ADHD

Effects of Atomoxetine and Methylphenidate on Sleep in Children With ADHD R. Bart Sangal, MD1; Judith Owens, MD2; Albert J. Allen, MD3; Virginia Sutton, PhD3; Kory Schuh, PhD3; Douglas Kelsey, MD3 Clinical Neurophysiology Services, PC, Troy, MI; 2Child and Family Psychiatry, Rhode Island Hospital, Providence, RI; 3Lilly Research Laboratories, Indianapolis, IN 1

irritable, had less difficulty getting ready for bed, and had less difficulty falling asleep with atomoxetine, compared with methylphenidate. There were no significant differences between medications using the main measures of efficacy for ADHD treatment. Atomoxetine was superior on some secondary ADHD treatment-efficacy measures, based on parent reports. The only significant differences in treatment-emergent adverse events were greater incidence of decreased appetite and greater incidence of insomnia with methylphenidate. Conclusions: Patients receiving twice-daily atomoxetine had shorter sleep-onset latencies, relative to thrice-daily methylphenidate, based on objective actigraphy and polysomnography data. Although both medications decreased nighttime awakenings, the decrease was greater for methylphenidate. Keywords: Atomoxetine, methylphenidate, attention-deficit/hyperactivity disorder, child, sleep Citation: Sangal RB; Owens J; Allen AJ et al. Effects of atomoxetine and methylphenidate on sleep in children with ADHD. SLEEP 2006;29(12):1573-1585.

Study Objectives: This study compared the effects of atomoxetine and methylphenidate on the sleep of children with attention-deficit/hyperactivity disorder (ADHD). This study also compared the efficacy of these medications for treating ADHD in these children. Design: Randomized, double-blind, crossover trial. Setting: Two sleep disorders centers in the United States; 1 in a privatepractice setting and 1 in a hospital setting. Patients: 85 children diagnosed with ADHD. Interventions: Twice-daily atomoxetine and thrice-daily methylphenidate, each for approximately 7 weeks. Measurements and Results: Relative to baseline, the actigraphy data indicated that methylphenidate increased sleep-onset latency significantly more than did atomoxetine (39.2 vs 12.1 minutes, p < .001). These results were consistent with the polysomnography data. Child diaries indicated that it was easier to get up in the morning, it took less time to fall asleep, and the children slept better with atomoxetine, compared with methylphenidate. Parents reported that it was less difficult getting their children up and getting them ready in the morning and that the children were less

ylphenidate are similar to those of amphetamine. They include nervousness, insomnia, and anorexia.3 Atomoxetine (Strattera®) is a nonstimulant treatment for ADHD. It is a highly specific norepinephrine reuptake inhibitor. Unlike methylphenidate, atomoxetine does not increase extracellular dopamine in the striatum and nucleus accumbens.4 Because methylphenidate and atomoxetine differ in their modes of action, they might have differing effects on sleep. Clinical experience suggests that stimulants often adversely affect sleep, either due to a direct drug effect or due to a secondary “rebound” effect as the medications wear off. These settling problems, or delayed sleep onset latency—often manifested clinically as difficulty falling asleep or bedtime resistance, or both—may result in significantly shortened sleep. Although dosing with either short-acting stimulants twice daily in the morning and at noon or with longer-acting stimulants once in the morning may result in a lack of coverage in the evening hours when patients need to concentrate and interact in social and family situations, additional late-day doses of stimulants may not be desirable because of these stimulant-related sleep effects.5 Previous studies that investigated the effects stimulants might have on sleep have produced mixed results. Studies that detected a negative impact on sleep include a 1983 study by Greenhill and colleagues6 that examined the sleep of 7 boys with ADHD at baseline and after 6 months of continuous twice-daily, methylphenidate treatment. Treatment with methylphenidate was associated with polysomnographically determined lengthened total sleep time, increased sleep-stage shifts, increased number of rapid eye movement (REM) periods, and elevated indexes of REM activity and REM-period fragmentation. A 1998 study by Ring and colleagues7 found that children (n = 13) with ADHD being treated with a single morning dose of methylphenidate were significantly more likely to experience sleep disturbances (61.5%),

