ORIGINAL ARTICLE

Reduced GABAA benzodiazepine receptor binding inveterans with post-traumatic stress disorderE Geuze1,2, BNM van Berckel2,3, AA Lammertsma3, R Boellaard3, CS de Kloet1, E Vermetten1,2

and HGM Westenberg2

1Research Centre—Military Mental Healthcare, Ministry of Defense, Utrecht, The Netherlands; 2Department of Psychiatry,Rudolf Magnus Institute of Neuroscience, Utrecht University Medical Centre, Utrecht, The Netherlands and 3Department ofNuclear Medicine & PET Research, VU University Medical Centre, Amsterdam, The Netherlands

c-Aminobutyric acid (GABAA) receptors are thought to play an important role in modulatingthe central nervous system in response to stress. Animal data have shown alterations in theGABAA receptor complex by uncontrollable stressors. SPECT imaging with benzodiazepineligands showed lower distribution volumes of the benzodiazepine-GABAA receptor in theprefrontal cortex of patients with post-traumatic stress disorder (PTSD) in one, but notin another study. The objective of the present study was to assess differences in thebenzodiazepine-GABAA receptor complex in veterans with and without PTSD using[11C]flumazenil and positron emission tomography (PET). Nine drug naive male Dutch veteranswith deployment related PTSD and seven male Dutch veterans without PTSD were recruited,and matched for age, region and year of deployment. Each subject received a [11C]flumazenilPET scan and a structural magnetic resonance imaging scan. Dynamic 3D PET scans with atotal duration of 60 min were acquired, and binding in template based and manually definedregions of interest (ROI) was quantified using validated plasma input and reference tissuemodels. In addition, parametric binding potential images were compared on a voxel-by-voxelbasis using statistical parametric mapping (SPM2). ROI analyses using both template basedand manual ROIs showed significantly reduced [11C]flumazenil binding in PTSD subjectsthroughout the cortex, hippocampus and thalamus. SPM analysis confirmed these results.The observed global reduction of [11C]flumazenil binding in patients with PTSD providescircumstantial evidence for the role of the benzodiazepine-GABAA receptor in the pathophy-siology of PTSD and is consistent with previous animal research and clinical psychopharma-cological studies.Molecular Psychiatry (2008) 13, 74–83; doi:10.1038/sj.mp.4002054; published online 31 July 2007

Keywords: PTSD; GABAA receptors; benzodiazepine; positron emission tomography; flumaze-nil; hippocampus

Introduction

Anxiety disorders are associated with alterations in anumber of neuroendocrine and neurotransmissionsystems, including the hypothalamic-pituitary-adre-nal axis, and serotonin, noradrenalin, and g-amino-butyric acid (GABA) systems. GABA is the principalinhibitory neurotransmitter in the brain, exertingcontrol over excitability in most brain areas. GABAplays an important role in homeostasis during stressand alterations in GABAergic systems have beenimplicated in the pathogenesis of anxiety disorders,including posttraumatic stress disorder (PTSD) and

depression. Low plasma GABA levels after a trau-matic event are predictive of subsequent developmentof PTSD,1 suggesting that the GABA system isinvolved in the pathophysiology of PTSD. Involve-ment of the GABA system in PTSD has beenexamined using a variety of different paradigms,including preclinical studies, pharmacologic studiesand neuroimaging techniques.

In the stress–restress paradigm, an animal modelthat shows resemblance to PTSD, stress evoked asustained decrease in hippocampal GABA levels.2

Exposure of rodents to inescapable foot shock, amodel for stress-related depression characterized bydeficits in learning and memory, resulted in de-creased GABAA function and binding in the cerebralcortex and hippocampus.3–5 Benzodiazepines, whichmodulate the GABAA receptor function, inhibit thestartle response induced by predator stress.6 Thesepredator stress paradigms model aspects of hyper-arousal seen in patients with PTSD.

Received 10 August 2006; revised 2 May 2007; accepted 7 May2007; published online 31 July 2007

Correspondence: Dr E Geuze, Research Centre—Military MentalHealthcare, Ministry of Defense, PO Box 90.000, 3509 AA Utrecht,The Netherlands.E-mail: s.g.geuze@umcutrecht.nl

Molecular Psychiatry (2008) 13, 74–83& 2008 Nature Publishing Group All rights reserved 1359-4184/08 $30.00

www.nature.com/mp

Pharmacological studies have also provided sup-port for the involvement of the GABA system inPTSD. Benzodiazepines have fast-acting anxietyreducing properties, which have led to their wide-spread use as treatment for anxiety disorders, includ-ing PTSD. Although benzodiazepines are effective inreducing anxiety in PTSD, they have no effect on thecore symptoms of this disorder, such as intrusivethoughts, numbing and hyperarousal.7–9 A putativerole of GABA in PTSD is also supported by treatmentstudies with other GABAergic compounds, such astiagabine, a selective GABA reuptake inhibitor.10,11

Neuroimaging techniques provide a unique oppor-tunity to investigate receptor binding and neuro-transmitter release in vivo. Only a few neuroimagingstudies of the benzodiazepine-GABAA receptor inPTSD have been performed and all these studies usedsingle photon emission computed tomography(SPECT) and [123I]iomazenil. Using this technique, ithas been reported that Vietnam veterans with PTSDhad lower volumes of distribution of [123I]iomazenilin the prefrontal cortex (Brodmann’s area 9) whencompared to healthy controls.12 A more recent[123I]iomazenil SPECT study, however, was unable toconfirm these results in Gulf War veterans comparedto undeployed military controls.13

Positron emission tomography (PET) using [11C]flu-mazenil is a more accurate technique for quantifyingbenzodiazepine-GABAA receptor binding than SPECTimaging using [123I]iomazenil. Apart from being fullyquantitative, PET also has superior spatial resolutioncompared to SPECT. PET imaging using [11C]fluma-zenil has been employed frequently in patients withepilepsy14–16 and in patients with schizophrenia.17

PET is also an important neuroimaging tool used instudies of psychiatric drug action and in novel drugdevelopment.18 To the best of our knowledge, thepresent study is the first to investigate GABAergicfunctioning in PTSD using PET and [11C]flumazenil.

Based on preclinical studies and clinical findings,the working hypothesis of the present study was thatveterans with PTSD have reduced [11C]flumazenil-binding potential in both cortical areas and thehippocampus compared to veterans without PTSD.

