Quantitative Analysis of EEG in Boys with Attention-deficit-hyperactivity Disorder: Controlled Study with Clinical Imp cations C h r i s t o p h e r A. M a n n , P h D * , Joel E Lubar, P h D * , A n d r e w W. Z i m m e r m a n , MD*:!:, C h r i s t o p h e r A. Miller, MDt*, a n d R o b e r t A. M u e n c h e n , MS~
Sixteen-channel topographic brain mapping of electroencephalograms of 25 right-handed males, 9-12 years of age, with attention-deficit-hyperactivity disorder revealed increased theta (4-7.75 Hz) and decreased beta 1 (12.75-21 Hz) when compared with 27 controls matched for age and grade level. The differences were greater when patients were tested for reading and drawing skills, but were decreased when they were at rest during visual f'Lxation. Although the differences in patients with attention-deficit-hyperactivity disorder were generalized, increased theta was more prominent in frontal regions, while beta 1 was significantly decreased in temporal regions. Principal component analysis was used to combine the variables into 2 components which accounted for 82% of the total variance. A discriminant function analysis using these components was able to predict group membership for attention-deficit-hyperactivity disorder patients 80% of the time and 74% for controls. These findings support the use of topographic electroencephalography for further elucidation of the neurophysiology of attention-deficithyperactivity disorder. Mann CA, Lubar JF, Zimmerman AW, Miller CA, Muenchen RA. Quantitative analysis of EEG in boys with attention-deficit-hyperactivity disorder: Controlled study with clinical implications. Pediatr Neurol 1992;8:30-6.
Introduction Attention-deficit-hyperactivity disorder (ADHD) is a common disorder in school children with uncertain and probably multiple etiologies. It is characterized clinically by decreased attention span, impulsivity and, for some children, increased motor activity. The broad nature of diagnostic criteria outlined in DSM III-R has resulted in a heterogenous population of children being classified with this disorder [1].
From the Departmentsof *Psychology,tPediatrics (Neurology),and §Computing Center; Universityof Tennessee; and :~Departmentof Medicine; EastTennesseeChildren's Hospital; Knoxville,Tennessee.
30 PEDIATRICNEUROLOGY Vol.8 No. 1
Among the investigational approaches to ADHD that have been used, the analog electroencephalogram (EEG) has been reported to demonstrate increased background slow activity for chronologic age, in addition to occasional paroxysmal activity [2-5]; however, no specific EEG patterns have been identified and this technique has not been regarded as generally useful for the evaluation of children with ADHD [6]. Evoked potential (EP) differences have also been reported for children with ADHD. In a previous study, we reported differences between normal, learning-disabled, and gifted children based upon an auditory evoked potential task [7]. The results replicated previous studies by demonstrating that children with ADHD and reading disabilities had a suppressed P300 component compared with matched controls in response to a target auditory stimulus [8,9]. In addition, they made more response errors to target stimuli. The P300 component is believed to represent the interpretation of stimulus meaning, while the P200 represents stimulus detection [ 10]. A newer technique, computerized power spectral analysis (PSA), permits topographic representation and statistical analysis of EEG for individual and group comparisons. Previous studies using PSA confirmed findings from analog EEG studies of excessive theta and lack of alpha attenuation in children with deficient attention in the absence of conduct disorders [1 t-14]. These findings suggest that underarousal or hypoactivation of the neocortex is present in children with ADHD which may lend support to postulated noradrenergic involvement in this disorder [l 5]. We hypothesized that children with ADHD may have distinctive findings on topographic PSA-EEG which can be used to distinguish groups with this disorder on a statistical basis from matched controls. Our purpose was to study a carefully selected group of boys diagnosed with attention deficit but not hyperactivity, leaming disability, or conduct disorder, in order to determine the predictability of averaged PSA data in ADHD. We further predicted that perturbation of the EEG by attention to reading and
Communications should be addressed to: Dr. Lubar; Departmentof Psychology;Universityof Tennes~e" Knoxville, TN 37996-0900. Received January 22, 1991; acceptedMay 24, 1991.
drawing tasks would further distinguish individuals with ADHD from controls.