INTRODUCTION ATTENTION-DEFICIT/HYPERACTIVITY DISORDER (ADHD) IS A CENTRAL NERVOUS SYSTEM DISORDER THAT HAS ITS ONSET IN CHILDHOOD AND IS estimated to occur in 3% to 7% of school-aged children.1 Treatment of ADHD traditionally has relied on stimulants, including methylphenidate and amphetamines. Methylphenidate, the most commonly used medication for treating ADHD, is a central nervous system stimulant. Methylphenidate blocks the reuptake, and increases release of, dopamine and norepinephrine from presynaptic vesicles into the synapse. These increases in extracellular dopamine and norepinephrine cause sympathomimetic effects and are thought to produce increased performance on tasks requiring vigilance and mental awareness, as well as decreased need for sleep and decreased awareness of fatigue.2 Common side effects of methDisclosure Statement This was an industry supported study sponsored by Eli Lilly. The data were analyzed by statisticians at Eli Lilly, including Dr. Sutton, one of the authors. The manuscript was written as a combined effort of all authors. Dr. Sangal has received research support from Eli Lilly, Merck, Organon, Cephalon, and Novartis. Dr. Owens has received research support from Cephalon, SanofiAventis, Johnson & Johnson, Sepracor, and Eli Lilly; is a member of the speakers’ bureau for Eli Lilly; is a consultant for Cephalon; and is an advisory board member for Cephalon, Pfizer, and Eli Lilly. Drs. Sutton, Allen, Schuh, and Kelsey are employees of Eli Lilly. Submitted for publication August 25, 2005 Accepted for publication August 5, 2006 Address correspondence to: Kory J. Schuh, PhD, Eli Lilly and Company, Lilly Corporate Center DC 4135 Indianapolis IN 46285; Tel: (317) 276-6627; Fax: (317) 651- 1726; E-mail: [email protected] SLEEP, Vol. 29, No. 12, 2006

1573

Atomoxetine and Methylphenidate Sleep Effects—Sangal et al

and untreated ADHD producing similar sleep disturbances. The current study compared methylphenidate with atomoxetine, a nonstimulant ADHD treatment medication, to determine the effects on sleep. Because atomoxetine is a highly selective inhibitor of the presynaptic norepinephrine transporter with little affinity for other neurotransmitter transporters or receptors, it might have fewer negative effects on sleep, compared with methylphenidate. Atomoxetine is approved by the Food and Drug Administration as an ADHD treatment in children, adolescents, and adults. In addition to the differences in the mechanisms of action of these 2 drugs that would suggest possible differences in effects, clinical trial data from an initial pediatric efficacy study that included both atomoxetine and methylphenidate also suggest the possibility of differential effects on sleep. Among children who received atomoxetine, methylphenidate, or placebo, spontaneously reported rates of insomnia were significantly higher for methylphenidate-treated patients compared with atomoxetine- or placebo-treated patients, whereas rates of insomnia among children receiving atomoxetine were similar to those of children on placebo (methylphenidate 10 of 37, 27%; atomoxetine 9 of 129, 7%; placebo 11 of 124, 8.9%; p < .05; difference between atomoxetine and methylphenidate; p = .002 [Eli Lilly, data on file]). This finding appears to reflect a subjective perception of differences in sleep patterns; however, the nature and severity of these differences, as well as the replicability of the findings, are unknown. The present study had 2 objectives: (1) to compare the effects of twice-daily atomoxetine and thrice-daily methylphenidate on the sleep of children with ADHD, as measured by both objective (actigraphy and polysomnography) and subjective (parent and child diaries) measures and (2) to compare the efficacy of atomoxetine and methylphenidate for treating ADHD in these children. Clinical trials have demonstrated significant efficacy compared with placebo when atomoxetine is given twice daily12,13 or once daily.14,15 In this study, atomoxetine was given twice a day because it had not been studied as a once-daily medication at the time this study was designed. Methylphenidate given 3 times a day was selected as the comparator medication because the halflife of methylphenidate is shorter than that of atomoxetine and this would produce an approximate equivalency in the duration of pharmacologic effects compared with twice-daily atomoxetine.

