preservation of p300 event-related potential topographic asymmetries in schizol ... -...

378 Electroencephalography and clinical Neurophysiology, 1990, 75: 378 - 391 Elsevier Scientific Publishers Ireland, Ltd.

EEG 89043

Preservation of P300 event-related potential topographic asymmetries in schizol enia with use of either linked-ear or nose reference sites

Steven F. Faux, M a r t h a E. Shenton , R o b e r t W. McCar l ey , Paul G. Nestor , Br ian M a r c y

a n d A m y L u d w i g Harvard Medical School, Department of Psychiatry, Brockton VAMC and Massachusetts Mental Health Center, Brockton, MA (U.S.A.)

(Accepted for publication: 5 June 1989)

Summary Previous studies of the auditory P300 event-related potential (ERP) from our laboratory have reported a left- greater than fight-sided attenuation in medicated chronic schizophrenics compared with normal controls. A possible confound in these studies has been the use of the linked-ear reference (LER), which has been criticized on the grounds that it might either induce or suppress topographic asymmetries. To test the effects of LER on P300 asymmetries in schizophrenia, we recorded ERPs with both LER and a nose reference (NR) in a group of 20 chronic medicated schizophrenics and in group of 20 age-matched normal controls. We here report: (1) confirmation of our previous P300 findings of left temporal scalp region deficit using both LER and NR with a 28-electrode montage; this feature was prominent in the wave form associated with the target stimulus, without the use of the wave form subtractions of our previous studies; (2) no statistically significant topographic differences between the LER and NR for either the schizophrenic or normal subjects; and (3) better performance of the LER in differentiating schizophrenics versus normal controls, due to lower wave form variability. We conclude that the LER is preferable for studies using subject groups and methodology similar to the present study.

Key words: Auditory event-related potentials; P300; Scalp topography; Reference electrode sites: Schizophrenia

Two previous studies of the topography of the auditory P300 event-related potential (ERP) from this laboratory have reported a left- greater than right-sided attenuation in schizophrenics compared with normal controls, with a P300 amplitude minimum in the left temporal scalp region (Morstyn et al. 1983; Faux et al. 1988a,b). Both of these studies used a linked-ears reference (LER).

The present study addresses whether the LER used in our previous studies might have distorted ERP scalp distribution and, hence, might have suppressed or induced hemispheric asymmetries. The use of LER in topographic studies has been strongly criticized in that: (1) unequal activity at

Correspondence to: Dr. Robert W. McCarley, Dept. of Psychiatry l l6A, Harvard Medical School, 940 Belmont St., Brockton, MA 02401 (U.S.A.).

the scalp could be induced by unequal electrode impedances at the ears (Beamont 1983; Mowery and Bennett 1957) and, conversely, (2) unequal activity at left and right temporal scalp regions could be reduced because 'forcing the same potential on the two ears ... tends to make the amplitude contours rather more symmetric about the rnidline' (from Nunez 1981; see also Rugg 1983; Davidson 1988; Kahn et al. 1988).

Unfortunately there have been few empirical data bearing on these important questions derived from theoretical concerns. Empirical studies of actual EEG or ERP asymmetries rarely compare multiple reference sites (for review see Davidson 1988). Moreover, we know of n o study assessing the effect of reference site on ERP asymmetries in psychopathological populations. The question ap- pears especially pertinent since, despite continuing criticisms of LER usage, the LER continues to be

0013-4649/90/$03.50 © 1990 Elsevier Scientific Publishers Ireland, Ltd.

ERP TOPOGRAPHY IN SCHIZOPHRENIA 379

one of the most popular choices among topographic investigators (cf., Kahn et al. 1988).

The purpose of this study was to compare systematically P300 topography using the LER with that of a reference site not likely to induce topographic distortion. The nose (NR) and balanced sternovertebral (SVR; Stephenson and Gibbs 1951) references have the advantage that they do not induce or reduce asymmetrical EP fields and are among the most widely used alternatives to the LER. Each reference has had its advocates among topographic investigators (e.g., NR: Vaughan and Ritter 1970; Vaughan et al. 1980; and SVR: Wolpaw and Wood 1982; Wood and Wolpaw 1982). Clearly, however, the best choice is dependent on the experimental situation and is a compromise between flawed alternatives. For example, Wolpaw and Wood (1982) have argued that the nose lies within the brain electrical fields of the auditory EP and produces voltage gradients which vary across subjects. Scherg and Von Cramon (1985), however, reanalyzed Wood and Wolpaw's data and showed that no reference site, including SVR, is 'inactive' with respect to the auditory EP field if more than two dipole sources are simultaneously active within the epoch of interest, a situation that likely holds for the P300, which may have multiple source generators (reviewed in Vaughan and Arezzo 1988).

The reference site controversy in normals is thus dearly not resolved. However, in working with psychopathological populations additional considerations must be weighed, namely which site will produce records most free of artifact. Pilot data have led us to use the NR to compare with the LER, since the SVR reference produced more EMG contamination from neck and body muscle activity in patients unable to relax fully.

With respect to reference site, the basic plan of this study was quite simple: to determine whether left- versus right-sided schizophrenic and normal group differences found under the LER and NR were the same or different, and to determine if either reference was preferable for P300 studies. A secondary purpose of this study was to evaluate whether our previous findings based on BEAM technology (e.g., Duffy et al. 1981; Duffy 1982) and a 20-electrode montage were replicable using

additional electrodes, a different topographic mapping technology, and in a subject group approximately a decade older.

We here report (1) general confirmation of our previous P300 findings by both LER and NR, and (2) an absence of topographic differences for the two reference sites for either schizophrenics or normals.

