Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Extended‐release methylphenidate for attention deficit hyperactivity disorder (ADHD) in adults

Authors' declarations of interest

Version published: 15 November 2017 Version history

https://doi.org/10.1002/14651858.CD012857

This is not the most recent version

view the current version 24 February 2022
Collapse all Expand all

Abstract

This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:

To assess the beneficial and harmful effects of extended‐release formulations of methylphenidate in adults diagnosed with ADHD.

Background

Description of the condition

Attention deficit hyperactivity disorder (ADHD) in childhood was introduced as a diagnosis in the American Diagnostic and Statistical Manual of Mental Disorders Third Revision (DSM‐III‐R) in 1987 (APA 1987; Lange 2010), and in adolescence and adulthood in 2013 with the fifth edition of the DSM (DSM‐5; APA 2013). A meta‐analysis of epidemiological data reported a prevalence of ADHD in adults of 2.5% (Simon 2009). A diagnosis of ADHD in adulthood requires the presence of five core symptoms that must have persisted for more than six months, and must have been present before 12 years of age, according to DSM‐5 criteria (APA 2013). The equivalent condition in the World Health Organization’s (WHO) tenth edition of the International Classification of Diseases (ICD‐10) is hyperkinetic disorder, with similar diagnostic criteria but a required age of onset of symptoms of seven years (WHO 1992).

Some argue that there is consensus that ADHD in childhood may persist into adulthood (Kooij 2010). This opinion has been based on long‐term follow‐up in cohort studies of ADHD in childhood, but a meta‐analysis of these studies estimated that this is rarely the case. Only 15% of children diagnosed with ADHD still fulfilled the diagnostic criteria for ADHD at 25 years of age (Faraone 2006). In agreement with this, the first population study that compared the prevalence of ADHD in children and adults found less than 10% overlap between those diagnosed with ADHD as children and those diagnosed as adults at the age of 38 years (Moffitt 2015). Furthermore, those diagnosed with ADHD as adults had normal neurocognitive tests both as children and adults, which is contradictory to the classification of ADHD in adulthood as a neurodevelopmental disorder (APA 2013). Recently, two studies reported data that support this observation. Longitudinal cohort studies in children with a follow‐up until the age of 18 to 19 years found that only 17% of those diagnosed with ADHD during childhood were still eligible for a diagnosis of ADHD as adults in Brazil (Caye 2016). The corresponding number was 22% in the UK (Agnew‐Blais 2016). These studies also found that more than two‐thirds of those adults with ADHD did not have ADHD symptoms at childhood assessments. This challenges the hypothesis that ADHD is a neurodevelopmental disorder that persists into adulthood (Agnew‐Blais 2016; Caye 2016; Moffitt 2015), and has launched a debate about adult‐onset ADHD (Faraone 2016), and the difference between symptom remission and continuous functional impairments in adulthood (Faraone 2006).

ADHD is a controversial diagnosis, as recognised in the Consensus Statement by the National Institutes of Health (NIH 1998), and the views are polarised regarding its etiology, diagnosis and treatment (Frances 2013; NICE 2009). ADHD in childhood is associated with low parental socioeconomic status (Russell 2014), and school children born in December are 50% more likely to get the diagnosis than those born in January in the same class who have had 11 months more for their brains to develop (Morrow 2012). A consensus statement was issued in 2002 by a group of 86 ADHD researchers, which supported the hypothesis that ADHD is a neurobiological disease with a high degree of heritability that can be effectively treated with central stimulants (Barkley 2002). A specific critique of this statement was issued by another group of 34 academics and practitioners in 2004 that questioned the validity and reliability of the evidence (Timimi 2004).