as reported by their parents using a structured sleep questionnaire, compared with their healthy siblings (37.5%). It is unclear, however, whether the sleep disturbances resulted from the children having ADHD or from their being treated with methylphenidate. Another study that detected a negative impact on sleep was a 1999 study by Stein.8 Parents of children with ADHD being treated with stimulants (types of stimulants not specified) reported almost a 2-fold increase in long latencies (> 30 minutes) to sleep onset or insomnia at least once per week and a higher prevalence of nightly “severe” sleep problems than did children with untreated ADHD. Furthermore, 29% of children with ADHD being treated with stimulants displayed increased sleep latency or insomnia every night, compared with only 10% of children with untreated ADHD. These results suggest that treatment with stimulants, and not necessarily ADHD per se, may be an important cause of sleep disturbances in children with ADHD. Two studies produced mixed results. A 1996 study by Stein and colleagues9 compared placebo with twice-daily and thrice-daily methylphenidate and found a significant dosing-schedule effect for sleep duration according to sleep diaries completed by parents. Based on the diaries, children (n = 25 boys) slept an average of 24 minutes less per night on the thrice-daily schedule, compared with placebo. Actigraphy recordings indicated that the average sleep duration was 8.8 hours on thrice-daily, methylphenidate, compared with 9.4 hours on placebo (p < .08). Actigraphic data also indicated that when children were administered placebo, they fell asleep 33 and 25 minutes faster compared with twice-daily and thrice-daily methylphenidate, respectively. Although these differences were not statistically significant, the authors noted that the analysis had limited power because the actigraphic data were obtained from only two thirds of the children. A 1997 study by Efron and colleagues10 compared the side effects of twice-daily methylphenidate and dexamphetamine in 125 children with ADHD using a double-blind, crossover trial. Parents were asked to complete the Side Effects Rating Scale, developed by Barkley, at the end of each medication period. Dexamphetamine, but not methylphenidate, produced significantly more parent-rated insomnia, relative to baseline. A 1995 study by Kent and colleagues11 did not detect a negative impact on sleep. This study compared the effects of placebo with the effects of methylphenidate (10 mg and 15 mg) taken at 4:00 pm. Patients also received methylphenidate at 7:00 am and noon. Although these 12 children were more often rated as less tired when awakening after nights they received 10 mg of methylphenidate, compared with nights they received 15 mg, methylphenidate did not produce statistically significant increases in sleep latency (observed by the nursing staff) compared with placebo. The average time to sleep onset was 49 minutes, with considerable variability. The authors stated that it is possible that clinically significant differences might exist between medication conditions but did not reach statistical significance because of the small sample size. In summary, previous studies that examined the effects of stimulants on sleep have produced mixed results. Many of these studies have had small numbers of patients, which can preclude drawing reliable conclusions. In addition, when studies have compared the sleep of children with ADHD being treated with methylphenidate to the children’s own nonmedicated baseline state or to the effects of placebo, the lack of reported significant differences in sleep parameters could result from methylphenidate treatment SLEEP, Vol. 29, No. 12, 2006

METHODS Patients Patients were 6 to 14 years old at study entry. They were diagnosed with ADHD using the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV™) criteria1 as well as severity criteria. Diagnosis was assessed by the investigator’s clinical evaluation and by the administration of several modules of the Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version structured interview. In addition, patients had an ADHD Rating Scale-IVParent Version: Investigator-Administered and Scored (ADHD RS)16 score at least 1.0 standard deviation above normative values for age and sex for either the inattentive or hyperactive/impulsive subscore, or for the combined score. All patients scored at least 80 rd on the Wechsler Intelligence Scale for Children®-3 edition. Important exclusion criteria included serious medical illness, a history of symptoms suggestive of a primary sleep disorder—such 1574

Atomoxetine and Methylphenidate Sleep Effects—Sangal et al

Period I Diagnostic Assessment Phase

Period III Acute Tre a t m en t Phase 2

Period II Acute T re a t m e n t Phase 1

ATX (n=37)

ATX (n=44)

Drugf re e

MPH (n=42)

MPH (n=41)

5-12 Days (9 Days Suggested)

~6-7 Weeks

Period IV Discontinuation Phase

10-20 Days

~ 6-7 Week s

7-10 Days (9 Days Suggested)

Figure 1—Study diagram.

as obstructive sleep apnea (OSA) (e.g., habitual snoring), periodic limb movement disorder (PLMD, eg, kicking movements during sleep), or insufficient sleep syndrome (e.g., voluntary sleep restriction resulting in sleep duration habitually significantly shorter than expected age norms)—that could potentially result in a daytime symptom constellation similar to ADHD, and abnormal laboratory values or electrocardiogram (ECG) readings. Patients agreed not to use caffeinated beverages during the duration of the study.

Patients who had been unable to consistently maintain the agreedupon “lights-out” time were discontinued from participation in the study. A consistent bedtime was defined as actual “lights out” occurring within 30 minutes before or after the agreed upon “lights-out” bedtime on at least 85% of nights (this corresponds to 6 out of 7 nights per week). Visit 3 took place 5 to 12 days after Visit 2. Laboratory results were reviewed; actigraph data, sleep diary data, and sleep habits were reviewed with the parents and children. Families who were unable to consistently provide data or who were unable to maintain a consistent “lights-out” time (within 30 minutes of the patient’s usual bedtime) were not eligible to participate in the study. Patients were allowed to go to bed 30 minutes later on weekends and holidays, and the usual bedtime during the summer vacation was allowed to be different from the usual bedtime during the school year. A subset of patients (n = 39) volunteered to participate in polysomnographic evaluation following Visit 2 at an American Academy of Sleep Medicine-accredited sleep laboratory. The baseline polysomnogram was also used to objectively assess for the presence of primary sleep disorders (OSA, PLMD). Two consecutive nights of polysomnography were recorded in order to minimize “first-night” effect. The bedtime chosen during polysomnography was the patient’s usual “light-out” time. Patients were allowed to sleep until spontaneous morning awakening at 1 laboratory, whereas the recording was truncated early in the morning at the other laboratory. Following the completion of the second polysomnographic evaluation (or immediately following Visit 3 for patients not undergoing polysomnography), patients were stratified (by whether or not they completed a polysomnographic evaluation) and randomly allocated to begin double-blind treatment with either atomoxetine or methylphenidate. Study Period II was approximately a 7-week period of active treatment with either atomoxetine or methylphenidate (beginning at the end of Visit 3 and continuing through Visit 7) and included a 10- to 20-day study-drug washout phase (Visit 8). It consisted

