Bupropion Hydrochloride in Attention Deficit Disorder

with Hyperactivity
C. KEITH CONNERS, PH.D., CHARLES D. CASAT, M.D., C. THOMAS GUALTIERI, M.D.,
ELIZABETH WELLER, M.D., MARK READER, PH.D., ALLAN REISS, M.D., RONALD A. WELLER, M.D.,
MOISE KHAYRALLAH, PH.D., AND JOHN ASCHER, M.D.

ABSTRACT
Objective: This is a multisite, double-blind, placebo-controlled trial to determine the safety and efficacy of bupropion
in the treatment of children with attention deficit disorder with hyperactivity (ADDH). Method: In a four-center, double-
blind comparison of bupropion (n = 72) and placebo (n = 37), children aged 6 to 12 years meeting DSM-III criteria for
ADDH were randomized to receive bupropion 3 to 6 mg/kg per day or placebo, administered twice daily, at 7 A.M. and
7 P.M. Measures of efficacy included the Conners Parent and Teacher Questionnaires (93-item, 39-item, and 10 item),
Clinical Global Impressions Scales of Severity and Improvement, the Sternberg Short-Term Memory Task, and the
Continuous Performance Test. Screen and posttreatment physical examinations, electrocardiograms, electroencephalo-
grams, and clinical laboratory evaluations were performed. Height, weight, and vital signs were measured and adverse
experiences were assessed weekly. Results: A significant treatment effect, apparent as early as day 3, was present
for both conduct problems and hyperactivity on the Conners 10-item and 39-item teacher's checklist, and at day 28 for
conduct problems and restless-impulsive behavior on the 93-item parent questionnaire. Findings were of smaller
magnitude for parent ratings than teacher ratings. Significant treatment effects were present on both the Continuous
Performance Test and memory retrieval test. Effect sizes of bupropion/placebo differences for teacher and parent
ratings in this study were somewhat smaller than for standard stimulant drugs used to treat ADDH. Bupropion appeared
to be well tolerated in most children. Dermatological reactions were twice as frequent in the drug group as the placebo
group, with four reactions involving rash and urticaria that were serious enough to require discontinuation of medication.
Conclusions: Bupropion may be a useful addition to available treatments for ADDH. Comparative trials with such
standard drugs as methylphenidate are warranted to determine the relative clinical merits of bupropion. J. Am. Acad.
Child Ado/esc. Psychiatry, 1996, 35(10):1314-1321. Key Words: bupropion, attention deficit disorder with hyperactivity,
attention-deficit hyperactivity disorder.

Bupropion hydrochloride is a monocyclic phenylami- ines, with significant antidepressant effects (Preskorn
noketone structurally related to the phenylisopropylam- and Othmer, 1984) (Fig. 1). Its mechanism of action
is unclear. Bupropion does not interact with neurore-
Accepted February 27, 1996. ceptors or inhibit monoamine oxidase. A weak indirect
Dr. Conners is Professor of Medical Psychology, Duke University Medical dopamine agonist effect has been reported from animal
Center, Durham, NC Dr. Casat is Director of Research, Center for Mental
Health, Charlotte, NC Dr. Gualtieri is in private practice in Chapel Hill, studies but does not seem sufficient to explain its
NC Dr. E. Weller is Professor and Director of Training, Division of Child antidepressant activity (Cooper et a1., 1980; Ferris
and Adolescent Psychiatry, and Dr. R.A. Weller is Professor ofPsychiatry, Ohio
State University, Columbus. Dr. Reader is in private practice in Baltimore. Dr. et a1., 1983). Recent data indicate that in humans,
Reiss is Director, Department ofBehavioral Neurogeneticsand Neuroimaging, bupropion decreases whole body norepinephrine turn-
Kennedy-Krieger Institute, Baltimore. Drs. Khayrallah and Ascher are Senior
Research Scientists, Division of Clinical Research, Glaxo-Wellcome, Inc., Re- over, but effects do not appear to depend on reuptake
search Triangle Park, NC blockade (Golden et a1., 1988). In rats, bupropion
Work on this project was partially supported by Career ScienceAward K05
MH01229-03 from the NIMH to Dr. Conners. The authors thank Dr. John decreases the firing rates of noradrenergic neurons in
Messenheimer at the University of North Carolina, Chapel Hill, for his the locus ceruleus (Shea and Wang, 1985) at doses
assistance in interpreting EEG data.
Reprint requests to Dr. Conners, Department ofPsychiatry, Box 3362, Duke
found to be active in animal models of antidepressant
University Medical Center, Durham, NC 27710; telephone: (919) 684-4152; activity. These data suggest that bupropion's effects
fax: (919) 684-4564; e-mail: Conners@Compuserve.72233.245
0890-8567/96/3510-1314$03.00/0©1996 by the American Academy
on the noradrenergic system may be responsible for
of Child and Adolescent Psychiatry. its antidepressant activity.

