Psychiatry Research: Neuroimaging 173 (2009) 88–93

Contents lists available at ScienceDirect

Psychiatry Research: Neuroimaging

j ourna l homepage: www.e lsev ie r.com/ locate /psychresns

Association between cerebral metabolic and structural abnormalities and cognitiveperformance in schizophrenia

Vicente Molinaa,⁎, Sara Solerab, Javier Sanzb, Fernando Sarrameac, Rogelio Luquec, Roberto Rodríguezb,Miguel Angel Jiménez-Arrierob, Tomás Palomob

aDept of Psychiatry, Hospital Universitario, Salamanca, SpainbDept of Psychiatry, Hospital Doce de Octubre, Madrid, SpaincDept of Psychiatry, Hospital Reina Sofía, Córdoba, Spain

⁎ Corresponding author. Dept. of Psychiatry, HospitalSan Vicente, 58-182, E-37007 Salamanca, Spain. Tel.923291383.

E-mail address: vmolina@usal.es (V. Molina).

0925-4927/$ – see front matter © 2008 Elsevier Irelandoi:10.1016/j.pscychresns.2008.09.009

a b s t r a c t

a r t i c l e i n f o

Article history:

The possible association in Received 4 December 2007Received in revised form 19 July 2008Accepted 23 September 2008

Keywords:SchizophreniaHypofrontalityProcessing speedWorking memoryNeuropsychologyPrefrontalAtrophy

schizophrenia between frontal abnormalities, such as hypofrontality and frontalgrey matter (GM) deficits, and neuropsychological deficits is not yet well defined. Our objective was to studysuch an association and to clarify the cognitive relevance of metabolic and anatomical variability acrossschizophrenia patients. To do so, we studied dorsolateral prefrontal (DLPF) metabolism during an attentiontest using fluoro-deoxy-glucose positron emission tomography and DLPF structure with magnetic resonanceimaging (MRI) in 22 schizophrenia patients [9 neuroleptic-naïve (NN) first episodes]. These patients alsounderwent a comprehensive battery of neuropsychological tests aimed at evaluating global intelligence andthe proposed domains of cognitive alteration in schizophrenia, i.e., attention, visual and verbal learning andmemory, working memory, problem solving and processing speed. The metabolic activity in the right DLPFregion was significantly and directly related to processing speed, and a measure of structural deficit in thesame area was directly related to working memory scores. In the NN group studied alone, these associationswere replicated. We may conclude that hypofrontality during cognitive activation, and the degree of DLPFstructural deficit may be associated to a particular profile of cognitive deficit, including lower processingspeed and working memory capacity.

© 2008 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

In schizophrenia, frontal structural (Shenton et al., 2001) andfunctional (Buchsbaum, 1995) deficits are replicated findings. How-ever, their relevance to the known cognitive problems in this disorderhas received less attention to date, although existing results suggestthat an association may indeed exist between the cognitive andcerebral abnormalities. Concerning functional abnormalities, it hasbeen reported that changes over time in prefrontal perfusion areassociated with an improvement in working-memory scores inschizophrenia patients (Wolf et al., 2007). From the structuralperspective, a variation in the DISC-1 gene was related in a twinstudy to both prefrontal volume deficit and short- and long-termmemory deficits in schizophrenia (Cannon et al., 2005).

An association between anatomical or functional frontal devia-tions, on the one hand, and neuropsychological deficit, on the otherhand, also seems likely in view of the role that prefrontal regions play

Clínico Universitario, Paseo de: +34 923291102; fax: +34

d Ltd. All rights reserved.

in executive and attention processes (Fuster, 1999). The frontal lobesare involved in many of the functions assessed by neuropsychologicaltests in which schizophrenia patients show a low performance(Cuesta and Peralta, 1995; Cuesta et al., 1998). In particular, asreviewed by Alvarez and Emory (2006), frontal regions are activatedin tests such as the Wisconsin Card Sorting Test (WCST, a classical testfor executive functions), the FAS (a test for verbal fluency) and theToulouse-Piéron tests (a test for the assessment of attention). Thus, arelationship would be predicted between impaired performance onthose and other similar tests and measures of frontal activation orstructure. However, it is clear that alterations in non-frontal regionsmay also contribute to the cognitive deficits in schizophrenia.

In this context, the evaluation of the relationship betweenperformance on neuropsychological tests and frontal anatomical andmetabolic measurements may be relevant to an understanding of thecerebral basis of at least some cognitive deficits in schizophrenia.Similarly, it could also help to identify the correlates of deviations inthe structure and function of frontal regions in schizophrenia. Withthis objective, we used structural magnetic resonance and positronemission tomography to assess anatomical and metabolic frontalparameters, together with a battery of neuropsychological testsdesigned to evaluate consensus-based cognitive dimensions in theillness (Nuechterlein et al., 2004).

