Content uploaded by Carlos Renato Moreira-Maia
Author content
All content in this area was uploaded by Carlos Renato Moreira-Maia on Jun 16, 2017
Content may be subject to copyright.
Content uploaded by Carlos Renato Moreira-Maia
Author content
All content in this area was uploaded by Carlos Renato Moreira-Maia on Nov 26, 2015
Content may be subject to copyright.
thebmj
BMJ
2015;351:h5203 doi: 10.1136/bmj.h5203
RESEARCH
1
open access
For numbered aliations see
end of article.
Correspondence to: O J Storebø
ojst@regionsjaelland.dk
Cite this as: BMJ ;:h
doi: 10.1136/bmj.h5203
Accepted: 13 September 2015
Methylphenidate for attention-decit/hyperactivity disorder in
children and adolescents: Cochrane systematic review with
meta-analyses and trial sequential analyses of randomised
clinical trials
Ole Jakob Storebø,1,2,3 Helle B Krogh,1,2 Erica Ramstad,1,2 Carlos R Moreira-Maia,4
MathildeHolmskov,1 Maria Skoog,5 Trine Danvad Nilausen,1 Frederik L Magnusson,1
Morris Zwi,6 Donna Gillies,7 Susanne Rosendal,8 Camilla Groth,9 Kirsten Buch Rasmussen,1
Dorothy Gauci,10 Richard Kirubakaran,11 Bente Forsbøl,2 Erik Simonsen,1,12 Christian Gluud5,13
ABSTRACT
STUDY QUESTION
Is methylphenidate benecial or harmful for the
treatment of attention-decit/hyperactivity disorder
(ADHD) in children and adolescents?
METHODS
Electronic databases were searched up to February
2015 for parallel and crossover randomised clinical
trials comparing methylphenidate with placebo or no
intervention in children and adolescents with ADHD.
Meta-analyses and trial sequential analyses (TSA)
were conducted. Quality was assessed using GRADE.
Teachers, parents, and observers rated ADHD
symptoms and general behaviour.
STUDY ANSWER AND LIMITATIONS
The analyses included 38 parallel group trials
(n=5111, median treatment duration 49 days) and 147
crossover trials (n=7134, 14 days). The average age
across all studies was 9.7 years. The analysis
suggested a benecial eect of methylphenidate on
teacher rated symptoms in 19 parallel group trials
(standardised mean dierence (SMD) −0.77,
n=1698), corresponding to a mean dierence of −9.6
points on the ADHD rating scale. There was no
evidence that methylphenidate was associated with
an increase in serious adverse events (risk ratio
0.98, nine trials, n=1532; TSA adjusted intervention
eect RR 0.91). Methylphenidate was associated with
an increased risk of non-serious adverse events
(1.29, 21 trials, n=3132; TSA adjusted RR 1.29).
Teacher rated general behaviour seemed to improve
with methylphenidate (SMD −0.87, ve trials, n=668)
A change of 7 points on the child health
questionnaire (CHQ) has been deemed a minimal
clinically relevant dierence. The change reported in
a meta-analysis of three trials corresponds to a mean
dierence of 8.0 points on the CHQ (range 0-100
points), which suggests that methylphenidate may
improve parent reported quality of life (SMD 0.61,
three trials, n=514). 96.8% of trials were considered
high risk of bias trials according to the Cochrane
guidelines. All outcomes were assessed very low
quality according to GRADE.
WHAT THIS STUDY ADDS
The results suggest that among children and
adolescents with a diagnosis of ADHD,
methylphenidate may improve teacher reported
symptoms of ADHD and general behaviour and
parent reported quality of life. However, given the
risk of bias in the included studies, and the very low
quality of outcomes, the magnitude of the eects is
uncertain. Methylphenidate is associated with an
increased risk of non-serious but not serious adverse
events.
FUNDING, COMPETING INTERESTS, DATA SHARING
Region Zealand Research Foundation and Copenhagen
Trial Unit. Competing interests are given in the full
paper on bmj.com. Full data are available in the
version of this review published in The Cochrane
Library.
Introduction
Attention-deficit/hyperactivity disorder (ADHD) is one
of the most commonly diagnosed and treated child-
hood psychiatric disorders,1 with a prevalence of 3.4%.2
It is increasingly seen as a developmental disorder,
which has high comorbidity with other psychiatric dis-
orders.3 Diagnosis is made through recognition of
excessive inattention, hyperactivity, and impulsivity in
children before 12 years of age, which impairs their
functioning or development.4 5
Methylphenidate has been used for the treatment of
ADHD for over 50 years and is now globally the most
common drug treatment for the disorder.6 7 Despite the
WHAT IS ALREADY KNOWN ON THIS TOPIC
Methylphenidate has been used for the treatment of attention-decit/hyperactivity
disorder for over 50 years and is globally the most common treatment for the
disorder
Despite the widespread use of methylphenidate, no comprehensive systematic
reviews of benets and harms have been done
WHAT THIS STUDY ADDS
The results of meta-analyses suggest that methylphenidate may improve teacher
reported ADHD symptoms, teacher reported general behaviour, and parent reported
quality of life among children and adolescents with a diagnosis of ADHD
The low quality of the underpinning evidence means that the magnitude of the
eects is uncertain
Within the short follow-up periods typical of the included trials, there is some
evidence that methylphenidate is associated with increased risk of non-serious
adverse events, such as sleep problems and decreased appetite, but no evidence
that it increases the risk of serious adverse events
doi: 10.1136/bmj.h5203
BMJ
2015;351:h5203 thebmj
RESEARCH
2
widespread use of methylphenidate no comprehensive
systematic review has been done of both benefits and
harms. Fifteen reviews of the eect of methylphenidate
on the symptoms of ADHD in children and adolescents
have been published.8-22 None of them were conducted
using Cochrane methodology and none prepublished a
peer reviewed protocol. Thirteen did not undertake sub-
group analyses on comorbidity influencing treatment
eects 8-16 18 19 21 22 nor did they control for the treatment
eect on subtypes of ADHD.8 10 11 15-19 21 22 Ten did not con-
sider dosage.9 10 12 13 15 16 18-20 22 Seven meta-analyses com-
bined outcome data across raters and observers8 9 10
15 16 17 20 and eight did not separate outcomes for inatten-
tion and hyperactivity or impulsivity.8 10-13 15 16 22 Nine
failed to present spontaneous adverse events10-16 18 22
and 14 did not report adverse events measured by rating
scales.8 10-22 Eleven reviews 8-14 16 17 21 22 did not follow
‘gold standard’ guidelines—that is, the Cochrane Hand-
book23 or the preferred reporting items for systematic
reviews and meta-analyses (PRISMA) guidelines.24 25
Risk of random errors, risk of bias, and trial quality
were not systematically assessed in 11 reviews.8-17 22
Language bias (exclusion of non-English publications)
was present in four reviews,10 14 19 22 and narrow or unre-
ported search strategies in four reviews8 11 16 20 may have
compromised data collection and, ultimately the
meta-analyses.
To avoid these flaws we conducted a systematic
review of the benefits and harms of methylphenidate in
children and adolescents with ADHD using the
Cochrane Handbook23 and PRISMA guidelines.24 25 This
article presents the results of a systematic review focus-
ing on the benefits and harms of methylphenidate in
randomised clinical trials.26 A second systematic review
will focus on harms in non-randomised studies.27
Methods
We used Cochrane methodology,23 following our pub-
lished protocol.28
Study selection
We included both parallel and crossover randomised
clinical trials comparing all types of methylphenidate
with placebo or no intervention in children and adoles-
cents from 3 to 18 years of age (two trials recruited a
small number of participants aged 19 to 21) with ADHD.
Trials were included irrespective of language, publica-
tion year, publication type, or publication status. After
the exclusion of duplicates and studies not meeting the
inclusion criteria, we obtained full text articles as per
protocol.28
Inclusion criteria
In one trial the diagnosis used for ADHD had to be
determined according to the Diagnostic and Statistical
Manual of Mental Disorders (third edition, third edition
revised, fourth edition, fourth edition revised, or fifth
edition),4 or according to International Classification of
Diseases ninth or 10th revisions (ICD-9 or ICD-10
codes).5 At least 75% of participants had to be aged less
than 19 years and the mean age of the study population
had to be less than 19 years. We included trials in which
participants had comorbidities; however, at least 75%
of the participants were required to have an intellectual
quotient in the normal range (IQ >70).
Search strategy and selection criteria
We searched the Cochrane Central Register of Con-
trolled Trials (The Cochrane Library 2015, Issue 2),
Medline, Embase, CINAHL, PsycINFO, ISI Conference
Proceedings Citation Index, Science and Conference
Proceedings Citation Index-Social Science & Human-
ities (Web of Science), ClincalTrials.gov, and WHO’s
International Clinical Trials Registry Platform up to
February 2015 using two dierent search strategies,
one for ecacy and one for adverse events. The com-
plete search strategy is available in the Cochrane
review.26 We screened reference lists of identified
reviews, meta-analyses, and a selection of included
trials for additional relevant articles. Furthermore, we
contacted pharmaceutical companies, including
Shire, Medice (represented in Denmark by HB
Pharma), Janssen-Cilag, and Novartis for published
and unpublished data. Emails were also sent to
experts in the discipline requesting data on unpub-
lished or ongoing studies.