Methods

Subjects The schizophrenic group was composed of 20

male chronic schizophrenics from the Brockton VAMC who met the criteria for schizophrenia of the 3 major diagnostic schemata: Washington University (WU; Feighner et al. 1972), DSM III-R (American Psychiatric Association 1987), and Re- search Diagnostic Criteria (RDC; Spitzer et al. 1978). All patients were administered the Schedule for Affective Disorders and Schizophrenia (Spitzer and Endicott 1978). This information, in conjunc- tion with chart reviews and videotaped interviews, was used to make WU, DSM III-R, and RDC diagnoses. The normal control group comprised 20 volunteers, primarily recruited from hospital staff. Controls had no personal or family history of psychiatric illness. Further criteria for subject selection were: ages between 20 and 55 years, male, right-handed, no neurological illness, no drug abuse, no history of alcohol abuse, and no medica- tions which would grossly affect the EEG (e.g., reserpine or barbiturates). Subjects with hearing impairments or verbal IQs below 70 were not selected 1. Subjects were given detailed information concerning the study protocol and gave in- formed consent.

All 20 schizophrenics were neuroleptic medicated with a mean dose equivalent to 1151 mg / day of chlorpromazine (equivalence calculated by

1 Twenty-nine patients, meeting the above exclusionary criteria were selected primarily from two inpatient psychiatric wards. Additional exclusions occurred because of inability to perform task accurately (4), refusal to complete the experiment (2), and excessive EMG/EOG artifacts (3).

380 S.F. FAUX ET AL.

average of Shader 1975; and Bassuk et al. 1983). The mean age (S.D.) of the schizophrenic group was 40.0 (+ 8.7) years and the control group was 37.7 (+ 11.6) years (t = 0.73, P > 0.47). Mean scores on the Information sub-scale of the WAIS-R (INFO) were 18.1 (+ 5.64) for schizophrenics and 22.3 (+3.91) for normal controls, yielding pro- jected verbal IQs (VIQ) of 99.5 and 111.0 (t (38) = 2.15, P < 0.04), respectively. Mean score on the Mini-Mental-State exam (Folstein et al. 1975) was 26.9 ( + 2.5) out of a possible 30 for schizophrenics, and was 28.5 ( + 1.1) for normal controls (t (38) = 2.55, P < 0.02). Mean years of schooling was 11.8 ( + 0.9) for schizophrenics and 15.2 ( + 3.2) for normal controls (t (38) = 4.53, P < 0.001).

Recording procedures Evoked potentials were recorded using either

linked-ears (LER) or nose (NR) reference sites in an auditory 'odd-ball paradigm.' ERPs were produced by tone-pips of 40 msec duration (10 msec rise time) presented at 1 sec inter-stimulus inter- vals using Etymotic insert headphones. The infrequent (15%) high-pitched tones (1500 Hz, 97 dB SPL) were presented pseudo-randomly, inter- spersed between frequent low-pitched tones (1000 Hz, 97 dB SPL). There was a continuous back- ground of 70 dB white noise. To allow subjects to adapt to the auditory environment, and for consistency with previous paradigms, subjects read a novel for 20 min while listening to the tones. No instructions for counting the tones were given, but subjects were told they would be asked questions about their reading.

After reading, subjects were instructed to stare at a central fixation point and silently count each high-pitched tone. The experimenter stopped the task after every 20-35 (randomly determined) target tone presentations and asked subjects to report their counts. Correct or near correct ( + 2) answers were rewarded with $1.00 or the equivalent in VA canteen coupons. An investigator sitting behind the patient monitored the patient for compliance with the counting task, assisted in coaching subjects to minimize muscle movement artifacts, and provided reward payments for accuracy. Schizophrenic subjects were able to perform this task with a median accuracy of 97%; no

subject scored lower than 82% accuracy on any one run of 20-35 infrequent tones. Normal con- trois performed the task with a median accuracy of 99% with the lowest performance at 89% 2

In the counting task, 600 tones, yielding between 80 and 100 (randomly determined) infrequent tones, were presented in each of the 2 reference site conditions. Order of reference site was counterbalanced within groups, such that one-half of the subjects from each group received the LER condition during the first block of 600 tone presentations, and the remaining received the N R first. The 2 reference sites were never simultaneously active, and electrical continuity between earlobe leads existed only in the LER condition.

ERPs were recorded from 28 scalp electrodes using an Electrode Caps International, Inc. electrode cap with tin plate electrodes. Scalp electrode placements included all electrodes in the international 10-20 system with 8 additional interpolated electrodes. Cap electrodes Fpl , Fp2, T3, Cz, and T4 were located by precise international 10-20 measurements, and all other 10-20 electrodes were positioned automatically at standard relative dis- tances. For the remaining interpolated electrodes: FTC1 was placed at the intersection between F3- T3 and C3-F7; and TCP1 was similarly placed between C3-T5 and P3-T3. CP1 and PO1 were placed at the midpoints of the diagonals formed by Cz-P3 and Pz-O1, respectively. Homologous electrodes were placed symmetrically on the con- tralateral side (FTC2, TCP2, CP2, PO2). Vertical EOG was recorded for each eye using supra- and infra-orbital electrodes. Electrodes on the external canthi recorded horizontal EOG. Placements on the left and right masseter muscles of the jaw were used to monitor EMG. Electrode impedance was maintained between 2 and 4 k fL The EEG was filtered using a bandpass of 0.3-40 Hz (Neurosci- ence, Inc., EEG amplifiers, 36 dB/octave roll-off for low-pass and 6 dB/oc tave for high-pass). Evoked potential averages for each channel were

2 Accuracy was defined by the formula: c=(1.0-[t p - a Va]), where c equals the proportion correct (overall accuracy), p equals the subject's count, and a equals the actual count.


constructed from 2 msec samples over a 616 msec time interval, yielding 308 sample points. Sam- piing began 16 msec prior to stimulus presentation (a duration fixed by the hardware manufacturer), and the average of this prestimulus interval was used to establish the baseline. Sample segments from any channel which had voltages in excess of + 50 #V were excluded from the average. Based upon this rejection criterion, individual and grand-averaged EMG and EOG wave forms from the artifact monitoring channels showed no activity > + 2 #V about the baseline. Evoked potentials were averaged separately for the following conditions: reference electrode site (LER and NR) and tone probability (infrequent and frequent) to produce infrequent attentive (Ia) and frequent attentive (Fa) wave forms.