Materials and methods

SubjectsNine male Dutch veterans with PTSD (patient group)and seven male Dutch veterans without PTSD (con-trol group) were recruited for this study. Veteranswith PTSD were recruited from the Department ofMilitary Psychiatry at the Central Military Hospital inUtrecht. Control veterans were recruited via directmail to veterans who were registered at the VeteransInstitute in The Netherlands. All veterans had servedin UN peacekeeping missions in Lebanon and Bosnia.Control veterans were matched to the patient groupwith respect to age, year of deployment and countryof deployment. None of the veterans included wasphysically injured during the time of deployment. At

study entry, all veterans were medication naive (thatis, they had never used psychotropic medication).In addition, none of the veterans had a history ofbenzodiazepine usage or alcohol abuse within 6months before the study. Urinary drug screeningwas performed in all veterans on the morning of theday on which the PET scan was performed. None ofthe veterans needed to be excluded from the studybecause of benzodiazepine or other drug use. Inaddition, veterans were excluded when they had ahistory of neurological illness or psychiatric illnessother than mood or anxiety disorders. In addition, forall veterans, no clinically significant abnormalitieswere visible on the MRI scan (as determined by aneuroradiologist). Finally, control veterans had tofulfil Research Diagnostic Criteria for never beingmentally ill.

All veterans were evaluated with the StructuredClinical Interview for DSM-IV (SCID;19 Clinician-Administered PTSD Scale for DSM-IV) CAPS;20

Hamilton Depression Scale (HAM-D), and HamiltonAnxiety Scale (HAM-A). In veterans with PTSD,PTSD was the primary diagnosis. PTSD was con-firmed by the CAPS and through consensus by threeclinicians (EG, EV, CdK). Control veterans withoutPTSD also met the A1 criterion for PTSD (that is, theyhad also experienced traumatic events). Controlveterans were only included if their CAPS score wasless than 20. Written informed consent was obtainedfrom all veterans after a complete written and verbaldescription of the study. A total of nine veterans withPTSD and seven veterans without PTSD were in-cluded. The study was performed between August2003 and July 2004, and was approved by the EthicalReview Boards of both the University Medical Centreof Utrecht, The Netherlands, and the VU UniversityMedical Centre, Amsterdam, The Netherlands. PETscanning was performed at the VU University Med-ical Centre in Amsterdam, The Netherlands, whileMRI scanning was performed at Utrecht UniversityMedical Centre in Utrecht, The Netherlands.

Production of [11C]flumazenil[11C]Flumazenil was produced according to GoodManufacturing Practice (GMP) guidelines (govern-ment license 107627A). Briefly, 11CO2 was producedby irradiation of 14N2/O2 (95.5/0.5%) with 18 MeVprotons using an IBA 18/9 cyclone cyclotron (IBA,Louvain-la Neuve, Belgium). After irradiation, thecollected 11CO2 was reacted with lithium aluminum-hydride in tetra-hydrofurane. The tetra-hydrofuranewas subsequently evaporated at 1301C. At thistemperature, hydrogen iodide was added to yield[11C]methyl iodide. [11C]Methyl iodide was distilledinto a second reaction vessel, containing a mixture ofRO15-5528 (des methyl flumazenil) and tetra-butylammonium hydroxide in dimethyl formamide. Thismixture was heated for 40 s at 401C to yield[11C]flumazenil. [11C]Flumazenil was purified fromthis reaction mixture by semi-preparative high-performance liquid chromatography (HPLC) on a

Reduced binding to benzodiazepine receptorE Geuze et al

75

Molecular Psychiatry

Nucleosil C18 (250� 16 mm) with 0.25 M phosphatebuffer/acetonitrile 70:30 (v/v) as eluent. [11C]fluma-zenil was extracted from the collected HPLC fractionby solid-phase extraction of a Waters Seppak tC18 andreformulated in 7.09 mM NaH2PO4 in saline and 7% ofethanol yielding a sterile, pyrogen-free and isotonicsolution of [11C]flumazenil. The average radio-chemical yield was 37% (1200–2800 MBq at time ofinjection) and, in all cases, radiochemical purity was> 99.5%, while the specific activity ranged from 23 to109 GBq/mmol at time of injection.

PETPET scans were performed on an ECAT EXACTHRþ scanner (Siemens/CTI, Knoxville, TN, USA),which is located at the Department of NuclearMedicine & PET Research of the VU UniversityMedical Centre in Amsterdam, The Netherlands. Thisscanner enables the acquisition of 63 transaxialplanes over a 15.5 cm axial field of view, thusallowing the whole brain to be imaged. Character-istics of this scanner have been described else-where.21,22 All subjects received an indwelling radialartery cannula for arterial blood sampling and avenous cannula for tracer injection.

First, a 10 min transmission scan was performed in2D acquisition mode using three retractable rotatingline sources. This scan was used to correct thesubsequent emission scan for photon attenuation.Then a dynamic emission scan in 3D acquisitionmode was started simultaneously with the intra-venous injection of 371756 MBq (mean7s.d.) of[11C]flumazenil using an infusion pump at 0.8 ml/s,after which the line was flushed with 42 ml salineat 2.0 ml/s (Med-Rad, Beek, The Netherlands). Thespecific activity (SA) was 0.03070.011 MBq/nM

(mean7s.d.). The estimated receptor binding asso-ciated with these high specific activity tracer doses is< 1%. This dynamic emission scan consisted of 16frames with progressive increase in frame duration(4�15, 4� 60, 2�150, 2�300, 4� 600 s) and a totalduration of 60 min. Arterial blood was withdrawncontinuously at a rate of 5 ml/min for the first 10 minand 2.5 ml/min thereafter, using an online detectionsystem (Veenstra Instruments, Joure, The Nether-lands), which was cross-calibrated against the PETscanner.23 At 2.5, 5, 10, 20, 30, 40, and 60 min post-injection, continuous withdrawal was briefly inter-rupted for manual collection of blood samples. Thesesamples were used to measure both whole bloodand plasma concentrations. In addition, the plasmasamples were used to determine parent [11C]flumaze-nil fractions. Motion of the subject in the PET camerawas checked visually at regular intervals (by checkingthe position of the head with laser beams).

Image reconstructionAll PET sinograms were corrected for dead time,tissue attenuation using the transmission scan, decay,scatter and randoms, and reconstructed using astandard filtered backprojection algorithm and a

Hanning filter with cutoff at 0.5 times the Nyquistfrequency. A zoom factor of 2 and a matrix size of256� 256� 63 were used, resulting in a voxel size of1.2�1.2� 2.4 mm and a spatial resolution of approxi-mately 7 mm full-width at half-maximum at the centreof the field of view. Images were then transferred toworkstations (Sun Microsystems, Santa Clara, CA,USA) for further analysis.

