A Meta-analysis of the Consistency of Atomoxetine Treatment Effects in Pediatric Patients with Attention-Deficit/Hyperactivity Disorder from 15 Clinical Trials Across Four Geographic Regions

Abstract

Objective:

Atomoxetine has been approved as a treatment for children and adolescents with attention-deficit/hyperactivity disorder (ADHD) in the United States, throughout Europe, and in other countries. This meta-analysis was to assess the consistency of the treatment effect of atomoxetine across four global geographic regions.

Methods:

Data from 15 acute, double-blind, placebo-controlled trials were pooled (2 in Asia, 4 in Europe, 8 in North America, and 1 in Russia), yielding 2569 pediatric patients with ADHD. Improvements during 6–10 weeks of atomoxetine treatment were evaluated using the ADHD Rating Scale-IV or the Swanson, Nolan, and Pelham Scale–Revised. Consistency across regions was assessed by an interaction test and Higgins I² . Consistency of one region versus other regions was assessed by effect sizes of individual regions and pairwise differences.

Results:

Patient demographics were generally similar across regions. More patients from Asia met diagnostic criteria for ADHD inattentive subtype and fewer for combined subtype compared with patients from Europe, North America, or Russia. Asian patients had a lower mean baseline ADHD total score and mean hyperactivity/impulsivity subscore. Treatment effects showed marginal inconsistency and moderate heterogeneity among the regions (percentage of patients achieving a 40% decrease from baseline ADHD scores, atomoxetine versus placebo: Asia 39.6%, 24.0%; Europe 40.2%, 12.1%; North America 45.3%, 21.7%; Russia 54.2%, 33.3%). Inconsistency was observed primarily in Asia versus the other regions. Completion rates with atomoxetine were higher in Asia and Russia (94.4% and 94.3%, respectively) than in Europe (84.3%) or North America (80.4%).

Conclusions:

Atomoxetine was demonstrated as an effective treatment for ADHD in 15 clinical trials from four global regions. The current meta-analysis has revealed a degree of heterogeneity in treatment efficacy across regions, most notably in the comparison of Asian patients relative to those from the other regions.

Introduction

A tomoxetine was first approved on November 26, 2002 as a nonstimulant treatment for attention-deficit/hyperactivity disorder (ADHD) in children between 6 and 12 years of age and in adolescents, ages 13–17 years. This selective noradrenergic reuptake inhibitor is currently marketed in 80 countries worldwide, with >10,000,000 patients estimated to have received atomoxetine since its first approval (data on file, Eli Lilly and Company and/or one of its subsidiaries). Although the efficacy and safety of atomoxetine has been established in numerous country-specific and international clinical trials (Michelson et al. 2001; Spencer et al. 2002; Weiss et al. 2005), potential regional variations in response rates and tolerability may exist because of biological, social, or economic factors or differences among countries in their clinical practices or the diagnostic criteria that they use.

It is often necessary to be able to determine to what extent the data from patients in one region can be generalized to patients in another region. The ICH (International Conference on Harmonization) E5 guideline (International Conference on Harmonization 1998) was adopted in 1998 to recommend a framework for evaluating the impact of ethnic factors on drug effects. For compounds that might be sensitive to ethnic factors, if a sponsor can show evidence of similarity of treatment effects across regions through a bridging study, results can be extrapolated across regions. It has been demonstrated, for example, that geographic variability exists in the prevalence of and treatment procedures for ADHD (Polanczyk et al. 2007; Hinshaw et al. 2011). Similarly, drug metabolism varies considerably among regions because of differences in the distribution of genetic polymorphisms of cytochrome P450, the principal oxidative enzyme for atomoxetine, with “poor metabolizers” representing 5–10% of Europeans but<1% of Japanese (Idle and Smith 1979; Wilson et al. 2001). It is well known that the rate of clearance of atomoxetine in cytochrome P450 2D6 poor metabolizers is approximately one tenth that of extensive metabolizers (Sauer et al. 2003) and are therefore exposed to greater concentrations of the drug and run a greater risk of adverse reactions. Cytochrome P450 2D6 poor metabolizers are much more prevalent among European Caucasians than among east Asians (Idle and Smith 1979). Because potential geographic differences may have a biological basis, some countries, for example Japan and Taiwan, require that some form of “bridging” study be carried out prior to market approval to demonstrate that patients from those countries exhibit treatment responsiveness and tolerability equivalent to those previously found among patients from elsewhere. In this regard, it is not clear to what degree patients with ADHD differ among different regions around the world with respect to their disease state or their responsiveness to treatment. As a step toward addressing this topic, the current exploratory meta-analysis was conducted using a number of approaches that have been summarized for possible application by the Consistency Workstream of the Pharmaceutical Research and Manufacturers Association (PhRMA) Multi-Regional Clinical Trials Key Issues team (see review by Chen et al. 2010) for examining the geographic consistency of treatment-effect data among patient samples taken from different regions around the world. Specifically, responses to atomoxetine were investigated among children and adolescents with ADHD who were from one of four geographic regions: Asia, Europe, North America, and Russia.

Methods

Studies in the analysis

Among the clinical trials conducted for atomoxetine by Eli Lilly and Company, the 15 double-blind, placebo-controlled, pediatric clinical trials in the clinical development database (Table 1) were selected, as these trials 1) included children or adolescents from 6 to 17 years of age who had been diagnosed with ADHD, per the Diagnostic and Statistical Manual of Mental Disorders, 4th ed. (DSM-IV) (American Psychiatric Association 1994), and 2) had a placebo-controlled treatment period of 6–10 weeks. Each of these trials had demonstrated significant reduction of ADHD symptoms by atomoxetine treatment relative to placebo during the acute treatment period. The data from each trial were independently extracted and pooled for this analysis. Representing the Asian region, one study was conducted in Taiwan and one in Japan, for a total sample of 288 patients. Two studies were conducted in Germany, one in Sweden, and one more broadly in European Union (EU) countries (Belgium, Denmark, Finland, Germany, Netherlands, Spain, and the United Kingdom). The latter study also included some patients from Australia. Because of their limited number (n=33), as well as the racial and cultural similarity of Australians and Europeans, the Australian patients were included in the sample from Europe, which thereby comprised a total of 629 patients. One study was conducted in Russia, involving 105 patients. Because of Russia's unique status as both a European and Asian country with large representations within its population from either continent, its data are analyzed separately from those of the European and Asian regions. Finally, eight studies were conducted in North America, primarily at sites located in the United States, but also some in Canada and Puerto Rico, with a total of 1605 patients.

