Abnormal EEG Slow Activity in Left Temporal Areas in Senile Dementia of the Alzheimer Type

Resting 32-channel topographical measures of EEG slow activity were compared in 12 elderly controls and 12 patients with senile dementia of the Alt.heimer type. The patients had higher amplitude delta and theta than controls, especially in the left temporal regions. This greater amount of low frequency EEG activity in the left temporal area is consistent with recent EEG, neuropsychological assessment, and positron emission tomography findings in SDAT patients. Five patients with mild-to-moderate dementia (as determined by the Fa/stein Mini-Mental State scale) primarily exhibited focal, abnormal slow activity in the left temporal regions. Seven patients with severe dementia exhibited increased slow activity across the head, which was still most abnormal in the left temporal regions.

J-\. in elderly patients was described in the aging literature in the 1950s ( 1-3). Other than the description of " diffuse slowing," there was no attempt to define the topographical properties of this abnormality in various subtypes of dementia. Since that time, EEG slow activity in Senile Dementia of the Alzheimer Type (SDA T) has been reported by a large number of studies (4--1 I). However, there has been little agreement in this literature on the topographical distribution . This might be related to the small number of electrodes and the lack of a standardized reference electrode across these studies.
M<;>st recently, three independent studies have reported that abnormal EEG slow activity (delta and theta frequency bands) in SDAT is greatest in the left temporal areas (12)(13)(14). These recent EEG findings overlap with the older positron emission tomography (PET) findings of temporal and parietal hypometabolic activity ( l 5, 16) and the more recent PET findings of predominantly left hemispheric hypometabolic activity in SDAT ( 17,18).
The first purpose of this study was to replicate the previous EEG left temporal findings with an independent group of SDAT patients and controls. The second purpose was to relate the left temporal EEG slow activity measure to the severity of dementia. The third purpose was to assess this left temporal EEG slow activity measure as an objective diagnostic instrument for early SDAT patients. METHOD Subjects. -The patients were 12 right-handed individuals (6 male, 6 female, mean age 72. 7, SD = 7. I) who had been diagnosed as having SDAT by a physician (A.S.) without knowledge of the EEG results. The diagnostic criteria were those recommended by the Work Group on the Diagnosis of Alzheimer's Disease ( 19). They included (a) a normal magnetic resonance image scan that showed no evidence of small vessel infarcts or other structurai abnormalities of the brain; (b) normal blood chemistries with particular attention to vitamin 8 12 , folic acid, BUN, and hemoglobin; (c) evidence or involvement of more than one cognitive deficit that would suggest the pathology to be more diffuse; and (d) a history of gradual progression of symptoms.
Seven of these patients (3 male, 4 female, mean age 71.0, SD = 8.9) were classified as "severely demented" on the basis of having a Folstein Mini-Mental State Score (MMSS) (20) of 7 or below. Five of these patients (3 male, 2 female, mean age 75.2, SD = 2.5) were classified as "mild-tomoderately demented" on the basis of having a MMSS in the range of 16-25. The duration of illness was estimated by asking the patient's caretaker about when the earliest symptoms appeared. The mean duration of illness was 7.8 years for the severely demented group and 5.6 years for the mildto-moderately demented group. Twelve control subjects (6 male, 6 female, mean age 71.4, SD = 4.6) were selected to participate. Six of these controls were spouses of the patients and the other six were recruited from senior centers in Orange County, CA. The MMSS range for these control subjects was 25-30. All patients and subjects were screened to exclude those with psychiatric and other neurological disorders, and with clinically significant cardiovascular disorders. In addition, these individuals were screened to exclude those with recent drug or alcohol dependence or a history of head injury. All of the subjects had been free of CNS active medication for at least 7 days before testing. In addition, all of the subjects had abstained from the use of caffeinated beverages on the day of the testing.
Procedure. -All recordings were _made in a soundattenuated and darkened room with subjects reclining at a 45° angle and resting their heads so as to minimize neck muscle tension. They were instructed to keep their eyes closed throughout the session and remain as relaxed as possible. EEGs were recorded in three 38-second periods. If the record showed evidence of drowsiness or excessive movement, the recording period was interrupted and the Ml 45 "M146 RICE ET AL. subject was asked to maintain a relaxed state of wakefulness. Electrodes were placed on 32 positions over the scalp surface, using the International I 0-20 system plus five extra · leads on each hemisphere and two extra midline positions with linked ears as reference ( Figure 1). These 32 channels were converted to an average reference record ing by computation after analog-to-digital conversion. All analyses were performed on average reference data. Data were collected with an on-line computer system. EEG activity for spectral analysis was amplified with a 3dB bandpass of 0.5-50 Hz and digitized for each channel at 118 Hz and recorded in I. 75 sec. blocks (207 points). All 32 leads for each epoch were visually inspected for eye blinks, eye movements, and other movement artifacts as defined by previous EEG topographical analysis (21). Any 1.75-second epoch containing artifacts was eliminated from further analysis.
Typically, 20-30 1.75 second epochs (mean = 27 .41, SD = l 0.40) were included in the analysis per subject. The amplifiers were calibrated by recording a 10 Hz standard signal through all chan nels and determining the calibration factor for each channel. Before analysis, each channel was thus proportionally adjusted. Piecewise quadratic interpolation was performed on the original 207-point, 1.75-second epochs to transform them into 256-point, I. 75-second epochs for compatibility with our spectral analysis program . A window function consisting of a l 0% cosine taper was obtained by weighting the 12 points at either end of each 256-point, 1. 75second epoch by a cosine bell. A standard fast Fourier transform was applied to each of the artifact-free, I. 75second epochs in a recording and the power estimates were computed at 0.57 Hz steps. The transform yielded a value representing the average magnitude, expressed in microvolts (square root of power). This was calculated as the square root of the sums of squares across the 0. 57 Hz steps and yields the sine wave equivalent in microvolts. The bandwidths for each frequency were: delta, 0 .57-3.99 cps; theta, 4.57-7.41 cps; alpha, 7.98-13. 11 cps; beta, 13.68-19.95 cps. Continuous surface density maps of the entire scalp surface were created from these values, using a 4-nearest neighbor interpolation algorithm (22).
Differences between the demented and control subjects were assessed for each frequency band separately using 3way ANOYA (BMDP2Y) with repeated measures (23). Diagnostic category was the grouping variable, and Hemisphericity and Longitudinality were repeated factors. These factors were obtained by sampling 25 electrodes (Table I) out of the original 32 and forming a 5 (Longitudinality) x 5 (Hemisphericity) grid. A technical error had resulted in the loss of 16 channels of information from one of the mild-to- Thus , only 10 patients could be included in the ANOV A.
However, for 15 of the channels there were 12 patients, and for 17 of the channels there were I I patients who could be included in the r-tests described below. Huynh-Feldt reduced degrees of freedom were used to avoid Type I errors caused by falsely inflated d.f. in repeated measures ANOVA.