89V. Molina et al. / Psychiatry Research: Neuroimaging 173 (2009) 88–93

Our hypothesis was that cognitive deficit in schizophrenia wouldbe associated with structural and/or metabolic deviations in theprefrontal region.

2. Methods

2.1. Patients

Our sample included 22 paranoid schizophrenia patients (12 males).The patients' clinical scores were assessed both upon admission and atthe time of the neuropsychological examinations with the Positive andNegative Syndrome Scale (PANSS) (Kay et al., 1987), Spanish version(Cuesta and Peralta, 1995). Among the patients, nine were neuroleptic-naïve (NN)prior to their inclusionand theother13were chronicpatients.

Before their inclusion in the study, the NN patients had had a firstpsychotic episode with symptoms lasting more than 1 month. Thediagnosis of schizophreniawas confirmed in these cases after a follow-upperiod of more than 2 years using a semi-structured interview (SCID).

The neuropsychological exams were given after clinical stabiliza-tion, defined as a greater than 50% decrease in the PANSS score(Rummel et al., 2003), when the patients had been recently dis-charged from a psychiatric unit where they had been admitted owingto a psychotic exacerbation of their illness. Manual dominance wasassessed with the Harris laterality subtest.

At the time of the neuropsychological exam, all patients werereceiving treatment with an atypical antipsychotic (3 with clozapine,14 with risperidone and 5 with olanzapine). The mean dose inchlorpromazine equivalents (Woods, 2003) was 335.63 (S.D. 98.44)mg/d. All the NN patients received risperidone following theirinclusion in the study.

The patients studied here partly overlapped with the sample usedin other studies conducted by our group with first-episode (Molinaet al., 2005b, 2006) and chronic (Molina et al., 2005a) patients. Thedata from the healthy controls in these studies with both MRI and PETdata available (n=41, 22 males, age 28.43, S.D. 8.55) were used toexplore the deviation in structural and metabolic values. Neuropsy-chological studies were not available for the healthy controls, andhence their imaging data were only used to corroborate that thepatients showed the expected patterns of deviation according to theliterature about schizophrenia.

No differences in age or parental socioeconomic level wereobserved between groups (Hollingshead and Frederick, 1953). Theeducational level (of both the patients and their parents) was scoredaccording to the subject's achievement: 1 no education, 2 primaryschool, 3 high school, 4 college. Educational level of controls accordingto this scale was 2.23 (S.D. 0.71).

The exclusion criteria were mental retardation and neurologicaldisease, MRI findings deemed clinically relevant (from a neurologicalperspective) by a radiologist blind to the diagnosis, a history of headtrauma with loss of consciousness, and a history of substance abuseduring the previous 6 months (other than occasional hypnotics,caffeine and nicotine). In all cases a urine analysis was done on the dayof admission to confirm that the consumption of toxic substances hadnot been a major factor in the psychotic episode.

2.2. Neuropsychological assessment

Neuropsychological assessmentwasperformedafter theacute symp-toms had remitted so that the bias related to the acute state could beminimised (22.3±S.D. 8.3 days later than PET studies). The psychologistwho administered the tests (SS) was blind to the psychopathologicalassessments of the patients and the neuroimaging results.

The neuropsychological tests included: the Trail-Making Test A andB(TMT A and B), the WAIS Digit-symbol and arithmetic subtests, theTolouse-Pièron task, the Verbal Learning “España-Complutense”(TAVEC) test, the Rey–Osterrieth Complex Figure test and theWisconsin

Card Sorting Test (WCST). The TAVEC is a validated Spanish version ofthe California Verbal Learning Test (CVLT).

This battery was designed to assess the separable cognitive domainsof processing speed, attention/vigilance, working memory, verballearning and memory, visual learning and memory, and problemsolving. These domains have been identified according to expertconsensus (Nuechterlein et al., 2004) andmay have different substrates.

Moreover, the complete Wechsler Adult Intelligence Scale (WAIS)(Wechsler, 1999) was used to evaluate the global intelligence quotient(IQ) and verbal and manipulative IQ.

2.2.1. Definition and assessment of cognitive domainsThe domains of interest were as follows, in agreement with those

identified by Nuechterlein et al. (2004).Processing speed: TMT-A time and digit symbol of WAIS score.Attention: final Tolouse Pièron score.Working memory: WAIS arithmetic subtest score.Verbal learning and memory: TAVEC short-term memory score.Visual learning and memory: accuracy and richness of reproduction

in Rey's Figure.Reasoning and problem solving: Categories and perseverative errors

in WCST.