Methods
Patients. All patients were male, right-handed, and 9-12 years of age. All children lived with their families and were enrolled in school; however, those under the care of juvenile or adoptive agencies were excluded. All ADHD children with the main complaints of inattention and impulsivity had been referred to one of two pediatric neurologists from pediatricians or local school districts. Other descriptions included consistently being "off task," unmotivated, "daydreamers," and having poor compliance with instructions; however, no child displayed symptoms of a conduct disorder, specific learning disability, or dyslexia in either the classroom or home settings. Standard neurologic examinations revealed no signs of gross neurologic impairment. Connors's Parent and Teacher Behavioral Rating Scales were administered and all test patients rated above 1.5 S.D. on both measures were included in this study [16,17]. A subpopulation of this group was chosen with ADHD criteria but without symptoms of hyperactivity [1]. Children diagnosed with any other neurologic or psychiatric disorder were excluded from the study. No child was receiving any medications during or prior to data acquisition. Test and control subjects were matched by demographic variables. Control children were randomly selected from the same school population as the ADHD group. Teachers were asked to choose several children whose performance they believed to fall within the normal range. Both Teacher and Parent Connors's scales were administered; the results from both scales were in agreement and supported inclusion in the control group. Procedure. All testing took place in an accredited pediatric neurodiagnostic center during school hours. Patients and parents were initially familiarized with the laboratory setting, with all procedures explained and consent forms signed. Parents were interviewed and questions pertaining to general attitude, school work, grades, and peer relations were assessed. Patients were administered the Wide Range Achievement Test-Revised (WRAT-R) [18]. Next, the Wechsler Intelligence Scale for Children-Revised (WISC-R) was administered, followed by a 10 min rest period [19]. Measuring and marking were performed using the International 10-20 System of electrode placement. Electrodes were attached with Nihon Khoden Elefix paste and reinforced with individual gauze patches over each electrode site. A referential montage with linked mastoid reference was used along with an extra EOG electrode placed below the left eye. All impedances measured below 3 kohms. A Nihon Kohden EEG 4217 polygraph was used with a low-frequency filter set at 0.1 and a high-frequency filter at 70 Hz. The sampling rate was 128/sec. The polygraph was interfaced with a DT-2901 data translation board which performed the analog to digital conversion. Stellate Topographic Brain Mapping Program computed spectral plots, Fast Fourier Transforms (FFT), and topographic maps [20]. Channel by channel calibrations were performed prior to the testing of each patient. Patients were seated in a reclining chair in a sound-attenuated room. Five minutes of EEG were collected during each of three conditions: baseline, reading, and drawing. During each condition every effort was made to reduce artifact due to eye movement and muscle tension. During the first condition patients were instructed to visually fixate on a red dot placed 91 cm in front of them. They were instructed not to speak and to remain still. For the next condition, patients were asked to read silently a story taken from a grade-appropriate text. The story was typewritten and single-spaced without pictures and placed on a lap desk; the patient then was asked a series of questions pertaining to the story's basic theme to ensure that he had read the material. For the drawing task, patients were asked to copy a series of drawings taken from the Bender-Gestalt Visual Motor Test. The figures were placed just above the drawing area to limit eye movements, while arm and hand movements were also restricted. All eye-blink artifact was carefully monitored and removed. Another qualified individual, blind to the
condition or group, performed the artifact rejection on randomly chosen patients. Agreement on rejected epochs was greater than 95%. Following removal of the artifacts, 90-100 sec of EEG remained for each condition. These data were subjected to Fast Fourier Transformation. Sixteen channels of EEG were recorded across 6 band passes: delta 1 = .75-1.75 Hz, delta 2 = 2.0-3.75 Hz, theta = 4.0-7.75 Hz, alpha = 8-12.75 Hz, beta 1 = 13-21.75 Hz, and beta 2 = 22-31.0 Hz. The delta and beta frequency bands were subdivided to help isolate artifactgenerated events from true EEG rhythms. Tables of absolute and relative amplitude and topographic color maps were generated for further analyses. The relative measure represents the percentage of the amplitude in a specific frequency band as compared with the total amplitude across all bands. Fast Fourier transforms (FFF) were performed and numeric values plotted for each patient over 16 standard locations and 6 frequency bands. The data were normalized by performing log transformations on absolute and relative amplitude values. The formulas for each transformation were as follows: log (x) for absolute amplitude, and log (X/lX) for relative amplitude [21]. Univariate plots of the transformed data revealed normal distributions. Multiple t tests were used to test both between-group and group-bycondition interactions. Individual comparisons were made by choosing a combination of frequency, amplitude, location, and condition and comparing ADHD patients with controls. The problem inherent in multiple t tests is that overall probability levels cannot be controlled. To control this, principal component analysis (PCA) was used to reduce the data to one linear combination for use in discriminant analysis.