1314 ]. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 34:10, OCTOBER 1996

 BUPROPION IN ADDH

Both dopaminergic and noradrenergic mechanisms For example, at the senior author's site there were 165 patients
screened, or about one fifth of the total entered.
have been postulated in attention deficit disorder with
Exclusion criteria were as follows: (1) WISC-R IQ < 70: (2)
hyperactivity (ADDH) (Shekim et aI. , 1983; Zametkin body weight <20 kg; (3) girls who had passed menarche; (4)
and Rapoport, 1987). Interest in the use of bupropion known hypersensitivity copsychotropi c medicat ions; and (5) hiscory
in children with ADDH derives from these putative or presence of seizure or tic disorders. All subjects had co be free
of psychotrop ic medications for a minimum of 14 days prior co
catecholamine actions. In a previous single-blind trial study entry. Permission for study participation was obtained from
of bupropion, Simeon et al. (1986) reported moderate the parents of all subjects prior co enrollment .
to marked improvement in global behavioral measures
Medication
in 12 to 17 inpatient child subjects with ADDH or
conduct disorder. This article will report on the results Subjects were randomly assigned to bupropion or placebo in a 2:1
ratio co allow increased assignment co active drug while maintaining
of a multicenter, double-blind trial ofbupropion hydro- statistical power and sound double-blind conditions. Bupropion,
chloride in children with ADDH and a possiblesecond- consisting of 50-mg and 75-mg tablets, or matching placebo tablets
ary diagnosis of conduct disorder. were dispensed weekly co the parent in bottles for supervised
administration. Three weight ranges (20 to 30 kg, 31 to 40 kg,
and >40 kg) were used CO determine medication dosage schedules.
METHOD All treatment-phase medication s were administered twice daily at
7 A.M . and 7 P.M . During the initial l-week single-blind phase,
The study was a 6-week, parallel-group, randomized, double- all subjects received placebo tablets once daily. Subjects were then
blind compari son of bupropion (n = 72) and placebo (n = 37) . randomized for the 4-week, double-blind, flexible-dose treatment
Four centers collaborated in the investigation . Subjects were children phase of the trial. Dosage was escalated from 3 mg/kg of body
aged 6 to 12 years who met the DSM-III criteria for ADDH. A weight from day 1 co day 3 to 6 mg/kg from day 15 co day 28.
l-we ek, single-blind washout period with placebo preceded the A maximum total dose of 150 mg/day was established for the
double-blind randomization co bupropion or placebo and was lowest weight range, 200 mg/d ay for the middle range, and 250
repeated at the conclusion of the active drug trial. mg/day for the heaviest weight range. A log of medication adminis-
tration was kept by the parent and reviewed weekly by the research
Subjects assistant and study physician, together with tablet counts of medica-
tion remaining in the returned bottle co monitor medication
The following inclusion criteria were used for subject selection: compliance.
(1) a score of moderate illness severity on the Child Diagnosti c
Scale: (2) a physician diagnosis of ADDH, based on history and
Efficacy Assessments
examination : (3) occurrence of mean parent and teacher scores of
at least 1.5 on the Conners Parent Questionnaire Hyperactive- Parent and Teacher Questionnaires. The 93-item version of the
Immature or Conduct Disorder factors, and the Hyperactive or Conners Parent Questionn aire and the 39-item version of the
Conduct Disorder factors from the Conners Teacher Questionnaire; Conners Teacher Questionnaire were used (Conners, 1969, 1970) .
and (4) in good physical health and without evidence oflaboracory, Each contains sympcoms rated by the observer on a 4-point Likert
electroencephalographic (EEG), or electrocardiographic (ECG) ab- scale (not at all, just a little, pretty much, very much). Three
normalities. Subjects were all recruited from university-based outpa- factors were used from the Parent Form (Hyperactive-Immature,
tient psychiatry clinics, and at one site some were recruited from Restless-Impulsive, and Conduct Disorder), and two factors were
child psychiatric inpatient adm issions (n = 6). All sites also placed chosen from the Teacher Form (H yperactivity and Conduct Disor-
local advertisements. Subjects were not reimbursed or charged for der). These questionnaires were used to establish initial symptom
services connected with the study. Recruitment cook place over a severity for subject selection and were completed at screening, day
3-year time period because of a temporary discontinuation of the 0, day 14, day 28, and day 35 by both parent and teacher.
stud y due to regulatory issues involving concurrent adult studies Abbreviated Parentand Teacher Questionnaires. Ten items known
with the drug. The number of subjects screened generally far as the " Hyperactivity Index" comprise the highest-loaded items
exceeded the number accepted, owing to the strict entrance criteria. from the other factor scales, and these were rated at all assessments
by parent and teacher. Items from the abbreviated teacher form
have been further separated into two distinct subfaccors for analysis
o purposes, one relating co aggression and the other to hyperactivity
(Loney and Milich , 1982; Pelham er al., 1989).
Clinical Global Impressions Scales. The Clinical Global Impres-
sions-Severity Scale (CGI-S) is a measure of sympcom severity on
a 7-point scale (1 = " normal, not at all ill" co 7 = "among the
most extremely ill patients"). It was completed by the study
physician at initial subject assessment and at each assessment
thereafter for each subject (NIMH, 1985). A separate global rating
of improvement, the Clinical Global Impressions-Improvement
CL Scale ,~CGI-I), was.made at e~~h ass~sment begi~,ning ",;,!th baseline
(1 = ..very much Improved, 4 = no change, 7 = very much
Fig. 1 Structure of bupropion hydrochloride. worse ).

