Stimulants

• Immediate-release compared with extended-release formulations
  • The evidence regarding immediate-release methylphenidate compared with methylphenidate OROS was conflicting, with 2 double-blind trials unable to identify differences, while 2 open-label studies found that methylphenidate OROS resulted in greater improvements on some but not all assessments.

    –

    Exploratory pooled analysis of the inattention/overactivity scores of the IOWA Conners' scale indicated that methylphenidate OROS may result in greater improvement (weighted mean difference −1.19; 95% CI, −1.78 to −0.60).

  • Limited evidence was available for the comparisons of immediate-release methylphenidate to other extended-release formulations. Overall, the studies were unable to identify differences between methylphenidate SR and immediate-release methylphenidate, and methylphenidate CD was found to be noninferior to immediate-release methylphenidate.
  • Database studies using intermediate outcomes reported greater persistence with methylphenidate OROS and methylphenidate SODAS compared with immediate-release methylphenidate. Methodologic concerns indicate caution in interpreting this evidence.
• Sustained-release compared with sustained-release formulations
  • Limited evidence from 2 small crossover studies suggested that methylphenidate LA is superior to methylphenidate OROS on some, but not all efficacy outcomes. However, these results should be interpreted with caution until higher quality evidence is available.
  • Methylphenidate CD was associated with significantly larger effect sizes than methylphenidate OROS in the morning, treatment effects were similar in the afternoon, and methylphenidate OROS was superior in the evening. Methodologic concerns indicate caution in interpreting these findings.

    –

    Methylphenidate OROS had statistically significantly higher rates of insomnia and decreased appetite than methylphenidate CD.

  • Dexmethylphenidate ER resulted in better response from 0 to 2 hours post dose compared with methylphenidate OROS (primary outcome measure). A difference was found up to 6 hours post dose, but methylphenidate OROS resulted in better scores later in the day; from 10 to 12 hours post dose.
  • There was no evidence of a difference in adverse events between immediate-release and sustained-release formulations.
• Dextroamphetamine compared with methylphenidate
  • The body of evidence clearly indicated no difference in efficacy between immediate-release dextroamphetamine and immediate-release methylphenidate. Evidence from short-term trials and observational studies suggested that weight loss is greater with immediate-release dextroamphetamine than immediate-release methylphenidate.
• Mixed amphetamine salts compared with methylphenidate
  • Immediate-release mixed amphetamine salts were superior to immediate-release methylphenidate on a few efficacy outcome measures in 2 trials, but clear evidence of superiority was lacking. Very limited evidence suggested that twice daily dosing of immediate-release mixed amphetamine salts led to higher rates of loss of appetite and sleep trouble than once daily dosing of immediate-release methylphenidate.
• Modafinil compared with methylphenidate
  • Differences were not found between modafinil and immediate-release methylphenidate over 6 weeks.

    –

    Response rate (>40% reduction in score): Modafinil 73% compared with immediate-release methylphenidate 70% for parents rating.

    –

    Rates of adverse events were similar between the drugs.

• Dextroamphetamine compared with mixed amphetamine salts
  • Evidence of immediate-release dextroamphetamine compared with dextroamphetamine SR compared with mixed amphetamine salts was limited and conflicting, but may suggest that measures in the morning find immediate-release dextroamphetamine superior to dextroamphetamine SR, and measures in the afternoon find dextroamphetamine SR superior to mixed amphetamine salts. Transient weight loss was greater with mixed amphetamine salts and dextroamphetamine SR than with immediate-release dextroamphetamine. However, this evidence should be interpreted with caution.
• Lisdexamfetamine compared with mixed amphetamine salts XR
  • Evidence from the US Food and Drug Administration medical review and manufacturer-submitted data dossier suggests that mean Swanson, Kotlin, Agler, M-Flynn, and Pelham-Deportment Subscale (SKAMP-DS) scores were similar in children following 1 week of lisdexamfetamine or Adderall XR^®. Rates of specific adverse events were not available for the individual treatment groups, but the data dossier did not specify any differences between them.
• Transdermal methylphenidate compared with methylphenidate OROS
  • Methylphenidate transdermal system was found to have similar efficacy to methylphenidate OROS over a 7-week period, based on investigator, teacher, and parent ratings starting 4 hours after administration of dose or application of patch and weekly. Methylphenidate transdermal system was found to have similar efficacy to immediate-release methylphenidate over 12 hours in a simulated classroom setting, starting 30 minutes after dosing.
  • Although rates of adverse events and discontinuations due to adverse events were greater with transdermal methylphenidate than with methylphenidate OROS, differences were not found to be statistically significant. Differences in adverse events were not found between methylphenidate transdermal system and immediate-release methylphenidate.
• Longer-term studies indicated that although the evidence is somewhat mixed, efficacy benefits seen with immediate-release methylphenidate can be maintained over periods of up to 24 months, but that deterioration in benefit is seen with longer follow-up.

Nonstimulants

Comparison of immediate-release and sustained-release formulations

Methylphenidate. We included 13 trials of immediate-release methylphenidate compared with methylphenidate SR.^39-49 Of these, 4 were poor quality due to either inadequate or undescribed methods of randomization and allocation concealment, combined with lack of description of an intent to treat analysis, lack of information on eligibility criteria, attrition, or post randomization exclusions (Evidence Table 3).^{39, 40, 44, 46} The remaining studies compared immediate-release methylphenidate to 5 extended-release formulations of methylphenidate (Biphentin^®, Concerta^®, Ritalin SR^®, Medikinet^®, or Metadate CD^®).^{41-43, 45, 47-49}

No trials comparing the other extended-release formulations of methylphenidate (Ritalin LA^®, Methylin ER^®, or Metadate ER^®) to immediate-release methylphenidate were found. Table 4, below, presents basic pharmacokinetic information on the methylphenidate products.

Table 4

Pharmacokinetic profiles of methylphenidate products.

Immediate-release methylphenidate compared with methylphenidate OROS (Concerta^®). Five studies have compared immediate-release methylphenidate with methylphenidate OROS once daily, enrolling a total of 561 children with ADHD (Table 5).^{41, 42, 48-50}

Table 5

Trials of immediate-release methylphenidate compared with methylphenidate OROS (Concerta^®).

In the 3 double-blind trials submitted to the US Food and Drug Administration, in which the primary comparison of interest was specified as methylphenidate OROS compared with placebo, methylphenidate OROS and immediate-release methylphenidate did not differ significantly on the majority of direct comparisons.^{41, 42, 50} In contrast, the 2 newer, open-label studies did find a significant difference favoring methylphenidate OROS.^{48, 49} There is a potential risk of selection bias in that only 1 of the studies⁴¹ reported the proportion of patients taking immediate-release methylphenidate or methylphenidate OROS prior to enrollment. The US Food and Drug Administration Statistical Review of the New Drug Application for methylphenidate OROS includes criticism of the trials submitted for product approve,^{41, 42, 50} indicating that an assumption of equivalence should not be made based on these studies alone. (http://www.fda.gov/cder/foi/nda/2000/21-121_Concerta_statr.pdf - page 32).⁵¹

In the largest, highest-quality study, there were no significant differences between the formulations on the primary outcome measure (IOWA Conners' scale) or on 11 secondary measures in a randomized controlled trial of 312 children.⁴² Similarly, a much smaller crossover trial (68 children) that was 7 days long and included behavioral treatment, found methylphenidate OROS to have lower scores on the Abbreviated Conners' Parents scale (total), and on the inattention/overactivity item (out of 16 items), however no differences were found based on assessments made by teachers and counselors.⁴¹ An additional study of 64 children was rated poor quality because it lacked adequate reporting on multiple measures to provide meaningful results.⁵⁰

The study by Steele, et al.⁴⁹ was open-label, comparing usual care to switching to methylphenidate OROS. Based on a definition of remission as a score of 0 or 1 (none or just a little) on the 18 items relating to ADHD symptoms only (excluding the items pertaining to oppositional defiant disorder) of the parent assessed SNAP-IV scale, methylphenidate OROS treatment resulted in more patients being classified as in remission at 8 weeks, with a number needed to treat near 4 (see Table 5). Similar results were found using other measures of parental assessment. This study did not include teacher ratings. Because the study was open to patients currently receiving treatment, including immediate-release methylphenidate, and it was unblinded, it is potentially biased against immediate-release methylphenidate. The proportion of patients taking immediate-release methylphenidate, methylphenidate OROS, or who were not taking drug therapy prior to study enrollment was not reported.

We undertook an exploratory analysis, pooling the parent ratings of inattention/overactivity subscale items of the IOWA Conners' scale from these 3 studies, as it was the only item reported across all 3 (see Table 5). While the Wolraich and Pelham studies did not find significant differences in the mean change on this item, the pooled analysis with the Steele study does result in a statistically significant finding, favoring methylphenidate OROS (weighted mean difference, −1.19; 95% CI, −1.78 to −0.60). However, we did consider this an exploratory analysis because standard deviations were not provided in the Pelham and Wolraich studies and we made an assumption that the baseline and final scores were moderately correlated (r² = 0.25).

A fourth study conducted in Taiwan found methylphenidate OROS superior to immediate-release methylphenidate, assessing the change in Conners' Teacher Rating Scale Revised Short-Form score by either teacher or parent over 5 time points using a linear mixed model, P<0.0001 (see Table 4). The absolute difference in individual scores were not large (Table 4), with the largest difference in teacher ratings being 1.12 for oppositional defiant behaviors (out of 5 possible), and 1.69 for hyperactivity/impulsivity (out of 7 possible) in the parent ratings. This study had the same potential for bias as the unblinded study by Steele, except that here all patients had previously been taking some form of methylphenidate, but again the proportions taking immediate-release methylphenidate compared with methylphenidate OROS or other formulations prior to enrollment was not reported.

In contrast, findings from a retrospective study of 92 children from a “real-life clinical situation” in the United Kingdom suggested that 32% (P<0.001) were considered treatment failures when switched to an extended-release form of methylphenidate (Concerta XL^®) from immediate-release methylphenidate of an unknown duration.⁵² The validity and generalizability of these findings were unclear, however, as the study was retrospective in nature, physicians' use of personal case load to identify patients may have introduced a selection bias, treatment failure was not precisely defined, and it was unclear whether the United Kingdom formulation is comparable to methylphenidate OROS as included in this review.

Immediate-release methylphenidate compared with methylphenidate SR (Ritalin SR^®). A small 2-week randomized controlled trial (34 children) of immediate-release methylphenidate compared with methylphenidate SR found mixed results.⁴³ The outcome measures included questionnaires (not validated) completed by a physician, a teacher, and a parent. The teacher questionnaires indicated significant differences in final total score and the “Conduct Problem” scores favored immediate-release methylphenidate. Parent questionnaires indicated a significant difference favoring methylphenidate SR on the “Conduct Problem” item final score, and the physician scores showed no difference.

Immediate-release methylphenidate compared with methylphenidate ER (Metadate CD^®, Equasym^®). A 3-week study using over-encapsulation for blinding enrolled 327 children, comparing immediate-release methylphenidate to Equasym^® (sold in the United States as Metadate CD^®). The study analyzed only 87% of patients in the main per-protocol analysis with unclear description of those excluded.⁴⁷ The study included a non-inferiority analysis, assuming a difference of ≤ 1.5 points on the I/O score of the Conners' IOWA teachers rating scale to indicate equivalence (non-inferiority). At weeks 1, 2, and 3 immediate-release methylphenidate was found equivalent to Equasym^®. Intent-to-treat analysis as well as subgroup analyses (country, dose, ADHD subtype) was reported in the discussion as supporting these results. Additional analysis examined the effects of the drugs in the morning and afternoon, but a direct comparison was made only to the placebo group as both methylphenidate groups were found similarly superior to placebo at both time points throughout the study.

Immediate-release methylphenidate compared with methylphenidate multilayer-release (Biphentin^®). Two small, fair-quality, crossover studies compared immediate-release methylphenidate to methylphenidate multilayer-release (Biphentin^®, available in Canada, not available in the United States as of September 2011).^{53, 54} In the first study, 90 children were randomized to either immediate-release methylphenidate or methylphenidate multilayer-release and had dose titration over 2-3 weeks, with observation by parent, teacher, and investigator over 2 weeks.⁵⁴ Discontinuations were similar between groups (86% methylphenidate multilayer-release, 89% immediate-release methylphenidate), and mean daily doses were similar between treatments (0.8 mg/kg daily). Using the Conners' scales, “normal” was defined as a final T-score of <65 on each of the 4 subscales. After 5 weeks of treatment, more children taking immediate-release methylphenidate had achieved a normal score on the ADHD Index compared with those taking methylphenidate multilayer-release (90% compared with 79% on the teacher scale and 81% compared with 77% on the parent scale). The authors reported that the mean ADHD Index T-scale score was statistically significantly better (lower) with immediate-release methylphenidate based on the teacher scale (mean differences, 3.12%; 95% CI, 1.51 to 4.73) but not significant on the parent scale (mean differences, 0.38%; 95% CI, −1.34 to 2.10). No other differences were found between treatment groups.

The second, smaller study (N=18) reported only single-day measurements after 1 week of immediate-release methylphenidate, methylphenidate multilayer-release, or placebo.⁵³ This study found no statistically significant differences between drug treatments on the Conners' IOWA scale, although baseline scores differed across treatment groups such that these findings should be interpreted with caution; the analyses attempted to control for differences in baseline scores, including assessing for carryover effects. Analyses of time-course responses were not able to identify consistent differences among the drugs compared with placebo.

Other measures of comparative effectiveness of immediate-release compared with sustained-release formulations

Clinical trials of extended-release compared with immediate-release formulations were too short to demonstrate differences in long-term health outcomes. However, the intermediate outcome measure of persistence (the proportion of patients continuing to take or refill prescriptions for a medication after some longer period of time) is thought to be a good proxy for extension of benefits seen in the short-term, or if none were found, evidence of a difference in longer-term, real-life settings. Persistence is an intermediate outcome with unknown validity because direct evidence of a relationship between persistence rates and long-term health outcomes with ADHD drugs is lacking.

In 5 observational studies (6 publications), persistence with treatment with long-acting stimulant formulations (methylphenidate OROS or methylphenidate ER) was significantly longer compared with shorter-acting formulations (immediate-release methylphenidate or immediate-release mixed amphetamine salts) over periods of 6 months and 12 months^55-58 following index prescription. One of these studies examined only adults treated with methylphenidate OROS (median duration of treatment 68 days; 95% CI, 65 to 71) compared with immediate-release methylphenidate (39 days; 95% CI, 33 to 52).⁵⁹ The findings of these studies should be interpreted with caution, however, until confirmed by a randomized controlled trial that would serve to rule out potential sources of bias, including between-group baseline differences in unmeasured clinical characteristics, physicians' prescribing preferences, and differences in reasons for discontinuation (e.g., change in insurance benefit and use of promotional samples). We rated these studies fair quality.

