Patients were children ages 6 through 12 recruited from 5 outpatient centers. To be eligible for the study, children had to meet the following criteria: a DSM-IV diagnosis of ADHD, any type; a rating on the ADHD Rating Scale Version IV Parent Reported-Investigator Rated version (ADHDRS-IV-Parent:Inv)  of at least 1.5 standard deviations above age and gender norms; and a severity rating of at least moderate on the Clinical Global Impressions Severity Scale (CGI-S). Additionally, they must have experienced a prior history (preceding 12 months) of insufficient response to an adequate stimulant trial, which was defined as a gradual titration of stimulant medication for at least two weeks at specified doses for each of the medication. Inadequate response was determined by the child's prescribing physician who also documented his or her opinion that a change in treatment was needed.
Children were excluded from participating in the study if they weighed less than 22 kg or more than 60 kg at study entry; had any other Axis I diagnosis, including pervasive developmental disorder, mood, or anxiety disorder; had any medical conditions that would contraindicate the use of either atomoxetine or extended-release methylphenidate, or used any concomitant psychotropic or excluded medications. The presence of comorbid oppositional defiant disorder was not an exclusion criterion. Children who had a history of intolerance or nonresponse to atomoxetine were excluded because of the ethical reason that they should not be enrolled in an augmentation study in which they have demonstrated inadequate response to both treatments. All patients were required to be free of any excluded medications for at least 5 days prior to baseline ratings and randomization.
Each site's institutional review board approved the conduct of the study, which was developed in accordance with the ethical standards of Good Clinical Practice (GCP) and the Declaration of Helsinki, as revised in 2000 . Parents or legal guardians of all subjects provided written informed consent, and the patients gave verbal assent to participate in the study.
At the initial screening visit, patients underwent a thorough diagnostic and medical examination. The diagnosis of ADHD was determined using the semi-structured clinical interview, KIDDIE-SADS-PL version , and was confirmed by a child psychiatrist. The medical evaluation included medical history, physical exam, routine chemistry, hematology, urinalysis, and electrocardiograms.
During the course of treatment, safety was assessed by obtaining vital signs, weight, spontaneous adverse event reports, and concomitant medications at each visit. Illness severity measures included the ADHDRS-IV-Parent:Inv; Clinical Global Impression Scale – Improvement (CGI-I)  ratings; the Weekly Parent Ratings of Evening and Morning Behavior-Revised (WREMB-R); and the Conners Parent Rating Scale Revised, Short-Form (CPRS-R:S) . With the exception of the WREMB-R, these outcome measures have been established as valid and reliable within the ADHD field, and they are scored so that higher scores indicate greater symptom severity. The WREMB-R is a more recent instrument in which a parent rates 11 behaviors for their severity during the morning and evening hours, and it has shown sensitivity to treatment effects in prior clinical trials . Overall improvement also was rated by clinicians using the CGI-I, which consists of a 7-point scale where 1 = "very much improved" and 7 = "very much worse". The safety and efficacy measures were repeated again at the end of treatment or when patients discontinued the study.
The study design consisted of 3 sequential phases: an evaluation/screening phase, a double-blind 4-week acute treatment (study phase 1); and a 6-week, double-blind, combination treatment phase (study phase 2). At visit 2 (treatment week 1), all patients were started on open-label atomoxetine and given a pill placebo. Atomoxetine was titrated to a target dose of 1.2 mg/kg/day (maximum dose 1.4 mg/kg/day). Neither the investigator nor the patients knew when the onset of active augmentation would occur as the investigator's protocol did not specify the timing of active augmentation. Another protocol that specified the onset of augmentation was mailed directly to the investigator's institutional review board for full disclosure. After 4 weeks on placebo, patients' illness severity was compared with their baseline using the above scales. If patients were rated on the CGI-I scale as 1 or 2 (much or very much improved), they were classified as remitters and were maintained on placebo. If patients continued to have substantial symptoms, they were then randomly assigned via an interactive voice response system to receive either extended-release methylphenidate or placebo. Dose for OROS methylphenidate was titrated to a target dose of 1.08 mg/kg/day (maximum dose 1.2 mg/kg/day). All patients continued with their open label atomoxetine dose during the active augmentation treatment phase. During study phase 1, patients were seen after 14 days and then weekly for 2 visits (4 weeks total). During study phase 2, they were seen weekly for 2 weeks, and then 1 month later (6 weeks).
The power calculations for the study were based on the primary objective of safety by estimating the number of patients who would be required to demonstrate a safety signal in categorical changes in vitals. Based on the assumption of 5% of patients would have a categorical change in vitals on atomoxetine, 85 subjects were estimated to be required for an 80% power to detect a eight-fold increase in categorical changes. However, it was difficult to find a sample of children who had stopped stimulant treatment, were not intolerant of stimulants, and had not already been treated with atomoxetine. Thus, the final sample size was 25, which caused the study to be underpowered to detect categorical differences between groups.
Safety and tolerability outcomes were reported with frequency counts. Paired t tests were used to examine whether mean changes from baseline to the end of study phase 1 and mean changes from baseline to study phase 2 were significantly different from zero. Categorical changes in vital signs were defined as follows: (1) for diastolic and systolic blood pressure, an increase of at least 5 mmHg to above the 95th percentile based on age, gender, and height-adjusted National Institute of Health norms ; (2) for pulse, an increase of at least 25 to a value of at least 110 bpm.
Efficacy outcome measures were conducted on the intent-to-treat sample using a last-observation-carried-forward method. Patients were classified into 3 groups: atomoxetine/methylphenidate, atomoxetine/placebo excluding early responders (who were not randomized), and atomoxetine/placebo including early responders. Efficacy was analyzed using a repeated measure analysis of covariance (ANCOVA) comparing changes on the ADHDRS-IV-Parent:Inv total score from initial baseline, at the end of study phase 1, and the end of study phase2. The initial baseline score was the covariate, and treatment and investigator were fixed effects. Effect sizes were also calculated on the secondary outcome measures to determine the overall treatment response (10-week treatment from baseline to study end-point) and the incremental effect size (6-week double-blind randomization to study end-point).
Efficacy was examined descriptively by classifying patients individually based on their T-score obtained from the ADHDRS-IV-Parent:Inv total at the end of study phase 1 and at the end of study phase 2. Patients were classified as normalizers if their scores at the end of the study phases were within 1 standard deviation of the normal range (i.e., T-score ≤ 60). The frequency of normalization was then summed to determine the rate of those who did not improve, who transiently improved (normalized at end of study phase 1, but not study phase 2), or obtained/maintained improvement (normalized at end of both study phases).