Magnetic resonance imaging

Magnetic resonance imaging (MRI) of all subjectswas performed at the Department of Radiology of theUniversity Medical Centre of Utrecht. MRI scans wereacquired using a 1.5T scanner (Philips Gyroscan;Philips Medical Systems, Best, The Netherlands).These scans were used for segmentation of grey andwhite matter and for delineation of regions of interest(ROI). T1-weighted, 3D, fast-field echo scans with160–180 1.2 mm contiguous coronal slices (echo time,4.6 ms; repetition time, 30 ms; flip angle 301; field ofview 256 mm) of the whole head were used forcoregistration with PET.

Region of interest definition

MRI images were aligned to corresponding PETimages using a mutual information algorithm in-cluded in MIRIT (multimodality image registrationusing information theory24,25). The data were analysedusing a probability-map-based automatic delineationof 35 ROIs that delineates the whole brain (bilateralareas were counted as one ROI).26 This method is afast, objective, reproducible, and safe way to assessregional brain values from PET or SPECT scans, whichhas previously been described and validated.26 The35 ROIs are based on manual definition of theseROIs in MRI images of 10 healthy controls. EachROI represents the volume that is present in eachindividual ROI after warping to a standard brain.Using a warping algorithm, template ROI sets definedfrom each individual were transferred to the MRimages of the individual new subject. The voxelsoverlapping after these individual warpings are usedas the final ROI. ROIs in this template include thebilateral cerebellum, bilateral orbital frontal cortex,bilateral medial inferior frontal cortex, bilateralanterior cingulate cortex, bilateral insula, bilateralthalamus, bilateral caudate nuclei, bilateral putamen,bilateral superior temporal cortex, bilateral parietalcortex, bilateral medial inferior temporal cortex,bilateral superior frontal cortex, bilateral occipitalcortex, bilateral sensory motor cortex, bilateral poster-ior cingulate cortex, bilateral enthorinal cortex,bilateral striatum, and midbrain. In addition to thistemplate, manual ROIs for bilateral hippocampi,bilateral amygdala, and pons were defined on thefused MRI scan using DISPLAY (http://www.bic.mni.mcgill.ca/) according to standard anatomicalcriteria.27,28

Reduced binding to benzodiazepine receptorE Geuze et al

76

Molecular Psychiatry

Kinetic analysisKinetic analyses were performed using dedicatedsoftware developed within Matlab 5.3 (The Math-works, Natick, MA, USA). Manual samples collectedduring scanning were used to calibrate the onlineblood curve, to determine plasma to whole bloodratios of radioactivity, and to measure plasma meta-bolite fractions. Metabolite fractions were determinedusing HPLC, as described previously.29 The onlineblood curve was calibrated using the manual wholeblood samples. Plasma to whole blood ratios werefitted to an exponential function and multiplied withthe whole blood (online sampler) curve to obtainthe corresponding (total) plasma curve. Finally, thisplasma curve was multiplied with the parent fractionusing a Hill function to obtain a metabolite-correctedplasma input function. [11C]flumazenil time-activitycurves were generated by projecting all ROIs onto allframes of the dynamic [11C]flumazenil scan. Metabo-lite-corrected plasma input curves were available forseven patients and seven controls due to technicalreasons. These data were analysed using a singletissue compartment model yielding the outcomeparameter volume of distribution (Vd). This was usedto confirm the use of the pons as reference tissue. Thesingle tissue compartment includes free, non-specificand specific compartments, and assumes that alltissue pools equilibrate rapidly with respect to blood–brain barrier transport rates.30 Data from all ninepatients and seven controls were then analysed usingthe simplified reference tissue model (SRTM) result-ing in the outcome parameter binding potential31

(BP). The use of the SRTM (with pons as referencetissue) for quantification of [11C]flumazenil BP hasrecently been validated.32 Owing to the small samplesize and because data were not normally distributed,the Mann–Whitney U-test was used for all ROI data.All statistical analyses were performed with SPSS12.0 for Windows (SPSS, Chicago, IL, USA). Thestatistical threshold for significance for all measureswas set at P < 0.05.

SPM analysisIn addition to above ROI analysis, a voxel-by-voxelSPM analysis was performed. This analysis was toassess whether there were clusters of voxels withaltered binding not identified in the ROI analysis, forexample due to heterogeneity within an ROI. BeforeSPM analysis, scans were checked visually for motionartefacts. Motion was checked by overlaying theoutline contour of the patient’s head on each frameof the dynamic study allowing assessment of inter-frame displacements. None of the scans needed to beexcluded. For the SPM analysis, BP images weregenerated using Ichise plots of dynamic data from 10to 60 min post injection.33 Before Ichise plot analysis,dynamic scans were smoothed using an additional10 mm full-width half-maximum Gaussian filter toreduce noise, thereby avoiding noise induced biasduring Ichise analysis. As the cortical distribution of[11C]flumazenil is similar to that of water, normal-

ization was performed using the standard watertemplate included within SPM. Correct normalizationwas checked visually by direct comparison of normal-ized BP images and the template used. Next, theseplots were used in a voxel-based comparison betweenthe two groups using SPM2 (http://www.fil.ion.ucl.ac.uk/spm). As images were already smoothed beforeIchise analysis, the usual smoothing within SPM wasomitted. SPM was performed without proportionalscaling. Proportional scaling may be omitted because,following smoothing, Ichise plots are quantitativelyaccurate even at lower noise levels. The images werethresholded at P < 0.001, uncorrected for multiplecomparisons, using a cluster size k > 50 voxels.34 Inaddition, the false discovery rate was used to providecorrected P-values.35

Results

Psychometric data

Veterans with PTSD had significantly greater CAPS,Hamilton A, and Hamilton D scores (Table 1).According to the SCID, the PTSD group met lifetime(past) DSM-IV (American Psychiatric Association1994) diagnostic criteria for the following disorders:major depressive disorder (n = 4), bipolar disorder(n = 1), alcohol abuse (n = 2), manic episode (n = 1),and panic disorder with agoraphobia (n = 1). Only oneveteran with PTSD met current diagnostic criteria forpanic disorder with agoraphobia. The two veteranswith PTSD who had formerly been diagnosed withalcohol abuse were abstinent for a period greater than6 months. Among the control subjects, the SCID didnot reveal any psychiatric disorders.