Data were derived from the Integrated Health Care Information Services National Managed Care Benchmark Database in 2 studies from the same group of researchers, with overlapping data. Using a definition of persistence as less than a 15-day gap in prescription refills, the studies found methylphenidate OROS to be associated with greater persistence rates than immediate-release methylphenidate (12% compared with 1%, P<0.0001⁵⁶ and 15% compared with 3%, P<0.0001).^{57, 58} The second study also reported persistence using less than a 30-day gap in refills as the definition and found 33% persistent with methylphenidate OROS and 5% with immediate-release methylphenidate.^{57, 58} There was uncertainty about how well this study population represented patients in actual practice as ethnicity and comorbidity characteristics were not reported and there were age and diagnosis differences between those receiving methylphenidate OROS compared with immediate-release methylphenidate.

California Medicaid claims files from a 3-year period were examined to identify youth prescribed methylphenidate (N=11 537).⁵⁵ This study population involved a lower than average proportion of White patients (45.3%) and higher proportions of Hispanic patients (26.1%). Total mean duration (days) of treatment without any 30-day gaps was greater for patients taking extended-release formulations (combined group of methylphenidate OROS = 83%, methylphenidate ER = 8.7%, methylphenidate SODAS = 8.3%) than for those taking immediate-release methylphenidate (140.3 compared with 103.4; survival time ratio, 1.37; 95% CI, 1.32 to 1.42). Subgroup analysis results suggested that persistence duration was greatest for methylphenidate OROS (147.2 days; 95% CI, 142.6 to 151.7 days) compared with methylphenidate SODAS (113 days; 95% CI, 100.9 to 125.1 days) or methylphenidate CD (101.1 days; 95% CI, 91.2 to 111.0 days). Together, extended-release formulations extended persistence duration regardless of ethnicity.

The Texas Medicaid Vendor Drug Program database was used to identify claims for newly started stimulants (2001-2002 school year).⁶⁰ Prescription refill patterns for children (75.7% male; mean age 9.93 years) with new claims for either immediate-release mixed amphetamine salts (n=3425), immediate-release methylphenidate (n=3343), or methylphenidate OROS (n=2781) were evaluated over 6-month assessment periods. Proportion of days of treatment without any 15-day gaps was greater for patients taking methylphenidate OROS than for immediate-release methylphenidate or immediate-release mixed amphetamine salts (0.5 compared with 37 compared with 42; P<0.001), as was proportion of patients that continued receiving therapy for 151-180 days (30.23% compared with 13.62% compared with 18.89%; P<0.001). Within those days of treatment, compliance rates, as measured using the Medication Possession Ratio, were higher in patients taking methylphenidate OROS compared with immediate-release methylphenidate or immediate-release mixed amphetamine salts (0.76 compared with 0.69 compared with 0.73; P<0.001).

Comparisons of sustained-release formulations

Methylphenidate OROS (Concerta^®) compared with methylphenidate CD (Metadate CD^®). Results from the fair-quality COMACS crossover study of 184 children suggested that relative improvements in SKAMP deportment and attention scale scores differed for the comparison of methylphenidate OROS 18-54 mg and methylphenidate CD 20-60 mg (both given once daily) depending on time of assessment.^{61, 62} Methylphenidate CD was associated with significantly larger effect sizes than methylphenidate OROS in the morning, while treatment effects were similar in the afternoon, and methylphenidate OROS was superior in the evening. This study presented several problems, however, in that the SKAMP scale has been criticized for lack of sensitivity to change in symptoms, and that ANOVA analysis found the interaction of site × treatment × sequence (the order to randomization within patients) was found to be statistically significant. This finding resulted in the authors conducting additional analyses; however the effect of sequence was not included in these subsequent analyses. Therefore, these findings should be interpreted with caution.

Methylphenidate OROS (Concerta^®) compared with methylphenidate SODAS (Ritalin LA^®). Two small crossover studies have found methylphenidate SODAS superior to methylphenidate OROS. A small 1-week crossover study of methylphenidate SODAS 20 mg compared with methylphenidate OROS 18 mg and 36 mg⁶³ found methylphenidate SODAS superior on the attention or deportment subscores of the SKAMP scale depending on the time-point and dose comparison. Secondary outcome assessment also found methylphenidate SODAS superior on 1 measure (proportion correct on math test). These limited differences were mitigated by concerns over the assessment tool (SKAMP) sensitivity, use of a simulated classroom, involvement of study sponsor in authorship, and differences in groups at baseline. A similar second crossover study of methylphenidate OROS (18 and 36 mg) and methylphenidate SODAS (20 and 40 mg) also assessed children in a simulated classroom setting after a single dose of the study medication using the SKAMP scale.⁶⁴ Here methylphenidate SODAS 40 mg was found superior to methylphenidate OROS 36 mg at all time points (0-4, 0-8, and 0-12 hours) based on the SKAMP attention subscale score area under the curve analyses, while methylphenidate SODAS 20 mg was not significantly different to either dose of methylphenidate OROS. Here, concerns over the clinical importance of the difference in area under the curve, involvement of study sponsor in authorship, and the impact of sequence of randomized treatment (analysis of treatment sequence was stated to be planned but results not reported) were present.

Dexmethylphenidate ER compared with methylphenidate OROS. A single, small (N=84) fair-quality crossover study compared 2 doses of dexmethylphenidate ER with 2 doses of methylphenidate OROS or placebo using a simulated classroom assessment.⁶⁵ The primary outcome was the mean change in the SKAMP combined score from zero to 2 hours post dose in the dexmethylphenidate ER 20 mg daily group compared with the methylphenidate OROS 36 mg daily group. Children were given the intervention for 7 days prior to the assessment. The mean change in SKAMP combined scores at 2 hours post dose was statistically significantly greater with dexmethylphenidate ER 20 mg daily compared with methylphenidate OROS 36 mg daily (adjusted mean change −11 compared with −6; P<0.001). Similar results were found comparing the higher doses (30 mg dexmethylphenidate ER and 54 mg methylphenidate OROS daily) to each other. At other time points, the drugs differed depending on the time of day. For time points up to 6 hours, dexmethylphenidate ER had statistically significantly superior change in SKAMP combined scores comparing either the 2 lower doses or the 2 higher doses to each other (P values ranged from <0.001 to =0.044). Similarly, a statistically significant difference was seen at the first time point, 0.5 hours (P=0.044). However, at later time points (10, 11, and 12 hours post dose), methylphenidate OROS had statistically significantly superior change in SKAMP combined scores (P values ranged from <0.001 to <0.05). At hours 7, 8, and 9 there was no statistically significant difference between the drugs at either dose levels and analysis by Area Under the Curve from 0-6 and 6-12 hours was unable to identify statistically significant differences between the drugs. Analysis of attention and deportment subscale scores showed similar results. Assessments of math scores and problems attempted showed dexmethylphenidate ER superior up to 4 hours post dose and methylphenidate OROS superior at 11 and 12 hours post dose. In comparison to placebo, dexmethylphenidate ER was superior on SKAMP combined scores starting at 0.5 hours but was not statistically different to placebo at 12 hours. Methylphenidate OROS was superior to placebo starting at 1 hour (not at 0.5 hours) and remained superior through 12 hours.

According to the US Food and Drug Administration Medical Review,⁶⁶ data from 2 short-term, randomized, placebo-controlled, double-blind efficacy trials were submitted to the US Food and Drug Administration in the New Drug Application for dexmethylphenidate ER.^{67, 68} Both were fair quality. Study 2301 was a 7-week, parallel-group, flexible-dosing trial of 103 children.⁶⁷ Study US08 was a 2-week, fixed-dose, crossover trial of 54 children.⁶⁸ Dexmethylphenidate ER was significantly superior to placebo for both primary outcomes of change from baseline to final visit in Conners' ADHD/DSM-IV Scale-Teacher version in Study 2301 (−16.3 compared with −5.7 points; P<0.001) and of mean change in SKAMP combined scores from predose to 1-hour post dose in Study US08 (−10.014 compared with 0.078 points, P<0.001).

Four small, fair-quality placebo-controlled trials have been conducted with dexmethylphenidate ER. ^67-70 A 7-week, parallel-group, flexible-dosing trial of 103 children found dexmethylphenidate ER significantly superior to placebo in change from baseline to final visit in Conners' ADHD/DSM-IV Scale-Teacher version (−16.3 compared with −5.7 points; P<0.001).⁶⁷ Three crossover studies of dexmethylphenidate ER 20 mg daily evaluated response on the SKAMP scale in a laboratory classroom setting. All found dexmethylphenidate ER superior to placebo on the primary outcome measure of mean change in SKAMP combined score over 1 to 8 or 12 hours post dose. Secondary analyses assessed differences at early time points; 2 studies found a statistically significant difference on mean change in the combined score at 0.5 hours (−2.2 dexmethylphenidate ER compared with 3.5 placebo; P=0.001⁷⁰ and −0.969 dexmethylphenidate ER compared with 3.336 placebo; P=0.001),^70,69 and the third found a difference starting at 1 hour post dose (−10.014 compared with 0.078; P<0.001).⁶⁸ Lack of adequate variance data prevented pooling of these results. Because these are crossover studies, carryover effects must be taken into account, however results of such analyses were not reported.

No direct comparisons of other extended-release formulations of methylphenidate or other ADHD drugs were found.

Methylphenidate CD (Metadate CD^®) compared with placebo. A 3-week trial of Metadate CD^® compared with placebo enrolled 314 children out of 507 screened.⁷¹ Twenty-four percent of those excluded at screening were because they responded to placebo during a 1-week washout period. This biases the study population towards the Metadate CD^® arm, reducing the applicability of the results. The mean change in the primary outcome measure, the teachers' Clinical Global Impression Scale ratings combined in the morning and afternoon, were significantly lower (better) in the Metadate CD^® group. Secondary measures also favored Metadate CD^®.

Immediate-release formulations: Efficacy outcomes

Dextroamphetamine compared with methylphenidate. We included 9 fair-quality studies (reported in 11 publications) of immediate-release dextroamphetamine compared with immediate-release methylphenidate.^72-82 Two poor-quality studies and 1 poor-quality subgroup analysis were found.^83-85 All 9 fair-quality studies were randomized, blinded crossover trials. Table 6 summarizes the study characteristics.

Table 6

Immediate-release dextroamphetamine compared with immediate-release methylphenidate study characteristics.

The 2 largest studies,^{73, 74} which used clear criteria for diagnosis, enrolled children with ADHD in order to test the hypothesis that some adverse events associated with stimulants are actually characteristics of ADHD and would be improved by drug treatment in 1 study,⁷³ and to test the differences between child and parent assessment of therapy in the other.⁷⁴ Neither study provides details on the efficacy results, other than summary statements that there were no differences between the 2 drugs based on children's self-assessment⁷⁴ and based on parent and teacher ratings.⁷³ These 2 studies had similar populations, primarily children with the Mixed subtype (82%), however comorbidities and ethnicity were not reported.

Of the 7 small studies (N=12 to 48), only 1 found a difference between the drugs.⁸¹ This study assessed attention to task and deviant behavior in the usual classroom settings using a modified version of the Werry-Quay Direct Observational System.⁸¹ The text of the paper reported that in a post-hoc analysis, immediate-release dextroamphetamine was the most effective drug in instances where a positive effect was seen. Because this study did not use a standardized tool for diagnosis, and ADHD subtypes, comorbidities, or ethnicity were not reported, it must be assumed that significant heterogeneity in the population may have lead to the discordant results.

Response rates

Very few studies attempted to make a comparison of the rate of response (defined a priori) between 2 drugs. Table 7 shows the studies that did. Overall, no differences in response rates, as defined below, were found between the comparisons of methylphenidate OROS, immediate-release dextroamphetamine, mixed amphetamine salts or clonidine to immediate-release methylphenidate. Additionally, the majority of these response rates were lower than those reported and quoted from placebo-controlled trials (rates of approximately 75%).

Table 7

Comparison of response rates to immediate-release methylphenidate.

Immediate-release formulations: Effectiveness outcomes

We found extremely limited information on effectiveness outcomes from the clinical trials. Therefore, we included observational studies of ≥6 month's duration that reported effectiveness outcomes (Evidence Tables 13 and 14).

Immediate-release methylphenidate compared with methylphenidate OROS (Concerta^®). Integrated Health Care Information Services managed care claims data (described above) suggest that methylphenidate OROS was associated with fewer outpatient visits/hospitalization for accidents/injury than immediate-release methylphenidate over a 12-month follow-up period (odds ratio, 0.58; 95% CI, 0.353 to 0.945).⁵⁶ The study population (N=1,775) was 75% male, with a mean age of 9.7 years; however no other information regarding ADHD subtypes, comorbidities, or race/ethnicity were provided. In a second study, reported in 2 publications,⁵⁷ that also used data from the Integrated Health Care Information Services database to derive a larger sample (N=5,939) of somewhat older children (mean age of 15 years) who were also mostly male (77%), findings also suggest that methylphenidate OROS was associated with a lower probability of an emergency room visit (odds ratio, 0.79; 95% CI, 0.60 to 0.95)⁵⁷ and a lower probability of being hospitalized (odds ratio, 0.67; 95% CI, 0.45 to 0.99) over a 12-month period.⁵⁸ This study also found that age, prior number of diagnoses, and drug or alcohol abuse were statistically significantly associated with the probability of being hospitalized⁵⁸ and that geographic region, total number of diagnoses, presence of drug or alcohol abuse, or accident or injury were statistically significantly associated with the probability of an emergency room visit and the number of visits.⁵⁷ However, the study also found that those taking immediate-release methylphenidate were statistically significantly younger (14 years compared with 17 years old), had more total diagnoses, and geographic differences in the proportions of patients taking methylphenidate OROS compared with immediate-release methylphenidate were present.

Immediate-release methylphenidate. In a 4-year follow-up study of 62 children treated with methylphenidate, the effect of duration of treatment on academic performance was assessed.⁸⁶ The duration of treatment was divided into <6 months, 6 months to 2 years, 2 to 3 years, 3 to 4 years, and those currently taking stimulants at follow-up. No differences were found between the groups on academic achievement as measured by teachers, the proportion repeating grades, in special education classes, or being tutored. Although the proportion of children repeating grades was lowest in the group continuing to take methylphenidate (8% compared with 46%, 50%, 36%, 31%), this difference was not statistically significant, possibly because of the small numbers of boys per group (n=10 to 14). Due to methodological limitations, this study provides no comparative information.