2.2.2. Transformation to z scores and cognitive domainsIn order to obtain a measure of the cognitive performance of the

patients and to compare it with normative data for the general Spanishpopulation, we transformed the direct scores on all these tests into zscores (defined as the relative position with respect to the mean of thenormative population) and centiles (percentage of the general popula-tionwith lower scores), basedon the scores in thenormative population,matched for age, sex and education. Then, the scores for each cognitivedomain were defined as the mean of the z scores of their components.

2.3. Imaging methods

In order to determine the structure and metabolic activity in theDLPF cortex, we used a procedure based on MRI/PET image fusion.

2.3.1. Acquisition

2.3.1.1. MRI protocol. Magnetic resonance imaging data on a PhilipsGyroscan 1.5T scanner using a T1-weighted 3D gradient echo sequencewith the following parameters: matrix size 256×256; pixel size0.9×0.9 mm (FOV≈256 mm); flip angle 30°; echo time 4.6 ms; slicethickness 1.1 mm. MRI scans were obtained in the first week followingadmission.

2.3.1.2. PET protocol. PET data were obtained with a SIEMENS Exact 47tomograph, 20 min after injecting 370 MBq of 18FDG, while subjectsperformed a contingent Continuous Performance Test. Subjects wereinstructed to push a button if T immediately followed the letter L, aspresented on a computer screen. After the placement of an intra-venous line for FDG administration, the subjects began the task, andFDG was administered 1 min after this.

Tracer activity values were proportionally normalized to the globalactivity of each PET scan, thus representing relative activity. The totalmetabolic activity for each region of interest (ROI) was divided by theROI volume, thus providing a measurement independent of theamount of tissue sampled.

PET studies were performed during the first 3 days following theadmission of the patients to a psychiatric unit due to an acuteexacerbation or a first psychotic episode. Prior to the PET study, patientsonly received haloperidol, except for the 12 h preceding it.

2.3.2. SegmentationTo obtain anatomical and metabolic measurements of the pre-

frontal cortex, we used amethod for semi-automated segmentation of

Table 1Neuropsychological (direct scores), neuroimaging, clinical and demographic values inthe whole sample and in the neuroleptic-naïve subset.

N=22 Neuroleptic-naïve subsetN=9

Age 27.1 (6.7) 25.63 (4.8)PANSS-Positive (admission/NP exams)

24.9 (7.7)/10.9 (3.6) 26.6 (7.0)/10.8 (3.4)

PANSS Negative 24.0 (9.3)/20.4 (8.3) 26.11 (8.7)/19.8 (6.6)PANSS General 53.2 (19.3)/30.5 (9.7) 55.0 (19.4)/29.3 (7.5)Education level 2.77 (0.68) 2.88 (0.64)Parental education 2.12 (0.69) 2.51 (0.86)Hand dominance 1.14 (0.64) 1.38 (0.74)DLPF volume residual (l/r) −1.51 (4.37)/−3.10 (30.06)⁎ −2.83 (2.88)/−3.98 (2.64)DLPF relative activity (l/r) 98.74 (60.3)⁎/101.49 (5.83) 100.61 (6.54)/104.28 (7.54)TMT A 58.41 (33.12) 61.0 (37.35)WAIS digit-symbol 38.86 (12.78) 38.56 (13.18)WAIS arithmetics 9.09 (3.06) 9.11 (4.04)Tolouse-Pièron 108.36 (63.83) 100.44 (61.97)TAVEC short-term memory 9.45 (3.97) 8.00 (4.72)Rey accuracy and richness 16.27 (8.44) 15.89 (8.22)WCST categories 5.55 (1.26) 5.44 (1.67)WCST, perseverative errors 14.59 (12.71) 12.33 (7.28)Verbal IQ 106.00 (14.40) 104.44 (17.15)Manipulative IQ 100.27 (13.19) 100.44 (13.88)Total IQ 103.41 (14.79) 102.22 (16.61)

Structural values are shown as residuals (deviation from the expected values in normalsubjects) and metabolic values as relative activity in comparison with total brain (seetext).⁎ Pb0.05 in comparison with healthy controls (see text).

90 V. Molina et al. / Psychiatry Research: Neuroimaging 173 (2009) 88–93

the brain based on the Talairach reference system. In brief, we used atwo-step procedure. The first step involved editing the MRI to removeskull and extracranial tissue, the registration of PET and MRI values,and an initial segmentation of cerebral tissues into grey matter (GM),white matter (WM), and cerebrospinal fluid (CSF).

The edited MRI was co-registered to the PET (Woods et al., 1993).An initial segmentationwas performed using an automated method, astandard processing tool in the SPM (Statistical Parametric Mapping)program (Ashburner and Friston, 1997). This method classifies all MRIpixels into four GM, WM, CSF, and “other tissues”. The algorithm alsoremoves the effect of inhomogeneities in the radiofrequency field(Ashburner and Friston, 1997). This segmentation was checked andcorrected manually whenever necessary.

