Atomoxetine in Adults with ADHD: Two Randomized,
Placebo-Controlled Studies
David Michelson, Lenard Adler, Thomas Spencer, Frederick W. Reimherr,
Scott A. West, Albert J. Allen, Douglas Kelsey, Joachim Wernicke,
Anthony Dietrich, and Dena´i Milton
Background: Attention-deficit/hyperactivity disorder
(ADHD) has been less studied in adults than in children,
and the treatment studies reported to date have been
small, single-center trials. To assess the efficacy of atom-
oxetine, a new and highly selective inhibitor of the
norepinephrine transporter, we conducted two large, mul-
ticenter treatment trials.
Methods: Two identical studies using randomized, dou-
ble-blind, placebo-controlled designs and a 10-week treat-
ment period were conducted in adults with DSM-IV-
defined ADHD as assessed by clinical history and
confirmed by a structured interview (study I, n � 280;
study II, n � 256). The primary outcome measure was a
comparison of atomoxetine and placebo using repeated
measures mixed model analysis of postbaseline values of
the Conners’ Adult ADHD Rating Scale.
Results: In each study, atomoxetine was statistically
superior to placebo in reducing both inattentive and
hyperactive and impulsive symptoms as assessed by pri-
mary and secondary measures. Discontinuations for ad-
verse events among atomoxetine patients were under 10%
in both studies.
Conclusions: Atomoxetine appears to be an efficacious
treatment for adult ADHD. Its lack of abuse potential may
be an advantage for many patients. Biol Psychiatry 2003;
53:112–120 © 2003 Society of Biological Psychiatry
Key Words: Atomoxetine, ADHD, adults, nonstimulant
Introduction
Attention-deficit/hyperactivity disorder (ADHD) is apsychiatric disorder characterized by difficulties sus-
taining attention and difficulties with impulse control.
Evidence suggests that genetic factors are important in its
pathogenesis (Hudziak 1997) and that its pathophysiology
involves alterations in central dopaminergic and noradren-
ergic tone (Biederman et al 1999a; Safer 2000). The
disorder begins early in life and has been studied primarily
in children, among whom its prevalence is 3–7% (Amer-
ican Psychiatric Association 2000). As children mature,
symptoms can become less problematic, most likely
through developmental changes in the brain, by learning to
compensate for deficits with adaptive behaviors, or by
gravitating toward environments that make fewer demands
in areas of impairment; however, the disorder often
persists into adulthood (Biederman et al 2000; Weiss
1985) and is associated with significant morbidity and
undesirable outcomes (Brown et al 1986; Hechtman 1996;
Klein and Mannuzza, 1989, 1991; Murphy and Barkley
1996; Seidman et al 1998; Thorley 1984; Weiss 1996;
Weiss and Hechtman 1993).
Pharmacotherapies effective in children appear to be
valuable in adults (Spencer et al 1995; Wender and
Reimherr, 1990; Wender et al 1985; Wilens et al 1996);
however, the literature on treatment of adult ADHD is
limited. Published studies have been small and have
differed in the methodologies used for diagnostic ascer-
tainment and symptom assessment. No large, parallel-
design, placebo-controlled trial using rigorous diagnostic
and assessment procedures has been reported to date for
any agent.
The most widely used medications for ADHD are the
psychostimulants methylphenidate and amphetamine
(Popper 2000). There is, however, considerable interest in
alternative, nonstimulant therapies, because some patients
respond poorly to stimulants or are unable to tolerate them.
In addition, some physicians are reluctant to use stimulants
because of concerns about misuse in a population at
increased risk for substance abuse (Wilens and Biederman
1992), although whether treatment with psychostimulants
increases risk for substance abuse in adult ADHD is
controversial (Biederman et al 1999b; Wilens 2000). The
current alternatives to stimulants are primarily the antide-
pressants desipramine and bupropion. These are effica-
cious in children (Biederman et al 1989; Conners et al
From Lilly Research Laboratories (DM, AJA, DK, JW, DM) and Indiana University
School of Medicine (DM), Indianapolis, Indiana; New York University (LA),
New York, New York; Massachusetts General Hospital (TS), Boston, Massa-
chusetts; University of Utah (FWR), Salt Lake City, Utah; Psychiatric Institute
of Florida (SAW), Orlando, Florida; Neuropsychiatric Associates (AD),
Woodstock, Vermont.
