Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Methylphenidate for core and ADHD‐like symptoms in children aged 6 to 18 years with autism spectrum disorders (ASDs)

Authors' declarations of interest

Version published: 13 August 2014 Version history

https://doi.org/10.1002/14651858.CD011144

This is not the most recent version

view the current version 21 November 2017
Collapse all Expand all

Abstract

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

To determine the efficacy and safety of methylphenidate for 1) features of ADHD such as hyperactivity, impulsivity, and concentration, and 2) the core features of ASD (social, communication, and behavioural problems) in children aged six to 18 years with ASD.

Background

Description of the condition

The autism spectrum is a group of developmental disorders described in the Diagnostic and Statistical Manual of Mental Disorders (DSM) DSM‐IV‐TR as autistic disorder, Asperger's disorder, Rett's disorder, childhood disintegrative disorder, and pervasive developmental disorder not otherwise specified (PDD‐NOS) (Wing 1997; AAP 2001). DSM‐V uses the diagnosis of autism spectrum disorder (ASD) for individuals with a DSM‐IV diagnosis of autistic disorder, Asperger's disorder or pervasive developmental disorder not otherwise specified (Hyman 2013). ASD encompasses disorders previously referred to as early infantile autism, childhood autism, Kanner’s autism, high‐functioning autism, atypical autism, pervasive developmental disorder not otherwise specified, childhood disintegrative disorder, and Asperger’s disorder (Hyman 2013). The 10th revision of the International Classification of Diseases (ICD‐10) classifies childhood autism as a disorder of psychological development, defined by 1) the presence of abnormal or impaired development that is manifest before three years of age, and 2) the characteristic type of abnormal functioning in all three areas of psychopathology: reciprocal social interaction, communication, and restricted, stereotyped, repetitive behaviour (WHO 2007).

DSM‐IV describes three characteristic manifestations of ASDs: 1) impaired social interaction; 2) impairment communication; and 3) restricted repetitive and stereotyped patterns of behaviour (Wing 1997; Filipek 1999). DSM‐V uses five diagnostic criteria for the diagnosis of ASD: 1) persistent deficits in social communication and social interaction across multiple contexts; 2) restricted, repetitive patterns of behaviour, interests, or activities; 3) symptoms are present in the early developmental period; 4) symptoms cause clinically significant impairment; and 5) these disturbances are not better explained by intellectual disability or global developmental delay (Hyman 2013). The studies that will be reviewed use the DSM‐IV or DSM‐IV‐TR Diagnostic Criteria, where they refer to the DSM, as the DSM‐V was only published in May 2013.

A systematic review of prevalence studies (Williams 2006) calculated an overall prevalence estimate of 7.1 per 10,000 (95% CI 1.6 to 30.6) for typical autism and a prevalence of 20.0 (95% CI 4.9 to 82.1) for ASD. The estimated prevalence of ASD in Australia is 62.5 per 10,000 for children aged six to 12 years (AABASD 2007). In recent years, reported frequencies for ASD across the US and other countries have approached 1% of the population, with similar estimates in child and adult samples (Brugha et al. 2011). Some of the variation in prevalence of ASD quoted in studies is likely to be due to the use of different diagnostic criteria and the age of children in the sample.

Since the early observations of autism by Kanner in 1943 (Kanner 1943), the autism spectrum is recognised as a heterogenous group of disorders with varying clinical presentations and severities (AAP 2001; NICHD 2014). The onset is typically before three years of age (APA 1994), although the diagnosis is often not made until two or three years after symptoms are evident (Filipek 2000). Children commonly present with speech delay, poor eye contact, social impairment, unusual or repetitive play, need for routine, difficulty coping with change, and obsessions. A significant proportion have intellectual impairment, but many have an intelligence quotient in the normal range. All groups may also present with abnormal movements or stereotypies, heightened levels of anxiety, and self injurious or aggressive behaviour. In addition to specific diagnostic features, common non‐specific problems include phobias, sleeping and eating disturbances, temper tantrums, and (self directed) aggression (WHO 2007). Abnormalities of attention (overly focused or easily distracted) are common in individuals with ASD, as is hyperactivity, and many meet the DSM criteria for attention deficit hyperactivity disorder (ADHD) (Handen 2000; Hyman 2013). Hyperactivity and inattention can lead to poor school performance and academic underachievement, further social impairment due to inappropriate and impulsive behaviours, and can place the child at risk of harm. These behaviours are not only a significant impairment for the child, but also affect the parents, carers, and siblings. There is often emotional and psychological stress, along with financial strain due to the time needed for supervision of the child, medical appointments, and therapy.