Two P300 peak component latency measures were used for each Ia wave form. The first measure (maximum amplitude latency (MAL)) was defined as the data point at the Cz electrode with the largest positive voltage between 300 and 500 msec. The second measure (Woody latency [WL]) was calculated using a variation of Woody cross- correlation filtering (Woody 1967). Each individual's wave form at Cz was cross-correlated with a template wave form consisting of the grand-averaged wave form of the normal controls. An individual wave form segment between 300 and 500 msec was shifted one point in time along the corresponding segment in the template wave form, and a cross-correlation was calculated with each time shift (lag). Woody latency was then defined as the maximum amplitude latency in the template wave form minus the lag having the largest cross- correlation.

P300 integrated voltage was defined as the mean voltage between 300 and 400 msec in the Ia condition. This time-window captured the rising phase and peak of the P300 component in most subjects and was similar to that of previous studies (Mors- tyn et al. 1983; Faux et al. 1988a).

The first step in analysis was to examine the voltage topography using color-coded maps (Duffy et al. 1981; Duffy 1982) of the voltages at the 28 scalp electrodes. We then examined mean differences as a function of the variance at each site using a t-statistic significance probability map (t-

SPM; Duffy et al. 1981) to determine the area of maximal t-statistic separation between schizophrenics and normal controls. SPM analysis was used to insure that the planned subsets of electrodes for hypothesis testing were sufficient to describe group differences at entire scalp.

The presence of left temporal P300 deficits in Morstyn et al. (1983) and Faux et al. (1988a) was the basis for selecting, prior to data collection, a subset of electrodes for MANOVA. Specifically, in the Ia condition and for each reference site condition we examined the mean P300 integrated voltages at electrodes T3, C3, Cz, C4, and T4. The set of analyses used paralleled our previous studies: a simple ratio of left versus right amplitudes and a series of tests derived from MANOVA. The first MANOVA consisted of a 'within-groups' analysis to test the effects of reference site on P300 distribution at the scalp, i.e., to test for the presence of a scalp site X reference site interaction. (Since MANOVA is conservative with respect to declaring statistical significance, we also performed paired t tests ( N R - L E R ) of P300 integrated amplitudes at each of the 28 scalp electrode sites, in effect biasing the outcome toward finding some statistically different sites as a result of multiple t tests and greater statistical power.) The second MANOVA focused on 'between'- group differences, where Hotelling's T 2 test provided information on 3 kinds of differences: (1) overall group amplitude differences; (2) group amplitude differences at each scalp electrode site ( 'protected' t contrasts); and (3) group topographic differences based on a test of a group x electrode site interaction (profile analysis of parallelism). We also used McCarthy and Wood's (1985) algorithm for normalizing (scaling) the data to eliminate amplitude effects which might be misin- terpreted by MANOVA as an interaction.

In a previous study (Faux et al. 1988a) a P300 voltage criterion of 2 /~V at the T3 electrode correctly categorized 7 / 9 normals and 9/11 schizophrenics. The same 2 #V criterion level at T3 was used to determine the degree of differentiation of schizophrenic and control individuals in the LER condition.

An important methodological point is that, in the present study, our analyses focus on the P300


component derived from the Ia condition, e.g., the wave form associated with the target stimulus; we will refer to this simply as the 'P300.' This is in contrast to the 'P300g,' a subtraction wave form previously used by us, and one that combined wave forms to frequent and infrequent stimuli and in counting and non-counting conditions (see Faux et al. 1988a,b).

Results

(1) P300 topographic distribution and t -SPM analyses

For the normal control group P300 component distribution was symmetrical throughout the P300 interval and was similar for both linked-ears (LER) and nose reference (NR) conditions (Fig. 1). In both LER and NR conditions maximum amplitude was at about 380 msec and showed a centroparietal midline maximum.

The schizophrenic group P300 topography was generally similar for both LER and NR conditions (Fig. 1). However, both conditions showed more positive activity in the right temporal scalp regions than in the left. We next computed t-SPMs to compare schizophrenic vs. control group difference topography for all scalp electrode sites (see Fig. 1). In the LER condition the largest t values were in the left frontotemporal region (F7, FTC1, and T3) and in the NR condition they were in the left frontotemporal and centroparietal regions (electrodes F7, FTC1, T3, C3, TCP1, and P3). In both NR and LER reference site conditions, left-sided electrodes (involving T3) showed

larger t values than did midline or right-sided electrodes. Since our exploratory analyses were consistent with left vs. right group differences, subsequent statistical analyses were reduced to the band of electrodes beginning at T3 and progress- ing contralaterally to include C3, Cz, C4 and T4, as in Faux et al. (1988a) 3

(2) P300 grand-averaged and individual time do- main wave forms and latency measurements

Grand-averaged wave forms are displayed for frontal, lateral, and midline electrode sites in Fig. 2 for the LER condition and in Fig. 3 for the NR condition. In both reference conditions normal controls showed P300 component maxima (300- 400 msec) at the centroparietal midline (Cz and Pz) with lateral placements showing reduced but approximately symmetrical amplitudes. In contrast, schizophrenics showed P300 component amplitude reductions at all sites compared to normal controls, with greater fight- than left-sided P300 component amplitudes. These findings were consistent with the LER and NR topographic maps of Fig. 1.