Address reprint requests to David Michelson, M.D., Lilly Research Laboratories,
DC 6026, Indianapolis Indiana 46285.
Received April 26, 2002; revised August 12, 2002; accepted August 26, 2002.
© 2003 Society of Biological Psychiatry 0006-3223/03/$30.00
doi:10.1016/S0006-3223(02)01671-2
1996) and have been superior to placebo in small studies
in adults (Wilens et al 1996, 2001), but neither drug is
approved for the treatment of ADHD, and desipramine has
a low therapeutic index and can adversely affect cardiac
conduction.
The investigational drug atomoxetine is a potent inhib-
itor of the presynaptic norepinephrine transporter
(Ki�4.5nM) with minimal affinity for other noradrenergic
receptors or for other neurotransmitter transporters. It is
efficacious in children and adolescents with ADHD
(Michelson et al 2001), and preliminary data suggest
efficacy in adults (Spencer et al 1998). Atomoxetine does
not appear to have abuse potential (Heil et al 2002), and
unlike desipramine, atomoxetine is not associated with
adverse effects on cardiac conduction (Michelson et al
2001). To assess the efficacy of atomoxetine for adult
ADHD, we conducted two large, prospective, double-
blind, placebo-controlled, randomized studies. We report
the results of these studies here.
Methods and Materials
Two identical randomized, double-blind, placebo-controlled
studies were conducted concurrently at 17 (study I) and 14 (study
II) outpatient sites in North America. Adults who met DSM-IV
(American Psychiatric Association 2000) criteria for ADHD as
assessed by clinical interview and confirmed by the Conners’
Adult ADHD Diagnostic Interview for DSM-IV (CAAR-D;
Conners et al 1999) were recruited from clinics and by adver-
tisement. Patients were required to have at least moderate
symptom severity, and the diagnosis had to be corroborated by a
second reporter for either current symptoms (by a significant
other) or childhood symptoms (by a parent or older sibling). If
the second reporter’s rating did not corroborate the patient’s
report, the patient was ineligible to participate in the study.
Comorbid psychiatric diagnoses were assessed by clinical
interview and by the Structured Clinical Interview for DSM-IV
(SCID; First et al 2000). Patients who met diagnostic criteria for
current major depression or anxiety disorder or for current or past
bipolar or psychotic disorders were excluded, as were patients
with serious medical illness and patients who met DSM-IV
criteria for alcohol dependence. A history of episodic recre-
ational drug use did not exclude patients, but patients actively
using drugs of abuse at the time of study entry were excluded.
Urine screening for drugs of abuse was performed at the initial
visit and could be repeated at any time during the trial at the
investigator’s discretion.
Following an initial 1-week medication washout and evalua-
tion period, patients entered a 2-week placebo lead-in phase
(modified double blind, because efficacy raters were blind to the
protocol, but others at the investigative sites were not). Patients
who maintained the initial severity criteria required for study
entry were randomized to receive atomoxetine or placebo for a
10-week period, during which visits were biweekly. Patients
were randomized according to computer-generated treatment
codes obtained from an interactive voice-response system. Study
drug materials for both treatment groups were identical in
appearance. Adherence was assessed by pill counts and history.
Each site’s institutional review board evaluated and approved
the study protocol. After description of the procedures and
purpose of the study and before the administration of any study
procedure or dispensing of study medication, written informed
consent was obtained from each patient. The study was con-
ducted in accordance with the ethical standards of each of the
investigative sites’ institutional review boards and with the
Declaration of Helsinki 1975, as revised in 2000.