Parr 2008 reports that "About 15 percent of adults with autism live independent lives, whereas 15 to 20 percent live alone with community support." For the ASD population "...verbal and overall cognitive capacities seem to be the most important predictors of an ability to live independently as an adult" (p.760). 

Diagnosis of ASD includes history from parents and teachers, clinical observation, psychological and often speech language assessment. Numerous behaviour rating scales have been developed both to aid the assessment of ASD as well as monitor response to therapy. Examples of commonly used rating scales include the Autism Diagnostic Observation Schedule ‐ generic (ADOS) (Lord 2000) and the Autism Diagnostic Interview, Revised (ADI‐R) (Scahill 2005; Bertoglio 2009). Other scales are used in the context of screening, including the Childhood Autism Rating Scale (CARS) (Handen 2000; AAP 2001; Scahill 2005; Scahill 2006; Bertoglio 2009); the Children's Global Assessment Scale (CGAS) or a modified version known as the Developmental Disabilities‐CGAS (for children under 18 years of age with pervasive developmental disorder) (Wagner 2007); the Australian Scale for Asperger's Syndrome (ASAS) (Filipek 1999); and the Autism Behaviour Checklist (ABC) (AAP 2001; Scahill 2005; Bertoglio 2009). There is much variability between the rating scales with no clear consensus regarding which measure provides the most reliable results to aid diagnosis. Some tools, such as ASAS, are targeted at primary school‐aged children with possible Asperger's. There are also a number of generic rating scales, such as the C‐GAS and the CBCL (Achenbach 1991), that can be used to monitor behaviour in individuals with ASD.

Description of the intervention

The mainstays of treatment for ASD are therapeutic, with pharmacological treatments targeted at specific behaviours. The SIGN 2007 guideline summarises the evidence for non‐pharmacological interventions and suggests that behavioural interventions should be considered to address a wide range of specific behaviours. There is also some evidence for tailored social communication and communication interventions (such as the use of visual augmentation) (SIGN 2007). Pharmacological therapy, including atypical antipsychotics and antidepressants, target specific behaviours and aim to achieve a global improvement. Pharmacological treatments have been widely used and, to date, much of the evidence has been from small randomised controlled trials (RCTs). NICE 2013 does not support the use of antipsychotics, antidepressants, and anticonvulsants. Recently, Cochrane reviews have looked into the use of selective serotonin reuptake inhibitors (SSRIs) in children with ASD and concluded that there is no evidence to support the use of SSRIs to reduce core features of ASD and anxiety in these children (Williams 2010). Another Cochrane review has looked into the use of risperidone in targeting behavioural problems in ASD and concludes that there is a limited role for risperidone, perhaps only short term and targeted at specific behaviours (Jesner 2007).

Psychostimulants are commonly prescribed to target ADHD behaviours (inattention, distractibility, and impulsivity) in children with ASD. Of the psychostimulants (that is, methylphenidate, dexamphetamine, pemoline, and modafenil), methylphenidate is by far the most widely used in the treatment of ADHD to improve concentration and reduce the symptoms of impulsivity and hyperactivity. In this review, therefore, we will focus on methylphenidate. Another Cochrane review has studied the effect of dexamphetamine in adults with ADHD (Castells 2011) and the other stimulants are much less commonly used in clinical practice.