Adherence rates as proxy measures of duration of effectiveness and caregiver satisfaction were reported for 307 Chinese children with ADHD taking immediate-release methylphenidate that were followed for 6 months of treatment.⁸⁷ Parents of 100 children (32.6%) were unsatisfied with their children's adherence to immediate-release methylphenidate and cited the following reasons for missing doses: forgetting to take immediate-release methylphenidate at school (72.9%), the medication having no effect (20%), forgetting to bring immediate-release methylphenidate to school (19.1%), refusing to take immediate-release methylphenidate (12.7%), bitterness (11.4%), side effect (11.4%), and teacher's objection (7.7%). Compared to families with children demonstrating good adherence, poor adherence was associated with increased risk of impairments in maternal psychological status and perceived family support.

Stimulants. In a birth cohort study of 5713 children born in Rochester, Minnesota during the years of 1976 to 1982, 370 children were diagnosed with ADHD. Two hundred ninety-five were treated with a stimulant and 84 were not.⁸⁸ Of those exposed to a stimulant, 66% took methylphenidate and 30% took dextroamphetamine (assumed to be immediate-release formulations). Median age of initiation of treatment was 10 years, median duration of treatment was 34 months, and median dose was 21 mg daily methylphenidate or methylphenidate equivalents. In addition to the 84 children diagnosed with ADHD but not receiving a stimulant at any time, the study also identified a control group from the birth cohort. Using a Poisson regression analysis, exposure at any time during follow-up was associated with lower rates of absenteeism (P=0.012) and duration of exposure was also significantly associated with lower absenteeism rates (P=0.041). Other factors were also found statistically significantly associated with number of days absent: comorbid conditions (P=0.006), type of educational interventions (P<0.001), and maternal education at birth (P=0.005). Reading scores were similar between groups, although among those treated with a stimulant there was a “mild correlation” between the mean dose of stimulant and final reading score recorded (r=0.15; P=0.012). Children who were exposed to a stimulant were 1.8 times (95% CI, 1.01 to 3.2) less likely to be retained a grade at any time; based on Kaplan-Meier analysis 66 children were retained a grade level. Dropout rate (based on 69 of 301 cases available for analysis) was significantly associated with maternal education at birth, comorbid conditions, and type of educational intervention, but not stimulant exposure, duration, or dose. While this study had some methodological advantages over other studies, the main limitation was the number of children included, particularly in the nonmedicated group, such that these findings should be interpreted cautiously.

Maintenance of short-term symptom response effects

Methylphenidate or immediate-release dextroamphetamine compared with placebo or non-drug therapy. All of the trials reported above were very short-term trials (range 1 to 9 weeks). Because of this serious limitation, the evidence does not provide information on the long-term benefits of these drugs in treating ADHD. To provide further evidence on duration of effect and longer-term outcomes, placebo- or non-drug therapy controlled trials of ADHD drugs with duration ≥6 months are reported here (Evidence Tables 7 and 8). We found 3 placebo-controlled trials of at least 6 months duration, 1 with immediate-release dextroamphetamine and 2 with immediate-release methylphenidate,^89-91 and 3 trials that randomized children to stimulant medication or nondrug therapy for 12 to 14 months.^92-94 Many of these studied indicated dissipation of medication effects over time, with unmedicated control groups having similar longer-term outcomes, particularly with follow-up of 2 years or greater.

Of these, the largest (N=579) and longest duration of follow-up is the Multimodal Treatment Study of Children with Attention Deficit/Hyperactivity Disorder (MTA). The MTA was a relatively large study funded by the NIMH assessing medication management, behavioral treatments, standard community care, and combined medication management and behavioral treatments over a 14-month period.⁹² Following the 14-month trial the groups had follow-up at 2, 3, and 8 years post randomization.^{92, 95-97} Medication management could involve any stimulant medication, but started with methylphenidate titration. At study end, 73% of those in 1 of the medication management groups were on methylphenidate and 10% on immediate-release dextroamphetamine, with small numbers of patients taking no medication, pemoline, imipramine, bupropion, or haloperidol, and 6% refusing to be in the medication arm assigned. All participants met DSM-IV criteria for ADHD combined type, had a mean age of 8.5 years, and 80% were males. The sample population was ethnically diverse, with White (61%), African American (20%), and Hispanic (8%) representation. Comorbidities included anxiety disorder (33.5%), conduct disorder (14.3%), oppositional-defiant disorder (39.9%), affective disorder (3.8%), tic disorder (10.9%), mania/hypomania (2.2%), and other (e.g., bulimia, enuresis) (0.2%). This study was a pragmatic trial in that the treatments were given openly (after blinded titration in the 2 drug treatment arms), and participants could refuse the assigned arm or add or change treatments. In the community care arm, for example, 68% were taking ADHD medications although the mean dose and number of daily doses of methylphenidate was lower in the community care arm than the medication arms. However, the outcome measures were not effectiveness outcomes, so the trial must still be viewed as an efficacy trial.

After 14 months, medication management alone resulted in better scores compared with behavioral therapy for the symptoms of inattention (rated by both parents and teachers) and hyperactive-impulsive symptoms (parent ratings). Medication alone resulted in better scores on all ADHD symptoms than community care, except as measured by a classroom observer. Aggression-oppositional defiant disorder symptoms scores were better with medication alone compared with community care in teacher ratings only. Combined therapy (medication and behavioral therapy) was not different to medication alone on any scale. The effect of medication management was maintained over the 14 month period.

Families were contacted 10 months after the end of the 14-month study (2 years post randomization) to assess longer-term persistence of treatment effects.⁹⁵ A total of 540 (93%) of the originally randomized 579 participated and 10 months after study end, 72% in the medication management alone group, 70% in the combined therapy group, 38% in the behavioral therapy group, and 62% in the community care group were taking medication for ADHD. At 2 years, medication alone still resulted in better scores on ADHD and oppositional defiant disorder symptoms than behavioral therapy and community care. Despite this, analyses of combined outcomes from the medication management alone and combined therapy groups compared with those of the behavioral therapy and community care groups suggest a reduction in the magnitude of benefit by half from the 14-month to 24-month time points; effect size changes for ADHD symptoms were 0.60 compared with 0.30 and oppositional defiant disorder symptoms were 0.39 compared with 0.21. At 3 years of follow-up, 485 children participated (84%) and the proportions taking medication had changed. There was a decrease from 91% to 71% in the medication only/combined therapy group, an increase from 14% to 45% in the behavioral therapy group; and about constant (60% to 62%) in the community care group.⁹⁶ Along with these changes, the difference between groups in outcome measures was no longer statistically significant although all groups had improved compared with baseline scores on all measures. Further analyses indicated a benefit of regular medication use during the 14 month and 24 month periods, but not at 36 months. At 6 and 8 years, follow-up was possible in 78% and 75%, respectively.⁹⁶ Regular medication use was reported in 42% at 6 years and in 31% at 8 years, with no significant differences among the groups. Among children taking a stimulant at 3 and 8 years follow-up, mean dose had increased from a mean of 31 mg daily to 43 mg daily. Small numbers of children were taking a nonstimulant. Again, no differences were found between groups in efficacy measures. This follow-up included questions about other outcomes, including police contacts and arrests; academic performance on reading and math tests; grade point average; use of school services; and grade retention, but no differences among groups were found.

The other smaller trials of immediate-release methylphenidate, compared with placebo^89-91 or other non-drug interventions,^92-94 reported a dissipation of effect at earlier time points, 9 months to 2 years. Although some of these studies do not report mean doses, of those that do, the doses used in the MTA study were higher. Two studies were poor quality due to serious flaws that represent significant potential for bias, primarily due to no details on the subject's characteristics at baseline and no details on those who discontinued the study.^{89, 98}

Remission rates: Immediate-release methylphenidate

Three studies assessed the effects of withdrawing immediate-release methylphenidate after periods of treatment.^99-101 Two of these were poor quality,^{99, 100} but the third study¹⁰¹ included a group of 21 boys who had been treated with methylphenidate for a mean of 1.75 years and randomized to 3 weeks of placebo or methylphenidate. Using the Conners' Teacher Rating Scale, this study found that on the Subscale items of hyperactivity and defiance the scores during the placebo period were significantly worse than during the methylphenidate period. No baseline assessments were presented, and the analyses are based on scores at week 3 of each condition only so there is no information about the effectiveness of their pre-existing methylphenidate regimen at baseline. In addition, the effect of order of drug/placebo was not analyzed in this crossover study, so the results must be interpreted with caution.

Atomoxetine

Atomoxetine compared with methylphenidate. While 4 studies have included both atomoxetine and immediate-release methylphenidate, only 2 made relevant comparisons for assessing comparative efficacy.^{128, 129} In a fair-quality, 8-week, noninferiority trial (N=330), atomoxetine was found noninferior to immediate-release methylphenidate based on ADHD rating scale response rates (>40% reduction in score; atomoxetine, 77%; immediate-release methylphenidate, 82%; P=0.4, assuming a margin [delta] of 18%).¹²⁹ The mean final doses of drug were somewhat imbalanced, with 44 mg daily of atomoxetine and 18 mg daily for immediate-release methylphenidate. Differences were not found between groups using other measures or through logistic regression controlling for multiple factors. Another study comparing atomoxetine and immediate-release methylphenidate found no differences between the drugs based on changes in the ADHD rating scale, the Conners' Parent Rating Scale Revised hyperactivity item, and the Clinical Global Impression-Severity Scale.¹²⁸ Concerns over the study quality indicating potential bias suggest caution in interpreting these findings (see Evidence Table 4).

A second study comparing immediate-release methylphenidate and atomoxetine primarily assessed the impact of each drug on sleep, using a crossover design and sleep labs.¹³⁰ This small study (N=75) evaluated sleep onset (latency) using actigraphy, a device worn on the wrist to measure activity over 7 weeks. The mean dose of immediate-release methylphenidate was 42.29 mg daily, and of atomoxetine was 58.27 mg daily. Only 50 of 85 patients (59%) randomized were included in the analysis, mostly due to inadequacy of actinography data, a number that does not reach the stated 60 needed to adequately power this analysis. Additionally, 21% of those screened (22 of 107) were excluded for a variety of reasons relating largely to not complying with a pre-specified “light-out” time consistently. The primary outcome was the comparison of the mean change in sleep-onset latency from baseline to endpoint. At baseline, 43.5% were not taking stimulants. Both groups experienced an increase in time to fall asleep, but the immediate-release methylphenidate group had a significantly longer increase (39.24 minutes) compared with atomoxetine (12.06 minutes). A similar decrease in overall sleep time was also seen. Differences were not found between the drugs in ratings of ADHD symptoms. Results of planned ANOVA analysis of sequence were not reported, so the impact of order of randomization cannot be assessed here but may be important. The study involved funding, data analysis, and authorship by the maker of atomoxetine. Because of the above concerns, we have rated this study poor quality.

Atomoxetine compared with methylphenidate OROS. In a 6-week fair-quality noninferiority trial, atomoxetine was not found noninferior to methylphenidate OROS.¹³¹ Using response (40% or more reduction of the ADHD–RS) as the primary outcome, and a margin of 15%, methylphenidate OROS was found superior to atomoxetine with an overall 56% response rate with methylphenidate OROS compared with 45% with atomoxetine (number needed to treat, 9; P=0.02). Analysis of the subgroup with prior stimulant exposure (n=310) found again a statistically significantly higher rate of response with methylphenidate OROS (51%) compared with atomoxetine (37%) (number needed to treat, 8; P=0.03). In this subgroup, atomoxetine was not found different than placebo. However, in the smaller subgroup without prior stimulant exposure, (n=191) the 2 drugs were not found to be statistically significantly different in response rates (57% atomoxetine compared with 64% methylphenidate OROS). Secondary outcome measures, such as the mean change in ADHD rating scale total and subscale scores, resulted in similar findings. This study used over-encapsulation of methylphenidate OROS. The authors reported that dissolution studies indicated no alteration in drug release but no data are reported. Also, atomoxetine was administered in a divided dose rather than given once daily.

The Formal Observation of Concerta^® compared with Strattera^® (FOCUS) trial compared open-label methylphenidate OROS and atomoxetine for 3 weeks in 1323 children with ADHD.¹³² The main findings from the FOCUS trial are summarized in Evidence Table 3, but will not be discussed here due to concerns about study quality. The FOCUS trial was rated poor quality based on a combination of flaws including undescribed methods of randomization and allocation concealment, significant between-groups baseline differences in ADHD severity, and lack of information about attrition and number of patients included in analyses (Evidence Table 4).

Atomoxetine compared with mixed amphetamine salts XR (Adderall SR^®). The extended-release form of mixed amphetamine salts (Adderall SR^®) 10-30 mg was superior to atomoxetine 0.5-1.2 mg/kg daily on most efficacy outcomes after 3 weeks in a fair-quality trial of 215 children (mean age, 8.7 years).¹³³ This trial, also known as Strattera^®/Adderall XR^® Randomized Trial, was conducted in a simulated classroom setting which involved 12 hours of observation daily. Participants were mostly male (71.9%) who were diagnosed with ADHD of either the hyperactive/impulsive or combined subtypes. Mixed amphetamine salts XR was associated with significantly greater reductions in the mean SKAMP deportment scale scores, which was prespecified as the primary outcome (−0.56 compared with −0.13; P<0.0001). Mixed amphetamine salts XR was also associated with superior outcomes on multiple secondary outcome measures including mean change in SKAMP Attention scale scores, proportions of SKAMP scale “responders” (≥ 25% improvement on deportment and/or attention scales), and numbers of math problems attempted and/or completed correctly.

Atomoxetine compared with standard therapy. A British study of atomoxetine compared with standard treatment assessed the child's function and health status using the final score on the Child Health and Illness Profile – Child Edition as the primary outcome measure.¹³⁴ The total score of the tool was stated to not have previously been used, but to have been validated by the owner (Riley and colleagues). This research was cited only as “submitted for publication,” and a recent search did not uncover such a publication, so it is considered an unvalidated tool here. A total of 201 patients were randomized to 10 weeks of treatment with either atomoxetine or whatever treatment (including no treatment) prescribed by the investigator or the treating physician. This was an open-label study, with parent making the assessments. This study is poor quality, with no description of randomization and allocation concealment procedures, and some imbalances between the groups at baseline (Inattentive ADHD subtype 11.5% compared with 3.1%, previous exposure to stimulants 59.6% compared with 70% in atomoxetine and control groups, respectively). Additional concerns were that the higher discontinuation rate in the atomoxetine group was not taken into account by the modified intent to treat analysis described (it appears only 75% of atomoxetine group is included in the analysis, compared with 94% of control group), the standard treatment group was described as having their treatment determined by unblinded investigators, and the primary author being an employee of the manufacturer of atomoxetine.