J. AM . ACAD. CHILD ADOLESC . PSYCHIATRY, 34:10, OCTOBER 1996 1315

CONNERS ET AL.

Short-TermMemory Test. A computerized test of memory retrieval in the third grade or lower (66%), with slightly more
time, requiring a "yes/no" response to a probe letter, was completed
at baseline on day 0, and again on day 28. The probe was randomly
than half performing in an average to superior range
present 50% of the time in the memory set that preceded it. in school. Drug and placebo groups were comparable
Memory set size was either one, three, or six letters. Each test was in birth order, paternal education, and previous
composed of 60 trials, 20 at each set size. This is a test widely
used in experimental cognitive psychology that is not normed. treatments.
However, it was included in the study because a comparison of The 10 to 1 ratio of males to females in this sample
the active drug and placebo effects on short-term memory processes was typical of clinic samples of DSM-III ADDH in
would be informative with respect to the question of possible
interactions between memory load and drug effect. This is similar many outpatient settings at the time recruitment for
to the testing paradigm used in Sprague and Sleator's (l977) classic the study began. Epidemiological studies indicate a
paper on methylphenidate and cognitive load.
Continuous Performance Test. A test of vigilance presented letters
somewhat smaller ratio of boys to girls than appear in
on a computer monitor at l-second intervals, using subject responses clinical settings. The fact that patients could be entered
to visual "X" or "B-X" targets occurring 20% of the time in a with either a high Hyperactivity or Conduct Problem
series ofletters (Conners, 1985; Swanson, 1985). Fifty targets were
presented in each segment of the test. Correct responses, and errors parent rating probably favored the recruitment of sub-
of omission and commission, were recorded on day 0 and day 28. jects with higher levels of externalizing symptomatol-
ogy, more of whom are males. The parents were
Safety Examination
predominantly high school or college educated (81%).
Physical Examination. A medical history, review of systems, and
complete physical examination were done at screening, including
Most children (85%) had had symptoms of ADDH
height, weight, and vital signs. Subsequently, height, weight, and for more than 2 years, and almost half (45%) had
vital signs were measured at each assessment throughout each received some form of psychiatric treatment for their
subject's study participation.
Laboratory Battery. Routine laboratory tests were obtained at problems. Tests of differences between the drug and
day 0 and again at day 28, and these included (l) hematology placebo groups for these variables by t test and X2
(platelet estimation, hemoglobin, hematocrit, total red blood cell showed no significant differences. There was, however,
count, and total white blood cell count with differential); (2)
blood chemistry (calcium, creatinine, glucose, uric acid, blood urea a trend for the bupropion group to contain more of
nitrogen, total cholesterol, total protein, albumin, total bilirubin, the children with above-average school performance
alkaline phosphatase, lactate dehydrogenase, aspartate aminotrans-
ferase, and alanine aminotransferase); (3) blood sample for plasma
(40% versus 15%; X2 = 3.15, P < .076).
prolactin and growth hormone levels; (4) routine urinalysis; (5)
ECG; and (6) EEG. Efficacy Measures