Table 1.

List of Studies Included in the Analysis

Study code	Registry number^a	Location	Start/End dates	n	Dose	Effect size
B4Z-MC-HFBD	b	US	Nov 1998/Feb 2000	ATX=65, PLA=62	1.6^c	0.79
B4Z-MC-HFBK	b	US	Feb 1999/Oct 1999	ATX=64, PLA=62	1.5^c	0.66
B4Z-MC-LYAC	b	US	Apr 2000/Feb 2002	ATX=213, PLA=84		0.60
B4Z-MC-LYAT	b	US	Feb 2001/May 2001	ATX=85, PLA=86	1.31^c	0.68
B4Z-MC-LYAW	b	US	Jan 2002/May 2005	ATX=101, PLA=52		0.63
B4Z-US-LYBG	b	US	Mar 2002/Apr 2003	ATX=133, PLA=64		0.72
B4Z-MC-LYBI	b	US	Aug 2002/Sep 2003	ATX=222, PLA=74	1.45^c	0.57
B4Z-US-LYCC	NCT00486122	US	Oct 2003/Jun 2004	ATX=195, PLA=93		0.43
B4Z-TW-S010	NCT00485459	Taiwan	Feb 2004/Nov 2004	ATX=72, PLA=34		0.46
B4Z-JE-LYBC	NCT00191295	Japan	Feb 2005/Sep 2006	ATX=183, PLA=62		0.41
B4Z-SB-LYDV	NCT00546910	Germany	Oct 2007/May 2009	ATX=63, PLA=62		0.87
B4Z-SB-LYDW	NCT00406354	Germany	Nov 2006/Jan 2009	ATX=121, PLA=59		0.69
B4Z-MC-LYBX	NCT00191698	EU/Australia^d	Dec 2003/Apr 2004	ATX=156, PLA=70		0.51
B4Z-SO-LY15	NCT00191542	Sweden	Mar 2005/Aug 2006	ATX=49, PLA=50		1.18
B4Z-MW-LYCZ	NCT00386581	Russia	Aug 2004/Feb 2005	ATX=72, PLA=33		0.61

ClinicalTrials.gov registry number.

Not available: predates CTR listing requirement.

Flexible dosing, mean final dose (mg/kg/day).

Australia, Belgium, Denmark, Finland, Germany, Netherlands, Spain, United Kingdom.

ATX, atomoxetine; CTR, Clinical Trial Registry; EU, European Union; PLA, placebo.

Atomoxetine was given in doses between 0.5 and 1.8 mg/kg/day, given once daily or as a split dose twice daily (Table 1). All but 2 of the 15 trials included a flexible dosing design that allowed the investigating physician to titrate doses of atomoxetine as needed for optimal treatment efficacy and tolerability. Prior to participation, patients and their parents or guardians were provided with a complete description of the study, and each patient signed a child assent form, if applicable, while the parent or legal guardian signed an informed consent document approved by the study site's institutional review board. All studies were conducted in accordance with the ethical standards of the Declaration of Helsinki (World Medical Association 2008).

Assessments

Evaluations of treatment efficacy were based on the data from the first 6 weeks of all studies. Treatment efficacy was evaluated using investigator-scored symptom-assessment measures. For most studies, the total score of the Attention-Deficit/Hyperactivity Disorder Rating Scale (ADHD-RS)-IV (DuPaul et al. 1998) was used. However, the European/Australian study and one of the two German studies used a very similar scale, the Swanson, Nolan, and Pelham Scale–Revised (SNAP-IV; Swanson 1992). Like the Attention-Deficit/Hyperactivity Disorder Rating Scale-IV, the SNAP-IV comprises 18 items based on the DSM-IV, each item scored from 0=“not at all” to 3=“very much.” Because the total scores from the two measures are considered equivalent, pooled analyses are based on the two measures. Accordingly, the term “ADHD-RS” is used throughout the rest of this report to denote both of these efficacy scales collectively.

Response rate analysis was based on a 40% decrease from baseline in ADHD-RS total scores. This level of reduction was chosen because the treatment response and effect was considered a priori to be stronger and more reliable using 40%, rather than 25% or 30%, and hence considered more robust for assessing regional consistency. In addition, a 40% improvement is closer to the standard used for defining a robust improvement (Spencer et al. 1996).

Statistical methods

In considering the assessment of geographic consistency, there are two different research issues to consider: an overall global consistency across all regions, and a region-specific assessment of consistency (that is, one region relative to all others). In this report, we explore each of these questions using different methods. Most of these various consistency assessment concepts and methods are already well described in detail by Chen et al. (2010). As this analysis is based on empirical observation and is largely exploratory, the interpretive emphasis is on a descriptive approach, with analytical statistics intended primarily to provide additional support for conclusions from the observed results.

Some methods are developed for meta-analysis and some are for multiregional clinical trials or within studies. Some methods are parameter based, which will be tested with p values, whereas others will be observed outcomes based (that is, no p values). Some are based on quantitative assessment (variation in the magnitude of treatment effect, but not the direction) or on qualitative assessment (variation in the direction of treatment effect). Caution needs to be exercised in interpretation from different methods, as some methods that are based on p values (for example, t tests and the interaction test) assess only whether there is a hint of heterogeneity, but the actual magnitude of heterogeneity may be of less importance.

Consistency across regions

To assess the consistency of treatment effects of atomoxetine across all four regions (Asia, Europe, North America, and Russia), three approaches were taken: 1) The guidelines on multiregional clinical trials, Method 2, provided by the Japan Ministry of Health, Labour, and Welfare (MHLW) in their Basic Principles on Global Clinical Trials (Ministry of Health Labour and Welfare of Japan 2007); 2) the Higgins I² quantity (Higgins et al. 2003); and 3) analysis of treatment-by-study interaction (Cox 1984).