DELTA Z-TRANSFORMATION
SOAT T-Test .. cation ANOV A interactions were found, the ANOV A was repeated using values normalized with a z-transformation to correct for a possible multiplicative effect produced by differences in source strength (24) and to eliminate differences due to overall reduction in EEG power. This was done by calculating the mean and SD across the values for the 32 leads and then reexpressing each lead as (lead value-mean)/ standard deviation. The z-transformed frequency bands yielding significant Hemisphericity x Group or Longitudinality x Group interactions were selected for calculation of group means and c-tests at each of the 32 leads.

RESULTS
Control vs SDAT effects. -The SDAT patients had a larger amount of delta activity than the controls all over the head, with the differences being largest in the left temporal areas. In addition, the SDAT patients had a larger amount of theta activity than the controls all over the head, but the differences were only slightly larger in the left temporal area. The ANOVA tests of the absolute amplitude in the four frequency bands for the main effect of diagnosis confirm that patientS had significantly more delta (F = 30.16, df = 1,20, p < .00001) and theta (F = 34.92, df = 1,20, p < .00001) than controls, while there were no significant differences for alpha and beta. The patients had more delta than the controls in the left hemisphere relative to the right (Hemisphere x Diagnosis interaction, F = 9.71, df =  of the /-tests on the individual leads were significant. The greater abnormal left temporal slow activity in the SDAT patients was further supported by the ANOV A in the topographically normalized data. Significant Hemisphere x Diagnosis x Longitudinal interactions for both the normalized delta (F = 2. 76, df = 16,320, p < .01) and normalized theta (F = 1.06, df = 16,320, p < .01) were found. Figure 2 demonstrates the topographically normalized amplitude and t-test maps for the delta frequency band. It shows that the largest effects were entirely concentrated in the left temporal regions. Figure 3 shows the topographically normalized amplitude and t-test maps for the theta frequency band. It shows that the effects were only slightly larger in the left temporal area.
To estimate the most important frequency range, we performed t-tests on the normalized EEG values at the four left temporal leads across all the individual 0.57 Hz steps in frequency within the delta and theta range. The strongest effect occurs around 1. 14 Hz (Table 2).
Also, there was a significant negative correlation with MMSS at all leads for absolute delta. These correlations

