Study Design

This open-label, non-randomized, parallel-design pilot study was conducted from October 2003 to March 2006 at 12 centers in the United States. The study protocol was approved and conducted in accordance with the principles of the Declaration of Helsinki, and all parents or legal guardian(s) of the patients provided written informed consent (and patients provided assent where applicable) after the procedure(s) and possible side effects were fully explained. All patients meeting criteria received open-label treatment with atomoxetine at doses ranging from 1.0-1.4 mg/kg once daily given orally as capsules (Strattera^®, Eli Lilly and Company, Indianapolis, IN, USA) for approximately 16 weeks. The maximum dose prescribed was 1.4 mg/kg or 100 mg, whichever dose was less. After initiating treatment, patients were assessed every 2 weeks for 8 weeks and then once a month for the remaining 8 weeks of the study.

Patients

Male and female patients, aged 10 to 16 years, meeting the DSM-IV diagnosis of ADHD and/or ADHD with comorbid dyslexia (ADHD+D) were enrolled. In addition to meeting DSM-IV diagnosis criteria for ADHD, patients in the ADHD-only and ADHD+D treatment groups were required to meet Kiddie Schedule for Affective Disorders and Schizophrenia for School-Aged Children-Present and Lifetime, Behavioral Disorders Supplement [17] module criteria for ADHD. Further, during the screening visits, patients were required to have an ADHD symptom severity score at least 1.5 standard deviations above age and gender norms for at least 1 of the diagnostic subtypes (inattentive or hyperactive/impulsive) or the total score for the combined subtype as assessed by the Attention-Deficit/Hyperactivity Disorder Rating Scale-IV-Parent Version: Investigator Administered and Scored (ADHDRS-IV) [18]. Patients diagnosed as ADHD+D were additionally required to have at least a 22-point discrepancy (representing a 1.5 standard deviation discrepancy) between ability (using the highest intelligence quotient score of the Vocabulary [verbal] subtest standard score [SS]; the Matrices [non-verbal/performance] subtest SS; or the IQ composite score on the Kaufman Brief Intelligence Test [K-BIT] [19]) and achievement (reading composite SS on the Kaufman Test of Educational Achievement [K-TEA] [20]). An IQ composite score of ≥80 was required on the K-BIT.

Patients were excluded for any of the following reasons: weight less than 25 kg or greater than 70 kg at study entry; any current or previous diagnosis of bipolar I or II disorder or psychosis; autism, Asperger's syndrome, or pervasive developmental disorder; serious suicidal risk; serious medical illness or clinically significant laboratory abnormalities, hospitalization, or an excluded medication during the course of the study; a history of substance abuse or dependence within the past 3 months (excluding nicotine and caffeine); a positive urine drug screen for any substances of abuse; and treatment with a monoamine oxidase inhibitor within 14 days prior to baseline.

Concomitant medications with primarily central nervous system activity were not allowed. Medications that are strong CYP2D6 inhibitors or substrates were not permitted. Chronic use of cough and cold medications containing pseudoephedrine or the sedating antihistamine diphenhydramine were not allowed. Narcotic use was not permitted unless special circumstances arose (e.g. limited use post-operatively, etc) and approval of the Lilly physician or designee was granted. Patients on methylphenidate or another prescribed stimulant for the treatment of ADHD were required to be stimulant-free 24 hours prior to obtaining baseline measures and to subsequently discontinue medication 1 day prior to the last screening visit before dispensation of study medication.

Efficacy Measures

The primary objective was to assess the effect of atomoxetine on patients with ADHD+D as measured by the mean change from baseline on the ADHDRS-IV. Secondary measures included comparison between the ADHD and ADHD+D groups on mean change in the ADHDRS-IV total and subscale scores; K-TEA measures (Reading Decoding, Reading Comprehension, Spelling subtests, and Reading Composite scale) [20]; the Working Memory Test Battery for Children (WMTB-C) [15]; and the Life Participation Scale for ADHD-Child Version: Investigator- and Parent-Rated versions (LPS-C) [21]; as well as the correlation between ADHDRS-IV and both the WMTB-C and K-TEA. The efficacy measure, ADHDRS-IV, was assessed at every scheduled visit, whereas the K-TEA and WMTB-C were assessed at approximately every other clinic visit in order to minimize the test/re-test phenomenon. Inter-rater reliability testing of clinicians administering the ADHDRS-IV was performed to ensure consistency among sites. Further, only psychologists experienced with the administration of the educational tests were permitted to administer and score the K-TEA, K-BIT, and WMTB-C.