Results
Psychometric There were no significant differences for age or grade ( T a b l e 1). S i g n i f i c a n t l y l o w e r v e r b a l I Q w a s f o u n d in t h e ADHD
group, as measured by WlSC-R.
A l t h o u g h this
finding was statistically significant, the mean difference w a s o f little c l i n i c a l s i g n i f i c a n c e . N o s i g n i f i c a n t d i f f e r ences were found on performance or full-scale mean scores. All subtests, reading, spelling, and arithmetic, on the Wide Range Achievement Test-Revised revealed signif i c a n t l y l o w e r m e a n s c o r e s in t h e A D H D g r o u p .
Neurometric Absolute Amplitude: Between-group Comparisons. W e f o u n d n o p a r o x y s m a l a c t i v i t y o n s t a n d a r d a n a l o g E E G in any patients. Each condition- baseline, reading, and drawing- contained 96 tests of significance between g r o u p s (16 c h a n n e l s x 6 f r e q u e n c i e s ) . During the baseline condition, the majority of differe n c e s (20 o f 25) w e r e l o c a l i z e d t o f r o n t a l r e g i o n s . S e v e n o f t h e s e d i f f e r e n c e s w e r e s i g n i f i c a n t at P < .01 ( F i g 1). A l t h o u g h s o m e o f t h e s e d i f f e r e n c e s m a y b e a t t r i b u t e d to e y e m o v e m e n t , all r e f l e c t e d i n c r e a s e d a b s o l u t e a m p l i t u d e in A D H D p a t i e n t s . It is i m p o r t a n t to e m p h a s i z e that a r t i f a c t r e j e c t i o n p r i o r to t h e p e r f o r m e d a n a l y s e s w a s e s p e c i a l l y s u i t e d to t h e r e m o v a l o f e y e b l i n k s a n d r a p i d h o r i z o n t a l m o v e m e n t , b u t s l o w m o v e m e n t s still m a y h a v e c o n t r i b u t e d to t h e d e l t a I b a n d - p a s s . D u r i n g t h e r e a d i n g c o n d i t i o n , (11 o f 15) s i g n i f i c a n t d i f f e r e n c e s w e r e f o u n d in t h e t h e m b a n d , 8 o f w h i c h w e r e n o t f o u n d d u r i n g b a s e l i n e . T h e i n c r e a s e d t h e m at F4 r e a c h e d
Mann et al: Attention Deficit and EEG Analysis
31
Table 1. Psychometric differences between 25 ADHD patients and 27 controls*
Age
ADHD Patients
Controls
Mean (S.D.) Range
10.6 (1.0) 9.0-12.t
10.5 (1.1) 9.1-12.6
4.6 (1.1) 3-7
4.9 (1.2) 3-7
Grade Mean (S.D.) Range
reading revealed larger absolute amplitude decreases in frontal locations particularly for delta I ia the ADHD group. Many of these changes may be interpreled as increased saccadic eye movements in the controls during reading. Location Fp2 revealed significantly larger decreases in beta 1 in the ADHD group. Changes from baseline to drawing revealed significantly larger theta increases in the ADHD group, especially in F3 and F4, with the greatest beta decreases in T3, T4, and Fp2. Relative Amplitude: Between-group Differences
WISC-R Verbal*
Mean (S.D.) Range
101.6 (11.2) 88-123
106.9 (9.3) 92-125
WISC-R Performance
Mean (S.D.) Range
103.8 (10.6) 88- 120
105.7 (9.3) 90-121
102.5 (11.4) 88-124
107.0 (9.3) 91- 121
94.9 (9.8) 65-109
103.0 (13.9) 85-138
WISC-R Full Scale
Mean (S.D.) Range WRAT-R Readingt
Mean (S.D.) Range
Group-by-condition Interaction (Table 2)
WRAT.R SpeUingt
Mean (S.D.) Range
92.6 (12.8) 69-114
99.9 (11.6) 82-130
WRAT.R Arithmetic~
Mean (S.D.) Range
There were no significant amplitude differences between groups of A D H D and control patients in relative amplitude during the baseline condition (Fig 1). Both groups produced similar amounts of percent amplitude for each frequency band relative to the total amplitude. This lack of significance further supports the validity of differences which were found during the cognitive tasks. The few differences in the reading condition can be attributed to decreased eye movement in the A D H D group. During the drawing condition we again observed generalized frontal and central theta increases and posterior and temporal beta decreases in the AHDH group. Five of 16 theta locations and 5 of 16 beta 1 locations achieved significance (P < .01).