According to the diagnostic criteria specified in the DSM‐5, adults are diagnosed retrospectively based on memories of childhood behaviour (APA 2013). Studies have concluded that making diagnoses retrospectively is not very reliable (e.g. due to recall and confirmation bias) (Mannuzza 2002; Suhr 2009), and it is therefore unsurprising that the validity of the diagnosis of ADHD in adulthood is subject to heated debate (Asherson 2010; Moncrieff 2010). Because of these problems, and the absence of a 'gold standard' for diagnosis, global prevalence estimates are uncertain. In the USA, the prevalence of adult ADHD increased three‐fold between 2002 and 2007 (Montejano 2011). This raises concerns about overdiagnosis (Paris 2015), and the chairman of the DSM‐IV Taskforce, Allen Frances, has stated that the current definition of ADHD in adulthood is too vague (Frances 2013).

People labelled with the adult‐ADHD diagnosis generally have poorer work performance (de Graaf 2008), excessive mortality (Dalsgaard 2015), higher frequencies of unemployment, job changes, divorces and lower life satisfaction (Biederman 2006). A major shortcoming of these studies, however, is the poor reliability of the retrospective diagnosis. Adults diagnosed with ADHD often suffer from additional comorbid psychopathology, such as affective and anxiety disorders, substance abuse and personality disorders (Cumyn 2009; Groß‐Lesch 2013; Jacob 2007; Kessler 2006; Philipsen 2014; Sobanski 2007a).

The UK National Institute for Health and Care Excellence (NICE) recommends medication as first‐line treatment for adults with ADHD, unless the patient prefers a psychotherapeutic approach, and methylphenidate should be tried first (NICE 2009). Additionally, the NICE guideline states that: "Drug treatment for adults with ADHD should always form part of a comprehensive treatment programme that addresses psychological, behavioural and educational or occupational needs" (NICE 2009).

Description of the intervention

Methylphenidate is available as immediate‐ or extended‐release formulations. The latter consist of both immediate‐ and extended‐release components leading to a bimodal pharmacokinetic profile (Maldonado 2013). Several extended‐release formulations are on the market, including a transdermal patch, and these formulations differ in their pharmacokinetic profiles with respect to peak plasma levels. They are effective for eight to 12 hours (Coghill 2013), compared to four hours for immediate‐release methylphenidate. The extended‐release formulations require one to two daily administrations compared to two to three administrations for immediate‐release methylphenidate, which has the advantage of better compliance, less social stigmatisation and less inconvenience (Coghill 2013). However, there has been only one head‐to‐head trial of immediate‐release methylphenidate versus extended‐release methylphenidate in adults, which could not demonstrate differences regarding their benefits or harms (McDonagh 2015; Spencer 2011).

Time trends for the use of methylphenidate in children in the UK and Spain show that extended‐release formulations are used more often compared to immediate‐release formulations and that the proportion of extended‐release users is rising (Beau‐Lejdstrom 2016; Treceño 2012). A similar shift towards extended‐release use is seen in adults in the Nordic countries (Karlstad 2016). A Cochrane Review has evaluated immediate‐release methylphenidate for adults with ADHD (Boesen 2017; Epstein 2016), but this was withdrawn (see: Why it is important to do this review). Another Cochrane Review has evaluated immediate‐ and extended‐release methylphenidate in children (Storebø 2015).

How the intervention might work

Methylphenidate is a central nervous system stimulant, often referred to as a 'sympathomimetic amine', and is structurally related to amphetamine (Nissen 2006). One of its actions is to block dopamine reuptake, which increases intrasynaptic concentrations of dopamine. Furthermore, methylphenidate is thought to enhance the ability to concentrate on one task, while filtering out other stimuli, which increases attention and motivation during the execution of specific tasks (Volkow 2005).

A systematic review of placebo controlled trials assessing methylphenidate's effects on cognition in children and adolescents with ADHD reported that cognitive functions improved with methylphenidate treatment (Coghill 2014). The first study to investigate the effects of methylphenidate on cognition in healthy adults found no differences compared to adults with ADHD (Agay 2010). This suggests that the effect of methylphenidate is non‐specific. As late as 2010, none of the published randomised controlled trials that assessed the benefits and harms of methylphenidate in adults with ADHD had investigated the effect on quality of life, which the author of the systematic review found to be a serious omission (Coghill 2010).

