Acute atomoxetine treatment of younger and older children with ADHD: A meta-analysis of tolerability and efficacy

This report is based upon a meta-analysis of 6 randomized, double-blind, placebo-controlled studies of ATX that were conducted in the United States [12] between 1998 and 2004[13, 14]. Subjects were 6–16 years of age, although this analysis will focus only on children 6–12 years of age.

Subjects were assessed using the Kiddie Schedule for Affective Disorders and Schizophrenia for School-aged Children, Present and Lifetime Versions (KSADS-PL) [16], a semi-structured interview for psychiatric disorders. All subjects were required to meet the DSM-IV [1] diagnostic criteria for ADHD on the KSADS-PL, which was confirmed as the primary diagnosis by clinical assessment.

Although learning disabilities were not exclusionary, subjects were required to be of normal intelligence (IQ ≥ 80) as assessed by one of the following means: four subtests (e.g. Block Design, Picture Arrangement, Similarities, and Vocabulary) of the Wechsler Intelligence Scale for Children-3rd Edition (WISC-III) [17], the full WISC-III, or the general assessment of the physician investigator (studies HFBK, HFBD and LYAT) [13, 14]. In three studies (studies LYBG, LYBI and LYCC) [12, 18, 19], the IQ requirement was ≥70 based on the investigator's assessment of the child. Potential subjects with any serious medical illness, comorbid psychosis or bipolar disorder, history of a seizure disorder, comorbid condition requiring use of excluded concomitant medications, or ongoing use of psychoactive medications other than the study drug, were excluded.

For each subject, a parent or guardian provided written informed consent to participate and the child provided written assent, prior to receiving any study treatment or undergoing any study procedure. These studies met all federal and local regulatory requirements and were conducted in accordance with the ethical standards of each investigative site's institutional review board and the Helsinki Declaration of 1975, as revised in 2000 [20].

The primary outcome measure for all 6 studies was the ADHD-RS [21], an investigator-administered and scored instrument that includes the 18 DSM-IV symptom criteria for ADHD. Each item was rated 0–3 by the investigator during a semi-structured interview with the parent or primary caregiver. Subjects were required to have a total score or subscale score that was ≥1.5 standard deviations above age and gender norms, depending on their diagnostic subtype (e.g., total score for combined, or subscale score for primarily inattentive or primarily hyperactive/impulsive). Other measures included the CPRS-R:S [22], which contains subscales for oppositional behavior, hyperactivity and cognition, as well as an ADHD Index, and the CGI-ADHD-S [23]. The ADHD-RS and the CGI-ADHD-S were administered at each visit, while the CPRS-R:S was administered at baseline and again at the final acute treatment visit in all studies.

In 3 studies [12, 13, 19], the subjects were randomly assigned to receive either once-daily ATX or placebo (PBO) for 6 to 8 weeks. In 2 of the studies [12, 19] subjects assigned to ATX received 0.8 mg/kg/day in the morning for 3 days, after which the dose was increased to 1.2 mg/kg/day. These subjects were maintained on an "optimal" dose for 2 to 8 weeks. In the third study [13], subjects assigned to receive ATX had treatment initiated at a dose of 0.5 mg/kg/day for 3 days, after which the dose was increased to 0.75 mg/kg/day for the remainder of the first week. The daily dose was increased to 1.0 mg/kg/day after 7 days of treatment and maintained on an optimal dose for 4 to 6 weeks. In all 3 trials, subjects with significant residual symptomatology (defined as having a CGI-ADHD score ≥3) after 3 to 4 weeks of ATX treatment and without safety or tolerability contraindications could have their dose increased at physician discretion to a maximum of 1.4/1.8 mg/kg/day.

In 2 of the studies [14], design was identical, PBO or double-blinded ATX was dosed twice daily for 9 weeks duration. The titration was flexible based on therapeutic response and tolerability. Atomoxetine doses ranged from 5 to 45 mg BID, with a maximum total daily dose of 90 mg/day permitted, and a maximum weight-adjusted daily dose of 2.0 mg/kg/day. Final visit mean and median doses of ATX in these combined studies were 1.5 and 1.7 mg/kg/day, respectively.

In the final study (study LYBI) [18], subjects were randomized to receive one of three treatments, ATX, PBO, or OROS MPH, for 6 weeks during the acute treatment phase of the trial (Note: Only data from the ATX and PBO treatment groups are included in the present meta-analysis). Subjects assigned to ATX initiated treatment at a dose of 0.8 mg/kg/day divided BID for 4 days, which was then increased to 1.2 mg/kg/day. Similar to the once-daily trials, subjects with significant residual symptomatology (defined as having CGI-ADHD-S score ≥3) after 3 weeks of ATX treatment and without safety or tolerability contraindications could have their dose increased to a maximum of 1.8 mg/kg/day.