90.4 (12.2) 53-112
101.1 (11.8) 88-133
* WRAT-Rvalues represent standard scores. + P < .05 2 tail. ~:P < .01 2 tail.
the P = .01 level of significance. Therefore, the A D H D group had a significant increase in frontal absolute theta amplitude during this cognitive task. During the drawing task, a similar pattern was found, in which the theta band revealed more differences than any other frequency (12 of 31) and in virtually identical locations as those found in the reading condition. Eight of these differences were significant at P > .01 and 6 of these were found in frontal locations. Delta 2 and alpha bands found in frontal locations also differed significantly (P = .01). All absolute differences were attributed to increased amplitude in the A D H D group. Increased theta activity could be seen by routine inspection of standard EEG as well as by power spectral analysis and topographic mapping. Group.by-condition Interaction (Table 2)
The group-by-condition interaction assesses group changes between conditions. Changes from baseline to
32 PEDIATRICNEUROLOGY Vol. 8 No. 1
In the boys with ADHD, the largest increase in theta, as well as the largest decrease in beta 1 (P < .01), was demonstrated in the right frontal region (F4) using baseline-toreading difference scores. Changes from the baseline to drawing task were responsible for more significant differences than any other comparison. Generalized, greater decreases in beta were clearly present in the ADHD group. Both delta frequencies demonstrated decreased activity at T3 and T4, and beta 2 followed beta 1 especially in the same temporal regions. As was true in the reading task, the largest significant differences were demonstrated at F4 (P < .001) for increased theta- and decreased beta-l-relative amplitudes. In order to simplify the large number of tests performed, PCA was used to combine the many variables into principal component scores according to Gasser's criteria [21 ]. For example, the relative amplitude measures of beta 1 from all 14 channels not including 01 and 02 where no differences were found were converted into principal components. The first 3 of these components represented 80% of the variance contained in the original 14 variables. Analyses then could be performed on the 3 uncorrelated components instead of the 14 original variables. Discriminant function analysis can be useful in performing clinical estimates of how well our model can predict whether a child meeting criteria for a given population can be classified as either A D H D or control. Discriminant analysis applied to the drawing condition values disclosed that the groups were significantly different (F = 6,48,
Figure 1. Topographic distribution of absolute and relative amplitude differences color coded as percent change and level of significance for 6 frequency bands matched between ADHD (N = 25) and control (N = 27) groups. Absolute amplitude represents EEG voltage differences between groups. Relative amplitude represents the difference in the proportion of activity in a chosen frequency band. Relative amplitude is computed by dividing the absolute amplitude in each frequency band by the total amplitude for all bands for each of the 16 channels.
P < .003) using the first two beta 1 principal components. The other component did not contribute to predictability, which does not mean that the theta component had no predictability, but rather that beta 1 was a better predictor and that theta offered nothing in addition to beta 1. The
fact that the location T6 had a significant t test at P < .001 may account for this finding. Discriminant functions demonstrated that for a child not included in the sample statistics, the predictability rating of group membership was 74% for the controls and 80% for the A D H D group.
Mann et al: Attention Deficit and EEG Analysis 33
"Fable 2.
A b s o l u t e and relative differences*
Delta 1 Absolute
Baseline R e a d i n g F r e q u e n c i e s Delta 2 Theta
Beta 1
Delta I
Baseline D r a w i n g F r e q u e n c i e s Delta 2 Theta Beta 1
Beta 2
Differences
F7
--
Fg
--
T3
T4 FpI
- -
Fp2
F3
+++
F4
+++
Relative Differences F8
++
--
T3
++
++
T4
++
++
Fpi
++
Fp2
++
F3
++
C4 P3
P4 * O n l y those locations a n d f r e q u e n c y b a n d s for which significant differences between groups o c c u r r e d are included. Frequencies: delta 1 = .75-1.75 Hz, delta 2 = 2-3.75 Hz, theta = 4-7.75 Hz, alpha = 8-12.75 Hz, beta 1 = 13-21.75 Hz, a n d beta 2 = 22-31 Hz. A D H D > Controls: ++ = P < . 0 1 . +++ = P < .001.
