Methylphenidate for children and adolescents with attention deficit hyperactivity disorder (ADHD)

Summary of findings for the main comparison. Methylphenidate compared with placebo or no intervention for ADHD

Methylphenidate compared with placebo or no intervention for ADHD
Patient or population: children and adolescents (up to and including 18 years of age) with ADHD Settings: out‐patient clinic, in‐patient hospital ward and summer school Intervention: methylphenidate Comparison: placebo or no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Placebo or no intervention	Methylphenidate
ADHD symptoms: all parallel‐group trials and first‐period cross‐over trials ADHD Rating Scale (Teacher‐rated) Average study duration: 74.8 days		Mean ADHD symptom score in the intervention groups corresponds to a mean difference of 9.6 (95% CI 11.25 to 8.00) on ADHD Rating Scale	SMD ‐0.77 (‐0.90 to ‐0.64)	1698 (19 studies)	⊕⊝⊝⊝ Very low^a,b	The analysis was conducted on a standardised scale with data from studies that used different teacher‐rated scales of symptoms (Conners' Teacher Rating Scale (CTRS), Strengths and Weaknesses of ADHD Symptoms and Normal Behaviour (SWAN) Scale, Schedule for Non‐adaptive and Adaptive Personality (SNAP) ‐ Teacher, Fremdbeurteilungsbogen für Hyperkinetische Störungen (FBB‐HKS)). The effect size has been translated on to the ADHD Rating Scale from the SMD
Total number of serious adverse events	Trial population		RR 0.98 (0.44 to 2.22)	1532 (9 studies)	⊕⊝⊝⊝ Very low^a,c	‐
	16 per 1000	16 per 1000 (7 to 36)
Total number of non‐serious adverse events	Trial population		RR 1.29 (1.10 to 1.51)	3132 (21 studies)	⊕⊝⊝⊝ Very low^a,b	‐
	408 per 1000	526 per 1000 (448 to 615)
General behaviour: all parallel‐group trials and first‐period cross‐over trials General behaviour rating scales (Teacher‐rated)		Mean general behaviour score in the intervention groups was 0.87 standard mean deviations lower (95% CI 1.04 to 0.71 lower)	SMD ‐0.87 (‐0.71 to ‐1.04)	668 (5 studies)	⊕⊝⊝⊝ Very low^a,d	‐
Quality of life (Parent‐rated)		Mean quality of life score in the intervention groups corresponds to a mean difference of 8.0 (95% CI 5.49 to 10.46) on the Child Health Questionnaire	SMD 0.61 (0.42 to 0.80)	514 (3 studies)	⊕⊝⊝⊝ Very low^a,e	‐
The basis for the assumed risk* (e.g. median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). ADHD: Attention deficit hyperactivity disorder; CI: Confidence interval; RR: Risk ratio; SMD: Standardised mean difference
GRADE Working Group grades of evidence. High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
^aDowngraded two levels due to high risk of bias (systematic errors causing overestimation of benefits and underestimation of harms) in several risk of bias domains, including lack of sufficient blinding and selective outcome reporting (many of the included trials did not report on this outcome). ^bDowngraded one level due to inconsistency: moderate statistical heterogeneity. ^c Downgraded one level due to imprecision: wide confidence intervals. ^dDowngraded one level due to indirectness: children's general behavior was assessed by different types of rating scales with different focus on behavior. e Downgraded one level due to indirectness: children's quality of life was assessed by their parents.

Background

Description of the condition

Attention deficit hyperactivity disorder (ADHD) is one of the most commonly diagnosed and treated childhood psychiatric disorders (Scahill 2000). The prevalence of ADHD in children and adolescents is estimated to be 3% to 5% (Polanczyk 2007), depending on the classification system used, with boys two to four times more likely to be diagnosed than girls (Schmidt 2009). Individuals with ADHD exhibit difficulty with attentional and cognitive functions such as solving problems, planning, orienting, maintaining flexibility, sustaining attention, inhibiting response and sustaining a working memory (Pasini 2007; Sergeant 2003). They also have difficulty handling affective features such as motivational delay and mood dysregulation (Castellanos 2006; Nigg 2005; Schmidt 2009).

Diagnosis of ADHD is confirmed through recognition of excessive inattention, hyperactivity and impulsivity in a child, before 12 years of age, that impair his or her functioning or development (APA 2013; WHO 1992). This diagnosis may be based on 18 symptoms indicative of inattention, hyperactivity and impulsivity, according to the principal diagnostic classification systems ‐ International Classification of Diseases, 10ͭth Revision (ICD‐10; WHO 1992), and the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM‐5; APA 2013). The criteria of the ICD‐10 and the DSM–5 require that inattention, hyperactivity and impulsivity are pervasive, that is, are seen in a range of situations for at least six months and are present before the age of six (ICD‐10; WHO 1992) or 12 years (DSM‐5; APA 2013), and that some impairment resulting from these symptoms is observed in two or more settings. Clinically significant impairment in social, academic or occupational functioning must also be evident (APA 1994; APA 2000; APA 2013; WHO 1992). The Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM‐IV; APA 1994), provides three different subtypes to identify and classify particular symptoms, namely 'predominantly inattentive type', 'predominantly hyperactive‐impulsive type' or 'combined type', the last of which presents with both hyperactive‐impulsive and inattentive symptoms (Willcut 2012).

ADHD is seen increasingly as a developmental psychiatric disorder that extends into adulthood and occurs with high heterogeneity and comorbidity with other psychiatric disorders (Schmidt 2009). Comorbid disorders are common in ADHD. The Multimodal Treatment of Attention Deficit Hyperactivity Disorder (MTA) trial identified one or more comorbid disorders in almost 40% of participants (MTA 1999). These included oppositional defiant disorder, conduct disorder, depression, anxiety, tics, learning difficulties and cognitive deficits (Jensen 2001; Kadesjö 2001).

Rising rates of ADHD diagnosis, possible harm to children resulting from drug treatment (Zito 2000), and variation in prevalence estimates are matters of increasing concern (Moffit 2007; Polanczyk 2014). The need for a validated diagnostic test to confirm the clinical diagnosis of ADHD has given rise to debate about its validity as a diagnosis (Timimi 2004). Professional and national bodies have developed guidelines on assessment, diagnosis and treatment of ADHD in an attempt to ensure that high standards are maintained in diagnostic and therapeutic practice (American Academy of Pediatrics 2011; CADDRA 2011; NICE 2009; Pliszka 2007a; SIGN 2009). Psychosocial interventions, such as parent management training, are recommended in the first instance for younger children and for those with mild to moderate symptoms (American Academy of Pediatrics 2011; NICE 2009; Pliszka 2007a), and stimulants (given alone or in combination with psychosocial interventions) are recommended for children with more severe ADHD (American Academy of Pediatrics 2011; CADDRA 2011; NICE 2009).

Description of the intervention

Methylphenidate, dexamphetamine and atomoxetine (a non‐stimulant selective noradrenaline reuptake inhibitor) are recommended medical treatments for children and adolescents with ADHD (Greenhill 2006b; NICE 2009). Globally, methylphenidate, the drug of choice, has been used for longer than 50 years for the treatment of children with ADHD (Kadesjö 2002; NICE 2009). Research suggests that the combination of behavioural therapy (e.g. behavioural parent training, school consultation, direct contingency management) and pharmacotherapy might benefit children with ADHD (Gilmore 2001; MTA 1999).

Methylphenidate is approved for the treatment of individuals with ADHD and narcolepsy (Kanjwal 2012). Pharmacological treatment of children and adolescents with ADHD is reported to have a beneficial effect on the major symptoms of hyperactivity, impulsivity and inattention. The dosage of the intervention can vary significantly between children, with some responding to relatively low dosages and others requiring larger doses to achieve the same effect (Stevenson 1989). Therefore, it is important that the dose of methylphenidate is titrated to an optimal level that maximises therapeutic benefits while producing minimal adverse events. The dose can range from 5 mg to 60 mg methylphenidate administered two to three times daily (Pliszka 2007a). Furthermore, treatment with methylphenidate should not be provided without pause. Medication‐free periods are recommended to reassess effects of treatment on symptoms (Kidd 2000; NICE 2009).

How the intervention might work

It is presumed that the effects of methylphenidate on ADHD symptoms are related to its effects on dopaminergic and noradrenergic neurotransmissions within the central nervous system (CNS) (Engert 2008). Methylphenidate acts by inhibiting catecholamine reuptake, primarily as a dopamine‐norepinephrine reuptake inhibitor, modulating levels of dopamine and to a lesser extent levels of norepinephrine. Methylphenidate binds to and blocks dopamine and norepinephrine transporters (Heal 2006; Iversen 2006), and increased concentrations of dopamine and norepinephrine in the synaptic cleft lead to escalated neurotransmission.

The bioavailability of oral methylphenidate is 11% to 52%. Instant‐release methylphenidate has a duration of action of around 2 to 4 hours, and sustained‐release and extended‐release formulations of methylphenidate have a duration of action of 3 to 8 hours and 8 to 12 hours, respectively (Kimko 1999). Methylphenidate is thought to activate self regulated control processes to ameliorate what are believed to be the core neurofunctional problems of ADHD (Barkley 1977a; Schulz 2012; Solanto 1998). Evidence suggests that symptom control is strongly related to functional improvement (Biederman 2003b; Cox 2004a; Swanson 2004a).

Studies indicate that methylphenidate is effective for treating both the core symptoms of ADHD (inattention, hyperactivity and impulsivity), and aggression (Connor 2002), with the result that children can manage their impulsivity better (Barkley 1981; Barkley 1989a; Shaw 2012). However, a child or adolescent may become less responsive to methylphenidate. In a supplementary analysis of the trial 'A Comparison of Methylphenidates in an Analog Classroom Setting' (COMACS), investigators found that girls had a superior response to methylphenidate (Sonuga‐Barke 2007; Swanson 2004a). In addition, Barkley 1991b noted differences in response to methylphenidate between ADHD inattentive and combined subtypes: children with the inattentive subtype were judged to have a less favourable response to methylphenidate than those diagnosed with the combined subtype. Furthermore, debate continues regarding whether it is valid to diagnose pre‐school children with ADHD, and whether methylphenidate is efficacious and safe for use by pre‐school children with ADHD (Greenhill 2006b).

Why it is important to do this review

Over the past 15 years, several published systematic reviews have investigated the efficacy of methylphenidate for ADHD (with or without meta‐analysis). Fifteen reviews have pooled results on methylphenidate treatment for children and adolescents with ADHD (Bloch 2009; Charach 2011; Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Faraone 2010; Hanwella 2011; Kambeitz 2014; King 2006; Maia 2014; Punja 2013; Reichow 2013; Schachter 2001; Van der Oord 2008). However, none of these were conducted as Cochrane systematic reviews, thus most (Bloch 2009; Charach 2011; Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Faraone 2010;Kambeitz 2014; King 2006; Schachter 2001; Van der Oord 2008) did not adhere to the guidelines of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) nor to the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines (Liberati 2009; Moher 2015), and for none of these reviews was a peer‐reviewed protocol published before the analyses were conducted. Thirteen did not undertake subgroup analyses examining the effects of comorbidity on treatment effects (Bloch 2009; Charach 2011; Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Faraone 2010; Hanwella 2011; Kambeitz 2014; Maia 2014; Punja 2013; Schachter 2001; Van der Oord 2008); some did not control for treatment effects by ADHD subtype (Bloch 2009; Charach 2013; Faraone 2002; Hanwella 2011; Kambeitz 2014; King 2006; Maia 2014; Punja 2013; Schachter 2001; Van der Oord 2008); and others did not consider effects according to dose of methylphenidate (Charach 2011; Charach 2013; Faraone 2006; Faraone 2009; Hanwella 2011; Kambeitz 2014; Maia 2014; Punja 2013; Reichow 2013; Van der Oord 2008). In addition, as regards the outcomes, most meta‐analyses pooled data from parents, teachers and independent assessors (Bloch 2009; Charach 2011; Charach 2013; Hanwella 2011; Kambeitz 2014; King 2006; Reichow 2013), and did not separate outcome measures for inattention and hyperactivity/impulsivity (Bloch 2009; Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Hanwella 2011; Kambeitz 2014; Van der Oord 2008). Moreover, most previous reviews investigated only the effects of methylphenidate on symptoms of ADHD; review authors did not present data on spontaneous adverse events (Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Faraone 2010; Hanwella 2011; Kambeitz 2014; Maia 2014; Van der Oord 2008), nor on adverse events as measured by rating scales (Bloch 2009; Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Faraone 2010; Hanwella 2011; Kambeitz 2014; King 2006; Maia 2014; Punja 2013; Reichow 2013; Schachter 2001; Van der Oord 2008), and they did not try to explain why such information was not provided. Finally, these reviews did not systematically assess risk of random errors, risk of bias and trial quality (Bloch 2009; Charach 2011; Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Faraone 2010; Hanwella 2011; Kambeitz 2014; King 2006; Van der Oord 2008). These shortcomings plus other methodological limitations, including potential bias in excluding non‐English publications (Charach 2013; Faraone 2010; Punja 2013; Van der Oord 2008), and in not searching the principal major international databases nor reporting search terms clearly (Bloch 2009; Faraone 2002; Kambeitz 2014; Reichow 2013), may have compromised data collection, consequently calling the results of these previous meta‐analyses into question.

Given mounting concerns regarding increasing use of methylphenidate in children younger than six years of age, it is vital that researchers explore risk versus benefit of treatment in this younger population (US FDA 2011). Although stimulant medications may have a favourable risk‐benefit profile, they might carry potential risks of both serious and non‐serious adverse events. Adverse events most often associated with methylphenidate include headache, sleep problems, tiredness and decreased appetite. Serious adverse reactions, such as psychotic symptoms and mood disorders, affect about 3% of children treated with methylphenidate (Block 1998; Cherland 1999; MTA 1999; NICE 2009; Pliszka 1998). Some studies indicate that methylphenidate can decrease children's height and weight (Schachar 1997b; Swanson 2004a; Swanson 2009). Other studies report sudden death though it remains unclear whether these deaths are directly related to methylphenidate treatment (Vitiello 2008); researchers are currently exploring the link between sudden death and treatment with methylphenidate (US FDA 2011).

Given the limitations of existing reviews, we conducted a systematic review of the benefits and harms of methylphenidate for children and adolescents with ADHD while adhering to the recommendations of The Cochrane Collaboration (Higgins 2011) and to PRISMA guidelines (Liberati 2009; Moher 2015).

This systematic review focuses on the beneficial and harmful effects of methylphenidate in randomised controlled trials and is the first of two systematic reviews. A second review focused on harms reported in non‐randomised studies is under way (Storebø in press).

Objectives

To assess the beneficial and harmful effects of methylphenidate for children and adolescents with ADHD.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) of methylphenidate for the treatment of children and adolescents with ADHD. We included trials irrespective of language, publication year, publication type or publication status.

Types of participants

Children and adolescents aged 18 years and younger with a diagnosis of ADHD, according to the Diagnostic and Statistical Manual of Mental Disorders (DSM), Third Edition (DSM‐III; APA 1980), Third Edition Revised (DSM‐III‐R; APA 1987), Fourth Edition (DSM‐IV; APA 1994), and Fifth Edition (DSM‐5; APA 2013), or with a diagnosis of hyperkinetic disorders according to the International Statistical Classification of Diseases and Related Health Problems, Ninth Revision (ICD‐9), and 10ͭth Revision (ICD‐10; WHO 1992). We included participants with ADHD with or without comorbid conditions such as conduct or oppositional disorders, tics, depression, attachment disorders or anxiety disorders. Trials eligible for inclusion were those in which at least 75% of participants were aged 18 years or younger, and the mean age of the trial population was 18 years or younger. We also required that at least 75% of participants had a normal intellectual quotient (IQ > 70).

Types of interventions

Methylphenidate, administered at any dosage or in any formulation as part of any medical treatment regimen, compared with placebo or with no intervention.

We permitted cointerventions if intervention groups received cointerventions similarly. Consequently, we did not permit polypharmacy as a cointervention in one of the groups.

Types of outcome measures

Primary outcomes

ADHD symptoms (attention, hyperactivity and impulsivity), measured over the short term (within six months) and over the long term (longer than six months) by psychometric instruments or by observations of behaviour, using, for example, Conners' Teacher Rating Scales (Conners 1998a; Conners 2008). Raters could be teachers, independent assessors or parents. We chose to report the results of teacher‐rated outcomes as primary outcomes (see Results).
Numbers of serious adverse events. We defined a serious adverse event as any event that led to death, was life‐threatening, required in‐patient hospitalisation or prolongation of existing hospitalisation or resulted in persistent or significant disability, or as any important medical event that may have jeopardised the patient's life or that required intervention for prevention. We considered all other adverse events as non‐serious (ICH 1996).

Secondary outcomes

Non‐serious adverse events. We assessed all adverse events, including, for example, growth retardation and cardiological, neurological and gastrointestinal events, as described in ICH (International Conference on Harmonisation of technical requirements for registration of pharmaceuticals for human use) Harmonised Tripartite Guideline. Guideline for Good Clinical Practice E6(R1) (ICH 1996).
General behaviour in school and at home, as rated by psychometric instruments such as the Child Behaviour Checklist (CBCL; Achenbach 1991), measured over the short term (within six months) and over the long term (longer than six months). Raters could be teachers, independent assessors or parents. We chose to report the results of teacher‐rated outcomes as primary outcomes (see Results).
Quality of life, as measured by psychometric instruments such as the Child Health Questionnaire (CHQ; Landgraf 1998). Raters could be teachers, independent assessors or parents.

Search methods for identification of studies

Electronic searches

We ran the first literature searches in October 2011 and updated them in November 2012, March 2014 and most recently between 26 February and 10 March 2015. We searched the following sources.

Cochrane Central Register of Controlled Trials (CENTRAL; 2015, Issue 2; part of The Cochrane Library, which includes the Specialised Register of the Cochrane Developmental, Psychosocial and Learning Problems Group), searched 10 March 2015.
Ovid MEDLINE (1948 to current), searched 10 March 2015.
EMBASE (Ovid; 1980 to current), searched 10 March 2015.
CINAHL (Cumulative Index to Nursing and Allied Health Literature; EBSCOhost; 1980 to current), searched 10 March 2015.
PsycINFO (Ovid; 1806 to current), searched 10 March 2015.
Conference Proceedings Citation Index ‐ Science (CPCI‐S) and Conference Proceedings Citation Index ‐ Social Science & Humanities (CPCI‐SS&H) (Web of Science; 1990 to 17 March 2015), searched 19 March 2015.
ClinicalTrials.gov (1999 to current), searched 26 February 2015.
World Health Organization International Clinical Trials Registry Platform (WHO ICTRP; who.int/ictrp/en; 1999 to current), searched 26 February 2015.

The search strategy for each database is shown in Appendix 1. We searched CENTRAL, Ovid MEDLINE, EMBASE and PsycINFO using two separate search strategies (one for efficacy of methylphenidate and one for adverse events of methylphenidate). For the remaining databases, we used one broad strategy to capture trials on efficacy and trials on adverse events. To overcome poor indexing and abstracting, we listed individual brand names within the search strategies. We did not limit searches by language, year of publication or type or status of publication. We sought translation of relevant sections of non‐English language articles.

Searching other resources

To find additional relevant trials not identified by electronic searches, we checked the bibliographic references of identified review articles, meta‐analyses and a selection of included trials. Furthermore, we requested published and unpublished data from pharmaceutical companies manufacturing methylphenidate, including Shire, Medice (represented in Denmark by HB Pharma), Janssen‐Cilag and Novartis (Appendix 2). We also requested data from unpublished trials from experts in the field.

Data collection and analysis

We conducted this review according to the recommendations provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and performed analyses using Review Manager (RevMan), Version 5.3 ‐ the statistical software of The Cochrane Collaboration (Review Manager 2014).

Selection of studies

Eleven review authors (ER, FLM, HBK, KBR, MH, MS, OJS, TDN, SR, TK and TB) worked together in groups of two and independently screened titles and abstracts of all publications obtained from the literature searches; uncertainty or disagreements were resolved by consensus or by consultation with a third review author. We obtained full‐text articles of trials presenting potentially relevant data and assessed them against our listed inclusion criteria. We discussed disagreements, and if agreement or consensus could not be reached, we consulted a third review author (OJS).

Data extraction and management

Working together in groups of two, 17 review authors extracted data (CGr, CRMM, DGa, DGi, ER, ES, FLM, HK, KBR, MH, MJR, MS, MZ, OJS, RK, SR and TDN). We resolved disagreements by discussion and we used an arbiter if required. When data were incomplete, or when data provided in published trial reports were unclear, we contacted trial authors to ask for clarification of missing information. We contacted the authors of all cross‐over trials to obtain first period data on ADHD symptoms.

We developed data extraction forms a priori (after performing data extraction pilots, we updated these forms to accommodate extraction of more detailed data and to facilitate standardised approaches to data extraction among review authors). These extraction forms were used by all data extractors (see Appendix 3; Appendix 4).

Six review authors (CRMM, FLM, MH, HK, ER and OJS) entered data into RevMan (Review Manager 2014).

Assessment of risk of bias in included studies

For each included trial, data extractors independently evaluated risk of bias domains (listed below), resolving disagreements by discussion. For each domain, we assigned each trial to one of the following three categories: low risk of bias, unclear (uncertain) risk of bias or high risk of bias, according to guidelines provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Given the risk of overestimation of beneficial intervention effects and underestimation of harmful intervention effects in RCTs with unclear or inadequate methodological quality (Kjaergard 2001; Lundh 2012; Moher 1998; Savović 2012a; Savović 2012b; Schulz 1995; Wood 2008), we assessed the influence of risk of bias on our results (see Subgroup analysis and investigation of heterogeneity). Risk of bias components were as follows.

Random sequence generation

Low risk of bias. The method used was adequate (e.g. computer‐generated random numbers, table of random numbers) or was unlikely to introduce selection bias.
Unclear risk of bias. Information was insufficient for assessment of whether the method used could introduce selection bias.
High risk of bias. The method used was likely to introduce bias.

Allocation concealment

Low risk of bias. The method used (e.g. central allocation) was unlikely to bias allocation to groups.
Unclear risk of bias. Information was insufficient for assessment of whether the method used could bias allocation to groups.
High risk of bias. The method used (e.g. open random allocation schedule) could bias allocation to groups.

Blinding of participants and personnel

Low risk of bias. The method of blinding was described sufficiently and blinding was conducted in a satisfactory way that was unlikely to introduce performance bias.
Unclear risk of bias. Information was insufficient for assessment of whether adequate blinding was used and whether it was likely to introduce performance bias.
High risk of bias. No blinding or incomplete blinding was described.

Blinding of outcome assessment

Low risk of bias. The method of blinding was described and blinding was conducted in a satisfactory way that was unlikely to introduce detection bias.
Unclear risk of bias. Information was insufficient for assessment of whether the type of blinding used was likely to bias the estimate of effect.
High risk of bias. No blinding or incomplete blinding was described.

Incomplete outcome data

Low risk of bias. Underlying reasons for missing data probably would not affect outcome measurement regarding effects of methylphenidate, as all missing data can be considered as missing at random or all data were reported.
Unclear risk of bias. Information was insufficient for assessment of whether missing data or the method used to handle missing data was likely to bias the estimate of effect.
High risk of bias. The crude estimate of effects could be biased given the reasons for the missing data.

Selective reporting

Low risk of bias. The trial protocol was available and all pre‐specified outcomes of interest were reported.
Unclear risk of bias. Information was insufficient for assessment of whether selective outcome reporting could have occurred.
High risk of bias. Not all of the primary outcomes specified beforehand were reported or participants were excluded after randomisation (selection bias).

Vested interest bias

Low risk of bias. The trial was not funded by any parties that might be considered to have a conflict of interest (e.g. a manufacturer of methylphenidate).
Unclear risk of bias. The source of funding was not clear.
High risk of bias. The trial was funded by parties that might have had a conflict of interest (e.g. a manufacturer of methylphenidate) or potential conflicts of interest were reported by trial authors.