Outcomes
The primary outcomes were symptoms of ADHD (inat-
tention, hyperactivity, and impulsivity), both short term
(six months or less) and long term (more than six
months), and serious adverse events. We defined the
latter as any event that led to death, was life threaten-
ing, required hospital admission or prolongation of
existing hospital stay, resulted in persistent or major
disability, and any important medical event that may
have jeopardised the participant’s life or required inter-
vention to prevent it. All other adverse events were con-
sidered non-serious.29
Secondary outcomes were non-serious adverse
events, general behaviour, and quality of life. We
assessed all non-serious adverse events, including
growth, cardiological, neurological, gastrointestinal,
and sleep events, and appetite. We rated general
behaviour at school and home using psychometric
validated instruments. Behaviour was classified
according to length of assessment as short (six
months or less) or long term (more than six months).
Quality of life was measured by psychometric vali-
dated instruments.
Patient involvement
No patients were involved in setting the research ques-
tion or the outcome measures, nor were they involved in
the design and implementation of the study. There are
no plans to involve patients in dissemination of the
results.
Data extraction and synthesis
Seventeen reviewers extracted the data independently
in the first of a two phase process.26 In the second
phase, a dierent reviewer checked the extracted data
thebmj
BMJ
2015;351:h5203 doi: 10.1136/bmj.h5203
RESEARCH
3
and disparities were resolved through discussion
between extractors, or consultation with the first
author (OJS) where consensus was not reached.26 For
additional questions and missing data we contacted
the authors of trials. Furthermore, we contacted all
authors of the crossover trials to obtain data for all
periods of the trial. We used Mendeley and Google
Drive online software programs for data exchange and
storage. Six authors entered the data into Review
Manager 5.3.26
Dichotomous data were summarised as risk ratios
with 95% confidence intervals. We used continuous
data to calculate the mean dierence between groups
(with 95% confidence intervals) if the same measure
was used in all trials, or we calculated the stan-
dardised mean dierence where dierent outcome
measures were used for the same construct in dierent
trials. To assess the minimal clinical relevant dier-
ence, we transformed the standardised mean dier-
ence into mean dierence using scales with published
minimal clinical relevant dierence. To our knowledge
the only published minimal clinical relevant dier-
ence on scales measuring our outcomes are 6.6 points
for the ADHD rating scale (ADHD symptoms, the scale
ranging from 0 to 72 points)30 and 7.0 points for the
child health questionnaire (quality of life, the scale
ranging from 0 to 100).31 Fixed eect and random
effects models were applied and discrepancies
between the results investigated.
Because crossover trials are more prone to bias
owing to carry-over eects, period eects, and errors
in unit of analysis,32 we conducted a subgroup analy-
sis. We analysed data from the first period in cross-
over trials with data from parallel trials. Our original
intent was to adjust for the eect of the unit of analy-
sis error in crossover trials by conducting a covariate
analysis; however, the data were insucient. We
tested for the possibility of a carry-over eect and
period eect and found similar treatment eects in
the parallel group trials plus the first period of the
crossover trials compared to the second period of the
crossover trials. We found no statistically significant
subgroup dierences between the two groups but
high heterogeneity in the subgroup analyses. We
therefore present the analyses separately.
The treatment eect was defined as an improvement
in the symptoms of ADHD, general behaviour, or quality
of life. Teachers, observers, or parents rated symptoms
and general behaviour. We considered these data as dif-
ferent outcomes and teacher rated measures as the pri-
mary outcome because symptoms of ADHD are more
readily detectable in the school setting.33
We used the teacher rated symptoms of ADHD from
parallel group or first period of crossover trials as our
primary analysis to test the robustness of this estimate
with several subgroup analyses:
• Type of scale.
• Dose of methylphenidate (low dose: ≤20 mg/day or
≤0.6 mg/kg/day compared to moderate or high dose:
>20 mg/day or >0.6 mg/kg/day).
• Design (parallel group trials compared to first phase
and end of trial of crossover trials).
• Drug status before randomisation—“drug naïve” (if
>80% of participants were naive) compared to
“previous use of drug” (>80% of participants used
the drug previously).
• Risk of bias (low risk of bias trials compared to high
risk of bias trials).
• Age—comparing trials of participants aged 2 to 6
years to those of participants aged 7 to 11 years and to
those aged 12 to 18 years.
• Sex—boys compared to girls.
• Comorbidity—participants with comorbid disorders
compared to participants without comorbid disor-
ders.
• Type of ADHD—predominantly participants with
inattentive type compared to participants with com-
bined type.
• Types of raters—parents compared to observers com-
pared to teachers.
• Trials with cohort selection bias of all participants
compared to trials without cohort selection bias of all
participants.
• Trials using fixed doses compared to trials using ini-
tial titration.
Sensitivity analyses
We conducted sensitivity analyses to ascertain whether
our findings were sensitive to decisions during the
review (for example, our assessment of clinical hetero-
geneity), the combination of both change scores and
end of trial scores in one meta-analysis, and inclusion
of studies with participants of IQ less than 70 or aged
more than 18 years.
Consequently we undertook sensitivity analyses
excluding trials using change scores,34-38 trials with
participants of IQ less than 70,39-42 and trials including
some participants aged more than 18 years.43 44
No valid method exists for combining the results of
trials rated as high risk and low risk of bias.23 We per-
formed sensitivity analyses grouping together the trials
with similar classifications of bias and investigated the
impact on intervention eects.
Trial sequential analysis
A meta-analysis should include a calculation of a
required information size at least as large as the sam-
ple size of an adequately powered single trial to reduce
the risks of random errors, taking into consideration
the heterogeneity of the meta-analysis.45 46 Trial
sequential analysis is a program that calculates the
required information size for a meta-analysis, provid-
ing adjusted statistical thresholds for benefits, harms,
or futility before the required information size is
reached.46-49 Trial sequential analysis can thereby con-
trol the risks of type I and type II errors due to sparse
data and repetitive testing of accumulating data.46-49
Meta-analyses not reaching the required information
size are analysed with trial sequential monitoring
boundaries analogous to interim monitoring boundar-
ies in a single trial, requiring more stringent Z values to
doi: 10.1136/bmj.h5203
BMJ
2015;351:h5203 thebmj
RESEARCH
4
declare benefits or harms.46 If a trial sequential analy-
sis results in insignificant findings before the required
information size has been reached (no Z curve crossing
of the trial sequential monitoring boundaries), the con-
clusion should be that more trials are needed to either
accept or reject the intervention eect used for calcu-
lating the required sample size. If the cumulated Z
curve enters the futility area, the anticipated interven-
tion eect can be rejected.
For the trial sequential analysis calculations of
binary outcomes we included trials with zero events by
substituting zero with 0.5.46 50
For the outcomes of total number of serious adverse
events and total number of non-serious adverse events,
we calculated the a priori diversity adjusted required
information size (that is, the number of participants
required to detect or reject a specific intervention eect
in the meta-analysis) on the following assump-
tions46-49 51: the proportion of participants in the control
group with adverse events, a relative risk reduction or
increase of 20% (25% for serious adverse events), a type
I error of 5%, a type II error of 20%, and the observed
diversity of the meta-analysis.
Quality of evidence
For each included trial, data extractors independently
evaluated all risk of bias domains (listed below),
resolving any disagreements by discussion. We
assigned each bias domain to one of three categories:
low risk of bias, uncertain risk of bias, or high risk of
bias, according to the Cochrane guidelines.23 Owing to
the risk of overestimation of beneficial intervention
eects and underestimation of harmful intervention
eects in randomised clinical trials with unclear or
high risk of bias,52-58 we assessed the influence of the
risk of bias on our results (see subgroup analyses). We
used the following domains to assess the risk of bias in
the included trials23 59: generation of allocation
sequence, allocation of concealment, blinding of par-
ticipants and sta, blinding of outcome assessors,
incomplete outcome data, selective outcome report-
ing, and vested interest (trial was funded by parties
that might have had a conflict of interest (for example,
a manufacturer of methylphenidate) or there were
potential conflicts of interests among authors because
they had been working for companies producing or
selling methylphenidate, or both). We considered tri-
als with low risk of bias in all domains to be at low risk
of bias and trials with one or more unclear or inade-
quate component to be at high risk of bias. For 32%
(59/185) of the included trials we noted a specific type
of bias occurring before randomisation. Non-respond-
ers to methylphenidate, responders to placebo, or par-
ticipants who had adverse events due to the drug were
excluded as a consequence of exclusion criteria or
after a titration phase. Such trials have limited exter-
nal validity and, to identify whether this cohort selec-
tion bias had an eect on estimates of eects, we did
subgroup analyses (see above).