Controls > A D H D : - P
Sixteen-channel topographic brain mapping of electroencephalograms of 25 right-handed males, 9-12 years of age, with attention-deficit-hyperactivity disorder revealed increased theta (4-7.75 Hz) and decreased beta 1 (12.75-21 Hz) when compared with 27 controls matched for age and grade level. The differences were greater when patients were tested for reading and drawing skills, but were decreased when they were at rest during visual f'Lxation. Although the differences in patients with attention-deficit-hyperactivity disorder were generalized, increased theta was more prominent in frontal regions, while beta 1 was significantly decreased in temporal regions. Principal component analysis was used to combine the variables into 2 components which accounted for 82% of the total variance. A discriminant function analysis using these components was able to predict group membership for attention-deficit-hyperactivity disorder patients 80% of the time and 74% for controls. These findings support the use of topographic electroencephalography for further elucidation of the neurophysiology of attention-deficithyperactivity disorder. Mann CA, Lubar JF, Zimmerman AW, Miller CA, Muenchen RA. Quantitative analysis of EEG in boys with attention-deficit-hyperactivity disorder: Controlled study with clinical implications. Pediatr Neurol 1992;8:30-6.
Introduction Attention-deficit-hyperactivity disorder (ADHD) is a common disorder in school children with uncertain and probably multiple etiologies. It is characterized clinically by decreased attention span, impulsivity and, for some children, increased motor activity. The broad nature of diagnostic criteria outlined in DSM III-R has resulted in a heterogenous population of children being classified with this disorder [1].
From the Departmentsof *Psychology,tPediatrics (Neurology),and §Computing Center; Universityof Tennessee; and :~Departmentof Medicine; EastTennesseeChildren's Hospital; Knoxville,Tennessee.
30 PEDIATRICNEUROLOGY Vol.8 No. 1
Among the investigational approaches to ADHD that have been used, the analog electroencephalogram (EEG) has been reported to demonstrate increased background slow activity for chronologic age, in addition to occasional paroxysmal activity [2-5]; however, no specific EEG patterns have been identified and this technique has not been regarded as generally useful for the evaluation of children with ADHD [6]. Evoked potential (EP) differences have also been reported for children with ADHD. In a previous study, we reported differences between normal, learning-disabled, and gifted children based upon an auditory evoked potential task [7]. The results replicated previous studies by demonstrating that children with ADHD and reading disabilities had a suppressed P300 component compared with matched controls in response to a target auditory stimulus [8,9]. In addition, they made more response errors to target stimuli. The P300 component is believed to represent the interpretation of stimulus meaning, while the P200 represents stimulus detection [ 10]. A newer technique, computerized power spectral analysis (PSA), permits topographic representation and statistical analysis of EEG for individual and group comparisons. Previous studies using PSA confirmed findings from analog EEG studies of excessive theta and lack of alpha attenuation in children with deficient attention in the absence of conduct disorders [1 t-14]. These findings suggest that underarousal or hypoactivation of the neocortex is present in children with ADHD which may lend support to postulated noradrenergic involvement in this disorder [l 5]. We hypothesized that children with ADHD may have distinctive findings on topographic PSA-EEG which can be used to distinguish groups with this disorder on a statistical basis from matched controls. Our purpose was to study a carefully selected group of boys diagnosed with attention deficit but not hyperactivity, leaming disability, or conduct disorder, in order to determine the predictability of averaged PSA data in ADHD. We further predicted that perturbation of the EEG by attention to reading and
Communications should be addressed to: Dr. Lubar; Departmentof Psychology;Universityof Tennes~e" Knoxville, TN 37996-0900. Received January 22, 1991; acceptedMay 24, 1991.
drawing tasks would further distinguish individuals with ADHD from controls.