Atomoxetine compared with placebo. Six placebo-controlled studies of atomoxetine in children and adolescents with ADHD found atomoxetine to be superior based on ADHD rating scale as the primary outcome measure and various scales as secondary measures.^135-140 Results of 2 of the 6 trials were described as identically-designed and were reported in 1 publication.¹³⁷ The mean change on ADHD rating scale in these 6 to 9 week studies ranged from −12.8 to −16.7 with atomoxetine compared with −5.0 to −7.0 for placebo. A study of once daily dosing reported response rates (defined as ≥25% reduction in ADHD rating scale score) in the atomoxetine group of 59.5% compared with 31.3% in the placebo group (P<0.001).¹³⁹ Remission rates (defined as an endpoint Clinical Global Impression-Severity Scale score of 1 or 2) were 28.6% and 9.6%, respectively (P=0.003). All studies were funded and coauthored by representatives of the manufacturer of atomoxetine. All used the DSM-IV criteria, however the proportions of ADHD subtypes varied, for example 52% to 79% of enrolled children had the Mixed subtype. More concerning is the variation in the proportions of children with each subtype per assigned group. Proportions of children with comorbidities also varied across the studies (e.g. 18% to 45% with oppositional defiant disorder). Results of a subgroup analysis from 2 identically-designed placebo-controlled trials¹³⁷ suggested that atomoxetine was associated with significantly greater reductions in ADHD rating scale total scores than placebo (−17.0 compared with −7.5; P<0.001) in 98 of the original 291 children with comorbid ADHD and oppositional defiant disorder.¹⁴¹ No subgroup analyzed based on ADHD subtypes or other comorbidities were reported. Based on what appeared to be post-hoc analysis of secondary outcome measures of 1 of these trials,¹⁴⁰ no statistically significant difference between atomoxetine and placebo was seen in academic performance (based on the Academic Performance Rating Scale) or quality of life (based on the Children's Health Questionnaire psychological summary score) after 7 weeks.¹⁴²

In a good-quality systematic review, these 6 trials and 3 additional trials with placebo and active arms were combined in a meta-analysis that indicated atomoxetine was superior to placebo in improving ADHD rating scale total score (standardized mean difference, −0.638; 95% CI, −0.76 to −0.516), as well as subscale scores on inattentive symptoms, hyperactivity/impulsive symptoms, and the Conners' scales with teacher and parent ratings.¹⁴³ Meta-regression identified study duration and number of study sites, male sex, ADHD hyperactive/impulsive subtype, oppositional defiant disorder, baseline ADHD rating scale total score, inattention score, and hyperactivity/impulsivity score to be negatively associated with response. After adjusting for these confounders, atomoxetine remained superior over placebo. Six adverse events were found to occur significantly more often with atomoxetine (numbers needed to harm; P value): decrease in appetite (8; P<0.05), somnolence (19; P<0.05), abdominal pain (22; P=0.02), vomiting (30; P=0 .02), dyspepsia (49; P<0.01), dizziness (53; P=0.01), fatigue (62; P=0.01), infection (72; P=0.02), and pruritus (120; P=0.04). Risk of adverse events was found to be negatively associated with mean age, ADHD inattentive subtype, baseline ADHD rating scale score, and hyperactivity/impulsivity score. Meta-regression identified high ADHD rating scale total and hyperactivity/impulsivity scores at baseline to be significantly associated with adverse events (P<0.01).

Based on the 6 placebo-controlled trials above, with data apparently provided by the manufacturer, meta-analysis was performed to assess differences in response between younger (ages 6-7) and older (ages 8-12) children. Atomoxetine was found statistically significantly superior to placebo on the ADHD-RS and Conners' scales, in both age groups, although the difference between atomoxetine and placebo was smaller in the older age group compared with the smaller age group.¹⁴⁴ This study also found that abdominal pain, decreased appetite, vomiting, and somnolence occurred significantly more often with atomoxetine than placebo in younger children, and decreased appetite, somnolence, irritability, and fatigue among older children. There was a significant treatment by age effect in abdominal pain (P=0.04), vomiting (P=0.053), pyrexia (P=0.058), and cough (P=0.007). Statistically significant but small increases in pulse were seen in both younger and older children, and older children experienced increases in both systolic and diastolic blood pressure. In these short-term studies, statistically significant weight decrease was seen in both age groups (−0.5 and −0.6 kg).

Atomoxetine was associated with less rapid times to relapse than placebo under double-blind conditions (218 days compared with 146 days; P<0.001) in a randomized subgroup of 416 children (out of 603) that were classified as “responders” following an initial 12-week, open-label period of treatment with atomoxetine.¹³⁹ The primary outcome measure was the number of days to relapse and relapse was defined as return to 90% of baseline ADHD rating scale score and Clinical Global Impression-Severity Scale score increase of at least 2 points. Similarly, fewer patients on atomoxetine relapsed than on placebo (22% compared with 38%; P<0.002). As a continuation of that study, subjects initially randomized to atomoxetine were rerandomized to an additional 6 months of either atomoxetine (n=81) or placebo (n=82), with mean time to relapse being 160 days for atomoxetine and 130.8 for placebo, P<0.008. Relapse rates were 2.5% for atomoxetine and 2% for placebo and the relative risk for relapse during placebo treatment was 5.6 (95% CI, 1.2 to 25.6).¹⁴⁵

Atomoxetine: Effectiveness outcomes. A few noncomparative observational studies evaluated duration of effectiveness for atomoxetine.^{146, 147} In 1 study, 229 children who had a ≥40% reduction in ADHD rating scale total score after a 7- to 9-week trial of atomoxetine (51% of original sample) were randomly assigned to continue treatment for 8 months at the same or lower dosages.¹⁴⁶ In the other study, stability of treatment response over time was examined in 312 children who had completed 24 months of open treatment with atomoxetine (34% of original sample).¹⁴⁷ Both studies were consistent in finding that improvements in ADHD symptoms and in aspects of health-related quality of life were maintained during longer-term treatment periods, even with reduced dosages of atomoxetine. Although encouraging, findings from these studies must be interpreted with caution, mainly due to the extremely high attrition rates.

In a pooled analysis of data from 714 children who received atomoxetine for at least 3 years in open-label studies, 1.7% of children and 2% of adolescents discontinued due to adverse events indicating high rates of persistence in both age groups.¹⁴⁸

Clonidine

Immediate-release clonidine compared with methylphenidate. Four parallel group randomized controlled trials of immediate-release clonidine compared with immediate-release methylphenidate were found.^149-152 Two small randomized controlled trials of immediate-release clonidine compared with immediate-release methylphenidate measured outcomes using scales or tests that have either been shown to have low validity (e.g., Home/School Situations Questionnaire [HSQ and SSQ] and Gordon Diagnostic System), or which validity could not be verified (e.g., Disruptive Behavior Scale and Grooved Pegboard) and were rated poor quality.^{150, 152} ClinicalTrials.gov lists 1 additional study of immediate-release clonidine and immediate-release methylphenidate that is completed but not yet published.

The remaining small trials (N = 122 and 132) reported no statistically significant differences in Conners' Abbreviated Symptom Questionnaire Teacher scale from baseline to endpoint (16 weeks) between immediate-release methylphenidate and clonidine. However, this comparison was not the primary aim of either study, and the studies have conflicting findings outside of this comparison. While the studies had similar 2 × 2 factorial designs, 1 enrolled children with both ADHD and Tourette's disorder and evaluated the effect of drugs on tics, while the other enrolled children without Tourette's disorder. The study of children with Tourette's disorder and ADHD enrolled 136 children (mean age 10.2 years) and assigned them to immediate-release methylphenidate, immediate-release clonidine, both drugs, or placebo for 8 weeks.¹⁵¹ Mean doses at the end of study were 0.25 mg clonidine and 26 mg immediate-release methylphenidate daily. All analyses made comparisons of each drug group to placebo, although it is stated that there was no difference between the immediate-release methylphenidate and immediate-release clonidine groups on the primary outcome measure of the Conners' Abbreviated Symptom Questionnaire Teacher scale (Table 8). Immediate-release clonidine was significantly better than placebo on more Tourette's outcome measures than immediate-release methylphenidate, and immediate-release methylphenidate was significantly better than placebo on more ADHD outcome measures than clonidine.

Table 8

Summary of differences in results of ADHD/Tourette's disorder study.

The second trial enrolled 122 children with ADHD (mean age 8 years) and also randomized them to immediate-release methylphenidate, immediate-release clonidine, both drugs, or placebo for 8 weeks.¹⁴⁹ Mean doses at the end of study were 0.24 mg clonidine and 30 mg immediate-release methylphenidate daily. Based on the primary outcome measure of mean change in the Conners' Abbreviated Symptom Questionnaire Teacher scale no difference was found between groups receiving clonidine and those receiving methylphenidate although the change was greater with methylphenidate (−3.35 vs. −5.07; P=0.16). Similar results were found with secondary outcome measures for the direct comparison of the drugs.

The difference in results of these 2 studies is shown in Table 9 below. Using the 2× 2 factorial designs, the analysis of treatment effect in groups receiving methylphenidate (alone or with clonidine) was statistically significant compared with groups not receiving methylphenidate (clonidine or placebo groups) in both studies. However, the treatment effect of clonidine (alone or with methylphenidate) was statistically significant only in the Tourette's syndrome study, and not in the Palumbo 2008 trial. The parent ratings in the Palumbo trial conflict with the results based on teacher ratings – showing methylphenidate to have no effect and clonidine to have an effect. In the Tourette's syndrome study, the parent and teacher ratings results were similar, finding both drugs beneficial. The reasons for the differences were not clear, but may have been related to differential attrition rates across groups in the Palumbo study, and while the both studies allowed psychological interventions, important variation across the studies may have occurred. Taken together, and considering all efficacy outcomes, we were not able to currently identify a difference in effect between immediate-release clonidine and immediate-release methylphenidate, but it may be that clonidine has a lesser effect.

Table 9

Change in Conners' Abbreviated Symptom Questionnaire Teacher scale.

Clonidine compared with placebo. Two placebo-controlled trials were not helpful in examining the comparative effectiveness of immediate-release clonidine. A small randomized placebo-controlled crossover trial was rated poor quality.^{153, 154} The other¹⁵⁵ did not report clear eligibility criteria but it appears that the children had to have a diagnosis of ADHD and Tourette's disorder based on DSM-IIIR. Outcomes were assessed using primarily Child Behavior Checklists for parents and teachers, and linear analogue scales of parent assessments of hyperactivity and tics, as compared with any time in the past. Based on the linear analogue assessments, clonidine was not significantly different to placebo. The assessment of tics, based on 4 scales, did not show a difference between placebo and clonidine. Multiple subgroup analyses of the checklists resulted in clonidine being superior to placebo on some subscale items. A poor quality systematic review included 11 studies of varying design. These studies did not meet inclusion criteria for this review, and were inappropriately pooled regardless of their varying study designs. ¹⁵⁶

Clonidine ER compared with placebo. Two studies of extended-release clonidine have been published to date, with additional completed studies not published to date according to ClinicalTrials.gov.^{157, 158} The trials enrolled children with the hyperactive or combined subtype of ADHD with less than adequate response (ADHD-RS-IV scale score ≥ 26 out of 54) to a stable dose of stimulant, and used the ADHD-RS-IV scale assessed by investigators as the primary outcome measure. Parent measurements on the Conners' Parent Rating Scale and the Parent Global Assessment were secondary outcomes.

In a fair-quality trial of 198 children randomized to flexible dose clonidine ER (0.1 to 0.4 mg daily) or placebo in addition to their stimulant, after 5 weeks children on clonidine ER had more improvement in their ADHD-RS-IV scores than those on placebo (−15.7 compared with −11.5; P=0.009), although the absolute difference was very small.¹⁵⁸ Parent assessments found the group with clonidine XR to have better improvement in the Conners' Parent Rating scale (− 40.2 and −27.1; P=0.017) and better final Global Assessment scores (2.7 and 3.4; P=0.001) compared with the placebo group. Subgroup analysis by specific stimulants failed to show statistical significant benefits, most likely due to inadequate sample sizes (Type II error).

In a fair-quality fixed-dose trial of 236 children, monotherapy clonidine ER 0.2 or 0.4 mg, given in 2 divided doses, resulted in greater improvement in ADHD-RS-IV scores at 5 weeks compared with placebo (−15.6, −16.5, and −7.5 respectively; P<0.0001 for each compared with placebo) with little difference between doses.¹⁵⁷ Week by week analysis indicated that the difference between drug (ether dose) and placebo started at week 2. Parent assessment using the Conners' rating scale and the Parent Global Assessment also indicated a significant difference between changes in score at either dose compared with placebo, but actual scores were not reported.

Guanfacine

Immediate-release guanfacine compared with placebo. A small study of 24 children with ADHD, all of the mixed type, and a tic disorder studied the effects of immediate-release guanfacine compared with placebo for 8 weeks.¹⁵⁹ Slightly more than half of enrolled children had Tourette's disorder (58.8%), and 35% had chronic motor tic disorder. Teachers and investigators rated immediate-release guanfacine superior to placebo, while parent ratings did not. Teacher ratings resulted in a 37% decrease on the ADHD Rating Scale at 8 weeks, compared with an 8% drop with placebo group (P<0.001) while parent ratings showed a 27% improvement with immediate-release guanfacine and 21% with placebo.

Guanfacine XR compared with placebo. Four fair-quality placebo-controlled trials 8-9 weeks in duration of 1341 children have been published with guanfacine XR. Two were monotherapy dose-ranging studies of 1 to 4 mg daily, included in the initial US Food and Drug Administration documents reviewed, ^160-162 and another monotherapy study included only children aged 6-12 years with comorbid oppositional symptoms using a flexible dose of 1 to 4 mg daily.¹⁶³ The fourth study assessing guanfacine XR as adjunctive therapy to stimulants was subsequently submitted to the US Food and Drug Administration.¹⁶⁴ Baseline mean ADHD-RS-IV scores (test results in score ranging from 0 to 54) were 37, 40, 37, and 42, respectively.