Study Design This study was a randomized, double-blind, crossover trial comparing atomoxetine and methylphenidate conducted at 2 outpatient sites in the United States. This study was approved by each site’s institutional review board and was conducted in accordance with the ethical principles that have their origin in the Declaration of Helsinki and are consistent with good clinical practices and applicable laws and regulations. Study Period I was a diagnostic assessment period used to screen patients and collect baseline measures (Figure 1). At Visit 1, the study was explained to the patient and his/her parent, and informed consent and assent, if applicable, were obtained. Initial evaluations were completed; use of the parent and child diaries and the actigraphy monitor were discussed; and the parent, patient, and investigator agreed upon a time during the duration of the study when the child would be in bed with “lights out.” “Lights out” was defined as the first time the parents turn out the lights for bedtime. Any awakenings, disruptions, or subsequent turning on or off of the lights were considered an event that should be recorded. On nonschool nights, this bedtime could be up to 30 minutes later than the predetermined school-night “lights-out” bedtime. Visit 2 took place 5 to 12 days after Visit 1. At Visit 2, actigraphy and diary data were reviewed with the child and parent. SLEEP, Vol. 29, No. 12, 2006

1575

Atomoxetine and Methylphenidate Sleep Effects—Sangal et al

of 5 visits separated by approximately 10 days (range, 7-12) for each of Visits 3 through 6, by approximately 15 days (range, 1218) between Visits 6 and 7, and by 10 to 20 days between Visits 7 and 8. The dosing schedules for atomoxetine and methylphenidate are shown in Table 1. Atomoxetine was given twice daily and methylphenidate was given 3 times daily. The morning dose was given before 8:00 AM, the midday dose at or about noon, and the afternoon dose between 4:00 and 5:00 PM. The absolute upper limit for the dose was 120 mg per day for atomoxetine and 60 mg per day for methylphenidate, regardless of the patient’s weight. If a dose increase during Visit 6 was not tolerated well, the dose could be decreased with the following limitations: only 1 decrease in medication dose was allowed during the study period, and the patients must have been able to tolerate at least a total daily dose of 1.0 mg/kg per day of atomoxetine or 0.9 mg/kg per day of methylphenidate. Patients who were unable to tolerate study medication at these doses were discontinued from the study. Between Visits 6 and 7, actigraphy was repeated and a sleep diary was completed in the same fashion as between Visits 1 and 3. At Visit 7, the sleep assessments of Visit 3, including polysomnography, were repeated in the same manner. In addition, an ECG was obtained at Visit 7. Study Period III mirrored Study Period II exactly, except that those patients who received atomoxetine during Study Period II received methylphenidate and vice versa. Patients ceased taking study medication after the final polysomnogram was obtained at Visit 12. Approximately 7 to 10 days after Visit 12, patients were seen for a final discontinuation visit (Visit 13) to assess safety and tolerability. No study medication was dispensed between Visits 12 and 13.

Table 1—Atomoxetine and Methylphenidate Dosing Schedulesa

Visit 3 Visit 4 Visit 5 Visit 6

Methylphenidate Doses divided into thirds and given thrice daily 0.45 0.9 1.35 Can be decreased to 0.9 for patients with tolerability problems or increased to 1.8 for patients with residual ADHD symptoms with an absolute upper limit of 60 mg/day

a All doses are in mg/kg per day. ADHD refers to attention-deficit/hyperactivity disorder.

last minute of that 10-minute period was recorded as the end of sleep. Actigraphy was also used to measure total nap time, assumed sleep time (sum of all epochs scored as sleep from sleep onset to wake time), interrupted sleep time (sum of all epochs scored as wake), number of sleep interruptions (any transition from epoch marked sleep to epoch marked wake), and total sleep interval (sleep onset to wake time, including the total duration of all sleep interruptions). Polysomnography Procedures As noted above, for a subset of patients who agreed to participate (n = 39), polysomnography was conducted on 2 consecutive nights 3 times during the study (at baseline and at the end of each active treatment period [Visits 7 and 12]). Polysomnography was conducted in a sleep-laboratory setting supervised by study-site personnel. The first night of the Visit 3 polysomnography included monitoring of electroencephalogram, electrooculogram, submental electromyogram, ECG, measures of airflow (nasal pressure or nasal thermistor) and respiratory effort, and leg movements to rule out previously undiagnosed primary sleep disorders (e.g., OSA, PLMD). Patients with a respiratory disturbance index of more than 5 per hour of sleep or periodic limb movements with an arousal index of more than 5 per hour of sleep were excluded from the study. The second night of the Visit 3 polysomnography and all other polysomnographies included monitoring of electroencephalography, electrooculography, submental electromyography, and ECG. Variables calculated were total time in bed, time to onset of first sleep epoch; time to onset of persistent sleep (defined as 20 continuous epochs of sleep); total sleep time; time in Stage 1, 2, and 3/4 sleep; number of awakenings; number of arousals; time in REM sleep; REM latency; and sleep efficiency (time asleep divided by time in bed). Arousals were defined using the American Academy of Sleep Medicine, and the arousal index (number of arousals per hour of sleep) was calculated using arousals divided by total sleep time (in minutes) multiplied by 60. An awakening was any epoch scored “wake” that was both preceded and followed by an epoch scored as any stage of sleep. Wake after sleep onset was calculated as the number of minutes awake following the onset of the first sleep epoch. Data from the first night of the baseline polysomnogram have been previously