In the second stage, ROIs were obtained by superimposing the 3Dtissue masks corresponding to WM, GM, and CSF onto each subject'sTalairach reference grid, where the ROIs were defined as sets of cells. Onthis MRI with the Talairach grid, the volumes for each tissue type weremeasured by totalling the data from the grid cells associated with eachROI. On the PET/MRI fused images, activity was measured by totallingthe data from the grid cells associated with each ROI.

All manual procedures were performed by a single operator.

Table 2Neuropsychological scores shown as z scores and centiles (position with respect to normal

N=22

Centiles z scores Interpretati

TMT-A 27.73 −0.60 Slight deficWAIS digit-symbol 37 −0.33 Slight deficWAIS Arithmetic 37 −0.33 Slight deficToulouse-Pieron 16,82 −1 Moderate dTMT-B 28,86 −0,58 Slight deficTAVEC short term memory 9 −1.36 Moderate dREY accuracy and richness 28.86 −0.58 Slight deficWCST categories 92.5 1.5 Normal-higWCST perseverative errors 39.95 −0.27 Slight deficVerbal IQ 65 0.39 NormalManipulative IQ 50 0 NormalTotal IQ 59 0.21 Normal

Normative data were extracted from the corresponding test manuals (in the Spanish versio

The effect of age and cranial size upon regional volumes wasremoved by using the residuals from the regression models obtainedfrom a group of 57 healthy individuals, following the procedure ofPfefferbaum et al. (1992). After this correction, the volume variableswere expressed as deviations from the expected volumes of healthyindividuals of the same age and brain size as the patients. Thus, negativeor positive residuals respectively represented a quantitative measure-ment of atrophy or hypertrophy in a given region for each subject.

The ROI variables included in the analysis were the left and rightdorsolateral prefrontal (DLPF) cortex. The DLPF was defined as thecortex encompassing Brodmann's areas 8, 9,10, and 46, according tothe Talairach atlas. A more detailed description can be foundelsewhere (Molina et al., 2003).

2.4. Statistical methods

We first assessed the statistical significance of the deviation in theMRI and metabolic values in the patients by comparing them withthose in the healthy controls using t tests for independent samples.We then assessed the relationship between the global IQ and thescores on each test and cognitive domain to evaluate global effects.Wealso assessed whether the antipsychotic doses in CPZ equivalents atthe time of the neuropsychological exams had a significant effect oncognitive performance (Pearson's r). Moreover, we calculated thedifferences between NN and chronic patients in each of the cognitivedomains by comparing their z scores (U test).

A priori, we also planned to carry out two evaluations to rule outthe potential effect of patients' failing to collaborate and/or state-dependent cognitive abnormalities related to acute symptoms onmetabolic activation. First, we calculated the correlation between theIQ scores and cerebral measurements, as well as the correlationsbetween the IQ scores and scores on each test in order to rule out aglobal effect. We then assessed the correlation coefficients betweenthe clinical and neuropsychological scores under the suspicion thatthe more paranoid and/or more negative patients might cooperateless in the testing.

Regarding the main objective of the study, we used partialcorrelation tests to assess the correlation coefficients between thePET (relative DLPF activation) and MRI (DLPF structural residuals)results and the z scores in the defined cognitive domains, partiallingout global IQ scores. We also planned to repeat the same comparisonsin the NN group in order to rule out the effect of chronicity, in this caseusing Spearman's ρ test owing to the small sample size.

3. Results

The patients (both the whole group and the NN patients) had asignificantly lower right DLPF residual (t=2.75, df=60, P=0.008)

population matched in sex, age and education).

N=9

on Centiles z scores Interpretation

it 24.44 −0.68 Slight deficitit 37 −0.33 Slight deficitit 37 −0.33 Slight deficiteficit 15,78 −1 Moderate deficitit 24,44 −0.68 Slight deficiteficit 2 −2.00 Moderate to severe deficitit 30.33 −0.56 Slight deficith 90.6 1.33 Normal-highit 34.56 −0.37 Slight deficit

61 0.27 Normal50 0 Normal57 0.15 Normal

n when available) C: centiles; ts: typical score and naïve subset.

Fig.1. Axial, view of the Talairach grid built for measuring volume andmetabolic activityof the cortex. ROI volume and metabolic activity are calculated as the portion of tissuemask contained in the group of cells that define the ROI (see Methods). This figurerepresents an editedMRI (i.e., after removing non-cerebral parts) fused with a PET scan.In this view, occipital is down and frontal is up. The voxels selected for structural andmetabolic analyses are highlighted.