Fig. 4 shows the evoked potential average from each individual which was used to compute the group grand averages at T3 and T4. The uniformly low amplitudes over the 300-400 msec interval in

3 In the N R condition, TCP1 had a slightly larger t value (t (38) = 4.13) than did T3 (t (38 )= 3.93); these sites accounted for nearly equal portions of the variance (29% and 25%, respectively; each P < 0.0005). For consistency with previous work and with the LER analysis we have used T3 for the ratio and multivariate analyses in the NR condition.

Fig. 1. Topographic distribution of P300 activity and t-SPMs (300-400 msec) in LER and N R conditions. Top, voltage maps: these voltage maps show in both LER and N R conditions that schizophrenics produced an L < R P300 asymmetry, compared to the more symmetrical amplitudes of normal controls. Color-coded voltages are scaled differently for schizophrenics and normal controls to make scalp distributions dearly visible. Voltage scale minima and maxima are the following: LER/cont ro l s = - 3 to 9 ~V; LER/schizophrenics = - 4 to 4 pV; NR/con t ro l s = - 2 to 10/~V; NR/schizophrenics = - 3 to 5/~V. Voltages are displayed at 300, 320, 340, 360, 380, and 400 msec after the presentation of the infrequent auditory stimulus; integrated voltages between 300 and 400 msec are also displayed. Bottom, t-SPMs: SPM scale (in t values) minima and maxima were the following: LER = 0 - 6 . 5 ; NR = 0-3.8. In the LER condition the largest t values were in the left frontotemporal region (F7, FTC1, and T3), with t values ranging from 6.45 to 7.07. The t values of the neighboring electrodes (FP1, F3, C3, TCP1, and T5) were 2.95-5.38, of the midline scalp electrodes were 3.38-4.15, and of the right scalp electrodes were 2.35-4.45. In the N R condition the t-SPM showed the largest t values in the left frontotemporal and centroparietal regions (electrodes F7, FTC1, T3, C3, TCP1, and P3), with t values of 3.84-4.09.

The t values of midline electrodes were 2.54-3.41, and of right scalp electrodes were 2.04-3.46.


LINKED-EARS REFERENCE ELECTRODE - Infrequent Attentive Condition

Fpl Fp2

_ _ Sch izophren ics , N - 2 0

. . . . . . . . . . . Con t ro l s , N=20

F7 F z

V : ......... k%

F8

i i 3.0 pV

T3 Cz • T4

3.0 IJV

Tcpl

- ~ , ~ ~ - i3"0~ v

0 100 200 300 400 500 600 MS

Pz Tcp2

; I I I I J * ~ [ L i ,

0 100 200 300 400 500 600 MS 0 100 200 300 400 500 600 MS

Fig. 2. Grand-averaged wave forms (la condition) for schizophrenics and normal controls referred to LER. Note that schizophrenic versus normal control group differences over 300-400 msec are generally larger at left-sided electrode sites than at homologous sites on the fight. Voltage origins at time zero were determined by a 16 msec prestimulus interval (not shown). Compare with Fig. 3, NR

values.

schizophrenics compared to normals, especially in the LER condition, was notable, as was the slightly greater wave form variability in N R compared with LER, especially for the schizophrenic group.

Scoring of P300 peak component latency for each subject's P300 wave form showed no statistical differences between groups in peak component latency at any electrode site for either the maximum amplitude latency (MAL) or the Woody latency (WL). For example, at Cz, mean WL and mean MAL (values given in parentheses) were: (1) in LER condition, 386.2 (383.8) msec for controls and 388.6 (371.6) msec for schizophrenics (WL:

t = 0.20, P < 0.84; MAL: t = 0.99, P < 0.33), and (2) in N R condition, 377.0 (385.8) msec for con- trois and 380,0 (385.0) msec for schizophrenics (WL: t = 0 . 2 8 , P < 0 . 7 8 ; MAL: t = 0 . 0 7 , P < 0.95). Moreover, there were no statistical differences between WL and MAL measurements within groups.

(3) Analysis of left versus right ampfitude ratios The first statistical measure testing for a left-

fight voltage asymmetry was a simple log ratio of left- versus right-sided electrodes for P300 integrated amplitudes. In Table I negative values of


NOSE REFERENCE ELECTRODE - Infrequent Attentive Condition

Fpl Fp2

f ~ , .

+ " Ti30,v i 3.0 pV

• . : . . / ' ,J . -

Sch izophren ics , N 20

. . . . . . . . . . Cont ro ls , N = 20

F7

~ - . ; ~ 3,0 pV

T3

/ ,.

3.0 ~V

Tcpl / / "\

h I l I I I I 0 100 200 300 400 500 600 MS

Fz ." ..

i / : i • / ............. -. j

Cz .,

" , j

f '",.,

Pz / '..

• ~ i 3.0 uV

I I I 1 I ,J 0 100 200 300 400 500 600 MS

F8

.,, ...

i . . . . _ ; , . . . . . . . . . . . . . . . . . . .

T 4 :.~-.

Tcp2 -

i 3.0 ~V

I i ~ I i i 0 100 200 300 400 500 600 MS

Fig. 3. Grand-averaged wave forms (Ia condition) for schizophrenics and normal controls referred to NR. Note that the general trends in scalp distribution and group differences are the same as those for LER displayed in Fig. 2.

TABLE I

P300 left/right voltage ratios in schizophrenics and normal controls. Figures in parentheses indicate the geometric mean. The probability levels are from a one-sided Mann-Whitney U test, since previous results predicted that schizophrenics would show smaller left versus right voltage ratios than controls.