Age was dichotomized into two categories: 6–7 years; and 8–12 years. Only subjects aged 6 to 12 years were included in this analysis, since 12 was the maximum age for inclusion in all but two of the studies (in studies LYBI and LYAT the maximum age was 16) [13, 18]. Patient demographics and baseline characteristics were summarized using descriptive statistics. Change from baseline to endpoint, using a last-observation-carried-forward (LOCF) approach, was computed for all subjects with baseline and at least one post-baseline measurement. For continuous efficacy variables, treatment difference within each age group was assessed by analysis of covariance (ANCOVA) with terms for baseline, protocol, and treatment. Using the ANCOVA model, effect size (ES) was computed by subtracting the least-squares (LS) means for the PBO group from the LS means from the ATX group and dividing by the square root of the mean-squared error. In addition, consistency of treatment effect between age groups for continuous measures was assessed using an ANCOVA model with terms for baseline, protocol, treatment, age group, and treatment-by-age-group interaction.

Response was defined in two different ways in this meta-analysis: 1) ≥25% reduction from baseline in ADHD-RS total score and 2) ADHD-RS total T-scores of < 65. In addition, remission was defined as ADHD-RS total T-scores of < 60 at endpoint. For each response and remission rate, treatment differences within each age group were determined using a Fisher's exact test, while the Breslow-Day test compared odds ratios between the age categories for consistency of treatment effect across the groups. All randomly assigned subjects who took at least one dose of study drug were included in the safety analysis. Treatment-emergent adverse events (AEs) were defined as events that had newly occurred or had worsened after initiating protocol treatment. Treatment-emergent AEs were analyzed similarly to that of response rate. Although height and weight were collected, only weight data are presented due to the short duration of the studies. Pulse and blood pressure were reported for 3 of the 4 studies, but methods of collection were varied (e.g. standing and supine in one study, seated in the other three studies). For change in weight, vital signs, corrected QT interval, and laboratory parameters, treatment difference within each age category was assessed using an ANOVA model with a treatment term. Consistency of treatment effect across age groups was assessed using an ANOVA model with terms for treatment, age group, and treatment-by-age-group interaction. Since laboratory data tended not to meet normality assumptions, ranked data were used instead of raw data in these ANOVA models for all laboratory measures.

All tests were performed using a 2-sided test at a 0.05 significance level, with the exception of the treatment-by-age-group interaction tests, which were performed at a 0.10 significance level. All statistical analyses were performed using SAS software, version 8.2 [24].

Demographic characteristics for patients by each age category are presented in Table 1. There were 1,140 subjects in the pooled analysis, of which 280 (25%) were 6–7 years of age (ATX, n = 184; PBO, n = 96) while 860 patients (75%) were 8–12 years of age (ATX, n = 544; PBO, n = 316). Seventy-four percent of the subjects were male, and 71% were Caucasian. The mean ages were 7.2 years in the 6- to 7-year-old group and 10.2 years for the 8- to 12-year-old group. Seventy-three percent of all subjects met criteria for ADHD, combined subtype; 24% were classified as inattentive subtype and 2% were classified as hyperactive/impulsive subtype. There were no statistically significant demographic differences found between ATX and PBO treatment groups within each age group.

ADHD symptom severity was similar at baseline for both treatment conditions within each age group, as measured by the ADHD-RS total T score and the CGI-ADHD-S score. However, between age groups, younger subjects (6–7 years) experienced more severe ADHD symptoms at baseline compared with the older subjects (8–12 years). Mean ADHD-RS total T scores were at least 3 standard deviations above normal in each group. A higher percentage of children in the older age group met criteria for the inattentive subtype compared with those in the younger group. More children aged 8–12 years had previously been treated with a stimulant compared with their younger counterparts. Comorbid conditions were comparable, with 34% of the subjects in both age groups meeting diagnostic criteria for oppositional defiant disorder (ODD).

Table 2 summarizes the change from baseline to endpoint for ADHD-RS total score, subscale scores, and total T-score, CGI-ADHD-S score, and all four CPRS-R:S subscale scores. With the exception of the CPRS-R:S Oppositional subscale, ATX-treated subjects in both the younger and older age groups demonstrated significant improvement compared with those treated with PBO on all efficacy measures: ADHD-RS total score: younger ES = .77, older ES = .65; total T-score: younger ES = .75, older ES = .63; Inattentive subscale: younger ES = .71, older ES = .59; Hyperactive/Impulsive subscale: younger ES = .76, older ES = .62; CGI-ADHD-S score: younger ES = .62, older ES = .59; CPRS-R:S ADHD Index: younger ES = .50, older ES = .74; CPRS-R:S Cognition: younger ES = .41, older ES = .69; CPRS-R:S Hyperactive: younger ES = .56, older ES = .72). In addition, only the CPRS-R:S Oppositional subscale had a statistically significant treatment-by-age-group interaction. However, significant age group differences were observed for ADHD-RS total and subscale scores and CGI-ADHD-S score, where older children (irrespective of whether they were treated with ATX or PBO) improved significantly more than their younger counterparts.