Methylphenidate causes many adverse effects, including weight loss, sleep disturbances and tics (Graham 2011). It leads to stomach ache, reduced appetite and headache in children (Storebø 2015). Stimulants also increase blood pressure and heart rate (Mick 2013), and two observational studies found associations with transient ischaemic attacks (Holick 2009), ventricular arrhythmias and sudden death in adults with ADHD (Schelleman 2012). Methylphenidate is associated with an increased risk of violence towards others (Moore 2010), and it may also trigger a psychosis (Kraemer 2010). Furthermore, methylphenidate increases the risk that patients receive an additional diagnosis of bipolar disorder, because the adverse effects of the drug are similar to the symptoms that define this disorder (Novartis 2007).

It is common to use a dose‐titration design in methylphenidate trials (Boesen 2017). A dose‐titration design implies that each participant is titrated to the highest tolerated dose. The use of a dose‐titration design indicates that there is a highly variable tolerance to the drug. Central stimulant drug effects are likely related to the pharmacokinetics and thus to the rise and fall in plasma levels (Angrist 1987). There is high variability in enzyme function related to methylphenidate metabolism, such as the carboxylesterases 1 and 2, which complicates interpretations of dose‐response relationships (Merali 2014; Stage 2017). The dose‐titration study design compounds the problem of maintaining blinding in trials of a central stimulant drug with well‐known characteristic adverse effects. The problem of maintaining double‐blind conditions in central stimulant trials were highlighted in previous Cochrane Reviews of amphetamines for adults with ADHD (Castells 2011b), amphetamines for children with ADHD (Punja 2016), and methylphenidate for children with ADHD (Storebø 2015).

It is doubtful whether stimulant treatment leads to long‐term benefit. A diminution of clinical long‐term effects was observed in the Multimodal Treatment Study (MTA), one of the largest trials in children and the one with the longest follow‐up, where there were no differences between the groups at three, six and eight years' follow‐up (Jensen 2007; Molina 2009). At 16 years' follow‐up medication was associated with suppressed growth but not with symptom reduction (Swanson 2017). Methylphenidate did not protect against delinquency or substance abuse; if anything, it caused this (Molina 2007). The patient sample was probably biased in favour of methylphenidate, as patients who had previously been on an ADHD drug were excluded if they had not tolerated the drug. The reason for the diminution of treatment effect over time could be adaptations in the brain. A systematic review of neuroimaging studies investigating the dopamine transporter density in adult patients with ADHD compared to healthy controls concluded that stimulant treatment leads to compensatory upregulation of dopamine transporters (Fusar‐Poli 2012). This was confirmed in a prospective study (Wang 2013), and the authors concluded that this adaptation could explain why the effects of methylphenidate wane over time.

Why it is important to do this review

Previous meta‐analyses of methylphenidate trials for adults with ADHD have reported standardised mean differences (SMDs) of 0.40 to 0.90 in favour of methylphenidate. Due to variations in methodology and included trials, these meta‐analyses are not comparable. Some have included all types of stimulants (Cunill 2016; Faraone 2010; Meszaros 2009; Peterson 2008), and others only methylphenidate (Castells 2011a; Castells 2013; Faraone 2004; Koesters 2009).

Extensive criticism of methylphenidate trials in both children and adults has been published. Some of the main criticisms are: short trial duration, extensive exclusion criteria (e.g. where those who do not tolerate the drug are not randomised, misleadingly called an 'enriched' design), lack of long‐term follow‐up, lack of long‐term cardiovascular monitoring, lack of patient relevant outcomes, heterogeneous study designs and outcome reporting bias (Boesen 2017; McDonagh 2011; Peterson 2008; Storebø 2015). All 185 trials included in the large Cochrane Review on methylphenidate for children and adolescents were assessed as being at high risk of bias, and all primary and secondary outcomes were graded as “very low‐quality” evidence (Storebø 2015). The authors argued that longer, better‐designed trials with an “active placebo” (a substance that gives patients side effects) to secure an effective blinding are much needed.