Table 4Correlation (Spearman's rho) coefficients between metabolic and anatomicalmeasurements and cognitive domains in NN patients (n=9).

Processingspeed

Attention Workingmemory

Verbalmemory

Visualmemory

Reasoning andproblem-solving

DLPFactivity, l

0.567 −0.017 0.072 −0.017 0.059 0.367

DLPFactivity, r

0.733⁎ −0.067 0.431 0.060 0.134 0.550

DLPFresidual, l

0.100 0.383 0.192 0.034 −0.192 −0.317

DLPFresidual, r

0.367 0.283 0.743⁎ 0.604 0.494 0.133

The associations between processing speed and right activity and between workingmemory and structural deviation remain significant.⁎The correlation is significant at the 0.05 level.

91V. Molina et al. / Psychiatry Research: Neuroimaging 173 (2009) 88–93

and less left DLPF relative activity (t=2.13, df=60, P=0.03; seeTable 1) as compared with the controls.

Performance in the CPT test during PET acquisitionwas good in thepatients (83.8±17.1% correct responses). Performance in mostcognitive tests of the battery was, however, in the lower centile or zscores, except for total IQ (Table 2; Fig. 1). Regarding the cognitivedomains, the patients scored low on processing speed (TMT-A anddigit-symbol test), attention (Toulose-Piéron), working memory(WAIS arithmetic subtest) and short-term visual memory (Rey'sFigure). Concerning problem solving, low scores were obtained inperseverative errors but not in completed categories on the WCST.

No significant association was observed between the clinical andneuropsychological scores (either with negative or with positivesymptom scores). A trend was only found for the association betweennegative symptoms at the time of the neuropsychological exams and theproblem-solving scores (r=−0.48, P=0.07). There was no significant

Table 3Correlation (Pearson's r) coefficients and level of significance between metabolic andanatomical measurements and cognitive domains in all patients (n=22).

Processingspeed

Attention Workingmemory

VerbalMemory

VisualMemory

Reasoning andproblem-solving

DLPFactivity, l

0.418 0.096 0.071 0.131 0.190 −0.001P=0.056 P=0.679 P=0.771 P=0.572 P=0.410 P=0.995

DLPFactivity, r

0.486 0.036 0.261 0.065 0.298 −0.050P=0.026 P=0.878 P=0.374 P=0.574 P=0.190 P=0.828

DLPFresidual, l

0.179 0.105 0.402 0.326 0.128 0.288P=0438 P=0.651 P=0.088 P=0.150 P=0.581 P=0.205

DLPFresidual, r

−0.157 − .229 0.474 0.459 −0.091 0.095P=0.496 P=0.318 P=0.03 P=0.037 P=0.696 P=0.683

Activity reflects relative metabolismwith respect to total brain. Residuals are a measureof structural deviation (see Methods).

association between cognitive performance in any domain and anti-psychotic doses.

There were no significant differences in the z scores in each of thecognitive domains between NN and chronic patients (in all cases,zb0.5, P=ns).

There was no association between total IQ scores and metabolicactivity or anatomical residuals on either side.

There was a significant association between the total WAIS scoresand TMT B time (r=0.686, n=22, Pb0.001), accuracy and richness ofreproduction (r=0.443, n=22, P=0.03). No association between IQand WCST, WAIS arithmetic and the TAVEC scores was observed.

3.1. Association between structural deviation and neuropsychologicalscores

We observed an association between right PF GM residuals and theverbal memory scores (Table 3). This pattern disappeared when theNN patients were studied alone (Spearman's ρ=0.569, n=9,P=0.14).

We also observed an association between working memory scoresand right PF GM residuals (Table 3). This correlation was stillsignificant in the NN patient studied alone (Table 4).

No association was found between IQ (total, verbal or manip-ulative) and anatomical residuals.

3.2. Association between functional and neuropsychological scores

Right and left frontal activation was directly and significantlyrelated to the processing speed (Table 3) scores. In the NN group, thisassociation was replicated in the right side (Table 4).

No association was detected between functional values and total,manipulative or verbal IQ scores.

4. Discussion

In this study, our schizophrenia patients showed a significantassociation between metabolic prefrontal activation and processingspeed scores and between structural deviation and working memorydeficits. Those associations did not depend on clinical (psychopatho-logical and CPZ equivalents) or general IQ factors.