Statistical Analysis
Table 1 presents the observed scores for the teacher
and parent 10-item scales, and Table 2 presents the
All variables were first examined for site by treatment by day
interactions. If interactions with site were nonsignificant, then data results of the ANCOVA for all parent and teacher
were pooled across sites in subsequent analyses. Efficacy measures rating scales. Figure 2 presents Hyperactivity factor
were examined in a repeated-measures analysis of covariance (AN-
CaVA), with treatment being a between-subjects factor (drug
scores for the 39-item teacher and 93-item parent scales.
versus placebo), with day (0, 3, 7, 14, 21, 28) and treatment by A significant treatment by day interaction indicates that
day interaction being within-subject factors. The baseline assessment the two treatments differ in their effects at different
(day 0) was used as a covariate. For the Conners Parent-Teacher
Abbreviated (lO-item) Questionnaire and the CGI scales, days 3, time points of the study.
7, 14, 21, and 28 were used for the within-subject repeated Teachers. The results indicate that a significant treat-
measure. For the longer forms of the Conners Parent and Teacher ment effect is present for both Conduct Problems and
Questionnaires, days 14 and 28 were used for the repeated measure.
Planned comparisons for examination of significant treatment by Hyperactivity on the l O-item teacher checklist and the
day interactions were carried out with unidirectional t tests. All longer Conners Teacher Questionnaire. However, this
analyses were carried out for both the final sample (observed), and
latter finding with the Hyperactivity factor scale is
for the original sample with last observation carried forward
(LOCF). Where results differ for the two methods, they are somewhat tempered by the failure of this measure to
reported separately. attain significance in the LOCF analysis (p < .08).
Nevertheless, the treatment effect noted on the 10-
RESULTS item form indicates that teachers clearly discerned a
positive bupropion effect. Analysis of differences for
Sample Description
each evaluation day using t tests indicates that teachers
The sample consisted predominantly of white (75%) noticed the effect as early as the third treatment day
male (90%) schoolchildren of about 8.5 years of age (p < .02). The teachers failed to detect any significant

1316 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 34:10, OCTOBER 1996

 BUPROPION IN ADDH

TABLE 1
10-Item Tea cher and Parent Ratings: Observed Scores
Placebo Bupropion
Period Day n Mean SO n Mean SO

Parent
Screen - 7 34 22.53 4.72 67 2 1.64 5.00
Baseline 0 36 20.67 6.07 69 19.55 5.62
Drug 3 31 16.68 6.86 61 16.10 6.23
7 34 17.32 7.27 68 15.79 6.35
14 35 16.97 7.28 64 16.02 6.53
21 33 16.33 6.53 63 15.19 6.78
28 34 16.91 7.57 62 13.81 6.83
Post 35 31 17.29 7.43 61 15.21 7.52
Teacher
Screen - 7 32 21.50 4.08 62 20.35 5.21
Baseline 0 36 20 .58 5.89 69 20.03 5.62
Drug 3 28 19.32 6.64 59 16.78 6.13
7 31 19.61 5.79 61 16.23 5.86
14 30 19.23 7.0 1 62 15.95 6.0 7
21 30 18.73 5.79 57 15.37 7.05
28 27 19.11 6. 15 54 14.67 6.97
Post 35 20 19.40 8.30 52 16.37 6.62

difference between treatment groups on day 35 , the questionnaire. Net changes were examined comparing
last day of a week of treatment in which both group s baseline and days 14, 28, and 35. Results indicate that
received placebo. significant treatment effects on the aggression subscale
Further analysis was undertaken of the Aggression were present at day 28 (p < .015 and p < .027 for
and H yperactivity subfactors for the l O-itern teacher observed and LOCF scores, respectively) and at day
14 and day 28 for the H yperactivity subscale (p <
TABLE 2 .0 1 on both days for observed scores; p < .01 on day
Teacher and Parent Ratings: Probabili ty Values for 14 and p < .06 on day 28 for LOCF scores).
Treatment and Day Effects
Parents. Parents did not distinguish any treatment
Treatment Treatment X D ay
effects on the weekly 10-item checklist. However , on
10-Item form (LOCF)
T eacher .0003 NS
Parent NS NS
10-ltem form (observed)
T eacher .00 1 NS
Parent NS NS
39-lte m teacher form (LOCF)
Conduct D isorder .05 NS
Hyperactivity .08 NS
39-ltem teacher form (observed)
Conduct D isorder .02 NS
Hyperactivity .03 NS
93-ltem parent form (LO CF)
Conduct Disorder NS .01
Restless-Impulsive NS .01
H yperative-Immature .06 NS Sc reen Baseline Oay1-4 Oay28 Day 35
93-lt em parent form (observed)
Co nduct Disorder .09 .017 Fig. 2 Hyperactivity-Impulsivity factorscoresfrom 39-item teacher scale
Restless-Impulsive .096 .013 (A) and 93-item parent rating scale (B) . Post hoc t tests indicate a significa nt
H yperactive-Immature .02 NS difference between drug and placebo at day 28 for the parent scale (p <
.02) and at days 14 and 28 for the teacher scale (p < .01 for both days
Note: LOCF = Last observation carried forward; NS not for observed scores, and P < .01 and P < .06 for last-observation-carried-
significant . forward scores).