Ratio Log e geometric mean (geometric mean in parentheses)

Linked-ears reference ' Nose reference

Sz Controls Mann-Whitney S z Controls Mann-Whimey

T 3 / T 4 - 0.13 0.02 P < 0.03 - 0.19 0.01 P < 0.01 (0.74) (1.04) (0.64) (1.03)

C 3 / C 4 - 0 . 0 7 0.00 P < 0.06, ns - 0 . 1 5 0.02 P < 0.005 (0.83) (1.00) (0.78) (1,04)

Note: The appropriate summary statistic for a ratio measurement is the geometric mean, computed by taking the exponential of the mean of the logarithms of the ratios. To prevent division by zero in computation of ratios, a constant value of + 0 . 1 ~V was added to the voltage at each electrode site.


Linked-Ears Reference Electrode

INFREQUENT ATTENTIVE CONDITION

N o s e Reference Electrode

T3 T4 T3 T4

Controls N 20 i 3.0 yv i 3.0 t=V i a.ofv

T3 T4 T3 T4

Schizophrenics N 20

i 3.or v

o x)o 2oo 3oo 40o 5oe emo Ms o loo 2oo 3oo 4oo soo eoo MS

Fig. 4. Superimposed P300 wave forms (Ia condition) for each subject by reference site and temporal scalp electrode site.

o/,v

mean ratios in logarithmic form indicate a left < right amplitude and positive values indicate a left > right amplitude. Schizophrenics showed a clear trend toward left < right P300 componen t amplitudes, whereas normal controls tended to show more symmetrical amplitudes, with left slightly greater than right for T3 versus T4. The N R appeared to have slightly more L -R asymmetry for schizophrenics than did the LER; t he appropria te statistical analyses of L E R / N R topographic differences within groups are reported in section 4 below.

(4) P300 multivariate analyses using Hotelling's T: test

Two types of M A N O V A were performed. The first M A N O V A consisted of a 'wi th in-groups ' analysis to test the effects of reference site on P300 distribution at the scalp, and the second set of analyses focused on 'be tween-group ' differences.

Within-groups analysis. Control and schizophrenic groups were tested separately for the effects of reference site on P300 distribution. Nor- mal controls showed no evidence of a scalp site x reference site interaction (Hotell ing's T 2 equiv-

alent F (4, 16) = 0.58, P > 0.68). Schizophrenics also showed no scalp site × reference site interaction ( F (4, 16) = 1.33, P > 0.30). Paired t tests ( N R - LER) of P300 integrated amplitudes for

I . . - O - _

o 5 / / ~" "~ I 0 " s" '~' Cont ro ls (N = 20) x "., ~,Q

0 ~ O - - 0 = Linked-ears Reference • o o" "o 4 e - - o = Nose Reference

~ 2

F i i

2 3 C 3 C z C 4 T 4

Scalp Electrode Sites

Fig. 5. Mean integrated P300 amplitude (300-400 msec) compared by groups, scalp region, and reference electrode site. Note that normal control amplitudes were symmetrical about the midline (Cz), but that schizophrenics showed an L < R amplitude profile. A statistical analysis of these values is

summarized in Table II.


TABLE II

P300 protected t test comparisons at 5 electrode sites for schizophrenic vs. normal control group: control-schizophrenic group difference (C- S), pooled standard deviation and Hotelling's T contrast values.

Electrode sites

T3 C3 Cz C4 T4

Linked-ears reference Difference C - S 3.03 3.54 3.03 3.11 2.44 Pooled S.D. 1.48 2.15 2.92 2.49 1.76 Hotelling's T 6.45 * 5.20 * 3.38 3.95 * 4.38 *

Nose reference Difference C - S 2.97 3.69 2.85 2.62 2.03 Pooled S.D. 2.39 3.03 3.50 3.01 2.34 Hotelling's T 3.93 * 3.85 * 2.58 2.74 2.75

Note: Asterisked HoteUing's T contrasts indicate C-'S differences which are statistically greater than zero; T contrast criterion for P < 0.05 is 3.74.

each group at each of the 28 scalp electrode sites were not statistically significantly different at any electrode site in either the schizophrenic or control group (t (19) values varied from 0.1 to 0.6 in schizophrenics, and from 0.1 to 0.5 in no rma l controls). Thus there was no statistical evidence to suggest that LER and N R were different in P300

voltage topography for either group. Between-groups analysis. (i) Overall group dif-

ferences. Fig. 5 plots the mean ampl i tude values (by group, electrode site, and reference) which were used in the 'be tween-groups ' M A N O V A analyses for bo th the LER and N R condit ions. Overall P300 ampl i tude differences compar ing schizophrenics and norma l controls were statistically significant for bo th LER (Hotel l ing 's T 2

equivalent F (5, 34) = 9.44, P < 0.05) and N R ( F (5, 34) = 5.23, P < 0.05). Fig. 5 also shows a rela- tively symmetrical L = R voltage d is t r ibut ion for no rma l controls, while schizophrenics showed an L < R pat tern, suggesting the presence of a group × scalp electrode site interaction.

(ii) Group amplitude differences at each scalp electrode site ('protected" t contrasts). Table II shows that protected t contrasts compar ing individual scalp regions under bo th reference condi- t ions were maximal at T3, a f inding consis tent with the t-SPM topography of Fig. 1. Wi th respect to reference site, t contrasts were un i formly smaller for N R than for LER, a f inding related to the slightly higher wave form variabil i ty of N R com-

pared to LER, as shown by the pooled s tandard deviat ions (Table II) and wave form va r i ab i l i ty (Fig. 4, see also below). However, F ( m a x / m i n ) tests showed that N R and L E R variances were only statistically different at T3 ( F (19, 19) = 2.61,

P < 0.05).