Other potential sources of bias

Low risk of bias. The trial appeared to be free of other sources of bias.
Unclear risk of bias. Information was inadequate for assessment of other possible sources of bias.
High risk of bias. Other sources of bias were identified.

Seven of the above domains are specified in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We added an eighth domain ‐ vested interest (Storebø 2012). Andreas Lundh and colleagues have shown that there are many subtle mechanisms through which sponsorship and conflict of interest may influence intervention effects on outcomes. The AMSTAR (A MeaSurement Tool to Assess systematic Reviews) tool for methodological quality assessment of systematic reviews also includes funding and conflicts of interest as a domain (amstar.ca). For more information, please see editorials by Bero 2013 and Sterne 2013, and the commentary by Gøtzsche 2015.

We defined low risk of bias trials as trials that had low risk of bias in all domains. We considered trials with one or more unclear or high risk of bias domains as trials with high risk of bias.

Some trials excluded methylphenidate non‐responders, placebo responders and/or participants who had adverse events due to the medication. We did not consider these trials to be at risk of bias as participants were excluded before randomisation. However, to identify whether this 'cohort selection bias of all participants' had an effect on estimates of effectiveness, we conducted subgroup analyses based on these criteria (Subgroup analysis and investigation of heterogeneity).

Measures of treatment effect

The treatment effect was defined as an improvement in ADHD symptoms, general behaviour and quality of life.

Dichotomous data

We summarised dichotomous data as risk ratios (RRs) with 95% confidence intervals (CIs). We calculated the risk difference (RD).

Continuous data

If the same measure of a given continuous outcome was used in all trials in a meta‐analysis, we calculated mean differences (MDs) with 95% CIs. If different measures were used, we calculated standardised mean differences (SMDs) with 95% CIs. If trials did not report means and standard deviations but did report other values (e.g. t‐tests, P values), we transformed these into standard deviations.

For primary analyses of teacher‐rated ADHD symptoms, teacher‐rated general behaviour and quality of life, we transformed SMDs into MDs on the following scales to assess whether results exceeded the minimum clinically important difference: ADHD Rating Scale (ADHD‐RS; DuPaul 1991a), Conners' Global Index (CGI; Conners 1998a) and Child Health Questionnaire (CHQ; Landgraf 1998). We identified a minimal clinically relevant difference (MIREDIF) of 6.6 points on the ADHD‐RS, ranging from 0 to 72 points, based on a trial by Zhang 2005, and a MIREDIF of 7.0 points on the CHQ, ranging from 0 to 100 points, based on a trial by Rentz 2005. We could find no references describing a MIREDIF on the CGI (range 0 to 30 points).

Unit of analysis issues

Many ADHD trials use cross‐over methods. We aimed to obtain data from the first period of these trials and to pool these data with data from parallel‐group trials, as they are similar (Curtin 2002). We requested these data from trial authors if they were not available in the published report. When we were not able to obtain first‐period data from cross‐over trials, we established another group comprising only endpoint data. Our original intention was to adjust for the effect of the unit of analysis error in cross‐over trials by conducting a covariate analysis, but data were insufficient for this. As cross‐over trials are more prone to bias from carry‐over effects, period effects and unit of analysis errors (Curtin 2002), we conducted a subgroup analysis to compare these two groups. We tested for the possibility of a carry‐over effect and a period effect (Subgroup analysis and investigation of heterogeneity). We found similar treatment effects in the two groups and no significant subgroup differences. However, we noted considerable heterogeneity, and so we presented the results of the analyses separately (Effects of interventions).

For dichotomous outcomes in cross‐over trials, we were unable to adjust the variance to account for the correlation coefficient as advised by Elbourne 2002 due to insufficient information, or to estimate the RR using the marginal probabilities as recommended by Becker 1993. Consequently, we used endpoint data for estimating RRs. As these effect estimates are prone to potential bias, we performed a sensitivity analysis by removing these trials to assess the robustness of the pooled results.

We used endpoint data when these were reported or could be obtained from trial authors. However, when RCTs reported only 'change scores', we pooled these with endpoint scores. We explored whether inclusion of change data affected outcomes by performing sensitivity analyses (see Sensitivity analysis).

Dealing with missing data

We obtained missing data by contacting trial authors. When we were not able to obtain missing data, we conducted analyses using available (incomplete) data. Although some trials reported that intention‐to‐treat (ITT) analyses were used, data were missing for many primary outcomes (Hollis 1999). We could not use 'best‐case scenario' and 'worst‐case scenario' analyses on our assessment of benefit as there were no dichotomous outcomes. Also, we decided not to use 'best‐case scenario' and 'worst‐case scenario' analyses in our assessment of adverse events because we evaluated these analyses to be imprecise due to the high number of trials not reporting adverse events, and due to the high number of dropouts in the trials reporting adverse events.

Assessment of heterogeneity

We identified three types of heterogeneity: clinical, methodological and statistical. Clinical heterogeneity reflects variability among participants, interventions and outcomes of trials. Methodological heterogeneity reflects variability in the design of trials, and statistical heterogeneity reflects differences in effect estimates between trials. We assessed clinical heterogeneity by comparing differences in trial populations, interventions and outcomes, and we evaluated methodological heterogeneity by comparing the design of trials. We identified potential reasons for clinical and methodological heterogeneity by examining individual trial characteristics and subgroups. Furthermore, we observed statistical heterogeneity in trials both by visual inspection of a forest plot and by use of a standard Chi² value with a significance level of α (alpha) = 0.1. We used I² to quantify inconsistency, with I² values between 30% and 60% indicating a moderate level of heterogeneity (Higgins 2011).

Assessment of reporting biases

We followed the recommendations for reporting bias, including publication bias and outcome reporting bias, provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We drew funnel plots (estimated differences in treatment effects against their standard error) and performed Egger's statistical test for small‐study effects; asymmetry could be due to publication bias or could indicate genuine heterogeneity between small and large trials (Higgins 2011). We did not visually inspect the funnel plot if fewer than 10 trials were included in the meta‐analysis, in accordance with the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Data synthesis

We performed statistical analyses as recommended by the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We synthesised data statistically when clinical heterogeneity was not excessive (e.g. variability in participant characteristics was minimal). Furthermore, we included and analysed trials undertaken in any configuration or setting (e.g. in groups, at home, at a centre).

We used the inverse variance method, which gives greater weight to larger trials, to generate more precise estimates. For some adverse events we combined dichotomous data and continuous data using the generic inverse variance method. We synthesised data using change from baseline scores or endpoint data. If data were available for several intervals, we used the longest period assessed. We used both the fixed‐effect and the random‐effects models in all meta‐analyses. However, we reported the results of the random‐effects model, which gives greater weight to smaller trials; statistical significance did not change when we applied a fixed‐effect model (Jakobsen 2014). We performed separate meta‐analyses for three types of raters (teachers, independent assessors, parents), for data from parallel‐group trials combined with data from the first period of cross‐over trials and for endpoint data derived from cross‐over trials.

ADHD symptom scales describe the severity of inattention, hyperactivity and impulsivity at home and at school; high scores indicate severe ADHD. We judged that, in spite of the diversity of psychometric instruments, they could be used for our outcomes, and we integrated different types of scales into the analyses. We used MDs if the same measure was used in all trials and SMDs when different outcome measures were used for the same construct in different trials.

When separate measures of hyperactivity, impulsivity and inattention were available, we used combined scores. When symptoms were measured and reported at different time points during the day (after ingestion of medication or placebo), we used the time point closest to noon.

Two outcomes ‐ ADHD symptoms and general behaviour ‐ were measured by three types of raters: teachers, independent assessors and parents. We considered these data as showing different outcomes. We presented the results of teacher‐rated measures as the primary outcome because symptoms of ADHD are more readily detectable in the school setting (Hartman 2007).

When trials reported data for different doses, we used data for the dose that we defined as moderate/high (> 20 mg/d) in our primary analyses.

We summarised adverse event data as RRs with 95% CIs for dichotomous outcomes. For the purposes of this review, we used only dichotomous outcomes that reflected the number of participants affected by the event per the total number of participants.

Heterogeneity‐adjusted required information size and Trial Sequential Analysis

Trial Sequential Analysis is a method that combines the required information size (RIS) for a meta‐analysis with the threshold for statistical significance (Brok 2008; Brok 2009; Thorlund 2009; Wetterslev 2008) to quantify the statistical reliability of data in a cumulative meta‐analysis, with P value thresholds controlled for sparse data and repetitive testing of accumulating data (Brok 2008; Brok 2009; Thorlund 2009; Wetterslev 2008).

Comparable with the a priori sample size estimation provided in a single RCT, a meta‐analysis should include an RIS at least as large as the sample size of an adequately powered single trial to reduce the risk of random error. A Trial Sequential Analysis calculates the RIS in a meta‐analysis and provides trial sequential monitoring boundaries with an adjusted P value.

When new trials emerge, multiple analyses of accumulating data lead to repeated significant testing and hence introduce multiplicity. Use of conventional P values exacerbates the risk of random error (Berkey 1996; Lau 1995). Meta‐analyses not reaching the RIS are analysed with trial sequential monitoring boundaries analogous to interim monitoring boundaries in a single trial (Wetterslev 2008). This approach will be crucial in coming updates of this review.

If a Trial Sequential Analysis does not result in significant findings (no Z‐curve crossing the trial sequential monitoring boundaries) before the RIS has been reached, the conclusion should be that more trials are needed to reject or accept an intervention effect that was used to calculate the required sample size, or when the cumulated Z‐curve enters the futility area, the anticipated intervention effect should be rejected.

For calculations with the Trial Sequential Analysis programme, we included trials with zero events by substituting 0.5 for zero (CTU 2011; Thorlund 2011).

For the outcomes 'total serious adverse events' and 'total non‐serious adverse events', we calculated the a priori diversity‐adjusted required information size (DARIS; i.e. number of participants in the meta‐analysis required to detect or reject a specific intervention effect) and performed a Trial Sequential Analysis for these outcomes based on the following assumptions (Brok 2008; Brok 2009; Thorlund 2009; Wetterslev 2008; Wetterslev 2009).

Proportion of participants in the control group with adverse events.
Relative risk reduction of 20% (25% on 'total serious adverse events').
Type I error of 5%.
Type II error of 20%.
Observed diversity of the meta‐analysis.

'Summary of findings' tables

We used the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) approach to construct a 'Summary of findings' table in which to document all review outcomes (GRADEpro 2014). The GRADE approach appraises the quality of a body of evidence based on the extent to which one can be confident that an estimate of effect or association reflects the item being assessed. Considerations are due to within‐trial risk of bias; directness of the evidence; heterogeneity of the data; precision of effect estimates; and risk of publication bias (Andrews 2013a; Andrews 2013b; Balshem 2011; Brunetti 2013; Guyatt 2011a; Guyatt 2011b; Guyatt 2011c; Guyatt 2011d; Guyatt 2011e; Guyatt 2011f; Guyatt 2011g; Guyatt 2011h; Guyatt 2013a; Guyatt 2013b; Guyatt 2013c; Mustafa 2013). When possible, that is when the MD or the RR was available, we used the results from the Trial Sequential Analysis as the rating for imprecision (Jakobsen 2014). We reported two primary outcomes (teacher‐rated ADHD symptoms and serious adverse events) and three secondary outcomes (non‐serious adverse events, teacher‐rated general behaviour, and quality of life) in summary of findings Table for the main comparison.

Subgroup analysis and investigation of heterogeneity

We performed the following subgroup analyses of teacher‐rated ADHD symptoms (primary outcome) to test the robustness of this estimate.

Age of participants (trials with participants aged 2 to 6 years versus those with participants aged 7 to 11 years versus those with participants aged 12 to 18 years).
Sex (boys versus girls).
Comorbidity (children with comorbid disorders versus children without comorbid disorders).
Type of ADHD (participants with predominantly inattentive subtype versus participants with predominantly combined subtype).

After learning about other factors that may affect the impact of methylphenidate, we performed the following additional post hoc subgroup analyses on teacher‐rated ADHD symptoms to test the robustness of the estimate.

Types of scales (e.g. Conners' Teacher Rating Scale (CTRS; Conners 1998a) versus Strengths and Weaknesses of ADHD Symptoms and Normal Behavior (SWAN) Scale (Swanson 2006).
Dose of methylphenidate (low dose (≤ 20 mg/d or ≤ 0.6 mg/kg/d) versus moderate/high dose (> 20 mg/d or > 0.6 mg/kg/d)).
Duration of treatment (short‐term trials (≤ six months) versus long‐term trials (> six months)).
Trial design (parallel‐group trials versus cross‐over trials (first period data and endpoint data)).
Medication status before randomisation (medication naive (> 80% of included participants were medication naive) versus not medication naive (< 20% of included participants were medication naive)).
Risk of bias (trials with low risk of bias versus trials with high risk of bias).
Cohort selection bias (trials with cohort selection bias of all participants versus trials without cohort selection bias).

Sensitivity analysis

We conducted sensitivity analyses to determine whether findings were sensitive to the following.

Decisions made during the review process such as our assessment of clinical heterogeneity (listed below).
Combined 'change scores' and 'endpoint data' in the meta‐analyses.

No sufficiently well‐designed method has been used to combine the results of trials with high risk of bias and trials with low risk of bias (Higgins 2011). We performed sensitivity analyses by grouping together trials with similar classifications of bias, as described above, and investigated the impact on intervention effects.

We excluded the following trials from the sensitivity analyses.

IQ < 70: Oesterheld 1998; Pearson 2013; Smith 1998; Taylor 1987.
Change scores: Carlson 2007; Findling 2007; Newcorn 2008; Palumbo 2008; Tucker 2009.
Older than 18 years of age: Green 2011; Szobot 2008.

Results

Description of studies

For more information, please see Characteristics of included studies, Characteristics of excluded studies, Characteristics of studies awaiting classification and Characteristics of ongoing studies.

Results of the search

We carried out electronic searches over four periods. Searches up to October 2011 produced 6358 records after duplicates were removed (10,249 initial records). Searches up to November 2012 produced an additional 713 records after duplicates were removed (1080 initial records). Searches up to March 2014 produced an additional 654 records after duplicates were removed (1274 initial records). Searches up to February 2015 (ClinicalTrials.gov and ICTRP) and March 2015 (all remaining databases) produced an additional 1178 records after duplicates were removed (1460 initial records). We identified 368 additional publications by reading the reference lists of included articles and reviews, and by corresponding with authors of relevant literature and with pharmaceutical companies. We contacted the authors of 161 trials twice for supplemental information and data; 92 responded.

From the 9271 screened records, we excluded 7811 clearly irrelevant reports on the basis of title and abstract. We retrieved the full texts of the remaining 1460 reports, which we assessed for eligibility. Of the retrieved publications, 69 were written in languages other than English, including Danish, Norwegian, German, Dutch, French, Italian, Spanish, Portuguese, Czech, Turkish, Farsi, Japanese and Chinese; we had these articles translated to assess their eligibility. We excluded 691 full‐text reports (please see subsection on Excluded studies and Characteristics of excluded studies tables), and identified 2 trials (from 2 reports) as awaiting classification (see Characteristics of studies awaiting classification), and 5 ongoing trials (from 6 reports; see Characteristics of ongoing studies). We included 761 reports, of which 449 described 185 RCTs and 312 described 243 non‐randomised studies. (We are currently assessing and synthesising data from the 243 non‐randomised studies. These findings will be published in the forthcoming review on adverse events in observational studies (Storebø in press). This review focuses on RCTs only (see Figure 1; Moher 2009). For more information on these trials, please see the Characteristics of included studies tables.

Figure 1

Flow chart.

Included studies

We included 185 trials (from 449 reports) in this review (Figure 1). Of these, 38 are parallel‐group trials (from 138 reports) and 147 are cross‐over trials (from 311 reports). One study ‐ Kollins 2006 (PATS) ‐ includes a parallel‐group trial and a cross‐over trial.

Included parallel‐group trials

We included 38 parallel‐group trials described in 138 reports.

Duration

Most trials (n = 34) were short‐term (< six months in duration). Only three were long‐term trials (conducted for ≥ six months; Jensen 1999 (MTA); Perez‐Alvarez 2009; Schachar 1997a). The duration of one trial was unclear (Tucker 2009). Average trial duration was 74.8 days (range 1 to 425 days).

Location

Twenty‐one of the 38 trials were conducted in the USA. Two trials were conducted in the USA and Canada (Biederman 2003; Jensen 1999 (MTA)); one in the USA, Canada and Australia (Findling 2006); and one in USA, Canada, Taiwan, Mexico and Puerto Rico (Lin 2014). Two trials each were conducted in Brazil (Martins 2004; Szobot 2004), Canada (Butter 1983; Schachar 1997a), Israel (Green 2011; Jacobi‐Polishook 2009) and Germany (Coghill 2013; Lehmkuhl 2002), although Coghill 2013 also included participants in Sweden. Single trials were conducted in New Zealand (Heriot 2008), Norway (Duric 2012) and the Netherlands (Van der Meere 1999a). The locations of two trials were not clear (Firestone 1981; Tourette's Syndrome Study Group 2002).

Setting

All but the following three trials were conducted in out‐patient clinics: one was carried out in a naturalistic classroom setting (Greenhill 2006), one in a research unit at a hospital (Schachar 1997a), and one provided no information on setting (Brown 1985).

Participants

The 38 trials included a total of 5111 participants (with the percentage of girls ranging from 0% to 50% (mean 20.7%), equivalent to a boy‐to‐girl ratio of 5:1). All participants were between 3 and 20 years of age (mean 9.7 years).

Twenty‐six trials described the percentage of methylphenidate‐naive participants (range 0% to 100%; mean 58.2%).

Twenty‐three trials described the proportion of participants with combined subtype ADHD (range 25% to 100%; mean 72%); 19 trials reported the proportion of participants with hyperactive subtype (range 0% to 56%; mean 5%) and 20 trials revealed the proportion with inattentive subtype (range 0% to 72.4%; mean 24%).

Five trials excluded children and adolescents with a comorbidity (Findling 2010; Greenhill 2002; Perez‐Alvarez 2009; Tucker 2009; Wigal 2004), but it was not clear whether such participants were included in 11 trials (Arnold 2004; Biederman 2003; Brown 1985; Butter 1983; Duric 2012; Findling 2006; Findling 2008; Firestone 1981; Greenhill 2006; Jacobi‐Polishook 2009; Wilens 2006b). Oppositional defiant disorder was the most common comorbidity (range 8% to 53%; mean 36.3%), followed by conduct disorder (range 2% to 32%; mean 11.6%).

Participants taking other medications were specifically excluded from 10 trials (Green 2011; Heriot 2008; Ialongo 1994; Jacobi‐Polishook 2009; Kollins 2006 (PATS); Martins 2004; Palumbo 2008; Perez‐Alvarez 2009; Schachar 1997a; Wigal 2004), and were permitted in two trials (Lehmkuhl 2002; Van der Meere 1999a). Riggs 2011 allowed inclusion of participants using drugs and alcohol.

Interventions

Twelve trials used extended‐ and modified‐release methylphenidate (Biederman 2003; Childress 2009; Coghill 2013; Green 2011; Greenhill 2002; Greenhill 2006; Lehmkuhl 2002; Lin 2014; Newcorn 2008; Riggs 2011; Tucker 2009; Wilens 2006b). Two trials used immediate‐ and extended‐release methylphenidate (Findling 2006; Wolraich 2001), and two trials used transdermal methylphenidate patches (Findling 2008; Findling 2010). All other trials used immediate‐release methylphenidate.

The method of reporting the dosage of methylphenidate varied considerably between trials, but overall daily dose ranged from 5 mg to 60 mg with a mean reported total daily dose of 28.2 mg/d or 0.73 mg/kg/d. The average dose of any type of modified‐ or extended‐release methylphenidate was 41.0 mg, and the average dose of immediate‐release methylphenidate was 23.1 mg.

Twenty‐eight trials used placebo as control, and 10 trials used no intervention as control (Brown 1985; Connor 2000; Duric 2012; Firestone 1981; Heriot 2008; Jensen 1999 (MTA); Perez‐Alvarez 2009; Riggs 2011; Schachar 1997a; Tucker 2009).

Four trials used clonidine (Connor 2000; Palumbo 2008; Tourette's Syndrome Study Group 2002) or atomoxetine (Carlson 2007) as a cointervention in both intervention and control groups. Three trials used parent training (Firestone 1981; Heriot 2008; Schachar 1997a), two used cognitive‐behavioural therapy (Brown 1985; Riggs 2011) and five used other behavioural therapies (Duric 2012; Horn 1991; Jensen 1999 (MTA); Perez‐Alvarez 2009; Tucker 2009) as cointerventions for intervention and control groups.

Included cross‐over trials

We included 147 cross‐over trials described in 311 reports.

Eighty‐three trials were described in a single publication. Ten trials yielded five or more publications (Brams 2012; Döpfner 2004; Gadow 1995; Grizenko 2012; Kollins 2006 (PATS); Rapport 1987; Sharp 1999; Swanson 2002a; Swanson 2002b; Swanson 2004b). The PATS (Preschool Attention Deficit Hyperactivity Disorder Treatment Study) trial reported the greatest number of publications per trial, with 23 publications (Kollins 2006 (PATS).

Duration

Three cross‐over trials did not report duration (Kelly 1989; Sunohara 1999; Tannock 1993). The remaining cross‐over trials had a duration of less than six months.

Location

A total of 103 trials were carried out in the USA; 21 in Canada; and one in both the USA and Canada (Quinn 2004). Five trials were conducted in Israel (Lufi 1997; Lufi 2007; Moshe 2012; Tirosh 1993a; Tirosh 1993b); five in Germany (Bliznakova 2007; Döpfner 2004; Konrad 2004; Konrad 2005; Schulz 2010); four in the Netherlands (Buitelaar 1995; Flapper 2008; Lijffijt 2006; Overtoom 2003); two each in the UK (Coghill 2007; Taylor 1987), Norway (Ramtvedt 2013; Zeiner 1999) and Brazil (Szobot 2008; Zeni 2009); and one in Australia (Nikles 2006). One trial did not specify the country of origin (Hicks 1985).

Setting

Seventeen trials were completed as a part of summer treatment programmes or summer schools. Five trials were conducted in in‐patient wards (Carlson 1995; Gonzalez‐Heydrich 2010; Kent 1995; Konrad 2005; Solanto 2009), and five in both out‐patient clinics and in‐patient wards (Garfinkel 1983; Hicks 1985; Kaplan 1990; Konrad 2004; Wallander 1987). Twelve trials were conducted in a laboratory classroom setting (Brams 2008; Lopez 2003; Oesterheld 1998; Schachar 2008; Sharp 1999; Silva 2006; Swanson 1998; Swanson 1999; Swanson 2002a; Swanson 2002b; Wigal 2003; Wigal 2014). Eight trials did not report the setting (Ben 2002; Bliznakova 2007; Froehlich 2011; Pliszka 2007; Stoner 1994; Tervo 2002; Ullmann 1985; Urman 1995). All remaining trials were conducted in out‐patient clinics only.

Participants

The 147 cross‐over trials included a total of 7134 participants (range 1 to 430 per trial; mean 48.5). In 131 of these trials, a total of 6597 participants were randomly assigned, 6018 of whom were followed up in analyses. In 13 trials, 537 participants were randomly assigned, but it was unclear how many participants were followed up (Ben 2002; Chronis 2003; Coghill 2007; Douglas 1995; Fitzpatrick 1992a; Gadow 2011; Hale 2011; Leddy 2009; Pelham 1989; Rapport 1987; Shiels 2009; Smithee 1998; Wallander 1987). In two trials, it was unclear how many participants were included, but 191 were followed up in analyses (Epstein 2011; Sharp 1999). In one trial, it was unclear how many participants were included or followed up (Rapport 1987).