ROI analysisThere was no statistically significant differencebetween the volume of distribution of [11C]flumazenil

Table 1 Subject demographic and psychometric data

PTSD Trauma controls

N 9 7Age 35.3 (6.3) 36.4 (4.7)NS

Year of deployment(range: 1980–1999)

1991 (2.1) 1992 (2.8)NS

Country of deployment Bosnia (n = 7) Bosnia (n = 5)Lebanon (n = 2) Lebanon (n = 2)

CAPS total 76.0 (15.5) 3.6 (6.1)*Hamilton A 18.6 (5.3) 0.4 (0.5)*Hamilton D 16.2 (4.2) 0.3 (0.5)*

Abbreviations: CAPS, Clinician Administered PTSD Scale;Hamilton A, Hamilton Anxiety Scale; Hamilton D, HamiltonDepression Scale; PTSD, post-traumatic stress disorder.Means for both groups are given. Standard deviations aregiven in parentheses.NSThis difference was not significant. *This difference wassignificant at P < 0.0001.

Reduced binding to benzodiazepine receptorE Geuze et al

77

Molecular Psychiatry

in the pons of veterans with PTSD and controlveterans (Vd(patients) =0.9470.08 vs Vd(controls) =0.8970.13(mean7s.d.); Mann–Whitney U = 19.00, P = 0.535).For this reason, SRTM with pons as referencetissue was used to compare the two groups. Resultsfrom SRTM revealed significantly decreased BP in alarge number of the template based ROIs in veteranswith PTSD compared to veterans without PTSD.Veterans with PTSD showed decreased BP throughoutthe brain including bilateral orbital frontal cortex,superior frontal cortex, medial inferior frontal cortex,superior temporal cortex, medial inferior temporalcortex, parietal cortex, occipital cortex, cerebellum,thalamus, insula, right anterior cingulate cortex, rightposterior cingulate cortex, left enthorinal cortex, andleft striatum (Table 2). In addition, patients alsoshowed decreased BP in several manually delineatedROIs including the left hippocampus and leftamygdala.

SPM analysisSPM analysis of Ichise-derived BP images showedstatistically significant decreased [11C]flumazenilbinding in veterans with PTSD compared to controlsin the bilateral prefrontal cortex, left temporal lobe,left hippocampus, right parietal cortex, left cuneus,bilateral occipital cortex, left caudate nucleus,bilateral thalamus and hippocampus (Figure 1 andTable 3). SPM and ROI analyses showed nearlysimilar results. In both analyses, decreased [11C]flu-mazenil binding in veterans with PTSD was foundthroughout cortical and subcortical areas. However,SPM analysis did not show significant differences inthe left striatum and the left amygdala.

Discussion

To the best of our knowledge, this is the first fullyquantitative PET study demonstrating reduced bind-

Table 2 [11C]Flumazenil-binding potential in veterans with PTSD and control veterans: results from SRTM

PTSD Controls Mann–Whitney

Mean S.d. Mean S.d. U P

Template based ROIsLeft superior frontal cortex 4.65 0.40 5.45 0.55 7.00 0.008Right superior frontal cortex 4.57 0.42 5.35 0.55 8.00 0.012Left orbital frontal cortex 4.98 0.37 5.77 0.68 11.00 0.031Right orbital frontal cortex 4.85 0.42 5.67 0.72 8.00 0.012Left medial inferior frontal cortex 5.07 0.47 5.94 0.64 7.00 0.008Right medial inferior frontal cortex 5.13 0.41 5.92 0.64 8.00 0.012Right anterior cingulate cortex 4.88 0.71 5.76 0.80 12.00 0.042Left sensory motor cortex 4.10 0.44 4.76 0.49 9.00 0.016Left Insula 4.99 0.58 5.67 0.63 12.00 0.042Right insula 5.14 0.48 5.89 0.66 8.50 0.012Left superior temporal cortex 5.16 0.45 6.02 0.71 8.00 0.012Right superior temporal cortex 5.23 0.41 6.01 0.73 9.00 0.016Left enthorinal cortex 3.96 0.31 4.53 0.58 9.00 0.016Left parietal cortex 4.56 0.37 5.46 0.64 7.00 0.008Right parietal cortex 4.58 0.35 5.36 0.61 8.00 0.012Right posterior parietal cortex 4.92 0.69 5.81 0.82 10.00 0.023Left medial inferior temporal cortex 5.07 0.39 5.92 0.67 7.00 0.008Right medial inferior temporal cortex 5.03 0.38 5.88 0.65 8.00 0.012Left occipital cortex 5.48 0.45 6.29 0.82 9.00 0.016Right occipital cortex 5.23 0.41 6.13 0.80 9.00 0.016Right striatum 1.82 0.25 2.11 0.30 11.00 0.031Left thalamus 1.95 0.23 2.37 0.40 10.00 0.023Right thalamus 1.95 0.19 2.36 0.35 8.00 0.012Left cerebellum 3.71 0.31 4.29 0.58 11.00 0.031Right cerebellum 3.74 0.28 4.30 0.52 11.00 0.031Total brain 3.55 0.41 4.22 0.49 8.00 0.012

Manual ROIsL Hippocampus 2.97 0.36 3.59 0.44 7.00 0.008R Hippocampus 2.93 0.34 3.46 0.59 11.00 0.031L Amygdala 3.22 0.30 3.87 0.64 10.00 0.023

Abbreviations: PTSD, post-traumatic stress disorder; ROI, region of interest; SRTM, simplified reference tissue model.SRTM analysis revealed that veterans with PTSD showed reduced [11C]flumazenil-binding potential compared to controlveterans without PTSD. Only ROIs, which showed significantly reduced [11C]flumazenil binding in the SRTM analysis, arereported. Means and s.d. for veterans with PTSD and veterans without PTSD are reported.

Reduced binding to benzodiazepine receptorE Geuze et al

78

Molecular Psychiatry

ing to the benzodiazepine–GABAA receptor complexin veterans with PTSD. Widespread reduced BP wasfound throughout the cortex, hippocampus andthalamus of veterans with PTSD in comparison tocontrol veterans without PTSD. Both SRTM (ROIanalysis) and Ichise plots (SPM analysis) showedsimilar results. In both analyses, decreased [11C]flu-mazenil binding in veterans with PTSD was found inthe frontal, temporal, parietal and occipital cortex,and in caudate nuclei, thalamus and left hippocam-pus. In the SRTM analysis, veterans with PTSD alsohad less [11C]flumazenil binding in the left amygdalaand left striatum.