Methods
Patients. All patients were male, right-handed, and 9-12 years of age. All children lived with their families and were enrolled in school; however, those under the care of juvenile or adoptive agencies were excluded. All ADHD children with the main complaints of inattention and impulsivity had been referred to one of two pediatric neurologists from pediatricians or local school districts. Other descriptions included consistently being "off task," unmotivated, "daydreamers," and having poor compliance with instructions; however, no child displayed symptoms of a conduct disorder, specific learning disability, or dyslexia in either the classroom or home settings. Standard neurologic examinations revealed no signs of gross neurologic impairment. Connors's Parent and Teacher Behavioral Rating Scales were administered and all test patients rated above 1.5 S.D. on both measures were included in this study [16,17]. A subpopulation of this group was chosen with ADHD criteria but without symptoms of hyperactivity [1]. Children diagnosed with any other neurologic or psychiatric disorder were excluded from the study. No child was receiving any medications during or prior to data acquisition. Test and control subjects were matched by demographic variables. Control children were randomly selected from the same school population as the ADHD group. Teachers were asked to choose several children whose performance they believed to fall within the normal range. Both Teacher and Parent Connors's scales were administered; the results from both scales were in agreement and supported inclusion in the control group. Procedure. All testing took place in an accredited pediatric neurodiagnostic center during school hours. Patients and parents were initially familiarized with the laboratory setting, with all procedures explained and consent forms signed. Parents were interviewed and questions pertaining to general attitude, school work, grades, and peer relations were assessed. Patients were administered the Wide Range Achievement Test-Revised (WRAT-R) [18]. Next, the Wechsler Intelligence Scale for Children-Revised (WISC-R) was administered, followed by a 10 min rest period [19]. Measuring and marking were performed using the International 10-20 System of electrode placement. Electrodes were attached with Nihon Khoden Elefix paste and reinforced with individual gauze patches over each electrode site. A referential montage with linked mastoid reference was used along with an extra EOG electrode placed below the left eye. All impedances measured below 3 kohms. A Nihon Kohden EEG 4217 polygraph was used with a low-frequency filter set at 0.1 and a high-frequency filter at 70 Hz. The sampling rate was 128/sec. The polygraph was interfaced with a DT-2901 data translation board which performed the analog to digital conversion. Stellate Topographic Brain Mapping Program computed spectral plots, Fast Fourier Transforms (FFT), and topographic maps [20]. Channel by channel calibrations were performed prior to the testing of each patient. Patients were seated in a reclining chair in a sound-attenuated room. Five minutes of EEG were collected during each of three conditions: baseline, reading, and drawing. During each condition every effort was made to reduce artifact due to eye movement and muscle tension. During the first condition patients were instructed to visually fixate on a red dot placed 91 cm in front of them. They were instructed not to speak and to remain still. For the next condition, patients were asked to read silently a story taken from a grade-appropriate text. The story was typewritten and single-spaced without pictures and placed on a lap desk; the patient then was asked a series of questions pertaining to the story's basic theme to ensure that he had read the material. For the drawing task, patients were asked to copy a series of drawings taken from the Bender-Gestalt Visual Motor Test. The figures were placed just above the drawing area to limit eye movements, while arm and hand movements were also restricted. All eye-blink artifact was carefully monitored and removed. Another qualified individual, blind to the
condition or group, performed the artifact rejection on randomly chosen patients. Agreement on rejected epochs was greater than 95%. Following removal of the artifacts, 90-100 sec of EEG remained for each condition. These data were subjected to Fast Fourier Transformation. Sixteen channels of EEG were recorded across 6 band passes: delta 1 = .75-1.75 Hz, delta 2 = 2.0-3.75 Hz, theta = 4.0-7.75 Hz, alpha = 8-12.75 Hz, beta 1 = 13-21.75 Hz, and beta 2 = 22-31.0 Hz. The delta and beta frequency bands were subdivided to help isolate artifactgenerated events from true EEG rhythms. Tables of absolute and relative amplitude and topographic color maps were generated for further analyses. The relative measure represents the percentage of the amplitude in a specific frequency band as compared with the total amplitude across all bands. Fast Fourier transforms (FFF) were performed and numeric values plotted for each patient over 16 standard locations and 6 frequency bands. The data were normalized by performing log transformations on absolute and relative amplitude values. The formulas for each transformation were as follows: log (x) for absolute amplitude, and log (X/lX) for relative amplitude [21]. Univariate plots of the transformed data revealed normal distributions. Multiple t tests were used to test both between-group and group-bycondition interactions. Individual comparisons were made by choosing a combination of frequency, amplitude, location, and condition and comparing ADHD patients with controls. The problem inherent in multiple t tests is that overall probability levels cannot be controlled. To control this, principal component analysis (PCA) was used to reduce the data to one linear combination for use in discriminant analysis.