We assessed and graded the evidence according to
the grading of recommendations assessment, develop-
ment, and evaluation (GRADE) for high risk of bias,
imprecision, indirectness, heterogeneity, and publica-
tion bias.60 The analyses were conducted with Review
Manager 5.3 (Review Manager 2014) and the trial
sequential analysis program.46 50
61
Results
We identified 14 431 records. After removal of dupli-
cates and irrelevant references we retrieved 1461 publi-
cations in full text for assessment of eligibility. Overall
we excluded 691 publications for not meeting the inclu-
sion criteria and eight because they were classified as
either awaiting classification or as ongoing studies. In
this review we included the remaining 761 publications
(69 in non-English languages) describing 185 ran-
domised clinical trials and 243 non-randomised stud-
ies (fig 1 ).62 When information to assess eligibility or
bias was missing or data were missing or unclear we
contacted the authors of the studies. Authors of 161 tri-
als were contacted up to two times and answers were
received for 92 trials.
Of the 185 randomised clinical trials, 38 were parallel
group (n=5111 participants) and 147 were crossover
(n=7134 participants) trials, with a total of 12 245 partic-
ipants. Participants of both sexes were included. Partic-
ipants were aged between 3 and 18 years, but two trials
included a small number of participants aged 19 to 21
years. The mean age was 9.7 years. The majority of these
trials were conducted in high income countries. The
median duration of treatment in the parallel group tri-
als was 49 days (range 1-425 days, mean 75 days) and in
the crossover trials was 14 days (1-56 days, mean 16
days). No parallel group trials and only six crossover
trials (3.2%, 285 participants) seemed to have low risk of
bias in all domains, and accordingly 179 (96.8%) trials
were considered high risk of bias trials. By using
GRADE, we rated the quality of evidence as being very
low for all outcomes.
Symptoms of ADHD
Data on symptoms of ADHD in our analyses were
available from 25 parallel group trials and 74 cross-
over trials (53.0%) reporting on such symptoms. The
results showed an effect of methylphenidate on
teacher rated symptoms in the parallel group trials
(standardised mean dierence −0.77, 95% confidence
interval −0.90 to −0.64, 19 trials, 1698 participants, fig
2 ). This corresponds to a mean dierence of −9.6
points (95% confidence interval −13.75 to −6.38) on
the ADHD rating scale, which was larger than the
minimal clinical relevant dierence of −6.6 points. No
publication bias was detected (Egger’s test P=0.81).
All the trials had high risk of bias, primarily as a result
of vested interest, lack of blinding of participants,
lack of outcome assessor blinding, selective outcome
reporting, or selection bias. Some but not all bias
risks were present in most studies. The result of the
GRADE assessment was “very low quality” owing to
high risks of bias and heterogeneity. The intervention
eect was significantly influenced by choice of scale
(test for subgroup dierences, P=0.006). Long term
thebmj
BMJ
2015;351:h5203 doi: 10.1136/bmj.h5203
RESEARCH
5
trials had a smaller eect (standardised mean dier-
ence −0.47, 95% confidence interval −0.72 to −0.22,
one trial, 253 participants) compared to short term
trials (−0.81, −0.94 to −0.68, 18 trials, 1445 partici-
pants; test for subgroup dierence, P=0.02). Trials
including participants with previous use of drugs
before randomisation resulted in a larger eect (−1.06,
−1.33 to −0.79, two trials, 286 participants) compared
to trials including participants naive to the drugs
(−0.63, −0.94 to −0.31, four trials, 431 participants; test
for subgroup dierence, P=0.04). However, no signif-
icant influences on the observed treatment eects
were found according to dose, trial design, cohort
selection bias (trials with optimal titration, exclusion
of non-responders, placebo responders, or partici-
pants with methylphenidate adverse events before
randomisation) and trials with initial titration or fixed
doses. One trial included in the meta-analysis
reported change from baselines scores37 rather than
endpoint data. Removing this trial did not change the
estimate noticeably.
The end of last period crossover trials also showed
a significant treatment eect, with a standardised
mean dierence of −0.93 (95% confidence interval
−1.06 to −0.80, 59 trials, 5145 participants) that was
not significantly influenced by risk of bias (test for
subgroup dierence, P=0.09). The benefit, however,
was significantly greater with higher doses of methyl-
phenidate (−0.98, −1.13 to −0.84, 36 trials, 3413 partic-
ipants) compared to low doses (−0.73, −0.89 to −0 . 57,
42 trials, 3408 participants; test for subgroup dier-
ence, P=0.02). Three trials included participants with
an IQ <70,40 41 42 but removing these did not change the
estimate noticeably.
Additional subgroup analyses
Additional subgroup analyses on symptoms of ADHD in
parallel group trials and first period crossover trials
showed that neither age nor comorbidity significantly
influenced the intervention eect. The intervention
eect was significantly influenced by subtype of ADHD,
with a higher intervention eect for the inattentive sub-
type (standardised mean dierence −1.31, 95% confi-
dence interval −1.61 to −1.01, one trial, 204 participants)
compared to the combined type (0.65, −1.30 to 2.60, two
trials, 559 participants; test for subgroup dierence,
P=0.05), but this dierence was based on sparse data.
We found no evidence of a carry-over eect in the cross-
over trials in a subgroup analysis between the first
period and the second period data from four crossover
trials. First period (−0.64, −0.85 to −0.44, four trials, 372
participants) and second period (−0.91, −1.18 to −0.65,
four trials, 372 participants; test for subgroup dier-
ence, P=0.1).
Records screened aer duplicates removed (n=9271)
Full text articles assessed for eligibility (n=1460)
Full text articles eligible (n=761)
185 included randomised studies (from 449 reports)
Parallel group trials (n=38); crossover trials (n=147) (parallel and crossover with 1st period data (n=42))
Fourth database search
(n=1460):
CINAHL (n=38)
Cochrane Library
(n=26)
Embase (n=627)
Medline (n=300)
PsycINFO (n=166)
ISI CPCI (n=4)
International Clinical
Trials Registry
Platform (n=265)
Clinical trials (n=34)
Additional records
identied through
other sources
(n=368):
Reference lists of
articles (n=70)
298 reviews (n=73)
Authors (n=39)
Pharmaceutical
companies (n=186)
Third database search
(n=1274):
CINAHL (n=133)
Cochrane Library
(n=184)
Embase (n=650)
Medline (n=182)
PsycINFO (n=124)
ISI CPCI (n=1)
Second database
search (n=1080):
CINAHL (n=64)
Cochrane Library
(n=230)
Embase (n=465)
Medline (n=212)
PsycINFO (n=100)
ISI CPCI (n=9)
First database search
(n=10 249):
CINAHL (n=772)
Cochrane Library
(n=925)
Embase (n=3964)
Medline (n=2787)
PsycINFO (n=1490)
ISI Conference
Proceedings Citation
Index (CPCI) (n=311)
Records excluded (n=7811)
243 non-randomised studies eligible for another review (from 312 reports)
Excluded (n=691):
Full text articles excluded (n=691):
Not randomised controlled trial (n=367)
No acceptable ADHD diagnosis (n=269)
No methylphenidate treatment (n=15)
Age >18 years (n=5)
IQ <70 (n=40)
No assessment of ADHD symptoms, general behaviour, quality of life, or adverse events (n=88)
Polypharmacy (n=31)
Other (n=14)
7 studies are ongoing or awaiting classication (from 8 reports)
Fig | Flow of studies through review
doi: 10.1136/bmj.h5203
BMJ
2015;351:h5203 thebmj
RESEARCH
6
Our analyses investigating the dierence between
raters showed no significant dierences: teacher rated
(−0.78, −0.93 to −0.63, 19 trials, 1689 participants),
observer rated (−0.61, −0.87 to −0.35, nine trials, 1826
participants), and parent rated (−0.65, −0.81 to −0.50, 21
trials, 2179 participants), test for subgroup dierence,
P=0.37.
Serious adverse events
We could only include nine parallel group trials (4.9%)
reporting serious adverse events. For these trials meth-
ylphenidate was not associated with an overall
increase in total number of serious adverse events
(risk ratio 0.98, 95% confidence interval 0.44 to 2.22,
1532 participants, fig 3). All the trials had high risk of
bias owing to vested interest, incomplete outcome
data, lack of blinding, and selective outcome report-
ing. The GRADE assessment was very low quality as a
result of high risk of bias and imprecision. Eight cross-
over trials reported serious adverse events at the end
of the last period. These trials did not seem to dier
between intervention groups (risk ratio 1.62, 95% con-
fidence interval 0.34 to 7.71, 1721 participants; I2=0%,
P=0.65).