Benzodiazepine binding sites in the brain arelocated predominantly on the GABAA receptors, andtherefore their distribution in the brain reflects thedistribution of GABAA receptors. Benzodiazepinespotentiate the function of GABA through conforma-tional changes in the receptor, thereby increasing theeffectiveness of GABA for opening the ion channel.Animal models support the notion that stress andtrauma can alter GABAA–benzodiazepine bindingdensity.2,6,36–38 In general, these animal models haveshown decreased GABAA function and binding in thecerebral cortex and hippocampus.3–5 This is the firstbenzodiazepine–GABAA receptor binding study inpatients with PTSD that supports these preclinical data.

The present finding of reduced [11C]flumazenilbinding throughout most of the cortical areas maybe indicative of an a priori difference in subunitcomposition of GABAA–benzodiazepine receptors, alower expression of the GABAA receptor in PTSDpatients, or a disease or trauma-induced modulation

or downregulation of the GABAA receptor complex.These explanations are consistent with other clinicalstudies that have suggested altered GABAergic func-tion in PTSD.1,7,10 A reduced GABAergic function canalso be explained by the presence of increased levelsof GABA.

Since the control subjects in this study had alsoexperienced traumatic events, the observed differencein [11C]flumazenil binding in this study is disorderrelated and not due to trauma. However, this does notpreclude the possibility that stress or trauma, intrauma-exposed individuals who do not developPTSD, can alter GABA function. In the absence of ahealthy control group of subjects who have neverexperienced psychological trauma, it cannot bedetermined whether the observed difference in bind-ing between veterans with and without PTSD isunique to PTSD or represents a quantitative differ-ence in binding density.

Decreased binding to benzodiazepine-GABAA re-ceptors is consistent with an [123I]iomazenil SPECTstudy in Vietnam veterans with PTSD.12 In that study,Vietnam veterans with PTSD had a decreased volumeof distribution (Vd) of [123I]iomazenil in the prefrontalcortex (one of the two a priori defined regions)compared to age-matched healthy controls. Another[123I]iomazenil SPECT study in Gulf War veteranswith PTSD, however, was unable to find anysignificant differences in Vd of [123I]iomazenil com-pared to age-matched undeployed military person-nel.13 Possible factors that could account for thediscrepancy in findings of these studies are, asmentioned by Fujita et al.,13 related to type of controls

Figure 1 Statistical parametric mapping (SPM) analysis of Ichise binding potential (BP) images, showing decreased[11C]flumazenil binding in post-traumatic stress disorder (PTSD) subjects compared to controls throughout a priorihypothesized regions, such as the occipital cortex, temporal cortex, parietal cortex, prefrontal cortex, insular cortex,thalamus and hippocampus. The sections are at x, y, z = 0, 0, 0. Coordinates of the peak voxels are given in Table 3.

Reduced binding to benzodiazepine receptorE Geuze et al

79

Molecular Psychiatry

used and the interval between traumatic experiencesand the study. In the study by Bremner et al.12 healthysubjects were used as controls, and the interval wasabout 25 years. In the second study by Fujita et al.,13

controls included were military personnel who hadserved at the same time, but who were not deployed.In addition, the interval between deployment andtime of study was about 10 years. These differences,however, cannot explain differences in findings withthe present study, where control subjects wereveterans who were matched to the patient group,and who had been deployed to the same countries atthe same time as the veterans with PTSD. In addition,in the present study the interval between traumaticexperiences and the PET study was also about 10–12years.

SPECT is a semi-quantitative method that requiresglobal normalization before SPM analysis. This mayalso explain why the SPECT studies performed inPTSD12,13 did not report a global reduction in Vd of[123I]iomazenil in PTSD. Any potential global differ-ences would have been removed by the globalnormalization. Interestingly, in panic disorder, whichis related to PTSD a global reduction of [11C]flumaze-nil binding was also found throughout the brain,39

suggesting that the GABAergic system is indeedinvolved in these anxiety disorders.

Several studies have shown that the prefrontalcortex is dysfunctional in PTSD.40–44 Across thesestudies, however, dysfunctional alterations occurredin different parts of the prefrontal cortex, indicatingthat failure of prefrontal inhibition in PTSD needs tobe clarified in the future. In the present study, areduction of GABAA–benzodiazepine BP throughoutthe frontal cortex was found. This may underliethe working memory deficit in patients with PTSD.45

Decreases in GABAA receptor binding are alsoassociated with alterations in working memoryperformance.37,38 In addition, appropriate GABAneurotransmission in the frontal cortex is requiredfor a normal working memory function.46

Another important brain area is the hippocampus,which plays a pivotal role in managing the stressresponse, novelty detection and memory processing.The hippocampus is also an important site forGABA and serotonin interaction, which is thoughtto modulate emotional behaviour.47 In patients withPTSD structural and functional alterations of thehippocampus have consistently been demon-strated.48–51 Benzodiazepine agonists enhance GABAA

Table 3 [11C]Flumazenil-binding potential in veterans with PTSD and control veterans: results from statistical parametricmapping

Region BA Z-score Talairach coordinates

X Y Z

Left superior frontal gyrus 8 3.91 �26 32 52Left superior frontal gyrus 9 3.95 0 56 34Right superior frontal gyrus 8 4.44 26 28 52Right precentral gyrus 4 3.27 46 �10 58Left middle frontal gyrus 10 3.62 �42 44 12Left middle frontal gyrus 9 3.89 �46 28 34Left middle frontal gyrus 8 3.73 �48 14 42Right middle frontal gyrus 6 4.69 38 0 52Left medial frontal gyrus 10 4.07 �4 60 10Right medial frontal gyrus 10 4.02 4 62 12Right middle temporal Gyrus 22 4.59 62 �30 0Right uncus 28 3.68 28 4 �30Right fusiform gyrus 36 4.45 42 �36 �20Left middle temporal gyrus 21 4.49 �56 �10 �8Left inferior temporal gyrus 20 4.67 �42 �24 �28Left hippocampus 3.70 �30 �28 �2Right superior parietal lobe 7 4.82 32 �52 48Left cuneus 18 4.85 �8 �84 18Left middle occipital gyrus 19 4.52 �32 �94 18Left inferior occipital gyrus 18 3.69 �32 �94 �8Right middle occipital gyrus 19 4.80 50 �76 �8Left caudate 3.59 �10 �4 16Left thalamus 3.66 �24 �34 8Right thalamus 4.70 8 �10 12

Abbreviations: BA, Brodmann area; PTSD, post-traumatic stress disorder.Statistical parametric mapping analysis revealed that veterans with PTSD showed reduced [11C]flumazenil-binding potentialcompared to control veterans without PTSD. Talairach coordinates, corrected P-values (false discovery rate-correctedP = 0.003) and the Z-score of peak voxels are reported. BAs corresponding to the Talairach coordinates of the peak voxel arealso reported.