1.0 g

g .;'5

~.50

~..25

E 0

A

~ l l ~ - s2a ~ ~ O

i i i i T3 C3 Cz C4 T4


B

1.0 ~

,75 - i t

.50 i II

/ • = Nose ref.

£ o r i i 1

T3 C3 CZ C4 T4


Fig. 6. Normalized P300 integrated amplitudes (300-400 msec) for LER and NR conditions. Data were normalized according to the following formula: Nijk=(Vijk-Min)/(Max-Min), where N equals a normalized voltage for the kth person, the jth electrode and the ith group; V equals the raw integrated voltages for each level of i, j, and k; Min equals the smallest mean voltage for each i and j; Max equals the largest mean voltage for each i and j. Group×electrode site interaction effects were tested using a MANOVA profile analysis of parallelism. Normal controls (panel A) showed an L = R symmetry in contrast to schizophrenics (panel B) who showed L < R

amplitudes.

388 S.F. FAUX ET A L

Distributions of normalized amplitudes for LER and NR were nearly identical within the normal control group (Fig. 6A) and within the schizophrenic group (Fig. 6B), but the schizophrenic groups' L < R topography was in contrast to the L = R topography of normals. Group differences in normalized topographic distribution tested by profile analysis of parallelism were statistically significant for both LER ( F (4, 35)= 3.72, P < 0.05) and NR conditions ( F (4, 35) = 3.89, P < 0.05), indicating the presence of a strong group x electrode site interaction. These results provide statistical confirmation that the topographic differences between schizophrenics and normal controls that are apparent in color maps (Fig. 1), grand-averaged wave forms (Figs. 2 and 3), and in ratio data are approximately the same for both reference site conditions.

(5) P300 separation of individual in the schizophrenic and control groups

Fig. 7 (left panel) shows that the LER correctly categorized 18/20 individuals in both groups using the previously established 2 t~V criterion level 4. In the right panel (Fig. 7) we have displayed the NR condition using the TCP1 electrode (adjacent to T3, and slightly posterior and medial), since, as noted in footnote 3, this electrode site had the largest t value. Comparison of LER and N R shows that the visual separation between groups is worse in the LER condition. This visual impres- sion is bourne out by the magnitude of respective t values (LER: t (38) = 6.45; NR: t (38) = 4.13). This difference in individual categorization is con- sonant with the slightly greater variability of the NR condition shown in Table II.

4 For comparison with previous studies, it is important to note that these data differ somewhat from those of Faux et al. (1988b), in which the P3OOg subtraction wave form [(infrequent attentive - frequent attentive) - (infrequent inattentive - frequent inattentive)] enhanced the statistical separations between schizophrenics and normal controls. In the present study, the unsubtractod P300 wave form to the target st imulus (Ia wave form) produced greater statistical separation between groups and greater topographic asymmetries than the P300g.

o o 9

o 7

o

3

0

l O • ÷

I F o • . . .

0 ~ • ~ ÷

i i i i i

Controls Schlzoohremcs Controls Schizophrenics

Linked-ears Ref. Nose Ref.

Fig. 7. Mean integrated P300 amplitude (300-400 msec) of individual members of the schizophrenic and normal control group. Left panel, LER condition values at the T3 electrode site and with the criterion level value of 2 gV (same values as in previous LER study, Faux et al. 1988b). Note the correct separation of all but 2 individuals in each group (90% accuracy). For the NR condition, groups overlapped to a greater extent, even though TCP1 had the largest t value in that

condition.

Discussion

These results confirm previous findings of P300 asymmetry in schizophrenia (Morstyn et al. 1983; Faux et al. 1988a) and suggest that the LER does not greatly influence the detection and measurement of P300 asymmetry relative to the NR. These results do not suggest that past criticism of the LER is unfounded, but they do suggest that in some studies the possibility of an experimental confound caused by the LER may be a greater theoretical than practical concern.

Within groups, direct comparison of LER and NR P300 integrated amplitudes using repeated- measures MANOVA (conserving type I errors) and paired t statistic maps (conserving type II errors) showed that there was no statistical difference between LER and NR scalp topography. Between-group differences also were similar for LER and NR. Both reference sites obtained the following results: (1) t statistic mapping showed that left temporal scalp regions maximized the statistical separation between groups. (2) Analysis of left-right log ratios showed a P300 asymmetry in schizophrenics' amplitudes, and MANOVA profile analysis of P300 'normalized' topography validated the presence of a group × scalp region


interaction. (3) At the region of maximal statistical separation in LER, schizophrenics were dis- criminated from normals with 90% accuracy using a 2 #V criterion. Since the NR resulted in less group discriminative power than the LER, it may be less preferable in these studies than the LER. (4) No statistical differences between groups were found in P300 peak component latency.

Additionally, these results extended the gener- alizability of our previous studies in that they were acquired from: (1) patients who were a decade older than patients previously studied; (2) patients in a VA hospital, as opposed to a state hospital setting; and (3) completely different EEG instru- mentation with an extended electrode montage. These data thus strengthen the argument that P300 asymmetries in schizophrenia are robust to a variety of experimental parameters.

There is suggestive evidence that the P300 left temporal feature may be related to lateraliTed brain limbic system abnormalities in schizophrenia. McCarley et al. (1989) report a high correlation of P300 amplitude with left sylvian fissure enlargement on CT scans, a presumptive indicator of left temporal lobe tissue loss. Depth recordings have indicated major generator loci for the P300 in temporal lobe limbic structures (reviewed in Halgren et al. 1986).