We conducted trial sequential analysis on the “total
serious adverse events” outcome, involving nine par-
allel group trials. We had planned to use a relative risk
reduction of 20% but owing to too large a distance
between the accrued information and the required
information the program rejected 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 a con-
stant of 0.5 for zero. We calculated the diversity
adjusted required information size (DARIS) on the
basis of serious adverse events in the control group of
2%; a relative risk reduction or increase in the experi-
mental 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
Arnold 2004
Biederman 2003
Brown 1985
Butter 1983
Childress 2009
Findling 2006
Findling 2008
Firestone 1981
Ialongo 1994
Jensen 1999 (MTA)
Kollins 2006 (PATS)
Lehmkuhl 2002
Moshe 2012
Palumbo 2008
Pliszka 2000
Schachar 1997
Taylor 1987
Van der Meere 1999
Wolraich 2001
Total (95% CI)
Test for heterogeneity:
τ2=0.03, χ2=28.78, df=18, P=0.05, I2=37%
Test for overall eect: z=11.23, P<0.001
-0.85 (-1.33 to -0.38)
-1.07 (-1.43 to -0.71)
-0.15 (-1.03 to 0.73)
-0.74 (-1.65 to 0.17)
-1.02 (-1.40 to -0.64)
-1.07 (-1.45 to -0.69)
-0.71 (-1.01 to -0.41)
-0.57 (-1.30 to 0.16)
-1.02 (-1.86 to -0.17)
-0.47 (-0.72 to -0.22)
-0.33 (-0.82 to 0.17)
-1.06 (-1.52 to -0.61)
-0.58 (-0.96 to -0.21)
-0.28 (-0.79 to 0.23)
-0.89 (-1.56 to -0.22)
-1.13 (-1.65 to -0.60)
-0.57 (-1.03 to -0.12)
-0.73 (-1.32 to -0.15)
-1.05 (-1.44 to -0.67)
-0.77 (-0.90 to -0.64)
5.3
7.2
2.0
1.9
6.8
6.9
8.7
2.8
2.2
10.0
5.0
5.5
7.0
4.7
3.2
4.6
5.6
3.9
6.8
100.0
-2 -1 012
Study
Favours
methylphenidate
Favours
control
Standardised mean
dierence, IV random
(95% CI)
Standardised mean
dierence, IV random
(95% CI)
Weight
(%)
0.7
16.3
15.1
30.5
16.4
4.3
18.3
8.9
7.5
0.8
1.1
0.9
58.0
-5.1
0.8
0.9
0.5
73.6
5.7
Mean
0.7
12.1
4.6
17.3
13.4
3.2
17.4
4.9
7.4
0.7
0.8
0.7
10.3
6.8
0.62
0.7
1.04
12.7
3.8
SD
35
63
10
10
57
120
94
18
13
134
32
43
57
29
20
37
39
24
81
916
No
Methylphenidate
1.4
31.3
15.7
42.7
30.0
7.7
31.6
11.8
15.3
1.1
1.4
1.6
64.7
-3.2
1.5
1.7
1.2
83.1
9.9
Mean
0.9
15.4
2.9
14.2
13.1
3.1
20.1
4.8
7.3
0.8
0.8
0.6
12.5
6.4
0.9
0.7
1.3
12.7
4.1
SD
40
71
10
10
63
39
88
13
12
119
32
42
57
30
18
29
39
24
46
782
No
Control
Random sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding of participants and personnel (performance bias)
Blinding of outcome assessment (detection bias)
Incomplete outcome data (attrition bias)
Selective reporting (reporting bias)
Vested interest bias
Fig | Teacher rated symptoms of attention-decit/hyperactivity disorder in parallel group trials.
thebmj
BMJ
2015;351:h5203 doi: 10.1136/bmj.h5203
RESEARCH
7
not cross the conventional or trial sequential monitor-
ing boundaries for benefit, harm, or futility (fig 4). As
only less than 7% of the DARIS was accured, risks of
random type II error cannot be excluded. The trial
sequential analysis adjusted intervention eect is risk
ratio 0.91% (95% confidence interval 0.02 to 33.2).
Therefore, the total sample size in the meta-analysis of
serious adverse events for 1532 participants was con-
siderably under powered to identify a dierence in
serious adverse events.
Non-serious adverse events
We could only include 26 parallel group trials (14.0%)
reporting non-serious adverse events. Methylpheni-
date was associated with an overall risk of total num-
ber of non-serious adverse events by 29% (risk ratio
1.29, 95% confidence interval 1.10 to 1.51, 21 trials, 3132
participants, fig 5). All trials had high risk of bias owing
to lack of blinding of participants, lack of outcome
assessor blinding, vested interest, selective outcome
reporting, and incomplete outcome data. The GRADE
assessment was very low quality as a result of high risk
of bias and heterogeneity. Heterogeneity was substan-
tial between trials (τ²=0.08, χ²=61.94, df=12, P<0.001);
I²=81%), which did not seem to be related to dose (test
for subgroup dierence between low dose and high
dose methylphenidate, P=0.57). The adverse events
reported over all trials included neurological, diges-
tive, urinary, circulatory, respiratory, reproductive,
skeletal, muscular, and immunological adverse events
as well as physical measures such as dierence in
height, weight, body mass index, and vital signs. The
most common non-serious adverse events were
decreased appetite (risk ratio 3.66, 95% confidence
interval 2.56 to 5.23, 16 trials, 2962 participants; I2=18%)
and sleep problems (1.60, 1.15 to 2.23, 13 trials, 2416 par-
ticipants; I2=0%).
Sixty seven crossover trials reported non-serious
adverse events at the end of the second period. For
these trials methylphenidate was associated with an
Carlson 2007
Childress 2009
Coghill 2013
Findling 2010
Jacobi-Polishook 2009
Lehmkuhl 2002
Palumbo 2008
Riggs 2011
Wolraich 2001
Total (95% CI)
Test for heterogeneity:
τ2=0.00, χ2=3.14, df=6, P=0.79, I2=0%
Test for overall eect: z=0.04, P=0.97
3.10 (0.13 to 75.14)
1.05 (0.04 to 25.43)
0.67 (0.11 to 3.95)
3.50 (0.18 to 66.86)
Not estimable
2.93 (0.12 to 70.00)
3.10 (0.13 to 73.14)
0.58 (0.17 to 1.92)
Not estimable
0.98 (0.44 to 2.22)
6.5
6.5
21.1
7.6
6.6
6.6
45.2
100.0
-2 -1 012
Study
Favours
methylphenidate
Favours
control
Risk ratio, IV
random (95% CI)
Risk ratio, IV
random (95% CI)
Weight
(%)
1
1
2
3
0
1
1
4
0
13
Events
87
182
110
145
12
43
29
151
160
919
No
Methylphenidate
0
0
3
0
0
0
0
7
0
10
Events
90
63
111
72
12
42
30
152
41
613
No
Control
Random sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding of participants and personnel (performance bias)
Blinding of outcome assessment (detection bias)
Incomplete outcome data (attrition bias)
Selective reporting (reporting bias)
Vested interest bias
Fig | Serious adverse events in parallel group trials. Green=low risk of bias; yellow=uncertain risk of bias; red=high risk of bias. See Cochrane review for
details of references
No of patients (linear scale)
Cumulative Z score
Favours
methylphenidate
Favours
control
DARIS Pc 2%; RRR 25%; α 5%; β 20%; diversity 0% = 21 593
-10
-6
-2
2
6
10
-8
-4
0
4
8
Z curve
1532
Fig | Trial sequential analysis on total number of serious adverse events. DARIS=diversity
adjusted required information size; RRR=relative risk reduction
doi: 10.1136/bmj.h5203
BMJ
2015;351:h5203 thebmj
RESEARCH
8
overall increase in total number of non-serious adverse
events (risk ratio 1.33, 95% confidence interval 1.11 to
1.58, 21 trials, 2072 participants). The most common
specific non-serious adverse events were decreased
appetite (3.04, 2.35 to 3.94, 35 trials, 3862 participants,
I2=40%) and sleep problems (1.57, 1.20 to 2.06, 31 trials,
3270 participants, I2=47%).
