Topographic abnormality of slow cortical potentials in ... · abnormality (topographic) in schizophrenic patients, and thus serve as ... individual topographic deviation from normality,

97

Braz J Med Biol Res 37(1) 2004

Slow potential topography in schizophreniaBrazilian Journal of Medical and Biological Research (2004) 37: 97-109ISSN 0100-879X

Topographic abnormality of slowcortical potentials in schizophrenia

Laboratório de Neurosciências (LIM-27), Departamento de Psiquiatria,Faculdade de Medicina, Universidade de São Paulo, São Paulo, SP, Brasil

L.F.H. Basile, J. Yacubian,B.L.C. Ferreira, A.C. Valim

and W.F. Gattaz

Abstract

A recent study from our laboratory has provided evidence for thegeneration of slow potentials occurring in anticipation to task-perfor-mance feedback stimuli, in multiple association cortical areas, consis-tently including two prefrontal areas. In the present study, we intendedto determine whether these slow potentials would indicate someabnormality (topographic) in schizophrenic patients, and thus serve asan indication of abnormal association cortex activity. We recordedslow potentials while subjects performed a paired-associates memorytask. A 123-channel EEG montage and common average referencewere used for 20 unmedicated schizophrenic (mean duration of ill-ness: 11.3 ± 9.2 years; mean number of previous hospitalizations: 1.2± 1.9) and 22 healthy control subjects during a visual paired-associatesmatching task. For the topographic analysis, we used a simple index ofindividual topographic deviation from normality, corrected for abso-lute potential intensities. Slow potentials were observed in all sub-jects. Control subjects showed a simple spatial pattern of voltageextrema (left central positive and right prefrontal negative), whereasschizophrenic patients presented a more complex, fragmented pattern.Topographic deviation was significantly different between groups(P < 0.001). The increased topographic complexity in schizophrenicscould be visualized in grand averages computed across subjects.Increased topographic complexity could also be seen when grandaverages were computed for subgroups of patients assembled eitheraccording to task-performance (high versus low) or by their scores onpsychopathological scales. There was no significant correlation be-tween topographic deviation and psychopathology scores. We con-clude that the slow potential topographic abnormalities of schizophre-nia indicate an abnormality in the configuration of large-scale electri-cal activity in association cortices.

CorrespondenceL.F.H. Basile

Laboratório de Neurosciências

Departamento de Psiquiatria

FM, USP

Av. Dr. Ovidio Pires de Campos, s/n

05403-010 São Paulo, SP

Brasil

Fax: +51-11-3673-2201

E-mail: [email protected]

Research supported by FAPESP

(Nos. 98/07640-3 and 97/11083-0).

Received April 9, 2003

Accepted September 16, 2003

Key words• High-resolution EEG• Slow potentials• Contingent negative

variations• Schizophrenia• Source localization• Functional brain mapping

Introduction

Since quantitative and averaging electro-encephalographic methods were introduced,various electrophysiological alterations havebeen reported in schizophrenia. Alterations

were found in the power of particular spec-tral bands of the ongoing electroencephalo-gram, or in latencies, intensities and topogra-phy of event-related potentials. Among theformer are studies indicating increased poweron delta and theta bands in schizophrenia, at

98


L.F.H. Basile et al.

frontal electrodes, originally considered torepresent the electrical version of‘hypofrontality’ (1,2), although recent sourcelocalization studies showed these alterationsin frontotemporal and posterior cortical ar-eas (3). With respect to the second type ofelectroencephalographic studies on strictlysensory or exogenous evoked potentials, theliterature is still controversial, and even find-ings that were considered among the mostconsistent in the literature (nonsuppressionof the P50 auditory evoked potential) arecurrently being reinterpreted after the devel-opment of new methods of recording andanalysis (4).

It is expected that pathophysiologicalstudies in schizophrenia should benefit morefrom indexes of association rather than sen-sory-motor cortical activity. The analysis ofendogenous potentials of the P300 class orof slow potentials (of the contingent nega-tive variation (CNV) type), which are phe-nomena more directly related to voluntarybehavior, offers a possibility for study. TheP300s are attention-related potentials (5)believed to originate in sensory plus poly-modal cortices (6). Increased peak latenciesand reduced amplitudes have been observedin schizophrenia for visual, auditory andsomatosensory P300 potentials (7-11).

The slow potentials of the CNV class areparticularly interesting, since they appear toreflect electrical activity from the prefrontalcortex. The CNV obtained during traditionaltasks, which require a simple timing of mo-tor responses, are negative shifts of the elec-troencephalogram that peak at fronto-cen-tral sites (12). However, it seems that themere anticipation of any type of task-rel-evant stimulus is sufficient to provoke slowpotentials. Since their discovery (13), CNVshave been assumed to be generated by the(pre)frontal cortex. Evidence from differenttypes of studies has confirmed the prefrontalareas at least as the main centers of slowpotential generation, i.e., studies with mod-eling of intracranial generators of CNVs or

their magnetoencephalographic equivalentsin the form of few equivalent current dipoles(14-18), lesion studies (19,20), and invasiveintracranial recordings (21,22).