Reduced binding to benzodiazepine receptorE Geuze et al

80

Molecular Psychiatry

receptor function in the CA1 region of the hippocam-pus and can thus disrupt memory formation andhippocampal synaptic plasticity.52 It is also possiblethat in patients with PTSD increased sensitivity tostress causes the alterations in GABAA receptors. In aprevious SPECT study, patients with panic disorderalso revealed decreased GABAA–benzodiazepinereceptor binding in the left hippocampus.53

Other brain areas, which play an important role inanxiety disorders, such as the left amygdala and theleft caudate nucleus, also showed a reduction inbenzodiazepine receptor binding. Several neuroima-ging studies in PTSD have revealed that the amygdalaplays an important role in the modulation ofemotional responses in PTSD.54–56 In addition, theamygdala also seems to play a role in nociceptiveresponses in veterans with PTSD.57 The caudatenucleus has been implicated in a number of anxietydisorders, including obsessive compulsive disorderand social phobia. In relation to PTSD, the caudatenucleus has attracted little attention from neurobio-logical research. It seems, however, that caudatefunction should also be examined in future PTSDstudies. Certainly, the relation of the dopaminergicsystem (hence also the caudate nucleus) to PTSDshould be examined in more detail, considering therecent use of atypical antipsychotics as pharma-cotherapy in PTSD.58,59

All research methods employed have both metho-dological advantages and limitations. In the presentstudy, SRTM with pons as reference tissue was usedfor analyzing the data. Apart from enabling inclusionof data from all subjects (due to technical reasons thearterial plasma curve could not be used in twosubjects), an important advantage of this method isthat it measures binding potential for specific bindingonly. In contrast, the generally accepted single tissuecompartment model with plasma input (requiringarterial cannulation) measures the volume of distri-bution of total binding (that is, both specific and non-specific binding). Reference tissue models such asSRTM require more assumptions than plasma inputmodels, but they exhibit less variance in determiningspecific binding making them more reproducible.60

Recently, the use of SRTM with pons as referencetissue in the quantification of [11C]flumazenil PET hasbeen validated in a large sample of normal controlsand patients with depression.32 This validation sup-ports the plausibility of the present results, as datawere analysed in a similar manner. Although headmotion was visually checked during scanning and avisual check for motion artefacts before analyses ofthe scans was performed, no additional motioncorrection was applied before SPM and ROI analysis.Watabe et al.61 have proposed an elegant method formotion correction during a dynamic PET scan.However, this implies acquisition of the data in list-mode, which is not commercially available onHRþ scanners. As such, data were acquired in thestandard frame-mode, making motion correction withthis method impossible.

Other potential confounders of the present studyare related to sample size and study population. Likein most PET studies the number of subjects includedwas relatively small. As a result, ROI data werenot normally distributed, requiring nonparametricmethods of analysis. Therefore, care should be takenin generalizing the present results. One of the mainstrengths of this study is the use of matched traumacontrols. Therefore, the results are likely to be relatedto the effect of PTSD itself, and not the effect ofhaving witnessed traumatic events or the stress ofdeployment. With two matched groups, one withPTSD and one without, the question may arise whysome subjects develop PTSD, while others do not.Therefore, it would be interesting for future researchto investigate the effects of personality differencesand genetic makeup on neurobiological differences inPTSD. As this study only included male veterans withPTSD, care should be taken in extrapolating thesefindings to females with PTSD, and to PTSD related toother types of trauma. Therefore, further studies arewarranted. Future research should also considerwhether the observed effects are basal or adaptivechanges, as the present study cannot distinguishbetween predisposing, mediating, or resultant factors.

In conclusion, this study has shown decreased[11C]flumazenil binding throughout the brain ofveterans with PTSD compared to veterans withoutPTSD. This provides support for the involvement ofGABAA benzodiazepine receptors in PTSD andwarrants further investigation of GABAergic function-ing in PTSD.

Acknowledgments

This work was financially supported by the DutchMinistry of Defence. We thank the staff of theRadiology Department of Utrecht University MedicalCentre for acquisition of MRI scans, the staff of theDepartment of Nuclear Medicine & PET Research ofthe VU University Medical Center for acquisitionof PET scans, Arthur Rademaker, MSc, for clinicalassessments, Reina Kloet, MSc, for help with dataanalysis, and the production chemists and the BVCyclotron VU for providing 11CO2 and [11C]flumazenilfor this study.

References

1 Vaiva G, Thomas P, Ducrocq F, Fontaine M, Boss V, Devos P et al.Low posttrauma GABA plasma levels as a predictive factor in thedevelopment of acute posttraumatic stress disorder. Biol Psychia-try 2004; 55: 250–254.

2 Harvey BH, Oosthuizen F, Brand L, Wegener G, Stein DJ. Stress-restress evokes sustained iNOS activity and altered GABA levelsand NMDA receptors in rat hippocampus. Psychopharmacology(Berl) 2004; 175: 494–502.

3 Lippa AS, Klepner CA, Yunger L, Sano MC, Smith WV, Beer B.Relationship between benzodiazepine receptors and experimentalanxiety in rats. Pharmacol Biochem Behav 1978; 9: 853–856.

4 Medina JH, Novas ML, Wolfman CN, Levi de Stein M, De RobertisE. Benzodiazepine receptors in rat cerebral cortex and hippocam-

Reduced binding to benzodiazepine receptorE Geuze et al

81

Molecular Psychiatry

pus undergo rapid and reversible changes after acute stress.Neuroscience 1983; 9: 331–335.

5 Drugan RC, Basile AS, Crawley JN, Paul SM, Skolnick P.Inescapable shock reduces. Pharmacol Biochem Behav 1986; 24:1673–1677.

6 Adamec R, Strasser K, Blundell J, Burton P, McKay DW. Proteinsynthesis and the mechanisms of lasting change in anxietyinduced by severe stress. Behav Brain Res 2006; 167: 270–286.

7 Davidson JR. Use of benzodiazepines in social anxiety disorder,generalized anxiety disorder, and posttraumatic stress disorder.J Clin Psychiatry 2004; 65(Suppl 5): 29–33.