J. AM. ACAD. CHILD ADOLESC . PSYCHIATRY, 34:10, OCT OB ER 19 96 1317

CONNERS ET AL.

the 93-item Conners Parent Questionnaire, there were interaction of this effect with set size, it appears that
significant treatment by day interactions for Conduct speed of memory retrieval was enhanced irrespective
Problems and Restless-Impulsive behavior at day 28 of memory load.
(p < .01).
Clinical Global Impressions. Both the CGI-S and Safety and Tolerance
CGI-I showed significant treatment by site interactions
Vital Signs. Vital signs from baseline to each return
throughout the course of treatment. While two sites
visit were examined by ttests for net difference between
showed a significant effect of bupropion on the CGI-
bupropion and placebo changes. There were no statisti-
S and one site reported a significant CGI-I, the pooled
cally significant effects on height, weight, or standing
results from the four sites failed to demonstrate a
systolic or diastolic blood pressure. However, statistical
significant treatment or treatment by day interaction
on either measure using repeated-measures ANCOVA. trends were noted for supine blood pressure (p < .06).
Results were identical for LOCF and observed data. There was a minor relative increase in supine systolic
blood pressure on day 28 (a decrease of 2.3 mm Hg
Cognitive Measures in the placebo group, compared with an increase of
2.4 mm Hg in the bupropion group). A relative increase
Continuous Performance Test. This test was adminis- of 4.7 beats per minute (bpm) in standing pulse was
tered at baseline and on day 28 only. Data were log- noted on day 28 (a decrease of 2.9 bpm in the placebo
transformed prior to analysisbecause of severeskewness. group, compared with an increase of 1.8 bpm in the
Pre- and posttest scores were used as repeated measures, bupropion group) (p < .04).
with treatment and site as between-subjects variables. Electrocardiogram. There was a mean heart rate in-
Age was also used as a covariate because of a strong
crease of approximately 2 bpm in the bupropion group
age correlation with scores. Of main interest is the
and a mean decrease of approximately 5 bpm in the
treatment by day interaction. There was a slight im-
placebo group. The net difference was statistically signi-
provement in the bupropion group, accompanied by
ficant (p < .009). There was no change in the QT
a slight worsening in the placebo group. This was
interval in the bupropion group, but a slight increase
significant for the more difficult "B-X" task (F[ 1,89] =
was observed in the placebo group (from 0.36 second
4.50, P < .037). Results did not differ for LOCF and
to 0.37 second); the net difference was significant at
observed data. There was a near-significant site effect
(p < .095) due to one site, which mistakenly adminis-
p < .02. There was a decrease of about 4 degrees in
tered only half as many trials as other sites, but no the R axis for the placebo group, and an increase of
treatment by site interaction. Part of the treatment about 5 degrees in the bupropion group. This net
effect appeared to be due to the offsetting of a perform- difference reached statistical significance (p < .015).
ance decline associated with repeated administration No statistically significant differences in PR interval
of the test. or QRS duration were observed.
Memory Retrieval Time. Analysis of errors indicates Hematology. Hematological results were compared
that there were no treatment effects on errors. (This between screening and drug termination days. There
is a desirable effect in this type of test because the was a slight drop in monocytes in the placebo group
main interest centers on the speed of response to the (mean change = 0.25%) and a slight increase in the
probe stimulus as a function of set size for accurate bupropion group (mean change = 0.41%), which re-
responses.) Reaction time scores were log-transformed. sulted in a significant net difference (p < .04). No other
Data were analyzed with treatment and site as between- measures showed statistically significant mean changes.
subjects factors, and pre/post and set size (one, three, Blood Chemistry. There was a slight mean increase
or six items per set) as repeated-measures factors. Age of blood urea nitrogen levels in the placebo group
was used as a covariate. There were no significant (mean change = 0.24 mg/dL) and a decrease of 1.55
treatment by site interactions, indicating similar pat- mg/dL in the bupropion group (p < .005). Bilirubin
terns of response at all four sites. Main interest centers was unchanged in the bupropion group but slightly
on interactions with treatment. There was a significant increased in the placebo group (p < .02). There were
treatment effect interaction (treatment by pre/post no treatment-related effects for either growth hormone
F[1,62] = 4.14, P < .046). Since there was no further or prolactin.