Results
Psychometric There were no significant differences for age or grade ( T a b l e 1). S i g n i f i c a n t l y l o w e r v e r b a l I Q w a s f o u n d in t h e ADHD
group, as measured by WlSC-R.
A l t h o u g h this
finding was statistically significant, the mean difference w a s o f little c l i n i c a l s i g n i f i c a n c e . N o s i g n i f i c a n t d i f f e r ences were found on performance or full-scale mean scores. All subtests, reading, spelling, and arithmetic, on the Wide Range Achievement Test-Revised revealed signif i c a n t l y l o w e r m e a n s c o r e s in t h e A D H D g r o u p .
Neurometric Absolute Amplitude: Between-group Comparisons. W e f o u n d n o p a r o x y s m a l a c t i v i t y o n s t a n d a r d a n a l o g E E G in any patients. Each condition- baseline, reading, and drawing- contained 96 tests of significance between g r o u p s (16 c h a n n e l s x 6 f r e q u e n c i e s ) . During the baseline condition, the majority of differe n c e s (20 o f 25) w e r e l o c a l i z e d t o f r o n t a l r e g i o n s . S e v e n o f t h e s e d i f f e r e n c e s w e r e s i g n i f i c a n t at P < .01 ( F i g 1). A l t h o u g h s o m e o f t h e s e d i f f e r e n c e s m a y b e a t t r i b u t e d to e y e m o v e m e n t , all r e f l e c t e d i n c r e a s e d a b s o l u t e a m p l i t u d e in A D H D p a t i e n t s . It is i m p o r t a n t to e m p h a s i z e that a r t i f a c t r e j e c t i o n p r i o r to t h e p e r f o r m e d a n a l y s e s w a s e s p e c i a l l y s u i t e d to t h e r e m o v a l o f e y e b l i n k s a n d r a p i d h o r i z o n t a l m o v e m e n t , b u t s l o w m o v e m e n t s still m a y h a v e c o n t r i b u t e d to t h e d e l t a I b a n d - p a s s . D u r i n g t h e r e a d i n g c o n d i t i o n , (11 o f 15) s i g n i f i c a n t d i f f e r e n c e s w e r e f o u n d in t h e t h e m b a n d , 8 o f w h i c h w e r e n o t f o u n d d u r i n g b a s e l i n e . T h e i n c r e a s e d t h e m at F4 r e a c h e d
Mann et al: Attention Deficit and EEG Analysis
31
Table 1. Psychometric differences between 25 ADHD patients and 27 controls*
Age
ADHD Patients
Controls
Mean (S.D.) Range
10.6 (1.0) 9.0-12.t
10.5 (1.1) 9.1-12.6
4.6 (1.1) 3-7
4.9 (1.2) 3-7
Grade Mean (S.D.) Range
reading revealed larger absolute amplitude decreases in frontal locations particularly for delta I ia the ADHD group. Many of these changes may be interpreled as increased saccadic eye movements in the controls during reading. Location Fp2 revealed significantly larger decreases in beta 1 in the ADHD group. Changes from baseline to drawing revealed significantly larger theta increases in the ADHD group, especially in F3 and F4, with the greatest beta decreases in T3, T4, and Fp2. Relative Amplitude: Between-group Differences
WISC-R Verbal*
Mean (S.D.) Range
101.6 (11.2) 88-123
106.9 (9.3) 92-125
WISC-R Performance
Mean (S.D.) Range
103.8 (10.6) 88- 120
105.7 (9.3) 90-121
102.5 (11.4) 88-124
107.0 (9.3) 91- 121
94.9 (9.8) 65-109
103.0 (13.9) 85-138
WISC-R Full Scale
Mean (S.D.) Range WRAT-R Readingt
Mean (S.D.) Range
Group-by-condition Interaction (Table 2)
WRAT.R SpeUingt
Mean (S.D.) Range
92.6 (12.8) 69-114
99.9 (11.6) 82-130
WRAT.R Arithmetic~
Mean (S.D.) Range
There were no significant amplitude differences between groups of A D H D and control patients in relative amplitude during the baseline condition (Fig 1). Both groups produced similar amounts of percent amplitude for each frequency band relative to the total amplitude. This lack of significance further supports the validity of differences which were found during the cognitive tasks. The few differences in the reading condition can be attributed to decreased eye movement in the A D H D group. During the drawing condition we again observed generalized frontal and central theta increases and posterior and temporal beta decreases in the AHDH group. Five of 16 theta locations and 5 of 16 beta 1 locations achieved significance (P < .01).