Actigraphy Procedures Following entry into the study, each patient wore a wrist actigraphy monitor (Actiwatch®, Mini Mitter Co., Inc., Bend, OR) on his or her nondominant wrist all day between Visits 1 and 3, Visits 6 and 7, and Visits 11 and 12 and completed a daily sleep diary during the same periods. The actigraphy monitor measures and stores data regarding body movements of several days to weeks. The raw movement data are then transformed using a standardized software program into approximate sleep-wake patterns. Movements are scored in 1-minute epochs; all epochs that record movement amplitude above a preset threshold (sensitivity level) are scored as “wake,” and those that are below the threshold are scored as “sleep.” Parents and children were trained by study-site staff in the proper methods of putting on the actigraphy monitor and filling out the sleep diary. Parents were asked to ensure that they noted the time of “lights out” each night and also to maintain the child’s agreed-upon bedtime. Actigraphy was used to measure sleep-onset latency, the main dependent measure. Actiware Sleep® software (Mini Mitter Co., Inc.,) was used to determine sleep onset automatically by searching for the first 10-minute interval in which there was activity in no more than 1 epoch that was above the threshold set for determining “wake.” The software then considers the first minute of this 10-minute period as the time of sleep onset. Sleep offset (or wake time) was determined in similar fashion. The record prior to “get up time” was searched for the last 10-minute period containing only 1 epoch with activity above the wake threshold. The SLEEP, Vol. 29, No. 12, 2006

Atomoxetine Doses divided and given twice daily with placebo given as the noontime dose 0.5 then increased to 0.8 1.0 1.5 Can be decreased to 1.0 for patients with tolerability problems or increased to 1.8 for patients with residual ADHD symptoms with an absolute upper limit of 120 mg/day

1576

Atomoxetine and Methylphenidate Sleep Effects—Sangal et al

reported by Sangal and colleagues,17 along with a discussion of the rationale for exclusion based on cut-off values of the respiratory disturbance index and the periodic limb movement arousal index.

sults reported in this paper exclude data from these 3 patients. The secondary objectives of the study were to compare the effects of atomoxetine and methylphenidate on various sleep parameters in children with ADHD, as assessed by actigraphy, polysomnography, and parent- and child-reported sleep diaries, and to compare the efficacy of atomoxetine and methylphenidate for treating ADHD. There were multiple secondary variables compared across treatments (6 actigraphic variables, 14 polysomnographic variables, 4 child-diary questions, and 14 parent-diary questions [results from 2 of the child-diary questions and 3 of the parent diary questions were excluded from analysis due to the large number of missing values for these questions]). Therefore, after a comparison of all variables, a Bonferroni adjustment was performed separately for actigraphy variables, the polysomnography variables, the child-diary variables, and the parent-diary variables. The primary outcome variable was the change from baseline to endpoint in sleep-onset latency, as measured by actigraphy, from both study periods. To test for treatment differences in the primary outcome variable and other continuous data, the sums and differences of endpoint (or change) scores were computed following the method described in Koch20 and Taulbee21 for 2period, 2-treatment crossover designs. Treatment difference was assessed by testing whether the sequence term was significant in the analysis of variance (ANOVA) model applied to difference scores incorporating terms for sequence and investigator. Effect sizes were computed using the Cohen d formula, as the ratio of the difference in raw mean change to pooled standard deviation. In addition, 95% confidence intervals were calculated. Due to nonnormality of the time-to-sleep-onset variable, the rank of the difference scores served as the dependent variable in the ANOVA model. To be included in the primary analysis, patients had to have actigraphy data on 70% of the days in each treatment interval. A posthoc analysis was performed to examine whether the medication responses might vary across age (at or below median age vs above median age) and sex. Treatment difference was assessed by using the same ANOVA model as in the primary analysis within these patient subgroups. Due to the extremely small sample size in some of these subgroups, a test for interaction of treatment and either age group or sex was not performed. Characteristics (demographics and baseline psychiatric and sleep measures) were summarized for all enrolled patients. For these analyses, only data collected during Visits 1 to 3 were considered baseline. Treatment-sequence differences in categorical baseline characteristics were assessed using the Fisher exact test, and treatment-sequence differences in continuous baseline characteristics were assessed using 1-way ANOVA with a term for sequence. In addition to change-from-baseline scores from the ADHD RS, a response analysis that determined the percentages of responders was also calculated using the analysis method proposed by Nagelkerke22 for crossover designs. Using this method, patients are classified according to their response during each treatment period (1= Responder, 0 = Nonresponder). So, a patient who responded in period II but not period III would be coded (1,0), whereas a patient who responded in both periods would be coded (1,1). A 2×3 table is constructed with 3 columns: patients who responded to both treatments, patients who failed to respond to both treatments, and patients who responded to only 1 of the 2