Nine trials did not state the ratio of boys to girls (Epstein 2011; Fine 1993; Leddy 2009; Pliszka 1990; Rapport 1987; Samuels 2006; Solanto 2009; Sumner 2010; Wallace 1994), and 30 trials did not specifically comprise girls. In the 108 trials that included participants of both sexes and reported proportions, the percentage of girls ranged from 4% to 50% (mean 21.8%). Participants were between 4 and 21 years of age with an average age of 9.7 years (17 trials did not report average age; however, all of these trials reported age range).

A total of 98 trials described the percentage of methylphenidate‐naive participants included (range 0% to 100%; mean 51%). In 26 trials, all participants were methylphenidate naive. Only participants previously treated with methylphenidate were included in 30 trials.

Sixty‐four trials described the proportion of participants with combined subtype ADHD (range 0% to 100%; mean 69.1%), 63 reported the proportion with hyperactive subtype (range 0% to 100%; mean 12.3%) and 63 reported the proportion with inattentive subtype (range 0% to 72%; mean 20.2%).

Eighty‐eight trials reported the presence of comorbidity. Oppositional defiant disorder was the most commonly reported comorbidity (61 trials; range 1.4% to 84%; mean 42%), followed by conduct disorder (49 trials; range 0% to 100%; mean 24%). Four trials reported participants with Tourette's syndrome (range 12% to 100%; mean 75%; Castellanos 1997; Gadow 2007; Gadow 2011; Kent 1999). One trial included only participants with epilepsy (Gonzalez‐Heydrich 2010), one included only participants with cerebral palsy (Symons 2007), and one included only participants with bipolar disorder (Findling 2007). Fourteen trials reported that participants had no comorbidity (Borcherding 1990; Brams 2012; Garfinkel 1983; Lufi 1997; Moshe 2012; Quinn 2004; Schachar 2008; Sumner 2010; Swanson 1998; Swanson 2002a; Tirosh 1993a; Tirosh 1993b; Wilens 2008; Wilkison 1995).

Thirteen trials reported that participants were taking other medications (Buitelaar 1995; Carlson 1995; Castellanos 1997; Cox 2006; Findling 2007; Gadow 2007; Gonzalez‐Heydrich 2010; Kaplan 1990; McBride 1988a; Pearson 2013; Pelham 1989; Szobot 2008; Zeni 2009), and 42 trials stated that comedication was not used. It was not clear whether the remaining 92 trials permitted comedication.

Interventions

In all, 101 trials reported use of a single type of methylphenidate (without specifying the type), whereas 12 trials reported use of a single type of immediate‐release methylphenidate (Ahmann 1993; Chronis 2003; Fabiano 2007; Gadow 2007; Johnston 1988; Kollins 2006 (PATS); Moshe 2012; Smith 2004; Solanto 2009; Swanson 1999; Swanson 2002a; Tervo 2002), and 19 reported use of a single type of extended‐release methylphenidate. Seven trials used both immediate‐ and extended‐release methylphenidate (Döpfner 2004; Fitzpatrick 1992a; Pelham 1990a; Pelham 2001a; Schachar 2008; Swanson 2002b; Wigal 2003), four trials used two different types of extended‐release methylphenidate (Lopez 2003; Schulz 2010; Silva 2005a; Swanson 2004b), and four used transdermal methylphenidate patches (Pelham 2005; Pelham 2011; Wilens 2008; Wilens 2010).

The method of reporting the dose of methylphenidate varied considerably between trials, and three trials did not report dose (Bliznakova 2007; Nikles 2006; Wallace 1994). Overall daily dose ranged from 4 mg to 77 mg, with mean reported total daily dose of 24.5 mg or 0.6 mg/kg. Doses of immediate‐release methylphenidate ranged from 4 mg to 54 mg, with mean reported total daily dose of 25 mg or 0.7 mg/kg. Doses of extended‐release methylphenidate ranged from 15 mg to 77 mg, with mean reported total daily dose of 36 mg or 1.1 mg/kg. Duration of methylphenidate treatment ranged from 1 to 56 days, with an average duration of 15.6 days.

All trials used placebos as control.

Three trials used concomitant cointerventions (antidepressant: Carlson 1995; behaviour modification: Kolko 1999; and antipsychotic: Zeni 2009) in both intervention and control groups.

Outcomes

Some psychometric ADHD instruments measured total score for ADHD symptoms, whereas others assessed only specific symptom domains of ADHD (e.g. inattention, hyperactivity, impulsivity). We categorised all scales into five subgroups: ADHD symptoms; serious adverse events; non‐serious adverse events; general behaviour; and quality of life. Some psychometric instruments are abbreviated versions or revised versions, but all have been validated.

ADHD symptoms

Conners' questionnaires were the most frequently used measures of ADHD symptoms; more than 30 different versions measured core symptoms of ADHD (normative data are generally well intercorrelated in revised versions; Goyette 1978).

Table 1 presents the list of all measures used to assess ADHD symptoms in the included trials. This list primarily refers to original articles describing the psychometric properties of measurement scales, but in a few cases, we refer to trials describing use of a specific measurement scale.

Table 1. ADHD symptoms ‐ rating scales

Name of scale	Abbreviation	Reference
Abbreviated Conners’ Rating Scales, Parent (ACPRS) and Teacher (ACTRS), including Abbreviated Parent Rating Scale (APRS) and Teacher Rating Scale, Hyperkinesis Index and ADHD and Emotional Lability subscales	ACRS	Conners 1997a
Abbreviated Symptom Questionnaire, including ASQ Teacher and ASQ Parent	ASQ	Conners 1995
Academic Performance Rating Scale	APRS	DuPaul 1991a
The ADD/H Comprehensive Teacher Rating Scale	ACTeRS	Ullmann 1984
ADHD/ODD Rating Scale, Parent‐ and Teacher‐Rated	ADHD‐RS	Barkley 1998
ADHD Rating Scale, including ADHD Rating Scale Parent and Teacher Ratings	ADHD‐RS	DuPaul 1991a
ADHD Rating Scale‐IV, including ADHD Rating Scale‐IV Parent and Teacher Versions	ADHD‐RS‐IV	DuPaul 1991a
Brief Psychiatric Rating Scale for Children	BPRS	Gale 1986
Child Attention Problems Rating Scale	CAP	Achenbach 1986
Child Attention Profile	CAP	Barkley 1988b
Child Behavior Rating Form	NCBHF	Aman 1996
Child Symptom Inventory	CSI	Gadow 1994
Children’s Psychiatric Rating Scale	CPRS	Pfefferbaum‐Levine 1983
Conners’ Abbreviated Hyperactivity Questionnaire	C‐HI	Conners 1997a
Conners’ Abbreviated Questionnaire	ASQ	Conners 1995
Conners’ Abbreviated Parent Teacher Questionnaire	APTQ	Rowe 1997
Conners’ Abbreviated Rating Scale	ABRS	Conners 1997a
Conners’ Abbreviated Symptom Questionnaire	ASQ	Conners 1995
Conners Abbreviated Symptom Questionnaire for Parents	ASQ‐Parent	Conners 1995
Conners’ Abbreviated Symptom Questionnaire for Teachers	ASQ‐Teacher	Conners 1997a
Conners’ Abbreviated Teacher Rating Scale	ABTRS	Conners 2001
Conners’ ADHD/DSM‐IV Scales Adolescent	CADS‐A	Conners 1997b
Conners’ ADHD/DSM‐IV Scales Parent	CADS–P, CADS‐P DSM‐IV	Conners 1997a
Conners’ ADHD/DSM‐IV Scale Teacher, including Inattentive and Hyperactive‐Impulsive subscales	CADS‐T, CADS‐T DSM‐IV	Conners 1997a
Conners’ Rating Scale ‐ Revised, Parent and Teacher: Hyperactivity and Conduct Factors score	CPRS‐R and CTRS‐R	Goyette 1978
Conners’ Hyperactivity Index, Parent and Teacher, including abbreviated versions	CPRS/CTRS‐Hyperactivity index	Conners 1997a
Conners’ Hyperkinesis Index	‐	Milich 1980
Conners, Loney and Milich Scale	CLAM	Milich 1980
Conners’ Parent and Teacher Rating Scale ‐ Revised, Short Form	CRS‐R:S	Conners 1997a
Conners’ Parent Rating Scale, including abbreviated versions	CPRS	Conners 1998b
Conners’ Parent Rating Scale ‐ Revised	CPRS‐R	Conners 1997a
Conners’ Parent Rating Scale ‐ Revised, Short Form	CPRS‐R:S	Conners 1997a
Conners’ Parent Rating Scale ‐ Revised, Long Version	CPRS‐R:L	Conners 1997a
Conners’ Rating Scale ‐ Revised	CRS‐R	Conners 1997a
Conners’ Short Form Rating Scale, Parent and Teacher	‐	Conners 1997a
Conners’ Teacher Rating Scale	CTRS	Conners 1998a
Conners’ Teacher Rating Scale ‐ Revised, Long Version	CTRS‐R:L	Conners 1998a
Diagnostic and Statistical Manual of Mental Disorders Total	DSM‐IV	APA 1994
Diagnostiksystem für Psychische Störungen im Kindes ‐ und Jugendalter nach ICD‐10 und DSM‐IV, Parental Questionnaire of ADHD symptoms	DISYPS	Döpfner 2000
Fremdbeurteilungsbogen für Hyperkinetische Störungen	FBB‐HKS	Döpfner 2008
German Teacher’s report on ADHD symptoms	FBB‐HKS of the DISYPS	Döpfner 2000
Hyperactivity Index of the Revised Conners Parent and Teacher Rating Scales	‐	Goyette 1978
IOWA Conners Parent Rating Scale, including abbreviated versions	IOWA CPRS	Loney 1982
IOWA Conners Teacher Rating Scale, including abbreviated versions	IOWA CTRS	Loney 1982
IOWA Conners Teacher Rating Scale, Inattention/Overactivity (I/O) and Oppositional/Defiant (O/D) subscales	IOWA‐I/O and O/D subscales	Loney 1982
IOWA Inattention/Overactivity and Aggression/Noncompliance scales ‐ Parent and Teacher rating	IOWA	Loney 1982
Lehrer‐Fragenbogen von Steinhausen	LF	Steinhausen 1993
Loney’s Time on Task Scale, Hyperactivity, Attention and Aggression subscales	TOTS	Fitzpatrick 1992b
Modified Conner Scale Parent and Teacher	ACR	Conners 1997a
Mothers’ Objective Method for Subgrouping	MOMS	Loney 1984
Parent Symptom Checklist	PSC ADHD	Döpfner 2000
Parental Account of Children’s Symptoms	PACS	Chen 2006
Restricted Academic Situation Scale	RASS	Fischer 1998
Schedule for Affective Disorders and Schizophrenia	K‐SADS/ K‐SADS‐E for diagnosis	Chambers 1985
Schedule for Non‐adaptive and Adaptive Personality	SNAP	Clark 1993; Clark 1996
Swanson, Nolan, and Pelham ‐ IV SNAP‐ADHD Rating scale	SNAP‐ADHD	Swanson 1992
Swanson, Nolan, and Pelham ‐ IV SNAP‐IV (Brazilian Version)	SNAP‐IV	Clark 1993; Clark 1996
Swanson, Kotkin, Atkins, M‐Flynn, Pelham Scale (SKAMP combined, SKAMP attention, and SKAMP deportment)	SKAMP (SKAMP combined, SKAMP attention, and SKAMP deportment)	Wigal 1998; Murray 2009
Teacher Self‐control Rating Scale	SCRS	Kendall 1979
Turgay ‐ DSM‐IV Scale, Parent	T‐DSM‐IV Scale, Parent	Turgay 1994; Ercan 2001
Turgay ‐ DSM‐IV Scale, Teacher	T‐DSM‐IV Scale, Teacher	Turgay 1994; Ercan 2001
Teacher Hyperactivity Index	THI	Achenbach 1991b
Teacher Symptom Checklist	TSC	Döpfner 2000
Vanderbilt ADHD Rating Scale	VADP(T)RS	Wolraich 2003
Wender Utah Rating Scale	WURS	Ward 1993
Wide Range Achievement Test	WRAT‐4	Wilkinson 2006
Wide Range Achievement Test Revised	WRAT‐R	Woodcock 2001

ADD/H: Attention deficit disorder with hyperactivity.
ADHD: Attention deficit hyperactivity disorder.
DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition.
ODD: Oppositional defiant disorder.

Serious and non‐serious adverse events

Adverse events were measured by rating scales or by spontaneous reports and/or were recorded by investigators at regular interviews or visits. Some trials included physical examinations or paraclinical examinations, or both, such as blood testing, electrocardiogram (ECG), blood pressure reading, measurement of heart rate and assessment of weight and height. Serious adverse events were recorded in accordance with the ICH classification (ICH 1996). However, when in doubt, we asked trial authors which classification or definition they had used in their trial.

Some trials combined all of the above modes of measurement; others used a single measure such as spontaneous reports or rating scales. Fifty‐two trials employed rating scales; the Barkley Side Effects Rating Scale (SERS) was used most frequently (Barkley 1990).

Other scales used included the Significant Adverse Event Reviews Questionnaire (SAERS; Barkley 1990; Zeni 2009), the Pittsburgh Side Effect Rating Scale (PSERS; Pelham 1993b; Pelham 2005a) and Subject’s Treatment Emergent Symptom Scale (STESS; Guy 1976a).

For the purpose of measuring specific adverse events, some trials used rating scales such as Paediatric Sleep Questionnaire (PSQ; Chervin 2000), Sleep Disturbances Scale for Children (SDSC; Bruni 1996), Children’s Sleep Habits Questionnaire (CSHQ; Owens 2000), Child Depression Rating Scale (CDRS‐R; Poznanski 1983), Young Mania Rating Scale (YMRS; Young 1978), Yale Global Tic Severity Scale (YGTSS; Leckman 1989), Tic Symptom Self Report Scale (TSSR; Leckman 1988), and the Massachusetts General Hospital (MGH) Abuse and Diversion Questionnaire (Wilens 2006a).

General behaviour

Many different scales can be used to assess general behaviour. These scales have different foci, such as aggression or oppositional behaviour, but all describe participants’ behaviour and the influence of methylphenidate. Higher scores on general behaviour symptom scales signify better outcomes.

Table 2 presents the list of all measures used to assess general behaviour. This list refers primarily to the original articles describing the psychometric properties of measurement scales used to measure general behaviour in the included trials. In a few cases, we refer to trials that describe use of a specific measurement scale.

Table 2. Table 2: General behaviour rating scales

Name of scale	Abbreviation	Reference
Achenbach Child Behaviour Checklist	CBCL	Achenbach 1991a
Achenbach’s Teacher Report	ATRF	Achenbach 1991b; Achenbach 2001
ADHD Rating Scale	ADHD‐RS	DuPaul 1991a
ADHD School Observation Code	ADHD‐SOC	Gadow 1996
Barkley Scales, Disruptive Behavior Disorders Rating Scale	‐	Barkley 1991a
Child Attention Problems Scale	CAP	Barkley 1991
Child Attention Profile	CAP	Barkley 1988b
Child Behavior Checklist	CBCL	Achenbach 1991a
Child Health Questionnaire	CHQ	Landgraf 1998
Child and Adolescent Psychiatric Assessment, selected items	CAPA	Angold 1995
Children’s Psychiatric Rating Scale	CPRS	Fish 1985
Classroom Observation Code (Abikoff Classroom Observational System)	COC	Abikoff 1980
Code for Observing Social Activity	COSA	Sprafkin 1986
Conners' Child Behavior Scale	UC‐CCBS	Ladd 1996
Conners' Global Index Scale	CGI‐S	Conners 1998a
Conners’ Global Index ‐ Parent	CGI‐P	Conners 1997a
Conners' Global Index ‐ Teacher	CGI‐T	Conners 1998a
Conners', Loney and Milich Scale	CLAM	Milich 1980
Conners’ Parent Questionnaire	CPQ	Conners 1995
Conners’ Parent Rating Scale	CPRS	Conners 1998b
Conners’ Teacher Rating Scale	CTRS	Conners 1998a
Conners’ Teacher Rating Conduct Problems	‐	Miller 1997
Disruptive Behavior Disorders Rating Scale, Parent‐ and Teacher‐Rated	DBS	Mendelsohn 1978
Disruptive Behavior Disorders Rating Scale	DBD	Silva 2005b
Groninger Behaviour Observation Scale	GOO and GBO	Van der Meere 1999b
Groninger Behaviour Checklists, Parent and Teacher Versions of the abbreviated Groninger	GGGS and GGBS	Van der Meere 1999b
Hillside Behavior Rating Scale	HBRS	Gittleman‐Klein 1976
Home Situations Questionnaire	HSQ	Barkley 1987
Home Situations Questionnaire ‐ Revised	HSQ‐R	DuPaul 1992
Humphrey’s Teacher Self‐Control Rating Scale	TSCRS	Humphrey 1982
Hyperactivity Index from the Conners Revised Teacher Rating Scale	CTRS‐R‐Hyperactivity Index	Goyette 1978
Impairment Rating Scale	IRS	Fabiano 2006
Inpatient Global Rating Scale, Revised	IGRS	Conners 1985
Inpatient Global Rating Scale, Somatic factor	IGRS‐S	Conners 1985
IOWA Conners' Rating Scale, Oppositional/Defiant (O/D) subscales	IOWA‐O/D subscales	Loney 1982
Nisonger Child Behavior Rating Form	NCBRF	Aman 1996
Paired Associates Learning	PAL	Wechsler 1945
Parent Global Assessment for Improvement	PGA	McGough 2006a
Peer Conflict Scale	PCS	Marsee 2007
Personality Inventory for Children	PIC	Lachar 1986
School Situations Questionnaire	SSQ	Barkley 1987
School Situations Questionnaire ‐ Revised	SSQ‐R	DuPaul 1992
Schedule for Nonadaptive and Adaptive Personality	SNAP	Clark 1993; Clark 1996
Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale, Parent and Teacher	SWAN	Swanson 2006; Polderman 2007
Subjective Treatment Emergent Symptom Scale	STESS‐R	Guy 1976
Swanson, Nolan and Pelham, Fourth Edition	SNAP‐IV	Bussing 2008
Teachers Report Form	TRF	Achenbach 1991b
Telephone Interview Probe (Parent and Teacher)	TIP	Corkum 2007
Vanderbilt ADHD rating scales: Vanderbilt ADHD Diagnostic Parent Rating Scale and Vanderbilt ADHD Diagnostic Teacher Rating Scale	VADPRS and VADTRS	Wolraich 2003
Wahler, House and Stambaugh’s Ecobehavioral Assessment System	ECO	Wahler 1976
The Weekly Parent Ratings of Evening and Morning Behaviour	WREMB‐R	Kelsey 2004
Werry‐Weiss‐Peters Activity Rating Scale	WWP	Routh 1978
Woodcock‐Johnson Achievement Battery	WJ‐III Ach	Woodcock 2001

ADHD: attention deficit hyperactivity disorder.

Quality of life

Seven scales measured quality of life in relation to both ADHD and life in general. All of these scales produced higher values equating to better health. Only three could be used in meta‐analyses: Child Health Questionnaire (CHQ; Landgraf 1998); Children's Global Assessment Scale (CGAS; Shaffer 1983); and Child Health and Illness Profile, Child Edition: Parent Report Form (CHIP‐CE:PRF; Riley 2004).

See Table 3 for additional information on types of rating scales used to assess quality of life in the included trials.

See: Summary of findings for the main comparison Methylphenidate compared with placebo or no intervention for ADHD

Table 3. Table 3: Quality of life ratings scales

Name of scale	Abbreviation	Reference
ADHD Impact Module‐Child	AIM‐C	AIM‐C 2013
Child Impact Scale and Home Impact Scale	CIS/HIS	Landgraf 2002
Child Health and Illness Profile, Child Edition: Parent Report Form	CHIP‐CE:PRF	Riley 2004
Child Health Questionnaire	CHQ‐P	Landgraf 1998
Children's Global Assessment Scale	CGAS	Shaffer 1983
Comprehensive Psychopathological Rating Scale	CPRS	Aasberg 1978
Health Utilities Index ‐ 2	HUI‐2	Torrance 1982

ADHD: attention deficit hyperactivity disorder.

Excluded studies

We have taken a very inclusive approach by reading 1460 full text reports, 603 of which were subsequently excluded for the following reasons: 367 reports of non‐randomised studies, 131 reports with no description of an acceptable ADHD diagnosis, 15 reports describing no treatment with methylphenidate, 5 reports that only included patients older than 18 years of age, 40 reports of patients with a IQ below 70, 31 reports with polypharmacy, and 14 reports excluded for various 'other' reasons (see Figure 1).

We formally excluded 78 trials (88 reports) that examined the impact of methylphenidate on very specific domains that were outside the focus of this review such as motor co‐ordination, reaction time, memory tasks, and reading skills. For more information on these trials, please see Characteristics of excluded studies tables.

Risk of bias in included studies

We assessed the risk of bias of each included trial using the Cochrane 'Risk of bias' tool (Higgins 2011). A summary of our assessment is displayed in Figure 2 and Figure 3. As shown, we assessed six cross‐over trials (3%) as having low risk of bias in all domains. We assessed the remaining 179 trials (97%) as trials with high risk of bias. However, even the six cross‐over trials had likely breaks in their blinding due to prevalent adverse events to methylphenidate (see below). Accordingly, we judged all 185 trials to be trials with high risk of bias.

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included trial.

Parallel‐group trials

None of the 38 included parallel‐group trials had low risk of bias in all bias domains.

Allocation

Random sequence generation

We considered random sequence generation to be at low risk of bias in 22 trials, at high risk of bias in 3 trials (Connor 2000; Green 2011; Heriot 2008), and at unclear risk of bias in 13 trials (Brown 1985; Butter 1983; Findling 2006; Firestone 1981; Greenhill 2002; Greenhill 2006; Ialongo 1994; Lin 2014 ; Newcorn 2008; Perez‐Alvarez 2009; Tucker 2009; Wigal 2004; Wilens 2006).

Allocation concealment

We considered allocation concealment to be at low risk of bias in 18 trials, at high risk of bias in 1 trial (Green 2011; Szobot 2008), and insufficiently reported in 19 trials.

Blinding

We considered the method of blinding of participants and personnel to be adequately described in 28 trials but insufficiently reported in 6 trials (Biederman 2003; Childress 2009; Findling 2010; Greenhill 2006; Lin 2014; Tucker 2009). Four trials were not blinded (Brown 1985; Duric 2012; Jensen 1999 (MTA); Perez‐Alvarez 2009).

We considered the method of blinding of outcome assessment to be adequately described in 23 trials but insufficiently reported in 11 trials (Biederman 2003; Butter 1983; Childress 2009; Findling 2006; Findling 2008; Findling 2010; Green 2011; Greenhill 2002; Greenhill 2006; Lin 2014; Martins 2004). Four trials did not include blinded outcome assessors (Brown 1985; Duric 2012; Jensen 1999 (MTA); Perez‐Alvarez 2009).

Incomplete outcome data

Twenty‐four trials adequately addressed incomplete data, and 10 trials did not (Findling 2006; Findling 2008; Findling 2010; Heriot 2008; Ialongo 1994; Lin 2014; Pliszka 2000; Schachar 1997; Tucker 2009; Wilens 2006). In four trials information was insufficient for assessment of whether the method used to handle missing data was likely to bias the estimate of effect (Coghill 2013; Firestone 1981; Palumbo 2008; Wigal 2004).

Selective reporting

Twenty‐five trials reported all pre‐defined or otherwise expected outcomes, and two trials did not (Greenhill 2006; Lin 2014). In 11 trials it was unclear whether trial authors reported all pre‐defined or otherwise expected outcomes (Butter 1983; Firestone 1981; Greenhill 2002; Heriot 2008; Horn 1991; Ialongo 1994; Perez‐Alvarez 2009; Schachar 1997; Szobot 2004; Tucker 2009; Wolraich 2001).