Response rates, defined as ≥25% reduction from baseline in ADHD-RS total score, were significantly different between ATX and PBO treatment groups for children 6–7 years (ATX, 55.7%; PBO, 22.0%; p < .001) and children aged 8–12 years (ATX, 63.5%; PBO, 35.6%; p < .001). No statistically significant differential treatment effects were observed between the age groups (p = .287). Response rates, defined as having endpoint T-scores of < 65, were significantly different between ATX and PBO treatment groups for the 6–7 year olds (ATX, 44.3%; PBO, 16.5%; p < .0001) and the 8–12 year olds (ATX, 51.9%; PBO, 28.4%; p < .0001). The treatment-by-age-group effect was not significant (p = .270). The percentage of subjects experiencing remission at endpoint, as defined by T-score < 60, were significantly different between ATX and PBO for both age groups (6–7 year old ATX, 36.4%, PBO, 8.8%, p < .0001; 8–12 year old ATX, 41.0%, PBO, 19.8%, p < .0001). A significant treatment-by-age effect was seen (p = .0830) in remission rates.

Atomoxetine was well tolerated by children in both age groups. The median and mean (standard deviation [SD]) final ATX doses were 1.47 mg/kg/day and 1.39 (0.38) mg/kg/day for younger children, and 1.44 mg/kg/day and 1.37 (0.40) mg/kg/day for older children. The difference in final dose was not statistically significant. Rates of study completion were similar between the two groups (younger children, 76.4%; older children, 78.5%).

Reasons for discontinuation for subjects receiving ATX or PBO did not significantly differ within age groups, with the exception of discontinuation due to lack of efficacy. Patients who received PBO had a significantly higher rate of study discontinuation due to lack of efficacy for both younger (ATX, 1.1%; PBO, 6.3%; p = .021) and older (ATX, 2.8%; PBO, 9.5%; p < .001) children. Conversely, incidence of study discontinuation due to AEs was not significantly different between treatment groups in younger (ATX, 1.1%; PBO, 4.2%; p = .186) versus older (ATX, 3.7%; PBO, 1.6%, p = .093) children. However, a significant differential treatment effect was observed between the age groups (p = .015).

Treatment-emergent AEs reported by at least 5% of patients are presented in Table 3. Younger children taking ATX versus PBO had significantly higher rates of upper abdominal pain, decreased appetite, vomiting, and somnolence. Among older children, there were significantly higher rates of decreased appetite, somnolence, irritability, and fatigue observed for those taking ATX versus PBO. Of these treatment-emergent AEs, upper abdominal pain and vomiting had a significant treatment-by-age-group interaction. The odds ratio (OR) for treatment-emergent upper abdominal pain in younger versus older children was 3.4 and 1.2, respectively (p = 0.044); for vomiting, the OR was 7.4 and 1.4, respectively, for younger versus older children (p = 0.053). Of note, a significant treatment differential was also observed for pyrexia and cough.

Atomoxetine was associated with a statistically significant increase in mean (SD) pulse rate for both younger (ATX, 8.7 [12.7]; PBO, 1.0 [13.7]; p = .001) and older (ATX, 6.8 [11.7]; PBO, 0.6 [11.3]; p < .001) subjects. Similarly, a statistically significant treatment-group difference in systolic blood pressure (ATX, 2.1 [9.8] mmHg; PBO, 0.3 [8.1] mmHg; p = .034) and diastolic blood pressure (ATX, 2.9 [8.2] mmHg; PBO, 0.6 [8.0] mmHg; p = .002) was observed for older children, but not for younger children. There was no significant treatment-by-age-group interaction observed for either pulse rate, systolic or diastolic blood pressure. For mean (SD) weight change from baseline to endpoint, a statistically significant decrease in weight was observed for children taking ATX compared with PBO in both age groups (younger: ATX, -0.5 [1.1] kg; PBO, +0.7 [0.7] kg; older: ATX, -0.6 [1.3] kg; PBO, +1.1 [1.4]; p < .001 for both). In addition, a significant differential treatment effect between the age groups was observed for mean weight change (p = .004). There were no significant differences between the age groups or significant treatment-by-age-group interaction for mean (SD) corrected QT interval (Fridericia's method) (younger: ATX, -1.0 [21.1] msec; PBO, 0.7 [16.3] msec; p = .510; older: ATX, -0.9 [18.1] msec; PBO, -1.1 [17.5] msec; p = .862; interaction p = .485), and no clinically meaningful differences in laboratory measures.