8 Gelpin E, Bonne O, Peri T, Brandes D, Shalev AY. Treatment ofrecent trauma survivors with benzodiazepines: a prospectivestudy. J Clin Psychiatry 1996; 57: 390–394.

9 Viola J, Ditzler T, Batzer W, Harazin J, Adams D, Lettich L et al.Pharmacological management of post-traumatic stress disorder:clinical summary of a five-year retrospective study, 1990–1995.Mil Med 1997; 162: 616–619.

10 Connor KM, Davidson JR, Weisler RH, Zhang W, Abraham K.Tiagabine for posttraumatic stress disorder: effects of open-labeland double-blind discontinuation treatment. Psychopharmacology(Berl) 2006; 184: 21–25.

11 Taylor FB. Tiagabine for posttraumatic stress disorder: a case seriesof 7 women. J Clin Psychiatry 2003; 64: 1421–1425.

12 Bremner JD, Innis RB, Southwick SM, Staib L, Zoghbi S, CharneyDS. Decreased benzodiazepine receptor binding in prefrontalcortex in combat-related posttraumatic stress disorder. Am JPsychiatry 2000; 157: 1120–1126.

13 Fujita M, Southwick SM, Denucci CC, Zoghbi SS, Dillon MS,Baldwin RM et al. Central type benzodiazepine receptors in GulfWar veterans with posttraumatic stress disorder. Biol Psychiatry2004; 56: 95–100.

14 Fedi M, Berkovic SF, Marini C, Mulligan R, Tochon-Danguy H,Reutens DC. A GABAA receptor mutation causing generalizedepilepsy reduces benzodiazepine receptor binding. Neuroimage2006; 32: 995–1000.

15 Katsifis A, Kassiou M. Development of radioligands for in vivoimaging of GABA(A)-benzodiazepine receptors. Mini Rev MedChem 2004; 4: 909–921.

16 Juhasz C, Chugani HT. Imaging the epileptic brain with positronemission tomography. Neuroimaging Clin N Am 2003; 13:705–716, viii.

17 Erritzoe D, Talbot P, Frankle WG, Abi-Dargham A. Positronemission tomography and single photon emission CT molecularimaging in schizophrenia. Neuroimaging Clin N Am 2003; 13:817–832.

18 Talbot PS, Laruelle M. The role of in vivo molecular imaging withPET and SPECT in the elucidation of psychiatric drug action andnew drug development. Eur Neuropsychopharmacol 2002; 12:503–511.

19 First MB, Spitzer RL, Gibbon M, Williams JBW. Structured ClinicalInterview for DSM-IV Axis I Disorders. SCID-I/P, 1997.

20 Blake D, Weathers F, Nagy L, Kaloupek D, Klauminzer G, CharneyD. A Clinican Rating Scale for Assessing Current and LifetimePTSD: The CAPS-1. Behav Ther 1990; 13: 187–188.

21 Brix G, Zaers J, Adam LE, Bellemann ME, Ostertag H, Trojan Het al. Performance evaluation of a whole-body PET scanner usingthe NEMA protocol. National Electrical Manufacturers Associa-tion. J Nucl Med 1997; 38: 1614–1623.

22 Adam LE, Jears J, Ostertag H, Trojan H, Bellemann ME, Brix G.Performance of the whole-body PET scanner ECAT EXACT HRþfollowing the IEC standard. IEEE Trans Nucl Sci 1997; 44:1172–1179.

23 Boellaard R, van Lingen A, van Balen SC, Hoving BG, LammertsmaAA. Characteristics of a new fully programmable blood samplingdevice for monitoring blood radioactivity during PET. Eur J NuclMed 2001; 28: 81–89.

24 Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P.Multimodality image registration by maximization of mutualinformation. IEEE Trans Med Imaging 1997; 16: 187–198.

25 West J, Fitzpatrick JM, Wang MY, Dawant BM, Maurer Jr CR,Kessler RM et al. Comparison and evaluation of retrospectiveintermodality brain image registration techniques. J Comput AssistTomogr 1997; 21: 554–566.

26 Svarer C, Madsen K, Hasselbalch SG, Pinborg LH, Haugbol S,Frokjaer VG et al. MR-based automatic delineation of volumes ofinterest in human brain PET images using probability maps.Neuroimage 2005; 24: 969–979.

27 Watson C, Andermann F, Gloor P, Jones-Gotman M, Peters T, EvansA et al. Anatomic basis of amygdaloid and hippocampal volumemeasurement by magnetic resonance imaging. Neurology 1992; 42:1743–1750.

28 Duvernoy HM, Bourgouin P. The Human Brain: Surface, Three-Dimensional Sectional Anatomy and MRI, and Blood Supply.Springer Verlag: Vienna, 1999.

29 Greuter HN, van Ophemert PL, Luurtsema G, Franssen EJ,Boellaard R, Lammertsma AA. Validation of a multiwell gamma-counter for measuring high-pressure liquid chromatographymetabolite profiles. J Nucl Med Technol 2004; 32: 28–32.

30 Koeppe RA, Holthoff VA, Frey KA, Kilbourn MR, Kuhl DE.Compartmental analysis of flumazenil kinetics for the estimationof ligand transport rate and receptor distribution using positronemission tomography. J Cereb Blood Flow Metab 1991; 11:735–744.

31 Lammertsma AA, Hume SP. Simplified reference tissue model forPET receptor studies. Neuroimage 1996; 4: 153–158.

32 Klumpers UMH, Veltman DJ, Comans EFI, Kropholler MA,Boellaard R, Hoogendijk WGJ et al. Comparison of various modelsfor analysing [11C] flumazenil studies. J Cereb Blood Flow Metab2005; 25: S660.

33 Ichise M, Fujita M, Seibyl JP, Verhoeff NP, Baldwin RM, Zoghbi SSet al. Graphical analysis and simplified quantification of striataland extrastriatal dopamine D2 receptor binding with epideprideSPECT. J Nucl Med 1999; 40: 1902–1912.

34 Friston KJ, Frith CD, Liddle PF, Frackowiak RS. Comparingfunctional (PET) images: the assessment of significant change.J Cereb Blood Flow Metab 1991; 11: 690–699.

35 Genovese CR, Lazar NA, Nichols T. Thresholding of statisticalmaps in functional neuroimaging using the false discovery rate.Neuroimage 2002; 15: 870–878.