We conducted trial sequential analysis on the total
number of non-serious adverse events outcome
including 21 parallel groups or end of first period
crossover trials (fig 6). The diversity adjusted
required information size was calculated based on a
proportion of adverse events in the control group of
47%, a relative risk reduction or increase of 20% in
the experimental group, a type I error of 5%, a type II
error of 20% (80% power), and a diversity (D2) of
79%. The diversity adjusted required information size
Biederman 2003
Carlson 2007
Childress 2009
Coghill 2013
Findling 2006
Findling 2008
Findling 2010
Greenhill 2002
Greenhill 2006
Jacobi-Polishook 2009
Lehmkuhl 2002
Newcorn 2008
Palumbo 2008
Pliszka 2000
Riggs 2011
TSSG* 2002
Tucker 2009
Wigal 2004
Wilens 2006
Wolraich 2001
Zeni 2009
Total (95% CI)
Test for heterogeneity:
τ2=0.08, χ2=72.18, df=19, P<0.001, I2=74%
Test for overall eect: z=3.13, P=0.002
1.37 (0.38 to 4.87)
0.89 (0.40 to 1.97)
1.12 (0.88 to 1.42)
1.13 (0.92 to 1.40)
0.65 (0.53 to 0.79)
1.20 (0.96 to 1.51)
1.35 (1.08 to 1.69)
1.36 (1.06 to 1.75)
1.31 (0.98 to 1.76)
Not estimable
1.76 (0.92 to 3.35)
1.23 (0.98 to 1.55)
1.47 (0.86 to 2.50)
1.70 (0.72 to 4.01)
1.10 (0.85 to 1.42)
1.78 (0.35 to 9.12)
3.31 (2.12 to 5.16)
1.78 (1.30 to 2.43)
1.21 (0.59 to 2.46)
1.61 (0.89 to 2.90)
0.50 (0.11 to 2.35)
1.29 (1.10 to 1.51)
1.3
2.8
7.2
7.5
7.5
7.3
7.4
7.1
6.7
3.6
7.3
4.4
2.5
7.1
0.9
5.2
6.5
3.2
4.0
0.9
100.0
0.2 0.5 12 5
Study
Favours
methylphenidate
Favours
control
Risk ratio, IV
random (95% CI)
Risk ratio, IV
random (95% CI)
Weight
(%)
5
5
116
72
71
63
112
80
40
0
18
146
17
10
70
4
47
80
14
56
2
1028
Events
65
9
182
111
133
91
145
155
53
12
43
219
29
20
151
37
53
90
87
160
16
1861
No
Methylphenidate
4
5
36
63
38
49
40
61
27
0
10
40
12
5
64
2
15
21
12
10
4
518
Events
71
8
63
110
46
85
70
161
47
12
42
74
30
17
152
33
56
42
90
46
16
1271
No
Control
Random sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding of participants and personnel (performance bias)
Blinding of outcome assessment (detection bias)
Incomplete outcome data (attrition bias)
Selective reporting (reporting bias)
Vested interest bias
*Tourette's-Syndrome-Study-Group
Fig | Non-serious adverse events in parallel group trials. Green=low risk of bias; yellow=uncertain risk of bias; red=high risk of bias. See Cochrane
review for details of references
No of patients (linear scale)
Cumulative Z score
Favours
methylphenidate
Favours
control
3132
DARIS Pc 47%; RRR 20%; α 5%; β 20%; diversity 79% = 4133
-10
-6
-2
2
6
10
-8
-4
0
4
8
Z curve
Fig | Trial sequential analysis on total number of non-serious adverse events.
DARIS=diversity adjusted required information size; RRR=relative risk reduction
thebmj
BMJ
2015;351:h5203 doi: 10.1136/bmj.h5203
RESEARCH
9
was 4133 participants. The cumulative Z curve (red
line) crossed the trial sequential monitoring bound-
ary for harm (blue inward sloping line) after the
seventh trial, then regressed, and crossed the bound-
ary again after the 17th trial. Thereafter it never
regressed. The trial sequential analysis adjusted risk
ratio was 1.29 (95% confidence interval 1.06 to 1.56).
Accordingly, we can exclude random error as a cause
of the finding.
General behaviour
We could only include seven parallel group trials and
19 crossover trials (13.0%) reporting general
behaviour. The standardised mean difference of
teacher rated general behaviour in parallel group tri-
als was −0.87 (95% confidence interval −1.04 to −0.71,
five trials, 668 participants, fig 7). It was not possible
to transform this estimate to a widely used validated
scale. All the trials were high risk of bias owing to
uncertainty about the method used for sequence gen-
eration and allocation of concealment, vested inter-
est, incomplete outcome data, and selective outcome
reporting. The GRADE assessment was very low qual-
ity owing to a high risk of bias and indirectness. Nei-
ther the type of scales nor the dose significantly
influenced the intervention eect. The crossover trial
analysis also showed a beneficial treatment eect
(standardised mean dierence −0.69, 95% confidence
interval −0.78 to −0.60, 16 trials, 2014 participants).
The intervention eect was not influenced by dose of
methylphenidate. All crossover trials were considered
high risk of bias.
Quality of life
Only three parallel group trials (1.6%) reported qual-
ity of life. There was a small beneficial eect on qual-
ity of life (standardised mean dierence 0.61, 95%
confidence interval 0.42 to 0.80, three trials, 514 par-
ticipants, fig 8 ), which corresponds on the child
health questionnaire scale to a mean dierence of 8.0
points (95% confidence interval 5.49 to 10.46), which
is larger than the minimal clinical relevant dierence
of 7.0.31 However, the estimate relies on only three tri-
als and all three had high risk of bias, primarily due to
lack of blinding of participants, selective outcome
reporting, and vested interests. The GRADE assess-
ment was very low quality owing to high risk of bias
and indirectness.
Discussion
In this meta-analysis and trial sequential analysis we
found that methylphenidate reduces the symptoms of
ADHD in children and adolescents. We also observed a
possible small beneficial eect on quality of life and
general behaviour. The apparent eects of methylphe-
nidate on both the ADHD rating scale and the child
health questionnaire should be considered clinically
relevant based on our predefined minimal relevant dif-
ferences. However, our present results are based on tri-
als that by GRADE are considered very low quality and
may be prone to bias.
The use of methylphenidate is associated with a rel-
atively high risk of non-serious adverse events in gen-
eral. Just over a quarter of children and adolescents
seemed to experience non-serious adverse events
Findling 2006
Greenhill 2002
Ialongo 1994
Van der Meere 1999
Wolraich 2001
Total (95% CI)
Test for heterogeneity:
τ2=0.00, χ2=0.40, df=4, P=0.98, I2=0%
Test for overall eect: z=10.39, P<0.001
-0.84 (-1.21 to -0.46)
-0.91 (-1.14 to -0.68)
-0.67 (-1.48 to 0.14)
-0.84 (-1.46 to -0.21)
-0.88 (-1.25 to -0.50)
-0.87 (-1.04 to -0.71)
19.6
50.3
4.1
6.9
19.0
100.0
-2 -1 012
Study
Favours
methylphenidate
Favours
control
Standardised mean
dierence, IV random
(95% CI)
Standardised mean
dierence, IV random
(95% CI)
Weight
(%)
2.3
4.9
3.1
37.8
1.8
Mean
2.74
4.66
2.78
5.85
2.99
SD
120
155
13
22
81
391
No
Methylphenidate
4.6
10.3
5.8
42.8
5.2
Mean
2.75
6.92
5.00
5.85
5.12
SD
39
159
12
21
46
277
No
Control
Random sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding of participants and personnel (performance bias)
Blinding of outcome assessment (detection bias)
Incomplete outcome data (attrition bias)
Selective reporting (reporting bias)
Vested interest bias
Fig | Teacher rated general behaviour in parallel group trials. Green=low risk of bias; yellow=uncertain risk of bias; red=high risk of bias. See Cochrane
review for details of references
doi: 10.1136/bmj.h5203
BMJ
2015;351:h5203 thebmj
RESEARCH
10
after methylphenidate treatment. Although methyl-
phenidate has been reported to cause rare but serious
cardiac events as well as sudden cardiac death,63 it
did not appear to cause an increase in serious adverse
events in the short term in our meta-analysis. How-
ever, the data on serious adverse events was under-
powered as shown by the trial sequential analysis
adjusted confidence interval and there were no data
available from randomised trials on the long term
incidence of such events. Our findings should be seen
in the context of the low quality of the included trials
owing to avoidable methodological limitations—for
example, inadequate sequence generation and allo-
cation concealment, lack of blinding, selection bias,
incomplete outcome data, reporting bias, and possi-
ble bias caused by vested interest.57 64 Only six of 185
trials with a total of 183 participants seemed to be at
low risk of bias in all domains. Even the trials origi-
nally considered at low risk of bias may in fact be tri-
als with high risk of bias owing to lack of blinding
despite the use of placebo, as methylphenidate gives
rise to several easily recognisable adverse events that
can lead to loss of blinding and influence the rating of
symptoms and adverse events. We found no trials
employing nocebo tablets (“active placebo”) in the
controls, thus the extent of this bias cannot be
assessed. Furthermore, heterogeneity might have
influenced the results.
