Effects of Atomoxetine on Growth After 2-Year Treatment Among Pediatric Patients With Attention-Deficit/Hyperactivity Disorder
Objective. Treatment for attention-deficit/hyperactivity disorder is maintained typically over periods of months or years and, as a result, the potential effects on growth of pharmacotherapy for this disorder have been an area of concern. This meta-analysis examined the effect on growth of atomoxetine, now approved in the United States for the treatment of attention-deficit/hyperactivity disorder.
Methods. Patients (N = 412) were 6 to 16 years of age at the start of the treatment period and received atomoxetine treatment (maximal dose: 1.8 mg/kg per day) for ≥2 years. Weight and height measurements were analyzed both as actual values and after conversion to percentiles and z scores with growth charts from the Centers for Disease Control and Prevention. Expected weight and height at the end point were calculated through extrapolation from patients' baseline percentiles with the growth charts.
Results. Results indicated that, after 2 years, observed weight and height were close to those predicted on the basis of the patients' baseline weight and height. Weight increased an average of 10.8 kg, a decrease relative to baseline normative weight of 2.7 percentiles, corresponding to 0.87 kg. Height increased an average of 13.3 cm, a decrease relative to baseline normative heights of 2.2 percentiles, corresponding to 0.44 cm. For both weight and height, the quartile of patients who were smallest at baseline had an increase in end-point percentile, whereas patients in the highest quartile had a decrease.
Conclusions. These findings suggested that, at the group level, there was only a minimal effect on height after 2 years of treatment with atomoxetine and, for patients most at risk (the lowest quartile), there seemed to be no effect.
Attention-deficit/hyperactivity disorder (ADHD) is among the most common psychiatric disorders of childhood, and pharmacotherapy is now widely regarded as the most effective intervention for most patients.1,2 Because ADHD is a chronic disorder, typically treatment is maintained for periods of months or years; as a result, the potential effects of pharmacologic treatments for ADHD on developmental outcomes, particularly growth, have been an area of interest and concern for many clinicians. Until recently, however, only limited systematic data about the effects of ADHD treatments on growth have been available.3–8 In recent years, more efforts have been made to investigate the issue of whether growth in height and weight is affected by long-term pharmacologic treatment, and the recent Multimodal Treatment Study of children with ADHD (MTA) results,9 together with the introduction of regulatory requirements for systematic investigations of drug effects during long-term treatment, have created a growing body of data to address this question.
Several recent studies have suggested that the initiation of psychostimulant treatment is associated with a modest decline in growth rates that abates with time and that, over longer periods, both weight and height gains are at or close to expected values.6,8 Data from the MTA trial of methylphenidate, dosed 3 times per day, showed somewhat more pronounced effects on growth by early adolescence.10
Atomoxetine, a nonstimulant, selective, norepinephrine reuptake inhibitor, was approved by the US Food and Drug Administration as a treatment for ADHD.11–14 The clinical development of atomoxetine was conducted primarily in pediatric populations and included a provision for children and adolescents who completed short-term studies (typically 6–12 weeks) to continue to receive treatment in an ongoing, long-term, extension study for up to 5 years. As part of all of these studies, weight and height were measured and recorded regularly, providing an opportunity to evaluate systematically the effects of long-term treatment with atomoxetine on growth. We report data on weight and height for patients who completed ≥2 years of treatment with atomoxetine.
Patient Group and Study Design
Data were pooled from 13 multicenter trials conducted at 90 sites across North America as part of the clinical development program for atomoxetine. Patients included children and adolescents between the ages of 6 and 16 years at the time of initial assessment. ADHD (as defined in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition) was assessed clinically and confirmed with the Kiddie Schedule for Affective Disorders and Schizophrenia.15 The studies allowed patients who entered and completed a short-term study the option to continue treatment for up to 5 additional years in an extension study. The studies that fed into the extension ranged from 6 weeks to 2 years in length and used similar dose ranges, with a target dose of 1.2 mg/kg per day and a maximal dose of 1.8 mg/kg per day in most studies. The total daily dose was administered either once daily or as a divided twice-daily dose. Each study was reviewed and approved by the institutional review boards for the participating study sites. After complete explanation of the procedures and risks of the studies, each patient's parent or legal guardian signed written informed consent forms and each patient signed assent forms. These studies were conducted in accordance with the ethical standards of the Declaration of Helsinki (1975, as revised in 2000).16
All patients underwent medical history recording and a complete physical examination at entry into the initial study and again at entry into the extension study. Routine laboratory examinations, including serum electrolyte assays, liver function tests, and complete blood counts, and electrocardiography were performed at the initial study entry and periodically thereafter, to monitor safety. Patients with serious medical illnesses were excluded, and administration of psychotropic medications other than atomoxetine was not allowed except for one 187-patient study in which some of the patients had an 8-week period of fluoxetine coadministration. Weight was measured at every study visit in all studies. Height was measured at the initial and final study visits for all studies. In the extension study and in the 2 initial studies that lasted >6 months, height was measured at study entry and then at regular intervals that did not exceed 6 months.