1318 J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 34:10, OCTOBER 1996

 BUPROPION IN ADDH

None of the statistically significant differences noted baseline. One patient showed spike-and-wave dis-
on the above safety parameters were considered to be charges at day 28 but showed no abnormalities at
clinically important. baseline or at a 3-month follow-up visit. Two other
Treatment-Emergent Adverse Experiences. In general, patients who had normal EEGs at baseline showed
side effectswere infrequent. Dermatological and gastro- rolandic discharges at day 28. Neither of these children
intestinal complaints were the most frequently encoun- achieved sleep at the baseline EEG. Since rolandic
tered symptoms in the bupropion group. Although discharges are very often only present in light sleep,
16.7% of the bupropion patients complained of nausea both patients may have had rolandic activity prior to
and vomiting, there was almost an equal rate of such drug administration that was not detected.
complaints (13.5%) in the placebo group. Twelve The regulatory agreement for this study did not
bupropion patients, or 16.7%, had a rash versus 8.1% permit continued open-label follow-up of the drug. At
in the placebo group. Four patients were discontinued the end of the study all subjects were discontinued
from the study because of apparent allergic reactions from bupropion and placebo. They were then given
to bupropion (skin rash with urticaria). One of these appropriate clinical referral information and no further
four patients had a low-grade fever with flushing, study follow-up was maintained.
swelling of the lower extremities, and pruritus. All
other adverse experiences were managed by lowering
the dose or adding a concomitant medication allowed DISCUSSION
in the protocol (such as mild topical anesthetics, aspi-
rin). Four additional bupropion patients discontinued The teacher rating scales in this study provide strong
the study prematurely, three because of ineffectiveness evidence that bupropion reduced hyperactivity and
and one who did not return for scheduled visits. Two aggression in a group of children with conduct and
placebo patients discontinued the study prematurely, attention problems. Parents also reported significant
one because of ineffectiveness and one who did not reductions of symptoms, but these effects were less
return. reliable and of less magnitude. Whereas teachers saw
Electroencephalogram. All patients received an EEG an immediate effect after day 3 of treatment, parents
at baseline and again on the final day of active treat- did not reliably detect improvement until the fourth
ment. No patients showed any evidence of clinical week of treatment (day 28). Part of this difference
seizure activity during treatment. Abnormal findings between parents and teachers may reflect the sizable
are reported in Table 3. placebo effect in the parent ratings. For example,
All EEGs were collected and read by local pediatric whereas the decrease in the placebo group for the
neurologists at each site; Ten cases with abnormal parents' l O-item ratings averaged about 7 points, the
findings that warranted further investigation were re- teachers' ratings in the placebo group never declined
viewed by a pediatric neurologist. Of the 10 cases with more than 2 points. This represents a difference of
abnormal findings, 4 were judged abnormal at both mor~ than three standard errors in the placebo effect.
baseline and day 28 and appeared not to be drug- The teacher and parent rating scale data stand in
related. Two of those four patients demonstrated diffuse contrast with the results of the investigator-rated CGI
background slowing at each recording, and one dis- scales. For the latter there were sizable treatment by site
played potentially biphasic wavesat baseline and occipi- interactions, indicating that some sites were detecting
tal sharp waves at both discontinuation and a follow- global effectsthat others did not. Stimulant drug studies
up EEG. One patient in the placebo group displayed of hyperactive and aggressive children have generally
rolandic discharges, which are common to this age shown that clinical global impressions are sensitive and
group and often not associated with clinical seizures. reliable indicators of drug effect (Werry et al., 1976).
EEGs of six patients receiving bupropion went from No satisfactory explanation can be offered for the
normal to abnormal. Slow waves appeared after treat- failure of the CGI results to be consistent across sites
ment in one patient, and though the patient may in this investigation. The study procedures should have
simply have been drowsy at the recording, a drug effect made it difficult to confound improvement ratings by
cannot be ruled out. Two patients showed spike-and- knowledge of side effects. While all investigators were
wave discharges at day 28 after a normal recording at experienced clinicians, the results underscore the need

]. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 34:10, OCTOBER 1996 1319

CONNERS ET AL.