Methylphenidate primarily acts as a dopamine‐noradrenaline reuptake inhibitor. Its effect on adrenaline is much weaker than its effect on dopamine. It acts by inhibiting dopamine‐noradrenaline reuptake transporters. Treatment with immediate‐release formulations should be initiated at 5 mg once or twice daily, up to a maximum of 60 mg per day. Treatment with modified‐release formulations of methylphenidate should be initiated at a dose of 18 mg once daily (in the morning), and increased, if necessary, up to a maximum of 54 mg once daily. (NICE 2013). Methylphenidate is also available in some countries as an extended‐release form, transdermal patch (Mayo Clinic 2014). Short‐release systems of methylphenidate are absorbed rapidly within 30 minutes and their effects last up to four to six hours. Long‐acting or modified‐release forms contain immediate and delayed‐release mechanisms, and are taken once daily (in the morning). They are available in eight‐hour and 12‐hour preparations. The advantages of these controlled‐release products include the possible decrease in likelihood and severity of rebound symptoms (Szymansk 2001), and increase in compliance in children due to reduced dosing frequency and fewer tablets.

The optimal dose of methylphenidate is based on observations of clinical response by the individual, as individual responses are variable and not dose predictable. The dose is initially titrated by increasing the daily dose every week until the effects are observable, or until adverse effects warrant dose reduction or cessation. In children with ADHD, the ability to attend and focus, particularly in busy environments, such as the classroom, indicates clinical effect. In short‐acting forms, the maximum recommended dose is 1.5 mg/kg/day or 60 mg in two to three divided doses (TGL 2012). This can be given as a once‐daily dose in the combination forms. Most side effects of methylphenidate are dose dependent (Rossi 2010). Common side effects include headaches, loss of appetite, abdominal discomfort, nausea, anxiety, and insomnia (Rossi 2010). These effects could lead to reduced tolerance of the drug. Methylphenidate also increases blood pressure and heart rate, so consultation with a cardiologist is required before starting the drug in children with cardiac abnormalities. It also has the potential to infrequently cause serious conditions, such as growth restriction, psychosis, liver dysfunction, and neuroleptic malignant syndrome, for which monitoring is needed (Medsafe 2010).

Clinical studies assessing the impact of treatments for hyperactivity symptoms in children with ASD are important to guide clinicians (Nicolson 2000; Siegel 2012). However, the variability of presentations and diagnostic criteria used in trials is high and it remains challenging to identify subgroups of children who will benefit most from treatment (Siegel 2012). In this review, we wish to assess the overall effectiveness of methylphenidate in these children, by evaluating the evidence both for changes in ADHD symptoms and also for changes in the core symptoms of ASD. This is because we cannot exclude a priori the possibility that the effect of methylphenidate on one of these sets of symptoms is in a different direction from its effect on the other set, which might negate any overall clinical benefit.

How the intervention might work

Methylphenidate blocks the dopamine and noradrenaline transporter, thereby increasing extracellular dopamine levels (Volkow 2001). In children with ADHD this has been shown to reduce impulsivity and increase their ability to concentrate. The mechanism of action has not been clearly defined but may be due to the role of dopamine in task‐specific signalling or in the motivation and reward pathway (Volkow 2001). The possible mechanisms underpinning the effect of methylphenidate in children with ASD, who display behaviours similar to those in ADHD (for example, impulsivity and overactivity), have not yet been established.

Why it is important to do this review

The use of psychostimulants, and methylphenidate in particular, for ASD has been controversial. Psychiatrists and paediatricians have questioned its efficacy and expressed concerns that the drug could worsen symptoms. Recently, a number of small trials have suggested that methylphenidate may have a role in the management of core symptoms as well as ADHD‐like symptoms in children with ASD (Santosh 2006). A review of the literature concludes that psychostimulants are effective in the treatment of ADHD‐like symptoms in individuals (children, adolescents, and adults) with ASD (Cortese 2012), however they did not investigate the effect on core symptoms of ASD.