MRI and PET studies were performed at a different time thanneuropsychological testing, and the activation paradigm during PETacquisition was not included in the battery. Despite this, metabolicactivation was significantly related to neuropsychological performanceafter clinical stabilization, at least to processing speed and workingmemory. The former association might simply be due to a lower extentof collaboration in the exams (i.e, the less the involvement of the subjectin the task, the lower the activation to be expected) or to an effectdependent on the psychopathological acute state. However, thispossibility seems unlikely since the IQ scores were within the normal

92 V. Molina et al. / Psychiatry Research: Neuroimaging 173 (2009) 88–93

limits, and the scores on the other tests were within the expected rangefor schizophrenia, implying good cooperation in cognitive testing.Moreover, although the symptom scores were low at the time of theneuropsychological exams, no association was found between theneuropsychological and clinical scores, and no significant relationshipwas observed between the metabolic measurements and the IQ scores.Taking all this into account, the present results seem to indicate thatcapacity of the prefrontal lobe to become activated during an attentionprobe may predict cognitive performance after clinical stabilization, atleast in tasks that require rapid processing.

The processing speed scores were directly correlated with metabolicactivity in the DLPF region, suggesting that hypofrontal patients wouldalso showa deficit in this cognitive domain. Nevertheless, other cognitiveproblems, such as problem-solving deficits as assessed by theWCST,wereunrelated to frontal metabolism, which seems coherent with theidentificationof separable cognitive factors in schizophreniawithpossiblydistinct substrates (Nuechterlein et al., 2004). Neuropsychologicalalterations other than processing speed might depend upon structuralor functional alterations different from frontal metabolic hypoactivation.

Processing speedandworkingmemoryscoreswere associated in ourpatients to different parameters of brain abnormalities, supporting acomplex relationship between brain abnormalities and cognitivefunction in schizophrenia. This idea seems compatible with the widelydistributed networks underlying cognitive functions in the human brain(Mesulam, 1990). In this vein, the association reported here betweenregionalmetabolismand cognitiveperformance suggests that an illness-related alteration at some point of a complex cognitive network mayhave a detectable effect on certain aspects of cognitive function.

Keeping this in mind, the association found between hypofrontal-ity and cognitive deficit is coherent with the poorer prognosis of first-episode patients with hypofrontality (MacDonald et al., 2005; Molinaet al., 2005b), since cognitive deficits may lead to a poorer outcome inschizophrenia (Green et al., 2004). In particular, in schizophreniaprocessing speed may have important prognostic consequences(Dickinson et al., 2007), in agreement with the prognostic relevanceof hypofrontality, since according to our data both seem to be related.

Our findings indicate that at least processing speed deficits inschizophrenia are associatedwith prefrontal functional rather thanwithstructuralmeasurements at the same level, althoughnothing canbe saidof other regions. This seems coherent with previous evidence.Neuropsychological deficits may already be present in first-episodeschizophrenia (Sweeney et al., 1992), and functional (MacDonald et al.,2005; Molina et al., 2005b) but not structural (Molina et al., 2006)deficits have been described in first episodes that evolve intoschizophrenia (in comparison with first episodes with a betterprognosis).

The observed association between metabolic and neuropsychologi-cal measurements is also consistent with the possible improvement inthe latterwith some treatments. Clozapinemay have a positive effect onattention, response speed, and fluency (Goldberg and Weinberger,1994), and at the same time it has been shown to increase the activationof prefrontal areas during a cognitive paradigm (Lahti et al., 2004).

In light of our results, it could be speculated that a decreasedrecruitment within frontal regions could contribute to the processingspeed deficit seen in schizophrenia; this decreased recruitmentwould inturn be reflected in lower glucose requirements and hence inhypofrontality as measured by FDG-PET. Such a decreased requirementmightbe related to adecreaseddopaminergic tone in the frontal cortex inschizophrenia, as has been proposed (Winterer and Weinberger, 2004).

Moreover, an associationwas found between thedegree of structuraldeviation (i.e., the amount of grey matter deficit given the age andintracranial volume of each subject) and the performance in a workingmemory test. This is in part consistent with association between long-and short-termmemory deficits, prefrontal greymatter atrophy and theDISC-1 gene in schizophrenia (Cannon et al, 2005). Considered togetherwith the relation between hypofrontality and speed processing, the

association between working memory and structural deviation mayindicate that both dimensions of frontal deviation may have differentconsequences. Furthermore, this dissociation seems to support adifferent substrate frontal atrophy and hypoactivation in schizophrenia,which is also supported by the lack of association between measures ofstructural and metabolic deviation in schizophrenia (Molina et al.,2005c).

Our patients did not show any significant association betweentheir attention, visual memory or problem-solving scores and theirfrontal metabolic or structural measurements. This may be due to thesmall sample size (a type II error) or could be related to the significantcontribution of other areas to performance in the underlying cognitiveabilities (Alvarez and Emory, 2006). Other explanations for this lack ofassociation are possible and deserve investigation. For example,schizophrenia patients may overactivate the prefrontal cortex toachieve a correct performance in cognitive tests (Callicott et al., 2000;Manoach et al., 1999). In other words, from a low baseline, increasedactivation may result in an apparently normal metabolic state toachieve even a low performance. Nevertheless, our approach (a singlePET study) does not allow us to investigate this possibility.