In a recent study, using a 123-channelelectroencephalographic montage and com-mon average reference, we analyzed the in-tracranial generators of the slow potentialsin healthy subjects, including the controlindividuals that participated in the presentstudy, by current density reconstruction. Wefound evidence for multifocal generation ofthese slow potentials, including prefrontalareas 9 and 10 in all cases, plus a number ofposterior association cortices varying acrosssubjects (23). In the present study, we usedthe same task as in the prior one applied to agroup of unmedicated schizophrenic patientsto compare their slow potentials with thosefrom the healthy individuals. We focused onthe scalp topography of the slow potentialsas a first approach to those putative noninva-sive indexes of association cortex activity.

We performed two types of statisticalcomparisons of the topographic data betweencontrols and patients. As a first approach, weused an analysis of a global nature, extract-ing a topographical index from each indivi-dual’s whole set of electrodes. Then, wecompared each electrode separately to deter-mine which subset of the montage accountedfor a possible difference in topography be-tween groups.

Material and Methods

Subjects

The sample comprised 20 schizophrenicpatients (DSM-IV; 14 males and 6 females,mean age 32.0 ± 7.9 years) and 22 healthycontrols (14 males and 8 females, mean age32.4 ± 9.3 years). All subjects had normal orcorrected to normal vision and hearing.Schizophrenic patients had been drug freefor at least 3 weeks, and 4 patients were drugnaive. Structured interviews were used, with

99


Slow potential topography in schizophrenia

10 patients being classified as paranoid, 6 asresidual and 4 as disorganized. Eleven pa-tients were hospitalized, 3 of whom withprevious hospitalizations. Psychopathologywas assessed by the anchored version of theBrief Psychiatric Rating Scale (BPRS-A) (24)and the Negative Symptoms Rating Scale(NSRS) (25). The mean psychopathologicalscores were 27.1 ± 6.6 in BPRS and 14.8 ±8.8 in NSRS. Patients were classified into(overlapping) subgroups based both on acategorical distinction (subtype diagnosisfrom structured interview) and on their rank-ing on the independent psychopathologicaldimensions (scores on scales). Mean dura-tion of illness was 11.3 ± 9.2 years, and themean number of previous hospitalizationswas 1.2 ± 1.9. All subjects were informedabout the experiment and signed consentforms approved by the Ethics Committee ofthe University Hospital.

Stimuli and task

All aspects of the task were controlledwith a commercial computer program (Stim;Neurosoft Inc., El Paso, TX, USA). Visualstimuli subtending less than three degrees inthe visual field were presented on a com-puter screen and formed pairs associated intime, separated by 2.5 s. The three types ofpairs comprised randomly intercalated ver-bal, pictorial and spatial subtasks (words:concrete nouns; abstract visual patterns andtraces; central arrows or circles indicating 8possible positions on the screen). Each stimu-lus lasted 0.5 s and the intertrial interval was6 s. Subjects were required to memorize 8pairs per block (each presented twice, withreversed order the second time), keeping inmind whether or not there was a categoricalmatch for each pair (food and nonfood -verbal task; patterns and traces - pictorialtask; same or different position - spatial task).During the test blocks, the same sequencejust memorized was presented. However,only the S1 would be initially presented and

remained on the screen until the subjectdecided whether or not its recently learnedassociated stimulus (S2) belonged to the samecategory as S1. They had to indicate a matchby pressing a button with the right indexfinger or a mismatch by using the middlefinger. After a button press, S1 disappearedfrom the screen, and was followed, 2.5 slater, by its corresponding S2. Two and ahalf seconds after S2, feedback stimuli (S3)were presented. In the present study we onlyanalyzed the slow potentials correspondingto the 5 s of anticipation of S3 stimuli duringthe test trials. The analyses of slow poten-tials corresponding to the memorization tri-als, including source reconstruction sepa-rated by subtask, were presented elsewhere(26, and Basile LFH, Yacubian J, Castro CCand Gattaz WF, unpublished results). Cor-rect trials were followed by visual S3 stimulicomposed of interesting pictures of photo-graphic models and artists that lasted on thescreen for 1.5 s, and incorrect trials werefollowed by an unpleasant sound (500 Hz,70 dB). A new trial (next S1) was started 3.5s after S3 offset. Eighteen blocks were pre-sented to each subject, thus resulting in atotal of 288 trials. We measured performanceby the percentage of correct trials. Our rea-sons to use such complex task instead of amotor-type CNV task are explained in thediscussion, and include the fact that the pres-ent study is part of a larger project in whichwe intend to analyze other potentials sepa-rately and study other psychiatric conditions(e.g., slow potentials from memorizationblocks and evoked potentials).

Recordings, computation of averagepotentials and topographic analysis

We used a fast Ag/AgCl electrode posi-tioning system consisting of an extended 10-20 system, a 123-channel montage (Quik-Cap; Neuromedical Supplies®, El Paso, TX,USA), and an impedance-reducing gel whicheliminated the need for skin abrasion (Quik-