36 Caldji C, Francis D, Sharma S, Plotsky PM, Meaney MJ. The effectsof early rearing environment on the development of GABAA andcentral benzodiazepine receptor levels and novelty-inducedfearfulness in the rat. Neuropsychopharmacology 2000; 22:219–229.

37 Weizman R, Weizman A, Kook KA, Vocci F, Deutsch SI, Paul SM.Repeated swim stress alters brain benzodiazepine receptorsmeasured in vivo. J Pharmacol Exp Ther 1989; 249: 701–707.

38 Drugan RC, Morrow AL, Weizman R, Weizman A, Deutsch SI,Crawley JN et al. Stress-induced behavioral depression in the rat isassociated with a decrease in GABA receptor-mediated chlorideion flux and brain benzodiazepine receptor occupancy. Brain Res1989; 487: 45–51.

39 Malizia AL, Cunningham VJ, Bell CJ, Liddle PF, Jones T, Nutt DJ.Decreased brain GABA(A)-benzodiazepine receptor binding inpanic disorder: preliminary results from a quantitative PET study.Arch Gen Psychiatry 1998; 55: 715–720.

40 Bremner JD, Vermetten E, Vythilingam M, Afzal N, Schmahl C,Elzinga B et al. Neural correlates of the classic color and emotionalstroop in women with abuse-related posttraumatic stress disorder.Biol Psychiatry 2004; 55: 612–620.

41 Bremner JD, Staib LH, Kaloupek D, Southwick SM, Soufer R,Charney DS. Neural correlates of exposure to traumatic picturesand sound in Vietnam combat veterans with and withoutposttraumatic stress disorder: a positron emission tomographystudy. Biol Psychiatry 1999; 45: 806–816.

42 Bremner JD, Innis RB, Ng CK, Staib LH, Salomon RM, Bronen RAet al. Positron emission tomography measurement of cerebralmetabolic correlates of yohimbine administration in combat-related posttraumatic stress disorder. Arch Gen Psychiatry 1997;54: 246–254.

43 Britton JC, Phan KL, Taylor SF, Fig LM, Liberzon I. Corticolimbicblood flow in posttraumatic stress disorder during script-drivenimagery. Biol Psychiatry 2005; 57: 832–840.

44 Shin LM, Wright CI, Cannistraro PA, Wedig MM, McMullin K,Martis B et al. functional magnetic resonance imaging study ofamygdala and medial prefrontal cortex responses to overtly

Reduced binding to benzodiazepine receptorE Geuze et al

82

Molecular Psychiatry

presented fearful faces in posttraumatic stress disorder. Arch GenPsychiatry 2005; 62: 273–281.

45 Weber DL, Clark CR, McFarlane AC, Moores KA, Morris P, EganGF. Abnormal frontal and parietal activity during working memoryupdating in post-traumatic stress disorder. Psychiatry Res 2005;140: 27–44.

46 Lewis DA, Volk DW, Hashimoto T. Selective alterations inprefrontal cortical GABA neurotransmission in schizophrenia: anovel target for the treatment of working memory dysfunction.Psychopharmacology (Berl) 2004; 174: 143–150.

47 Nazar M, Siemiatkowski M, Czlonkowska A, Sienkiewicz-JaroszH, Plaznik A. The role of the hippocampus and 5-HT/GABAinteraction in the central effects of benzodiazepine receptorligands. J Neural Transm 1999; 106: 369–381.

48 Geuze E, Vermetten E, Bremner JD. MR-based in vivo hippocampalvolumetrics: 2. Findings in neuropsychiatric disorders. MolPsychiatry 2005; 10: 160–184.

49 Nemeroff CB, Bremner JD, Foa EB, Mayberg HS, North CS, SteinMB. Posttraumatic stress disorder: a state-of-the-science review.J Psychiatr Res 2006; 40: 1–21.

50 Bremner JD, Vythilingam M, Vermetten E, Southwick SM,McGlashan T, Staib LH et al. Neural correlates of declarativememory for emotionally valenced words in women with posttrau-matic stress disorder related to early childhood sexual abuse.Biol Psychiatry 2003; 53: 879–889.

51 Sakamoto H, Fukuda R, Okuaki T, Rogers M, Kasai K, Machida Tet al. Parahippocampal activation evoked by masked traumaticimages in posttraumatic stress disorder: a functional MRI study.Neuroimage 2005; 26: 813–821.

52 Seabrook GR, Easter A, Dawson GR, Bowery BJ. Modulation oflong-term potentiation in CA1 region of mouse hippocampal brain

slices by GABAA receptor benzodiazepine site ligands. Neuro-pharmacology 1997; 36: 823–830.

53 Bremner JD, Innis RB, White T, Fujita M, Silbersweig D, GoddardAW et al. SPECT. Biol Psychiatry 2000; 47: 96–106.

54 Debiec J, LeDoux JE. Oradrenergic signaling in the amygdalacontributes to the reconsolidation of fear memory:treatment implications for PTSD. Ann N Y Acad Sci 2006; 1071:521–524.

55 Shin LM, Rauch SL, Pitman RK. Mygdala, medial prefrontalcortex, and hippocampal function in PTSD. Ann N Y Acad Sci2006; 1071: 67–79.

56 Liberzon I, Martis B. Euroimaging studies of emotional responsesin PTSD. Ann N Y Acad Sci 2006; 1071: 87–109.

57 Geuze E, Westenberg HGM, Jochims A, de Kloet CS, Vermetten E,Schmahl C. Altered pain processing in veterans withposttraumatic stress disorder. Arch Gen Psychiatry 2007; 64:76–85.

58 David D, De Faria L, Lapeyra O, Mellman TA. Adjunctiverisperidone treatment in combat veterans with chronic PTSD.J Clin Psychopharmacol 2004; 24: 556–559.

59 Robert S, Hamner MB, Kose S, Ulmer HG, Deitsch SE, LorberbaumJP. Quetiapine improves sleep disturbances in combat veteranswith PTSD: sleep data from a prospective, open-label study. J ClinPsychopharmacol 2005; 25: 387–388.

60 Ichise M, Meyer JH, Yonekura Y. An introduction to PET andSPECT neuroreceptor quantification models. J Nucl Med 2001; 42:755–763.

61 Watabe H, Matsumoto K, Senda M, Iida H. Performance of listmode data acquisition with ECAT EXACT HR and ECAT EXACTHRþ positron emission scanners. Ann Nucl Med 2006; 20:189–194.

Reduced binding to benzodiazepine receptorE Geuze et al

83

Molecular Psychiatry