Two study populations were defined for analytical purposes. The first population included all patients who completed ≥2 years of treatment and had a weight or height measurement performed after 2 years of drug exposure. For these patients, changes in weight and height and changes in weight and height percentiles, relative to population normative values for age and gender,17 were calculated for the overall period of drug exposure and for selected time points up to 2 years. Two additional analyses were performed with a second population, which included all patients who received atomoxetine therapy. This was done to assess the effect of treatment discontinuation on the results of the analyses that included only patients with 2 years of drug exposure. These analyses included t tests, comparing results for patients who discontinued treatment after 0 to 3, 3 to 6, 6 to12, or 12 to 18 months of exposure with results for patients with ≥2 years of exposure, and repeated-measures analyses. These repeated-measures analyses modeled weight and height z scores (taking into account both the mean and the SD of the sample to provide information about the relative weight and height of subgroups) and included effects of the duration of atomoxetine exposure and the duration of exposure squared.
Changes from baseline to end point in weight, height, and BMI z scores were summarized separately for subgroups of patients, on the basis of gender, age, and modal atomoxetine daily dosage. Subgroup differences were analyzed with an analysis of covariance model with effects for baseline value and subgroup. Repeated-measures analyses with data for all patients were also performed and included effects for subgroup, duration of atomoxetine exposure, and duration of exposure squared and interaction terms for subgroup, duration of exposure, and duration of exposure squared (a model arrived at through backward selection from models with higher-order polynomial terms).
A total of 2551 patients received atomoxetine and had baseline weight and height measurements. Of these, 2327 had ≥1 other height measurement and 2532 had ≥1 other weight measurement. A total of 419 patients received atomoxetine for ≥2 years. Of these, 412 had a weight measurement recorded after 2 years and 382 had a height measurement recorded after 2 years. The numbers of patients included in the weight and height analyses are different because weight was measured at every visit but height was measured less frequently. Therefore, some patients in these ongoing trials had not yet had a height measurement after passing the 2-year treatment date at the cutoff time for inclusion in these data analyses.
Patient characteristics for those receiving atomoxetine therapy (N = 2551) are summarized in Table 1. Mean changes in weight, height, and BMI after ≥2 years of treatment are summarized in Table 2. At baseline, the mean weight percentile was 60.1 (z score: 0.41) and the mean height percentile was 51.8 (z score: 0.06). As is typical in the North American population (from which the patients were drawn), the patients weighed more than expected, given their height (mean BMI percentile: 62.2; z score: 0.44).
For the 412 patients with weight measurements recorded after 2 years, there was a marked absolute mean weight gain of ∼10.79 kg at the end point. These values corresponded to slight mean decreases, relative to baseline normative weights (−2.7 percentiles, P = .002). Compared with final height, however, final weight was still slightly above that expected (mean BMI percentile: 59.6; z score: 0.35). The decrease from the predicted weight, assuming maintenance of the baseline weight percentile, was 0.87 kg at the end point. Weight percentiles plotted against time (a measure of growth velocity) are shown in Fig 1, A. These data points reveal an initial decrease followed by maintenance of weight percentiles over time. A repeated-measures analysis of z scores that included data for all enrolled patients with ≥1 postbaseline weight measurement showed results similar to those observed in the completer analysis (Fig 1, B).
For the 382 patients with height measurements after 2 years, there was a marked absolute mean height gain (13.3 cm at the end point) (Table 2). This value corresponded to a slight decrease, relative to the baseline mean normative height value (−2.2 percentiles, P = .02). The decrease from the height predicted by assuming maintenance of the baseline height percentile was 0.44 cm at the 2-year end point. Height percentile changes from baseline plotted against time are shown in Fig 2, A. These data points reveal an initial decrease but maintained height percentiles over time. A repeated-measures analysis of z scores that included data for all enrolled patients with ≥1 postbaseline height measurement showed results similar to those observed in the completer analysis (Fig 2, B).