Sleep Diaries Sleep diaries included items that were used to record the parent’s and child’s assessment of the child’s sleep and the child’s morning and evening behavior. The parent diary included 17 items. Most items used a 0- to 4-point scale. The child diary included 6 items. Most items used 100-mm visual analog scales. The child put a mark on the line indicating his/her response on this continuum, and the response was converted to a 0- to 100-point score. Efficacy Measures Efficacy measures were collected using the ADHD RS16 (Visit 1 and at the end of each study period), the Clinical Global Impression-Severity scale18 (Visits 1 and 3-12), the Conners’ Parent Rating Scale-Revised: Short Form (CPRS-R:S)19 (Visit 1 and at the end of each study period), and the Daily Parent Ratings of Evening and Morning Behavior (DPREMB) (Visits 1-3, 6, 7, 11, and 12). The DPREMB is a 13-item, parent-rated daily diary14 that assesses ADHD symptoms during evening and early-morning periods. Symptoms assessed included late-day and early-morning inattentiveness/distractibility, ability to concentrate on structured tasks, hyperactivity/impulsivity, and oppositionality. Each item was rated on a 5-point scale (0 = not present, to 4 = extremely problematic). Data Analysis The primary objective of this study was to test the hypothesis that the effects of atomoxetine on sleep differ from those of methylphenidate, as assessed by sleep-onset latency, as measured by actigraphy. Only the data from patients who completed both treatment periods were included in the sleep and efficacy analyses. For safety measures, all patients who took at least 1 dose of study drug and at least entered Study Period III were included. The planned sample size of 86 patients was selected to ensure that 60 patients would complete both study periods and their data would be included in the primary-efficacy analysis. This sample size provides at least 85% power for detecting a treatment difference in change from baseline of approximately 14 minutes in time to onset of persistent sleep, as measured by actigraphy, assuming a within-subject standard deviation of 25.5 minutes and a 2-sided test at the .05 α level. This sample size also ensures at least 70% power for the primary analysis if only the first period treatment data were used. It was assumed that approximately 30 patients would complete the polysomnography measurements, providing approximately 80% power to detect treatment-effect sizes of at least 0.72. Subsequent to unblinding, it was noted that 3 of the patients had unreliable actigraphy data, perhaps due to a problem with their Actiwatches®. At the primary investigator’s request, the actigraphy analyses were rerun excluding data from these 3 patients. The numerical means and standard deviations were affected by the exclusion of these data; however, conclusions from the analyses without the data from these 3 patients are identical to those with the data from these 3 patients included. Therefore, actigraphy reSLEEP, Vol. 29, No. 12, 2006

1577

Atomoxetine and Methylphenidate Sleep Effects—Sangal et al

treatments. A χ2 test of homogeneity with 2 degrees of freedom is then performed on this 2×3 table. A response criterion can be set to include robust responders only,23 or to include partial as well as robust responders. In this protocol, the a priori criterion for response was a decrease of at least 25% of their baseline ADHD RS Total score. This criterion would include partial responders as well as robust responders. For the diary and actigraph data, patient scores were averaged over the visit during which the scores were collected. The averages were considered missing if there were fewer than 70% of the scores present. This restriction was to ensure that data from only patients who provided sufficient scores to get an accurate reflection of their response to treatment were included in the analysis. This eligibility criterion was specified a priori in the analysis plan. The DPREMB totals from the sleep diary were analyzed as follows: for the average score to be computed, a patient must have nonmissing scores for at least 9 of 14 days for the 14-day window average or 4 of the 7 days for a 7-day window average. This completion rate was lowered from the 70% in the original protocol due to the lower number of patients with 70% or more responses. The weekly window approach was determined after examining the variability in the visit interval for patients prior to unblinding. Although the DPREMB questions were also collected as a scale separate from the sleep diary, they were not analyzed due to a large proportion of missing data.