100



Gel, Neuromedical Supplies®). Impedanceusually remained below 3 kOhms, and chan-nels that did not reach these levels wereeliminated from the analysis. Two bipolarchannels of the 123 channels in the montagewere used for recording both horizontal(HEOG) and vertical (VEOG) electro-oculo-grams. Artifact elimination was automatic:epochs containing signals in either HEOG orVEOG channels above +50 or below -50 µVwere eliminated. In our montage, the VEOGdetected blinks as deflections above 130 µVin the positive direction. In a pilot study, weobserved that the magnitudes of small sac-cades in various directions, directed at stimuliof approximately 3 degrees of eccentricity,were associated with deflections above 70µV in one or both EOG channels. Therefore,given our 50-Hz low-pass filter and epochsize of 5 s, the 50 µV criterion in the EOGchannels proved to be empirically sufficientfor the total elimination of epochs contain-ing blinks and small eye movements. Weused a minimum of 40 epochs of the total of96 from each task to compute the averageslow potentials. Linked mastoids served asreference only for data collection (commonaverage reference was used for mapping),and an electrode anterior to Fz (AFz) was theground. The impedances of reference elec-trodes were measured separately to be within1 kOhm each, but an internal circuit of theamplifiers summed them (the overall effectof our linked mastoid montage was mostly tolower event-related potential (ERP) ampli-tudes - with respect to a unilateral montage -but also to make topographic maps less proneto distortions from a single hyperactive tem-poral lobe). We used four 32-channel DCamplifiers (Synamps; Neuroscan Inc., ElPaso, TX, USA) for data collection and theScan 4.0 software package (Neurosoft Inc.)for computation of potential averages. Thefilter settings for acquisition were from DCto 50 Hz, and the digitization rate was 250Hz. The electroencephalogram was collectedin the continuous mode, and epochs for aver-

aging spanned the interval from 300 ms be-fore button press to 400 ms after S3 presen-tation. Baseline was defined as the 300 mspreceding button press. Artifact eliminationwas automatic: epochs corresponding to sig-nals in the EOG channel above +50 or below-50 µV were eliminated. The diagonal EOGdetected blinks as deflections of above 130µV in the positive direction. In a pilot study,we found that the magnitudes of small sac-cades in response to stimuli of approximately3 degrees of eccentricity were deflections ofabove 70 µV in either direction. Therefore,given our 50-Hz low-pass filter and epochsize of 5 s, in practice, our 50-µV criterion(in either direction) in the EOG channelalone proved to be sufficient for the totalelimination of epochs containing blinks andeye movements. We used a minimum of 60epochs out of a total of 288 to compute theaverage CNVs. Since the slow potentialspresently studied preceded and were relatedto the feedback anticipation and not to trialoutcome itself, both correct and incorrecttrials were included in the averages. Ourdesign permitted some subjects who haddifficulties to avoid blinking or small eyemovements to still produce a sufficient num-ber of uncontaminated epochs. However,even using such criteria, more than 5 pa-tients as well as 3 controls had to be elimi-nated from the originally recruited groupdue to artifact contamination (some patientswere not able to start the experiment due torestlessness or inability to learn the taskinstructions).

For comparison purposes, that is, to com-pute grand averages across subjects that al-low subtractions between individual andgroup data, we retained only the channelsdevoid of artifacts across all recording ses-sions. Thus, for computation of topographicmaps, our used montage was reduced from123 original electrodes to 87 good electrodes.The bad electrodes were distributed ran-domly over the montage, except for a row ofthe four lowermost occipital sites (neck move-

101



ment artifacts), so that no important gapsresulted in the electrode distribution. Theelectrical data were collapsed across the 22healthy subjects using the common montagefor computation of a grand average controlmap. To allow for an equal contribution byeach subject to the final pattern, where theshape of potential distribution was empha-sized (instead of field power, highly variableacross subjects), we rescaled each subject’sdata to an (arbitrary) common power value,choosing the value of the power of thenonscaled, absolute control group averagedata for this purpose (2.8 µV). To visualizepossible qualitative differences betweengroups, we computed a grand average for thepatient data as well. We also computed regu-lar grand average maps for performance andfor clinical subgroups, separating individu-als by the median values for task perfor-mance (below or above 71% for controls and58% for patients) and for scores in BPRS-positive symptoms (median = 11), BPRS-disorganization (median = 1.5) and NSRS(median = 14.5).

For visual inspection, all topographicmaps were computed in a three-dimensionalfashion. The discrete voltage distributionwas transformed into a continuous distribu-tion by linear interpolation, and the resultingdistribution was projected onto each subject’sskin surface. The skin surface was automati-cally segmented from individual magneticresonance image (MRI) sets by a commer-cially available program (Curry V4; Neuro-soft Inc.). Group averaged maps were pro-jected onto the segmented skin surface ofan individual of median head size. MRIswere obtained by a 1.5 Tesla GE machinemodel Horizon LX (Walkesha, WI, USA).Image sets consisted of 124 T-1 weighedsagittal images of 256 by 256 pixels spaced1.5 mm apart. Acquisition parameters were:standard echo sequence, three dimensions,fast spoiled gradient echo, two excitations,repetition time = 6.6 ms, echo time = 1.6 ms,flip angle of 15 degrees, and field of view =

26 x 26 cm. Total acquisition time was about8 min.

For quantitative analysis of the topo-graphic patterns, i.e., of how much eachindividual scalp distribution of potentialsdiffered from the control pattern, we meas-ured the topographic deviation as theEuclidian distance (quadratic vector norm)on an electrode-by-electrode basis betweeneach subject’s rescaled data and the controldata. That is, we computed the squared rootof the sum of squared differences in voltagebetween each pair of electrodes (group aver-age minus individual) divided by the numberof used channels. Thus, the resulting devia-tion index was measured in µVs, and bydefinition was identical to the mean globalfield power (MGFP) of the potentials ob-tained by subtraction between individual andgroup averaged data. MGFP was computedas the squared root of the sum of squaredpotentials, divided by the number of chan-nels. The rationale for using the control groupas a ‘normal data bank’ reference for com-paring patient data could lead to a trivialdifference, since the control group was notlarge and may not be representative of alarge population. Thus, we also computedthe relative deviation index between eachindividual and a grand averaged data setobtained from all subjects, controls and pa-tients.