Not only does ASD affect the individual, but there is also a significant impact on the family. A study evaluating the coping strategies of families who have children with ASD reports that "...the parents of children with ASD have somewhat lower marital happiness and family cohesion.." (Higgins 2005 P. 125). In addition, another study reports that "..siblings [are faced with] precocious responsibility whilst they also have a greater risk of developing internalising behaviour problems" (Benderix 2007 P. 416). Furthermore, the financial cost of caring for a child with ASD is twice that of caring for a child without ASD (Croen 2006).

There is considerable recognition of the role of attention deficits in ASD and how such deficits might impact a range of behaviours and impairments associated with the disorders (Burack 1997; RUPP Autism Network 2005). Studies have shown that up to 12% of children with autism are prescribed stimulant medications to treat these secondary features of impaired concentration, impulsivity, and hyperactivity (Handen 2000), indicating that behaviours typically seen in ADHD are frequently treated in this population. But if children with ASD respond to traditional pharmacological treatments for ADHD‐like symptoms at rates similar to typically developing children, there is little available data to support such an assumption or to guide clinicians (Handen 2000). Since psychostimulants are frequently used, it is important to determine their effectiveness.

In this review, we aim to explore whether methylphenidate has an impact on core ASD features of communication skills, social interactions, and stereotypical or repetitive behaviours, and on global function, concentration, the ADHD‐like behaviours, and the rate of adverse effects. It is important to determine the rate of side effects experienced by the participants and whether this outweighs any significant benefit or influences treatment compliance.

Objectives

To determine the efficacy and safety of methylphenidate for 1) features of ADHD such as hyperactivity, impulsivity, and concentration, and 2) the core features of ASD (social, communication, and behavioural problems) in children aged six to 18 years with ASD.

Methods

Criteria for considering studies for this review

Types of studies

Randomised (double blinded) controlled trials (RCTs) with a parallel group design.

Types of participants

Children and adolescents between six and 18 years of age, diagnosed with ASD using DSM‐IV‐TR (APA 2000) or DSM III (APA 1987) criteria, or diagnosed with pervasive developmental disorder (PDD) according to the ICD‐10 (WHO 2007). We will include children with concurrent diagnoses such as an anxiety disorder, intellectual impairment, learning delays, speech impairment, and congenital syndromes such as fragile X syndrome. We will include children receiving cointerventions, including psychotropic medications, dietary modifications, and therapist‐based interventions such as speech, occupational, and psychological therapy. We will exclude children with other established causes of cognitive or behavioural problems such as acquired brain injury.

Types of interventions

We will include trials if they use methylphenidate, irrespective of formulation or dose, and compare it to a placebo. We will discuss the comparability of the placebo intervention, that is if methylphenidate is delivered as a transdermal patch, the placebo intervention will have to be a similar formulation. All other oral formulations will be compared with similar oral placebo. We will include trials in which methylphenidate is given in addition to other psychotropic medications if these are provided to participants in both arms (or as adjunctive treatment). The treatment can be administered in any setting, including the home, hospital, or residential care.

Types of outcome measures

Primary outcomes

We will define clinical efficacy as:

  1. an improvement of features of ADHD (including attention, impulsivity, and hyperactivity); and

  2. an improvement of the core symptoms of ASD:

  • impaired social interaction

  • impairment in communication

  • restricted, repetitive, and stereotyped patterns of behaviour.

We will assess short term (one to three months) and long term (six to 12 months) outcomes.

Secondary outcomes

  1. Rate of adverse side effects, for example, common side effects of nausea, insomnia, and hypertension, and also more serious side effects such as growth retardation.

  2. Carer well‐being: levels of parental stress, using scales such as the Parental Stress Index (Abidin 1983).

  3. Need for institutionalisation of children or adolescents, or both, or special schooling options, for example, if therapy is required or to achieve learning outcomes due to decreased attention, or disturbance of home life due to hyperactive or impulsive behaviours.

  4. Overall quality of life of the child or adolescent, or both.

We will assess short term (one to three months) and long term (six to 12 months) outcomes.

We will list the primary outcomes and adverse effects in a 'Summary of findings' table indicating levels of evidence for the findings. We will use GRADEpro software (GRADEpro) to create the table.