Vested interests (industry bias)

Twenty‐two trials were funded by industry. In seven trials information was insufficient for assessment of whether potential vested interests were likely to bias the estimate of effect (Horn 1991; Ialongo 1994; Jacobi‐Polishook 2009; Martins 2004; Schachar 1997a; Tourette's Syndrome Study Group 2002; Riggs 2011). No vested interests were described in the remaining nine trials (Brown 1985; Butter 1983; Connor 2000; Duric 2012; Firestone 1981; Green 2011; Heriot 2008; Jensen 1999 (MTA); Perez‐Alvarez 2009).

Other potential sources of bias

We identified no other sources of bias.

Cross‐over trials

Only 6 of the 147 included cross‐over trials had low risk of bias in all bias domains (DuPaul 1996; Flapper 2008; Gorman 2006; Rapport 2008; Stein 1996; Wilkison 1995), and even these were considered at risk of deblinding due to prevalent adverse events (see below).

Allocation (selection bias)

Random sequence generation

We considered random sequence generation to be at low risk of bias in 58 trials, at high risk of bias in 10 trials (Carlson 1995; Fitzpatrick 1992a; Kaplan 1990; Kelly 1989; Manos 1999; McBride 1988a; Szobot 2008; Tirosh 1993b; Whalen 1990; Wigal 2014), and at unclear risk of bias in 79 trials.

Allocation concealment

We considered allocation concealment to be at low risk of bias in 53 trials, at high risk of bias in 5 trials (Carlson 1995; Fitzpatrick 1992a; Szobot 2008; Ullmann 1985; Wigal 2014), and not sufficiently reported in 89 trials.

Blinding (performance bias and detection bias)

We judged the method of blinding of participants and personnel to be adequately described in 109 trials but unclear in 29 trials. Nine trials were not blinded (Lopez 2003; Manos 1999; Pearson 2013; Pelham 2014; Ramtvedt 2013; Stein 2003; Ullmann 1985; Whalen 1990; Wigal 2013).

We judged the method of blinding of outcome assessment to be adequately described in 79 trials but unclear in 62 trials. Six trials did not include blinded outcome assessors (Cox 2006; Douglas 1986; Manos 1999; Whalen 1990; Wigal 2013; Wodrich 1998).

Incomplete outcome data (attrition bias)

Sixty‐four trials adequately addressed incomplete data. In 54 trials information was insufficient for assessment of whether the method used to handle missing data was likely to bias the estimate of effect. Twenty‐nine trials had incomplete outcome data.

Selective reporting (reporting bias)

Forty‐nine trials reported all pre‐defined or otherwise expected outcomes, and eight trials did not (Castellanos 1997; Chacko 2005; Gonzalez‐Heydrich 2010; McGough 2006; McInnes 2007; Stein 2003; Sunohara 1999; Taylor 1993). In 90 trials it was unclear whether trial authors reported all pre‐defined or otherwise expected outcomes.

Vested interests (industry bias)

Fifty trials were funded by industry. In 41 trials we judged information to be insufficient for assessment of whether potential vested interests were likely to bias the estimate of effect. The remaining 56 trials reported no vested interests.

Other potential sources of bias

We identified no other sources of bias.

Cohort selection bias of all participants

Ten parallel‐group trials and 49 cross‐over trials excluded methylphenidate non‐responders, placebo responders and/or patients with methylphenidate adverse events before randomisation. Such trials have limited external validity.

Effects of interventions

Below, we present the results of meta‐analyses performed for two primary outcomes (ADHD symptoms and serious adverse events) and three secondary outcomes (non‐serious adverse events, general behaviour and quality of life). For a summary of key results, please see summary of findings Table for the main comparison.

Comparison: methylphenidate versus placebo or no intervention

Primary outcomes

ADHD symptoms

We were able to combine data on ADHD symptoms from 25 parallel‐group trials and 74 cross‐over trials. The evidence was assessed to be of very low quality (see GRADE assessment below).

Teacher‐rated ADHD symptoms

Parallel‐group trials and cross‐over trials (first period data only)

Meta‐analysis suggested that methylphenidate significantly reduced teacher‐rated ADHD symptoms compared with placebo (standardised mean difference (SMD) ‐0.77, 95% confidence interval (CI) ‐0.90 to ‐0.64; I² = 37%; 19 trials, 1698 participants; Analysis 1.1). The SMD of ‐0.77 for ADHD symptoms corresponds to a mean difference (MD) of ‐9.6 points (95% CI ‐13.7 to ‐6.4) on the ADHD Rating Scale (DuPaul 1991a).

Our subgroup analyses revealed that the intervention effect varied according to the following.

Types of scales used (test for subgroup differences: Chi² = 24.81, df = 10 (P value = 0.006); I² = 59.7%; Analysis 1.2).
Medication status before randomisation, with a seemingly better effect in trials with no medication‐naive patients (SMD ‐1.06, 95% CI ‐1.33 to ‐0.79; I² = 0%; 2 trials, 286 participants) compared to trials with medication‐naive patients (SMD ‐0.63, 95% CI ‐0.94 to ‐0.31; I² = 51%; 4 trials, 431 participants (test for subgroup differences: Chi² = 4.18, df = 1 (P value = 0.04); I² = 76.1%; Analysis 1.3).
Duration of treatment, with seemingly lesser effect in long‐term trials (SMD ‐0.47, 95% CI ‐0.72 to ‐0.22; 1 trial, 253 participants) compared to short‐term trials (SMD ‐0.81, 95% CI ‐0.94 to ‐0.68; I² = 25%; 18 trials, 1445 participants) (test for subgroup difference: Chi² = 5.37, df = 1 (P value = 0.02); I² = 81.4%; Analysis 1.4). The SMD effect of ‐0.47 for ADHD long‐term trials corresponds to an MD of only ‐5.75 points (95% CI ‐4.2 to ‐9.2) on the ADHD‐RS (DuPaul 1991a).

No evidence suggested that dose (Analysis 1.5), trial design (Analysis 1.6) or risk of cohort selection bias (Analysis 1.7) influenced the estimated intervention effect. All 19 trials were at high risk of bias (Analysis 1.1). Inspection of the funnel plot in Figure 4 suggested potential bias (asymmetry), although we found no evidence of significant publication bias: Egger’s regression intercept (bias) was ‐0.2260 (two‐tailed, P value = 0.81). Finally, one of the trials in this meta‐analysis used change from baselines scores (Palumbo 2008), but removing this trial did not significantly change the estimate.

Figure 4

Funnel plot of comparison: 1. Teacher‐rated ADHD symptoms, outcome: 1.8 All data at low and high risk of bias (parallel‐group and cross‐over trials).

Cross‐over trials (endpoint data)

Meta‐analysis suggested that methylphenidate significantly reduced teacher‐rated ADHD symptoms compared with placebo (SMD ‐0.93, 95% CI ‐1.06 to ‐0.80; I² = 77%; 59 trials, 5145 participants; Analysis 1.8). The estimated intervention effect did not vary according to risk of bias (Analysis 1.8), but did vary according to dose of methylphenidate (Analysis 1.9). Three of these trials included some participants with an IQ less than 70 (Pearson 2013; Smith 1998; Taylor 1987). Removing these trials did not significantly change the estimate.

Parallel‐group trials and cross‐over trials (endpoint data)

Meta‐analysis suggested that methylphenidate significantly reduced teacher‐rated ADHD symptoms compared with placebo (SMD ‐0.91, 95% CI ‐1.01 to ‐0.80; I² = 72%; 75 trials, 6344 participants; Analysis 1.10). No evidence suggested that the intervention effect varied according to risk of bias (low risk of bias versus high risk of bias; Analysis 1.11).

Independent assessor‐rated ADHD symptoms

Most independent assessors were clinicians.

Parallel‐group trials and cross‐over trials (first period data only)

Meta‐analysis suggested that methylphenidate significantly reduced independent assessor‐rated ADHD symptoms compared with placebo (SMD ‐0.64, 95% CI ‐0.89 to ‐0.39; I² = 85%; 10 trials, 1907 participants; Analysis 2.1). The SMD effect of ‐0.64 for ADHD symptoms corresponds to an MD of ‐8.0 points (95% CI ‐5.8 to ‐12.5) on the ADHD‐RS (DuPaul 1991a). No evidence indicated that the intervention effect was influenced by types of scales (Analysis 2.2), duration of treatment (Analysis 2.3), trial design (Analysis 2.4) or risk of cohort selection bias (Analysis 2.5). We were unable to test for subgroup differences according to dose, as no low‐dose methylphenidate trials were identified (Analysis 2.6). All 10 trials were at high risk of bias (Analysis 2.1). Two trials reported change from baseline scores (Findling 2008; Newcorn 2008). Removing these trials did not significantly change the estimate.

Cross‐over trials (endpoint data)

Meta‐analysis suggested that methylphenidate significantly reduced independent assessor‐rated ADHD symptoms compared with placebo (SMD ‐1.00, 95% CI ‐1.16 to ‐0.84; I² = 69%; 19 trials, 2471 participants; Analysis 2.7). All 19 trials were at high risk of bias (Analysis 2.7). The intervention effect was significantly influenced by the dose of methylphenidate, with the highest estimate reported in the high‐dose group (SMD ‐1.13, 95% CI ‐1.31 to ‐0.95; I² = 66%; 12 trials, 1853 participants; Analysis 2.8).

Parallel‐group trials and cross‐over trials (endpoint data)

Meta‐analysis suggested that methylphenidate significantly reduced independent assessor‐rated ADHD symptoms compared with placebo (SMD ‐0.83, 95% CI ‐0.98 to ‐0.67; I² = 82%; 28 trials, 4215 participants; Analysis 2.9).

Parent‐rated ADHD symptoms

Parallel‐group trials and cross‐over trials (first period data only)

Meta‐analysis suggested that methylphenidate significantly reduced parent‐rated ADHD symptoms compared with placebo (SMD ‐0.66, 95% CI ‐0.82 to ‐0.51; I² = 60%; 21 trials, 2187 participants; Analysis 3.1). The SMD effect of ‐0.66 for ADHD symptoms corresponds to an MD of ‐8.2 points (95% CI ‐6.0 to 12.9) on the ADHD‐RS (DuPaul 1991a). We found evidence that the intervention effect varied by types of scales (test for subgroup differences: Chi² = 22.34, df = 10 (P value = 0.01); I² = 55.2%; Analysis 3.2). However, duration of treatment (Analysis 3.3), dose of methylphenidate (Analysis 3.4), medication status before randomisation (Analysis 3.5), risk of cohort selection bias (Analysis 3.6) and trial design (Analysis 3.7) did not seem to significantly influence the intervention effect. All 21 trials were at high risk of bias (Analysis 3.1). Two trials in the meta‐analysis reported change from baseline scores (Carlson 2007; Tucker 2009), but removing these trials did not significantly change the estimate.

Cross‐over trials (endpoint data)

Meta‐analysis suggested that methylphenidate significantly reduced parent‐rated ADHD symptoms compared with placebo (SMD ‐0.78 (95% CI ‐0.90 to ‐0.67; I² = 59%; 41 trials, 3734 participants; Analysis 3.8). This effect was not significantly influenced by risk of bias (Analysis 3.8) or dose of methylphenidate (Analysis 3.9). Two trials included some participants with an IQ less than 70 (Pearson 2013; Taylor 1987), but removing these trials did not significantly change the estimate.

Parallel‐group trials and cross‐over trials (endpoint data)

We observed comparable findings when combining data from parallel‐group trials with endpoint data from cross‐over trials (SMD ‐0.74, 95% CI ‐0.83 to ‐0.65; I² = 58%; 59 trials, 5861 participants; Analysis 3.10).

Additional subgroup analyses

We tested for differences between raters (teachers, independent assessors and parents) and found no significant differences (test for subgroup differences: Chi² = 2.00, df = 2 (P value = 0.37); I² = 0.0%; Analysis 4.1).

We found no evidence suggesting that age (Analysis 4.2) or comorbidity influenced the intervention effect (Analysis 4.3). However, the intervention effect was significantly influenced by ADHD subtype, with a greater intervention effect noted for the inattentive subtype (SMD ‐1.31, 95% CI ‐1.61 to ‐1.01; 1 trial, 204 participants) compared with the combined subtype (SMD 0.65, 95% CI ‐1.30 to 2.60; I² = 99%; 2 trials, 559 participants) (test for subgroup differences: Chi² = 3.79, df = 1 (P value = 0.05); I² = 73.6%; Analysis 4.4). This difference rested upon one single trial.

We found no evidence of a 'carry‐over effect' in the cross‐over trials. We conducted a subgroup analysis to investigate the difference between first period data and endpoint data from four cross‐over trials (372 participants), and we found no significant subgroup differences (test for subgroup differences: Chi² = 2.47, df = 1 (P value = 0.12); I² = 59.6%; Analysis 4.5).

Numbers of serious adverse events

We were able to combine data on serious adverse events from nine parallel‐group trials (Carlson 2007; Childress 2009; Coghill 2013; Findling 2010; Jacobi‐Polishook 2009; Lehmkuhl 2002; Palumbo 2008; Riggs 2011; Wolraich 2001) and eight cross‐over trials (Brams 2008; Brams 2012; Buitelaar 1995; Cox 2006; Grizenko 2012; Schachar 2008; Silva 2008; Wigal 2013).

The evidence was assessed to be of very low quality (see GRADE assessment below). Therefore, we cannot exclude that harms may be worse than reported.

Parallel‐group trials and cross‐over trials (first period data only)

We found no significant differences between participants in the methylphenidate group and those in the control group as regards the numbers of serious adverse events (RR 0.98, 95% CI 0.44 to 2.22; 9 trials, 1532 participants; Analysis 5.1), and we found no evidence of heterogeneity between trials (I² = 0%).

Serious adverse events reported by participants in the methylphenidate group were cyst rupture (n = 1), kidney infection (n = 1) and psychosis (n = 2). Serious adverse events reported by participants in the control group included loss of consciousness (n = 1), drug toxicity (n = 1), asthma (n = 1) and concussion (n = 1). Severe adverse events reported by both groups included syncope (methylphenidate n = 3, control n = 1) and aggression (methylphenidate n = 1, control n = 2) (Analysis 5.1).

Cross‐over trials (endpoint data)

We found no differences between those in the methylphenidate group and individuals in the control group regarding the incidence of serious adverse events (RR 1.62, 95% CI 0.34 to 7.71; I² = 0%; 8 trials; n = 1721; Analysis 6.1). However, data from only four trials contributed to this meta‐analysis because zero events were reported in the other four trials.

Trial Sequential Analysis

We conducted a Trial Sequential Analysis on the 'total serious adverse events’ outcome, involving nine parallel‐group trials. We had planned to use a relative risk reduction of 20%, but the distance between the accrued information and the required information was too large, and the program failed to calculate and draw an interpretable figure. We therefore increased the relative risk reduction to 25%. We included trials with zero serious adverse events by substituting zero with a constant of 0.5 (Jacobi‐Polishook 2009; Wolraich 2001). We calculated the DARIS on the basis of serious adverse events in the control group of 2%; a relative risk reduction or increase in the experimental group of 25%; type I error of 5%; type II error of 20% (80% power); and diversity (D²) of 0%. The DARIS was 21,593 participants. The cumulative Z‐curve did not cross the conventional or trial sequential monitoring boundaries for benefit, harm, or futility (see Figure 5). As only less than 7% of the DARIS was accrued, risks of random type II error cannot be excluded. The Trial Sequential Analysis‐adjusted intervention effect was RR 0.91 (CI 0.02 to 33.2).

Figure 5

Trial Sequential Analysis: serious adverse events.

Secondary outcomes

Non‐serious adverse events

We were able to combine data on non‐serious adverse events from 26 parallel‐group trials and 67 cross‐over trials in a meta‐analysis. The evidence was assessed to be of very low quality (see GRADE assessment below). Therefore, we cannot exclude that harms may be worse than reported.

Parallel‐group trials and cross‐over trials (first period data only)

Overall adverse events

Participants receiving methylphenidate were significantly more likely to experience non‐serious adverse events overall (RR 1.29, 95% CI 1.10 to 1.51; I² = 73%; 21 trials, 3132 participants; Analysis 7.1). Substantial heterogeneity between trials was observed (Tau² = 0.07; Chi² = 69.75; df = 19 (P value < 0.00001); I² = 73%). However, this heterogeneity did not appear to be related to dose, as no differences were reported between low‐dose and high‐dose methylphenidate trials (test for subgroup differences: Chi² = 1.11; df = 2 (P value = 0.57); I² = 0%; Analysis 7.2).

Non‐serious adverse events included those affecting the nervous system (Analysis 7.3), as well as the digestive (Analysis 7.4), urinary (no data), circulatory and respiratory (Analysis 7.5), reproductive (no data), skeletal and muscular (Analysis 7.6), and immune systems (Analysis 7.7). Other reported adverse events included physical differences in height (Analysis 7.8), weight (Analysis 7.9), body mass index (BMI) (Analysis 7.10), vital signs (Analysis 7.11), and several less common events (Analysis 7.12).

Compared with those in the control group, participants in the methylphenidate group were more likely to:

report trouble sleeping or sleep problems (RR 1.60, 95% CI 1.15 to 2.23; I² = 0%; 13 trials, 2416 participants; Analysis 7.3);
report a decrease in appetite (RR 3.66, 95% CI 2.56 to 5.23; I² = 28%; 16 trials, 2962 participants; Analysis 7.4) and a decrease in weight (RR 3.89, 95% CI 1.43 to 10.59; I² = 0%; 6 trials, 859 participants; Analysis 7.4);
weigh significantly less (SMD ‐1.12, 95% CI ‐1.55 to ‐0.70; I² = 83%; 5 trials, 805 participants; Analysis 7.9);
have a lower body mass index (BMI) (MD ‐0.60, 95% CI ‐0.84 to ‐0.36; 1 trial, 215 participants; Analysis 7.10); and
have a higher pulse (MD 3.41, 95% CI 0.87 to 5.94; I² = 70%; 8 trials, 1240 participants; Analysis 7.11).

Cross‐over trials (endpoint data)

Overall adverse events

Significantly more adverse events were reported in the methylphenidate group (RR 1.33, 95% CI 1.11 to 1.58; I² = 18%; 21 trials, 2072 participants; Analysis 8.1). In addition, differences were noted in the numbers of events reported in trials of low doses of methylphenidate compared with trials of high doses of methylphenidate (test for subgroup differences: Chi² = 5.68, df = 2; P value = 0.06, I² = 64.8%; Analysis 8.2).

Categories of non‐serious adverse events included those affecting the nervous system (Analysis 8.3), as well as the digestive (Analysis 8.4), urinary (Analysis 8.5), skeletal and muscular (Analysis 8.6), and immune systems (Analysis 8.7), and skin (Analysis 8.8). Other reported adverse events included physical differences in vital signs (Analysis 8.9), height (Analysis 8.10) and weight (Analysis 8.11).

Compared with the control group, participants in the methylphenidate group were less likely to report:

anger (RR 0.45, 95% CI 0.26 to 0.77; I² = 0%; 3 trials, 264 participants; Analysis 8.3);
behavioural complaints (RR 0.55, 95% CI 0.35 to 0.86; 1 trial, 82 participants; Analysis 8.3); and
an increase in appetite (RR 0.20, 95% CI 0.08 to 0.50; 1 trial, 136 participants; Analysis 8.4).

However, they were more likely to report:

compulsive acts (RR 2.57, 95% CI 1.45 to 4.56; 1 trial, 90 participants; Analysis 8.3);
daydreaming (RR 0.66, 95% CI 0.44 to 0.98; I² = 0%; 3 trials, 222 participants; Analysis 8.3);
headache (RR 1.21, 95% CI 1.01 to 1.45; I² = 0%; 37 trials, 3752 participants; Analysis 8.3);
insomnia or sleep problems (RR 1.57, 95% CI 1.20 to 2.06; I² = 47%; 31 trials, 3270 participants; Analysis 8.3);
being overly meticulous (RR 40.77, 95% CI 2.35 to 706.72; 1 trial, 96 participants; Analysis 8.3);
obsessive thinking (RR 2.35, 95% CI 1.53 to 3.62; 1 trial, 90 participants; Analysis 8.3);
tics or nervous movements (RR 1.33, 95% CI 1.03 to 1.72; I² = 10%; 19 trials, 1403 participants) (Analysis 8.3);
a decrease in appetite (RR 3.04, 95% CI 2.35 to 3.94; I² = 40%; 35 trials, 3862 participants; Analysis 8.4);
stomachache (RR 1.61, 95% CI 1.27 to 2.04; I² = 22%; 33 trials, 3777 participants; Analysis 8.4);
somatic complaints (MD 0.85, 95% CI 0.79 to 0.91; 1 trial, 82 participants; Analysis 8.6); and
higher pulse/heart rate (MD 5.06, 95% CI 2.88 to 7.24; I² = 57%; 14 trials, 939 participants; Analysis 8.9).

Trial Sequential Analysis

We conducted a Trial Sequential Analysis on the 'total non‐serious adverse events’ outcome involving 21 parallel‐group trials. We included one trial with zero non‐serious adverse events by substituting zero with a constant of 0.5 (Jacobi‐Polishook 2009). We calculated the DARIS on the basis of adverse events in the control group of 47%; relative risk reduction in the intervention group of 20%; type I error of 5%; type II error of 20% (80% power); and diversity (D‐square) of 79%. The DARIS was 4133 participants. The cumulative Z‐curve (blue line) crossed the trial sequential monitoring boundaries for harm (red inward sloping line) after the 7th trial, and again after the 17th trial (Figure 6). Accordingly, risk of random error in the finding can be excluded according to the Lan‐DeMetz‐O’Brien‐Fleming monitoring boundary. Trial Sequential Analysis‐adjusted intervention effect was RR 1.29 (CI 1.06 to 1.56).

Figure 6

Trial Sequential Analysis: non‐serious adverse events.

General behaviour

We were able to include in our analyses data on general behaviour from 7 parallel‐group trials and from 19 cross‐over trials. The evidence was assessed to be of very low quality (see GRADE assessment below).

Teacher‐rated general behaviour

Parallel‐group trials and cross‐over trials (first period data only)

Meta‐analysis suggested that methylphenidate significantly reduced teacher‐rated general behaviour compared with placebo (SMD ‐0.87, 95% CI ‐1.04 to ‐0.71; I² = 0%; 5 trials, 668 participants; Analysis 9.1). The SMD effect of ‐0.87 for general behaviour corresponds to an MD of 5.0 points on the CGI (Conners 1998a). However, the evidence was assessed to be of very low quality. All five trials were at high risk of bias (Analysis 9.1). Neither type of scale (Analysis 9.2) nor dose (Analysis 9.3) seemed to significantly influence the intervention effect. We were not able to test for subgroup differences according to duration, as all trials were of short duration, that is, less than six months (Analysis 9.4), or according to trial design, as all trials in the analysis were parallel‐group trials (Analysis 9.5).

Cross‐over trials (endpoint data)

Meta‐analysis suggested that methylphenidate significantly reduced teacher‐rated general behaviour compared with placebo (SMD ‐0.69, 95% CI ‐0.78 to ‐0.60; I² = 0%; 16 trials, 2014 participants). The intervention effect was not influenced by the dose of methylphenidate (Analysis 9.6). However, the evidence was assessed to be of very low quality. All 16 trials were at high risk of bias (Analysis 9.7).

Parallel‐group trials and cross‐over trials (endpoint data)

We observed comparable findings when combining data from parallel‐group trials with endpoint data from cross‐over trials (SMD ‐0.79, 95% CI ‐0.88 to ‐0.70; I² = 0%; 21 trials, 1976 participants; Analysis 9.8).

Independent assessor‐rated general behaviour

We found no parallel‐group trials that provided data on independent assessor‐rated general behaviour. However, for cross‐over trials, methylphenidate significantly reduced independent assessor‐rated general behaviour symptoms compared with placebo (SMD ‐0.60, 95% CI ‐0.75 to ‐0.46; I² = 32%; 8 trials, 1241 participants; Analysis 10.1). All trials were at high risk of bias (Analysis 10.2). We were not able to test for subgroup differences based on trial design, as all trials in the analysis were cross‐over trials (Analysis 10.3).