THETA Z-TRANSFORMATION
SOAT T-Test Figure 3 . Left: mean normalized topographic maps of theta in elderly controls and SDA T patients. Scale is in standard deviation units. Right: maps of 32 exploratory Hest comparisons between elderly controls and ranged from (r = -.59, p = .05) at TCP! to (r = -.85, p < .01) at F3. However, there were no significant correlations between the normalized delta band EEG leads and the MMSS measures. In the theta band, the absolute theta at the EEG leads FP2 (r = -.61, p < .05) and C4 (r = -·.S8, p < .05) were significantly correlated with the MMSS measure. When the normalized theta measure was correlated with the MMSS, there were significant positive correlations at Cz (r = .63, p < .05), T3 (r = .76, p < .01), TS (r = .64, p <.OS) . and TT ! (r = .84, p < .01).
A relationship of EEG delta to the severity of dementia was also illustrated by a comparison of the number of abnormal EEG leads in the moderately demented group vs the severely demented group. An abnormal lead was defined as one where the value at a lead is greater than two standard deviations above the mean value at that lead for the control group distribution. For this absolute delta measure of abnormality , all five of the moderately demented patients had six or fewer abnormal leads, whereas all seven of the severely demented patients had 16 or more abnormal leads()( = 12.99, df = I, p < .01). In fact, all of the severely demented patients had 30 or more abnormal leads, but 16 was chosen for the upper cutoff because of the 16 missing channels in the one of the moderately demented patients. In all five of the moderately demented patients, at least one anterior left temporal lead (T3 or TT!) was included as an abnormal lead. As a group, the mean activity for the moderately demented patients was only abnormal in four (T3, TT!, TS, FF) of the 32 leads, and three of these were left temporal leads. As a group, the mean activity for the severely demented patients was abnormal in 3 I of the 32 leads; FC was the only normal lead.
For the absolute theta f!l~asure, the chi-square test of the number of patients in each group which had 6 or fewer abnormal leads vs 16 or more abnormal leads was not significant (X 2 = 3.33, df = I, p > .05). As a group, the mean activity for the moderately demented patients was abnormal in a total of 9 leads, and 3 of these (T3, TCP 1, and TT 1) were left temporal. As a group, tl"'c mean activity for the severely demented group was abnormal in 24 leads. These results suggest that the theta abnormalities showed much less .di  and mild-to-moderately demented patients (below). Black is p < .05.