We performed statistical analysis (Stu-dent t-test) to compare the deviation index,MGFP and task performance across the pa-tient and healthy control groups. We alsocomputed Pearson’s correlation coefficientsbetween these physiological indexes and thepsychopathological measures (NSRS, BPRSand subscales). Finally, we performed amultivariate statistical analysis (MANOVA,with electrodes as dependent variables andgroup as fixed factor) using the potentialvalues at all electrode locations in the mon-tage, to determine whether particular regions(electrode subsets) accounted for possiblegroup differences.

102



Results

Task performance and mean global fieldpower

Individual reports on task difficulty werevariable, in agreement with performancescores. Average performance (± standarddeviation) was 72.4 ± 11.8% for the controlgroup and 59.7 ± 10.1%, for the patientgroup. The difference in performance wasstatistically significant (P = 0.001). Perfor-mance was significantly correlated with to-tal BPRS score (r = -0.472, P = 0.036), withNSRS (r = -0.524, P = 0.018) and withBPRS-subscore affective blunting/negativesymptoms (r = -0.454, P = 0.044). We di-vided the patients into pairs of subgroupsbased on high versus low scores on eachpsychopathological scale, and tested whethermean performance was different betweensubgroups using the Student t-test. Only thesubgroups defined by a total BPRS scorediffered significantly in performance (highBPRS subgroup: mean performance = 54.4 ±6.00%; low BPRS subgroup: mean perfor-mance = 64.2 ± 11.0%; P = 0.023).

All subjects presented slow potentials,regardless of task performance, althoughthree patients had potentials of very lowamplitudes (MGFP below 3 µV). The aver-age MGFP was 5.8 ± 2.3 µV for the controlgroup and 5.1 ± 2.2 µV for the patient group,at the slow potential peak amplitude (5.0 safter button press, or at the moment of S3onset), and this difference was not signifi-cant. Signal-to-noise ratios, initially com-puted on a channel-wise basis (from the 300ms baselines) and then as an average acrosschannels, were also similar and not signifi-cantly different between groups (7.88 and7.35, respectively). Among subgroups ofpatients, divided by scores on BPRS andNSRS, MGFP was significantly different onlybetween subgroups of high and low scoreson BPRS for positive symptoms (6.0 ± 1.5and 4.11 ± 0.9 µV, respectively; P = 0.05).

Slow potential topography at visualinspection

The main finding and focus of the pres-ent study concerned the spatial distributionof slow potentials. Whereas most healthysubjects presented a simple pattern of poten-tial extrema, with maximum positive voltageclose to the C3 electrode and maximum nega-tive close to FP2, the patients presented amore complex pattern, with a higher degreeof variability in slow potential topography.Typically, in patients, the regions of ex-trema, especially the positive one, were frag-mented into many disconnected minor peaks,with a more folded, sinusoidal appearanceof the zero isopotential contour lines. Suchfragmentation of potential extrema can bevisualized on maps computed from eitherindividual or group averaged data (Figure 1).Due to the more complex appearance ofpatient maps, we counted the total number ofdisconnected positive voltage regions in eachsubject. There was no significant differencein such numbers between groups. Therefore,we concluded that the difference in mapsbetween groups was due to the shape and notto the number of same polarity regions, withpatients having more interdigitation betweenpositive and negative regions.

We also analyzed the appearance of grandaverages resulting from the division of pa-tient and control groups into subgroups basedon performance and on scores on BPRS(subscores for positive symptoms and disor-ganization) and NSRS. In controls, the mapsobtained from high and low task perfor-mance subgroups were similar in topogra-phy. In patients, the maps from both low andhigh task performance subgroups showedthe fragmented pattern (Figure 2). In all cases,maps from patient subgroups deviated fromthe control map, sharing the fragmented pat-tern. Among all patient subgroups, the mapscorresponding to high BPRS subscores weremore grossly fragmented and presented highpositive voltage in the right hemisphere.

103



3.85

2.57

1.28

0 -1.2

8-2

.57

-3.8

5

6.22

4.15

2.07

0 -2.0

7-4

.15

-6.2

2

[uv]

16.1

10.7 5.35

0 -5.3

5-1

0.7

-16.

1

[uv]

3.51

2.34

1.17

0 -1.1

7-2

.34

-3.5

1

5 µv

5 µv

5 µv

5 µv

Gra

nd a

vera

ge d

ata

Indi

vidu

al d

ata

Sch

izop

hren

ic g

roup

Sch

izop

hren

ic s

ubje

ct

Con

trol

gro

upC

ontr

ol s

ubje

ct

S1

S2

S3

S1

S2

02

40

24

Figure 1. Group averages andindividual examples of slow po-tentials and corresponding iso-potential contour maps. Mapsshown were at the onset of S3 -performance feedback stimu-lus. Angle of view: 40º from thenasion-right-left preauricularplane. Maps were projected onthe skin surface and were auto-matically segmented from indi-vidual magnetic resonance im-ages. Group average mapswere normalized so that eachindividual contributed equallywith respect to electricalpower, and were projected ontothe segmented skin of a sub-ject with medium head size.Numbers in color scale: voltagein µV.