The increased use of methylphenidate in adults has been drastic; insurance companies in the USA reported a 53% increase of adults using ADHD medication from 2008 to 2012 (Express 2014), and methylphenidate prescriptions for adults rose about nine‐fold in Sweden from 2006 to 2014 (SMER 2015). This has raised concerns of harms, including drug abuse (Swanson 2008), and cardiovascular harms. A systematic review about misuse and selling of prescription stimulants for treatment of ADHD to others suggested that methylphenidate was misused, both in people with and without an ADHD diagnosis (Wilens 2008), and the drug carries a 'black‐box warning' that warns of drug dependence. The US Food and Drug Administration (FDA) was also recommended to give all stimulants a "black‐box warning" in 2006 due to concerns about cardiovascular harms (Nissen 2006), but this was not followed. The global rise in methylphenidate consumption has concerned the International Narcotics Control Board for more than 20 years (INCB 1995; INCB 2014).

A Cochrane Review of immediate‐release methylphenidate in adults with ADHD was published in 2014 (Epstein 2014), but it was withdrawn in May 2016 as the authors of the review were unable to respond to extensive criticisms that several research groups had submitted (Epstein 2016). The criticisms focused on the Cochrane Review's conclusions, as it gave a misleading impression of certainty related to benefits and absence of harms, and many important methodological flaws were also highlighted. A summary of the criticism that led to the withdrawal has been published (Boesen 2017), and the original comments can be seen in the original review (Epstein 2014).

Objectives

To assess the beneficial and harmful effects of extended‐release formulations of methylphenidate in adults diagnosed with ADHD.

Methods

Criteria for considering studies for this review

Types of studies

Double‐blind, randomised controlled trials (RCTs) comparing extended‐release methylphenidate with placebo. We will exclude cross‐over trials due to the risk of carry‐over effects and the loss of blinding (Gualtieri 1985; Higgins 2011b), and because long‐term follow‐up is not possible. We will exclude cluster‐randomised trials (i.e. studies where the allocation is based on medical centres, communities, schools or similar) because the effect of the treatment is likely to vary considerably depending on the setting and local variation in other interventions offered.

Types of participants

Adults diagnosed with ADHD, according to DSM‐III‐R (APA 1987), DSM‐IV‐TR (fourth edition, text revision; APA 2000), ICD‐10 (WHO 1992), or the present definition according to DSM‐5 (APA 2013). Adults diagnosed according to the criteria of ADHD in childhood, before the introduction of the ADHD diagnosis in adolescence and adulthood with the DSM‐5 (APA 2013).

Types of interventions

Methylphenidate in any extended‐release formulation, at any dose. We will include trials with one or more comparators, either an active control or placebo, or both. We will permit cointerventions, such as cognitive behavioural therapy, provided that they are delivered to both groups.

Types of outcome measures

Primary outcomes

  1. Functional outcomes (daily functioning), for example, academic and job adherence measured as days of lost work or study activities (Philipsen 2008), marital status, delinquency and traffic accidents.

  2. Self‐rated ADHD symptoms measured on validated rating scales such as the Connors' Adult ADHD Rating Scale (Conners 1999).

  3. Serious adverse events as defined according to the International Council for Harmonisation (ICH) guideline as any untoward medical occurrence that, at any dose, results in death, is life‐threatening, requires inpatient hospitalisation or results in prolongation of existing hospitalisation, results in persistent or significant disability/incapacity, is a congenital anomaly/birth defect or is a medically important event or reaction (ICH 2003).

Secondary outcomes

  1. Investigator‐rated ADHD symptoms measured on validated rating scales such as the Connors' Adult ADHD Rating Scale (Conners 1999).

  2. Observer‐rated ADHD symptoms (partner, family or peer) measured on validated rating scales such as the Connors' Adult ADHD Rating Scale (Conners 1999).