Parent‐rated general behaviour

Parallel‐group trials and cross‐over trials (first period data only)

Meta‐analysis suggested that methylphenidate significantly reduced parent‐rated general behaviour compared with placebo (SMD ‐0.53, 95% CI ‐0.78 to ‐0.27; I² = 42%; 6 trials, 670 participants; Analysis 11.1). All trials were deemed to have high risk of bias (Analysis 11.2). Furthermore, the intervention effect was significantly influenced by the type of scale (test for subgroup differences: Chi² = 8.52; df = 3 (P value = 0.04); I² = 64.8%; Analysis 11.3). However, no evidence was found to suggest that trial design influenced the intervention effect (Analysis 11.4). We were not able to test for subgroup differences according to duration, as all trials were of short duration, that is, less than six months (Analysis 11.5), or according to dose, as no low‐dose methylphenidate trials were conducted (Analysis 11.6).

Cross‐over trials (endpoint data)

Meta‐analysis suggested that methylphenidate significantly reduced parent‐rated general behaviour compared with placebo (SMD ‐0.75, 95% CI ‐0.93 to ‐0.56; I² = 7%; 6 trials, 550 participants; Analysis 11.7). The effect was not significantly influenced by the dose of methylphenidate (Analysis 11.7). All trials were at high risk of bias (Analysis 11.8).

Parallel‐group trials and cross‐over trials (endpoint data)

We observed comparable findings when combining data from parallel‐group trials with endpoint data from cross‐over trials (SMD ‐0.68, 95% CI ‐0.86 to ‐0.50; I² = 39%; 12 trials, 1054 participants; Analysis 11.9).

Additional subgroup analyses

We found significant differences between raters, with a higher intervention effect for teacher‐rated trials (SMD ‐0.87, 95% CI ‐1.04 to ‐0.71, I² = 0%; 5 trials, 668 participants) compared with parent‐rated trials (SMD ‐0.53, 95% CI ‐0.78 to ‐0.27; I² = 42%; 6 trials, 670 participants) (test for subgroup differences: Chi² = 5.10, df = 1 (P value = 0.02), I² = 80.4%; Analysis 12.1).

We found no evidence that comorbidity influences the intervention effect (Analysis 12.2), and we found no evidence of a 'carry‐over effect' in the cross‐over trials (Analysis 12.3).

No data were available for subgroup analyses by age, sex or type of ADHD.

Quality of life

We could include data on quality of life from only three parallel‐group trials in our analyses. The evidence was assessed to be of very low quality (see GRADE assessment below).

Meta‐analysis suggested that methylphenidate significantly improved quality of life compared with placebo (SMD 0.61, 95% CI 0.42 to 0.80; I² = 0%; 3 trials, 514 participants), and no evidence indicated that the type of rating scale ‐ all of which were parent‐ or clinician‐rated ‐ influenced the effect of the intervention (Analysis 13.1). The SMD of 0.61 for quality of life corresponds to an MD of 8.0 (95% CI 5.49 to 10.46) on the Child Health Questionnaire (CHQ; Landgraf 1998), which ranges from 0 to 100 points. All trials were at high risk of bias.

Discussion

Summary of main results

We included 38 parallel‐group trials and 147 cross‐over trials in this review. Altogether, these trials randomised more than 12,000 participants, and were reported in 449 publications. The majority compared methylphenidate with placebo in short‐term trials less than six months duration. The average trial duration in the 38 parallel trials was 75 days. Most were conducted in out‐patient clinics in high income countries, particularly the USA. Participants' ages ranged from 3 to 18 years across most studies (in two studies ages ranged from 3 to 21 years (Green 2011; Szobot 2008). Both boys and girls were recruited, in a ratio of 5:1 respectively.

All included trials were judged to be at high risk of bias. This raises important concerns, which are discussed following a summary of the results.

The two primary outcomes of this review were ADHD symptoms (with teacher‐rated outcomes taking precedence over outcomes reported by parents or independent raters), and number of serious adverse events. Secondary outcomes were non‐serious adverse events, general behaviour in school and at home, and quality of life.

Primary outcomes

ADHD symptoms

A meta‐analysis of data from parallel‐group trials combined with data from the first period of cross‐over trials suggests that methylphenidate may improve ADHD symptoms as reported by teachers (SMD ‐0.77, 95% CI ‐0.90 to ‐0.64; I² = 37%; 19 trials, 1698 participants; Analysis 1.1). This corresponds to a MD of ‐9.6 (95% CI ‐13.75 to ‐6.38) on the ADHD‐RS (DuPaul 1991a), which ranges from 0 to 72. Clinically, this represents a modest improvement in ADHD symptoms.

Subgroup analyses revealed no differences in the effects of methylphenidate based on age, comorbidity or subtypes of ADHD, though the paucity of trials meant we lacked the power to investigate these thoroughly. The quality of the evidence was judged to be ‘very low’ (see Quality of the evidence).

Serious adverse events

Methylphenidate does not appear to be associated with an increased occurrence of serious adverse events. However, data for this outcome were only available in 9 of the 185 included trials (4.9%) and the quality of the underpinning evidence was judged to be ‘very low’ (see Quality of the evidence).

Secondary outcomes

Non‐serious adverse events

Amongst those in the experimental groups, 526 per 1000 (range 448 to 615) experienced non‐serious adverse events, compared with 408 per 1000 of those in the control group. This equates to a 29% increase in the overall risk of any non‐serious adverse events (RR 1.29, 95% CI 1.10 to 1.51; I² = 73%; 21 trials, 3132 participants; very low‐quality evidence; Analysis 7.1). The most common non‐serious adverse events were sleep problems and decreased appetite. Children in the methylphenidate group were at 60% greater risk for trouble sleeping/sleep problems (RR 1.60, 95% CI 1.15 to 2.23; I² = 0%; 13 trials, 2416 participants; Analysis 7.3.18 in Analysis 7.3), and 266% greater risk for reduced appetite (RR 3.66, 95% CI 2.56 to 5.23; I² = 28%; 16 trials, 2962 participants; Analysis 7.4.1 in Analysis 7.4) than children in the control group.

The overall quality of the evidence for this outcome was also judged to be ‘very low’, and as a result, we are uncertain of the magnitude of the harmful effects. Further, for methodological reasons, we used only dichotomous outcomes reflecting the number of participants affected by the event per the total number of participants. As most participants reported more than one adverse event, the actual increase in risk of non‐serious adverse events may well be higher than the 29% calculated.

General behaviour

Meta‐analyses of data from five parallel‐group trials indicated that methylphenidate was associated with a significant improvement in children’s general behaviour, as reported by teachers (SMD ‐0.87, 95% CI ‐1.04 to ‐0.71; I² = 0%; 668 participants; Analysis 9.1). This effect corresponds to an MD of 5.0 points on the CGI (Conners 1998a). We could find no references describing a MIREDIF on this scale (range 0 to 30 points), and therefore cannot state anything about the clinical significance of this result. Comparable findings emerged from meta‐analyses of cross‐over trials (endpoint data) as reported by teachers, and from meta‐analyses of eight cross‐over trials (endpoint data) as rated by independent assessors. This evidence was judged also to be of ‘very low’ quality (see Quality of the evidence).

Quality of life

Changes in children’s quality of life were assessed in just three trials (n = 514), each of which used a different scale. Measures were scored by parents or clinicians, rather than children. The result (SMD 0.61, 95% CI 0.42 to 0.80; I² = 0%; Analysis 13.1) is equivalent to a MD of 8.0 points (95% CI 5.49 to 10.46) on the CHQ (Landgraf 1998), a change estimated to represent a clinically important improvement on this scale, which ranges from 0 to 100. The quality of the evidence was judged to be ‘very low’ (see Quality of the evidence).

The very low quality of the evidence, as assessed using the GRADE approach, undermines the confidence that can be placed in the magnitude of any effect. In particular, the prevalence of non‐serious adverse events raises questions about the effectiveness of blinding in these trials. If blinding was broken in just 20% or 30% of patients given methylphenidate, the resulting bias might well account for the small but statistical significant findings concerning the possible benefits of methylphenidate.

Overall completeness and applicability of evidence

This review highlights two major issues concerning the overall completeness and applicability of the evidence of the benefits and harms of methylphenidate for children with ADHD: the dearth of trials conducted in children and adolescents in low‐ and middle‐income countries, and the lack of follow‐up beyond six months. Here, we focus on the impact on the applicability of findings of decisions taken as part of this review (choice of rater for assessing change in ADHD symptoms and quality of life, choice of dose and diagnosis), together with issues relating to rating scales, diagnostic criteria, choice of comparators and adverse events.

ADHD symptoms ‐ choice of teacher report

We chose to use teacher‐rated outcomes as the primary measure for both ADHD symptoms and general behaviour, although a number of trials used or relied on parent reports. Some researchers have argued that parent evaluations of ADHD symptoms may not be as reliable as those of other raters such as teachers of pre‐school children (Murray 2007) or college students (Lavigne 2012). For example, Caye 2013 suggests inconsistency in ratings between parents, and in the MTA trial, information provided by parents was not always thought to be strong (Efstratopoulou 2013). We tested the robustness of our decision by conducting subgroup analyses and found no significant differences between this score and those of other raters.

Importantly, we do not really know what a lower score on an ADHD symptom scale (like that reported in this review) means for a child’s quality of life and ability to live, learn and function with other people.

Short‐term versus long‐term effects

Based on a subgroup analysis comparing 18 short‐term trials (≤ six months) with a single long‐term trial (> six months), we found that the treatment effect for teacher‐rated ADHD symptoms decreased over time (test for subgroup differences: P value = 0.02). This was not the case for independent assessor‐ and parent‐rated ADHD symptoms, for which no significant differences between short‐term and long‐term duration were found (P value = 0.09 and 0.53, respectively), but here the power of the subgroup analysis was limited.

We identified no trial that examined the effects of more extended exposure on children's general behaviour. Overall, evidence on the long‐term effects of methylphenidate for children and young people with ADHD is lacking, and it is possible that when used for longer periods, any beneficial effects may be diminished or offset by an increase in the risk of harm (Light 2015). Decisions to initiate and persist with treatment will need to weigh potential improvement in ADHD symptoms against adverse events, such as lack of sleep, since this may impact effects on quality of life and learning abilities. This review indicates that these important issues have not been studied sufficiently.

Quality of life

ADHD can exert a significant, negative impact on children’s quality of life, broadly defined. Yet only 7 of the 185 included trials measured quality of life in relation both to ADHD and life in general, and it was only possible to synthesize data from three of these trials. In each case the assessments were made by parents, teachers or independent assessors, rather than by children themselves. These external assessors observed small beneficial effects of methylphenidate on quality of life. Children might well have had different views on their own quality of life, and the failure to include child‐reported ratings of quality of life is a significant limitation on the completeness of the evidence. Furthermore, observations of quality of life reported by parents, teachers and independent assessors may be subject to both systematic and random errors.

Dose – choice of moderate/high dose

For children weighing 25 kg or less, the maximum recommended dose is 30 mg/d compared to 60 mg/d for children weighing more than 25 kg. After careful consideration, we renamed the high‐dose group as 'moderate/high' dose because doses are not always 'high' in heavier children.

Guidelines from the National Institute for Health and Care Excellence (NICE) recommend that methylphenidate can be increased to 0.7 mg/kg per dose up to three times a day, or a total daily dose of 2.1 mg/kg/d. European guidelines recommend that dosage should begin at a low level of 0.2 mg/kg per dose up to three times a day and should increase according to response, to a ceiling of 0.7 mg/kg per dose (up to three times a day), or a total daily dose of 60 mg/d.

In the parallel‐group trials included in this review, the mean dose of extended‐release methylphenidate was 41.0 mg/d, and the mean dose of immediate‐release methylphenidate was 23.1 mg/d. In the cross‐over trials included in this review, the mean reported total daily dose of immediate‐release methylphenidate was 25 mg/d and extended‐release methylphenidate 36 mg/d.

With average doses ranging between 23.1 mg/d and 41.0 mg/d, our cutoff of 20 mg/d seems low. However, many of the included trials were short‐term trials, involving medication‐naive children who consequently received lower doses. Furthermore, many of the cross‐over trials used only morning and midday doses to achieve a cross‐over for trial purposes, with no afternoon dose given. However, extended‐release methylphenidate is designed to reduce symptoms in the late afternoon too, so the average expected daily dose would be higher.

We performed subgroup analyses to test differences in the estimate of effect based on differences in dosage. These analyses revealed no differences between low doses (≤ 20 mg/d) and moderate/high doses (> 20 mg/d) of methylphenidate. Given the many adverse events that can result when this medication is used, evidence suggests that higher doses may not be needed.

Rating scales

This review included trials from several countries conducted between 1981 and 2014. Pioneers in ADHD research conduct trials in different countries, and psychometric instruments change with trends over time; this is reflected in the variety of rating scales used by investigators in the included trials. Scales based on the diagnostic criteria of the DSM and the ICD measure slightly different constructs. We found significant differences between scales measuring ADHD symptoms, but not between scales measuring general behaviour; we found fewer differences when we performed sensitivity analyses in which we pooled subgroups of scales measuring the same ADHD subtype (e.g. scales measuring the inattentive subtype). All trials using subjective rating scales as proxy measures of outcomes are affected by these problems.

Diagnostic criteria

The diagnostic criterion as regards the age by which ADHD symptoms should first be observed has been much debated (APA 2013; Todd 2008). As regards DSM‐IV age of onset (APA 1994), trials have found no differences in phenotype of neuropsychological impairment, course or treatment response, whether children were diagnosed before or after seven years of age (Kieling 2010). By limiting age of onset to six years or younger, the ICD‐10 criteria for hyperkinetic disorder may under‐identify children with persistent ADHD symptoms and related impairment (Lahey 2006). In addition, extending age of onset is not implicated in prevalence changes (Polanczyk 2010). It is important to highlight that age of onset is considered only for diagnostic scales, not for rating scales.

The criteria of both the DSM‐IV (APA 1994) and the current DSM‐5 (APA 2013) encompass a broader spectrum of children with ADHD when compared with the criteria for hyperkinetic disorder in the ICD‐10 (WHO 1992); whilst the criteria are virtually identical (Lee 2008; Tripp 1999), the ICD‐10 criteria are more restrictively applied. Nonetheless, all have an equal ability to predict ADHD/hyperkinetic disorder, even in the presence of comorbidity (Schachar 2007). Consequently, we believe that our efficacy results would not be changed by any change in diagnostic criteria.

Comparators

The majority of trials in this review compared methylphenidate with placebo, and we earlier highlighted the problems of threats to blinding in these trials, due to the prevalence of non‐serious adverse events of methylphenidate. Trials that assess methylphenidate using an ‘active placebo’ (or 'nocebo tablets' ‐ tablets with a placebo‐like substance that causes similar adverse events as in the experimental drug arm), can strengthen double blinding and are recommended (Jakobsen 2013; Jakobsen 2014; Moncrieff 2004). We identified no such trials, and as the use of nocebo tablets in all diseases is ethically questionable, any decision to conduct nocebo tablet controlled trials in children should be deferred pending the results of such trials in adults. If these show that methylphenidate is superior to nocebo in treating ADHD symptoms, a rationale would exist for conducting such trials in children. (If another rationale could be mounted, nocebo controlled trials in children would not be ethically questionable, but currently no such case is evident to us).

Adverse events

Methylphenidate non‐responders, placebo responders and/or participants with methylphenidate adverse events before randomisation were excluded in 10 parallel‐group trials and in 49 cross‐over trials. The intervention effect of methylphenidate in these trials was compared with that in the remaining trials in subgroup analyses (Analysis 1.7; Analysis 2.5; Analysis 3.6), which found no differences in terms of the intervention effect of methylphenidate ‐ a surprising finding. However, some of our included trials involved participants who were not medication naive before randomisation, which may have exaggerated the benefits of methylphenidate. They might have detected the physiological effects (for example, improved concentration or adverse events such as appetite suppression) through prior exposure to the effects of methylphenidate. To investigate this, we performed post hoc subgroup analyses and found that effects of methylphenidate were different in trials involving medication‐naive participants (> 80% of included participants were medication naive) than in trials involving participants already taking methylphenidate before randomisation (< 20% of included participants were medication naive) for teacher‐rated ADHD symptoms (P value = 0.04), but not for parent‐rated ADHD symptoms (P value = 0.52). One might expect the issue of prior exposure to be of greatest concern in cross‐over trials. However, we found no differences between parallel‐group trials and cross‐over trials in both ADHD teacher‐rated and parent‐rated outcomes. Consequently, we believe that prior exposure is not a major concern when effects of methylphenidate are assessed.

We are currently preparing a second systematic review for the purpose of assessing harms of methylphenidate in observational trials with a duration of up to 36 months (Storebø in press). Preliminary results from this review show a small proportion of serious adverse events reported after treatment with methylphenidate, but one case control trial highlights the risk of sudden death for adolescents (Gould 2009). Just over a quarter of children appear to experience non‐serious adverse events after methylphenidate treatment.

Many claims have been made about significant increases in global rates of methylphenidate prescribing; this drug usually is prescribed for long‐term use and seldom with medication‐free periods. However, a recent paper reports that children in primary care in the UK did not continue methylphenidate treatment for longer than six months (Raman 2015). Furthermore, the incidence of ADHD diagnoses in the UK fell between 1998 and 2010 (Holden 2013). In the USA, however, almost 70% of children with ADHD, estimated at 6.4 million children, take medication (Visser 2014). This might mean that clinicians in the UK are more cautious about using methylphenidate, while US clinicians assume that evidence for the safe use of methylphenidate is sound.

Our assessment of the evidence does not preclude that individual patients may benefit from intervention with methylphenidate. However, despite more than 50 years of research in this field, we do not yet know how to identify those patients that may obtain more benefits than harms. Individual patient data meta‐analyses are needed to try to identify such patient characteristics.

Quality of the evidence

As assessed by the GRADE approach, the overall quality of evidence in this review is 'very low' because of high risk of bias (including loss of blinding (explained below) and outcome reporting bias), inconsistency, and indirectness; this led to uncertainty in the robustness of our estimates. We rated the risk of outcome reporting bias for adverse events to be high, as we only managed to obtain data on total serious adverse events from 9 of the 185 included trials, and on total non‐serious adverse events from 21 of the 185 included trials. We found no evidence for publication bias.

It is likely that the trials initially judged to be at low risk of bias may, in fact, be trials at high risk of bias because methylphenidate gives rise to various prevalent and easily recognisable adverse events, which can lead to loss of blinding and hence can bias the ratings of symptoms, resulting in an overestimation of benefits and an underestimation of harms (Kjaergard 2001; Savović 2012b; Wood 2008). To ensure adequate blinding, it is therefore important for researchers to use a nocebo ('active placebo') in the control group. As we found no trials employing nocebo tablets in the control group, the extent of this bias cannot be assessed. The fact that the intervention effect of methylphenidate on ADHD symptoms did not differ significantly between trials at low risk of bias compared with trials at high risk of bias may be taken as an indication that deblinding has occurred among former trials.

Also, the average duration of treatment was no longer than about two months. Therefore, little can be concluded about the benefits and harms of methylphenidate used for longer than six months.

Potential biases in the review process

The present systematic review has many strengths. We developed a protocol for this review according to instructions provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Our protocol was published before we embarked on the review itself. We conducted extensive searches of relevant databases, and we requested published and unpublished data from pharmaceutical companies manufacturing methylphenidate, including Shire, Medice (represented in Denmark by HB Pharma), Janssen‐Cilag and Novartis. Two review authors, working independently, selected trials for inclusion and extracted data. Disagreements were resolved by discussion with team members. We assessed risk of bias in all trials according to the recommendations provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We conducted Trial Sequential Analyses to control the risk of type I errors and to estimate how far we were from obtaining the DARIS to detect or reject a certain plausible intervention effect. In the meta‐analyses on non‐serious adverse events, the Trial Sequential Analysis showed that observed intervention effects were not likely to be due to type I error and confirmed that sufficient data had been obtained.

We excluded 78 trials described in 88 reports, which assessed the effects of methylphenidate on specialised outcomes (e.g. experimental/neurocognitive/functional outcomes) in children or adolescents with ADHD (see Characteristics of excluded studies). This raises the issue of bias in our review process as we did not write to these authors asking whether they collected data on other outcomes. This potential bias, however, is not likely to change our conclusions. Another limitation of this review is that we did not search the Food and Drug Administration (FDA) and European Medicines Agency (EMA) homepages for unpublished data (Schroll 2015). However, it is also unlikely that these searches would have changed our results.

Agreements and disagreements with other studies or reviews

Over the past 15 years, several reviews investigating the efficacy of methylphenidate for ADHD (with or without meta‐analyses) have been published. Some 15 reviews have pooled results derived from ADHD rating scales on the efficacy of methylphenidate treatment for children and adolescents with ADHD (Bloch 2009; Charach 2011; Charach 2013; Faraone 2002; Faraone 2006; Faraone 2009; Faraone 2010; Hanwella 2011; Kambeitz 2014; King 2006; Maia 2014; Punja 2013; Reichow 2013; Schachter 2001; Van der Oord 2008). Each of these reviews, however, has several shortcomings and these are described in detail in the subsection Why it is important to do this review. The most important concerns are that none of these reviews are based on a pre‐published protocol, and most assessed neither the risk of bias of included trials nor adverse events. Moreover, none of these reviews considered the risk of random errors. Therefore, their interpretation of findings is unlikely to have taken into account the poor reporting of adverse events, the impact of combining data from small trial samples, or the impact of risk of bias on their analyses; information about adverse events is also missing from several RCTs.

A recent Cochrane systematic review evaluated the effects of methylphenidate in adults with ADHD (Epstein 2014). The effect sizes across the different assessments of symptoms were similar to those found in our analyses (SMD 0.60). The authors noted that data on adverse events were limited by the short duration of the included trials (Epstein 2014). Despite the similar effects of methylphenidate on symptoms observed in our reviews, readers will notice that we have judged the quality of evidence in our review, as well as that in Epstein 2014 (Storebø 2015b [pers comm]), to be lower than Epstein and colleagues.

In order to address uncertainties over the frequency and severity of long‐term adverse events of methylphenidate, we are currently preparing a second systematic review of observational studies (Storebø in press), which will investigate the long‐term harms of methylphenidate as reported by non‐randomised studies with a duration of up to 36 months. We hope this will help us to understand the long‐term safety profile of the drug.

Figure 1

Flow chart.

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included trial.

Figure 4

Funnel plot of comparison: 1. Teacher‐rated ADHD symptoms, outcome: 1.8 All data at low and high risk of bias (parallel‐group and cross‐over trials).

Figure 5

Trial Sequential Analysis: serious adverse events.

Figure 6

Trial Sequential Analysis: non‐serious adverse events.

Analysis 1.1

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 1 All parallel‐group trials and first‐period cross‐over trials.

Analysis 1.2

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 2 Subgroup analysis: types of scales.

Analysis 1.3

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 3 Medication status: medication naive versus not medication naive.

Analysis 1.4

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 4 Subgroup analysis: duration of treatment.

Analysis 1.5

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 5 Subgroup analysis: dose.

Analysis 1.6

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 6 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials.

Analysis 1.7

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 7 Subgroup analysis: trials with cohort selection bias of all participants compared with trials without cohort selection bias of all participants.

Analysis 1.8

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 8 ADHD symptoms, cross‐over trial (endpoint data).

Analysis 1.9

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 9 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: dose.

Analysis 1.10

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 10 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials compared with cross‐over trials (endpoint data).

Analysis 1.11

Comparison 1 Teacher‐rated ADHD symptoms, Outcome 11 All parallel‐group trials and cross‐over trials: risk of bias.

Analysis 2.1

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 1 All parallel‐group trials and first‐period cross‐over trials.

Analysis 2.2

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 2 Subgroup analysis: types of scales.

Analysis 2.3

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 3 Subgroup analysis: duration of treatment.

Analysis 2.4

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 4 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials.