DELTA Z-TRANSFORMATION
I-tailed, patients higher than elderly controls. of the trend which was observed for delta, where moderately demented patients had a predominantly focal left temporal abnonnality pattern in comparison to severely demented patients.
In a further effort to get a view of where the Largest abnormalities were for both the moderately demented and the severely demented groups, r-tests were performed on the topographically normalized delta and theta band data. Because of the SDA T vs Control delta effects in the previous literature, one-tailed t-tests were done on the delta band data, whereas only exploratory (two-tailed) t-tests were done on the theta band data. Both the severe and the mild-to-moder· ate groups showed the greatest concentration of topographically nonnalized delta effects in the left temporal areas (Figure 4). However, this normalized delta effect was much more restricted to the left temporal area in the mild-tomoderate group, where the left temporal leads TT 1 , FTC 1, T3, and TCPl and the right temporal lead TCP2 were all significant . The normalized delta effect was not as restricted to the left temporal area in the severely demented patients, where the left temporal leads Tri , FTC! , and TCPI as well as P3, PO 1, and FTC2 were significant. In the theta comparisons, only the mild-to-moderate group showed a predominantly left-temporal abnonnality pattern ( Figure 5).
As a final illustration of the greater left temporal effect for delta, Figure 6 presents a scatter plot of the absolute delta activity at electrodes T3 and Cz in the patients and the controls. There is no overlap between the patients and the controls at T3, while there is some overlap at Cz. All of the patients that overlapped with the controls at Cz were in the mild-to-moderately demented group. DISCUSSION This is now the fourth recent study to suggest that abnormal EEG slow activity in SDA T patients is greatest in the left temporal areas (12)(13)(14). While this study is based upon a relatively small number of patients, it does suggest that this left temporal slow activity effect might be related to the severity of the disease process. Patients with mild-tomoderate dementia primarily exhibited a focal abnormal left temporal slow activity pattern , whereas patients with severe dementia exhibited a more diffuse slow activity pattern that was most abnormal in the left temporal regions. These effects were particularly clear in the delta frequency band.
While the abnormal left temporal EEG slow activity effect is consistent with recent PET reports of greater left hemispheric dysfunction in SDAT patients, a relationship with the severity of SDA Twas not found in the Loewenstein et al. PET study ( 18). One explanation for this discrepancy could be the different severity range of the patients in these studies. The Loewenstein et al. study had a much more restricted MMSS range without any patients in the severely demented group range (0-7) of this present study.
These data suggest that focal, left anterior temporal EEG slow activity might be a sensitive indicator of the milder stages of SDA T . At the T3 electrode, there was no overlap between the patients and the controls in the distribution of absolute delta activity ( Figure 6), while there was overlap at Cz. All of the overlap at Cz was due to the mild-tomoderately demented patients. This could imply that patients in the milder stages of SDAT might have much more focal left temporal slow activity with a relative sparing of other areas. The diagnostic implications of these data are tempered presently by the small number of patients, along with the lack of confirmatory neuropathological information on these patients.
The diagnostic implications of these data are also complicated by the fact that EEG slow activity is not specific to SDAT. Asymmetric EEG slow activity has been reported to be prevalent in the elderly in qualitative EEG studies of cerebrovascular disease (6,25). However, the one qualitative study (25) which documented the locality of these abnormalities suggests that they are more right temporal and more diffuse than those of the mild-to-moderate SDAT patients of the present study. EEG slow activity in multiinfarct patients has also been reported by two recent quantitative EEG studies which looked at both multi-infarct and Alzheimer-type dementia patients (12,14). Both of these recent studies used far fewer electrodes than the present study and, therefore, might not have been sensitive to a more focal pattern of left temporal EEG slow activity which could be more specifically related to the milder forms of SDA T . In spite of this, both of these studies report that their multiinfarct patients had more diffuse slow activity all over the head than their Alzheimer patients.
While a number of qualitative EEG studies report that focal left temporal EEG slow activity is found in a significant number of normal elderly individuals (4 ,25,26,27), the present data indicate that there are definite quantitative differences between SDAT patients and norinal elderly controls in the degree of focal left temporal slow activity. These present findings are consistent with the neuropathological observation of marked abnormalities of the temporal areas in Alzheimer's disease (28,29). It should be remembered that along with increased numbers of senile plaques and tangles: temporal lobe atrophy was also first suggested to be a neuropathological marker for senile dementia of the Alzheimer type (29). ..

T3 SOAT
ELDERLY CONTROLS VS SDAT PATIENTS Figure 6. Scatter plot demonstrating differences between elderly controls and SDA T patients in absolute delta . There was no overlap between the distributions of these two groups at T3. but there was overlap al Cz. All of the overlap at Cz was due lo patients who were mild-lo-moderately demented.
These findings also raise the possibility that the focal left temporal slow EEG activity often reported to be prevalent in normal elderly ind ividuals (4,(25)(26)(27) could be an early manifestation of Alzheimer's disease. Word-naming deficits are prevalent in the milder forms of clinically diagnosed Alzheimer's disease (30)(31)(32)(33) and word-naming deficits have also been related to severe left temporal slow EEG activity in normal elderly individuals (27) . However, memory deficits have never been associated with left temporal slow EEG abnormalities in the normal elderly. Research is needed which addresses the possibility of Alzheimer-like cognitive deficits in the normal elderly with left temporal slow EEG abnormalities.

ACKNOWLEDOMENTS
This work was funded, in part, by a project award from the Fidia Pharmaceutical Corporation, through the support of the UCI Brain Imaging CenterCommillee. and through the support of the John D. and Catherine T. MacArthur Foundation. Dr. Rice was suppo1•ed by a National Research Service Award (AG-05419) from the National institute on Aging under the sponsorship of Dr. Buchsbaum.
Address correspondence to Dr. Daniel M . Rice, Department of Psychia-