Tim

e (s

)Ti

me

(s)

104



Figure 2. Topographic mapsfrom group averaged data corre-sponding to subgroups of pa-tients with task-performanceabove (left, labeled good) andbelow the median (right, labeledpoor). Scales are as in Figure 1.At first impression, the good per-formance average map re-sembles the control group mapfrom Figure 1, but notice the lowpotential gradients on the righthemisphere between the volt-age extrema.

6.614.42.20-2.2-4.4-6.61

10.56.993.490

-3.49-6.99

-10.5Good Bad

Task performance

Figure 3. Topographic deviation scatterplot, with electrical power (originally in µV)transformed into z-scores. Each score was obtained from the root mean square ofthe channel-by-channel differences between individual slow potential amplitudeand the group averaged potentials (controls plus patients). The control groupconsisted of 22 subjects and the schizophrenic group of 20 patients.

Figure 4. Electrode positions where statistically significant differ-ences in mean voltage between patients and controls wereobserved (P < 0.01, MANOVA). The fronto-central set was cen-tered at the left of FCz and the frontotemporal set at FT8. Elec-trodes were plotted on the original control group average map toshow relation with regions of high slow potential voltage.

5.83.861.93

-1.93-3.86-5.8

0

z-S

core

(de

viat

ion

inde

x)

3.00

2.00

1.00

0.00

-1.00

Control Schizophrenia

[µv]

105



Topographic deviation index

The deviation of individual electrical datafrom the control group averaged data wastabulated in the form of the topographic devia-tion index, as described in Methods, for groupand subgroup comparisons. The mean topo-graphic deviation was higher in patients thanin controls (P < 0.001). When topographicdeviation was computed with respect to grandaveraged data obtained from the complete setof subjects (controls plus patients), the result-ing individual indexes were virtually indistin-guishable from those obtained using only thecontrol subjects as a reference. Typically, dif-ferences in the indexes (electrical power) werenot altered beyond 0.2 µV for any given indi-vidual, with respect to the method using onlythe controls as the ‘normal data bank’. In thiscase also, average indexes were significantlydifferent between controls and patients (P <0.001). The resulting deviation indexes arepresented in Figure 3, with electrical power inµV transformed into z-scores. We also com-puted correlations between MGFP and devia-tion index with scores on the BPRS (includingsubscores) and NSRS. There was a significantcorrelation between topographic deviation andperformance (r = -0.358, P = 0.020), but notwith scores on the psychopathological scales(the correlation between topographic devia-tion and the BPRS subscore for thought disor-der almost reached significance: r = 0.430, P =0.058). Finally, we divided the group of pa-tients into halves, i.e., subgroups defined byhigh versus low task performance and highversus low scores on psychopathology, to testwhether any pair of subgroups significantlydiffered in topographic deviation. We foundthat topographic deviation was significantlydifferent between subgroups of high and lowscores on BPRS for positive symptoms (P =0.042; Table 1).

Individual electrode analysis

We performed a MANOVA to test for

differences in mean slow potential ampli-tudes in individual electrodes across the twogroups. By using a significance cutoff levelof P = 0.01, a simple spatial pattern of elec-trodes emerged as the subset differing acrossgroups. They were grouped in only twoplaces, 7 in the (left) fronto-central and 4 inthe right frontotemporal regions. The firstregion was centered at the left of the elec-trode between Fz and Cz (FCz) and thesecond at a right frontotemporal electrode(FT8). In all cases, the electrodes were thosecorresponding to high voltage positions fromthe control map, following a line of highestoverall voltage gradient (Figure 4). In allcases, the differences in voltage were in theform of low amplitudes on the patients’ av-erage potentials as compared to the controldata. That is, patients had low electricalpower on the (left) fronto-central and rightfrontotemporal regions.

Discussion

We observed slow potentials in the re-cordings from all individuals. During trials,when a given subject did not remember astimulus paired-association, the task wouldbe reduced to mere guessing and anticipa-tion of the feedback stimuli: subjects had atleast to initiate trials at will, and wait without

Table 1. Topographic deviation index, in µV, forsubgroups of patients defined by task performanceand psychopathology scores.

Subgroup High Low

Performance 3.8 ± 1.0 4.5 ± 1.5NSRS 3.9 ± 0.9 4.3 ± 1.6BPRS total 4.7 ± 1.5 3.6 ± 0.6BPRS-positive symptoms 4.5 ± 1.5 3.8 ± 0.9BPRS disorganization 4.6 ± 1.4 3.7 ± 1.0

Patients were divided, in each case using the me-dian performance or score value, into two halvescontaining subjects scoring above (high) or below(low) the median values. Data are reported asmeans ± SD. BPRS and NSRS = Brief Psychiatricand Negative Symptoms Rating Scales, respectively.