The primary outcome measure was the sum of the Inattention
and Hyperactivity/Impulsivity subscales of the investigator-rated
CAARS, for which psychometric data have been reported (Con-
ners et al 1999). Each of the 18 items of these subscales
corresponds to one of the 18 DSM-IV symptoms for ADHD and
is rated on a 4-point scale. At each visit clinicians also rated a
Clinician Global Impression of Severity Scale (CGI-S; Guy
1976). Before starting the study, efficacy raters were required to
attend a training session using observed interviews and group
discussion to standardize rating practices for the CAARS. Effi-
cacy raters for the primary outcome measure were blind to all
details of the study design, including severity criteria for entry,
dose titration, and timing of the initiation of therapy, and were
not allowed to evaluate or ask about adverse events. At baseline
and end point, patients completed a self-rated version of the
CAARS and the Wender–Reimherr Adult Attention Deficit
Disorder Scale (WRAADDS; Wender et al 1985). Anxiety and
depressive symptoms were assessed with the Hamilton Anxiety
and 17-item Hamilton Depression Rating scales (Hamilton
1960), respectively, whereas changes in social and occupational
functioning were assessed using the Sheehan Disability scale (a
self-report scale that assesses work, family, and social function-
ing).
Atomoxetine was administered in evenly divided doses in the
morning and late afternoon–early evening beginning at a total
daily dosage of 60 mg. Patients with residual symptoms had their
dosage increased to 90 mg/day after 2 weeks and to 120 mg/day
after 4 weeks. If tolerability problems developed, dosage could
be decreased to the last tolerated dosage or an increase omitted.
Safety and tolerability were assessed at each visit by open-
ended questioning for adverse events and by monitoring of vital
signs and laboratory data.
Statistical Methods
Results were analyzed on an intent-to-treat basis. The primary
analysis was a comparison between treatment groups using a
repeated measures mixed model with the MIXED procedure in
SAS (SAS Institute 1997) that contained fixed class effect terms
for treatment, investigative site, visit, and interaction between
treatment and visit. The model included postbaseline values of
the investigator-rated CAARS Total ADHD Symptom Score as
the dependent variable with a random patient effect and baseline
investigator-rated CAARS Total ADHD Symptom Score as a
covariate and used an unstructured covariance. In addition to the
sum of the 18-item score, investigator-rated CAARS Inattentive
and Hyperactive/Impulsive subscales were also computed. If
more than one item of a subscale was missing, the score for the
Atomoxetine in Adults with ADHD 113BIOL PSYCHIATRY
2003;53:112–120
subscale (and the total score) was also considered missing. If
only a single item was missing, the mean score for all other items
in the subscale was imputed as the score for the missing item. For
the self-reported CAARS, outcomes are presented as t scores to
allow comparison of symptom severity in the study population
relative to healthy adults; t scores are transformations of raw
scores based on normative data adjusted for age and gender, such
that the normative mean is assigned a t score of 50, and a change
of 1 SD is represented by a change in t score of 10 points. Thus,
for example, a patient with a raw score 3.2 SD above the
population mean would have a T score of 82.
Secondary efficacy analyses of primary and secondary out-
comes included all patients with at least 1 postbaseline measure-
ment. Safety analysis included all patients who took at least 1
dose of study drug. Secondary efficacy analyses and safety
analysis of continuous measures were performed using a last
observation carried forward (LOCF) approach to compare mean
change values from baseline to end point using an analysis of
variance (ANOVA). Treatment differences in binary measures
were assessed using Fisher’s Exact Test. All tests used a
two-sided significance level of .05.
Results
Of 448 patients initially assessed in study I, 280 met entry
criteria and were randomized to atomoxetine (n � 141) or
placebo (n� 139). In study II, 388 patients were assessed,
and 256 met entry criteria and were randomized to
atomoxetine (n � 129) or placebo (n � 127). Details of
reasons for failure to be randomized (including all patients
who signed consent but discontinued before the placebo
lead-in as well as those who entered the placebo lead-in)
are provided in Figures 1 and 2. Among patients who met
initial screening criteria and entered the placebo lead-in
phase, 19 of 318 (6.0%) in study I and 12 of 287 (4.2%)
in study II were not randomized because their symptoms
improved in response to placebo. Patient characteristics
Figure 1. Patient flow diagram: Study I.