In the dose-ranging studies, the placebo-adjusted change from baseline to endpoint in the ADHD-RS-IV score (primary outcome) was statistically significantly greater than placebo for all doses, although the absolute difference in score was not large (least squares mean difference −5.41 to −9.99). In both studies the largest difference in score change was with the 4 mg daily dose (−9.99 and −7.88). ^160-162 For 2, 3, and 4 mg daily doses the difference compared with placebo was statistically significant starting at week 2. Revised Conners' Parent and Conners' Teacher Rating Scales also showed guanfacine XR superior to placebo in mean change from baseline to endpoint scores. Assessment of the duration of effect using these measures throughout the day showed all doses to have effect through 8 hours, but variable effects by dose and study at 12, 14, and 24 hours. Post-hoc analysis of weight-based dosing and outcome indicated a greater response with 0.09- 0.12 mg/kg and 0.13-0.17 mg/kg compared with lower doses, but differentiating between these dosing ranges is not possible from these data.

A study of 217 children aged 6-12 years with comorbid ADHD and oppositional symptoms using flexible dosing (1-4 mg daily) over 8 weeks found that the mean least squares mean change on the ADHD-RS-IV scale was −23.8 in the drug group and −11.5 in the placebo group (P<0.001 from a baseline of 42 in both groups).¹⁶³ The subscale scores on the CPRS-RS-L oppositional defiant subscale also improved more with guanfacine XR (−10.9 compared with − 6.8; P<0.001) and the change in the 2 scores was found to be highly correlated (r=0.74). Slightly more patients were taking 3 mg daily, and only a few took 1 mg daily.

As adjunctive therapy to stimulant therapy with suboptimal response (ADHD-RS-IV ≥ 24 and CGI-S ≥ 3 after at least 4 weeks treatment), guanfacine XR (given at night or in the morning) was found superior to placebo on the ADHD-RS-IV scale, least squares mean difference in Total score of −4.5 (P=0.02) for morning dosing and −5.3 (P=0.001) for evening dosing compared with placebo.¹⁶⁴ Following dose-optimization, a mean dose of 0.09 mg/kg was reached in both drug groups (median doses were 3 mg daily in the evening group and 3 mg daily in the morning group). As a secondary outcome measure, response (defined as reduction of ADHD-RS-IV Total score of ≥ 25% from baseline), the evening dosing of guanfacine XR was superior to placebo (83.1% vs. 69.7%; P=0.007), while the morning dosing was not statistically different to placebo (P=0.062). Multiple subgroup and secondary analyses were conducted but are not presented here because they are comparisons to placebo.

Methylphenidate OROS compared with immediate-release methylphenidate

In a fair-quality, single-blind trial of 53 adults, there was no significant difference in maintenance of response (much or very much improved on the Clinical Global Impression Scale-I) 6 weeks after switching from stable treatment with immediate-release methylphenidate administered 3 times daily to methylphenidate OROS compared with remaining on immediate-release methylphenidate 3 times daily (63% compared with 58%; P=0.8).¹⁸⁶ The similarity in efficacy was maintained in spite of the significantly greater average number of missed doses over the 6-week study with the immediate-release formulation of methylphenidate 3 times daily (7.3 compared with 3.3; P=0.02). Mean dosage was not reported, but was not to exceed 1.3 mg/kg/day or 144 mg/day total. Results of this trial are primarily applicable to patients who have been receiving a stable dose of immediate-release methylphenidate for at least 4 weeks and who have achieved a good clinical response, have good tolerability, and are satisfied with their treatment experience.

Immediate-release guanfacine compared with immediate-release dextroamphetamine

In a very small, 2-week, fair-quality, crossover trial of 17 adults, there was no significant difference between immediate-release guanfacine and immediate-release dextroamphetamine in the DSM-IV ADHD Behavior Checklist for Adults mean total symptom score at endpoint.(23.3 points compared with 24.2 points).¹⁸⁷ The average dose was 1.10 mg for immediate-release guanfacine and 10.2 mg for immediate-release dextroamphetamine. However, these results should be interpreted with caution as these drugs were administered only once daily in the morning and the small sample size may have lacked adequate statistical power to detect a significant difference. Baseline total symptom scores were not reported, preventing calculation of mean change, and response rate was also not reported.

Modafinil compared with immediate-release dextroamphetamine

In a fair-quality, crossover trial of 22 adults, identical proportions responded to 2 weeks of treatment with modafinil 206.8 mg and immediate-release dextroamphetamine 21.8 mg (48% compared with 48%; P=NS).¹⁸⁸ Response was defined as a 30% or greater mean improvement in ADHD Rating Scale total scores. Patients in this trial were mostly male (59%) and had a mean age of 40.8 years.¹⁸⁸

Atomoxetine

We included 10 placebo-controlled trials of atomoxetine (N=2493).^189-197 Two trials focused on subgroups of patients with either marijuana dependency¹⁹⁷ or comorbid social anxiety¹⁹⁶ and will be discussed under Key Question 3. All of the remaining trials were rated fair quality and ranged in duration from 3 weeks to 6 months. Mean age ranged from 34 years to 41 years and proportions of male patients ranged from 44% to 64%.

Response. Response rate was only reported in 3 trials^{192, 193, 195} and was significantly greater for atomoxetine in 2 of them.^{192, 195} The 2 trials that found significantly greater response rates for atomoxetine differed in duration and definition of response. In the shorter-term trial, 21 adults (48% male, mean age of 34 years) were randomized to tomoxetine 40 mg twice daily or placebo for 3 weeks of treatment and response was defined as at least a 30% reduction in the ADHD Rating Scale score.¹⁹² Response rate was 52% for tomoxetine and 10% for placebo (P<0.01).¹⁹² The longer-term trial randomized 510 adults (48% male, mean age of 41 years) to 24 weeks of treatment with atomoxetine, at a mean dose of 90.3 mg given once daily, or placebo and response was measured using the Conners' Adult ADHD Rating Scale, Self-Report Screening Version .¹⁹⁵ Response rates were significantly greater with atomoxetine, both when defined as at least a 25% decrease from baseline (68% compared with 42%, P<0.001) and when defined as at least a 40% decrease from baseline (47% compared with 28%, P<0.001).

The trial that did not find a significant difference in response between atomoxetine 40-80 mg (mean dosage not reported) and placebo had the smallest sample size (N=16) and was comprised of somewhat younger patients (mean age of 22.4 years), more of which who were male (87%).¹⁹³ Response was defined as at least a 30% reduction in the ADHD Rating Scale score and no significant difference was found between atomoxetine and placebo (40% compared with 25%, P=not significant).

Other ADHD symptom outcomes. In the other trials,^189-191 ADHD symptoms were measured based on mean change from baseline on either the Conners' Adult ADHD Rating Scale, Self-Report Screening Version or the ADHD-RS. Improvements in self-ratings were significantly greater for atomoxetine in all 4 trials. Improvements based on investigator or other ratings were significantly for atomoxetine in 3 of 4 trials.^{190, 191}

Quality of life. We identified 1 placebo-controlled trial that examined the effects of atomoxetine on quality of life in adults.¹⁸⁹ A 6-month trial of atomoxetine compared with placebo (N=410; mean age, 36.5; 60% male; 82% Caucasian), dose flexible from 40 mg to 100 mg daily based on tolerability, found no difference in change from baseline between treatment groups in relationships, psychological health, productivity, and work productivity.

Driving performance. The majority of evidence from 3 placebo-controlled trials found that atomoxetine was not significantly superior to placebo in improving driving outcomes. One large trial and 2 smaller trials assessed simulator driving performance among subjects taking atomoxetine compared with placebo. A 24-week trial of 410 subjects (mean age 36.5; 60% male) of atomoxetine (dose flexible from 40 mg to 100 mg daily based on tolerability) compared with placebo¹⁸⁹ found no differences in self report of the Driving Behavior Survey between treatment groups. Driving behavior was rated as significantly more improved for the atomoxetine group compared with the placebo group in a subsample of 252 of 410 patients for which observer ratings were available (mean improvement 6.1 compared with 2.0; P=0.01). A smaller, 3-week trial of twenty subjects (mean age 36; 44% male) comparing atomoxetine (titrated up to 1.2 mg/kg for 3 weeks) to placebo¹⁹⁰ reported mixed results. Self-ratings, but not other-ratings (such as friends or spouse) or examiner-ratings, were significantly greater for atomoxetine on the Safe Driving Behavior Scale and on simulator driving performance. Finally, a small 3-week trial of young adults (ages 19-25) found that atomoxetine (titrated up to 80 mg daily) was not statistically different than placebo in mean total driving score.¹⁹³

Immediate-release dextroamphetamine

The percentage of participants rated as treatment responders was significantly greater for immediate-release dextroamphetamine in both of 2 fair-quality trials.^{198, 199} The first trial randomized 51 adults (53% male, mean age of 35.5 years) to 6 weeks of immediate-release dextroamphetamine (mean dose 23.8 mg) or placebo. The percentage of patients in the immediate-release dextroamphetamine group who were much improved or very much improved on the Global Improvement Scale was 58% (P<0.001) whereas in the placebo group, 90% of patients were rated as unchanged or minimally worse and no patients were very much improved.¹⁹⁹ In the second trial, 98 adults (64% male, mean age of 37.5 years) were randomly assigned to psychotherapy plus a maximum dosage of immediate-release dextroamphetamine 20 mg twice daily (average dosage not reported), paroxetine, both or placebo for 20 weeks. Response was defined as much or very much improved on the Clinical Global Impressions-Improvement Scale for ADHD and rates were 64% for immediate-release dextroamphetamine and 16% for placebo (P not reported).¹⁹⁸

Extended-release dexmethylphenidate

We included 1 fair-quality trial of extended-release dexmethylphenidate.²⁰⁰ Compared with placebo (34%), after 5 weeks, treatment response was significantly greater for extended-release dexmethylphenidate 20 mg (58%, P=0.017) and 40 mg (61%, P=0.007), but not with the 30 mg dosage (54%, P=0.054).²⁰⁰ This trial randomized 221 adults (57% male, mean age of 38.7 years) and treatment response was defined as a 30% or greater improvement on the DSM-IV ADHD-RS Total Score. The comparison between extended-release dexmethylphenidate and placebo was not influenced by gender, age or ADHD subtype. There was no significant difference between any dose of extended-release dexmethylphenidate and placebo in quality of life as measured by the patient-rated Quality of Life Enjoyment and Satisfaction Questionnaire.

Lisdexamfetamine

We included 1 fair-quality trial²⁰¹ and 1 poor-quality trial of lisdexamfetamine.^{202, 203} The more recent trial of lisdexamfetamine used a crossover design and was rated poor quality due to a failure to meet a combination of criteria which may be related in indicating the presence of bias, including incomplete outcome data, as well as the inadequate reporting of the allocation concealment method and comparability of baseline patient characteristics based on order of randomization.²⁰² In the fair-quality trial, post-hoc analysis found that, compared with placebo (34%), treatment response at endpoint (≥ 30% reduction in ADHD-RS Total Score, estimated from Figure 3) was significantly greater for lisdexamfetamine 30 mg (60%; P<0.001), 50 mg (64%, P<0.0001), and 70 mg (65%; P<0.0001).²⁰¹ This trial randomized 420 adults for 4 weeks of treatment. The mean age was 35.1 years, 54% were men, and 82% were White.

Mixed amphetamine salts immediate-release

We included 1 fair-quality, crossover trial that compared 3 weeks of treatment with mixed amphetamine salts immediate-release 54 mg to placebo in 30 adults (50% male, mean age of 38 years).²⁰⁴ Compared to placebo, significantly more patients taking mixed amphetamine salts immediate-release achieved clinical improvement, both when defined as more than a 30% reduction on the ADHD Rating Scale (70% compared with 7%, P<0.001) and when defined as much or very much improved on the Clinical Global Improvement Scale (67% compared with 4%, P<0.001).

Mixed amphetamine salts extended-release

We included 2 fair-quality trials of mixed amphetamine salts extended-release.^{193, 205} Compared with placebo, the proportion of adults with at least a 30% reduction in the ADHD Rating Scale total score was significantly greater with mixed amphetamine salts extended-release in 1 of 2 trials.¹⁹³ The first trial randomized 255 adults to 4 weeks of mixed amphetamine salts extended-release 20 mg, 40 mg, 60 mg, or placebo.²⁰⁵ Mean age ranged from 39 years to 40 years and proportion of males ranged from 48% to 68%. Proportion of responders was only analyzed for the subgroup of completers (72%) and reported as “substantially higher” for mixed amphetamine salts extended-release 20 mg (74%), 40 mg (80%) and 60 mg (82%) compared with placebo (61%), but the P values were not reported.

The second trial used a crossover design to compare 3 weeks each of mixed amphetamine salts extended-release 50 mg to placebo in 19 young adults (mean age of 22 years, 89% male).¹⁹³ It was primarily designed to assess simulated driving changes, but also reported that there was a significantly greater proportion of responders with mixed amphetamine salts extended-release compared with placebo (80% compared with 13%; P=0.0004). Improvement in driving ability was measured based on lowering of the numerical overall Driving Safety Score, which reflects the mean of z-scores from each of 8 simulator-derived variables including total citations, total collisions, time to collision, driving out-of-lane incidents, percentage of time above excessive speed threshold, number of times overcornering, number of times tailgating, and response to crash-likely events. Greater improvements in overall simulated driving performances were found for mixed amphetamine salts extended release than for placebo both at 7-hours post dose (−0.31 compared with +0.33; P=0.013) and at 12-hours post dose (−0.29 compared with +0.31; P=0.005).¹⁹³

Immediate-release methylphenidate

We included 17 placebo-controlled trials of immediate-release methylphenidate (N=593).^206-222 We rated 1 trial poor quality²¹⁶ and the rest were rated fair quality. The majority of trials enrolled small samples of adults (N<50), used crossover designs, and were very short-term (≤ 3 weeks). The longest trial involved 12 weeks of treatment, but focused only on a subgroup of men who were cocaine dependent.²¹⁵ Overall, in trials of methylphenidate immediate-release, mean age ranged from 22 years to 42 years and the proportion of males ranged from 40% to 100%. Along with the trial of cocaine dependent men, another trial focused only on a subgroup of patients with substance abuse disorders²⁰⁸ and results from both trials will be discussed in Key Question 3. Additionally, results of placebo-controlled trials relating to effects on pre-existing anxiety will be discussed in Key Question 3 as comorbidity.