TABLE 3
Electroencephalogram Results
Baseline EEG Discontinuation EEG Follow-up
Patient Tx Result Description Result Description Result Description

102 B Normal Abnormal Slow waves Abnormal

303 B Abnormal Posterior polyphasic; Abnormal Occipital sharp waves Abnormal Sharp waves lOL;
potentially biphasic R>L no change from
previous EEG
311 B Normal No sleep achieved Abnormal
325 B Normal Abnormal Spike and wave Normal
complexes
405 B Normal No sleep achieved Abnormal Rolandic spike
410 B Normal Abnormal Spike and wave
413 B Abnormal Diffuse background slowing Abnormal Diffuse background
slowing
415 B Abnormal Diffuse background slowing Abnormal Diffuse background
slowing
421 B Normal Abnormal 3/sec spike and wave
423 P Abnormal Rolandic spikes Abnormal Spike & slow wave
considered to be
rolandic

Note: B = bupropion; P = placebo.

for all rnultisite investigations to require rater training- size X probe size) is relatively small compared with
in-common prior to the enrollment of subjects, to the load employed in Sprague and Slearor's classic
ensure acceptable interrater reliability. Furthermore, study, which showed a curvilinear relation of set size
these between-site findings of varying sensitivity on to stimulant dosage (Sprague and Sleator, 1977).
the CGI scale emphasize the importance of relying on An important issue is the magnitude of the clinical
standardized instruments in drug trials. effect achieved by bupropion. First, it should be noted
The results of the cognitive studies are weakened that the bupropion group improved below the subject
to some extent by the smaller number of occasions selection cutoff of 15 points of the l G-itern Conners
on which they were administered and by substantial scale commonly used for clinical identification of pa-
amounts of missing data due to technical problems or tients. The observed improvement of about 5 to 6
administrative errors. Nevertheless, the most de- points on this scale with bupropion may be compared
manding part of the CPT, the "B-X" task, showed a with the l O-point change seen by teachers with pemo-
significant drug effect in reducing errors of omissions. line (Conners et al., 1972). The change with bupropion
Thus, like stimulant drugs, which often reduce errors of is approximately the same as the effects of a good
omission in studies of hyperactive children, bupropion behavior therapy program (Pelham et al., 1980). The
appeared to do more than simply calm the child changes on the Aggression and Hyperactivity subfactors
behaviorally. of the 10-item scale are approximately half a standard
Results of the Sternberg Short-Term Memory Task deviation. In behavioral research, this is considered a
are consistent with the demonstrations by Sergeant moderate size effect (Cohen, 1987). Meta-analyses of
and colleagues (Sergeant and van der Meere, 1988) stimulant drug effects consistently show effect sizes of
that stimulants alter the intercept of the response 0.8 to 0.9 for behavioral measures-a very large effect
function relating reaction time to memory load, not in Cohen's terminology (Kavale, 1982; Ottenbacher
its slope. This suggests bupropion is primarily acting and Cooper, 1983; Thurber and Walker, 1983). Al-
upon response processes, rather than stimulus encoding though comparisons across different methodologies,
or memory capacity. There is an overall improvement subjects, dosing, etc., are hazardous, these studies sug-
in response efficiency, without change in performance gest that the effects in the present study are somewhat
in relation to the size of the memory load. However, less robust than the standard stimulant drugs used to
it should be noted that the memory load (load = set treat ADDH.