Search methods for identification of studies

Electronic searches

We will search for studies in the databases listed below. We will not limit searches by publication date or language.

  • Cochrane Central Register of Controlled Studies (CENTRAL)

  • Ovid MEDLINE

  • Ovid MEDLINE In‐Process & Other Non‐indexed Citations

  • EMBASE

  • Cochrane Database of Systematic Reviews (CDSR), part of the Cochrane Library

  • Database of Abstracts of Reviews of Effects (DARE)

  • CINAHL

  • PsycINFO

  • ERIC

  • Science Citation Index (SCI)

  • Social Sciences Citation Index (SSCI)

  • Conference Proceedings Citation Index: Science (CPCI‐S)

  • Conference Proceedings Citation Index: Social Sciences & Humanities (CPCI‐SSH)

  • AutismData (autism.org.uk/autismdata)

  • ClinicalTrials.gov (clinicaltrials.gov)

  • WHO IRCTP (apps.who.int/trialsearch/)

  • Proquest Dissertations & Theses

We will search Ovid MEDLINE using the following search strategy and adapt it for use with the other databases listed above.

1 Central Nervous System Stimulants/

2 Methylphenidate/

3 Methylphenidat$.mp.

4 (Attenta$ or Biphentin$ or Centedrin$ or Concerta$ or Daytrana$ or dexmethylphenidat$ or Equasym$).mp.

5 (Focalin$ or Medikinet$ or Metadate$ or Methylin$ or Methylphenidat$ or Penid$ or Phenidyl$ or Ritalin$ or Rubifen or tranquilyn$ or Tsentedrin$).mp.

6 or/1‐5

7 exp child development disorders, pervasive/

8 Developmental Disabilities/

9 pervasive development$ disorder$.tw.

10 (pervasive adj3 child$).tw.

11 (PDD or PDDs or PDD‐NOS or ASD or ASDs).tw.

12 autis$.tw.

13 asperger$.tw.

14 kanner$.tw.

15 childhood schizophrenia.tw.

16 Rett$.tw.

17 or/7‐16

18 6 and 17

Searching other resources

  1. We will contact the drug manufacturers by email, including Mallingkrodt, Novartis, Janssen Pharmaceuticals, Shire, and Medice (Medikinet) to obtain unpublished data.

  2. We will search the references of relevant studies and (systematic) reviews to identify additional studies.

  3. We will contact the first author of included RCTs to enquire about other relevant studies.

  4. We will contact known specialists in developmental paediatrics for additional studies.

Data collection and analysis

Selection of studies

Two review authors (ES, MVD) will read the title and abstracts to determine suitability according to the criteria mentioned above. Any disagreements will be resolved by discussion between the two review authors; otherwise, a third review author (TR) will act as arbiter. We will list any excluded studies that appear applicable, along with the reasons for exclusion. We will create a PRISMA flow diagram illustrating the selection process (Liberati 2009).

Data extraction and management

Two review authors (ES, MVD) will extract the following data using the piloted data extraction sheet: type of study, participants, type of intervention (including dose and administration form), and reported outcomes. A third review author (TR) will check the extracted data in case of discrepancies that cannot be resolved by discussion.

Assessment of risk of bias in included studies

We will use a checklist to assess risk of bias and it will include the criteria described below (Higgins 2011). Two review authors (ES and MVD) will independently assess the studies. If there are any disagreements, a third review author (TR) will evaluate the study and a consensus will be reached. We will refer to the Cochrane 'Risk of bias' assessment tool (Higgins 2011) to classify the studies as either 'low', 'unclear', or 'high' risk of bias for each of the following domains.

1. Sequence generation ‐ determines whether generation of the random numbers was adequate. We will assess the risk of bias of sequence generation as low, high or unclear.

    • Low – computer generated random numbers or random number tables.

    • High – random numbers are generated by sequentially allocating groups.

    • Unclear – when information about the generation of random numbers is described inadequately or not at all.