Strength and limitations of this study
This systematic review has several strengths. We devel-
oped a protocol for this review according to instruc-
tions provided in the Cochrane Handbook for Systematic
Reviews of Interventions.23 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 phar-
maceutical companies manufacturing methylpheni-
date, 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
14.4.1 CHQ
Newcorn 2008
Subtotal (95% CI)
Test for heterogeneity: Not applicable
Test for overall eect: z=3.70, P<0.001
14.4.2 CGAS
Szobot 2004
Subtotal (95% CI)
Test for heterogeneity: Not applicable
Test for overall eect: z=10.39, P<0.001
14.4.3 CHIP-CE:PRF
Coghill 2013
Subtotal (95% CI)
Test for heterogeneity: Not applicable
Test for overall eect: z=10.39, P<0.001
Total (95% CI)
Test for heterogeneity:
τ2=0.00, χ2=0.52, df=2, P=0.77, I2=0%
Test for overall eect: z=6.30, P<0.001
Test for subgroup dierences:
χ2=0.52, df=2, P=0.77, I2=0%
0.54 (0.25 to 0.83)
0.54 (0.25 to 0.83)
0.79 (0.10 to 1.47)
0.79 (0.10 to 1.47)
0.64 (0.37 to 0.91)
0.64 (0.37 to 0.91)
0.61 (0.42 to 0.80)
43.5
43.5
7.7
7.7
48.8
48.8
100.0
-2 -1 012
Study
Favours
control
Favours
methylphenidate
Standardised mean
difference, IV random
(95% CI)
Standardised mean
difference, IV random
(95% CI)
Weight
(%)
7.8
69.1
7.1
Mean
12.7
11.3
11.6
SD
193
193
19
19
111
111
323
No
Methylphenidate
1.0
59.7
-0.2
Mean
12.0
12.1
11.2
SD
64
64
17
17
110
110
191
No
Control
Random sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding of participants and personnel (performance bias)
Blinding of outcome assessment (detection bias)
Incomplete outcome data (attrition bias)
Selective reporting (reporting bias)
Vested interest bias
Fig | Quality of life in parallel group trials. Green=low risk of bias; yellow=uncertain risk of bias; red=high risk of bias. See Cochrane review for details of
references
thebmj
BMJ
2015;351:h5203 doi: 10.1136/bmj.h5203
RESEARCH
11
resolved by discussion with team members. We
assessed risk of bias in all trials according to recom-
mendations provided in the Cochrane Handbook for
Systematic Reviews of Interventions.23 In addition, this
review meta-analysed adverse events for the same
intervention and used trial sequential analysis meth-
odology to control the risks of false positive results in
meta-analysis owing to sparse data and repetitive anal-
yses of data.46 47 50 61
It may be considered a drawback that we did not
search the databases of the US Food and Drug Admin-
istration and European Medicines Agency for unpub-
lished trials. 65The median duration of drug treatment
was less than two months and few trials had a dura-
tion of more than six months. Therefore there is little
that can be concluded about the benefits and harms
of methylphenidate use for longer than six months.
When comparing short term trials (six months or less)
with long term trials (more than six months), we
found that the treatment eect for teacher rated
symptoms of ADHD decreased over time. This was not
the case for independent assessor rated and parent
rated symptoms of ADHD, where there were no signif-
icant dierences between short term and long term
duration of trials. We could identify no trials that
examined the eect of more extended use of treat-
ment on young people’s general behaviour. Overall,
there is a lack of evidence about the long term eects
of methylphenidate in children and adolescents with
ADHD.
Some researchers have argued that parents’ evalua-
tions of the symptoms of ADHD may not be as reliable
as those of other raters such as teachers of preschool
children66 or college students.67 One study suggested
that there was inconsistency in ratings between par-
ents.68 In the Multimodal ADHD Treatment (MTA) trial,
information provided by parents was not always
thought to be strong.69 However, we found no dier-
ences between dierent raters.
Agreements and disagreements with other studies
or reviews
During the past 15 years, several reviews investigat-
ing the efficacy of methylphenidate for ADHD (with
or without meta-analyses) have been published. Fif-
teen reviews on the efficacy of methylphenidate
treatment for children and adolescents with ADHD
have pooled the results of ADHD rating scales.8-22
These reviews have several shortcomings, as
described in our introduction. Most did not assess
the risk of bias of the included studies or adverse
events. Moreover, none of these reviews considered
the risks of random errors. Therefore, the true esti-
mate of the treatment is not known and information
about adverse events from several randomised clini-
cal trials is missing. All of these reviews and
meta-analyses reported a large effect of methylpheni-
date. The meta-analysis in our review showed an
effect of methylphenidate on some outcomes but
because we identified that almost all of the trials
have a high risk of bias, we do not know the true
intervention effect. We also found a large risk of out-
come reporting bias in the outcomes for both serious
and non-serious adverse events
A recent Cochrane systematic review evaluated the
effects of methylphenidate in adults with ADHD.70
The effect sizes across the different assessments of
symptoms were similar to those found in our analy-
ses (standardised mean difference 0.60). The authors
noted that data on adverse events were limited by the
short duration of the included trials.70 Despite the
similar effects of methylphenidate on symptoms
observed in our reviews, we have judged the quality
of evidence in our own review, as well as that in
Epstein 2014,71 to be lower than that of Epstein and
colleagues’.
Implications for practice
The results of meta-analyses suggest that methylpheni-
date may improve teacher reported ADHD symptoms,
teacher reported general behaviour, and parent
reported quality of life among children and adolescents
with a diagnosis of ADHD. However, the low quality of
the underpinning evidence means that we cannot be
certain of the magnitude of the eects.
Within the short follow-up periods typical of the
included trials, there is some evidence that methylphe-
nidate is associated with increased risk of non-serious
adverse events, such as sleep problems and decreased
appetite, but no evidence that it increases the risk of
serious adverse events.
Implications for research
Better designed trials are needed to assess the benefits
of methylphenidate. Because of the frequency of
non-serious adverse events associated with methylphe-
nidate, the particular diculties for blinding of partici-
pants and outcome assessors point to the advantage of
large, “nocebo tablet” controlled trials. These use a pla-
cebo-like substance that causes adverse events in the
control arm that are comparable to those associated
with methylphenidate. Such trials ought first to be con-
ducted in adults with ADHD. We also acknowledge that
investigators can directly carry out nocebo controlled
trials in children and adolescents if they can argue that
young people with ADHD are dierent from adults with
the disorder.
Future trials should publish depersonalised individ-
ual participant data and report all outcomes, including
adverse events. This will enable researchers conducting
systematic reviews to assess dierences between inter-
vention eects according to sex, age, type of ADHD,
presence of co-morbidities and dose. Finally, the find-
ings highlight the urgent need for large randomised tri-
als of non-pharmacological treatments.
Conclusions
Methylphenidate use in children and adolescents may
improve the symptoms of ADHD, general behaviour,
and quality of life. It does not seem to cause an
increased risk of serious adverse events in the short
term but was associated with a relatively high risks of
doi: 10.1136/bmj.h5203
BMJ
2015;351:h5203 thebmj
RESEARCH
12
non-serious adverse events. These findings should be
interpreted in the light of several limitations, includ-
ing the lack of blinding, outcome reporting bias, het-
erogeneity, and the consequent very low quality of
evidence for all outcomes. More long term randomised
nocebo tablet (active placebo) controlled clinical tri-
als without risks of bias are necessary to allow firm
decisions on methylphenidate treatment in children
and adolescents with ADHD. We believe that nocebo
controlled trials should be conducted first in adults
with ADHD.
AUTHOR AFFILIATIONS
1Psychiatric Research Unit, Region Zealand Psychiatry, Denmark
2Child and Adolescent Psychiatric Department, Region Zealand,
Denmark
3Department of Psychology, Faculty of Health Science, University of
Southern Denmark
4Department of Psychiatry, Federal University of Rio Grande do Sul,
Porto Alegre, Brazil
5Copenhagen Trial Unit, Centre for Clinical Intervention Research,
Rigshospitalet, Copenhagen University Hospital, Copenhagen,
Denmark
6Islington CAMHS, Whittington Health, London, UK
7Western Sydney Local Health District; Mental Health, Parramatta,
Australia
8Psychiatric Centre North Zealand, The Capital Region of Denmark,
Denmark
9Pediatric Department, Herlev University Hospital, Herlev, Denmark
10Directorate for Health Information and Research, Department of
Health, G’Mangia, Malta
11South Asian Cochrane Network & Center, Prof BV Moses Center for
Evidence-Informed Health Care and Health Policy, Christian Medical
College, Vellore, India
12Institute of Clinical Medicine, Faculty of Health and Medical
Sciences, Copenhagen University, Copenhagen, Denmark
13The Cochrane Hepato-Biliary Group, Copenhagen Trial Unit, Centre
for Clinical Intervention Research, Rigshospitalet, Copenhagen
University Hospital, Copenhagen, Denmark
We thank Janus Christian Jacobsen (Copenhagen Trial Unit) for
elaborating the idea of conducting this review; Trine Lacoppidan
Kæstel (research librarian, Psychiatric Research Unit, Region
Zealand, Denmark) for her help with the search of studies and
description of the measurement scales; Lise Aagaard (University of
Southern Denmark) for the advice given during the work of this
review; Jesper Pedersen (Department of Childrens and Youths
Psychiatry, Region Zealand, Denmark) for backing up this project;
Torben Bille (Pediatric Department, Holbaek Hospital, Copenhagen,
Denmark) for helping to write the protocol and for performing the
selection of studies; Maria Gaardahl, Kim Boesen, Farhad Shokraneh,
and Rene Spijker for helping with the translation of articles in
Japanese, Italian, Turkish, Farsi and Dutch; Nadia Pedersen
(Psychiatric Research Unit) for helping with nalising the review;
Martina Riegl (senior medical assessor, Medicine and Healthcare
products Regulatory Agency, Special Populations Unit (Paediatrics),
London) for helping with data extraction and assessment of risk of
bias; Jacob Riis (user experience lead, the Nordic Cochrane Centre,
Copenhagen, Denmark) and Rasmus Moustgaard (senior systems
architect, the Nordic Cochrane Centre, Copenhagen, Denmark) for
help on issues regarding Review Manager. We thank Geraldine
McDonald (coordinating editor), Joanne Wilson (managing editor),
Gemma O’Loughlin (assistant managing editor), and Margaret
Anderson (trials search coordinator) of the Cochrane Developmental,
Psychosocial and Learning Problems Group for providing help and
support. We are grateful for the advice and support of Toby
Lasserson (senior editor) and David Tovey (editor in chief) of the
Cochrane Central Editorial Unit.