Response. Only 3 fair-quality trials of average adults with ADHD reported response (N=98)^{212, 213, 217} and rates were significantly greater for immediate-release methylphenidate compared with placebo in 2 of them.^{212, 217} In all 3 trials, mean dosages were right around 0.9 mg/kg. In the first trial of 23 adults (43% male), response was defined as at least a 30% reduction in the ADHD Rating Scale plus a Clinical Global Impression score of 2 or less (much or very much improved) and 3-week rates were 78% for immediate-release methylphenidate and 4% for placebo (P<0.0001).²¹⁷ In the second trial of 30 adults, with a higher proportion of males (73%), response was defined as a Clinical Global Impression score of 2 or less, but 7-week rates were not significantly difference for immediate-release methylphenidate and placebo (50% compared with 27%).²¹³ Finally, in the third trial of 45 adults, response was defined as at least a 30% reduction in the ADHD Rating Scale and 3-week rates were 42% for immediate-release methylphenidate and 13% for placebo (P=0.011).²¹²

Driving. Immediate-release methylphenidate was also 1 of 3 drugs with data available regarding driving behavior. Driving performance was assessed in 3 small single-dose, placebo-controlled trials.^{206, 209, 220} A recent placebo-controlled crossover study of 18 ADHD patients (mean age 38, male 61%) performed on a primary highway during normal traffic assessed a single mean dose of 14.7 mg methylphenidate.²²⁰ In order to test the primary outcome measure, a camera was mounted on the roof of the test vehicle to measure the amount of weaving of the car (standard deviation of lateral position). Drivers were instructed to drive with a steady lateral position while maintaining a constant speed of 95 km/hr (60 mph). The study found that amount of weaving was significantly less with immediate-release methylphenidate (18.8 cm) compared with placebo (21.1 cm; P=0.004). Self-reports on various measures of driving quality and driving style were also superior for methylphenidate relative to placebo. However, the study also found that mean lateral position, standard deviation of speed (km/h), and mean speed were not significantly different between the 2 groups.²²⁰ Two additional studies have examined simulator driving performance trials. Results found that immediate-release methylphenidate 10 mg significantly improved an Impaired Driving Score (P=0.05),²⁰⁹ immediate-release methylphenidate 40 mg significantly reduced steering variability,²⁰⁶ and immediate-release methylphenidate 20 mg significantly improved appropriate use of turn signals.²⁰⁶ Although promising, results from driving methylphenidate performance trials should be considered preliminary and would be strengthened by further confirmation based on assessment of effects in patients driving their own vehicles in every-day traffic settings, across multiple occasions.

Methylphenidate extended-release

We included 1 fair-quality trial that compared 24 weeks of treatment with methylphenidate extended-release 41.3 mg to placebo in 363 adults (50% male, mean age of 34 years).²²³ Compared with placebo, when defined as 30% reduction of psychopathology by the Wender-Reimherr Adult Attention Deficit Disorder Scale (WRAADDS), significantly more patients taking methylphenidate extended-release achieved treatment response (61% compared with 41%, P=0.001). The difference between methylphenidate extended-release and placebo in treatment response was not statistically significant when defined as much or very much on the Clinical Global Impression scale (55% compared with 37%; P not reported).

Methylphenidate OROS

We included 7 placebo-controlled trials of methylphenidate OROS, which randomized a total of 1316 adults.^224-230 One trial was rated good quality,²²⁹ 1 was rated poor quality,²²⁴ and the rest were rated fair quality. Treatment duration ranged from 2 weeks to 11 weeks. Dosage regimen was flexible in all but 1 trial, with mean levels ranging from 68 mg to 84 mg. In the remaining trial, adults were randomized to fixed dosages of 18 mg, 36 mg, or 72 mg.²²⁶ Mean age ranged from 31 years to 40 years. One trial only enrolled mother-child dyads in which both were diagnosed with ADHD.²²⁵ Otherwise, the proportion of males ranged from 54% to 66%.

Response rates were reported in all 6 fair-to-good quality trials (Table 11). Regardless of the heterogenous definitions, response rates were generally significantly greater in the methylphenidate OROS groups compared with the placebo groups.

Table 11

Response rates in placebo-controlled trials of methylphenidate OROS.

In the trial that focused on ADHD mothers who had children with ADHD, parenting skills were also measured using the 42-item, validated Alabama Parenting Questionnaire based on mother self-report and collateral reports from individuals who lived with or were close to the mothers.²²⁵ During Phase 1, all mothers were titrated on methylphenidate OROS over 5 weeks for identification of a maximally effective dose. During Phase 2, mothers were then randomized to 2 weeks of treatment with their maximally effective dose of methylphenidate OROS (mean dose 83.7 mg daily) or placebo. Compared with placebo, maximally effective doses of methylphenidate OROS were superior in decreasing the frequency with which mothers used corporal punishment methods and inconsistent discipline. Significant differences were not found between methylphenidate OROS and placebo in effects on involvement, positive parenting, or poor monitoring/supervision behaviors.

Additionally, in 1 of the trials, adults judged to be responders after the first 6 weeks were eligible for a double-blind, 24-week maintenance phase to assess relapse rates.²³⁰ Among the 96 adults who entered the maintenance phase (42% of original group), there was no significant differences between methylphenidate OROS and placebo in the relapse rate, defined as a deterioration of at least 2 points on the Clinical Global Impression-Improvement Scale or decline of the ADHD Investigator Symptom Rating Scale to below a 15% improvement (18% compared with 18%).

Methylphenidate sustained-release

We included 4 fair-quality trials of methylphenidate sustained-release.^231-234 Three of the trials focused on subgroups of adults who were methadone-maintained,²³³ cocaine-dependent,²³¹ or amphetamine abusers,²³⁴and results from these will be discussed in Key Question 3.

The remaining trial compared 4 weeks of treatment with methylphenidate sustained-release 20 mg to nicotine, nicotine plus methylphenidate sustained-release, or placebo in 40 stimulant-naïve, nonsmoking adults (62.5% male, mean age of 37 years).²³² This trial did not report the proportions of adults who achieved a predefined level of clinical response. ADHD symptoms were assessed based on mean scores on the Clinical Global Impressions severity score and the group administered sustained-release methylphenidate alone did not demonstrate significantly greater improvement than those administered placebo. In fact, severity of ADHD symptoms was rated as somewhat higher in the group taking sustained-release methylphenidate alone.

Methylphenidate transdermal system

We found one 4-week, randomized controlled crossover trial that evaluated the effect of methylphenidate transdermal system in 67 adults.²³⁵ However, we rated this trial as poor quality due to an unacceptable level of attrition (22%) and inadequate reporting of methods of randomization and allocation concealment, comparability of baseline patient characteristics based on order of randomization, and numbers analyzed.

Modafinil

The effects of modafinil on core ADHD symptoms in adults remain unknown. We only found 1 crossover trial that compared a single 200 mg dose of modafinil to placebo in 20 adults with ADHD.²³⁶ However, this trial focused on cognitive outcomes and did not evaluate the efficacy of modafinil in treating ADHD symptoms.

Growth effects

An analysis of growth data from the Preschool ADHD Treatment Study found that ADHD patients (N=140; mean age 4.4 years) enrolled in the study were in general larger than average at baseline, based on Centers for Disease Control growth charts (73.1% for height; 79.7% for weight). Use of methylphenidate (mean 337 days) was associated with a reduction in growth rate based on a mixed-effect regression analysis, with a mean loss of −6.35 percentiles in height and −14.42 percentiles in weight. When completers (n=95; mean duration of exposure to methylphenidate, 401 days) were compared with non-completers (n=45; mean duration of exposure to methylphenidate, 202 days) the trend toward reduced growth rate remained. For height, completers had a mean loss of −7.53 percentiles, while non-completers had a mean loss of −3.84 percentiles, while for weight, completers had a mean loss of −13.18 percentiles and non completers had a loss of −17.19 percentile points. Subgroup analysis found that sex, initial height, and initial methylphenidate dose did not moderate the growth reductions. However, initial weight at screening was a significant predictor of greater weight loss during time on trial (F_1,137=7.89; P<0.06).

Direct evidence

Indirect evidence

Direct evidence

Indirect evidence

Methylphenidate OROS

Adverse events were reported in all 6 fair-to good-quality included placebo-controlled trials of methylphenidate OROS (Table 13).^225-230 However, statistical comparison of methylphenidate OROS and placebo was only undertaken in 3 trials.^228-230 In those trials, rates of decreased appetite were consistently significantly greater for methylphenidate OROS compared with placebo. Otherwise, for adverse event withdrawals and insomnia, differences between methylphenidate OROS and placebo did not consistently reach statistical significance.

Table 13

Adverse events in placebo-controlled trials of methylphenidate OROS.

Suicide

Cardiovascular deaths and events

Three retrospective cohort studies evaluated cardiovascular cardiac death (e.g. sudden death) and cardiovascular events in children exposed, or not exposed, to ADHD medications,^{251, 273-275} and 1 retrospective cohort study evaluated cardiovascular events in adults exposed to ADHD medications.²⁷¹

The largest study evaluating cardiac death in children included 241 417 incident (new) users of ADHD medications (specifically, amphetamines, methylphenidate, and atomoxetine) and evaluated only validated cases of sudden death or ventricular arrhythmia.²⁷⁴ Estimates for risk of sudden death or ventricular arrhythmia associated with methylphenidate current use compared with never use was elevated but not statistically significant, with a wide confidence interval (hazard ratio, 2.36; 95% CI, 0.29 to 23.69). Risk with use of any of the 3 types of medication was lower but also not found to be statistically significantly elevated (hazard ratio, 1.60; 95% CI, 0.19 to 13.60). Unfortunately, this study did not make direct comparisons among the different drugs, but the crude incidence rates per 10 000 person years was higher in methylphenidate users (hazard ratio, 0.16; 95% CI, 0.00 to 0.89) than either amphetamines or atomoxetine (both hazard ratios, 0.00; 95% CIs, 0.00 to 0.79 and 0.00 to 5.8, respectively). This study also examined cardiovascular outcomes such as stroke, myocardial infarction and composite outcomes. For all of these outcomes the hazard ratios were less than 1, indicated a reduced risk with exposure to the ADHD medication, but again the confidence intervals indicated a nonstatistically significant finding. Only “nonaccidental death” resulted in a (barely) statistically significant lower risk among users of any ADHD medication compared with nonusers (hazard ratio, 0.53; 95% CI, 0.29 to 0.99). In a much smaller study based on the GPRD database in the United Kingdom, 5321 patients (age 2 to 21 years) who had received at least 1 prescription for methylphenidate, dexamphetamine, or atomoxetine, 7 deaths occurred and none were deemed to be caused by “sudden death”.²⁷³

A good-quality retrospective cohort study based on 10 years of Florida Medicaid claims data and the Vital Statistics Death Registry data identified 55 383 patients with newly diagnosed ADHD.^{251, 275} Of these, 32 807 had used a stimulant (either currently or formerly) and 22 576 had never used a stimulant medication. Of 73 children who died over the study period, 5 died of circulatory causes (4 per 100 000 person-years); none of these were sudden cardiac death and numbers were too small to make reliable comparisons among groups. Emergency department and physician office visits due to cardiac causes occurred significantly more often in the group currently using a stimulant compared with non-users (hazard ratio, 1.20; 95% CI, 1.04 to 1.38 and hazard ratio, 1.21; 95% CI, 1.06 to 1.39, respectively). Former use of stimulants was not significantly associated. Comparison of current users of methylphenidate products to those currently using amphetamine products showed no statistically significant difference in the rate of emergency department visits for cardiac reasons (hazard ratio, 1.01; 95% CI, 0.80 to 1.28). Comparison of former use of these products also resulted in a nonsignificant finding.

In a good-quality case-control study, children (ages 7 to 19) who had died from “sudden unexplained death” during the years of 1985 to 1996 were identified from state vital statistics from each of the 50 United States.²⁷⁶ A control group was selected from children who died from motor vehicle traffic accidents, with 1:1 matching resulting in 564 in each group. The exposure was defined as stimulant use immediately prior to death, based on survey of parents. Ten (1.8%) of those with sudden death were reported to have been taking immediate-release methylphenidate at the time of their death, compared with 2 (0.4%) in the motor vehicle death group, resulting in an odds ratio of 7.4 (95% CI, 1.4 to 74.9). Sensitivity analyses altering the way exposure was identified or removing children also taking a tricyclic antidepressant did not meaningfully alter the results. Recall bias is the greatest risk in this study, as the time since the child's death to the survey of the parent was longer in the sudden death group (13 years) compared with the motor vehicle death group (10 years).

In a fair-quality retrospective cohort database study, the risk of stroke or transient ischemic attack was compared in adults taking initiating atomoxetine or stimulant therapy for ADHD/ADD (N=42 993). Using propensity score matching and only validated cases of stroke or transient ischemic attack, no statistically significant difference was found between the drugs for either outcome (relative risks, 1.38; 95% CI, 0.42 to 4.54 for stroke, and 0.31; 95% CI, 0.04 to 2.63 for transient ischemic attack).²⁷¹ Overall the numbers of cases were small, limiting the ability to determine statistically significant differences if they were to exist.

Blood pressure, pulse, electrocardiographic changes

Lisdexamfetamine. Three trials of lisdexamfetamine 30, 50, or 70 mg daily compared with placebo reported various intermediate outcome measures of cardiovascular function. Small increases in blood pressure and heart rate were seen in all 3 studies, with none being deemed clinically important. A small, open-label, uncontrolled 11-month study of lisdexamfetamine in 272 children did not find any cases of “clinically relevant” changes in blood pressure or electrocardiographic parameters.²⁷⁷ A 4-week, placebo-controlled trial of 314 adolescents reported increases in heart rate of 2 to 3 beats per minute with lisdexamfetamine (1 beat per minute increase with placebo).¹⁸⁴ Change in blood pressure was negligible. Two patients, both taking 70 mg daily doses of lisdexamfetamine, discontinued drug after 1 week due to increased QTc intervals. Additionally, a 4-week randomized, placebo-controlled trial of 30 mg, 50 mg, and 70 mg of lisdexamfetamine in 420 adults (54% male, mean age was 35.1 years) reported cardiovascular adverse events.²⁷⁸ Although statistical analysis of between-group differences was not reported, for lisdexamfetamine 30 mg, 50 mg, 70 mg, and placebo, respectively, rates for blood pressure increase were 1%, 3%, 4% and 0%, for increased heart rate were 1%, 3%, 3%, and 0%, for palpitations were 2%, 1%, 3%, and 0%, and for tachycardia were 1%, 3%, 0%, and 0%.²⁷⁸

Methylphenidate OROS. An open-extension of a trial of methylphenidate OROS reported small changes in blood pressure (3.3 mmHg systolic and 1.5 mmHg diastolic) and heart rate (3.9 bpm) over a 1 year study period.²⁷⁹ During this time, 33% discontinued treatment, but only 1 withdrew due to systolic blood pressure >130 mmHg. ANOVA analyses showed no relationship to dose or age and no tolerance development over time was found, but those children with the lowest blood pressure at baseline had the greatest increases. The final report from this 2-year study found no additional withdrawals due to cardiovascular adverse events.²⁸⁰

In a 7-week study of 226 adults (56% male, mean age of 39 years), similar proportions of participants in the methylphenidate OROS and placebo groups, respectively, had systolic blood pressure greater than 140 mmHg at any post baseline visit (8% compared with 6%), but greater proportions of participant in the methylphenidate OROS group had diastolic blood pressure greater than 90 mm Hg (10% compared with 3%; P not reported) and a pulse rate of greater than 100 bpm (7% compared with 2%; P not reported).²²⁷

Mixed amphetamine salts XR. Four open-label extension studies of mixed amphetamine salts XR, 1 each in children,^{281, 282} adolescents,²⁸³ and adults examined the cardiovascular effects over periods of 6 to 24 months.²⁸⁴ In each of these studies the subjects were populations of patients who were highly selected and were described as being healthy other than the diagnosis of ADHD. The studies in children and adolescents also included a short-term placebo-controlled phase. While no statistically significant differences compared with placebo in any electrocardiogram measure were found in children in the short-term trial, 2% (11/568) had diastolic blood pressure >90 mmHg, and 9% (50/568) had a systolic blood pressure >130 mmHg at some point during follow-up. Overall, 0.7% (4/586) withdrew from the study due to a cardiovascular adverse event; 1 due to tachycardia (max 121 bpm compared with 108 bpm at baseline), 2 due to chest pain (both had sinus bradycardia at baseline), and 1 due to elevated blood pressure (130/90 mmHg that resolved to 115/80mmHg after 1 month without drug). In a shorter duration open-label study, 2968 children were given mixed amphetamine salts XR for a period of up to 15 weeks.²⁸² The absolute numbers of patients with cardiovascular adverse events were not clearly reported. It was reported that 0.2% (7/2968) discontinued mixed amphetamine salts XR due to cardiovascular adverse events. Nine patients had treatment emergent cardiovascular adverse events that were moderate or serious in intensity, 5 of which were deemed probably related to mixed amphetamine salts XR.