2. Allocation concealment ‐ determines whether the method used to conceal allocation was adequate to prevent selection bias during the randomisation process before allocation. We will assess the risk of bias of allocation and concealment as low, high or unclear.

    • Low – used methods such as central allocation, or sealed opaque envelopes.

    • High – participants or investigators could possibly foresee the allocated treatment. For example, numbering participants and only including even numbers in the control group.

    • Unclear – when the method of allocation concealment is described inadequately or not at all.

3. The method of blinding participants and personnel. We will assess the risk of bias related to blinding of participants and personnel as low, high or unclear.

    • Low – participants or investigators are unable to determine the treatment allocated.

    • High – participants or investigators could possibly determine the treatment allocated.

    • Unclear – information about blinding is insufficient.

4. The method of blinding outcome assessors. We will assess the risk of bias related to blinding of outcome assessors as low, high or unclear.

    • Low – outcome assessors are unable to determine the treatment allocated.

    • High – outcome assessors have knowledge or could have knowledge of the allocated treatment and this could have influenced their assessment (for example, in the case of instruments assessed through interview).

    • Unclear – information about blinding of outcome assessors is insufficient.

5. Evaluate whether missing data was accounted for. Criteria will include:

    • whether attrition and exclusions were reported;

    • reasons for the attrition and exclusion;

    • number of patients in each intervention group;

    • reinclusions in analyses performed by review authors; and

    • whether imputations were used for missing outcomes.

Systematic differences between groups in a study (for example, selective drop out or withdrawal) can bias the study results. Therefore, we will carefully assess if attrition between groups is different and if so how this is explained. We will classify studies as "high risk of bias" in this section if no information is provided about attrition or missing data and no attempt has been made to address this in the analysis of the study. In case imputations were used for missing outcomes we will assess if these imputations are in line with the expected outcomes and how this was performed.

6. Selective reporting, that is, assess if all planned outcomes are reported. We will assess this by comparing the reported outcomes with those published in the protocol of the study, if available. If a previously published protocol is not available, we will compare the outcomes described in the methods section of the paper with the reported outcomes in the same paper. We will assess risk of bias related to selective reporting as low, high or unclear.

    • Low – all planned outcomes are reported.

    • High – not all planned outcomes are reported and no reasons given.

    • Unclear – insufficient information on planned outcomes available.

7. Other sources of bias, for example, funding of the trial and conflicts of interest of the authors or investigators.

We will assess studies funded by a manufacturer or studies authored by one or more employees of a manufacturer as "high risk of bias" in this category unless there is an explicit and sufficient description of the independence of the funding source and employees in the analysis and reporting of the study results. We will assess conflicts of interest declared by the authors in a similar way. We will scrutinise and discuss any undeclared, but known (through other resources) conflicts of interest of authors.

Measures of treatment effect

The outcomes in many of the studies relevant to the review are measured using psychometric scales, which are a form of continuous data. If the same scales are used to measure outcomes across studies, we will use the mean difference (with standard deviation SD) between groups to estimate the effect.

If the scales used in studies are different but outcomes are conceptually the same, we will calculate the standardised mean difference (SMD). The SMD expresses the size of the intervention effect in each included study relative to the variability observed in that study (Higgins 2011). This method is based on the assumption that the differences in SDs between trials reflect differences in scales and not real differences in variability between the populations included in the trials. The SMD is 'scale free', that is, since the dependent variable is standardised, the original units are replaced by standardised units. We will use the Hedges' g formulation to calculate the SMD. It is calculated by dividing the means between groups by the SD. Hedges' g uses a weighting to account for the population sizes in the different studies (Egger 2001). In case information on SDs in each of the studies is missing, we will calculate the SD from the standard error or confidence interval (CI) of the mean in each of the studies as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). If none of the required variables are available, we will not pool the data.

We will report dichotomous outcomes (such as adverse effects and institutionalisation) as risk ratios (RRs), which are calculated as the proportion of patients in the treatment group who experience the outcome (or event) divided by the proportion of patients in the control group who experience the outcome (or event). If there are no events in either the treatment group or the control group, we will not use the RR; we will report these outcomes separately and will not pool them. We will report the pooled result as RRs using Review Manager software (RevMan 2012). We will calculate 95% CIs for all dichotomous outcomes. We will calculate the risk difference (RD) and the number needed to treat (NNT) in case there is a significant effect and trials are sufficiently homogeneous.