Analysis 2.5

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 5 Subgroup analysis: trials with cohort selection bias of all participants compared with trials without cohort selection bias of all participants.

Analysis 2.6

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 6 Subgroup analysis: dose.

Analysis 2.7

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 7 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: risk of bias.

Analysis 2.8

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 8 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: dose.

Analysis 2.9

Comparison 2 Independent assessor‐rated ADHD symptoms, Outcome 9 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials compared with cross‐over trials (endpoint data).

Analysis 3.1

Comparison 3 Parent‐rated ADHD symptoms, Outcome 1 All parallel‐group trials and first‐period cross‐over trials.

Analysis 3.2

Comparison 3 Parent‐rated ADHD symptoms, Outcome 2 Subgroup analysis: types of scales.

Analysis 3.3

Comparison 3 Parent‐rated ADHD symptoms, Outcome 3 Subgroup analysis: duration of treatment.

Analysis 3.4

Comparison 3 Parent‐rated ADHD symptoms, Outcome 4 Subgroup analysis: dose.

Analysis 3.5

Comparison 3 Parent‐rated ADHD symptoms, Outcome 5 Medication status: medication naive versus not medication naive.

Analysis 3.6

Comparison 3 Parent‐rated ADHD symptoms, Outcome 6 Subgroup analysis: trials with cohort selection bias of all participants compared with trials without cohort selection bias of all participants.

Analysis 3.7

Comparison 3 Parent‐rated ADHD symptoms, Outcome 7 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials.

Analysis 3.8

Comparison 3 Parent‐rated ADHD symptoms, Outcome 8 ADHD symptoms, cross‐over trials (endpoint data).

Analysis 3.9

Comparison 3 Parent‐rated ADHD symptoms, Outcome 9 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: dose.

Analysis 3.10

Comparison 3 Parent‐rated ADHD symptoms, Outcome 10 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials compared with cross‐over trials (endpoint data).

Analysis 4.1

Comparison 4 Additional subgroup analyses of ADHD symptoms: parallel‐group trials and first‐period cross‐over trials, Outcome 1 Comparision of raters.

Analysis 4.2

Comparison 4 Additional subgroup analyses of ADHD symptoms: parallel‐group trials and first‐period cross‐over trials, Outcome 2 Age.

Analysis 4.3

Comparison 4 Additional subgroup analyses of ADHD symptoms: parallel‐group trials and first‐period cross‐over trials, Outcome 3 Comorbidity versus no comorbidity.

Analysis 4.4

Comparison 4 Additional subgroup analyses of ADHD symptoms: parallel‐group trials and first‐period cross‐over trials, Outcome 4 Subtypes ADHD: ADHD Rating Scale (parent‐, teacher‐ or independent assessor‐rated).

Analysis 4.5

Comparison 4 Additional subgroup analyses of ADHD symptoms: parallel‐group trials and first‐period cross‐over trials, Outcome 5 Cross‐over trials: first‐period data versus endpoint data (parent‐, independent assessor‐ and teacher‐rated).

Analysis 5.1

Comparison 5 Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 1 Number of serious adverse events (SAE).

Analysis 5.2

Comparison 5 Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 2 Nervous system.

Analysis 5.3

Comparison 5 Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 3 Digestive system: gastrointestinal disorders.

Analysis 5.4

Comparison 5 Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 4 Urinary system: kidney infection.

Analysis 5.5

Comparison 5 Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 5 Circulatory and respiratory systems: asthma.

Analysis 5.6

Comparison 5 Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 6 Immune system: cyst rupture.

Analysis 5.7

Comparison 5 Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 7 Other: drug toxicity.

Analysis 6.1

Comparison 6 Number of serious adverse events: cross‐over trials (endpoint data), Outcome 1 Number of serious adverse events (SAE).

Analysis 6.2

Comparison 6 Number of serious adverse events: cross‐over trials (endpoint data), Outcome 2 Hallucinations/psychosis.

Analysis 7.1

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 1 Total number of non‐serious adverse events.

Analysis 7.2

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 2 Subgroup analysis: total number of non‐serious adverse events according to dose.

Analysis 7.3

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 3 Nervous system.

Analysis 7.4

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 4 Digestive system.

Analysis 7.5

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 5 Circulatory and respiratory systems.

Analysis 7.6

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 6 Skeletal and muscular systems.

Analysis 7.7

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 7 Immune system.

Analysis 7.8

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 8 Height.

Analysis 7.9

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 9 Weight.

Analysis 7.10

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 10 BMI.

Analysis 7.11

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 11 Vital signs.

Analysis 7.12

Comparison 7 Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials, Outcome 12 Other.

Analysis 8.1

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 1 Total number of non‐serious adverse events.

Analysis 8.2

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 2 Subgroup analysis: total number of non‐serious adverse events according to dose.

Analysis 8.3

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 3 Nervous system.

Analysis 8.4

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 4 Digestive system.

Analysis 8.5

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 5 Urinary system: urinary incontinence.

Analysis 8.6

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 6 Skeletal and muscular system: somatic complaints.

Analysis 8.7

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 7 Immune system.

Analysis 8.8

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 8 Skin.

Analysis 8.9

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 9 Vital signs.

Analysis 8.10

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 10 Height (cm).

Analysis 8.11

Comparison 8 Number of non‐serious adverse events: cross‐over trials (endpoint data), Outcome 11 Weight.

Analysis 9.1

Comparison 9 Teacher‐rated general behaviour, Outcome 1 All parallel‐group trials and first‐period cross‐over trials: risk of bias.

Analysis 9.2

Comparison 9 Teacher‐rated general behaviour, Outcome 2 Subgroup analysis: types of scales.

Analysis 9.3

Comparison 9 Teacher‐rated general behaviour, Outcome 3 Subgroup analysis: dose.

Analysis 9.4

Comparison 9 Teacher‐rated general behaviour, Outcome 4 Subgroup analysis: duration of treatment.

Analysis 9.5

Comparison 9 Teacher‐rated general behaviour, Outcome 5 Subgroup analysis: parallel‐group trials versus first‐period cross‐over trials.

Analysis 9.6

Comparison 9 Teacher‐rated general behaviour, Outcome 6 General behaviour, cross‐over trials (endpoint data).

Analysis 9.7

Comparison 9 Teacher‐rated general behaviour, Outcome 7 Subgroup analysis: general behaviour, cross‐over trials (endpoint data): risk of bias.

Analysis 9.8

Comparison 9 Teacher‐rated general behaviour, Outcome 8 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials (teacher‐rated) versus cross‐over trials (endpoint data).

Analysis 10.1

Comparison 10 Independent assessor‐rated general behaviour, Outcome 1 General behaviour, cross‐over trials (endpoint data).

Analysis 10.2

Comparison 10 Independent assessor‐rated general behaviour, Outcome 2 Subgroup analysis: general behaviour, cross‐over trials (endpoint data): risk of bias.

Analysis 10.3

Comparison 10 Independent assessor‐rated general behaviour, Outcome 3 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials (independent assessor‐rated) compared with cross‐over trials (endpoint data).

Analysis 11.1

Comparison 11 Parent‐rated general behaviour, Outcome 1 All parallel‐group trials and first‐period cross‐over trials.

Analysis 11.2

Comparison 11 Parent‐rated general behaviour, Outcome 2 Subgroup analysis: risk of bias.

Analysis 11.3

Comparison 11 Parent‐rated general behaviour, Outcome 3 Subgroup analysis: types of scales.

Analysis 11.4

Comparison 11 Parent‐rated general behaviour, Outcome 4 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials.

Analysis 11.5

Comparison 11 Parent‐rated general behaviour, Outcome 5 Subgroup analysis: duration of treatment.

Analysis 11.6

Comparison 11 Parent‐rated general behaviour, Outcome 6 Subgroup analysis: dose.

Analysis 11.7

Comparison 11 Parent‐rated general behaviour, Outcome 7 General behaviour, cross‐over trials (endpoint data).

Analysis 11.8

Comparison 11 Parent‐rated general behaviour, Outcome 8 Subgroup analysis: general behaviour, cross‐over trials (endpoint data): risk of bias.

Analysis 11.9

Comparison 11 Parent‐rated general behaviour, Outcome 9 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials (parent‐rated) compared with cross‐over trials (endpoint data).

Analysis 12.1

Comparison 12 Additional subgroup analyses of general behaviour: parallel‐group trials and first‐period cross‐over trials, Outcome 1 Comparisions of raters.

Analysis 12.2

Comparison 12 Additional subgroup analyses of general behaviour: parallel‐group trials and first‐period cross‐over trials, Outcome 2 Comorbidity versus no comorbidity.

Analysis 12.3

Comparison 12 Additional subgroup analyses of general behaviour: parallel‐group trials and first‐period cross‐over trials, Outcome 3 Cross‐over trials: first‐period data versus endpoint data (teacher‐, parent‐, and independent assessor‐rated).

Analysis 13.1

Comparison 13 Quality of life: parallel‐group trials and first‐period cross‐over trials, Outcome 1 Subgroup analysis: types of scales.

Summary of findings for the main comparison. Methylphenidate compared with placebo or no intervention for ADHD

Methylphenidate compared with placebo or no intervention for ADHD
Patient or population: children and adolescents (up to and including 18 years of age) with ADHD Settings: out‐patient clinic, in‐patient hospital ward and summer school Intervention: methylphenidate Comparison: placebo or no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Placebo or no intervention	Methylphenidate
ADHD symptoms: all parallel‐group trials and first‐period cross‐over trials ADHD Rating Scale (Teacher‐rated) Average study duration: 74.8 days		Mean ADHD symptom score in the intervention groups corresponds to a mean difference of 9.6 (95% CI 11.25 to 8.00) on ADHD Rating Scale	SMD ‐0.77 (‐0.90 to ‐0.64)	1698 (19 studies)	⊕⊝⊝⊝ Very low^a,b	The analysis was conducted on a standardised scale with data from studies that used different teacher‐rated scales of symptoms (Conners' Teacher Rating Scale (CTRS), Strengths and Weaknesses of ADHD Symptoms and Normal Behaviour (SWAN) Scale, Schedule for Non‐adaptive and Adaptive Personality (SNAP) ‐ Teacher, Fremdbeurteilungsbogen für Hyperkinetische Störungen (FBB‐HKS)). The effect size has been translated on to the ADHD Rating Scale from the SMD
Total number of serious adverse events	Trial population		RR 0.98 (0.44 to 2.22)	1532 (9 studies)	⊕⊝⊝⊝ Very low^a,c	‐
	16 per 1000	16 per 1000 (7 to 36)
Total number of non‐serious adverse events	Trial population		RR 1.29 (1.10 to 1.51)	3132 (21 studies)	⊕⊝⊝⊝ Very low^a,b	‐
	408 per 1000	526 per 1000 (448 to 615)
General behaviour: all parallel‐group trials and first‐period cross‐over trials General behaviour rating scales (Teacher‐rated)		Mean general behaviour score in the intervention groups was 0.87 standard mean deviations lower (95% CI 1.04 to 0.71 lower)	SMD ‐0.87 (‐0.71 to ‐1.04)	668 (5 studies)	⊕⊝⊝⊝ Very low^a,d	‐
Quality of life (Parent‐rated)		Mean quality of life score in the intervention groups corresponds to a mean difference of 8.0 (95% CI 5.49 to 10.46) on the Child Health Questionnaire	SMD 0.61 (0.42 to 0.80)	514 (3 studies)	⊕⊝⊝⊝ Very low^a,e	‐
The basis for the assumed risk* (e.g. median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). ADHD: Attention deficit hyperactivity disorder; CI: Confidence interval; RR: Risk ratio; SMD: Standardised mean difference
GRADE Working Group grades of evidence. High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
^aDowngraded two levels due to high risk of bias (systematic errors causing overestimation of benefits and underestimation of harms) in several risk of bias domains, including lack of sufficient blinding and selective outcome reporting (many of the included trials did not report on this outcome). ^bDowngraded one level due to inconsistency: moderate statistical heterogeneity. ^c Downgraded one level due to imprecision: wide confidence intervals. ^dDowngraded one level due to indirectness: children's general behavior was assessed by different types of rating scales with different focus on behavior. e Downgraded one level due to indirectness: children's quality of life was assessed by their parents.

Summary of findings for the main comparison. Methylphenidate compared with placebo or no intervention for ADHD

Table 1. ADHD symptoms ‐ rating scales

Name of scale	Abbreviation	Reference
Abbreviated Conners’ Rating Scales, Parent (ACPRS) and Teacher (ACTRS), including Abbreviated Parent Rating Scale (APRS) and Teacher Rating Scale, Hyperkinesis Index and ADHD and Emotional Lability subscales	ACRS	Conners 1997a
Abbreviated Symptom Questionnaire, including ASQ Teacher and ASQ Parent	ASQ	Conners 1995
Academic Performance Rating Scale	APRS	DuPaul 1991a
The ADD/H Comprehensive Teacher Rating Scale	ACTeRS	Ullmann 1984
ADHD/ODD Rating Scale, Parent‐ and Teacher‐Rated	ADHD‐RS	Barkley 1998
ADHD Rating Scale, including ADHD Rating Scale Parent and Teacher Ratings	ADHD‐RS	DuPaul 1991a
ADHD Rating Scale‐IV, including ADHD Rating Scale‐IV Parent and Teacher Versions	ADHD‐RS‐IV	DuPaul 1991a
Brief Psychiatric Rating Scale for Children	BPRS	Gale 1986
Child Attention Problems Rating Scale	CAP	Achenbach 1986
Child Attention Profile	CAP	Barkley 1988b
Child Behavior Rating Form	NCBHF	Aman 1996
Child Symptom Inventory	CSI	Gadow 1994
Children’s Psychiatric Rating Scale	CPRS	Pfefferbaum‐Levine 1983
Conners’ Abbreviated Hyperactivity Questionnaire	C‐HI	Conners 1997a
Conners’ Abbreviated Questionnaire	ASQ	Conners 1995
Conners’ Abbreviated Parent Teacher Questionnaire	APTQ	Rowe 1997
Conners’ Abbreviated Rating Scale	ABRS	Conners 1997a
Conners’ Abbreviated Symptom Questionnaire	ASQ	Conners 1995
Conners Abbreviated Symptom Questionnaire for Parents	ASQ‐Parent	Conners 1995
Conners’ Abbreviated Symptom Questionnaire for Teachers	ASQ‐Teacher	Conners 1997a
Conners’ Abbreviated Teacher Rating Scale	ABTRS	Conners 2001
Conners’ ADHD/DSM‐IV Scales Adolescent	CADS‐A	Conners 1997b
Conners’ ADHD/DSM‐IV Scales Parent	CADS–P, CADS‐P DSM‐IV	Conners 1997a
Conners’ ADHD/DSM‐IV Scale Teacher, including Inattentive and Hyperactive‐Impulsive subscales	CADS‐T, CADS‐T DSM‐IV	Conners 1997a
Conners’ Rating Scale ‐ Revised, Parent and Teacher: Hyperactivity and Conduct Factors score	CPRS‐R and CTRS‐R	Goyette 1978
Conners’ Hyperactivity Index, Parent and Teacher, including abbreviated versions	CPRS/CTRS‐Hyperactivity index	Conners 1997a
Conners’ Hyperkinesis Index	‐	Milich 1980
Conners, Loney and Milich Scale	CLAM	Milich 1980
Conners’ Parent and Teacher Rating Scale ‐ Revised, Short Form	CRS‐R:S	Conners 1997a
Conners’ Parent Rating Scale, including abbreviated versions	CPRS	Conners 1998b
Conners’ Parent Rating Scale ‐ Revised	CPRS‐R	Conners 1997a
Conners’ Parent Rating Scale ‐ Revised, Short Form	CPRS‐R:S	Conners 1997a
Conners’ Parent Rating Scale ‐ Revised, Long Version	CPRS‐R:L	Conners 1997a
Conners’ Rating Scale ‐ Revised	CRS‐R	Conners 1997a
Conners’ Short Form Rating Scale, Parent and Teacher	‐	Conners 1997a
Conners’ Teacher Rating Scale	CTRS	Conners 1998a
Conners’ Teacher Rating Scale ‐ Revised, Long Version	CTRS‐R:L	Conners 1998a
Diagnostic and Statistical Manual of Mental Disorders Total	DSM‐IV	APA 1994
Diagnostiksystem für Psychische Störungen im Kindes ‐ und Jugendalter nach ICD‐10 und DSM‐IV, Parental Questionnaire of ADHD symptoms	DISYPS	Döpfner 2000
Fremdbeurteilungsbogen für Hyperkinetische Störungen	FBB‐HKS	Döpfner 2008
German Teacher’s report on ADHD symptoms	FBB‐HKS of the DISYPS	Döpfner 2000
Hyperactivity Index of the Revised Conners Parent and Teacher Rating Scales	‐	Goyette 1978
IOWA Conners Parent Rating Scale, including abbreviated versions	IOWA CPRS	Loney 1982
IOWA Conners Teacher Rating Scale, including abbreviated versions	IOWA CTRS	Loney 1982
IOWA Conners Teacher Rating Scale, Inattention/Overactivity (I/O) and Oppositional/Defiant (O/D) subscales	IOWA‐I/O and O/D subscales	Loney 1982
IOWA Inattention/Overactivity and Aggression/Noncompliance scales ‐ Parent and Teacher rating	IOWA	Loney 1982
Lehrer‐Fragenbogen von Steinhausen	LF	Steinhausen 1993
Loney’s Time on Task Scale, Hyperactivity, Attention and Aggression subscales	TOTS	Fitzpatrick 1992b
Modified Conner Scale Parent and Teacher	ACR	Conners 1997a
Mothers’ Objective Method for Subgrouping	MOMS	Loney 1984
Parent Symptom Checklist	PSC ADHD	Döpfner 2000
Parental Account of Children’s Symptoms	PACS	Chen 2006
Restricted Academic Situation Scale	RASS	Fischer 1998
Schedule for Affective Disorders and Schizophrenia	K‐SADS/ K‐SADS‐E for diagnosis	Chambers 1985
Schedule for Non‐adaptive and Adaptive Personality	SNAP	Clark 1993; Clark 1996
Swanson, Nolan, and Pelham ‐ IV SNAP‐ADHD Rating scale	SNAP‐ADHD	Swanson 1992
Swanson, Nolan, and Pelham ‐ IV SNAP‐IV (Brazilian Version)	SNAP‐IV	Clark 1993; Clark 1996
Swanson, Kotkin, Atkins, M‐Flynn, Pelham Scale (SKAMP combined, SKAMP attention, and SKAMP deportment)	SKAMP (SKAMP combined, SKAMP attention, and SKAMP deportment)	Wigal 1998; Murray 2009
Teacher Self‐control Rating Scale	SCRS	Kendall 1979
Turgay ‐ DSM‐IV Scale, Parent	T‐DSM‐IV Scale, Parent	Turgay 1994; Ercan 2001
Turgay ‐ DSM‐IV Scale, Teacher	T‐DSM‐IV Scale, Teacher	Turgay 1994; Ercan 2001
Teacher Hyperactivity Index	THI	Achenbach 1991b
Teacher Symptom Checklist	TSC	Döpfner 2000
Vanderbilt ADHD Rating Scale	VADP(T)RS	Wolraich 2003
Wender Utah Rating Scale	WURS	Ward 1993
Wide Range Achievement Test	WRAT‐4	Wilkinson 2006
Wide Range Achievement Test Revised	WRAT‐R	Woodcock 2001
ADD/H: Attention deficit disorder with hyperactivity. ADHD: Attention deficit hyperactivity disorder. DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition. ODD: Oppositional defiant disorder.

Table 1. ADHD symptoms ‐ rating scales

Table 2. Table 2: General behaviour rating scales

Name of scale	Abbreviation	Reference
Achenbach Child Behaviour Checklist	CBCL	Achenbach 1991a
Achenbach’s Teacher Report	ATRF	Achenbach 1991b; Achenbach 2001
ADHD Rating Scale	ADHD‐RS	DuPaul 1991a
ADHD School Observation Code	ADHD‐SOC	Gadow 1996
Barkley Scales, Disruptive Behavior Disorders Rating Scale	‐	Barkley 1991a
Child Attention Problems Scale	CAP	Barkley 1991
Child Attention Profile	CAP	Barkley 1988b
Child Behavior Checklist	CBCL	Achenbach 1991a
Child Health Questionnaire	CHQ	Landgraf 1998
Child and Adolescent Psychiatric Assessment, selected items	CAPA	Angold 1995
Children’s Psychiatric Rating Scale	CPRS	Fish 1985
Classroom Observation Code (Abikoff Classroom Observational System)	COC	Abikoff 1980
Code for Observing Social Activity	COSA	Sprafkin 1986
Conners' Child Behavior Scale	UC‐CCBS	Ladd 1996
Conners' Global Index Scale	CGI‐S	Conners 1998a
Conners’ Global Index ‐ Parent	CGI‐P	Conners 1997a
Conners' Global Index ‐ Teacher	CGI‐T	Conners 1998a
Conners', Loney and Milich Scale	CLAM	Milich 1980
Conners’ Parent Questionnaire	CPQ	Conners 1995
Conners’ Parent Rating Scale	CPRS	Conners 1998b
Conners’ Teacher Rating Scale	CTRS	Conners 1998a
Conners’ Teacher Rating Conduct Problems	‐	Miller 1997
Disruptive Behavior Disorders Rating Scale, Parent‐ and Teacher‐Rated	DBS	Mendelsohn 1978
Disruptive Behavior Disorders Rating Scale	DBD	Silva 2005b
Groninger Behaviour Observation Scale	GOO and GBO	Van der Meere 1999b
Groninger Behaviour Checklists, Parent and Teacher Versions of the abbreviated Groninger	GGGS and GGBS	Van der Meere 1999b
Hillside Behavior Rating Scale	HBRS	Gittleman‐Klein 1976
Home Situations Questionnaire	HSQ	Barkley 1987
Home Situations Questionnaire ‐ Revised	HSQ‐R	DuPaul 1992
Humphrey’s Teacher Self‐Control Rating Scale	TSCRS	Humphrey 1982
Hyperactivity Index from the Conners Revised Teacher Rating Scale	CTRS‐R‐Hyperactivity Index	Goyette 1978
Impairment Rating Scale	IRS	Fabiano 2006
Inpatient Global Rating Scale, Revised	IGRS	Conners 1985
Inpatient Global Rating Scale, Somatic factor	IGRS‐S	Conners 1985
IOWA Conners' Rating Scale, Oppositional/Defiant (O/D) subscales	IOWA‐O/D subscales	Loney 1982
Nisonger Child Behavior Rating Form	NCBRF	Aman 1996
Paired Associates Learning	PAL	Wechsler 1945
Parent Global Assessment for Improvement	PGA	McGough 2006a
Peer Conflict Scale	PCS	Marsee 2007
Personality Inventory for Children	PIC	Lachar 1986
School Situations Questionnaire	SSQ	Barkley 1987
School Situations Questionnaire ‐ Revised	SSQ‐R	DuPaul 1992
Schedule for Nonadaptive and Adaptive Personality	SNAP	Clark 1993; Clark 1996
Strengths and Weaknesses of ADHD Symptoms and Normal Behavior Scale, Parent and Teacher	SWAN	Swanson 2006; Polderman 2007
Subjective Treatment Emergent Symptom Scale	STESS‐R	Guy 1976
Swanson, Nolan and Pelham, Fourth Edition	SNAP‐IV	Bussing 2008
Teachers Report Form	TRF	Achenbach 1991b
Telephone Interview Probe (Parent and Teacher)	TIP	Corkum 2007
Vanderbilt ADHD rating scales: Vanderbilt ADHD Diagnostic Parent Rating Scale and Vanderbilt ADHD Diagnostic Teacher Rating Scale	VADPRS and VADTRS	Wolraich 2003
Wahler, House and Stambaugh’s Ecobehavioral Assessment System	ECO	Wahler 1976
The Weekly Parent Ratings of Evening and Morning Behaviour	WREMB‐R	Kelsey 2004
Werry‐Weiss‐Peters Activity Rating Scale	WWP	Routh 1978
Woodcock‐Johnson Achievement Battery	WJ‐III Ach	Woodcock 2001
ADHD: attention deficit hyperactivity disorder.