We also thank the many authors who responded to our requests for
further information on their trials, and the editors and peer reviewers of
the Cochrane Group and the BMJ for helpful comments on our Cochrane
review, which also influenced the wording of the present version.
ER and HBK are co-second authors on this review. This review is
an abridged version of a Cochrane systematic review: Storebø OJ,
Ramstad E, Krogh HB, etal. Methylphenidate for attention decit
hyperactivity disorder (ADHD) in children and adolescents. Cochrane
Database Syst Rev 2015 (In press).
Contributors: OJS, CGl, MS,SR, CGr, KBR, and ES wrote the protocol.
KBR developed the search strategy. OJS, ER, HK,TDN, MS, MH, FLM, SR,
and KBR carried out the study selection. OJS, ER, HK, TDN, MS, SR, MH,
CGJ, FLM, CMM, DG, KBR, DG, MZ, RK, and ES carried out the data
extraction and evaluation of bias. OJS and CGl developed the
analytical strategy. OJS, ER, HBK, MH, FLM, and CRMM entered data
into RevMan. OJS, ER, HBK, MH, FLM, and CRMM conducted the
statistical analysis. All authors participated in the discussion and
writing of the nal review. OJS is the guarantor.
Funding: This study received funding from Region Zealand Research
Foundation, Psychiatric Research Unit, Region Zealand Psychiatry,
Roskilde, Denmark and the Copenhagen Trial Unit, Centre for Clinical
Intervention Research, Copenhagen University Hospital, Copenhagen,
Denmark.
Competing interests: All authors have completed the ICMJE uniform
disclosure for at www.icmje.org/coi_disclosure.pdf (available on
request from the corresponding author) and declare: CRMM receives
nancial research support from the government agencies:
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
(CAPES) and Conselho Nacional de Desenvolvimento Cientíco e
Tecnológico (CNPq); has served as speaker to Novartis, developed
educational material for Novartis; received travel awards from the
Health Technology Assessment Institute (IATS), Universidade Federal
do Rio Grande do Sul (UFRGS), and travel and registration support to
the 4th World Congress on ADHD from the World Federation of ADHD;
MZ sits on the Paediatric Medicines Expert Advisory Group at the
Medicines and Healthcare Regulatory Agency, which considers
applications regarding the licensing of paediatric medicines. Payment
for MZ’s attendance at this meeting goes to his NHS organization. RK
is currently employed by South Asian Cochrane Centre, funded by
Indian Council for Medical Research, India and Eective Healthcare
Research Consortium (DFID), UK. CG received funds from the Lundbeck
Foundation to nance part of her Ph.D in the paediatric eld on
Tourette Syndrome. CG conrms that none of these funds were used to
work on this review.
Ethical approval: Not required.
Data sharing: Full data are available in the version of this Cochrane
review published by The Cochrane Library (www.cochrane library.com).
Transparency: The lead author (OJS) affirms that this manuscript is
an honest, accurate, and transparent account of the study being
reported; that no important aspects of the study have been
omitted; and that any discrepancies from the study as planned
(and, if relevant, registered) have been explained.
This is an Open Access article distributed in accordance with the
Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license,
which permits others to distribute, remix, adapt, build upon this work
non-commercially, and license their derivative works on dierent
terms, provided the original work is properly cited and the use is
non-commercial. See: http://creativecommons.org/licenses/
by-nc/4.0/.
1 Scahill L , Schwab-Stone M. Epidemiolog y of ADHD in scho ol-age
children. Child Adolesc Psychi atr Clin N Am 2000;9:541-55.
2 Polanczyk GV, Salum GA, Sugaya LS, Caye A, Rohde LA. Annual
research review: a meta-analysis of the worldwide prevalence of
mental disorders in children and adolescents. J Child Psychol
Psychiatry 2015;56:345-65.
3 Schmidt S, Petermann F. Developmental psychopathology:
Attention Decit Hyperactivity Disorder (ADHD). BMC Psychiatry
2009;9:58.
4 American Psychiatric Association. Diagnostic and statistical manual of
mental disorders : DSM-5. APA; 2013.
5 World Health Organization. The ICD-10 classication of mental and
behavioural disorders: clinical descriptions and diagnostic guidelines.
WHO; 1992.
6 Kadesjö B. ADHD hos barn och vuxna [ADHD in children and adults].
Socialstyrelsen; 2002.
7 National Institute for Health and Care Excellence. Attention decit
hyperactivity disorder: diagnosis and management of ADHD in
children, young people and adults. (Clinical guideline CG72.) 2008.
www.nice.org.uk/guidance/cg72.
8 Bloch MH, Panza KE, Landeros-Weisenberger A, Leckman JF.
Meta-analysis: treatment of attention-decit/hyperactivity disorder in
children with comorbid tic disorders. J Am Acad Child Adolesc
Psychiatry 2009;48:884-93.
thebmj
BMJ
2015;351:h5203 doi: 10.1136/bmj.h5203
RESEARCH
13
9 Charach A, Yeung E, Climans T, Lillie E. Childhood attention-decit/
hyperactivity disorder and future substance use disorders:
comparative meta-analyses. J Am Acad Child Adolesc Psychiatry
2011;50:9-21.
10 Charach A, Carson P, Fox S, Ali MU, Beckett J, Lim CG. Interventions for
preschool children at high risk for ADHD: a comparative eectiveness
review. Pediatrics 2013;131:e1584-e1604.
11 Faraone SV, Biederman J, Roe C. Comparative ecacy of Adderall and
methylphenidate in attention-decit/hyperactivity disorder: a
meta-analysis. J Clin Psychopharmacol 2002;22:468-73.
12 Faraone SV, Biederman J, Spencer TJ, Aleardi M. Comparing the
ecacy of medications for ADHD using meta-analysis. Med Gen Med
2006;8:4.
13 Faraone SV. Using Meta-analysis to compare the ecacy of
medications for attention-decit/hyperactivity disorder in youths.
P T 2009;34:678-94.
14 Faraone SV, Glatt SJ. A comparison of the ecacy of medications for
adult attention-decit/hyperactivity disorder using meta-analysis of
eect sizes. J Clin Psychiatry 2010;71:754-63.
15 Hanwella R, Senanayake M, de Silva V. Comparative ecacy and
acceptability of methylphenidate and atomoxetine in treatment of
attention decit hyperactivity disorder in children and adolescents: a
meta-analysis. BMC Psychiatry 2011;11:176.
16 Kambeitz J, Romanos M, Ettinger U. Meta-analysis of the association
between dopamine transporter genotype and response to
methylphenidate treatment in ADHD. Pharmacogenomics J
2014;14:77-84.
17 King S, Grin S, Hodges Z, etal. A systematic review and economic
model of the eectiveness and cost-eectiveness of methylphenidate,
dexamfetamine and atomoxetine for the treatment of attention decit
hyperactivity disorder in children and adolescents. Health Technol
Assess 2006;10:iii-146.
18 Maia CR, Cortese S, Caye A, etal. Long-term ecacy of
methylphenidate immediate-release for the treatment of childhood
ADHD: a systematic review and meta-analysis. J Atten Disord 2014;
published online 10 Dec.
19 Punja S, Zorzela L, Hartling L, Urichuk L, Vohra S. Long-acting versus
short-acting methylphenidate for paediatric ADHD: a systematic
review and meta-analysis of comparative ecacy. BMJ Open
2013;3:e002312.
20 Reichow B, Volkmar FR, Bloch MH. Systematic review and
meta-analysis of pharmacological treatment of the symptoms of
attention-decit/hyperactivity disorder in children with pervasive
developmental disorders. J Autism Dev Disord 2013;43:2435-41.
21 Schachter HM, Pham B, King J, Langford S, Moher D. How ecacious
and safe is short-acting methylphenidate for the treatment of
attention-decit disorder in children and adolescents? A meta-
analysis. CMAJ 2001;165:1475-88.
22 Van der Oord S, Prins PJ, Oosterlaan J, Emmelkamp PM. Ecacy of
methylphenidate, psychosocial treatments and their combination in
school-aged children with ADHD: a meta-analysis. Clin Psychol Rev
2008;28:783-800.