In case the same outcomes are reported as continuous variables in some studies and dichotomous variables in others (for example, when outcomes are expressed as a score on a scale in one study and as a proportion of responders on the same scale in another study), we will attempt to combine data in three ways as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will pool the means and SDs as continuous outcomes, pool the counts as dichotomous outcomes, and describe the data in text form as ‘other data’ outcomes. We will not combine them into one outcome estimate as this may jeopardise the accuracy of the clinical interpretation of the results.

Unit of analysis issues

The unit of analysis will usually be the individual participant. In situations where this is not the case, for example repeated observations on participants or in cluster‐randomised trials, we will undertake the appropriate analysis (see below) that takes these variations into account (Higgins 2011).

Cluster‐randomised trials

In cluster‐randomised trials, groups rather than individuals are randomised, which requires an adjustment to be made to account for the clustering effect. If trials have used cluster randomisation, we expect that cluster effects have been appropriately controlled for (robust standard errors or hierarchical linear models). If it is unclear whether appropriate controls for clustering were applied, we will contact the investigators for further details. If appropriate controlling was not used, we will request and reanalyse individual participant data using appropriate multilevel models. Following this, we will analyse effect sizes and standard errors in Review Manager (RevMan 2012) using the generic inverse method (Higgins 2011). If there is insufficient information to control for clustering, we will enter outcome data using individuals as the units of analysis, and then use sensitivity analysis to assess the impact of inadequately controlled cluster‐randomised trials on the effect estimate.

The method for adjustment for clustering suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) is based on reducing the size of effect of each clustered trial to its ‘effective sample size’, which is the original sample size of the cluster‐randomised trial divided by the ‘design effect’ (i.e. 1 + (M – 1) *ICC, where M is the average cluster size and ICC is the intracluster correlation coefficient) (Higgins 2011). The ICC will be estimated based upon similar studies (in similar populations, but including other treatments). This will enable us to pool non‐clustered and clustered trials to obtain an overall estimate of effect.

Trials with repeated measurements

In trials with repeated measurements for the same patient (for example, measurements at different time points), we will attempt to reduce the impact of multiple analysis by analysing the most frequently reported or the most clinically relevant time points (usually the longest duration of follow‐up, which is likely to be the best indication of a clinically sustainable effect) or both (Higgins 2011).

Trials with multiple treatment arms

We will attempt to combine arms to create a single pair‐wise comparison where appropriate (for example, slow‐ or controlled‐release and immediate‐release methylphenidate formulations). If this is not possible, we will use all treatment groups but split the comparison (placebo) group evenly across the intervention groups (Higgins 2011). Criteria for assessing the relevance of the treatment arms for each comparison will include clinical relevance (is the expected clinical effect likely to be different or not?) and clinical availability or use of the treatment in question (for example, transdermal formulations of methylphenidate are not commonly used, therefore if a trial has two active treatment arms with one of them a transdermal patch we may choose to consider only the oral formulation).

Dealing with missing data

We will contact the authors of the included studies for information regarding missing data, drop‐outs, or data not included in the study report, but relevant to the review. For example, outcomes of interest and summary data (such as number of participants and events) will be sought if they are not included in the published study report. We will report our attempt to obtain additional data and the results of these attempts.

We will perform an intention‐to‐treat (ITT) analysis to account for missing data. The ITT analysis considers all missing participant data of all randomised patients as a treatment failure. We will compare ITT analysis results with the results of "on‐treatment" or "complete case analysis" (all participants completing treatment) or per protocol (all participants following protocol or at least one dose of the allocated treatment) results to assess the impact of missing data on the overall estimate of effect. We will discuss the impact of missing data and perform a Sensitivity analysis, which we will discuss in the Discussion section of the review.