Analyses of weight and height according to baseline normative quartile are shown in Table 3. For both weight and height, the quartile of patients who were smallest at baseline increased in final percentile, whereas patients in the highest quartile had a decrease in end-point percentile. A concomitant change occurred in BMI, such that patients in the lower quartiles at baseline had an increase at the end point, whereas patients in the upper quartiles at baseline had a decrease at the end point. Overall, weight and height changes from baseline to end point (z scores) were correlated significantly, even when baseline weight and height z scores were taken into account (partial correlation: 0.56; P < .001).
The numbers and percentages of patients of lowest weight and height (relative to the general population) at baseline are presented in Table 4. The percentage of patients with weights 1.5 SDs lower than the population mean increased significantly from baseline (3.6%) to end point (7.3%, P = .008). The percentage of patients with weights 2.0 SDs lower than the population mean at baseline did not change significantly at end point. In addition, there were no significant changes from baseline to end point in the percentages of patients with heights 1.5 or 2.0 SDs below the population mean.
Changes from baseline in weight and height according to baseline BMI quartile were also calculated. Patients whose baseline BMI values were in the lower quartile increased in final percentile for both weight (+2.8%) and height (+0.1%), whereas patients with baseline BMI values in the highest quartile had a decrease in end-point percentile for both weight (−6.1%) and height (−3.1%).
Analyses of growth data according to atomoxetine dose, conducted with subgroups of patients with modal doses of atomoxetine that were low (<1.0 mg/kg per day), intermediate (1.0–1.4 mg/kg per day), or high (>1.4 mg/kg per day), revealed no statistically significant effects of atomoxetine dose on changes from baseline to final standardized weight, height, or BMI (data not shown). There were also no statistically significant gender differences in changes from baseline to final standardized weight, height, or BMI. However, repeated-measures analyses with all patient data showed a statistically significant difference between genders in normative weight values, suggesting that, at least initially, female patients showed greater decreases in normative weight but this difference was made up for with time (data not shown).
Analysis of standardized growth data according to age at first atomoxetine dose is shown in Table 5 and indicates, for weight, a significant effect of age (P = .005), such that <9-year-old patients showed decreases after atomoxetine treatment (change in z score: −0.26), 9- to 13-year-old patients showed smaller changes (z = −0.07), and >13-year-old patients were not affected at all (z = 0.0). For height, there was also a significant effect of age (P = .048), such that <9-year-old patients showed decreases after atomoxetine treatment (change in z score: −0.17), 9- to 13-year-old patients showed smaller decreases (z = −0.05), and >13-year-old patients actually showed increases (z = 0.07). For BMI, there was not a statistically significant effect of age. Repeated-measures analyses with all patient data showed a statistically significant difference between age groups in both normative weight and height values, with younger children at least initially showing greater decreases in normative values. All age groups resumed normal growth velocity as treatment progressed.
To assess the impact of patients who discontinued treatment, normative weight and height observations at the time points described below for patients who discontinued treatment at selected time points were compared with observations at the same time points for patients who completed ≥24 months of treatment with a 2-sample t test. The last weight measurements obtained before the start of a time interval for patients who discontinued treatment were compared with weight measurements taken at a similar time point for 24-month exposure patients (data not shown). The only statistically significant change was seen when patients who discontinued treatment after 3 to 6 months of exposure were compared with the 24-month exposure patients; 24-month exposure patients showed a greater decrease in normative weight. This observation does not suggest a clinically meaningful trend.
The last height measurements for patients who discontinued treatment after various intervals were compared with height measurements taken at similar time points for patients who completed 24 months of treatment (data not shown). One statistically significant difference was observed between groups (12–18 months); however, this difference was minimal and was not observed at any other time point. There was no pattern suggesting meaningful differences between patients who completed 24 months of treatment and those who discontinued treatment at earlier time points.