  3. Adverse events other than serious adverse events. We will pay special attention to the limitations of adverse event reporting, such as not reporting events if their frequency is below some threshold or the use of rating scales such as Barkley’s Side Effects Rating Scale, since they only cover few adverse events, which means that rare, important events are not included.

  4. Cardiovascular variables, for example, blood pressure and heart rate.

  5. Severe psychiatric adverse events, for example, changes of mental state, such as acute psychoses or depression, leading to hospitalisation or initiation of treatment (may overlap with serious adverse events) or psychiatric adverse events leading to dropout or dose‐reductions.

  6. Quality of life measured on validated psychometric scales such as the Quality of Life Enjoyment and Satisfaction Questionnaire ‐ Short Form (Mick 2008).

We will use data with the longest possible follow‐up within three outcome periods: short term (up to six months), medium term (up to 12 months) and long term (more than 12 months). See also Unit of analysis issues.

Search methods for identification of studies

Electronic searches

We will search for eligible studies in the following electronic sources.

  1. Cochrane Central Register of Controlled Trials (CENTRAL; current issue), in the Cochrane Library, which includes the Cochrane Developmental, Psychosocial and Learning Problems Specialised Register.

  2. MEDLINE Ovid (1946 onwards).

  3. MEDLINE In‐Process & Other Non‐Indexed Citations Ovid (current issue).

  4. MEDLINE Epub Ahead of Print Ovid (current issue).

  5. Embase Ovid (1980 onwards)

  6. PsycINFO Ovid (1806 onwards).

  7. Cochrane Database of Systematic Reviews (CDSR; current issue), part of the Cochrane Library.

  8. Database of Abstracts of Reviews of Effects (DARE; current issue), part of the Cochrane Library.

  9. CINAHL Plus EBSCOhost (Cumulative Index to Nursing and Allied Health Literature; 1937 onwards).

  10. Conference Proceedings Citation Index ‐ Science Web of Science (CPCI‐S; 1990 onwards).

  11. LILACS (Latin American and Caribbean Health Science Information database; www.lilacs.bvsalud.org/en).

  12. ClinicalTrials.gov (clinicaltrials.gov).

  13. EU Clinical Trials Register (clinicaltrialsregister.eu).

  14. ISRCTN Registry (www.isrctn.com).

  15. WHO International Clinical Trials Registry Platform (apps.who.int/trialsearch; all available years).

We will search Ovid MEDLINE using the strategy in Appendix 1, which we will adapt for the other databases, as necessary. We will not limit the search by publication date or language.

Searching other resources

We will search for bibliographic references in the included papers, relevant systematic reviews and national clinical guidelines including NICE guidelines (nice.org.uk) and the Danish National Clinical Guidelines (sundhedsstyrelsen.dk). We will contact authors of the included studies and pharmaceutical companies producing extended‐release formulations of methylphenidate for additional and unpublished data. Other sources for clinical trial data will include the Yale University Open Data Access (YODA) project (yoda.yale.edu).

Data collection and analysis

Selection of studies

Two review authors (KB and PD) will independently screen titles and abstracts obtained from the literature search for eligibility. They will then obtain and compare full‐text reports of potentially eligible studies against the inclusion criteria (Criteria for considering studies for this review). Both review authors will resolve disagreements by discussion, seeking consensus with a third review author (KJ), if necessary. We will use either Reference Manager 12 (Reference Manager 2008) or Covidence (Covidence 2015) to manage the screening process. We will list the reasons for exclusion of full‐text reports in a 'Characteristics of excluded studies' table. We will record the study selection process in a PRISMA diagram (Moher 2009).

Data extraction and management

Two review authors (KB and PD) will independently extract information about study methodology, population characteristics and outcomes using a prespecified data extraction sheet. They will compare the extracted data and resolve disagreements by discussion. Consensus will be sought with a third author (KJ or PG), if necessary. One review author (KB) will enter data into Review Manager 5 (Review Manager 2014), and a second review author (PD) will check it for accuracy.