106



blinking for each outcome. Thus, it is impor-tant to insist that the present potentials werenot related to task performance (correct ver-sus incorrect outcomes), and especially notrelated to remembering each pair of stimulisince remembering their categorical relationwas sufficient for correct performance. Theslow potentials corresponded to the antici-pation of the feedback stimuli, and not to thepresented feedback, since they preceded eachS3. However, we chose the present task foradditional reasons: by requiring buttonpresses at the start of the trials, we avoidedthe presence of movement-related fields onour recordings, so that the slow potentialswould predominantly reflect association cor-tex activity (17,23); second, some schizo-phrenic patients may functionally resemblepatients with medial frontal lesions, who areknown to have low performance on ‘experi-mental gambling’, which has a guessing as-pect similar to our task (27,28); finally, weoriginally expected that both the guessingand the anticipation of the particular type ofS3 stimuli that we chose could engage thesemedial, phylogenetically older prefrontal ar-eas. That is, the S3 stimuli, both for signalingtrial outcomes and for being interesting inthemselves, were expected to recruit pre-frontal areas more densely interconnectedwith structures such as the expanded amyg-dala. However, according to our recent cur-rent density reconstruction results, only ar-eas on the prefrontal convexity were active,a result that we interpret to be due to theinsufficiently arousing or biologically ‘rel-evant’ nature of the stimuli (23).

The present type of slow potential pro-voking task thus appears to be well suited forpatients, since task performance is not strictlyrelated to either the presence or topographyof the potentials. We did expect more diffi-culties for the patients due to primary memorydeficits, which are believed to be part of ageneral intellectual impairment in schizo-phrenia (29-31). In our sample, some pa-tients fluctuated in time with respect to keep-

ing in mind the task goal, others were rest-less or clearly delusional, and two had im-portant thought disorder. In spite of thesedifficulties, which were more important thanremembering the pairs of stimuli, signal tonoise ratios and mean global field powerwere comparable across groups. Only threepatients had potentials of very low ampli-tude.

The most important finding of the pres-ent study was the difference between schizo-phrenics and controls regarding the spatialpattern of slow potentials over the scalp.Regions of voltage extrema over the scalp inschizophrenics were fragmented and of in-creased complexity when compared to thecontrol pattern. This could be observed inmost cases by visual inspection of isopoten-tial maps, and was reflected in the topo-graphic deviation index, whose mean valuewas significantly higher for the patient group.Moreover, the position of the highest electri-cal gradients (usually close to the zero isopo-tential lines) was more anterior in patientsthan in controls. This may be equivalent tothe findings of a more anterior conventionalCNV maximum in schizophrenic patients instudies using few electrodes and reference toparticular electrodes (32,33). It is possiblethat some controversies in the literature re-garding the topography of event-related po-tentials, for instance P3s or conventionalCNVs, may be clarified by the use of high-density sensory arrays, as used in the presentstudy. If in the present study fewer elec-trodes had been used, it is likely that a widevariability in potential amplitude would havebeen observed, as usually reported in theliterature. This was due to spatial undersam-pling of the complex potential distribution ofpatients; for instance, different choices foranalysis among neighbor electrodes fromthe 10-20 system can lead to opposite resultsfor a given patient.

It is of interest to emphasize that thetopographic deviation observed in our pa-tients was independent of psychopathologi-

107



cal status, clinical subtypes or task perfor-mance. By computing grand averaged mapsfor subgroups of patients, divided by themedian values of task performance or ofpsychopathological scores, we found that allsubgroups still presented the fragmentationof voltage extrema and altered position ofhighest gradients. This independence wasalso reflected in the lack of significant corre-lations between the topographic deviationindex and either NSRS, BPRS (including itssubscores) or task performance. However,we found a significant difference in topo-graphic deviation between subgroups of highversus low scores on BPRS for positive symp-toms. This may indicate that, although topo-graphic fragmentation does not correlate withany psychopathological dimension of analy-sis, patients with predominantly positivesymptoms are the ones who exhibit it notexclusively but more strongly. Because ourpatients were unmedicated or drug naive, itis unlikely that the topographic fragmenta-tion of slow potentials was secondary to anonspecific drug artefact. Further longitudi-nal studies are necessary to determinewhether the fragmentation observed may beseen as a trait marker of schizophrenia.

Only two regions on the scalp containedthe electrodes that had significantly differentvoltages at the slow potential peak acrosspatients and controls, namely, the (slightlyleft) fronto-central area containing the slowpotential-positive voltage peak and the rightfrontotemporal area corresponding to part ofthe high negative area. The overall lowerslow potential electrical power in such fron-tal scalp regions in schizophrenia is compat-ible with the ‘hypofrontality’ idea. How-ever, one should not expect a trivial corre-spondence between the (group) overall scalpregions of low voltages and individual pat-terns of intracranial generators.

Various underlying facts could explainthe topographic findings in schizophrenia.We speculate that the large scale functionalcortical areas, whose sustained electrical

activity is reflected as scalp slow potentials,present a lower definition in patients, due toa lower cohesion with each other or withsubcortical cell fields and nuclei (such as thecholinergic or dopaminergic ascending pro-jections). The latter may be the very struc-tures that define cortical functional areas,i.e., dynamic ensembles of areas establishedafter task learning. Alternatively, the genera-tors of average slow potentials in patientsmay be simply more dispersed, possibly dueto a loss of normal cortico-cortical connec-tivity, as predicted from models of abnormalcortical pruning (34). Finally, it is also pos-sible that an overall different set of areasbecome active when patients perform a giventask, as compared to healthy subjects. Thesehypotheses might be confirmed or refutedafter the development of quantitative meth-ods to compare our source reconstructionresults across groups. The available quanti-tative methods (computation of current den-sity vector sums restricted to chosen percen-tile bands of current distribution) when ap-plied to the reconstruction results of thepresent data from patients (results not shown),did not distinguish the groups in a statisti-cally significant way. However, at visualinspection, they are in agreement with thedispersion hypothesis. We are currently ana-lyzing the generators of long-latency endog-enous evoked potentials corresponding tothe three subtasks, and have already ob-served a significantly lower focality (currentdensity/volume occupied by currents) in pa-tients.