Among patients treated for ≥2 years with atomoxetine at usual doses, mean growth rates slowed moderately for weight and slightly for height during the initial 6 months of treatment; however, during the subsequent ≥18 months, rates were as expected for weight and height. After ≥2 years of treatment, mean weight and height increased to values close to those predicted by age and gender normative values. There was only a ∼2-percentile change in height, from the mean 52nd percentile to the 50th percentile, and this was related largely to effects early in treatment. This change would correspond to a mean decrease from the height predicted by the initial height of only 0.44 cm. In addition, the analyses of changes according to quartile for both weight and height suggested that the smallest patients, who were presumably most at risk for clinically important effects, were least affected. These data suggest that atomoxetine has a modest initial effect on growth rates that does not persist during periods of extended treatment.
Interestingly, the analyses of changes relative to baseline quartiles for weight and height suggested some regression to the mean, with the most marked changes being observed for the patients in the highest and lowest quartiles. Similar findings were reported for children treated with methylphenidate18 and amphetamine,19 with long-term treatment of children <13 years of age. We cannot fully account for this finding but speculate that it may be related to inherent variability in growth processes. Growth curves represent average growth and do not fully reflect individual growth trajectories. This problem may be exacerbated by issues of pubertal tempo among adolescents. In our study, the mean age at study entry was 10.6 years. Therefore, the 2-year period of observation coincided with the onset and development of puberty for many patients. Tanner and Davies20 suggested that children who experience relatively early onset of puberty tend to reach their final height faster and are thus tall for their age early but not later. Similarly, children who reach puberty later tend to be short for their age early but not later. It is likely that the sample distribution of weight and height was affected at the beginning of the study because of early or late pubertal development and, over the course of the study, moved back toward the population mean as the group became older and differential effects related to the timing of puberty decreased.
Results indicated a statistically significant difference between genders, suggesting a slightly greater effect on growth velocity for weight among boys (consistent with a recent report by Lisska and Rivkees4). However, the overall patterns of responses, with modest early decreases in growth velocity followed by resumption of normal growth rates during ongoing treatment, were similar for the 2 genders. Similarly, there was a statistically significant change across age groups for both weight and height, with greater slowing of growth velocity early in treatment among younger children. However, all groups resumed normal growth velocity as the length of treatment increased. Because the observed differences between groups (male/female, younger/older) for these effects were small and because all growth velocity returned toward a normal trajectory over time, these findings, although statistically significant, seem unlikely to represent clinically meaningful differences.
Also of interest, the modal dose did not seem to be related to outcomes. This could reflect confounding related to pharmacokinetic variability, because low doses could be associated with high plasma drug exposures for some patients, whereas other patients could receive higher doses but have relatively low plasma drug exposures. However, the lack of relationship of dose to apparent growth effects provides additional evidence of the generally benign relationship of atomoxetine to growth processes.
The finding of an acute effect that resolved during chronic treatment could be related to several factors. Initiation of treatment with atomoxetine is associated with a decrease in appetite, which could result in decreased caloric intake and lead to the initial decrease in weight. This hypothesis is consistent with the observation in clinical studies that, over time, decreased appetite tends to diminish as a reported adverse effect. It also suggests that the addition of a caloric supplement for a period of time when treatment is initiated could blunt the initial deceleration in growth velocity. This is consistent with a previous report on hypopituitarism that suggested that changes in appetite modulate changes in growth rates.21 Alternatively, because it is known that norepinephrine affects neuroendocrine systems that modulate growth,22–24 it is possible that increased noradrenergic tone related to atomoxetine could affect the regulation of the hypothalamic-pituitary-growth axis. Because neuroendocrine systems are quite resilient and tend to return to homeostatic set-points, this mechanism could also be consistent with the observed pattern of early effects on growth that diminish with time.
Several factors limit the interpretation of these data. The comparison data used for prediction of expected growth rates were taken from normative population growth data, rather than from a matched, randomized sample. This choice reflects the fact that randomizing a matched group of patients to placebo or to a psychosocial intervention for an extended period for comparison purposes would have been unacceptable to most clinicians, particularly in view of the results of the MTA that demonstrated superior outcomes for children randomly assigned to pharmacologic treatment, compared with those randomly assigned to a psychosocial intervention. The study also cannot control for effects on growth related to ADHD itself, which were reported by some researchers.6,7
It is noted that collection of weight and height data at multiple sites over a long period of time might have resulted in variability related to measurement errors, which could have affected the observed outcomes. However, given the size of the sample and the likelihood that most measurement errors were random, errors would likely increase the random noise in the sample but not systematically bias the results. Errors made consistently across visits (for example, weighing patients with their shoes on) would not affect greatly examinations of changes over multiple visits. Therefore, we consider it unlikely that measurement errors seriously affected the results.