Thirteen of 79 adolescent patients (16%) experienced adverse events during a 4-week study of mixed amphetamine salts XR compared with placebo that included cardiovascular symptoms such as syncope, tachycardia, and electrocardiogram abnormality.²⁸³ Of these, 2 were withdrawn from study drug, 1 with palpitations and 1 with severe migraine and syncope. During 6-month follow-up there were no serious cardiovascular adverse events reported, although 4% (6/138) reported adverse events with cardiovascular symptoms, however none withdrew due to these adverse events. In a 2-year extension study in adults with ADHD, two-thirds discontinued the study prior to completing 2 years, 22% because of adverse events.²⁸⁴ Statistically significant, but not considered clinically meaningful, increases in systolic blood pressure and diastolic blood pressure were seen at various points throughout the study (mean increase in systolic blood pressure, 2.3 mmHg; diastolic blood pressure, 1.3 mmHg at endpoint). While a statistically significant increase in QTcB (7.2 msec; P<0.001) was found, no patient had a QTcB >480 msec. Three percent withdrew due to cardiovascular events (2 due to palpitations or tachycardia – extent not reported, and 5 due to hypertension).

Atomoxetine. Open-label extension studies of atomoxetine have reported on cardiovascular adverse events in children or teens²⁵⁷ and in adults.²⁸⁵ One report involved 169 children and adolescents that continued on open or blinded atomoxetine (maximum dose of 2 mg/kg divided into twice daily) for at least 1 year following 3 short-term, placebo-controlled trials.²⁵⁷ The timing of electrocardiogram measurements was not stated, but was presented by increasing dose. Linear regression suggested that there was no evidence of an increase in QTc with increasing dosage of atomoxetine.²⁵⁷ An interim analysis of an open-label extension study in adults reported no “clinically relevant changes in QTc” after a mean of 97 months of follow-up.²⁸⁵

Growth effects

A non-systematic review, using estimation techniques, graphing, and qualitative synthesis, found that stimulants (amphetamines and methylphenidate) caused growth delays in both height and weight but that these were attenuated over time.²⁴⁶ The qualitative analysis indicated that there may be a dose effect, that there are no important differences between amphetamines and methylphenidate, and that discontinuing treatment results in resumption of normal growth. Because this review was not systematic and pooled data from a wide variety of study designs, we suggest caution in interpreting these findings.

A frequently cited nonsystematic review concluded that effects on weight and height associated with immediate-release methylphenidate vary across short-term clinical trials and long-term observational studies and are mostly transient.²⁸⁶ We reached similar conclusions based on our analysis of a larger number of primarily long-term observational studies that compared immediate-release methylphenidate to immediate-release dextroamphetamine,^{258, 259, 265} or unmedicated hyperactive control groups.^{261, 265, 266} Height and weight changes associated with immediate-release methylphenidate^{254, 256, 260, 262, 263} and OROS were also observed in long-term noncomparative studies.²⁶² A noncomparative study of mixed amphetamine salts (Adderall XR^®) found a low overall rate of withdrawal due to weight loss (4.8%), however weight loss was the most common reason for withdrawal from this 24-month extension of placebo-controlled trials.²⁸⁷ Multiple noncomparative study findings provide inconclusive evidence regarding immediate-release methylphenidate effects on children's height and weight. Analysis of 2- and 5-year data from open-label extensions of 13 trials of atomoxetine assessed the effect on height and weight.^{249, 252}

We did not analyze results from a poor-quality, comparative study of growth rebound in methylphenidate and immediate-release dextroamphetamine due to our concerns about how possible additional biases may have affected the results.²⁶⁷ We cannot rule out the possibility of between-groups differences in baseline characteristics because no information/analysis was provided. We also cannot rule out the possibility that the results were confounded by time and other relevant factors. An additional study related to growth reported on tooth maturation in children taking immediate-release methylphenidate compared with an unexposed control group, finding no difference (Table 14).²⁸⁸

Table 14

Direct comparisons of long-term height and weight outcomes.

Insomnia, decreased appetite, and headaches

A small (N=150), 24-month, retrospective cohort study examined rates of insomnia, decreased appetite, and headache reported by children attending a single clinic database.²⁴⁸ Using a one-way ANOVA analysis, the rates of insomnia across immediate-release methylphenidate, methylphenidate OROS, mixed amphetamine salts, mixed amphetamine salts XR, and atomoxetine were not statistically significantly different, although the crude rate in the mixed amphetamine salts group (22%) was numerically greater than in the other groups (range 4% to 13%). Similarly, rates of decreased appetite were not found to be different, although the rates in the immediate-release mixed amphetamine salts, mixed amphetamine salts XR, and methylphenidate OROS groups (range 15% to 22%) were also higher than the atomoxetine and immediate-release methylphenidate groups (range 9% to 10%). Atomoxetine had lower rates of headache compared with mixed amphetamine salts XR (0% and 12%, P=0.001), immediate-release mixed amphetamine salts (0% and 11%, P=0.001), or methylphenidate OROS (0% and 10%, P=0.002). Dose was not controlled for in these analyses, and because the data were sparse a Bonferroni correction was used, thus we suggest caution in interpreting these findings.

Tics

Five observational studies and 2 meta-analyses reported tic-related outcomes.^{248, 254, 262, 264, 280, 294-296} One of these was a long-term placebo-controlled trial²⁹⁴ of immediate-release methylphenidate. Table 15 summarizes the characteristics and outcomes from these studies. Although the 1-year study started out with similar numbers assigned to placebo and methylphenidate, by the study end 72 were on methylphenidate and only 18 on placebo. Development of new tics or worsening of pre-existing tics was not different between the 2 groups. The studies do not provide any information about how different pharmacologic treatments for ADHD compare in safety with regard to tic-related outcomes.

Table 15

Tic-related outcomes in observational studies.

A meta-analysis of data from 3 short-term trials found similar rates of tics reported as an adverse event among Immediate-release methylphenidate, methylphenidate OROS and placebo.²⁹⁵ This same publication also reported on 2 open-label studies of methylphenidate OROS, 1 of which was already included here,²⁶² the other is a report on a 9-month community-use study in children, adolescents, and adults, for which no reference is given (see table 15). A second meta-analysis included 9 trials involving dextroamphetamine, methylphenidate “derivatives”, clonidine, guanfacine, and atomoxetine among other drugs not included in this review.²⁹⁶ This analysis found that methylphenidate “derivatives” (primarily immediate-release methylphenidate) did not increase tic severity, but improved ADHD symptoms, compared with placebo. Clonidine and guanfacine combined improved tic severity (Table 15) compared with placebo.

The rate of treatment emergent tics varied widely across the studies. Because these studies lack comparative elements and vary in design, higher quality evidence is needed to establish the risk of developing treatment emergent tics with ADHD medications.

Seizures

In an analysis of post marketing data and clinical trials data, the manufacturer of atomoxetine found that the rate of seizure was 01.% to 0.2%, with no statistically significant difference in rate between atomoxetine, methylphenidate, and placebo, although the comparative data were limited.²⁵⁰ In a good-quality retrospective cohort study of claims data for 34 727 children (age 6 to 17) with at least 2 codes for ADHD diagnosis, the risk of new onset seizure was higher with atomoxetine (relative risk, 2.5; 95% CI, 0.9 to 7.1) than “other ADHD therapy” (stimulants and bupropion combined, relative risk, 0.7; 95% CI, 0.2 to 2.0) compared with “nonuse”, although neither was statically significant.²⁷² The confidence intervals are quite wide and unfortunately, the study did not make direct statistical comparisons of the drugs and drawing conclusions from these findings was not possible.

Injuries

A retrospective database study analyzed an association between childhood behavioral disorders and common childhood injuries by using the British Columbia Linked Health Data Set to identify injuries. Children with behavioral disorders were identified using methylphenidate prescriptions as a proxy for diagnosis using data in a Triplicate Prescription Program.²⁵³ Injury frequencies in children prescribed methylphenidate at least once between January 1, 1990 and December 31, 1996 (n=16 806) were compared with those in children not taking methylphenidate (n=1,010,067). Children were 51.4% male and less than 19 years in age. Mean duration of exposure was not identified. Odds of any injury (fractures, open wounds, poisoning/toxic effect, intracranial, concussion, and burns) were significantly higher in children taking methylphenidate than for those not taking methylphenidate (odds ratio, 1.67; 95% CI, 1.54 to 1.81), even after adjusting for baseline age, sex, socioeconomic status, and region. This study design clearly suffers from lack of sensitivity to diagnosis, in that an unknown number of children with behavioral disorders are included in the group not taking methylphenidate. Since methylphenidate was used simply as a proxy for behavioral disorders, the relationship between the drug and the increase in injuries is not necessarily clear.

Association between treatment of ADHD with drug therapy in childhood and later development of substance abuse

This was a much-discussed topic in the literature,^301-309 but a clear conclusion has not yet been reached. The evidence is largely limited to longitudinal studies assessing the relationship of treatment with a stimulant during childhood and later substance use in adolescence or adulthood. None of these studies were comparative in terms of the specific stimulant drugs used during treatment, with most reporting immediate-release methylphenidate as the most commonly used drug. We did not find any evidence assessing the impact of nonstimulant drugs or extended-release stimulants on later substance use/abuse in patients with ADHD. In general these studies suffered from methodologic flaws that hinder clear conclusions from being drawn. Some depend on data that appear to have been collected for other purposes, or at least not for the specific purpose of assessing future substance abuse, and definition or methods of determination of substance abuse is not consistent across studies. There is general agreement that the rate of substance use in adolescence or adulthood is higher among those diagnosed with ADHD in childhood, compared with healthy controls, and that age of diagnosis (younger ages), severity of symptoms, and presence of conduct disorder increase the likelihood of later substance use. However, the impact of drug treatment during childhood on later substance use is not clear, and in fact there is distinctly conflicting evidence. The major concern raised regarding these studies is the lack of controlling for potential confounding, particularly severity of ADHD, age at follow-up (assessment during adolescence not allowing enough time for exposure to illicit substances), the definition of substance use (for example “ever use” compared with substance use disorder), and exposure to substances during childhood (for example cigarette smoking by parents or other relatives). We have rated all of these studies as fair quality and suggest caution in interpreting the results of any 1 study as conclusive.

We found a total of 12 fully published studies,^{166, 310-323} 3 of which had follow-up publications with additional analysis.^{310-315, 318} Additional studies were cited by others, many of which are only published as abstracts, do not address stimulant use, or were not available to us.^324-330 Several of these made comparisons to groups of children without ADHD.^{311, 313-315, 317, 318, 331, 332} Because these comparisons were less relevant than those comparing adolescents and adults with ADHD as children who did and did not received stimulants, these were not considered further. There were 7 studies that made relevant comparisons.^{166, 310, 319-321, 323, 331} Below is a summary of the findings of these studies (Table 16). These were generally small studies, with limited ability to control for all important confounding factors. Importantly, differences among the groups at baseline that may have lead to 1 group receiving a stimulant and the other not well identified particularly in the older studies where these data may not have been recorded. Therefore, the findings should be interpreted as suggestive and require further research to confirm. Overall, the studies of stimulant use in childhood and later abuse or dependence on nicotine, alcohol, or illicit drugs compared with children with ADHD but not exposed to stimulants did not indicate an increased risk. Some indicate a protective effect, but it appears that conduct disorder may be an important modifying factor.

Table 16

Relationship between stimulant treatment for ADHD and later substance abuse and dependence.

Rates of nicotine abuse and dependence were assessed in 4 studies,^{319, 321, 323, 331} with 1 finding stimulant use in girls protective against nicotine abuse (regular smoking) as adolescents (hazard ratio, 0.28; 95% CI, 0.14 to 0.60).³²³ Another found no association with the rate of or timing of the first cigarette, but did find that stimulant exposure delayed the time to regular smoking by 2 years and 1 month.³²¹ Two other studies, in males, however found no associations after controlling for confounding factors including conduct disorder.^{319, 331}

Four studies found no associations between alcohol abuse during teen and young adult years and stimulant exposure during childhood.^{166, 310, 319, 320} Earlier examinations of data from 1 longitudinal follow-up study had indicated a protective effect at 5 years of follow-up,³¹¹ but this association was no longer seen with 10 years of follow-up.³¹⁹

In examining substance abuse, 2 studies found stimulant use to be protective,^{320, 323} but a third study found that controlling for conduct disorder resulted in a nonsignificant finding.³¹⁹

Direct comparisons

Based on post-hoc subgroup analyses, differences in ADHD symptom response between boys and girls were not found in 5 trials of various drugs. Subgroup analyses based on gender were performed based on data from 2 double-blind, randomized controlled trials of lisdexamfetamine.²⁴⁰ The average SKAMP-DS scores for lisdexamfetamine were similar to mixed amphetamine salts XR and superior to placebo regardless of gender in the 1-week, crossover study (#201). In the 4-week, parallel-group trial, treatment effects appeared less robust in subgroups of girls for all dosage groups of lisdexamfetamine compared with placebo, but changes in ADHD rating scale IV lost statistical significance only in the 30 mg treatment group (−19 compared with −8.1; P=0.0537). Results from the subgroups of girls in study #301 must be interpreted with caution, however, due to the small sample sizes (N=88).