The K-BIT is a brief, individually administered measure of verbal and non-verbal intelligence. The test is composed of 2 subtests: Vocabulary and Matrices. Vocabulary measures verbal, school-related skills (crystallized thinking) by assessing a person's word knowledge and verbal concept formation. Matrices measure non-verbal skills and the ability to solve new problems (fluid thinking) by assessing an individual's ability to perceive relationships and complete analogies [19].

The K-TEA is an individually administered measure of the school achievement of children and adolescents in grades 1 through 12. Age equivalents were used for this study, as they provided a more appropriate indicator of improvement for children with disabilities. The K-TEA Comprehensive Form measures reading decoding and comprehension, spelling, and mathematics applications and computation. The Comprehensive Form subtests used for this study were Reading/Decoding, Reading/Comprehension, and Spelling. The sum of the subtest raw scores (Reading Decoding and Reading Comprehension) make up the Reading Composite score, which is transformed to a standard score that indicates age equivalency [20].

The WMTB-C consists of 9 subtests designed to reflect 3 main components of working memory: central executive, phonological loop, and visuo-spatial sketchpad. The frontal regions of both hemispheres of the brain are associated with the central executive functions of coordinating, processing and storage, controlling flow of information through working memory, and attentional control. The 3 central executive (CE) subtests include: Backward Digit Recall, Listening Recall, and Counting Recall. The phonological loop, located in the temporal lobes of left hemisphere, is associated with functions of temporary storage of material in a phonological (sound-based) form, which includes spoken language and both written language and pictures. The 4 subtests designed to measure phonological loop function are: Digit Recall, Word List Matching, Word List Recall, and Non-word List Recall. The visuo-spatial sketchpad, located in the right hemisphere, is associated with functions of storage of materials in terms of visual or spatial features (non-verbal information). Two subtests tap visuo-spatial sketchpad function: Block Recall and Mazes Memory [15].

Safety

Safety measures recorded at every visit included spontaneously reported treatment-emergent adverse events (TEAEs) and vital signs. Blood for chemistry and hematology laboratories were collected at baseline, after 4, 6, and 10 weeks of treatment, and at the end of the 16-week treatment period. Electrocardiograms were collected at baseline, after 4 weeks of treatment, and at discontinuation of the study.

Statistical Methods

The primary measure, the determination of significant improvement from baseline on the ADHDRS-IV total score for patients with ADHD+D, was analyzed using a Student's t-test applied to the least squares mean change from baseline score. Change scores were computed for each patient as the difference between the last observation carried forward (LOCF) score and baseline score. The least squares mean change and associated standard error used in the Student's t-test were derived from an analysis of covariance (ANCOVA) model, with terms for diagnostic group, investigator, gender, age, baseline score, and baseline score-by-diagnostic group interaction. All patients with comorbid ADHD+D and at least 1 baseline and 1 post-baseline score were included in the primary analysis.

Between-group changes from baseline to endpoint in efficacy measure variables were analyzed using a fixed-effects ANCOVA model, with terms for diagnostic group, investigator, gender, baseline score, age, and baseline score-by-diagnostic group interaction. Type III sums of squares were used for between-group tests. Changes within diagnostic group were assessed using Student's t-test applied to the least squares mean for the diagnostic group from the ANCOVA model. There were no adjustments made for the number of tests conducted.

Between-group changes in efficacy measure variables over time were analyzed using the relevant contrast from a restricted maximum likelihood repeated measures model, with terms for diagnostic group, investigator, visit, baseline score, diagnostic group-by-visit interaction, and baseline score-by-diagnostic group interaction. This model used the covariance structure that maximizes Schwartz's Bayesian Criterion and the Kenward-Roger method for estimating denominator degrees of freedom. Student's t-test was applied to the least squares mean and standard error to estimate within-group change to endpoint based on the relevant contrast from the diagnostic group-by-visit interaction from this model.

Incidence of categorical response variables were compared across diagnostic groups using Fisher's exact test. Correlations between change and baseline (Visit 1) scores for LPS-C Investigator- and Parent-rated scales were computed to assess the consistency of responses using alternate sources for rating the patient's behavior. Pearson correlation coefficients were calculated to examine the relationships between the changes of selected efficacy measures. Tests were two-tailed.