Method 2 from the Japan MHLW was explored for the qualitative assessment of consistency in this article, as this was a recent Japanese regulatory guideline on considering how to justify the sizes of patient samples from Japan in a multiregional clinical trial based on this criterion. Within the context of the current analysis, this method seeks to answer the question, “Is the intervention effective in all regions?” To determine the appropriate number of subjects for a Japanese sample, the overall difference between the treatment group and the placebo group is defined as D_all , and the differences between the treatment and placebo groups in each region are D₁ , D₂ , D₃ , and so forth. Sample sizes are determined such that D₁ , D₂ , D₃ … show a similar tendency; that is, D_k >0, where k=1,…, K for a total of n=K regions. Consistency of treatment effect is concluded if a positive trend is observed; that is, that D_k >0 for all samples.

The Higgins I² method is an alternative to the use of Cochran's Q (Patil 1975), which follows a χ² distribution, and seeks to answer the question, “Is there heterogeneity across these multiple samples?” I² describes the percentage of total variation across samples that is the result of heterogeneity, rather than chance, and is calculated as I² =100%×(Q − df)/Q, where Q is Cochran's heterogeneity statistic, and df is the degrees of freedom. Negative values of I² are set to zero so that I² falls between 0% and 100%, with 0% indicating no observed heterogeneity and 100% indicating maximum possible heterogeneity.

Analysis of treatment-by-study interaction is based on a χ² test that combines individual estimates from all regions. This approach seeks to answer the question, “What is the statistical significance of the detected heterogeneity?” Consistency of treatment effects is concluded when no significant interaction is detected.

Consistency of one region relative to others

The consistency of one region versus the other regions was assessed by analysis of effect sizes of individual regions and of pairwise differences. This involved five analytical approaches: 1) Method 1 from the Japan MHLW's Basic Principles on Global Clinical Trials (Ministry of Health Labour and Welfare of Japan 2007); 2) overlapping coefficients (proportions of similar responses [Rom and Hwang 1996]); 3) non-inferiority (Snapinn 2000); 4) bioequivalence (Rani and Pargal 2004); and 5) t test (Student 1908). These approaches examine each observation (that is, region) successively, comparing the data for that observation versus all of the others combined. The first four approaches seek to answer the question, “How homogeneous are these observations?”.

Method 1 from the Japan MHLW examines the estimated treatment effects D_all and D_J for all patients and the Japanese cohort, respectively, where D is again the difference between the placebo and study drug groups, D_all is the difference in the entire study population across regions, and D_J is the difference within the Japanese subpopulation. The number of Japanese subjects is determined such that D_J /D_all >π will occur with a probability of ≥80%. Accordingly, π should be set as an appropriate value, with a value of ≥0.5 generally recommended; that is, the observed treatment effect for Japanese patients is at least half of that observed for all other patients.

Analysis of overlapping coefficients examines the distributions of the responses under two conditions (for example, treatments or locations). When the two distribution curves are plotted, the area common to the two conditions, referred to as the proportion of similar responses (PSR), will be closer to 1 as the two treatment effects (or, in our situation, country effects) become more similar, and closer to 0 as they become less similar.

To further explore the assessment of consistency, a non-inferiority approach was applied. Non-inferiority examines the lower and upper limits of a one-sided confidence interval and determines whether it is less than or greater than the equivalence margin, δ, where δ is prespecified.

Similarly, the bioequivalence approach (U.S.Food and Drug Administration 2001) was applied. The general approach is to construct a two-sided confidence interval for the quantity μ_T–μ_R, where μ_T is the population average response for a test drug, and μ_R is the population average response for the reference drug. From this, an average bioequivalence can be concluded if this confidence interval is contained within the specified interval [–θ_A, θ_A], where θ_A is prespecified.

Results

Patient characteristics

Patient demographics were generally similar across regions, but some variation in patient characteristics was noted. A significant difference in the ratio of females to males was noted (p<0.001, χ² test), accounting for 25.4% of patients in North America, compared with a range of 13.2–14.3% in the other regions. A difference (p<0.001, χ² test) was also detected in the proportion of children (6–12 years) to adolescents (13–17 years), with adolescents accounting for 7.2% of patients in North America and 19.1% in Russia, with the proportions in Asia (13.9%) and Europe (12.2%) being intermediate. The distribution of patients by diagnostic category also differed among regions (p<0.001, χ² test), with 49.3% of Asian patients having the predominantly inattentive subtype, compared with 16.4–26.6% in the other regions, and 47.9% having the combined subtype, compared with 71.3–77.9% among the other regions. Overall baseline mean ADHD-RS scores differed significantly (p<0.001, analysis of variance). Although Asian patients displayed a mean baseline ADHD-RS Inattention subscale score (mean: 20.0, SD: 3.8) comparable to those of patients from the other regions (range of means: 20.3– 22.1), their ADHD-RS Hyperactivity/Impulsivity subscale scores (mean: 13.9, SD: 6.5) were lower than in the other regions (range of means: 17.4–19.8). Consequently, their ADHD-RS total scores (mean: 33.9, SD: 8.3) were also lower than in the other regions (range of means: 37.7–40.2). Unfortunately, prior use of stimulants was not consistently characterized in the studies conducted in the EU region or Russia, and arriving at definitive conclusions about regional differences in stimulant use was therefore not possible.

Efficacy

After 6 weeks of acute treatment, atomoxetine improved ADHD symptoms to a significantly greater extent relative to placebo in all four regions (Tables 2 and 3). Baseline-to-endpoint changes on the ADHD-RS and treatment effect sizes were notably smaller among Asian patients relative to the other regions, particularly for ADHD-RS total and Hyperactivity/Impulsivity subscale scores. Conversely, North America and especially Europe appeared to have larger treatment responses than the other regions. Of note, the magnitude of responses to atomoxetine treatment were numerically highest in the Russian patients, but responses to placebo were likewise numerically highest among Russian patients, with the result that the treatment effect was similar to those seen in the European and North American patients. Examination of confidence intervals of the differences in change scores (Table 3), particularly the most conservative 90% intervals, underscored the contrast between the responses of patients from Europe and North America, at the “most responsive” end of the scale, and Asia at the other end, with Russia taking an intermediary position and slightly overlapping with Asia.