Finally, we must stress that, although thegenerators of the slow potentials of the pres-ent type are complex in distribution even inhealthy individuals (23,26), and did not al-low for a quantitative demonstration of thedispersion hypothesis, the present study mustbe interpreted as indirect evidence in favorof an abnormality in multiple cortical asso-ciation areas in schizophrenia. This agreeswith a growing body of independent evi-dence against a mere ‘hypofrontality’ hypo-

108



thesis. Neuropathological, anatomical andfunctional studies have emphasized suchwidespread nature of cortical alterations (35-38). In particular, the concept of decreasedprefrontal activity in schizophrenia itself isbeing revised: metabolic tracing studies(fMRI, PET and SPECT) show opposite re-sults depending especially on task condi-tions (for a review, see Ref. 35). Our ownrecent reconstruction results show bothhyper- and hypoactive cortical generators ofslow potentials in patients (Basile LFH,Yacubian J, Castro CC and Gattaz WF, un-published results). Thus, electrical and meta-bolic dysfunction in multiple cortical areas(with neuropsychological loss) is what onewould expect from the progressive tissueloss observed in anatomical longitudinal stud-ies using high-resolution MRI (36). The dis-

tributed cortical abnormalities in schizophre-nia, in turn, may reflect a still larger scale setof encephalic alterations, including thethalamo-cortical-basal ganglium circuits andtheir interconnections with the dopaminer-gic projections, according to findings fromneuropathological studies (37,38).

Acknowledgments

We wish to thank Dr. Cláudio C. deCastro (Incor, FM, University of São Paulo)for providing the NMR images for all sub-jects, and Dr. Milkes Y. Alvarenga and Prof.Marcus V. Baldo (Department of Physiolo-gy and Biophysics, ICB, University of SãoPaulo) for their critical suggestions aboutdata analysis.

References

1. Gattaz WF, Mayer S, Ziegler P, Platz M & Gasser T (1991). Hypo-frontality on topographic EEG in schizophrenia. European Archivesof Psychiatry and Clinical Neuroscience, 241: 328-332.

2. Sponheim SR, Clementz BA, Iacono WG & Beiser M (1994). RestingEEG in first-episode and chronic schizophrenia. Psychophysiology,31: 37-43.

3. Fehr T, Kissler J, Moratti S, Wienbruch C, Rockstroh B & Elbert T(2001). Source distribution of neuromagnetic slow waves and MEG-delta activity in schizophrenic patients. Biological Psychiatry, 50:108-116.

4. Clementz BA & Blumenfeld LD (2001). Multichannel electroen-cephalographic assessment of auditory evoked response suppres-sion in schizophrenia. Experimental Brain Research, 139: 377-390.

5. Verleger R (1988). Event-related potentials and cognition: A critiqueof the context updating hypothesis and an alternative interpretationof P3. Behavioral and Brain Sciences, 11: 343-427.

6. Basile LFH, Rogers RL, Simos PG & Papanicolaou AC (1997).Magnetoencephalographic evidence for common sources of longlatency fields to rare target and rare novel stimuli. InternationalJournal of Psychophysiology, 25: 123-137.

7. Blackwood DH, Whalley LJ, Christie JE, Blackburn IM, St Clair DM& McInnes A (1987). Changes in auditory P3 event-related potentialin schizophrenia and depression. British Journal of Psychiatry, 150:154-160.

8. Louzã MR & Maurer K (1989). Differences between paranoid andnonparanoid schizophrenic patients on the somatosensory P300event-related potential. Neuropsychobiology, 21: 59-66.

9. Pfefferbaum A, Ford JM, White PM & Roth WT (1989). P3 inschizophrenia is affected by stimulus modality, response require-ments, medication status, and negative symptoms. Archives ofGeneral Psychiatry, 46: 1035-1044.

10. Niwa S, Hiramatsu K, Saitoh O, Fukuda M, Kameyama T, Itoh K &

Hayashida S (1992). Information dysregulation and event-relatedpotentials in schizophrenia. Schizophrenia Bulletin, 18: 95-105.

11. Iwanami A, Kanamori R, Isono H, Okajima Y & Kamijima K (1996).Impairment of inhibition of unattended stimuli in schizophrenic pa-tients: event-related potential correlates during selective attention.Neuropsychobiology, 34: 57-62.

12. McCallum WC (1988). Potentials related to expectancy, preparationand motor activity. In: Picton TW (Editor), Handbook of Electroen-cephalography and Clinical Neurophysiology. Human Event-RelatedPotentials. Revised series, Vol. 3. Elsevier Science Publishers,Kidlington, Oxford, UK, 427-534.

13. Walter WG, Cooper R, Aldridge VJ, McCallum WC & Winter AL(1964). Contingent negative variation: an electric sign of sensorimo-tor association and expectancy in the human brain. Nature, 203:380-384.

14. Basile LFH, Rogers RL, Bourbon WT & Papanicolaou AC (1994).Slow magnetic fields from human frontal cortex. Electroencepha-lography and Clinical Neurophysiology, 90: 157-165.