Table 2—Summary of Demographics and Other Patient Characteristics for the 85 Patients Randomly Assigned to Treatment Characteristica Sex Boys Girls Age, y Origin Caucasian Non-Caucasian ADHD Subtypec Hyperactive/Impulsive Inattentive Combined Prior Stimulant Exposure No Yes Present Comorbid Conditions ODD Conduct Disorder Anxiety Agoraphobia

No. (%)b 64 (75.3) 21 (24.7) 10.1 ± 2.0 62 (72.9) 23 (27.1) 2 (2.4) 25 (29.8) 57 (67.9) 37 (43.5) 48 (56.5) 41 (48.2) 3 (3.5) 1 (1.2)

There were no statistically significant between-group differences for any characteristic at baseline. b Data are presented as number (percentages) except age, which is presented as mean ± SD. ADHD refers to attention-deficit/hyperactivity disorder; ODD, oppositional defiant disorder. c Subtype based on K-SADS-PL; 1 patient did not meet KSADS ADHD criteria. a

RESULTS Of 107 patients screened, 85 met inclusion criteria and were randomly assigned to treatment (44 to atomoxetine/methylphenidate, 41 to methylphenidate/atomoxetine; Figure 2). Of the 22 patients who failed to meet enrollment criteria, 15 failed due to patient conflict or patient/caregiver decision, 5 failed to meet protocol entry criteria, 1 was lost to follow-up, and 1 was not enrolled due to a physician decision. One patient was excluded on the basis of a history or symptoms suggestive of a primary sleep disorder such as OSA or periodic limb movement disorder. Baseline characteristics for the atomoxetine/methylphenidate and methylphenidate/atomoxetine groups were similar, and the baseline characteristics for all randomly assigned patients are summarized in Table 2. The mean final atomoxetine dose was 58.27 mg/day (range = 15-100), or 1.56 mg/kg per day. The mean final methylphenidate dose was 42.29 mg/day (range = 15-60), or 1.12 mg/kg per day. Drug compliance, measured by pill counts at each visit, ranged

from 83.3% to 100% for atomoxetine and from 87.5% to 100% for methylphenidate. Sleep Measures Actigraphy There was a significant increase in sleep-onset latency, relative to baseline, when patients were being treated with methylphenidate (Table 3; Figure 3). Treatment with atomoxetine increased sleeponset latency by 12.06 minutes, compared with 39.24 minutes for methylphenidate (p < .001). Total sleep interval decreased from baseline significantly more for methylphenidate (-35.89 minutes), as compared with atomoxetine (-15.00 minutes; p = .004). The number of sleep interruptions was decreased from baseline with both atomoxetine and methylphenidate (-1.31 vs -4.36 for atomoxetine and methylphenidate, respectively; p = .011). Interrupted

Table 3—Actigraphic Sleep Measures During Atomoxetine and Methylphenidate Treatmenta Sleep Measure

Baseline

Atomoxetine

Methylphenidate

Endpoint Change Endpoint Sleep-onset latency, min 30.11 ± 24.84 42.17 ± 31.61 12.06 ± 27.07 69.35 ± 43.86 Total nap time, min 3.47 ± 5.32 7.97 ± 10.11 4.49 ± 10.41 6.51 ± 7.30 Total sleep interval, min 518.82 ± 44.13 503.82 ± 50.97 -15.00 ± 45.10 482.93 ± 62.64 Assumed sleep time, min 457.41 ± 47.34 442.14 ± 50.63 -15.26 ± 44.25 427.80 ± 57.20 Interrupted sleep time, min 61.41 ± 20.85 61.67 ± 20.00 0.26 ± 15.04 55.13 ± 20.61 Sleep interruptions, no. 31.78 ± 7.79 30.47 ± 10.42 -1.31 ± 6.83 27.42 ± 9.62

Atomoxetine vs Methylphenidate Change p Value Effect Size 95% CI 39.24 ± 40.77 < .001b -.79 -12.82, -6.49 3.04 ± 7.92 .475 .16 -1.68, 3.55 .41 6.81, 34.15 -35.89 ± 56.10 .004b -29.61 ± 53.00 .016 .29 2.73, 25.73 -6.28 ± 17.48 .025 .40 0.80, 11.69 -4.36 ± 6.33 .011 .46 0.70, 5.19

Data are from 50 subjects, except sleep interruptions, which were from 48 subjects both effect sizes; 95% confidence intervals computed based on methylphenidate subtracted from atomoxetine. Baseline, endpoint, and change data are presented as mean ± SD. CI refers to confidence interval. b p Value remained significant after a Bonferroni adjustment for multiple comparisons. a