Regarding themain limitations of the study, our sample size is smalland the treatments differed among the patients. However, a similarsignificant associationwas found in the NN patients receiving the sametreatment. Moreover, cognitive domains are presumably inter-related,and hence some support seems to be available for a relationshipbetween hypofrontality and non-specific cognitive performance,although large-scale studies have identified the domains used here asseparate dimensions. Additionally, we did not assess the same relation-ships between cognition and brain parameters in healthy controls;thereforewe cannot drawany conclusion regarding the specificity of theassociations found here. Finally, multiple correlations were performedbetween cognitive and cerebral parameters, although the hypothesiswas an association between neuropsychological and, on the other side,structural and/or functional PF deviation. In this context, the presentresults have to be considered exploratory. The advantages of this studyare that a comprehensive cognitive battery and two different imagingtechniqueswere used in each patient, allowing relevant complementaryinformation to be obtained that explains cerebral alterations better thanstudies using a single technique.

Wemay conclude fromour data that at least processing speed can berelated to hypofrontality in schizophrenia, and perhaps also workingmemory deficits may relate to structural atrophy. Taken together, ourdata suggest the existence of an association between cognitive and brainparameters, independently of the most conspicuous symptoms. Largersample sizes, in particular including patients without previous treat-ment, are needed to confirm these results.

Acknowledgments

We are grateful to Dr Manuel Cuesta for his kind and extremelyvaluable comments on our work.

This research was supported in part by the Spanish Ministry ofHealth, Instituto de Salud Carlos III, Red de Enfermedades Mentales(REM-TAP Network).

References

Alvarez, J.A., Emory, E., 2006. Executive function and the frontal lobes: a meta-analyticreview. Neuropsychological Rehabilitation 16, 17–42.

Ashburner, J., Friston, K.J., 1997. Multimodal image coregistration and partitioning — aunified framework. Neuroimage 6, 209–217.

Buchsbaum, M.S., 1995. Positron emission tomography studies of abnormal glucosemetabolism in schizophrenic illness. Clinical Neuroscience 3, 122–130.

Callicott, J.H., Bertolino, A., Mattay, V.S., Langheim, F.J., Duyn, J., Coppola, R., Goldberg, T.E.,Weinberger, D.R., 2000. Physiological dysfunction of the dorsolateral prefrontal cortexin schizophrenia revisited. Cerebral Cortex 10, 1078–1092.

Cannon, T.D., Hennah, W., van Erp, T.G., Thompson, P.M., Lonnqvist, J., Huttunen, M.,Gasperoni, T., Tuulio-Henriksson, A., Pirkola, T., Toga, A.W., Kaprio, J., Mazziotta, J.,Peltonen, L., 2005. Association of DISC1/TRAX haplotypes with schizophrenia,

93V. Molina et al. / Psychiatry Research: Neuroimaging 173 (2009) 88–93

reduced prefrontal gray matter, and impaired short- and long-term memory.Archives of General Psychiatry 62, 1205–1213.

Cuesta, M.J., Peralta, V., 1995. Cognitive disorders in the positive, negative, anddisorganization syndromes of schizophrenia. Psychiatry Research 58, 227–235.

Cuesta, M.J., Peralta, V., Zarzuela, A., 1998. Illness duration and neuropsychologicalimpairments in schizophrenia. Schizophrenia Research 33, 141–150.

Dickinson, D., Ramsey, M.E., Gold, J.M., 2007. Overlooking the obvious: a meta-analyticcomparison of digit symbol coding tasks and other cognitive measures inschizophrenia. Archives of General Psychiatry 64, 532–542.

Fuster, J.M., 1999. Synopsis of function and dysfunction of the frontal lobe. ActaPsychiatrica Scandinavica. Supplementum 395, 51–57.

Goldberg, T.E., Weinberger, D.R., 1994. The effects of clozapine on neurocognition: anoverview. Journal of Clinical Psychiatry 55, 88–90 (Suppl B).

Green, M.F., Kern, R.S., Heaton, R.K., 2004. Longitudinal studies of cognition andfunctional outcome in schizophrenia: implications for MATRICS. SchizophreniaResearch 72, 41–51.

Hollingshead, A., Frederick, R., 1953. Social stratification and psychiatric disorders.American Sociological Review 18, 163–189.

Kay, S.R., Fiszbein, A., Opler, L.A., 1987. The positive and negative syndrome scale(PANSS) for schizophrenia. Schizophrenia Research 13, 261–276.