LOE, loss of efficacy.
Figure 2. Patient flow diagram: Study II.
LOE, loss of efficacy.
114 D. Michelson et alBIOL PSYCHIATRY
2003;53:112–120
and baseline symptom severity for each study are provided
in Table 1. A majority of patients were men, and most
patients met criteria for the combined subtype of ADHD.
Overall symptom severity in all groups was approximately
3 SD above mean normative total CAARS self-report
scores, with inattentive symptoms more prominent than
hyperactive symptoms (Table 1). There were no statisti-
cally significant differences in demographics and baseline
severity between treatment groups in either study.
Efficacy results are summarized in Table 2. In both
studies, atomoxetine was superior to placebo in reduction
of ADHD symptoms as assessed by the primary outcome
measure. No treatment-by-site interaction was observed in
either study. Both the Attention and Hyperactive/Impul-
sive subscales improved significantly in the atomoxetine
groups compared with placebo groups in each study, and
outcomes were similar among patients with the combined
ADHD subtype and the inattentive subtype. No statisti-
cally significant interactions between treatment and gen-
der or treatment and age (dichotomized in each study by
median age) were observed in either study. The treatment
effect size (defined as the difference between treatment
groups in least squares means divided by the square root of
the mean square error for the entire sample) for the
primary outcome measure was 0.35 in study I and 0.40 in
study II.
Statistically significant change favoring atomoxetine
was also observed in both studies on secondary assess-
ments including the CAARS self-reports, the WRAADS,
and the CGI-S. In both studies atomoxetine was statisti-
cally significantly superior to placebo at the first postran-
domization visit; from the third postrandomization visit it
was superior to placebo at every visit in each study. There
was no difference between groups in change in Hamilton
Anxiety rating scale scores for either study. In study II but
not study I a small but statistically significant reduction
favoring placebo was observed in the 17-Item Hamilton
Depression Rating Scale score.
At end point, the most frequently prescribed dose was
90 mg (study I: 40.4%; study II: 38.8%), followed by 120
mg (study I: 39.7%; study II: 34.9%) and 60 mg (study I:
19.9%; study II: 26.4%).
Patient disposition by study is summarized in Table 3.
Compared with placebo, a greater proportion of patients
taking atomoxetine discontinued from study II but not
study I due to adverse events (Table 3). No serious safety
concerns were observed in either study. Atomoxetine was
associated with modest increases in blood pressure (all
values expressed as mean [SD] change from baseline to
end point in mm Hg: diastolic: study I: placebo 0.5 [7.8],
atomoxetine 2.3 [8.1], p � .063; study II: placebo 0.6
[7.7], atomoxetine 1.2 [9.0], p � .556; systolic: study I:
Table 1. Patient Characteristics and Baseline Symptom Severity by Study
Study I Study II
Placebo
(n � 139)
Atomoxetine
(n � 141) p Value
Placebo
(n � 127)
Atomoxetine
(n � 129) p Value
Gender n (%)
Male 87 (62.6) 91 (64.5) .804a 87 (68.5) 83 (64.3) .510a
Female 52 (37.4) 50 (35.5) 40 (31.5) 46 (35.7)
Age Mean (SD) 40.3 (11.6) 40.2 (11.7) .976b 41.2 (11.2) 43.0 (10.3) .186b