Assessment of risk of bias in included studies

Two review authors (KB and PD) will independently assess each trial for risk of bias using the Cochrane 'Risk of bias' tool (Higgins 2011a). Consensus will be sought, and disagreements will be resolved through discussion. If necessary, consensus will be sought with a third author (KJ or PG). We will use Covidence 2015 for the 'Risk of bias' assessments and consensus. The 'Risk of bias' tool consists of seven domains, as described below, and each domain is attributed a rating of low risk of bias, unclear risk of bias or high risk of bias (Higgins 2011a).

  1. Random sequence generation (selection bias): was the sequence generated adequately (e.g. computer‐generated) or inadequately (e.g. using the day of study enrolment to allocate participants)?

  2. Allocation sequence concealment (selection bias): was the implementation of the randomisation sequence adequate (e.g. central allocation by a third party) or inadequate (e.g. open allocation or using non‐opaque envelopes)?

  3. Blinding of participants and personnel (performance bias): were the methods to maintain the blinding of participants and personnel other than those measuring outcomes during the study adequate or inadequate (i.e. due to the drug's side effects)?

  4. Blinding of outcome assessment (detection bias): were the methods to maintain the blinding of those measuring the outcomes during the study adequate or inadequate (i.e. due to the drug's side effects)?

  5. Incomplete outcome data (attrition bias): were missing data adequately addressed, and were dropout rates balanced?

  6. Selective outcome reporting (reporting bias): were the primary and secondary outcomes fully reported, or not?

  7. Other potential sources of bias: was the study free of other sources of bias such as baseline differences?

Specific limitations to the external validity

We will describe three study design features with special importance for ADHD drug trials. These features are covered by the 'indirectness' domain in the GRADE tool, but we want to highlight these limitations separately. They will be evaluated as additional items within the Cochrane 'Risk of bias' tool (Higgins 2011a), and rated as high, low or unclear risk of limiting the external validity of each trial. These domains must not be mistaken for additional 'Risk of bias' domains (internal validity) as they assess the generalisability (external validity) of the trial.

1. Psychiatric comorbidity

Adults with an ADHD diagnosis often suffer from psychiatric comorbidity (Sobanski 2006). The exclusion of participants with comorbidity before randomisation will not only reduce the trials’ external validity (Surman 2010), but may also lead to an overestimation of the treatment effect (Pliszka 1989; Sobanski 2007b).

  1. Low risk: participants with psychiatric, comorbid psychopathology were included unless methylphenidate was contraindicated (e.g. bipolar affective disorder or suicidality).

  2. Unclear risk: there was no, or an unclear, description of whether or not participants with psychiatric comorbidity were allowed.

  3. High risk: participants with psychiatric comorbidities were excluded before, or after, randomisation.

2. Responder selection ('enriched design')

Some participants in methylphenidate trials do not benefit, or they get worse. These participants are referred to as "non‐responders" (Cho 2007; de Sonneville 1994; Elia 1991; Wilens 2002). It may be ethically unacceptable to include patients known not to benefit or who experience harm, but their exclusion leads to an overestimation of benefits and an underestimation of harms compared to a treatment‐naïve population, especially if participants known to benefit from previous exposure to the intervention are included. Despite the negative consequences for the external validity of the trial, such a study design is called an 'enriched design' (FDA 2012).

  1. Low risk: participants were treatment‐naïve.

  2. Unclear risk: there was no, or an unclear, description of exclusion criteria or of previous use of stimulants.

  3. High risk: 'non‐responders' (or similar) were excluded before randomisation. Similarly, trials that allowed previously‐medicated participants to enrol will be rated at high risk of responder selection.

3. Withdrawal effects

Withdrawal of stimulant drugs can lead to rebound symptoms (Cox 2008). Such withdrawal effects can occur also after a washout period, that is, when the patients have been randomised. Thus, patients in the placebo group may suffer from abstinence symptoms, which might be interpreted as a deterioration of the disorder.