Table 2.

Changes in ADHD-RS Scores from Baseline to Endpoint

		ATX		PLA
ADHD-RS measure	Region	n	Change mean (SD)	n	Change mean (SD)	Difference ATX – PLA	Effect size (SE)	Treatment p value
Total score	Asia	192	−10.9 (8.6)	96	−7.1 (9.3)	−3.7	0.42 (0.39)	<0.001
	Europe	383	−13.2 (12.6)	239	−5.1 (9.8)	−8.1	0.70 (0.34)	<0.001
	N. Am.	997	−14.6 (13.4)	557	−6.7 (11.3)	−7.9	0.63 (0.23)	<0.001
	Russia	72	−16.5 (10.5)	33	−10.6 (7.6)	−5.9	0.61 (0.70)	<0.001
	Overall	1644	−13.9 (12.7)	925	−6.4 (10.6)	−7.5	0.62 (0.17)	<0.001
Hyperactivity/impulsivity subscore	Asia	192	−4.8 (4.5)	96	−2.9 (4.6)	−1.8	0.40 (0.23)	<0.001
	Europe	383	−6.6 (6.8)	239	−2.3 (5.5)	−4.2	0.67 (0.20)	<0.001
	N. Am.	997	−7.1 (7.0)	557	−3.1 (5.9)	−4.0	0.60 (0.13)	<0.001
	Russia	72	−7.9 (5.7)	33	−4.1 (4.5)	−3.8	0.72 (0.42)	<0.001
	Overall	1644	−6.7 (6.7)	925	−2.9 (5.6)	−3.8	0.60 (0.10)	<0.001
Inattention subscore	Asia	192	−6.1 (5.4)	96	−4.2 (5.5)	−1.9	0.35 (0.26)	<0.001
	Europe	383	−6.6 (6.5)	239	−2.7 (5.2)	−3.9	0.64 (0.19)	<0.001
	N. Am.	997	−7.5 (7.5)	557	−3.6 (6.5)	−4.0	0.55 (0.14)	<0.001
	Russia	72	−8.6 (5.6)	33	−6.6 (4.1)	−2.1	0.40 (0.41)	<0.001
	Overall	1644	−7.2 (7.0)	925	−3.5 (6.0)	−3.7	0.55 (0.10)	<0.001

ADHD-RS, Attention-Deficit/Hyperactivity Disorder Rating Scale; ATX, atomoxetine; PLA, placebo; SD, standard deviation of the mean change from baseline to endpoint; SE, standard error of the effect size.

Table 3.

Confidence Intervals of Changes in ADHD-RS Total Scores from Baseline to Endpoint, “One Versus Others” Approach

Region	Mean change ATX – PLA Study (1)	Mean change ATX – PLA Rest (2)	Difference (1) – (2)	SE	90% C.I.	95% C.I.	97.5% C.I.
Asia	−3.74	−7.95	4.21	0.58	3.26, 5.17	3.07, 5.35	2.91, 5.51
Europe	−8.11	−7.22	−0.89	0.54	−1.78, −0.01	−1.95, 0.16	−2.11, 0.32
North America	−7.93	−6.76	−1.17	0.47	−1.94, −0.40	−2.08, −0.25	−2.21, −0.12
Russia	−5.89	−7.51	1.61	0.97	0.01, 3.21	−0.29, 3.52	−0.57, 3.79

ATX, atomoxetine; C.I., confidence interval of the difference, study versus rest; PLA, placebo; SE, standard error of the difference, study versus rest.

Consistency across regions

Although the four geographic regions were in accord with respect to the direction of treatment responses, some slight differences were seen in the conclusions implied by the different analytical methods. The Japanese MHLW guidance Method 2 indicated consistency of treatment effects across all studies with respect to the presence of positive treatment responses. However, the Higgins I² method indicated moderate levels of heterogeneity (I² =54.3), and analysis of treatment-by-region interaction, based on ADHD-RS changes from baseline and using an α threshold of 0.10, indicated a slight but significant interaction (F=2.40, df=3, p=0.066), suggesting at least some regional variability due to quantitative interaction among the magnitudes of treatment responses.

Consistency of one region relative to others

Comparison of treatment effects using the “one-region-versus-others” methods pointed collectively to a slight, but nevertheless detectable, difference between Asia and the other regions. The Japan MHLW guidance Method 1, using π=0.5, indicated that all regions except Asia met the criteria for consistency. Analysis of the PSR (overlapping coefficients) showed overlap among the treatment effects for the four regions, but with somewhat less overlap between Asia and the other three regions. Non-inferiority analysis based on a δ of 3.735, equivalent to half the total treatment effect, indicated that all regions except for Asia met the criteria for consistency (see Table 3). Similarly, the bioequivalence approach, using a range of (–3.735, 3.735), indicated that all regions except Asia met the criteria for consistency. Finally, the traditional t test showed that the difference relative to the other regions was highly significant not only for Asia (mean change difference versus other regions,+4.2; |t|=5.64, df=2567; p<0.001) but also North America (mean change difference versus other regions, −1.2; |t|=2.42, df=2567; p=0.008), but narrowly missed achieving significance for Europe (mean change difference versus other regions, −0.9; |t|=1.62, df=2567; p=0.052).

Analysis of response rates, based on a baseline-to-endpoint decrease of ≥40% on the ADHD-RS total, led to odds ratios that varied considerably, ranging from a low of 2.08 in Asia to a high of 4.87 in Europe (Table 4), with North America and Russia being intermediate. Responses to atomoxetine treatment were largely similar across regions, ranging from ∼40% in Europe and Asia to nearly 55% in Russia. Rates of responding to placebo, however, varied to a greater extent, ranging from 12.1% in Europe to 33.3% in Russia. The 95% confidence intervals for the odds ratios largely overlapped, but the interval for Asia was clearly at the lower end of the range, whereas that for Europe was at the upper end.