SLEEP, Vol. 29, No. 12, 2006

1578

Atomoxetine and Methylphenidate Sleep Effects—Sangal et al

N=107 Patients Entered

n = 22 Screening Failures

n = 85 Patients Randomized

Reasons: pat conflict; pat/crgvr decis (15) entry criteria not met (5) lost to follow-up (1) phys decis (1)

n=44 ATX

n = 42 (95.5%) Completed

n = 41 MPH

n = 37 (90.2%) Completed

n=2 Withdrawn Reason: pat decis (2)

n = 37 ATX

n = 42 MPH

n = 41 (97.6%) Completed

n=4 Withdrawn Reason: lost to follow-up (2) adverse event (1) prot viol (1)

n=1 Withdrawn Reason: prot viol (1)

Analyzed (n=27) Excluded from analysis (n=14) < 70% of actigraphy scores (n=13) Actigraphy watch malfunction (n=1)

n = 34 (91.9%) Completed

n=3 Withdrawn Reason: lost to follow-up (2) pat decis (1)

Analyzed (n=23) Excluded from analysis (n=11) < 70% of actigraphy scores (n=9) Actigraphy watch malfunction (n=2)

Figure 2—Overview of patient disposition. pat conflict refers to patient conflict; pat/crgvr decis, patient/caregiver decision; prot viol, protocol violation; loe, lack of efficacy; pat decis, patient’s decision; phys decis, physician’s decision. ATX refers to atomoxetine; MPH, methyphenidate.

sleep time was slightly increased for atomoxetine, compared with a small decrease for methylphenidate (0.26 vs -6.28 minutes for atomoxetine and methylphenidate, respectively; p = .025). Atomoxetine decreased assumed sleep time to a lesser extent than did methylphenidate (-15.26 vs -29.61 minutes for atomoxetine and methylphenidate, respectively; p = .016). After performing a Bonferroni adjustment for comparing multiple variables, only the difference in sleep-onset latency and total sleep interval remained statistically significant. Children on atomoxetine fell asleep faster and had less decrease in total sleep interval. A subgroup analysis that examined the effects of atomoxetine SLEEP, Vol. 29, No. 12, 2006

and methylphenidate on sleep-onset latency in patients at or below the median age indicated that methylphenidate produced a significantly greater increase, compared with atomoxetine (p = .001). The same was true for the patients older than the median age (p < .001). However, total sleep interval, interrupted sleep time, and number of sleep interruptions were decreased significantly more with methylphenidate than with atomoxetine in the older group only. There were no significant treatment differences in nap time or assumed sleep time in either age group. An analysis that examined the effects of atomoxetine and methylphenidate, looking at each sex individually, indicated that methylphenidate 1579

Atomoxetine and Methylphenidate Sleep Effects—Sangal et al

Mean Change from Baselin) e (min)

45

movement disorder (periodic limb movement arousal index ≥ 5) on the Baseline polysomnogram. Consistent with the actigraphy data, the polysomnography data (Table 4; Figure 3) indicated that the mean time to onset of first sleep epoch was significantly increased relative to baseline for methylphenidate (-0.31 vs 16.79 minutes for atomoxetine and methylphenidate, respectively; p < .001), as was the time to onset of persistent sleep (-0.22 vs 16.81 minutes for atomoxetine and methylphenidate, respectively; p < .001). Relative to methylphenidate, atomoxetine slightly increased the percentage of time in Stage 2 sleep (6% vs 3%; p = .042). These percentages correspond to an increase of 29.87 minutes with atomoxetine and 13.27 minutes with methylphenidate (p = .020). Similar to the actigraphy results, the number of awakenings was decreased from baseline more for methylphenidate than for atomoxetine (-4.41 vs -6.71 for atomoxetine and methylphenidate, respectively; p = .002). Wake after sleep onset was also decreased from baseline more for methylphenidate than for atomoxetine (-3.62 vs -11.10 for atomoxetine and methylphenidate, respectively; p = .009). Atomoxetine significantly increased latency to REM sleep (35.17 vs -3.55 min; p = .001), and there was a minor decrease in the percentage of time in REM sleep with atomoxetine (-1% atomoxetine vs 1% methylphenidate; p = .007). There were no other significant differences between atomoxetine and methylphenidate in terms of percentage of time in other sleep stages (Stages 1 and 3/4) relative to baseline. Although total sleep time did not differ by treatment condition, sleep efficiency was statistically increased relative to baseline in the atomoxetine condition (p = .020), although this slight difference is unlikely to be clinically significant. After performing a Bonferroni adjustment for comparing multiple polysomnographic variables, differences in (1) time to onset of first sleep epoch, (2) time to onset of persistent sleep, (3) number of awakenings, and (4) REM latency remained statistically

*** ATX MPH

35

25

*** 15

5

-5

Actigraphy

Polysomnography

Figure 3—Effects of atomoxetine (ATX) and methylphenidate (MPH) on mean change from baseline in sleep onset, in minutes. ***p < .001 for comparison of change scores between ATX and MPH.

produced significantly greater increases in sleep-onset latency compared, with atomoxetine, for both boys and girls (p values