humanbrain.Neuron. 2011;72(4):665-678.

61. Pressman PS, Miller BL. Diagnosis andmanagement of behavioral variant frontotemporaldementia. Biol Psychiatry. 2014;75(7):574-581.

62. Fusar-Poli P, Radua J, McGuire P, Borgwardt S.Neuroanatomical maps of psychosis onset:

voxel-wise meta-analysis of antipsychotic-naiveVBM studies. Schizophr Bull. 2012;38(6):1297-1307.

63. Palaniyappan L, Balain V, Liddle PF. Theneuroanatomyof psychotic diathesis: ameta-analyticreview. J Psychiatr Res. 2012;46(10):1249-1256.

64. Chen Y, Fu K, Feng C, et al. Different regionalgray matter loss in recent onset PTSD and nonPTSD after a single prolonged trauma exposure.PLoS One. 2012;7(11):e48298.

65. Corbo V, Clment MH, Armony JL, PruessnerJC, Brunet A. Size versus shape differences:contrasting voxel-based and volumetric analyses ofthe anterior cingulate cortex in individuals withacute posttraumatic stress disorder. Biol Psychiatry.2005;58(2):119-124.

66. Kasai K, Yamasue H, GilbertsonMW, ShentonME, Rauch SL, Pitman RK. Evidence for acquiredpregenual anterior cingulate gray matter loss from atwin study of combat-related posttraumatic stressdisorder. Biol Psychiatry. 2008;63(6):550-556.

67. Dannlowski U, Stuhrmann A, Beutelmann V,et al. Limbic scars: long-term consequences ofchildhoodmaltreatment revealed by functional andstructural magnetic resonance imaging. BiolPsychiatry. 2012;71(4):286-293.

68. Tomoda A, Suzuki H, Rabi K, Sheu YS, Polcari A,TeicherMH. Reduced prefrontal cortical graymattervolume in young adults exposed to harsh corporalpunishment.Neuroimage. 2009;47(suppl 2):T66-T71.

69. Bora E, Fornito A, Pantelis C, Ycel M. Graymatter abnormalities in major depressive disorder:a meta-analysis of voxel basedmorphometrystudies. J Affect Disord. 2012;138(1-2):9-18.

70. Bora E, Fornito A, Ycel M, Pantelis C.Voxelwise meta-analysis of gray matterabnormalities in bipolar disorder. Biol Psychiatry.2010;67(11):1097-1105.

71. Bora E, Fornito A, Ycel M, Pantelis C. Theeffects of gender on greymatter abnormalities inmajor psychoses: a comparative voxelwisemeta-analysis of schizophrenia and bipolardisorder. Psychol Med. 2012;42(2):295-307.

72. DuMY,Wu QZ, Yue Q, et al. Voxelwisemeta-analysis of gray matter reduction in majordepressive disorder. Prog NeuropsychopharmacolBiol Psychiatry. 2012;36(1):11-16.

73. Lai CH. Gray matter deficits in panic disorder:a pilot study ofmeta-analysis. J Clin Psychopharmacol.2011;31(3):287-293.

74. Lai CH. Gray matter volume in major depressivedisorder: a meta-analysis of voxel-basedmorphometry studies. Psychiatry Res. 2013;211(1):37-46.

75. Peng Z, Lui SS, Cheung EF, et al. Brain structuralabnormalities in obsessive-compulsive disorder:converging evidence fromwhite matter and greymatter. Asian J Psychiatr. 2012;5(4):290-296.

76. Radua J, Mataix-Cols D. Voxel-wisemeta-analysis of grey matter changes inobsessive-compulsive disorder. Br J Psychiatry.2009;195(5):393-402.

77. Radua J, van den Heuvel OA, Surguladze S,Mataix-Cols D. Meta-analytical comparison ofvoxel-basedmorphometry studies inobsessive-compulsive disorder vs other anxietydisorders. Arch Gen Psychiatry. 2010;67(7):701-711.

78. Rotge JY, Langbour N, Guehl D, et al. Graymatter alterations in obsessive-compulsivedisorder: an anatomic likelihood estimationmeta-analysis.Neuropsychopharmacology. 2010;35(3):686-691.

79. Selvaraj S, Arnone D, Job D, et al. Grey matterdifferences in bipolar disorder: a meta-analysis ofvoxel-basedmorphometry studies. Bipolar Disord.2012;14(2):135-145.

80. Ersche KD,Williams GB, Robbins TW, BullmoreET. Meta-analysis of structural brain abnormalitiesassociated with stimulant drug dependence andneuroimaging of addiction vulnerability andresilience. Curr Opin Neurobiol. 2013;23(4):615-624.

81. Khn S, Gallinat J. Gray matter correlates ofposttraumatic stress disorder: a quantitativemeta-analysis. Biol Psychiatry. 2013;73(1):70-74.

82. Bora E, Fornito A, Radua J, et al.Neuroanatomical abnormalities in schizophrenia:a multimodal voxelwise meta-analysis andmeta-regression analysis. Schizophr Res. 2011;127(1-3):46-57.

83. Houenou J, Frommberger J, Carde S, et al.Neuroimaging-basedmarkers of bipolar disorder:evidence from twometa-analyses. J Affect Disord.2011;132(3):344-355.

84. KemptonMJ, Salvador Z, Munaf MR, et al.Structural neuroimaging studies in majordepressive disorder. Meta-analysis and comparisonwith bipolar disorder. Arch Gen Psychiatry. 2011;68(7):675-690.

85. Pan P, Shi H, Zhong J, et al. Chronic smokingand brain gray matter changes: evidence frommeta-analysis of voxel-basedmorphometrystudies. Neurol Sci. 2013;34(6):813-817.

86. Insel T, Cuthbert B, GarveyM, et al. Researchdomain criteria (RDoC): toward a new classificationframework for research onmental disorders. Am JPsychiatry. 2010;167(7):748-751.





goodkind 2015

Documents