Table 4.

Response Rates ^a

	ATX		PLA
Region	n/N	Response rate (%)	n/N	Response rate (%)	Odds ratio	95% C.I.
Asia	76/192	39.6	23/96	24.0	2.08	1.20, 3.61
Europe	154/383	40.2	29/239	12.1	4.87	3.14, 7.55
N. Am.	452/997	45.3	121/557	21.7	2.99	2.36, 3.79
Russia	39/72	54.2	11/33	33.3	2.36	1.00, 5.58
Overall	721/1644	43.9	184/925	19.9	3.15	2.61, 3.80

Based on ≥40% improvement from baseline in ADHD-RS score.

ADHD-RS, Attention-Deficit/Hyperactivity Disorder Rating Scale; ATX, atomoxetine; C.I., confidence interval of the odds ratio; n, number of patients responding; N, number of patients assessed; PLA, placebo.

Disposition

Completion rates varied significantly among the four regions (Table 5), with Asia and Russia showing higher rates (both 94%) than those in Europe (84%) and North America (80%). Rates of discontinuation because of adverse events did not differ appreciably by region. However, rates of discontinuation because of lack of efficacy differed significantly, being highest in Europe (7.6%), intermediate in North America (4.2%), and lowest in Asia (2.1%) and Russia (0%).

Table 5.

Patient Disposition

	Region
Reason for discontinuation	Asia (N=288) n (%)	Europe (N=629) n (%)	North America (N=1605) n (%)	Russia (N=105) n (%)	Overall (N=2627) n (%)	p value
Adverse event	3 (1.0)	20 (3.2)	43 (2.7)	1 (1.0)	67 (2.6)	0.203
Lack of efficacy	6 (2.1)	48 (7.6)	67 (4.2)	0 (0.0)	121 (4.6)	<0.001
Lost to follow-up	0 (0.0)	1 (0.2)	41 (2.6)	0 (0.0)	42 (1.6)	<0.001
Personal conflict/other patient decision	1 (0.4)	20 (3.2)	63 (3.9)	5 (4.8)	89 (3.4)	0.003
Physician decision	1 (0.4)	3 (0.5)	6 (0.4)	0 (0.0)	10 (0.4)	0.892
Protocol violation	3 (1.0)	7 (1.1)	53 (3.3)	0 (0.0)	63 (2.4)	0.003
Sponsor decision	1 (0.4)	0 (0.0)	2 (0.1)	0 (0.0)	3 (0.1)	0.391
Other	1 (0.4)	0 (0.0)	37 (2.3)	0 (0.0)	38 (1.5)	0.022
Protocol/study period completed	272 (94.4)	530 (84.3)	1291 (80.4)	99 (94.3)	2192 (83.4)	<0.001

n, number of patients discontinuing for the given reason, or completing the study; N, total number of patients enrolled.

Discussion

The principal finding from this analysis is that atomoxetine was effective across all four regions examined, but the findings also underscore the importance of articulating the question to be answered, whether the focus is on overall consistency versus a difference between one region and all others in the analysis, as well as whether the consistency should be tested statistically, assessed directionally (positive or negative treatment effect), or quantified. For the most part, our findings suggested a general consistency among all four regions examined, with differences in findings being primarily a matter of degree, that is, variations were quantitative in nature, rather than qualitative. That being said, a trend was noted for Asia to be somewhat of an outlier from the other three regions with respect to the magnitude of responses to treatment. The “one versus others” methods were somewhat more sensitive to regional differences in this regard, relative to the “across regions” analyses, but again, each of these methods answers slightly different questions.

Although not definitive, some of the other findings from this analysis may provide some indication of the reasons for the observed heterogeneity of treatment responses, particularly with respect to the differences between Asia and the other three regions. Of note was a difference in the proportion of patients in Asia categorized by diagnostic category, with more patients of the predominantly inattentive subtype and fewer of the combined subtype. However, it is important to note that there is no evidence in the literature that atomoxetine is superior in treating ADHD combined cases than predominantly inattentive cases. Of note, the predominantly inattentive subtype of ADHD has been associated with greater responses to placebo (Newcorn et al. 2009); and yet, the magnitude and rates of responses to placebo were highest among Russian patients, lowest among European patients, and intermediate among Asian and North American patients. To our knowledge, the geographic, racial, or ethnic variability of placebo responses has not been well characterized, but in this analysis there does not appear to be a noticeable impact of diagnostic subtype on placebo responses. Moreover, given the composition of the regional populations, racial factors seem an unlikely explanation for the differences between Asian patients and the rest of the patient sample. Interestingly, Asian baseline ADHD-RS Inattention subscale scores appeared comparable to those in the other regions, but baseline Hyperactivity/Impulsivity subscale scores, and therefore ADHD-RS total scores, appeared to be less severe. This may be a reflection of cultural aspects of ADHD development in Asia and other regions, where Asian culture tends to prohibit behaviors that are typical of hyperactivity and impulsivity (Leung et al. 1996). This may also explain why patients in Asia appeared to have had less room for improvement and why their responses to treatment were therefore of a lower magnitude than those in the other regions.

Both Asian and Russian patients showed lower rates of discontinuation than did patients in Europe or North America. Rates of discontinuation because of lack of efficacy were notably highest in Europe and lowest in Asia and Russia. The reasons for this variability are not clear. One can only speculate on causes at this juncture, as, for example, patients were slightly older in Russia, but it is possible that geographic differences in both intrinsic and extrinsic factors may be involved.