Table 2. Table 2: General behaviour rating scales

Table 3. Table 3: Quality of life ratings scales

Name of scale	Abbreviation	Reference
ADHD Impact Module‐Child	AIM‐C	AIM‐C 2013
Child Impact Scale and Home Impact Scale	CIS/HIS	Landgraf 2002
Child Health and Illness Profile, Child Edition: Parent Report Form	CHIP‐CE:PRF	Riley 2004
Child Health Questionnaire	CHQ‐P	Landgraf 1998
Children's Global Assessment Scale	CGAS	Shaffer 1983
Comprehensive Psychopathological Rating Scale	CPRS	Aasberg 1978
Health Utilities Index ‐ 2	HUI‐2	Torrance 1982
ADHD: attention deficit hyperactivity disorder.

Table 3. Table 3: Quality of life ratings scales

Comparison 1. Teacher‐rated ADHD symptoms

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 All parallel‐group trials and first‐period cross‐over trials Show forest plot	19	1698	Std. Mean Difference (IV, Random, 95% CI)	‐0.77 [‐0.90, ‐0.64]

1.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
1.2 High risk of bias	19	1698	Std. Mean Difference (IV, Random, 95% CI)	‐0.77 [‐0.90, ‐0.64]
2 Subgroup analysis: types of scales Show forest plot	19		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

2.1 Conners' Teacher Rating Scale (CTRS)	8	518	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.82, ‐0.47]
2.2 Abbreviated Conners' Rating Scale (ACRS) ‐ Teacher	2	105	Std. Mean Difference (IV, Random, 95% CI)	‐0.75 [‐1.79, 0.29]
2.3 Conners' Abbreviated Symptom Questionnaire for Teachers (ASQ‐Teacher)	1	59	Std. Mean Difference (IV, Random, 95% CI)	‐0.28 [‐0.79, 0.23]
2.4 IOWA Conners' Teacher Rating Scale (IOWA CTRS) ‐ hyperactivity	2	193	Std. Mean Difference (IV, Random, 95% CI)	‐1.08 [‐1.39, ‐0.77]
2.5 Schedule for Non‐adaptive and Adaptive Personality (SNAP) ‐ Teacher	2	328	Std. Mean Difference (IV, Random, 95% CI)	‐0.61 [‐0.96, ‐0.25]
2.6 Teacher ratings of attention	1	20	Std. Mean Difference (IV, Random, 95% CI)	‐0.55 [‐1.45, 0.35]
2.7 Teacher ratings of impulsivity	1	20	Std. Mean Difference (IV, Random, 95% CI)	0.04 [‐0.83, 0.92]
2.8 IOWA Conners' Teacher Rating Scale ‐ Inattention/Overactivity (IOWA‐I/O)	2	197	Std. Mean Difference (IV, Random, 95% CI)	‐1.03 [‐1.36, ‐0.69]
2.9 Fremdbeurteilungsbogen für Hyperkinetische Störungen (FBB‐HKS)	1	85	Std. Mean Difference (IV, Random, 95% CI)	‐1.06 [‐1.52, ‐0.61]
2.10 Conners’ ADHD/DSM‐IV Scales ‐ Teacher (CADS‐T)	2	254	Std. Mean Difference (IV, Random, 95% CI)	‐1.05 [‐1.31, ‐0.78]
2.11 Strengths and Weaknesses of ADHD Symptoms and Normal Behaviour (SWAN) Scale	1	64	Std. Mean Difference (IV, Random, 95% CI)	‐0.33 [‐0.82, 0.17]
3 Medication status: medication naive versus not medication naive Show forest plot	6	717	Std. Mean Difference (IV, Random, 95% CI)	‐0.79 [‐1.08, ‐0.50]

3.1 Medication naive	4	431	Std. Mean Difference (IV, Random, 95% CI)	‐0.63 [‐0.94, ‐0.31]
3.2 Not medication naive	2	286	Std. Mean Difference (IV, Random, 95% CI)	‐1.06 [‐1.33, ‐0.79]
4 Subgroup analysis: duration of treatment Show forest plot	19	1698	Std. Mean Difference (IV, Random, 95% CI)	‐0.77 [‐0.90, ‐0.64]

4.1 Short term (up to 6 months)	18	1445	Std. Mean Difference (IV, Random, 95% CI)	‐0.81 [‐0.94, ‐0.68]
4.2 Long term (over 6 months)	1	253	Std. Mean Difference (IV, Random, 95% CI)	‐0.47 [‐0.72, ‐0.22]
5 Subgroup analysis: dose Show forest plot	19		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

5.1 Low dose	8	493	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.82, ‐0.46]
5.2 High dose	7	688	Std. Mean Difference (IV, Random, 95% CI)	‐0.81 [‐1.08, ‐0.54]
5.3 Unknown dose	6	669	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.06, ‐0.68]
6 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials Show forest plot	19	1698	Std. Mean Difference (IV, Random, 95% CI)	‐0.77 [‐0.90, ‐0.64]

6.1 Parallel‐group trials	17	1506	Std. Mean Difference (IV, Random, 95% CI)	‐0.80 [‐0.95, ‐0.65]
6.2 First‐period cross‐over trials	2	192	Std. Mean Difference (IV, Random, 95% CI)	‐0.58 [‐0.87, ‐0.29]
7 Subgroup analysis: trials with cohort selection bias of all participants compared with trials without cohort selection bias of all participants Show forest plot	19	1698	Std. Mean Difference (IV, Random, 95% CI)	‐0.77 [‐0.90, ‐0.64]

7.1 Trials with cohort selection bias of all participants	7	994	Std. Mean Difference (IV, Random, 95% CI)	‐0.79 [‐1.01, ‐0.57]
7.2 Trials without cohort selection bias of all participants	12	704	Std. Mean Difference (IV, Random, 95% CI)	‐0.76 [‐0.93, ‐0.59]
8 ADHD symptoms, cross‐over trial (endpoint data) Show forest plot	59	5145	Std. Mean Difference (IV, Random, 95% CI)	‐0.93 [‐1.06, ‐0.80]

8.1 Low risk of bias	4	204	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.97, ‐0.30]
8.2 High risk of bias	55	4941	Std. Mean Difference (IV, Random, 95% CI)	‐0.95 [‐1.09, ‐0.82]
9 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: dose Show forest plot	59	6821	Std. Mean Difference (IV, Random, 95% CI)	‐0.85 [‐0.96, ‐0.74]

9.1 Low dose	42	3408	Std. Mean Difference (IV, Random, 95% CI)	‐0.73 [‐0.89, ‐0.57]
9.2 High dose	36	3413	Std. Mean Difference (IV, Random, 95% CI)	‐0.98 [‐1.13, ‐0.84]
10 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials compared with cross‐over trials (endpoint data) Show forest plot	75	6344	Std. Mean Difference (IV, Random, 95% CI)	‐0.91 [‐1.01, ‐0.80]

10.1 All parallel‐group trials and first‐period cross‐over trials	19	1698	Std. Mean Difference (IV, Random, 95% CI)	‐0.77 [‐0.90, ‐0.64]
10.2 Cross‐over trials (endpoint data)	56	4646	Std. Mean Difference (IV, Random, 95% CI)	‐0.95 [‐1.09, ‐0.82]
11 All parallel‐group trials and cross‐over trials: risk of bias Show forest plot	75	6344	Std. Mean Difference (IV, Random, 95% CI)	‐0.91 [‐1.01, ‐0.80]

11.1 Low risk of bias	4	204	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.97, ‐0.30]
11.2 High risk of bias	71	6140	Std. Mean Difference (IV, Random, 95% CI)	‐0.92 [‐1.03, ‐0.81]

Comparison 1. Teacher‐rated ADHD symptoms

Comparison 2. Independent assessor‐rated ADHD symptoms

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 All parallel‐group trials and first‐period cross‐over trials Show forest plot	10	1907	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.89, ‐0.39]

1.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
1.2 High risk of bias	10	1907	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.89, ‐0.39]
2 Subgroup analysis: types of scales Show forest plot	10		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

2.1 Swanson, Kotkin, Agler, M‐Glynn and Pelham (SKAMP) Scale	1	164	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐1.04, ‐0.41]
2.2 ADHD Rating Scale (ADHD‐RS )	7	1442	Std. Mean Difference (IV, Random, 95% CI)	‐0.71 [‐1.09, ‐0.32]
2.3 Swanson, Nolan and Pelham (SNAP) Scale	1	221	Std. Mean Difference (IV, Random, 95% CI)	‐0.35 [‐0.61, ‐0.08]
2.4 Unknown	1	78	Std. Mean Difference (IV, Random, 95% CI)	‐0.94 [‐1.41, ‐0.47]
3 Subgroup analysis: duration of treatment Show forest plot	10	1907	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.89, ‐0.39]

3.1 Short term (up to 6 months)	9	1686	Std. Mean Difference (IV, Random, 95% CI)	‐0.68 [‐0.95, ‐0.40]
3.2 Long term (over 6 months)	1	221	Std. Mean Difference (IV, Random, 95% CI)	‐0.35 [‐0.61, ‐0.08]
4 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials Show forest plot	10	1907	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.89, ‐0.39]

4.1 Parallel‐group trials	7	1609	Std. Mean Difference (IV, Random, 95% CI)	‐0.60 [‐0.91, ‐0.28]
4.2 First‐period cross‐over trials	3	298	Std. Mean Difference (IV, Random, 95% CI)	‐0.76 [‐0.99, ‐0.52]
5 Subgroup analysis: trials with cohort selection bias of all participants compared with trials without cohort selection bias of all participants Show forest plot	10	1907	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.89, ‐0.39]

5.1 Trials with cohort selection bias of all participants	4	630	Std. Mean Difference (IV, Random, 95% CI)	‐0.51 [‐0.73, ‐0.29]
5.2 Trials without cohort selection bias of all participants	6	1277	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐1.11, ‐0.33]
6 Subgroup analysis: dose Show forest plot	10	1907	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.89, ‐0.39]

6.1 Low dose	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
6.2 High dose	6	1380	Std. Mean Difference (IV, Random, 95% CI)	‐0.55 [‐0.91, ‐0.20]
6.3 Unknown dose	4	527	Std. Mean Difference (IV, Random, 95% CI)	‐0.80 [‐0.98, ‐0.61]
7 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: risk of bias Show forest plot	19	2471	Std. Mean Difference (IV, Random, 95% CI)	1.00 [‐1.16, ‐0.84]

7.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
7.2 High risk of bias	19	2471	Std. Mean Difference (IV, Random, 95% CI)	1.00 [‐1.16, ‐0.84]
8 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: dose Show forest plot	19	3874	Std. Mean Difference (IV, Random, 95% CI)	‐0.89 [‐1.02, ‐0.75]

8.1 Low dose	16	2021	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.82, ‐0.55]
8.2 High dose	12	1853	Std. Mean Difference (IV, Random, 95% CI)	‐1.13 [‐1.31, ‐0.95]
9 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials compared with cross‐over trials (endpoint data) Show forest plot	28	4215	Std. Mean Difference (IV, Random, 95% CI)	‐0.83 [‐0.98, ‐0.67]

9.1 All parallel‐group trials and first‐period cross‐over trials	10	1907	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.89, ‐0.39]
9.2 Cross‐over trials (endpoint data)	18	2308	Std. Mean Difference (IV, Random, 95% CI)	‐0.95 [‐1.13, ‐0.77]

Comparison 2. Independent assessor‐rated ADHD symptoms

Comparison 3. Parent‐rated ADHD symptoms

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 All parallel‐group trials and first‐period cross‐over trials Show forest plot	21	2187	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.82, ‐0.51]

1.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
1.2 High risk of bias	21	2187	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.82, ‐0.51]
2 Subgroup analysis: types of scales Show forest plot	21		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

2.1 Conners' Parent Rating Scale (CPRS)	7	751	Std. Mean Difference (IV, Random, 95% CI)	‐0.58 [‐0.87, ‐0.30]
2.2 ADHD Rating Scale ‐ Fourth Edition (ADHD‐RS‐IV)	2	194	Std. Mean Difference (IV, Random, 95% CI)	‐0.30 [‐0.58, ‐0.02]
2.3 Fremdbeurteilungsbogen für Hyperkinetische Störungen (FBB‐HKS)	1	85	Std. Mean Difference (IV, Random, 95% CI)	‐0.91 [‐1.36, ‐0.46]
2.4 Conners’ ADHD/DSM‐IV Scales ‐ Parent (CADS‐P)	1	120	Std. Mean Difference (IV, Random, 95% CI)	‐1.26 [‐1.65, ‐0.86]
2.5 CADS‐P Inattentive subscale	1	109	Std. Mean Difference (IV, Random, 95% CI)	‐0.78 [‐1.17, ‐0.39]
2.6 CADS‐P Hyperactivity subscale	1	109	Std. Mean Difference (IV, Random, 95% CI)	‐0.93 [‐1.32, ‐0.53]
2.7 Clinican's Manual for the Assesment of Disruptive Behavior Disorders Rating Scale for Parents (Barkley)	1	41	Std. Mean Difference (IV, Random, 95% CI)	‐0.21 [‐0.82, 0.41]
2.8 Abbreviated Conners' Rating Scale (ACRS) ‐ Parent	2	121	Std. Mean Difference (IV, Random, 95% CI)	‐0.62 [‐0.99, ‐0.25]
2.9 Swanson, Nolan, and Pelham, Fourth Edition ‐ Parent (SNAP‐IV‐Parent) Scale	4	430	Std. Mean Difference (IV, Random, 95% CI)	‐0.60 [‐0.79, ‐0.40]
2.10 Strengths and Weaknesses of ADHD Symptoms and Normal Behavior (SWAN) Scale	1	86	Std. Mean Difference (IV, Random, 95% CI)	‐0.43 [‐0.86, 0.00]
2.11 IOWA Conners' Rating Scale ‐ Inattention/Overactivity (IOWA‐I/O)	3	352	Std. Mean Difference (IV, Random, 95% CI)	‐0.78 [‐1.35, ‐0.21]
3 Subgroup analysis: duration of treatment Show forest plot	21	2187	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.82, ‐0.51]

3.1 Short term (up to 6 months)	20	1925	Std. Mean Difference (IV, Random, 95% CI)	‐0.67 [‐0.84, ‐0.50]
3.2 Long term (over 6 months)	1	262	Std. Mean Difference (IV, Random, 95% CI)	‐0.58 [‐0.82, ‐0.33]
4 Subgroup analysis: dose Show forest plot	21	2335	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.79, ‐0.48]

4.1 Low dose	5	329	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [1.00, ‐0.07]
4.2 High dose	10	1132	Std. Mean Difference (IV, Random, 95% CI)	‐0.60 [‐0.86, ‐0.33]
4.3 Unknown dose	8	874	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐0.92, ‐0.52]
5 Medication status: medication naive versus not medication naive Show forest plot	7	795	Std. Mean Difference (IV, Random, 95% CI)	‐0.76 [‐1.05, ‐0.48]

5.1 Medication naive	4	492	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐1.03, ‐0.35]
5.2 Not medication naive	3	303	Std. Mean Difference (IV, Random, 95% CI)	‐0.88 [‐1.33, ‐0.42]
6 Subgroup analysis: trials with cohort selection bias of all participants compared with trials without cohort selection bias of all participants Show forest plot	21	2187	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.82, ‐0.51]

6.1 Trials with selection bias of all participants	10	1559	Std. Mean Difference (IV, Random, 95% CI)	‐0.68 [‐0.86, ‐0.51]
6.2 Trials without cohort selection bias of all participants	11	628	Std. Mean Difference (IV, Random, 95% CI)	‐0.62 [‐0.90, ‐0.33]
7 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials Show forest plot	21	2187	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.82, ‐0.51]

7.1 Parallel‐group trials	19	2094	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.83, ‐0.50]
7.2 First‐period cross‐over trials	2	93	Std. Mean Difference (IV, Random, 95% CI)	‐0.65 [‐1.07, ‐0.23]
8 ADHD symptoms, cross‐over trials (endpoint data) Show forest plot	41	3734	Std. Mean Difference (IV, Random, 95% CI)	‐0.78 [‐0.90, ‐0.67]

8.1 Low risk of bias	4	204	Std. Mean Difference (IV, Random, 95% CI)	‐0.54 [‐0.96, ‐0.13]
8.2 High risk of bias	37	3530	Std. Mean Difference (IV, Random, 95% CI)	‐0.81 [‐0.92, ‐0.69]
9 ADHD symptoms, cross‐over trials (endpoint data), subgroup analysis: dose Show forest plot	41	4918	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.79, ‐0.58]

9.1 Low dose	26	2272	Std. Mean Difference (IV, Random, 95% CI)	‐0.65 [‐0.82, ‐0.48]
9.2 High dose	28	2646	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐0.84, ‐0.60]
10 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials compared with cross‐over trials (endpoint data) Show forest plot	59	5861	Std. Mean Difference (IV, Random, 95% CI)	‐0.74 [‐0.83, ‐0.65]

10.1 All parallel‐group trials and first‐period cross‐over trials	21	2215	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.82, ‐0.51]
10.2 Cross‐over trials (endpoint data)	39	3646	Std. Mean Difference (IV, Random, 95% CI)	‐0.79 [‐0.90, ‐0.67]

Comparison 3. Parent‐rated ADHD symptoms

Comparison 4. Additional subgroup analyses of ADHD symptoms: parallel‐group trials and first‐period cross‐over trials

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Comparision of raters Show forest plot	31	5697	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.79, ‐0.59]

1.1 Teacher‐rated	19	1689	Std. Mean Difference (IV, Random, 95% CI)	‐0.78 [‐0.93, ‐0.63]
1.2 Independent assessor‐rated	9	1829	Std. Mean Difference (IV, Random, 95% CI)	‐0.61 [‐0.87, ‐0.35]
1.3 Parent‐rated	21	2179	Std. Mean Difference (IV, Random, 95% CI)	‐0.65 [‐0.81, ‐0.50]
2 Age Show forest plot	6	1039	Std. Mean Difference (IV, Random, 95% CI)	‐0.44 [‐0.74, ‐0.14]

2.1 2 to 6 years	1	64	Std. Mean Difference (IV, Random, 95% CI)	‐0.33 [‐0.82, 0.17]
2.2 7 to 11 years	2	278	Std. Mean Difference (IV, Random, 95% CI)	‐0.59 [‐1.03, ‐0.15]
2.3 12 to 18 years	3	697	Std. Mean Difference (IV, Random, 95% CI)	‐0.38 [‐0.88, 0.12]
3 Comorbidity versus no comorbidity Show forest plot	20	2310	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐0.91, ‐0.53]

3.1 ADHD with comorbidity	18	1981	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐0.94, ‐0.51]
3.2 ADHD without comorbidity	2	329	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.92, ‐0.46]
4 Subtypes ADHD: ADHD Rating Scale (parent‐, teacher‐ or independent assessor‐rated) Show forest plot	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

4.1 Combined ADHD	2	559	Std. Mean Difference (IV, Random, 95% CI)	0.65 [‐1.30, 2.60]
4.2 Inattentive ADHD	1	204	Std. Mean Difference (IV, Random, 95% CI)	‐1.31 [‐1.61, ‐1.01]
5 Cross‐over trials: first‐period data versus endpoint data (parent‐, independent assessor‐ and teacher‐rated) Show forest plot	4		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

5.1 First‐period data	4	372	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.85, ‐0.44]
5.2 Endpoint data	4	372	Std. Mean Difference (IV, Random, 95% CI)	‐0.91 [‐1.18, ‐0.65]

Comparison 4. Additional subgroup analyses of ADHD symptoms: parallel‐group trials and first‐period cross‐over trials

Comparison 5. Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Number of serious adverse events (SAE) Show forest plot	9	1532	Risk Ratio (IV, Random, 95% CI)	0.98 [0.44, 2.22]

2 Nervous system Show forest plot	6	2280	Risk Ratio (IV, Random, 95% CI)	0.87 [0.30, 2.53]

2.1 Aggression	1	303	Risk Ratio (IV, Random, 95% CI)	0.50 [0.05, 5.49]
2.2 Concussion	1	303	Risk Ratio (IV, Random, 95% CI)	0.34 [0.01, 8.17]
2.3 Loss of consciousness	1	221	Risk Ratio (IV, Random, 95% CI)	0.33 [0.01, 8.02]
2.4 Psychosis	4	712	Risk Ratio (IV, Random, 95% CI)	1.78 [0.19, 16.96]
2.5 Syncope	3	741	Risk Ratio (IV, Random, 95% CI)	1.39 [0.23, 8.47]
3 Digestive system: gastrointestinal disorders Show forest plot	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

4 Urinary system: kidney infection Show forest plot	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

5 Circulatory and respiratory systems: asthma Show forest plot	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

6 Immune system: cyst rupture Show forest plot	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

7 Other: drug toxicity Show forest plot	1	303	Risk Ratio (IV, Random, 95% CI)	0.34 [0.01, 8.17]

Comparison 5. Number of serious adverse events: parallel‐group trials and first‐period cross‐over trials

Comparison 6. Number of serious adverse events: cross‐over trials (endpoint data)

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Number of serious adverse events (SAE) Show forest plot	8	1721	Risk Ratio (IV, Random, 95% CI)	1.62 [0.34, 7.71]

2 Hallucinations/psychosis Show forest plot	4	187	Risk Ratio (IV, Random, 95% CI)	1.10 [0.18, 6.72]

Comparison 6. Number of serious adverse events: cross‐over trials (endpoint data)

Comparison 7. Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Total number of non‐serious adverse events Show forest plot	21	3132	Risk Ratio (IV, Random, 95% CI)	1.29 [1.10, 1.51]

2 Subgroup analysis: total number of non‐serious adverse events according to dose Show forest plot	21	3135	Risk Ratio (IV, Random, 95% CI)	1.28 [1.11, 1.49]

2.1 Low dose	2	151	Risk Ratio (IV, Random, 95% CI)	1.09 [0.82, 1.46]
2.2 High dose	10	1761	Risk Ratio (IV, Random, 95% CI)	1.22 [1.10, 1.35]
2.3 Unknown dose	10	1223	Risk Ratio (IV, Random, 95% CI)	1.37 [1.01, 1.87]
3 Nervous system Show forest plot	21		Risk Ratio (Random, 95% CI)	Subtotals only

3.1 Affective	4	390	Risk Ratio (Random, 95% CI)	2.39 [0.48, 11.96]
3.2 Aggression	2	417	Risk Ratio (Random, 95% CI)	1.16 [0.17, 7.80]
3.3 Apathy	1	59	Risk Ratio (Random, 95% CI)	0.80 [0.19, 3.33]
3.4 Confusion	2	548	Risk Ratio (Random, 95% CI)	1.01 [0.22, 4.73]
3.5 Depression	1	59	Risk Ratio (Random, 95% CI)	0.83 [0.22, 3.10]
3.6 Dizziness	3	683	Risk Ratio (Random, 95% CI)	2.50 [0.70, 8.99]
3.7 Drowsiness	4	811	Risk Ratio (Random, 95% CI)	1.27 [0.82, 1.98]
3.8 Emotional lability	1	132	Risk Ratio (Random, 95% CI)	2.41 [0.27, 21.32]
3.9 Fatigue	7	858	Risk Ratio (Random, 95% CI)	0.76 [0.36, 1.63]
3.10 Headache	17	2724	Risk Ratio (Random, 95% CI)	1.22 [0.90, 1.64]
3.11 Insomnia	3	349	Risk Ratio (Random, 95% CI)	1.31 [0.35, 4.93]
3.12 Irritability	11	1721	Risk Ratio (Random, 95% CI)	1.11 [0.77, 1.60]
3.13 Nervousness	2	362	Risk Ratio (Random, 95% CI)	2.52 [0.82, 7.76]
3.14 Pain	1	132	Risk Ratio (Random, 95% CI)	1.91 [0.21, 17.60]
3.15 Picking at skin or fingers, nail biting, lip or cheek chewing	1	316	Risk Ratio (Random, 95% CI)	1.01 [0.60, 1.70]
3.16 Sad, tearful or depressed	4	707	Risk Ratio (Random, 95% CI)	1.41 [0.86, 2.29]
3.17 Somnolence	2	173	Risk Ratio (Random, 95% CI)	0.59 [0.11, 3.11]
3.18 Trouble sleeping or sleep problems	13	2416	Risk Ratio (Random, 95% CI)	1.60 [1.15, 2.23]
3.19 Tics or nervous movements	8	1231	Risk Ratio (Random, 95% CI)	0.85 [0.26, 2.79]
3.20 Worried or anxious	3	596	Risk Ratio (Random, 95% CI)	1.37 [0.84, 2.25]
4 Digestive system Show forest plot	18		Risk Ratio (IV, Random, 95% CI)	Subtotals only