ADHD Subtype n (%)
Combined 100 (71.9) 101 (71.6) 1.00a 75 (59.1) 80 (62.0) .326a
Inattention 38 (27.3) 39 (27.7) 44 (34.6) 46 (35.7)
Hyperactive/Impulsive 1 (.7) 1 (.7) 8 (6.3) 3 (2.3)
Previous Stimulant Exposure 68 (48.9) 62 (44.0) .472a 55 (43.3) 65 (50.4) .263a
CAARS-INV Mean (SD)
Total ADHD Symptom Score 33.2 (7.8) 33.6 (7.2) .603b 34.2 (7.5) 34.9 (6.9) .597b
Inattentive 18.6 (4.4) 18.4 (4.2) .736b 19.3 (4.3) 20.0 (4.1) .223b
Hyperactive/Impulsive 14.5 (5.4) 15.2 (5.0) .309b 14.9 (5.2) 14.8 (4.8) .785b
CAARS-Self (t Score) Mean (SD)
Total ADHD Symptom Score 80.8 (12.3) 82.6 (12.7) .291b 80.0 (13.8) 82.6 (12.2) .179b
Inattentive 85.6 (12.7) 87.5 (12.5) .249b 84.4 (15.4) 87.4 (12.7) .161b
Hyperactive/Impulsive 68.4 (13.0) 69.7 (13.6) .517b 68.3 (14.1) 69.7 (11.7) .511b
CGI-ADHD-S Mean (SD) 4.7 (.7) 4.7 (.8) .886b 4.6 (.7) 4.6 (.6) .769b
WRAADDS Mean (SD) 17.6 (4.2) 18.3 (4.7) .109b 15.5 (5.7) 16.5 (5.0) .275b
HAMD-17 Mean (SD) 5.9 (3.9) 5.1 (3.6) .073b 4.6 (3.3) 4.7 (3.7) .918b
HAMA Mean (SD) 8.2 (4.8) 7.4 (5.2) .169b 5.9 (4.5) 6.5 (5.1) .333b
ADHD, attention-defict/hyperactivity disorder; CAARS, Conners’ Adult Attention Rating Scale; INV, Investigator rated; CGI-ADHD-S, Clinical Global Impressions of
Severity of ADHD Symptoms; WRAADDS, Wender–Reimherr Adult Attention Deficit Disorder Scale; HAMD-17, 17 Item Hamilton Depression Rating Scale; HAMA,
Hamilton Anxiety Rating Scale.
aTreatment comparisons conducted using Fisher’s Exact Test.
bTreatment comparisons conducted using analysis of variance.
Atomoxetine in Adults with ADHD 115BIOL PSYCHIATRY
2003;53:112–120
placebo �0.8 [9.8], atomoxetine 2.3 [11.1], p � .015;
study II: placebo 0.9 [11.1], atomoxetine 3.5 [10.6], p �
.059) as well as heart rate (mean [SD] change in beats/min:
study I: placebo �0.5 [9.3], atomoxetine 6.7 [11.6], p �
.001; study II: placebo �0.1[9.6], atomoxetine � 3.8
[10.2], p � .002). Pooled adverse event data are presented
in Table 4. No meaningful differences between groups in
laboratory results were observed.
Discussion
In two large, identical studies conducted concurrently,
atomoxetine was superior to placebo in reduction of
ADHD symptoms in adults. Outcomes were similar for
most measures across the two studies, and both attention
and hyperactive/impulsive symptoms improved with drug
treatment.
Psychostimulants (methylphenidate and amphetamine),
desipramine, and bupropion have been reported to be effec-
tive in adults with ADHD (Wilens et al 2001); however, all
studies reported to date have been small, several used
crossover rather than parallel designs, and most did not
exclude patients with comorbid psychiatric disorders that
could have confounded efficacy assessments. No study has
been large enough to provide reliable estimates of treat-
ment effect sizes or evidence that adult ADHD can be
reliably identified and studied in large populations by
different investigators. We are aware of only one study
Table 2. Efficacy Outcomes by Study, Mean (SD) Change from Baseline to End-Point
Study I Study II
Placebo
(n � 134)
Atomoxetine
(n � 133) (95% CI)
p
Valuea
Placebo
(n � 124)
Atomoxetine
(n � 124) (95% CI)
p
Valuea
CAARS-INV
Total ADHD Symptom
Score
�6.0 (9.3) �9.5 (10.1) (�5.61, �.99) .005 �6.7 (9.3) �10.5 (10.9) (�6.40, �1.49) .002
Inattentive �3.1 (5.8) �5.0 (5.7) (�3.