Intrinsic factors affecting expression of ADHD have been fairly well documented, as ADHD is perhaps one of the best characterized mental disorders, from a biological standpoint. However, regional differences in child and adolescent ADHD severity and treatment response have received little attention, and the current exploratory results will, therefore, require considerable additional research before these findings and interpretations can be accepted. The worldwide prevalence of ADHD as a distinct mental disorder is not in question (Rohde et al. 2005; Moffitt and Melchior 2007; Polanczyk et al. 2007), but geographic differences in the proportions of the ADHD hyperactive/impulsive, inattentive, and combined subtypes are not well documented. Moreover, the regional distributions of the different polymorphisms of the dopamine transporter gene, DAT1, or of the dopamine D₄ receptor gene, DRD4, have not been fully characterized, and some studies cast doubt on the purported association between the DRD4 polymorphism, particularly the seven repeat allele, and ADHD in Chinese patients (Leung et al. 2005; Cheuk et al. 2006). Both of these sets of gene variants have been implicated in differential responses to pharmacological treatments for ADHD (Tahir et al. 2000; Roman et al. 2004), although this has not been a consistent finding for DAT1 (for review, see Froehlich et al. 2010).

Most comorbid conditions appear to be unlikely factors in the observed geographic differences, as prevalence rates for such conditions as oppositional defiant disorder or conduct disorder, mood disorders, anxiety, and substance use do not appear to be appreciably different, at least among Chinese patients relative to their Western counterparts (Gau et al. 2010), although the rate of comorbid tic disorder may be lower among Asian patients with ADHD (Yang et al. 2011). Clarification of the role of differential expression of comorbidities, however, will require more rigorous statistical testing.

Potential extrinsic reasons for geographic variation in treatment outcomes have been hinted at in the ADHD literature. As an example, diagnostic and prescribing practices are known to differ, even within a region. In China, for example, it is common for pediatricians to make the initial diagnosis and treatment decision, whereas in Taiwan and Korea, psychiatrists typically make such assessments (Martényi et al. 2010). The assessment tool used to make the diagnosis will also affect the findings, as differences in definitions, for example between the DSM-IV and the International Classification of Disease and Related Health Problems, version 10 (ICD-10) guidelines, may lead to different diagnoses (Döpfner et al. 2008; Lee et al. 2008; Cardo et al. 2011). Admittedly, these are not likely reasons for the geographic differences observed in the current study, as the methodologies should have been similar for all of the clinical trials in this meta-analysis.

Other extrinsic factors could have had a role in bringing about the observed variations in treatment outcomes. For example, it has been suggested that the influence on east Asian cultures of Confucian, Taoist, and Buddhist philosophies that emphasize social harmony and behavioral compliance could result in less expression of hyperactive/impulsive behavior (Tseng et al. 2005). This could potentially explain the lower baseline ADHD-RS Hyperactivity/Impulsivity scores in our results from Asia, although the prevalence rate of Asian patients who were diagnosed as primarily hyperactive/impulsive was comparable to what was seen in the other regions. Other possible extrinsic factors that could be proposed include regional, local, or societal/cultural differences in referral bias (Cuffe et al. 1995; Wierdsma and Kamperman 2011), patient management practices (Svendsen 1996; Caldwell et al. 2004), structure of medical payment systems (Hinshaw et al. 2011), use of nonpharmacological treatment interventions (Snape et al. 2009), prescribing patterns of primary and concomitant medications (Bitter et al. 2003; Zullino et al. 2005; Tiainen et al. 2008), and patients' self-reporting behaviors (Wang et al. 2010). The contribution of any of these potential factors is at this point still conjectural and will require rigorous testing for validation.

Limitations

A number of limitations of the current analysis bear mention, not the least of which being the post-hoc nature of the design. The individual studies used in this highly exploratory meta-analysis were not optimized for combining into pooled sets of data, despite similarities in many design features, and individual studies varied somewhat with respect to both the baseline characteristics of the overall patient samples and the criteria for separation into different treatment subgroups (for example, stimulant-naïve versus previous users). The length of time patients underwent an initial medication washout, doses, and patient mean symptom severities all varied from study to study. Despite the known drug metabolism difference caused by cytochrome P450 2D6 metabolizers, we were unable to analyze contributions of 2D6 polymorphisms to the observed regional variability. The metabolizer status was not captured consistently in the studies analyzed in this article. Moreover, all studies on which this analysis was based were industry sponsored, although any potential for introduction of publication bias should have been neutralized by the use of raw data from all double-blind, placebo-controlled, pediatric and/or adolescent clinical trials of 6–8 weeks in duration that could be found in the development database for atomoxetine, regardless of outcome. As for the analytical approach, other than the obvious use of different methods to characterize the geographic consistency of treatment effects, it must be acknowledged that the methods involved are perhaps best used to answer slightly different questions: If it matters more whether an overall regional difference exists, or if it matters more whether one region differs from the others. Both qualitative and quantitative approaches have been taken, and no adjustment was made when multiple comparisons were tested. Not all of the more obvious variables have been examined, such as genetic differences or differences between rural and metropolitan populations with respect to social attitudes or access to clinical services. Finally, some leeway was taken with the method of grouping some of the data, such as the inclusion of a small number of Australian patients in the “European” region, the necessary separation of Russian patients from the European and Asian regions, and the pooling of ADHD-RS and SNAP-IV data, in itself a potential limitation of the analytical approach. It must again be stressed that this was strictly an exploratory analysis based on an empirical approach, and the findings are to be interpreted in that light.

Clinical Significance

Treatment effects were generally consistent among the four regions, but some regional variability was detected in the analyses across all regions. More noticeable differences were detected with the analyses comparing individual regions against the others, suggesting that treatment effects were slightly lower in Asia than in the other regions. The results also suggest that there may be some slight regional variability in the proportions of patients meeting the diagnostic criteria for the three currently acknowledged subtypes of ADHD (primarily inattentive, primarily hyperactive/impulsive, or combined), the severity of hyperactive/impulsive and inattentive symptoms, or both. These findings may be of use in the planning of global clinical trials that include investigative sites in Asia, or clinical trials that are to be used as part of a “bridging” strategy for submission for market approval in Asia.

Footnotes

Acknowledgment

We thank Traci A. Hoopingarner of PharmaNet/i3, a part of the inVentiv Health Company, for editorial assistance with the manuscript.