4.1 Decreased appetite	16	2962	Risk Ratio (IV, Random, 95% CI)	3.66 [2.56, 5.23]
4.2 Decreased weight	6	859	Risk Ratio (IV, Random, 95% CI)	3.89 [1.43, 10.59]
4.3 Diarrhoea	5	857	Risk Ratio (IV, Random, 95% CI)	1.07 [0.41, 2.74]
4.4 Dyspepsia	2	159	Risk Ratio (IV, Random, 95% CI)	1.80 [0.71, 4.54]
4.5 Increased appetite	1	179	Risk Ratio (IV, Random, 95% CI)	0.07 [0.00, 1.43]
4.6 Nausea	11	1995	Risk Ratio (IV, Random, 95% CI)	1.30 [0.85, 1.99]
4.7 Stomachache	13	2341	Risk Ratio (IV, Random, 95% CI)	1.30 [1.00, 1.69]
4.8 Vomiting	11	1916	Risk Ratio (IV, Random, 95% CI)	1.17 [0.76, 1.79]
5 Circulatory and respiratory systems Show forest plot	8		Risk Ratio (Random, 95% CI)	Subtotals only

5.1 ECG: prolonged QT‐interval	2	466	Risk Ratio (Random, 95% CI)	0.81 [0.13, 5.00]
5.2 ECG: tachycardia	1	245	Risk Ratio (Random, 95% CI)	1.04 [0.11, 10.18]
5.3 Cough	4	996	Risk Ratio (Random, 95% CI)	0.95 [0.41, 2.18]
5.4 Nasal congestion	2	479	Risk Ratio (Random, 95% CI)	1.19 [0.59, 2.41]
5.5 Pharyngolaryngeal pain	1	303	Risk Ratio (Random, 95% CI)	1.12 [0.59, 2.13]
5.6 Supraventricular extrasystoles	1	17	Risk Ratio (Random, 95% CI)	3.00 [0.11, 84.55]
5.7 Upper respiratory tract infection	1	217	Risk Ratio (Random, 95% CI)	1.07 [0.42, 2.76]
6 Skeletal and muscular systems Show forest plot	2		Risk Ratio (IV, Random, 95% CI)	Subtotals only

6.1 Arthralgia	1	303	Risk Ratio (IV, Random, 95% CI)	0.67 [0.24, 1.84]
6.2 Asthenia	1	177	Risk Ratio (IV, Random, 95% CI)	0.21 [0.01, 4.25]
6.3 Back pain	1	303	Risk Ratio (IV, Random, 95% CI)	0.81 [0.39, 1.66]
6.4 Myalgia	1	303	Risk Ratio (IV, Random, 95% CI)	0.60 [0.23, 1.62]
6.5 Toothache	1	303	Risk Ratio (IV, Random, 95% CI)	1.01 [0.43, 2.35]
7 Immune system Show forest plot	7		Risk Ratio (IV, Random, 95% CI)	Subtotals only

7.1 Gastroenteritis	3	435	Risk Ratio (IV, Random, 95% CI)	4.63 [0.99, 21.72]
7.2 Influenza	3	624	Risk Ratio (IV, Random, 95% CI)	0.65 [0.20, 2.10]
7.3 Nasopharyngitis	5	979	Risk Ratio (IV, Random, 95% CI)	1.15 [0.70, 1.87]
7.4 Otitis media	1	100	Risk Ratio (IV, Random, 95% CI)	1.77 [0.17, 18.94]
7.5 Pharyngitis	2	293	Risk Ratio (IV, Random, 95% CI)	2.43 [0.49, 12.05]
7.6 Pyrexia	2	400	Risk Ratio (IV, Random, 95% CI)	1.02 [0.01, 87.72]
7.7 Rhinitis	1	132	Risk Ratio (IV, Random, 95% CI)	1.28 [0.43, 3.79]
7.8 Upper respiratory tract infection ‐ not otherwise specified (NOS)	5	917	Risk Ratio (IV, Random, 95% CI)	1.19 [0.68, 2.06]
7.9 Viral infection NOS	3	614	Risk Ratio (IV, Random, 95% CI)	0.70 [0.23, 2.15]
8 Height Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

9 Weight Show forest plot	5	805	Std. Mean Difference (IV, Random, 95% CI)	‐1.12 [‐1.55, ‐0.70]

10 BMI Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

11 Vital signs Show forest plot	9	3374	Mean Difference (IV, Random, 95% CI)	1.41 [0.30, 2.52]

11.1 Diastolic blood pressure (mmHg)	8	1067	Mean Difference (IV, Random, 95% CI)	0.94 [‐0.12, 2.01]
11.2 Systolic blood pressure (mmHg)	8	1067	Mean Difference (IV, Random, 95% CI)	‐0.05 [‐1.25, 1.16]
11.3 Pulse or heart rate (bpm)	8	1240	Mean Difference (IV, Random, 95% CI)	3.41 [0.87, 5.94]
12 Other Show forest plot	5	1815	Risk Ratio (IV, Random, 95% CI)	1.20 [0.56, 2.57]

12.1 Accidental injury	3	656	Risk Ratio (IV, Random, 95% CI)	0.99 [0.48, 2.07]
12.2 Epistasis	1	132	Risk Ratio (IV, Random, 95% CI)	4.25 [0.23, 77.22]
12.3 Excoriation	1	303	Risk Ratio (IV, Random, 95% CI)	3.52 [1.19, 10.46]
12.4 Overdose	1	221	Risk Ratio (IV, Random, 95% CI)	2.97 [0.12, 72.20]
12.5 Skin disorder (rash)	2	200	Risk Ratio (IV, Random, 95% CI)	0.52 [0.01, 26.44]
12.6 Skin laceration	1	303	Risk Ratio (IV, Random, 95% CI)	0.42 [0.15, 1.16]

Comparison 7. Number of non‐serious adverse events: parallel‐group trials and first‐period cross‐over trials

Comparison 8. Number of non‐serious adverse events: cross‐over trials (endpoint data)

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Total number of non‐serious adverse events Show forest plot	21	2072	Risk Ratio (Random, 95% CI)	1.33 [1.11, 1.58]

2 Subgroup analysis: total number of non‐serious adverse events according to dose Show forest plot	21	2859	Risk Ratio (Random, 95% CI)	1.26 [1.11, 1.44]

2.1 Low dose	16	1539	Risk Ratio (Random, 95% CI)	1.11 [0.94, 1.31]
2.2 High dose	12	1080	Risk Ratio (Random, 95% CI)	1.57 [1.22, 2.01]
2.3 Unknown dose	2	240	Risk Ratio (Random, 95% CI)	0.98 [0.51, 1.86]
3 Nervous system Show forest plot	50		Risk Ratio (Random, 95% CI)	Subtotals only

3.1 Aggression	2	589	Risk Ratio (Random, 95% CI)	0.52 [0.17, 1.60]
3.2 Agitation	1	62	Risk Ratio (Random, 95% CI)	1.18 [0.38, 3.60]
3.3 Anger	3	264	Risk Ratio (Random, 95% CI)	0.45 [0.26, 0.77]
3.4 Behavioural complaints	1	82	Risk Ratio (Random, 95% CI)	0.55 [0.35, 0.86]
3.5 Buccal or lingual movements	4	302	Risk Ratio (Random, 95% CI)	1.06 [0.62, 1.79]
3.6 Compulsive acts	1	90	Risk Ratio (Random, 95% CI)	2.57 [1.45, 4.56]
3.7 Daydreaming	3	222	Risk Ratio (Random, 95% CI)	0.66 [0.44, 0.98]
3.8 Dizziness	9	746	Risk Ratio (Random, 95% CI)	1.17 [0.89, 1.55]
3.9 Drowsiness: dull, tired, listless or sleepy	21	1350	Risk Ratio (Random, 95% CI)	0.97 [0.73, 1.28]
3.10 Euphoria	6	405	Risk Ratio (Random, 95% CI)	1.06 [0.72, 1.57]
3.11 Headache	37	3752	Risk Ratio (Random, 95% CI)	1.21 [1.01, 1.45]
3.12 Insomnia or sleep problems	31	3270	Risk Ratio (Random, 95% CI)	1.57 [1.20, 2.06]
3.13 Irritability	23	2238	Risk Ratio (Random, 95% CI)	0.92 [0.66, 1.27]
3.14 Nightmares	10	686	Risk Ratio (Random, 95% CI)	0.97 [0.66, 1.42]
3.15 Overly meticulous	1	96	Risk Ratio (Random, 95% CI)	40.77 [2.35, 706.72]
3.16 Obsessive thinking	1	90	Risk Ratio (Random, 95% CI)	2.35 [1.53, 3.62]
3.17 Picking at skin or fingers, nail biting, lip or cheek chewing	15	888	Risk Ratio (Random, 95% CI)	1.12 [0.88, 1.41]
3.18 Repetitive language	1	48	Risk Ratio (Random, 95% CI)	1.0 [0.32, 3.10]
3.19 Sad, tearful or depressed	23	1849	Risk Ratio (Random, 95% CI)	1.15 [0.94, 1.41]
3.20 Socially withdrawn ‐ decreased interaction with others	12	771	Risk Ratio (Random, 95% CI)	1.24 [0.82, 1.87]
3.21 Sleep efficiency (SEF)	2	108	Risk Ratio (Random, 95% CI)	0.48 [0.02, 14.28]
3.22 Stares a lot	9	904	Risk Ratio (Random, 95% CI)	1.03 [0.75, 1.40]
3.23 Tics or nervous movements	19	1403	Risk Ratio (Random, 95% CI)	1.33 [1.03, 1.72]
3.24 Unusual blinking	1	48	Risk Ratio (Random, 95% CI)	3.13 [0.12, 80.68]
3.25 Worried or anxious	20	1673	Risk Ratio (Random, 95% CI)	0.82 [0.60, 1.12]
4 Digestive system Show forest plot	42		Risk Ratio (Random, 95% CI)	Subtotals only

4.1 Decreased appetite or loss of appetite	35	3862	Risk Ratio (Random, 95% CI)	3.04 [2.35, 3.94]
4.2 Diarrhoea	3	402	Risk Ratio (Random, 95% CI)	0.58 [0.19, 1.74]
4.3 Dry mouth	5	342	Risk Ratio (Random, 95% CI)	1.25 [0.54, 2.90]
4.4 Dyspepsia	1	62	Risk Ratio (Random, 95% CI)	0.22 [0.02, 2.14]
4.5 Nausea	9	768	Risk Ratio (Random, 95% CI)	1.52 [1.00, 2.30]
4.6 Increased appetite	1	136	Risk Ratio (Random, 95% CI)	0.20 [0.08, 0.50]
4.7 Stomachache	33	3777	Risk Ratio (Random, 95% CI)	1.61 [1.27, 2.04]
4.8 Vomiting	4	710	Risk Ratio (Random, 95% CI)	0.90 [0.26, 3.11]
5 Urinary system: urinary incontinence Show forest plot	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected

6 Skeletal and muscular system: somatic complaints Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

7 Immune system Show forest plot	7		Risk Ratio (IV, Random, 95% CI)	Subtotals only

7.1 Allergic rhinitis	4	475	Risk Ratio (IV, Random, 95% CI)	1.38 [0.35, 5.51]
7.2 Fever	2	91	Risk Ratio (IV, Random, 95% CI)	1.39 [0.09, 20.56]
7.3 Lymphadenitis	2	296	Risk Ratio (IV, Random, 95% CI)	3.93 [0.44, 35.11]
7.4 Pharyngolaryngeal pain	1	160	Risk Ratio (IV, Random, 95% CI)	2.0 [0.19, 21.62]
7.5 Pharyngitis	4	754	Risk Ratio (IV, Random, 95% CI)	0.71 [0.19, 2.62]
7.6 Upper respiratory tract infection	5	888	Risk Ratio (IV, Random, 95% CI)	0.61 [0.27, 1.37]
8 Skin Show forest plot	3		Risk Ratio (IV, Random, 95% CI)	Subtotals only

8.1 Rash	2	208	Risk Ratio (IV, Random, 95% CI)	0.96 [0.14, 6.41]
8.2 Skin laceration	1	167	Risk Ratio (IV, Random, 95% CI)	2.96 [0.12, 71.75]
9 Vital signs Show forest plot	14		Mean Difference (IV, Random, 95% CI)	Subtotals only

9.1 Diastolic blood pressure (mmHg)	11	755	Mean Difference (IV, Random, 95% CI)	1.23 [‐0.39, 2.86]
9.2 Systolic blood pressure (mmHg)	11	755	Mean Difference (IV, Random, 95% CI)	0.53 [‐1.30, 2.36]
9.3 Pulse or heart rate (bpm)	14	939	Mean Difference (IV, Random, 95% CI)	5.06 [2.88, 7.24]
10 Height (cm) Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

11 Weight Show forest plot	6	530	Std. Mean Difference (IV, Random, 95% CI)	‐0.06 [‐0.23, 0.11]

Comparison 8. Number of non‐serious adverse events: cross‐over trials (endpoint data)

Comparison 9. Teacher‐rated general behaviour

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 All parallel‐group trials and first‐period cross‐over trials: risk of bias Show forest plot	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]

1.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
1.2 High risk of bias	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]
2 Subgroup analysis: types of scales Show forest plot	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]

2.1 Conners' Global Index ‐ Teacher (CGI‐T)	1	314	Std. Mean Difference (IV, Random, 95% CI)	‐0.91 [‐1.14, ‐0.68]
2.2 Groninger Behaviour Observation Scale (GBOS)	1	43	Std. Mean Difference (IV, Random, 95% CI)	‐0.84 [‐1.46, ‐0.21]
2.3 Conners' Teacher Rating Scale ‐ Conduct problems	1	25	Std. Mean Difference (IV, Random, 95% CI)	‐0.67 [‐1.48, 0.14]
2.4 IOWA Conners' Rating Scale ‐ Oppositional/Defiant (IOWA‐O/D)	2	286	Std. Mean Difference (IV, Random, 95% CI)	‐0.86 [‐1.12, ‐0.59]
3 Subgroup analysis: dose Show forest plot	5	696	Std. Mean Difference (IV, Random, 95% CI)	‐0.85 [‐1.02, ‐0.69]

3.1 Low dose	2	71	Std. Mean Difference (IV, Random, 95% CI)	‐0.67 [‐1.16, ‐0.19]
3.2 High dose	3	466	Std. Mean Difference (IV, Random, 95% CI)	‐0.89 [‐1.08, ‐0.70]
3.3 Unknown dose	1	159	Std. Mean Difference (IV, Random, 95% CI)	‐0.84 [‐1.21, ‐0.46]
4 Subgroup analysis: duration of treatment Show forest plot	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]

4.1 Short term (up to 6 months)	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]
4.2 Long term (over 6 months)	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
5 Subgroup analysis: parallel‐group trials versus first‐period cross‐over trials Show forest plot	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]

5.1 Parallel‐group trials	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]
5.2 First‐period cross‐over trials	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
6 General behaviour, cross‐over trials (endpoint data) Show forest plot	16	2014	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.78, ‐0.60]

6.1 Low dose	13	1110	Std. Mean Difference (IV, Random, 95% CI)	‐0.60 [‐0.72, ‐0.48]
6.2 High dose	12	904	Std. Mean Difference (IV, Random, 95% CI)	‐0.82 [‐0.95, ‐0.68]
7 Subgroup analysis: general behaviour, cross‐over trials (endpoint data): risk of bias Show forest plot	16	1308	Std. Mean Difference (IV, Random, 95% CI)	‐0.75 [‐0.87, ‐0.63]

7.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
7.2 High risk of bias	16	1308	Std. Mean Difference (IV, Random, 95% CI)	‐0.75 [‐0.87, ‐0.63]
8 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials (teacher‐rated) versus cross‐over trials (endpoint data) Show forest plot	21	1976	Std. Mean Difference (IV, Random, 95% CI)	‐0.79 [‐0.88, ‐0.70]

8.1 All parallel‐group trials and first‐period cross‐over trials	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]
8.2 Cross‐over trials (endpoint data)	16	1308	Std. Mean Difference (IV, Random, 95% CI)	‐0.75 [‐0.87, ‐0.63]

Comparison 9. Teacher‐rated general behaviour

Comparison 10. Independent assessor‐rated general behaviour

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 General behaviour, cross‐over trials (endpoint data) Show forest plot	8	1241	Std. Mean Difference (IV, Random, 95% CI)	‐0.60 [‐0.75, ‐0.46]

1.1 Low dose	7	903	Std. Mean Difference (IV, Random, 95% CI)	‐0.56 [‐0.76, ‐0.36]
1.2 High dose	5	338	Std. Mean Difference (IV, Random, 95% CI)	‐0.71 [‐0.93, ‐0.49]
2 Subgroup analysis: general behaviour, cross‐over trials (endpoint data): risk of bias Show forest plot	8	951	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.88, ‐0.49]

2.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
2.2 High risk of bias	8	951	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.88, ‐0.49]
3 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials (independent assessor‐rated) compared with cross‐over trials (endpoint data) Show forest plot	8	951	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.88, ‐0.49]

3.1 Parallel‐group trials and first‐period cross‐over trials	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
3.2 Cross‐over trials (endpoint data)	8	951	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.88, ‐0.49]

Comparison 10. Independent assessor‐rated general behaviour

Comparison 11. Parent‐rated general behaviour

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 All parallel‐group trials and first‐period cross‐over trials Show forest plot	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]

2 Subgroup analysis: risk of bias Show forest plot	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]

2.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
2.2 High risk of bias	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]
3 Subgroup analysis: types of scales Show forest plot	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]

3.1 The Weekly Parent Ratings of Evening and Morning Behaviour (WPREMB) ‐ Revised	1	17	Std. Mean Difference (IV, Random, 95% CI)	0.50 [‐0.47, 1.47]
3.2 Conners' Global Index (CGI) ‐ Parent	2	352	Std. Mean Difference (IV, Random, 95% CI)	‐0.41 [‐0.63, ‐0.20]
3.3 Swanson, Nolan and Pelham, Fourth Edition ‐ Oppositional (SNAP‐IV‐Oppositional)	1	15	Std. Mean Difference (IV, Random, 95% CI)	‐0.86 [‐1.95, 0.23]
3.4 IOWA Conners' Rating Scale ‐ Oppositional/Defiant (IOWA‐I/O)	2	286	Std. Mean Difference (IV, Random, 95% CI)	‐0.75 [‐1.01, ‐0.49]
4 Subgroup analysis: parallel‐group trials compared with first‐period cross‐over trials Show forest plot	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]

4.1 Parallel‐group trials	5	655	Std. Mean Difference (IV, Random, 95% CI)	‐0.51 [‐0.78, ‐0.23]
4.2 First‐period cross‐over trials	1	15	Std. Mean Difference (IV, Random, 95% CI)	‐0.86 [‐1.95, 0.23]
5 Subgroup analysis: duration of treatment Show forest plot	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]

5.1 Short term (up to 6 months)	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]
5.2 Long term (over 6 months)	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
6 Subgroup analysis: dose Show forest plot	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]

6.1 Low dose	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
6.2 High dose	4	496	Std. Mean Difference (IV, Random, 95% CI)	‐0.42 [‐0.77, ‐0.08]
6.3 Unknown dose	2	174	Std. Mean Difference (IV, Random, 95% CI)	‐0.73 [‐1.08, ‐0.38]
7 General behaviour, cross‐over trials (endpoint data) Show forest plot	6	550	Std. Mean Difference (IV, Random, 95% CI)	‐0.75 [‐0.93, ‐0.56]

7.1 Low dose	5	248	Std. Mean Difference (IV, Random, 95% CI)	‐0.65 [‐0.93, ‐0.38]
7.2 High dose	4	302	Std. Mean Difference (IV, Random, 95% CI)	‐0.83 [‐1.07, ‐0.60]
8 Subgroup analysis: general behaviour, cross‐over trials (endpoint data): risk of bias Show forest plot	6	384	Std. Mean Difference (IV, Random, 95% CI)	‐0.84 [‐1.05, ‐0.63]

8.1 Low risk of bias	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
8.2 High risk of bias	6	384	Std. Mean Difference (IV, Random, 95% CI)	‐0.84 [‐1.05, ‐0.63]
9 Subgroup analysis: all parallel‐group trials and first‐period cross‐over trials (parent‐rated) compared with cross‐over trials (endpoint data) Show forest plot	12	1054	Std. Mean Difference (IV, Random, 95% CI)	‐0.68 [‐0.86, ‐0.50]

9.1 All parallel‐group trials and first‐period cross‐over trials	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]
9.2 Cross‐over trials (endpoint data)	6	384	Std. Mean Difference (IV, Random, 95% CI)	‐0.84 [‐1.05, ‐0.63]

Comparison 11. Parent‐rated general behaviour

Comparison 12. Additional subgroup analyses of general behaviour: parallel‐group trials and first‐period cross‐over trials

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Comparisions of raters Show forest plot	8	1338	Std. Mean Difference (IV, Random, 95% CI)	‐0.69 [‐0.86, ‐0.52]

1.1 Teacher‐rated	5	668	Std. Mean Difference (IV, Random, 95% CI)	‐0.87 [‐1.04, ‐0.71]
1.2 Independent assessor‐rated	0	0	Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
1.3 Parent‐rated	6	670	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐0.78, ‐0.27]
2 Comorbidity versus no comorbidity Show forest plot	7	579	Std. Mean Difference (IV, Random, 95% CI)	‐0.70 [‐0.98, ‐0.43]

2.1 ADHD with comorbidity	6	265	Std. Mean Difference (IV, Random, 95% CI)	‐0.59 [‐0.95, ‐0.23]
2.2 ADHD without comorbidity	1	314	Std. Mean Difference (IV, Random, 95% CI)	‐0.91 [‐1.14, ‐0.68]
3 Cross‐over trials: first‐period data versus endpoint data (teacher‐, parent‐, and independent assessor‐rated) Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Subtotals only

3.1 First‐period data	1	16	Mean Difference (IV, Random, 95% CI)	‐0.81 [‐1.75, 0.13]
3.2 Endpoint data	1	14	Mean Difference (IV, Random, 95% CI)	0.14 [‐0.71, 1.00]

Comparison 12. Additional subgroup analyses of general behaviour: parallel‐group trials and first‐period cross‐over trials

Comparison 13. Quality of life: parallel‐group trials and first‐period cross‐over trials

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Subgroup analysis: types of scales Show forest plot	3	514	Std. Mean Difference (IV, Random, 95% CI)	0.61 [0.42, 0.80]

1.1 Child Health Questionnaire (CHQ)	1	257	Std. Mean Difference (IV, Random, 95% CI)	0.54 [0.25, 0.83]
1.2 Children´s Global Assessment Scale (CGAS)	1	36	Std. Mean Difference (IV, Random, 95% CI)	0.79 [0.10, 1.47]
1.3 Child Health and Illness Profile, Child Edition: Parent Report Form (CHIP‐CE:PRF)	1	221	Std. Mean Difference (IV, Random, 95% CI)	0.64 [0.37, 0.91]

Comparison 13. Quality of life: parallel‐group trials and first‐period cross‐over trials