Finally, it is possible that the patients who did not remain in the study for the full 2 years (either because they discontinued treatment or because the original study protocol was <2 years) would have had an effect on this analysis had they been included. The dropout analysis, however, indicated that this would not have been the case. For weight, in the only dropout time interval that showed a significant difference, if the patients who dropped out had been included, then actually the weight loss noted in this study would have decreased. For height, there was a statistically significant difference at 1 isolated time point, but this was not indicative of a clinically significant trend.
The data presented here suggest that, at the group level, there was only a minimal effect on height after 2 years of treatment with atomoxetine; for those most at risk (those in the lowest quartile), there seemed to be no effect. Individual patients, however, could show more (or less) pronounced effects. Therefore, it is important for clinicians to assess growth periodically during treatment and, for patients who seem to be growing more slowly than expected, to consider whether treatment with atomoxetine is a factor.
This research was funded by Eli Lilly and Company.
We thank Dr Rodney Moore for contributions to the preparation of this manuscript.
- Accepted December 29, 2004.
- Address correspondence to Thomas J. Spencer, MD, Massachusetts General Hospital, 15 Parkman St, WACC 725, Boston, MA 02114. E-mail:
Conflict of interest: Drs Spencer, Newcorn, Kratochvil, and Biederman have functioned as paid consultants and/or investigators for studies sponsored by Eli Lilly and Company. Drs Ruff and Michelson are employees and shareholders of Eli Lilly and Company.
- ↵American Academy of Pediatrics. Clinical practice guideline: treatment of the school-aged child with attention-deficit/hyperactivity disorder. Pediatrics.2001;108 :1033– 1044
- ↵Biederman J, Faraone SV, Monuteaux MC, Plunkett EA, Gifford J, Spencer T. Growth deficits and attention-deficit/hyperactivity disorder revisited: impact of gender, development, and treatment. Pediatrics.2003;111 :1– 6
- ↵MTA Cooperative Group. National Institute of Mental Health Multimodal Treatment Study of ADHD follow-up: changes in effectiveness and growth after the end of treatment. Pediatrics.2004;113 :762– 769
- ↵Swanson J, Elliott G, Greenhill L, Newcorn J, Jensen P. 24-month stimulant effects on growth in MTA follow-up. Presented at the 50th Annual Meeting of the American Academy of Child and Adolescent Psychiatry; October 14–19, 2003; Miami, FL
- ↵Michelson D, Faries D, Wernicke J, et al. Atomoxetine in the treatment of children and adolescents with attention-deficit/hyperactivity disorder: a randomized, placebo-controlled, dose-response study. Pediatrics.2001;108 (5). Available at: www.pediatrics.org/cgi/content/full/108/5/e83
- ↵Kelsey DK, Sumner CR, Casat CD, et al. Once-daily atomoxetine treatment for children with attention-deficit/hyperactivity disorder, including an assessment of evening and morning behavior: a double-blind, placebo-controlled trial. Pediatrics.2004;114 (1). Available at: www.pediatrics.org/cgi/content/full/114/1/e1
- ↵Kaufman J, Birmaher B, Brent D, Rao U, Ryan N. Kiddie-SADS-Present and Lifetime Version (K-SADS-PL): Version 1.0 of October 1996 Edition. Pittsburgh, PA: University of Pittsburgh Press; 1996
- ↵World Medical Association. Declaration of Helsinki: Recommendations Guiding Medical Doctors in Biomedical Research Involving Human Subjects. Ferney-Voltaire, France: World Medical Association; 2000
- ↵Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, et al. CDC growth charts: United States. Adv Data.2000;(314) :1– 27
- ↵Spencer TJ. Treatment for ADHD with OROS methylphenidate: long-term effect on growth. Presented at the 156th Annual Meeting of the American Psychiatric Association; May 21, 2003; San Francisco, CA
- ↵Biederman J, Faraone S, Spencer TJ. Extended release mixed amphetamine salts in ADHD: growth parameter analysis. Presented at the 156th Annual Meeting of the American Psychiatric Association; May 17–22, 2003; San Francisco, CA
- Laakmann G, Schoen HW, Blaschke D, Wittmann M. Dose-dependent growth hormone, prolactin and cortisol stimulation after i.v. administration of desimipramine in human subjects. Psychoneuroendocrinology.1985;10 :83– 93
- Copyright © 2005 by the American Academy of Pediatrics