Disclosures

This project was sponsored by Eli Lilly and Company. Drs. Tanaka, Jin, Feldman, and Upadhyaya are employees and stakeholders in Eli Lilly and Company. Dr. Rohde was on the speakers' bureau and/or acted as consultant for Janssen-Cilag, Lilly, Novartis, and Shire in the last 3 years (<U.S.$10,000 per year and reflecting <5% of his gross income per year). He also received travel awards from Janssen-Cilag and Novartis for taking part in two child psychiatric meetings in 2010. The ADHD and Juvenile Bipolar Disorder Outpatient Programs chaired by him received unrestricted educational and research support from the following pharmaceutical companies in the last 3 years: Abbott, Lilly, Janssen-Cilag, Novartis, and Shire.

The statistical analysis was performed by Dr. Tanaka, and the first draft of the manuscript was written by Dr. Feldman. All authors have contributed to the interpretation of the analysis results, have provided substantial revision of the manuscript, and take responsibility for its entire contents.

References

American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th. Washington, DC: American Psychiatric Association, 1994.

Bitter

, Chou

JCY

, Ungvari

, Tang

, Xiang

, Iwanami

, Gaszner

. Prescribing for inpatients with schizophrenia: An international multi-center comparative study. Pharmacopsychiatry, 36:143–149. 2003.

Caldwell

, Jorm

, Knox

, Braddock

, Dear

KBG

, Britt

. General practice encounters for psychological problems in rural, remote and metropolitan areas in Australia. Aust N Z J Psychiatry, 38:774–780. 2004.

Cardo

, Servera

, Vidal

, de Azua

, Redondo

, Riutort

. The influence of different diagnostic criteria and the culture on the prevalence of attention deficit hyperactivity disorder [in Spanish] Rev Neurol, 52,Suppl 1:S109–S117. 2011.

Chen

, Quan

, Binkowitz

, Ouyang

, Tanaka

, Li

, Menjoge

, Ibia

. Consistency Workstream of the PhRMA MRCT Key Issue Team. Assessing consistent treatment effect in a multi-regional clinical trial: a systematic review. Pharm Stat, 9:242–253. 2010.

Cheuk

, Li

, Wong

. Exon 3 polymorphisms of dopamine D4 receptor (DRD4) gene, attention deficit hyperactivity disorder in Chinese children. Am J Med Genet B Neuropsychiatr Genet, 141B:907–911. 2006.

Cox

. Interaction. Int Stat Rev, 52:1–25. 1984.

Cuffe

, Waller

, Cuccaro

, Pumariega

, Garrison

. Race and gender differences in the treatment of psychiatric disorders in young adolescents. J Am Acad Child Adolesc Psychiatry, 34:1536–1543. 1995.

Döpfner

, Breuer

, Wille

, Erhart

, Ravens–Sieberer

. the BELLA study group. How often do children meet ICD-10/DSM-IV criteria of attention deficit-/hyperactivity disorder and hyperkinetic disorder? Parent-based prevalence rates in a national sample—results of the BELLA study. Eur Child Adolesc Psychiatry, 17,Suppl 1:59–70. 2008.

10.

DuPaul

, Power

, Anastopoulos

, Reid

. ADHD Rating Scale-IV: Checklists, Norms, Clinical Interpretation. New York: The Guilford Press, 1998.

11.

Froehlich

, McGough

, Stein

. Progress and promise of attention-deficit hyperactivity disorder pharmacogenetics. CNS Drugs, 24:99–117. 2010.

12.

Gau

SSF

, Ni

, Shang

, Soong

, Wu

, Lin

, Chiu

. Psychiatric comorbidity among children and adolescents with and without persistent attention-deficit hyperactivity disorder. Aust N Z J Psychiatry, 44:135–143. 2010.

13.

Higgins

JPT

, Thompson

, Deeks

, Altman

. Measuring inconsistency in meta-analyses. BMJ, 327:557–560. 2003.

14.

Hinshaw

, Scheffler

, Fulton

, Aase

, Banaschewski

, Cheng

, Mattos

, Holte

, Levy

, Sadeh

, Sergeant

, Taylor

, Weiss

. International variation in treatment procedures for ADHD: social context and recent trends. Psychiatr Serv, 62:459–464. 2011.

15.

Idle

, Smith

. Polymorphisms of oxidation at carbon centers of drugs and their clinical significance. Drug Metab Rev, 9:301–317. 1979.

16.

International Conference on Harmonization: ICH International Conference on Harmonization Tripartite Guidance E5: Ethnic Factor in the Acceptability of Foreign Data. Federal Register, 83,31:790–317961998.

17.

Lee

, Schachar

, Chen

, Ornstein

, Charach

, Barr

, Ickowicz

. Predictive validity of DSM-IV and ICD-10 criteria for ADHD and hyperkinetic disorder. J Child Psychol Psychiatry, 49:70–78. 2008.

18.

Leung

, Lee

, Hung

, Ho

, Tang

, Kwong

, Leung

, Yuen

, Lieh–Mak

, Oosterlaan

, Grady

, Harxhi

, Ding

, Chi

, Flodman

, Schuck

, Spence

, Moyzis

, Swanson

. Dopamine receptor D4 (DRD4) gene in Han Chinese children with attention-deficit/hyperactivity disorder (ADHD): Increased prevalence of the 2-repeat allele. Am J Med Genet B Neuropsychiatr Genet, 133B:54–56. 2005.

19.

Leung

, Luk

, Ho

, Taylor

, Mak

, Bacon–Shone

. The diagnosis and prevalence of hyperactivity in Chinese schoolboys. Br J Psychiatry, 168:486–496. 1996.

20.

Martényi

, Zheng

, Huang

, Lee

, Walton

, Harrison

, Treuer

. A prospective observational study of attention-deficit hyperactivity disorder in Asia: Baseline characteristics of symptom severity and treatment options in a paediatric population. East Asian Arch Psychiatry, 20:76–86. 2010.

21.

Michelson

, Faries

, Wernicke

, Kelsey

, Kendrick

, Sallee

, Spencer

. the Atomoxetine ADHD Study Group. Atomoxetine in the treatment of children, adolescents with attention-deficit/hyperactivity disorder: A randomized, placebo-controlled, dose-response study. Pediatrics, 108:e83. 2001.