Other data support the presence of temporal lobe limbic system abnormalities in schizophrenia, including several recent morphometric studies (Bogerts et al. 1985; Brown et al. 1986; Crow 1986), with Brown et al. (1986) reporting laterali- zation and greater left-sided pathology. Converg- ing lines of evidence suggesting left hemisphere abnormality in schizophrenia have come from measures as diverse as cerebral blood flow (Gur et al. 1985) and reaction time to a visual spatial task (Posner et al. 1988).

Despite the statistical strength of the P300 findings, several cautions must be emphasized regard- ing their clinical and electrophysiological interpre- tation.

(1) We have not yet compared data in the same patients on and off medication, nor have we tested patients without previous exposure to neuro- leptics. However, current evidence suggests that the P300 abnormalities we have observed in

schizophrenia are not an effect of neuroleptic medication. We have now observed similar topographic asymmetries (P < 0.05) in 10 unmedicated (> 2 weeks) schizophrenics compared with 8 age- matched normal controls (Faux et al. 1989; see also Blackwood et al. 1987; Brecher et al. 1987 for midline electrode data).

(2) We have also not compared the P300 topography of schizophrenics with other psychopathological populations. Thus, the specificity of our P300 findings remains a question.

(3) ERP scalp distributions cannot unambigu- ously localize underlying source generators. How- ever, the presence of an asymmetric P300 scalp distribution is consistent with the notion that a pathological process has modified, perhaps uni- laterally, the characteristics of source generators in schizophrenia. These characteristics might include (i) unilateral changes in dipole orientation, (ii) unilateral tissue reduction at the source, (iii) the activation of different or fewer generators, and/or (iv) changes in the synchronization of normally coactive generators. Thus, we believe that the presence of a P300 asymmetry in schizophrenics implies a pathological process, even if that process cannot be specifically localized by ERP techniques alone.

In sum, these data suggest: (1) LER does not influence the measurement of P300 asymmetries in schizophrenia; (2) P300 asymmetries in schizophrenia are a robust finding, now having been seen in 3 different studies; (3) the LER is prob- ably more suitable than the NR in these studies, since it maximized schizophrenic vs. normal control group differences; and (4) the P300 differences found in previous studies using subjects with a mean age in late twenties also hold for subjects who are a decade older and for a different montage and topographic mapping technology.

The authors gratefully acknowledge the technical assistance of Erik Mondrow and Virginia Penhune, and the administra- tive assistance of Marie Fairbanks and Carla Kohberger.

Supported by the Veterans Administration, NIMH 40.799 the Commonwealth of Massachusetts Research Center, NIMH Research Training Grant T32MH16259 (PGN), NIMH Re- search Scientist Development Award MHK-K01-MH00746-02 (MES), and the National Alliance for Research on Schizophre- nia and Depression (NARSAD). Dr. McCarley is a Medical Investigator of the VA Medical Research Service.

390

References

American Psychiatric Association, Committee on Nomencla- ture and Statistics. Diagnostic and Statistical Manual of Mental Disorders (3rd Edition-Revised). American Psychi- attic Association, Washington, DC, 1987.

Bassuk, E.L., Schoonover, S.C. and Gelenberg, A.J. The Practi- tioner's Guide to Psychoactive Drugs. Plenum Medical Book Co., New York, 1983.

Beaumont, J.G. The EEG and task performance: a tutorial review. In: A.W.K. Gaillard and W. Ritter (Eds.), Tutorials in Event Related Potential Research: Endogenous Compo- nents. North Holland, Amsterdam, 1983.

Blackwood, D.H.R., Whalley, L.J., Christie, J.E., Blackburn, I.M., St. Clair, D.M. and McInnes, A. Changes in auditory P3 event-related potential in schizophrenia and depression. Br. J. Psychiat., 1987, 150: 154-160.

Bogerts, B., Meertz, E. and Schonfeldt-Bausch, R. Basal ganglia and limbic system pathology in schizophrenia: a morphometric study of brain volume and shrinkage. Arch. Gen. Psychiat., 1985, 42: 784-790.

Brecher, M., Porjesz, B. and Begleiter, H. Late positive component amplitude in schizophrenics and alcoholics in two different paradigms. Biol. Psychiat., 1987, 22: 848-856.

Brown, R., Colter, N., Nicholas-CorseUis, J.A.N., Crow, T.J., Frith, C.D., Jagoe, R., Johnstone, E.C. and Marsh, L. Postmortem evidence of structural brain changes in schizophrenia: differences in brain weight, temporal horn area, and parahippocampal gyrus compared with affective dis- order. Arch. Gen. Psychiat., 1986, 43: 36-42.

Crow, T.J. Left brain, retrotransposons, and schizophrenia. Br. J. Med., 1986, 293: 3-4.

Davidson, R.J. EEG measures of cerebral asymmetry: concep- tual and methodological issues. Int. J. Neurosci., 1988, 39: 71-89.

Dully, F.H. Topographic display of evoked potentials: clinical applications of brain electrical activity mapping (BEAM). Ann. NY Acad. Sci., 1982, 388: 183-196.

Duffy, F.H., Bartels, P.H. and Burchfiel, J.L. Significance probability mapping: an aid in the topographic analysis of brain electrical activity. Electroenceph. clin. Neurophysiol., 1981, 51: 455-462.

Faux, S.F., Torello, M., McCarley, R., Shenton, M.E. and Duffy, F.H. P300 in schizophrenia: confirmation and statistical validation of temporal region deficit in P300 topography. Biol. Psychiat., 1988a, 23: 776-790.

Faux, S.F., Shenton, M.E., McCarley, R.W., Torello, M.W. and Duffy, F.H. Differentiation of schizophrenics and normal controls is enhanced by the Goodin subtraction procedure. Int. J. Neurosci., 1988b, 39: 117-135.