Assessment of heterogeneity

We will assess heterogeneity in two steps. First, we will assess clinical heterogeneity by comparing the populations included in the studies, the settings, the treatment modalities, and the outcomes. Clinical heterogeneity will be considered sufficient to preclude the pooling of studies if: the participant ages are obviously different (for example, we will not combine data from studies of teenage participants aged 16 to 18 years with data from studies of children aged four to six years); the severity of the ASD is obviously different (for example, we will not combine data from studies of children who require institutional care with data from studies of those with mild symptoms causing little impairment); or the outcome measures are not clinically comparable (for example, we will not combine data from a study that only measures impulsivity with data from a study that only measures hyperactivity).

Second, we will assess statistical heterogeneity by performing a Chi2 test and calculating the I2 value according to the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will assess statistical heterogeneity by using the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions for interpretation of the I2 value as follows (Higgins 2011):

  • 0% to 40%: might not be important;

  • 30% to 60%: may represent moderate heterogeneity;

  • 50% to 90%: may represent substantial heterogeneity; and

  • 75% to 100%: considerable heterogeneity.

If clinical heterogeneity is present, we will not pool the studies but only describe them. If statistical heterogeneity is present, we will perform a Sensitivity analysis to explore the source of heterogeneity. We will use appropriate analysis (that is, apply a random‐effects model for pooling) to account for statistical heterogeneity when indicated (see Data synthesis).

Assessment of reporting biases

If there are 10 or more studies included in the review, we will draw a funnel plot. The horizontal axis of the plot describes the intervention effect estimate of a study, and the vertical axis indicates the standard error of the intervention effect estimate of that same study. The funnel plot facilitates the identification of "small study effects" and, in the presence of such small study effects, publication bias should be considered. We will first visually examine the graph. An asymmetrical funnel plot implies that either smaller studies with no significant effect have remained unpublished or smaller studies of lower methodological quality have produced exaggerated effect estimates. If there appears to be asymmetry, we will assess whether the association between estimated intervention effects and study size is greater than what could be attributed to chance. This will be done by using tests described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will use the ‘trim and fill’ method, which entails first ‘trimming’ (removing) the smaller studies causing funnel plot asymmetry, then using the trimmed funnel plot to estimate the true ‘centre’ of the funnel, and subsequently replacing the omitted studies and their missing ‘counterparts’ around the centre (filling). As well as providing an estimate of the number of missing studies, we will calculate an adjusted intervention effect by performing a meta‐analysis, which includes the filled studies (Higgins 2011).

Data synthesis

We will pool the available data using a random‐effects model (Higgins 2011). In the absence of clinical or statistical heterogeneity (see Assessment of heterogeneity), we will also apply a fixed‐effect model for pooling and compare the effect estimates obtained from each of the two methods in order to assess the robustness of the estimates.

We will address the existence of reporting bias as identified by funnel plot analysis as described in Assessment of reporting biases.

Subgroup analysis and investigation of heterogeneity

We will perform the following subgroup analyses.

  1. Subgroups based on the dose of methylphenidate (low, medium, and high dose).

  2. Subgroups based on ages six to 12 years, and 12 to 18 years.

  3. Subgroups according to whether the formulation is a short‐acting or a controlled‐release form.

Sensitivity analysis

We will perform a sensitivity analysis to assess the impact of risk of bias on the overall result by adding or removing studies with a high risk of bias to the meta‐analysis. We will classify studies as "high risk of bias" if one or more of the following risk of bias assessment items is assessed as "high risk": random number generation, allocation concealment, blinding of participants and personnel, or outcome assessor.

We will also explore the impact of heterogeneity on the overall pooled effect estimate by adding or removing studies that are contributing to the heterogeneity. By eyeballing the forest plot, studies that are "outliers" and are potential sources for heterogeneity can be identified. We will remove the “outliers” one by one and assess the impact on the overall outcome.

We will perform a sensitivity analysis to explore the impact of missing data on the overall outcome by comparing the analyses with available outcome data with those following the ITT principle (see Dealing with missing data).