A subgroup analysis of the START study, comparing mixed amphetamine salts XR and atomoxetine, examined the effects in the 57 girls enrolled.³⁴⁶ Similar to the overall study analysis, mixed amphetamine salts XR was found to have greater improvements in symptoms based on the SKAMP deportment and attention subscale scores compared with atomoxetine. In the original analysis, 71.9% of the children enrolled were boys. A post-hoc analysis of data from the COMACS study, comparing methylphenidate OROS and methylphenidate CD, found differences between boys and girls, but not between drugs. At baseline, more girls had comorbid anxiety disorder and girls had superior response rates at 1.5 hours post dose, but inferior response rates at 12 hours post dose compared with boys.³⁴⁷ A post-hoc analysis of a small crossover study of 35 adolescents with ADHD comparing methylphenidate OROS and mixed amphetamine salts XR (and placebo) found that while females had lower symptoms scores, statistically significant interaction between drug and sex were not found based on self-report, parent report, or simulated driving skill.³⁴⁸ This study was small, and may have not had adequate statistical power (Type II error).

Indirect comparisons

We found 3 studies examining differences in response to stimulants (primarily immediate-release methylphenidate) between boys and girls.^{80, 349, 350} Two found no differences between boys and girls,^{80, 350} while the third found that during the task period, boys were significantly more compliant and mothers gave fewer commands and more praise comments than in the girls group.³⁴⁹ All 3 studies suffer from design and conduct flaws, including important differences between the groups at baseline and not accounted for in the analysis, and comparison to historical controls.

Data from girls enrolled in 2 separate placebo-controlled trials of atomoxetine with identical protocols were analyzed post-hoc to assess the effects in this subgroup of children.³⁵¹ This analysis of 52 girls reported similar efficacy to that reported for the whole trial group (atomoxetine superior to placebo on most measures) but did not make a comparison of the effects in boys compared with girls.

A pooled analysis of two 10-week, placebo-controlled trials (N=536; 35% female, 65% male) of atomoxetine in adults examined gender differences.³⁵² The study found that when compared with baseline, a statistically significant change favoring atomoxetine was observed among both genders on the multiple ADHD rating scales (P<0.05). This study conducted multiple exploratory analyses of differences in gender based on treatment effects. At endpoint, atomoxetine resulted in better scores in women on emotional dysregulation (temper + mood lability + emotional overactivity) items on the Wender-Reimherr Adults Attention Deficit Disorder Scale compared with men. The Sheehan Disability social life subscale demonstrated a significant gender-by-treatment effect (P=0.042), with women showing a stronger treatment effect than men, but there was no significant difference on the total score. No other analyses showed a gender difference. Considering the post-hoc, exploratory nature of these analyses and the smaller number of women than men in these studies, these findings are preliminary.

Post-hoc analyses of data from the COMACS study, combining methylphenidate OROS and methylphenidate CD adverse event data compared with placebo, found that sex was not a predictor of appetite/sleep disturbance adverse events.²³⁸ A small (N = 35; 19 males and 16 females), fair-quality, crossover study of methylphenidate OROS and extended-release mixed amphetamine salts in adolescents reported analyses of the differences in effects based on sex and drug assigned.³⁴⁸ Multiple ANOVA analyses were conducted on parent and student assessed Connors scale, a modified Connors scale for use by adolescents using a hand-held computer, and simulated driving scores comparing medication to either placebo or standard care (minimal or no medication). While this study had limitations, the analyses did not show a correlation between sex and effect with medication.

Guanfacine XR

A study of 217 children with comorbid ADHD and oppositional symptoms using flexible dosing (1-4 mg daily) over 8 weeks found that the mean least squares mean change on the ADHD-RS-IV scale was −23.8 in the drug group and −11.5 in the placebo group (P<0.001 from a baseline of 42 in both groups).¹⁶³ The subscale scores on the CPRS-RS-L oppositional defiant subscale also improved more with guanfacine XR (−10.9 compared with −6.8; P<0.001) and the change in the 2 scores was found to be highly correlated (r=0.74). Slightly more patients were taking 3 mg daily doses, and only few were taking 1 mg daily.

Immediate-release methylphenidate. Two placebo-controlled trials of immediate-release methylphenidate given twice daily studied children with oppositional defiant disorder and ADHD.^{367, 368} In both studies, immediate-release methylphenidate was effective in reducing ADHD symptoms relative to placebo. In the larger study (N=267), the presence of oppositional defiant disorder as a comorbidity did not affect the response to immediate-release methylphenidate 0.5 mg/kg compared with those without this comorbidity.³⁶⁸ In the smaller study (N=31), 3 doses were studied and only the 0.5 mg/kg dose was consistently found effective in the presence of oppositional defiant disorder as a comorbidity.³⁶⁷

Mixed amphetamine salts XR. The efficacy and adverse effects of mixed amphetamine salts XR 10-40 mg has also been studied in 235 children with ADHD and oppositional defiant disorder.³⁶⁹ This 4-week placebo-controlled trial focused on oppositional defiant disorder as the primary diagnosis, with only 79.2% of the original 308 children having comorbid ADHD. In the oppositional defiant disorder plus ADHD subgroup population, improvements in ADHD symptoms were significantly greater for mixed amphetamine salts XR compared with placebo on the parent- and teacher-ratings on the ADHD subscale of the SNAP-IV. Although these findings are encouraging, there are some limitations to consider. Mean change from baseline on the ADHD subscale of the SNAP-IV was included as a secondary outcome measure and it is unclear if the analysis was adequately powered to measure between-group differences.

Methylphenidate OROS and methylphenidate CD

Post-hoc analyses of data from the COMACS study, combining methylphenidate OROS and methylphenidate CD adverse event data compared with placebo, found that comorbidity with oppositional defiant disorder was not a predictor of appetite/sleep disturbance adverse events.²³⁸

Children

Overall, 7 head-to-head trials and 10 placebo-controlled trials reported symptoms of anxiety or nervousness as an adverse event and 1 head-to-head comparison and 3 placebo-controlled trials reported it as a symptom of ADHD or comorbidity.

In the head-to-head comparisons (immediate-release methylphenidate compared with immediate-release dextroamphetamine, mixed amphetamine salts, methylphenidate SR, methylphenidate OROS, or atomoxetine), no statistically significant differences in the rate of reporting anxiety as an adverse event were found, although for some comparisons numerical differences were apparent.^{43, 48, 49, 102, 106, 128, 371} For example, compared with immediate-release methylphenidate, rates were higher with atomoxetine (15.8% compared with 10% nervousness) and immediate-release dextroamphetamine (68% compared with 61%), but lower compared with Adderall^® (10% compared with 5%) or methylphenidate OROS (31.3% compared with 18.7% in 1 study, 12% compared with 13% in another). Two trials assessing anxiety symptoms as part of ADHD did not find a difference in anxiety between immediate-release methylphenidate and methylphenidate SR in children with minimal brain dysfunction⁴³ or between immediate-release methylphenidate and placebo in children with ADHD and mental retardation.³⁷²

Placebo-controlled trial evidence was conflicting; some studies showed higher rates of anxiety or nervousness with methylphenidate, indicating a dose-dependent effect, while others showed no increase over placebo rates.^{30, 125, 126, 135, 137, 372-377} Reports of anxiety were similar between placebo and atomoxetine in 2 studies^{135, 137} and placebo and modafinil in 2 others.^{125, 126} Because most of these studies reported these as spontaneously reported adverse events, we do not believe that the quality of the data warrants a conclusion.

A placebo-controlled trial examining the impact of comorbidities on ADHD response to immediate-release methylphenidate found that the small subgroup of children with symptoms of anxiety at baseline (N=28) had statistically significantly lower response rate, based on a clinical consensus of response compared with those who did not have anxiety at baseline (N=239; 50% compared with 71.9%; P=0.02).³⁶⁸ A second study of children with ADHD, tic disorders, and anxiety was rated poor quality due to inadequate reporting on multiple methods including blinding, comparison of patient characteristics at baseline between exposure groups, attrition, and handling of missing data.³⁶⁷

A 12-week fair-quality placebo-controlled study of atomoxetine in children with both ADHD and anxiety disorder diagnoses examined the affect on both ADHD and anxiety symptoms.³⁷⁸ In the intent to treat analysis, atomoxetine was superior to placebo in both improvements on ADHD symptoms and anxiety symptoms (−4.5 compared with −2.4 points on the Pediatric Anxiety Rating Scale; P<0.010). This study had a high drop-out rate; 25% overall; 10% dropped out during a 2-week placebo run-in phase, and another 16% dropped out during the 10-week treatment phase. With a high drop-out rate, we recommend caution in interpreting these findings.

Adults

The effect of atomoxetine on ADHD and comorbid anxiety has been evaluated in 2 trials.^{196, 379} One trial prospectively enrolled only adults with ADHD and comorbid social anxiety disorder (N=442).¹⁹⁶ After 14 weeks, there was a significantly greater improvement with atomoxetine 82.9 mg than placebo on the primary efficacy measure of mean change on the Conners' Adult ADHD Rating Scale, Self-Report Screening Version (−8.7 points compared with −5.6 points; P<0.001). Response rates were not reported.

The second publication reported findings from exploratory, post-hoc analyses of the effects of lifetime, but not current, diagnoses of DSM-IV comorbidity on response to atomoxetine compared with placebo.³⁷⁹ The main finding of these subanalyses was that compared with adults with “pure” ADHD (no comorbidities), adults with ADHD and post-traumatic stress disorder had greater improvements on atomoxetine compared with placebo when based on Investigator ratings, but not when based on patient self-report measures. While these findings provide rationale for design of future prospective research, they must be viewed in light of their limitations. These were post-hoc analyses of subgroups of unknown size and it was unclear as to whether they involved comparisons of atomoxetine and placebo groups that were well-matched on important baseline characteristics or whether there was any adjustment for potential confounders. Results from the primary analyses of these data were reported in an earlier, separate publication ¹⁹¹ and are discussed under Key Question 1.

Additionally, numerous placebo-controlled trials examined whether treatment with ADHD drugs improves comorbid anxiety symptoms.^{181,191, 198, 206, 207, 210-212, 218, 221, 232, 236} However, only immediate-release methylphenidate was generally consistently associated with improvements in anxiety symptoms in adults with ADHD.^{207, 218, 380} The only exception was that in a trial of 45 adults, similar numbers of participants with immediate-release methylphenidate compared with placebo (7% compared with 4%), had anxiety as defined as a Hamilton Anxiety Scale score above 21 points.^{206, 207, 210-212, 218, 221} Finally, in terms of adverse effects, only methylphenidate OROS has been associated with significantly greater adverse anxiety effects in adults than placebo across 2 trials.^{224, 228}

Direct evidence

In a fair-quality randomized, double-blind, double-dummy trial of 136 children (mean ages 9.7 to 10.7 years) with preexisting tic disorders, mainly Tourette's disorder, children were assessed for improvement in ADHD symptoms and worsening tics in response to immediate-release methylphenidate, clonidine, and the combination over 16 weeks.¹⁵¹ Mean doses at the end of study were 0.25 mg clonidine and 26 mg immediate-release methylphenidate daily. All analyses made comparisons of each drug group to placebo; although it was stated that there is no difference between the immediate-release methylphenidate and clonidine groups on the primary outcome measure of the Conner-ASQ-Teacher scale and the combination therapy provided the largest effect size compared with placebo. It was also noted that immediate-release methylphenidate resulted in better scores than clonidine on attentiveness, while clonidine had better scores on improving more emotional items (e.g. “demands must be met immediately, easily frustrated”). The rate of worsening tics was similar between the drug groups: 22% with immediate-release methylphenidate alone, 26% with clonidine, 18% with the combination, and 22% with placebo. However, 35% assigned to methylphenidate alone had to limit dosage increases due to tics, compared with 18% with clonidine alone or 15% with the combination. After a small worsening in score with methylphenidate alone at 8 weeks on the Yale Global Tic Severity Scale, scores on 3 scales assessing tic symptoms were significantly reduced at endpoint in all drug groups with no direct comparisons across groups presented.

Indirect evidence

There is concern that stimulant drugs may be contraindicated in ADHD patients with comorbid tic disorders due to possible tic exacerbation. There has also been uncertainty about whether stimulants treat ADHD symptoms as well in children with ADHD and established tic disorders as they do in children with primary ADHD. Several placebo-controlled trials of primarily immediate-release methylphenidate have examined these issues.^{85, 151, 386-392} Immediate-release dextroamphetamine and atomoxetine treatments for ADHD have also been studied in children with tic disorders.^{85, 389, 391, 393}

The majority of these trials were of short duration and involved very small numbers of children.^{85, 386-388, 390, 394} Children participating in these trials were mostly male (≥ 85%), with a range in age of 8.3 years to 11.2 years. Motor and verbal tic frequency and severity were assessed in classroom, lunchroom, and playground settings using a variety of different rating scales. The most common tic rating scale used was the Yale Global Tic Severity Scale.

Overall, there was very little evidence across these trials to indicate that immediate-release methylphenidate, immediate-release dextroamphetamine, or atomoxetine were associated with any tic exacerbation effects. Paradoxically, in one 2-week trial of 34 children, only the lowest dose of immediate-release methylphenidate (0.1 mg/kg daily) was associated with any tic worsening, characterized by an increase in motor tics only in the classroom setting.^{386, 388} In another 3-week trial of 12 children, only the higher dosages of immediate-release methylphenidate (0.67 mg/kg daily or 1.20 mg/kg daily) were associated with tic exacerbations.⁸⁵ Otherwise, compared with placebo, immediate-release methylphenidate, immediate-release dextroamphetamine, and atomoxetine were all consistently associated with improved tic severity in these trials. Furthermore, children also showed greater improvements in ADHD symptoms with immediate-release methylphenidate, immediate-release dextroamphetamine, and atomoxetine compared with placebo. Observational evidence of the impact of immediate-release methylphenidate treatment indicates that the baseline frequency and severity of motor and vocal tics was significantly higher than during the placebo phase of the study, and no differences were found among the placebo and 12, 18, and 24 month immediate-release methylphenidate treatment follow-up periods.³⁹⁵

Substance use disorder