15. Basile LFH, Simos PG, Brunder DG, Tarkka IM & Papanicolaou AC(1996). Task-specific magnetic fields from the left human frontalcortex. Brain Topography, 9: 31-37.

16. Basile LFH, Brunder DG, Tarkka IM & Papanicolaou AC (1997).Magnetic fields from human prefrontal cortex differ during tworecognition tasks. International Journal of Psychophysiology, 27:29-41.

17. Bocker KB, Brunia CH & van den Berg-Lenssen MM (1994). Aspatiotemporal dipole model of the stimulus preceding negativity(SPN) prior to feedback stimuli. Brain Topography, 7: 71-88.

18. Tarkka IM & Basile LFH (1998). Electrical source localization addsevidence for task-specific CNVs. Behavioural Neurology, 11: 21-28.

19. Rosahl SK & Knight RT (1995). Role of the prefrontal cortex ingeneration of the contingent negative variation. Cerebral Cortex, 2:

109



123-134.20. Zappoli R, Zappoli F, Versari A, Arnetoli G, Paganini M, Arneodo MG,

Poggiolini D & Thyrion EZ (1995). Cognitive potentials: ipsilateralcorticocortical interconnections in prefrontal human cortex abla-tions. Neuroscience Letters, 193: 140-144.

21. Ikeda A, Luders HO, Collura TF, Burgess RC, Morris HH, Hamano T& Shibasaki H (1996). Subdural potentials at orbitofrontal and mesialprefrontal areas accompanying anticipation and decision making inhumans: a comparison with Bereischaftspotential. Electroencepha-lography and Clinical Neurophysiology, 98: 206-212.

22. Hamano T, Luders HO, Ikeda A, Collura TF, Comair YG & ShibasakiH (1997). The cortical generators of the contingent negative varia-tion in humans: a study with subdural electrodes. Electroencepha-lography and Clinical Neurophysiology, 104: 257-268.

23. Basile LFH, Ballester G, Castro CC & Gattaz WF (2002). Prefrontalcortex activity assessed by high-resolution EEG and current densityreconstruction. International Journal of Psychophysiology, 45: 227-240.

24. Woerner MF, Manuzza S & Kane JM (1988). Anchoring the BPRS:An aid to improved reliability. Psychopharmacology Bulletin, 24:112-117.

25. Iager AC, Kirsh DG & Wyatt RJ (1985). A negative symptom ratingscale. Psychiatry Research, 16: 27-36.

26. Basile LFH, Baldo MV, Castro CC & Gattaz WF (2003). The genera-tors of slow potentials obtained during verbal, pictorial and spatialtasks. International Journal of Psychophysiology, 48: 55-65.

27. Eslinger PJ & Damasio AR (1985). Severe disturbance of highercognition after bilateral frontal lobe ablation: Patient EVR. Neurol-ogy, 35: 1731-1741.

28. Damasio AR (1996). The somatic marker hypothesis and the pos-sible functions of the prefrontal cortex. Philosophical Transactionsof the Royal Society of London, Series B. Biological Sciences, 351:1413-1420.

29. Stip E (1996). Memory impairment in schizophrenia: perspectivesfrom psychopathology and pharmacotherapy. Canadian Jounal ofPsychiatry, 41 (Suppl 2): S27-S34.

30. Heinrichs RW & Zakzanis KK (1998). Neurocognitive deficit in schizo-phrenia: a quantitative review of the evidence. Neuropsychology,12: 426-445.

31. Hawkins KA (1999). Memory deficits in patients with schizophrenia:preliminary data from the Wechsler Memory Scale - Third Editionsupport earlier findings. Journal of Psychiatry and Neuroscience,24: 341-347.

32. Rockstroh B, Müller M, Wagner M, Cohen R & Elbert T (1994).Event-related and motor responses to probes in a forewarned reac-tion time task in schizophrenic patients. Schizophrenia Research,13: 23-34.

33. Wagner M, Rendtorff N, Kathmann N & Engel RR (1996). CNV, PINVand probe-evoked potentials in schizophrenics. Electroencephalog-raphy and Clinical Neurophysiology, 98: 130-143.

34. Hoffman RE & Dobscha SK (1989). Cortical pruning and the devel-opment of schizophrenia: a computer model. Schizophrenia Bulle-tin, 15: 477-490.

35. Manoach D (2003). Prefrontal cortex dysfunction during workingmemory performance in schizophrenia: reconciling discrepant find-ings. Schizophrenia Research, 60: 285-298.

36. Thompson PM, Vidal C, Giedd JN, Gochman P, Blumenthal J,Nicolson R, Toga AW & Rapoport JL (2001). Mapping adolescentbrain change reveals dynamic wave of accelerated gray matter lossin very early-onset schizophrenia. Proceedings of the National Acad-emy of Sciences, USA, 98: 11650-11655.

37. Arnold SE & Trojanowski JQ (1996). Recent advances in definingthe neuropathology of schizophrenia. Acta Neuropathologica, 92:217-231.

38. Powers RE (1999). The neuropathology of schizophrenia. Journal ofNeuropathology and Experimental Neurology, 58: 679-390.

Documents

Topographic abnormality of slow cortical potentials in ... · abnormality (topographic) in schizophrenic patients, and thus serve as ... individual topographic deviation from normality,