Lahti, A.C., Holcomb, H.H.,Weiler, M.A.,Medoff, D.R., Frey, K.N., Hardin,M., Tamminga, C.A.,2004. Clozapine but not haloperidol re-establishes normal task-activated rCBFpatterns in schizophrenia within the anterior cingulate cortex. Neuropsychopharma-cology 29, 171–178.

MacDonald 3rd, A.W., Carter, C.S., Kerns, J.G., Ursu, S., Barch, D.M., Holmes, A.J., Stenger,V.A., Cohen, J.D., 2005. Specificity of prefrontal dysfunction and context processingdeficits to schizophrenia in never-medicated patients with first-episode psychosis.American Journal of Psychiatry 162, 475–484.

Manoach, D.S., Press, D.Z., Thangaraj, V., Searl, M.M., Goff, D.C., Halpern, E., Saper, C.B.,Warach, S.,1999. Schizophrenic subjects activate dorsolateral prefrontal cortex duringa working memory task, as measured by fMRI. Biological Psychiatry 45, 1128–1137.

Mesulam, M.M., 1990. Large-scale neurocognitive networks and distributed processingfor attention, language, and memory. Annals of Neurology 28, 597–613.

Molina, V., Reig, S., Sarramea, F., Sanz, J., Artaloytia, J.F., Luque, R., Aragüés, M., Pascau, J.,Benito, C., Palomo, T., Desco, M., 2003. Anatomical and functional brain variablesassociated to clozapine response in treatment-resistant schizophrenia. PsychiatryResearch: Neuroimaging 124, 153–161.

Molina, V., Gispert, J.D., Reig, S., Pascau, J., Palomo, T., Martínez, R., Desco, M., 2005a.Olanzapine-induced cerebral metabolic changes. Relation to symptom changes inschizophrenia. International Clinical Psychopharmacology 20, 13–18.

Molina, V., Sanz, J., Reig, S., Martínez, R., Sarramea, F., Luque, R., Benito, C., Gispert, J.D.,Pascau, J., Desco, M., 2005b. Hypofrontality in males with first-episode psychosis.British Journal of Psychiatry 186, 203–208.

Molina, V., Sarramea, F., Sanz, J., Benito, C., Palomo, T., 2005c. Prefrontal atrophy in firstepisodes of schizophrenia associated with limbic hyperactivity. Journal ofPsychiatric Research 39, 117–127.

Molina, V., Sanz, J., Sarramea, F., Luque, R., Benito, C., Palomo, T., 2006. Dorsolateralprefrontal and superior temporal volume deficits infirst episode psychosis that evolveinto schizophrenia. European Archives of Psychiatry and Clinical Neuroscience 256,106–111.

Nuechterlein, K.H., Barch, D.M., Gold, J.M., Goldberg, T.E., Green, M.F., Heaton, R.K., 2004.Identification of separable cognitive factors in schizophrenia. Schizophrenia Research72, 29–39.

Pfefferbaum, A., Lim, K.O., Zipursky, R.B., Mathalon, D.H., Rosenbloom, M.J., Lane, B., Ha,C.N., Sullivan, E.V., 1992. Brain gray and white matter volume loss accelerates withaging in chronic alcoholics: a quantitative MRI study. Alcoholism: Clinical andExperimental Research 16, 1078–1089.

Rummel, C., Hamann, J., Kissling, W., Leucht, S., 2003. New generation antipsychotics forfirst episode schizophrenia. Cochrane Database of Systematic Reviews CD004410.

Shenton, M.E., Dickey, C.C., Frumin, M., McCarley, R.W., 2001. A review of MRI findings inschizophrenia. Schizophrenia Research 49, 1–52.

Sweeney, J.A., Haas, G.L., Li, S., 1992. Neuropsychological and eyemovement abnormalitiesin first-episode and chronic schizophrenia. Schizophrenia Bulletin 18, 283–293.

Wechsler, D., 1999. Wechsler Adult Intelligence Scale. Ediciones TEA, S.A, Madrid.Winterer, G., Weinberger, D.R., 2004. Genes, dopamine and cortical signal-to-noise ratio

in schizophrenia. Trends in Neurosciences 27, 683–690.Wolf, R.C., Vasic, N., Hose, A., Spitzer, M., Walter, H., 2007. Changes over time in

frontotemporal activation during a working memory task in patients with schizo-phrenia. Schizophrenia Research 91, 141–150.

Woods, S.W., 2003. Chlorpromazine equivalent doses for the newer atypicalantipsychotics. Journal of Clinical Psychiatry 64, 663–667.

Woods, R.P., Mazziotta, J.C., Cherry, S.R., 1993. MRI-PET registration with automatedalgorithm. Journal of Computer Assisted Tomography 17, 536–546.