  1. Low risk: participants were treatment‐naïve.

  2. Unclear risk: there was no, or an unclear, description of whether only treatment‐naïve participants were randomised.

  3. High risk: stimulants were stopped before randomisation, with or without a washout.

Measures of treatment effect

Dichotomous data

We will summarize dichotomous data as risk ratios (RR) with corresponding 95% confidence intervals (CIs). We will not extract composite outcomes such as dichotomous responder rates that consist of several outcomes combined.

Continuous data

For continuous data measured on the same scale, we will extract mean change or endpoint data (preferred) in order to calculate a mean difference (MD) with 95% CIs. For continuous outcomes measured on different scales, we will extract mean change from baseline or endpoint data (preferred) and the corresponding standard deviations (SDs) or standard errors (SEs), to calculate a standardised mean difference (SMD) with 95% CIs.

Unit of analysis issues

Cross‐over trials

We will exclude cross‐over trials due to the risk of carry‐over effects and the ensuing loss of blinding. For example, in a study of immediate‐release methylphenidate (Gualtieri 1985), all eight participants were able to guess their allocation, most likely due to the adverse events. Furthermore, we will exclude these trials because long‐term follow‐up is not possible.

Studies with multiple time points

We will reduce the impact of multiple analyses and split up outcome periods into short term (up to six months), medium term (up to 12 months), and long term (more than 12 months), to identify any waning of treatment effects over time. If a trial reports multiple sets of study data within the same predefined outcome period, we will extract data with the longest follow‐up time.

Studies with multiple treatment groups

In studies with multiple methylphenidate groups, we will combine the experimental groups into a single group, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b), making a single, pair‐wise comparison.

Dealing with missing data

A drug class review from 2011 reported that a considerable amount of data were missing in ADHD medication trials (McDonagh 2011). We will highlight 'not missing at random' data such as omission of blood pressure or adverse events data. As we do not expect to get access to individual patient data, we will not be able to impute missing data; instead we will use best available (incomplete) data and take the following precautions:

  1. Substitute missing SDs based on trials of similar size using the same rating scale.

  2. Extract and present dropout rates in a meta‐analysis. We will be careful to use the correct, full population sizes instead of limited and otherwise restricted population sizes (e.g. per‐protocol populations etc.).

  3. Extract and discuss each trial's methodology for dealing with missing continuous data (e.g. last observation carried forward or modified intention‐to‐treat).

  4. Perform sensitivity analyses comparing trials with high and low dropout rates (see: Sensitivity analysis). We will use our assessments of attrition bias for each individual trial to differentiate between high and low dropout (see: Assessment of risk of bias in included studies). We will define trials at high and unclear risk of attrition bias as having a high dropout rate. We will define trials at low risk of attrition bias as having a low dropout rate. By using the attrition bias judgment we can dichotomise between high and low dropout based on an overall assessment of the dropout rate, reasons for dropouts and the statistical methodology to account for missing continuous data in each study.

  5. We will prefer analyses based on imputed rather than completer's data, and make it clear if such choice has been made.

Assessment of heterogeneity

We will assess clinical heterogeneity by exploring whether individual trial results are related to differences in trial populations, interventions or settings, with a special emphasis on three specific study features; the exclusion of participants with psychiatric comorbidity, the exclusion of participants who have previously not tolerated or not responded well to methylphenidate (called non‐responders), and the use of wash‐out phases prior to randomisation. We have described these study features in detail elsewhere (see: Assessment of risk of bias in included studies). We will assess methodological heterogeneity using the Cochrane 'Risk of bias' tool (see: Assessment of risk of bias in included studies). We will assess statistical heterogeneity using the I2 statistic and Chi2 test (Deek 2011). We will explore reasons for heterogeneity through subgroup (see: Subgroup analysis and investigation of heterogeneity) and sensitivity (see: Sensitivity analysis) analyses when the I2 statistic is greater than 50%. We will also report Tau2 as an estimate of between‐study variance as we will be using the random‐effects model.