1320 ]. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 34:10, OCTOBER 1996

 BUPROPION IN ADDH

Bupropion appeared to be well tolerated in most such as academic progress, when these are available.
children receiving the drug. The most frequently en- It will now be important to undertake comparative drug
countered adverse effects were dermatological (16.7% trials with standard drugs such as methylphenidate,
in the bupropion group versus 8.1% in the placebo dextroamphetamine, and pemoline in order that the
group) and nausea/vomiting (17% for bupropion versus relative clinical merits ofbupropion can be determined.
13.5% for placebo). That four children had an allergic
response to the medicine underlines the importance REFERENCES
of careful follow-up evaluations. Minor differences were Cohen J (1987), Statistical Power Analysis for the Behavioral Sciences.
noted for ECG, hematological, and liver function Hillsdale, NJ: Erlbaum
Conners CK (1969), A teacher rating scale for use in drug studies with
findings, but these were thought to be of no clinical children. Am J Psychiatry 126:884-888
significance. Six patients showed an abnormal EEG at Conners CK (1970), Symptom patterns in hyperkinetic, neurotic, and
normal children. Child Dev 41:667-682
day 28 that was not present at baseline. Three of these Conners CK (1985), The computerized continuous performance test.
were spike-and-wave discharges. Since these abnormali- Psychopharmacol Bull 21:891-892
Conners CK, Taylor E, Meo G, KurtzMA, Fournier M (1972), Magnesium
ties are often missed in a single 20-minute recording, pemoline and dextroamphetamine: a controlled study in children with
it is possible that they were present but missed in the minimal brain dysfunction. Psychopharmacologia 26:321-336
Cooper BR, Hester TJ, Maxwell RA (1980), Behavioral and biochemical
baseline sampling. One patient showed an increase in effects of the antidepressant bupropion (Wellbutrin): evidence for
slow waves at day 28. This was a child in a single- selective blockade of dopamine uptake in vivo. J Pharmacol Exp
Ther 215:127-134
parent family whose biological mother had a history Ferris RM, Cooper BR, Maxwell RA (1983), Studies of bupropion's
of diazepam and alcohol abuse during pregnancy. The mechanism of antidepressant activity. J Clin Psychiatry 44:74-78
Golden RN, Rudorfer MV, Sherer MA, Linnoila M, Potter WZ (1988),
three patients with spike-and-wave discharges had no Bupropion in depression 1. Biochemical effects and clinical response.
family psychiatric or medical history factors related to Arch Gen Psychiatry 45:139-143
Kavale K (1982), The efficacyof stimulant drug treatment for hyperactivity:
their condition. a meta-analysis. J Learn DisabiI15:280-289
Experience with seizures in depressed adults receiving Loney J, Milich R (1982), Hyperactivity, inattention, and aggression in
clinical practice. In: Advances in Developmental and Behavioral Pediatrics,
bupropion has shown the rate to be approximately Wolraich M, Routh D, eds. Greenwich, CT: JAl Press, pp 113-147
0.4%. Becausethe current trial in children with ADDH NIMH (1985), Clinical Global Impressions. Psychopharmacol Bull
21:839-844
was limited to 4 weeks of bupropion treatment, no Ottenbacher J], Cooper HM (1983), Drug treatment of hyperactivity in
generalization to possible occurrence of seizures with children. Dev Med Child NeuroI25:358-366
Pelham WE, Milich R, Murphy DA, Murphy HA (1989), Normative
treatment over a longer period of time can be made. data on the IOWA Conners Teacher Rating Scale. J Clin Child
Although the role of the drug in inducing observed PsychoI18:259-262
Pelham WE, Schnedler RW, Bologna NC, Contreras JA (1980), Behavioral
EEG changes must remain uncertain because of the and stimulant treatment of hyperactive children: a therapy study with
many factors that influence recordings in children, the methylphenidate probes in a within-subject design. J Appl Behav
Anal 13:221-236
overall rate of such findings in the study is quite Preskorn SH, Orhmer SC (1984), Evaluation of bupropion hydrochloride:
small. Most importantly, no actual clinical seizure the first of a new class of atypical antidepressants. Pharmacotherapy
4:20-34
abnormalities were observed in any of the study Sergeant JA, van der Meere J (1988), What happens after a hyperactive
subjects. child commits an error? Psychiatry Res 24:157-164
Shea VK, Wang CM (1985), Bupropion (Wellbutrin) inhibits the firing rate
The results of this investigation indicate that bupro- of noradrenergic but not serotonergic neurons. SocNeurosciAbstr 11:185
pion may be a potentially useful addition to the avail- Shekim WO, JavaidJ, Davis JM, Bylund DB (1983), Urinary MHPG and
HVA excretion in boys with attention deficit disorder and hyperactivity
able clinical treatments for ADDH and, perhaps, for treated with d-amphetamine. Biol Psychiatry 18:707-714
conduct disorder in children. The present study offers Simeon J, Ferguson HB, Van Wyck Fleet J (1986), Can J Psychiatry
31:581-585
evidence that clinically meaningful changes are detected Sprague RL, Sleator EK (1977), Methylphenidate in hyperkinetic children:
by teachers, and to a lesser extent by parents, within differences in dose effects on learning and social behavior. Science
198:1274-1276
a few days of initiating treatment. There are also signs Swanson JM (1985), Measures of cognitive functioning appropriate for
of modest improvements in cognitive functions of use in pediatric psychopharmacological researchstudies. Psychopharmacol
Bull 2 I:887-890
attention and memory retrieval. As with all new treat- Thurber S, Walker CE (1983), Medication and hyperactivity: a meta-
ments, however, a conservative approach is advised, analysis. J Gen Psycholl 08:79-86
Werry JS, Arnan MG, Lampen E (1976), Haloperidol and methylphenidate
involving careful physical examinations at baseline and in hyperactive children. Acta Paedopsychiatr42:26-40
during treatment, continuous monitoring of parent Zamerkin AJ, Rapoport JL (1987), Noradrenergic hypothesis of attention
deficit with hyperactivity: a critical review. In: Psychopharmacology: The
and teacher observations, frequent recording of side Third Generation ofProgress, Meltzer HY, ed. New York: Raven Press,
effects, and other objective measures of improvement, pp 837-842

J. AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 34:10, OCTOBER 1996 1321