90.4 (12.2) 53-112
101.1 (11.8) 88-133
* WRAT-Rvalues represent standard scores. + P < .05 2 tail. ~:P < .01 2 tail.
the P = .01 level of significance. Therefore, the A D H D group had a significant increase in frontal absolute theta amplitude during this cognitive task. During the drawing task, a similar pattern was found, in which the theta band revealed more differences than any other frequency (12 of 31) and in virtually identical locations as those found in the reading condition. Eight of these differences were significant at P > .01 and 6 of these were found in frontal locations. Delta 2 and alpha bands found in frontal locations also differed significantly (P = .01). All absolute differences were attributed to increased amplitude in the A D H D group. Increased theta activity could be seen by routine inspection of standard EEG as well as by power spectral analysis and topographic mapping. Group.by-condition Interaction (Table 2)
The group-by-condition interaction assesses group changes between conditions. Changes from baseline to
32 PEDIATRICNEUROLOGY Vol. 8 No. 1
In the boys with ADHD, the largest increase in theta, as well as the largest decrease in beta 1 (P < .01), was demonstrated in the right frontal region (F4) using baseline-toreading difference scores. Changes from the baseline to drawing task were responsible for more significant differences than any other comparison. Generalized, greater decreases in beta were clearly present in the ADHD group. Both delta frequencies demonstrated decreased activity at T3 and T4, and beta 2 followed beta 1 especially in the same temporal regions. As was true in the reading task, the largest significant differences were demonstrated at F4 (P < .001) for increased theta- and decreased beta-l-relative amplitudes. In order to simplify the large number of tests performed, PCA was used to combine the many variables into principal component scores according to Gasser's criteria [21 ]. For example, the relative amplitude measures of beta 1 from all 14 channels not including 01 and 02 where no differences were found were converted into principal components. The first 3 of these components represented 80% of the variance contained in the original 14 variables. Analyses then could be performed on the 3 uncorrelated components instead of the 14 original variables. Discriminant function analysis can be useful in performing clinical estimates of how well our model can predict whether a child meeting criteria for a given population can be classified as either A D H D or control. Discriminant analysis applied to the drawing condition values disclosed that the groups were significantly different (F = 6,48,
Figure 1. Topographic distribution of absolute and relative amplitude differences color coded as percent change and level of significance for 6 frequency bands matched between ADHD (N = 25) and control (N = 27) groups. Absolute amplitude represents EEG voltage differences between groups. Relative amplitude represents the difference in the proportion of activity in a chosen frequency band. Relative amplitude is computed by dividing the absolute amplitude in each frequency band by the total amplitude for all bands for each of the 16 channels.
P < .003) using the first two beta 1 principal components. The other component did not contribute to predictability, which does not mean that the theta component had no predictability, but rather that beta 1 was a better predictor and that theta offered nothing in addition to beta 1. The
fact that the location T6 had a significant t test at P < .001 may account for this finding. Discriminant functions demonstrated that for a child not included in the sample statistics, the predictability rating of group membership was 74% for the controls and 80% for the A D H D group.
Mann et al: Attention Deficit and EEG Analysis 33
"Fable 2.
A b s o l u t e and relative differences*
Delta 1 Absolute
Baseline R e a d i n g F r e q u e n c i e s Delta 2 Theta
Beta 1
Delta I
Baseline D r a w i n g F r e q u e n c i e s Delta 2 Theta Beta 1
Beta 2
Differences
F7
--
Fg
--
T3
T4 FpI
- -
Fp2
F3
+++
F4
+++
Relative Differences F8
++
--
T3
++
++
T4
++
++
Fpi
++
Fp2
++
F3
++
C4 P3
P4 * O n l y those locations a n d f r e q u e n c y b a n d s for which significant differences between groups o c c u r r e d are included. Frequencies: delta 1 = .75-1.75 Hz, delta 2 = 2-3.75 Hz, theta = 4-7.75 Hz, alpha = 8-12.75 Hz, beta 1 = 13-21.75 Hz, a n d beta 2 = 22-31 Hz. A D H D > Controls: ++ = P < . 0 1 . +++ = P < .001.
Controls > A D H D : - P