Assessment of reporting biases

We will assess selective outcome reporting using the Outcome Reporting Bias in Randomised Controlled Trials (ORBIT) tool (Kirkham 2010). Arbitrary thresholds that exclude adverse event reporting (i.e. that a certain percentage of participants must experience an adverse event before it is reported) will be highlighted and considered as selective outcome reporting in our 'Risk of bias' assessment (see: Assessment of risk of bias in included studies). Furthermore, we will compare predefined primary and secondary outcomes as stated in trial registries and published protocols with the actual published outcomes, to test for outcome switching and outcome reporting bias (Goldacre 2016). We will assess publication bias, and other small study effects, using funnel plots with estimated effect sizes plotted against their standard error (Egger 1997), but only if more than 10 studies report a given outcome (Sterne 2011). Asymmetry in the funnel plot could be due to publication bias or a real relationship between trial size and effect sizes. We will use Egger’s statistical test in the case of asymmetry. We will test for publication bias and selective outcome reporting in our subgroup analysis of published versus unpublished data that we may obtain from trial registries, trialists, databases such as YODA, and other relevant sources (see: Subgroup analysis and investigation of heterogeneity).

Data synthesis

We will summarise data using random‐effects meta‐analyses with inverse‐variance weighting due to anticipated clinical, methodological and statistical heterogeneity (McDonagh 2011; Storebø 2015), and present data in 'Summary of findings' table(s). All studies are eligible for meta‐analyses regardless of their risk of bias and methylphenidate dosage, as there is no established dose‐response relationship for methylphenidate (NICE 2009). We will combine the methylphenidate groups in studies with multiple arms using different dosages of the drug, as already described (see: Unit of analysis issues).

Trial Sequential Analysis

We will apply the Trial Sequential Analysis model to the dichotomous outcomes of adverse and serious adverse events, but not to the continuous outcomes because the Trial Sequential Analysis currently assumes MDs and not SMDs. It is a statistical model, similar to interim analyses in clinical trials, used to quantify the reliability of data in cumulative meta‐analyses, adjusting the P values for sparse data and multiplicity (Brok 2008; Wetterslev 2008). The required information size (the number of participants required to accept, or reject, the hypothesis of a certain a priori anticipated effect) is calculated using the following five components.

  1. Alpha = 0.05 (type 1 error).

  2. Power = 0.90 (type 2 error 0.10).

  3. Proportion (frequency) of participants experiencing serious adverse events and adverse events (based on observations).

  4. Relative risk reduction (RRR) or increase of 20%.

  5. Diversity (heterogeneity based on our observations. In case D = 0%, we include an analysis with D = 25%).

Preferably, this model should be applied to studies with a low risk of bias only, but we will conduct analyses that also include studies with high risk of bias. It could establish firm evidence of harms, even in the absence of low risk of bias studies, as the bias most likely leads to an underestimation of harms.

Summary of findings

We will create a 'Summary of findings' table for the primary (Primary outcomes) and secondary outcomes (Secondary outcomes) with placebo comparisons. We will use the GRADE approach to evaluate the confidence in the effect estimate for each outcome according to its five elements: risk of bias, publication bias, imprecision, inconsistency and indirectness (Schünemann 2013). Two review authors (KB and PD) will independently assess the quality of evidence for each outcome as either high, moderate, low or very low.

Subgroup analysis and investigation of heterogeneity

  1. Psychiatric comorbidity: trials that excluded participants with psychiatric comorbidity compared to trials that did not.

  2. Responder selection: trials with responder selection compared to trials without responder selection ('enriched design').

  3. Treatment status: trials with treatment‐naïve participants compared to trials with previously‐treated participants.

  4. Washout: trials containing a washout phase of stimulant medication prior to randomisation compared to trials with no washout phase.

  5. Trial data: unpublished trial data compared to published trial data.

  6. Vested interests: industry‐sponsored trials compared to other trials.

Sensitivity analysis

  1. Restricting the analyses to trials with an overall low risk of bias.

  2. Restricting the analyses to trials with low dropout rates, as defined in the 'Dealing with missing data' section.