Faux, S.F., Nestor, P., McCarley, R.W., Shenton, M.E., Horvath, T. and Davis, K. P300 asymmetries in unmedicated schizophrenics. Presented at EPIC IX, The Nether- lands, May 1989.

Feighner, J.P., Robins, E., Guze, S., Woodruff, R.A., Winokur, G. and Monuz, R. Diagnostic criteria used in psychiatric research. Arch. Gen. Psychiat., 1972, 26: 57-63.

S.F. FAUX ET AL.

Folstein, M.F., Folstein, S.E. and McHugh, P.R. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J. Psychiat. Res., 1975, 12:189 198.

Gur, R.E., Gur, R.C., Skolnick, B.E., Caroff, S., Obrist, W.D., Resnick, S. and Reivich, M. Brain function in psychiatric disorders. III. Regional cerebral blood flow in unmedicated schizophrenics. Arch. Gen. Psychiat., 1985, 42: 329-334.

Halgren, E., Stapleton, J.M., Smith, M. and Altafullah, I. Generators of the human scalp P300. In: R. Cracco and I. Bodis-Wollner (Eds.), Frontiers of Clinical Science: Evoked Potentials. 1986: 269-284.

Kahn, E.M., Weiner, R.D., Brenner, R.P. and Coppola, R Topographic maps of brain electrical activity: pitfalls and precautions. Biol. Psychiat., 1988, 23: 628-636.

McCarley, R.W., Faux, S.F., Shenton, M.E., LeMay, M., Cane, M., Ballinger, R. and Duffy, F.H. CT abnormalities in schizophrenia: a preliminary study of their correlations with P300/P200 electrophysiological features and positive/ negative symptoms. Arch. Gen. Psychiat., 1989, 46: 699- 708.

McCarthy, G. and Wood, C.C. Scalp distributions of event-related potentials: an ambiguity associated with analysis of variance models. Eleetroenceph. clin. Neurophysiol., 1985, 62: 203-208.

Morstyn, R., Duffy, F.H. and McCarley, R.W. Altered P300 topography in schizophrenia. Arch. Gen. Psychiat., 1983, 62: 203-208.

Mowery, G.L. and Bennett, A.E. Some technical notes on monopolar and bipolar recordings. Electroenceph. clin. Neurophysiol., 1957, 9: 377.

Nunez, P. Electric Fields of the Brain: the Neurophysics of EEG. Oxford University Press, Oxford, 1981.

Posner, M.I., Early, T.S., Reiman, E., Pardo, P.J. and Dhawan, M. Asymmetries in hemispheric control of attention in schizophrenia. Arch. Gen. Psychiat., 1988, 45: 814-821.

Rugg, M.D. The relationship between evoked potentials and lateral asymmetries of processing. In: A.W.K. Gaillard and W. Ritter (Eds.), Tutorials in Event Related Potential Re- search: Endogenous Components. North Holland, Amster- dam, 1983.

Scherg, M. and Von Cramon, P. Two bilateral sources of late AEP as identified by a spatial-temporal dipole model. Electroenceph. clin. Neurophysiol., 1985, 62: 32-44.

Shader, R.I. Manual of Psychatric Therapeutics. Little, Brown and Co., Boston, MA, 1975.

Shenton, M.E., Faux, S.F., McCarley, R.W., Baltinger, R., Coleman, M., Torello, M.W. and Duffy, F.H. Correlations between abnormal auditory P300 topography and positive symptoms in schizophrenia: a preliminary report. Biol. Psychiat., 1989, 25: 19-26.

Spitzer, R.L. and Endicott, J. Schedule for Affective Disorders and Schizophrenia. Biometric Research, New York State Psychiatric Institute, New York, 1978.

Spitzer, R.L., Endicott, J. and Robins, E. Research Diagnostic Criteria (RDC) for a Selected Group of Functional Dis- orders. Biometric Research, New York State Psychiatric Institute, New York, 1978.


Stephenson, W.A. and Gibbs, F.A. A balanced noncephalic reference electrode. Electroenceph. clin. Neurophysiol., 1951, 3: 237-240.

Vaughan, Jr., H.G. and Arezzo, J.C. The neural basis of event- related potentials. In: T.W. Picton (Ed.), Handbook of Encephalography and Clinical Neurophysiology (revised series). Vol. 3. Human Event-Related Potentials. Elsevier, Amsterdam, 1988: 45-96.

Vaughan, Jr., H.G. and Ritter, W. The sources of auditory evoked responses recorded from the human scalp. Electro- enceph, din. Neurophysiol., 1970, 28: 360-367.

Vaughan, Jr., H.G., Ritter, W. and Simson, R. Topographic analysis of auditory event-related potentials. In: H.H. Kornhuber and L. Deecke (Eds.), Motivation, Motor and

Sensory Processes of the Brain: Electrical Potentials, Be- havior and Clinical Use. Progr. in Brain Res., Vol. 54. Elsevier, Amsterdam, 1980: 279-290.

Wolpaw, J.R. and Wood, C.C. Scalp distribution of human auditory evoked potentials. I. Evaluation of reference electrode sites. Electroenceph. clin. Nenrophysiol., 1982, 54: 15-24.

Wood, C.C. and Wolpaw, J.R. Scalp distribution of human auditory evoked potentials. II. Evidence for multiple sources and involvement of auditory cortex. Electroenceph. clin. Neurophysiol., 1982, 54: 25-38.

Woody, C.D. Characterization of an adaptive filter for the analysis of variable latency neuroelectric signals. Med. Biol. Eng. Comput., 1967, 5: 539-553.

preservation of p300 event-related potential topographic asymmetries in schizol ... -...

Documents