Baseline Characteristics

A total of 134 patients were screened, and 56 patients met criteria for ADHD (n = 20) and ADHD+D (n = 36). A total of 47 patients, 16 in the ADHD group and 31 in the ADHD+D group, completed the study. Figure 1 displays the patient disposition. The patient population for this study was predominantly male (70%), and the median age was 12.6 years and 11.4 years for the ADHD and ADHD+D groups, respectively. Demographic characteristics, including weight, origin, and age, were similar between both groups, with the exception of height. The ADHD subtype was comparable between groups, with most patients diagnosed as predominantly combined (54%) or inattentive (43%) subtype. Whereas the K-BIT IQ composite mean baseline score was comparable between both groups, the ADHD group had a statistically significantly lower K-BIT matrices mean subtest standard score (99.9 ± 13.6) compared with the ADHD+D group (109.5 ± 10.3; p = .004). The K-TEA mean baseline scores were numerically similar between groups (p ≥ .118). The mean final prescribed dose was 1.29 mg/kg/day. Table 1 presents the patient baseline demographics.

Efficacy

The primary efficacy measure was the ADHDRS-IV total score. Similar statistically significant mean change improvement was demonstrated in both the ADHD and ADHD+D groups on ADHDRS-IV total score (-20.2 ± 2.8 and -17.7 ± 2.5, respectively; p < .001 for both groups) and the subscores for inattention (-11.0 ± 1.6 and -10.4 ± 1.4, respectively; p < .001 for both groups) and hyperactive/impulsivity (-8.5 ± 1.4 and -7.7 ± 1.2, respectively; p < .001 for both groups). There was no differentiation of response between the 2 groups on total score and the inattentive and hyperactive/impulsivity subscores (p = .503, p = .769, p = .660, respectively). Figure 2 presents the ADHDRS-IV total score mean change over time, which further supports improvements in both groups at every time-point throughout the study (p < .001).

The secondary efficacy outcome measure of the LPS-C showed statistically significant adaptive function improvement and agreement between the investigator-rated and parent-rated versions of the scale. Mean change from baseline to endpoint was statistically significant for both groups, though there were no differences between groups.

A by-visit analysis of the least squares (LS) mean changes for the WMTB-C revealed a statistically significantly greater improvement at the last week for patients with ADHD over the ADHD+D group on the CE standard (p = .003) and component scores (p = .012; Figure 3). Subtests of the CE function tests for the ADHD group revealed statistically significant improvements on the listening recall after 8 and 16 weeks of treatment (p ≤ .04) and backward digit recall mean scores at the end of treatment (p = .002). The ADHD group also gained improvements on the PL subtests for non-word list recall after 8 weeks of treatment, which continued to improve until the end of treatment (p < .04). Conversely, patients with ADHD+D demonstrated statistically significant improvement on the PL total standard and component score after only 4 weeks of treatment and continued to improve at every visit throughout the study. PL subtests for the ADHD+D group showed statistically significant gains for the ADHD+D group after 8 weeks of treatment that continued to the end of treatment for word list recall (p < .02) and non-word list recall (p < .01) tests, and statistically significant improvement on the PL digit recall subtest was realized by the end of the study (p < .04). Finally, the baseline-to-endpoint analyses did not reveal significant changes for the VSP component or standard scores for either group. However, the repeated measure analyses (by visit) of the VSP component and standard scores demonstrated statistically significant improvement for patients in the ADHD+D group but only at the end of 16 weeks of treatment (p < .03).

Pearson correlations between the ADHDRS-IV and K-TEA and WMTB-C were calculated. Only the ADHDRS-IV total score, and hyperactivity and inattentive subscores were statistically significantly correlated to the K-TEA Reading Comprehension standard score, though the correlation coefficients were weak (r = .33, r = .31, and r = .28, respectively). There were no statistically significant correlations between the ADHDRS-IV total score or subscores and any other K-TEA measures or to any WMTB-C components.

Safety

There were no serious adverse events reported. A total of 4 (7.1%) patients of the 56 randomized to the trial discontinued due to adverse events, which included nausea, mood swings, and abdominal pain. Adverse events occurring in at least 5% of all patients in the study were somnolence (n = 19), nausea (n = 17), decreased appetite (n = 12), headache (n = 11), nasopharyngitis (n = 7), upper abdominal pain (n = 11), vomiting (n = 9), cough (n = 4), upper respiratory tract infection (n = 3), constipation (n = 3), irritability (n = 5), psychomotor hyperactivity (n = 3), fatigue (n = 6), and abdominal pain (n = 3). There were no differences between groups in the occurrence of any reported adverse event. Likewise, ECGs, laboratory analytes, vital signs, height, and weight evaluations revealed no clinically significant changes from baseline to endpoint.