22.

Ministry of Health Labour and Welfare of Japan: Basic Principles on Global Clinical Trials. Tokyo: PMDA, 2007.

23.

Moffitt

, Melchior

. Why does the worldwide prevalence of childhood attention deficit hyperactivity disorder matter? Am J Psychiatry, 164:856–858. 2007.

24.

Newcorn

, Sutton

, Zhang

, Wilens

, Kratochvil

, Emslie

, D'souza

, Schuh

, Allen

. Characteristics of placebo responders in pediatric clinical trials of attention-deficit/hyperactivity disorder. J Am Acad Child Adolesc Psychiatry, 48:1165–1172. 2009.

25.

Patil

. Cochran's Q Test: Exact distribution. J Am Stat Assoc, 70:186–189. 1975.

26.

Polanczyk

, de Lima

, Horta

, Biederman

, Rohde

. The worldwide prevalence of ADHD: A systematic review and metaregression analysis. Am J Psychiatry, 164:942–948. 2007.

27.

Rani

, Pargal

. Bioequivalence: An overview of statistical concepts. Indian J Pharmacol, 36:209–216. 2004.

28.

Rohde

, Szobot

, Polanczyk

, Schmitz

, Martins

, Tramontina

. Attention-deficit/hyperactivity disorder in a diverse culture: Do research and clinical findings support the notion of a cultural construct for the disorder? Biol Psychiatry, 57:1436–1441. 2005.

29.

Rom

, Hwang

. Testing for individual and population equivalence based on the proportion of similar responses. Stat Med, 15:1489–1505. 1996.

30.

Roman

, Rohde

, Hutz

. Polymorphisms of the dopamine transporter gene: Influence on response to methylphenidate in attention deficit-hyperactivity disorder. Am J Pharmacogenomics, 4:83–92. 2004.

31.

Sauer

, Ponsler

, Mattiuz

, Long

, Witcher

, Thomasson

, Desante

. Disposition and metabolic fate of atomoxetine hydrochloride: The role of CYP2D6 in human disposition and metabolism. Drug Metab Dispos, 31:98–107. 2003.

32.

Snape

, Wang

, Wu

, Jacoby

, Baker

. Knowledge gaps and uncertainties about epilepsy: Findings from an ethnographic study in China. Epilepsy Behav, 14:172–178. 2009.

33.

Snapinn

. Noninferiority trials. Curr Control Trials Cardiovasc Med, 1:19–21. 2000.

34.

Spencer

, Biederman

, Wilens

, Harding

, O'Donnell

, Griffin

. Pharmacotherapy of attention-deficit hyperactivity disorder across the life cycle. J Am Acad Child Adolesc Psychiatry, 35:409–432. 1996.

35.

Spencer

, Heiligenstein

, Biederman

, Faries

, Kratochvil

, Conners

, Potter

. Results from 2 proof-of-concept, placebo-controlled studies of atomoxetine in children with attention-deficit/hyperactivity disorder. J Clin Psychiatry, 63:1140–1147. 2002.

36.

Student (Gosset WS). The probable error of a mean. Biometrika, 6:1–25. 1908.

37.

Svendsen

. Regional variations in the incidence of manic-depressive psychoses in the Danish psychiatric hospital system 1977–1993 [in Danish] Ugeskr Læger, 158:598–602. 1996.

38.

Swanson

. chool-Based Assessments and Interventions for ADD Students. Irvine, CA: KC Publishing, 1992.

39.

Tahir

, Yazgan

, Cirakoglu

, Ozbay

, Waldman

, Asherson

. Association and linkage of DRD4 and DRD5 with attention deficit hyperactivity disorder (ADHD) in a sample of Turkish children. Mol Psychiatry, 5:396–404. 2000.

40.

Tiainen

, Edman

, Flyckt

, Tomson

, Rehnberg

. Regional variations and determinants of direct psychiatric costs in Sweden. Scand J Public Health, 36:483–492. 2008.

41.

Tseng

, Chang

, Nishizono

. Asian culture, psychotherapy: An overview. Asian Culture and Psychotherapy: Implications for East and West. Tseng

, Chang

S.C.

, Nishizono

Honolulu: University of Hawai'i Press, 265–279. 2005.

42.

U.S. Food and Drug Administration. Guidance for Industry. Statistical Approaches to Establishing Bioequivalence. Washington, D.C.: Government Printing Office, 2001.

43.

Wang

, Cleeland

, Mendoza

, Yun

, Wang

, Okuyama

, Johnson

. Impact of cultural and linguistic factors on symptom reporting by patients with cancer. J Natl Cancer Inst, 102:732–738. 2010.

44.

Weiss

, Tannock

, Kratochvil

, Dunn

, Velez–Borras

, Thomason

, Tamura

, Kelsey

, Stevens

, Allen

. A randomized, placebo-controlled study of once-daily atomoxetine in the school setting in children with ADHD. J Am Acad Child Adolesc Psychiatry, 44:647–655. 2005.

45.

Wierdsma

, Kamperman

. Contact with the child and adolescent psychiatric services is determined by cultural and socio-economic differences [in Dutch] Tijdschr Psychiatr, 53:857–863. 2011.

46.

Wilson

, Weale

, Smith

, Gratrix

, Fletcher

, Thomas

, Bradman

, Goldstein

. Population genetic structure of variable drug response. Nat Genet, 29:265–269. 2001.

47.

World Medical Association:. WMA Declaration of Helsinki – Ethical Principles for Medical Research Involving Human Subjects. Bull WHO, 86:650–651. 2008.

48.

Yang

, Shang

, Gau

SSF

. Psychiatric comorbidities in adolescents with attention-deficit hyperactivity disorder and their siblings. Can J Psychiatry, 56:281–292. 2011.

49.

Zullino

, Mayland

, Schmidt

, Fähndrich

, Greil

, Horvath

, Rüther

, Engel

, Hippius

, Baumann

. Prescribing practices in German and Swiss psychiatric university and in non-university hospitals: National differences. Int J Clin Pharmacol Ther, 43:339–349. 2005.