23 Higgins JPT, Green S. Cochrane handbook for systematic reviews of
interventions. Version 5.1.0 [updated March 2011]. www.cochrane-
handbook.org.
24 Moher D, Shamseer L, Clarke M, etal. Preferred reporting items for
systematic review and meta-analysis protocols (PRISMA-P) 2015
statement. Syst Rev 2015;4:1.
25 Liberati A, Altman DG, Tetzla J, etal. The PRISMA statement for
reporting systematic reviews and meta-analyses of studies that
evaluate healthcare interventions: explanation and elaboration. BMJ
2009;339:b2700.
26 Storebø O, Ramstad E, Krogh H, etal. Methylphenidate for attention
decit hyperactivity disorder (ADHD) in children and adolescents.
Cochrane Database Syst Rev 2015 (In press).
27 Storebø O, Pedersen N, Krogh H, etal. Methylphenidate for attention
decit hyperactivity disorder (ADHD) in children and adolescents—
assessment of harms in observational studies. Cochrane Database
Syst Rev 2015; (In press).
28 Storebø OJ, Rosendal S, Skoog M, etal. Methylphenidate for
attention decit hyperactivity disorder (ADHD) in children and
adolescents [Protocol]. Cochrane Database Syst Rev 2012;5:
CD009885.
29 ICH Expert Working Group. ICH harmonized tripartite guideline:
guideline for good clinical practice E6(R1) 1996. http://bit.ly/1B0jeJg.
30 Zhang S, Faries DE, Vowles M, Michelson D. ADHD Rating Scale IV:
psychometric properties from a multinational study as a clinician-
administered instrument. Int J Methods Psychiatr Res
2005;14:186-201.
31 Rentz AM, Matza LS, Secnik K, Swensen A, Revicki DA. Psychometric
validation of the child health questionnaire (CHQ) in a sample of
children and adolescents with attention-decit/hyperactivity disorder.
Qual Life Res 2005;14:719-34.
32 Curtin F, Elbourne D, Altman DG. Meta-analysis combining parallel and
cross-over clinical trials. III: The issue of carry-over. Stat Med
2002;21:2161-73.
33 Hartman CA, Rhee SH, Willcutt EG, Pennington B. Modeling Rater
Disagreement for ADHD: Are Parents or Teachers Biased? J Abnorm
Child Psychol 2007;35:536-42.
34 Carlson GA, Dunn D, Kelsey D, etal. A pilot study for augmenting
atomoxetine with methylphenidate: safety of concomitant therapy in
children with attention-decit/hyperactivity disorder. Child Adolesc
Psychiatry Ment Health 2007;1:10.
35 Findling RL, Short EJ, McNamara NK, etal. Methylphenidate in the
treatment of children and adolescents with bipolar disorder and
attention-decit/hyperactivity disorder. J Am Acad Child Adolesc
Psychiatry 2007;46:1445-53.
36 Newcorn JH, Kratochvil CJ, Allen AJ, etal. Atomoxetine and osmotically
released methylphenidate for the treatment of attention decit
hyperactivity disorder: acute comparison and dierential response.
Am J Psychiatry 2008;165:721-30.
37 Palumbo DR, Sallee FR, Pelham WE, Bukstein OG, Daviss WB,
McDermott MP. Clonidine for attention-decit/hyperactivity disorder:
I. Ecacy and tolerability outcomes. J Am Acad Child Adolesc
Psychiatry 2008;47:180-8.
38 Tucker JD, Suter W, Petibone DM, etal. Cytogenetic assessment of
methylphenidate treatment in pediatric patients treated for attention
decit hyperactivity disorder. Mutat Res 2009;677:53-8.
39 Oesterheld JR, Kofoed L, Tervo R, Fogas B, Wilson A, Fiechtner H.
Eectiveness of methylphenidate in Native American children with
fetal alcohol syndrome and attention decit/hyperactivity disorder:
a controlled pilot study. J Child Adolesc Psychopharmacol
1998;8:39-48.
40 Pearson DA, Santos CW, Aman MG, etal. Eects of extended release
methylphenidate treatment on ratings of attention-decit/
hyperactivity disorder (ADHD) and associated behavior in children
with autism spectrum disorders and ADHD symptoms. J Child Adolesc
Psychopharmacol 2013;23:337-51.
41 Smith BH, Pelham WE, Evans S, etal. Dosage eects of
methylphenidate on the social behavior of adolescents diagnosed
with attention-decit hyperactivity disorder. Exp Clin
Psychopharmacol 1998;6:187-204.
42 Taylor E, Schachar R, Thorley G, Wieselberg HM, Everitt B, Rutter M.
Which boys respond to stimulant medication? A controlled trial of
methylphenidate in boys with disruptive behaviour. Psychol Med
1987;17:121-43.
43 Green T, Weinberger R, Diamond A, etal. The eect of
methylphenidate on prefrontal cognitive functioning, inattention, and
hyperactivity in velocardiofacial syndrome. J Child Adolesc
Psychopharmacol 2011;21:589-95.
44 Szobot CM, Rohde LA, Katz B, etal. A randomized crossover
clinicalstudy showing that methylphenidate-SODAS improves
attention-decit/hyperactivity disorder symptoms in
adolescentswith substance use disorder. Braz J Med Biol Res
2008;41:250-7.
45 Roberts I, Ker K, Edwards P, Beecher D, Manno D, Sydenham E. The
knowledge system underpinning healthcare is not t for purpose and
must change. BMJ 2015;350:h2463.
46 Wetterslev J, Thorlund K, Brok J, Gluud C. Trial sequential analysis may
establish when rm evidence is reached in cumulative meta-analysis.
J Clin Epidemiol 2008;61:64-75.
47 Brok J, Thorlund K, Gluud C, Wetterslev J. Trial sequential analysis
reveals insufficient information size and potentially false
positiveresults in many meta-analyses. J Clin Epidemiol 2008;61:
763-9.
48 Brok J, Thorlund K, Wetterslev J, Gluud C. Apparently conclusive
meta-analysis may be inconclusive—trial sequential analysis
adjustment of random error risk due to repetitive testing of
accumulating data in apparently conclusive neonatal meta-analysis.
Int J Epidemiol 2009;38:287-98.
49 Thorlund K, Devereaux PJ, Wetterslev J, etal. Can trial sequential
monitoring boundaries reduce spurious inferences from meta-
analysis? Int J Epidemiol 2009;38:276-86.
50 Copenhagen Trial Unit. TSA trial sequential analysis. 0.9 Beta.
[Computer program].Copenhagen Trial Unit 2011. http://ctu.dk/tsa/
downloads.aspx.
51 Wetterslev J, Thorlund K, Brok J, Gluud C. Estimating required
information size by quantifying diversity in random-eects model
meta-analyses. BMC Med Res Methodol 2009;9:86.
52 Kjaergard LL, Villumsen J, Gluud C. Reported methodologic quality
and discrepancies between large and small randomized trials in
meta-analyses. Ann Intern Med 2001;135:982-9.
53 Lundh A, Sismondo S, Lexchin J, Busuioc OA, Bero L. Industry
sponsorship and research outcome. Cochrane Database Syst Rev
2012;12:MR000033.
54 Moher D, Pham B, Jones A, etal. Does quality of reports of
randomised trials aect estimates of intervention ecacy reported in
meta-analysis? Lancet 1998;352:609-13.
55 Savovic J, Jones HE, Altman DG, etal. Influence of reported
studydesign characteristics on intervention eect estimates
fromrandomized, controlled trials. Ann Intern Med 2012;157:
429-38.
RESEARCH
No commercial reuse: See rights and reprints http://www.bmj.com/permissions Subscribe: http://www.bmj.com/subscribe
56 Savovic J, Jones H, Altman D, etal. Influence of reported study design
characteristics on intervention eect estimates from randomised
controlled trials: combined analysis of meta-epidemiological studies.
Health Technol Assess 2012;16:1-82.
57 Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of
bias. Dimensions of methodological quality associated with estimates
of treatment eects in controlled trials. JAMA 1995;273:408-12.
58 Wood L, Egger M, Gluud LL, etal. Empirical evidence of bias in treatment
eect estimates in controlled trials with dierent interventions and
outcomes: meta-epidemiological study. BMJ 2008;336:601-5.
59 Gluud C, Nikolova D, Klingenberg SL. Cochrane Hepato-Biliary About
the Cochrane Collaboration (Cochrane Review Groups (CRGs).
Cochrane Rev Groups 2015;1(Art.No:LIVER). www.cochranelibrary.
com/about/cochrane-review-groups.html.
60 Andrews J, Guyatt G, Oxman AD, etal. GRADE guidelines: 14. Going
from evidence to recommendations: the signicance and
presentation of recommendations. J Clin Epidemiol 2013;66:719-25.
61 Thorlund K, Engstrøm J, Wetterslev J, Brok J, Imberger G, Gluud C. User
manual for Trial Sequential Analysis (TSA). Copenhagen Trial Unit,
Centre for Clinical Intervention Research. Copenhagen, Denmark
2011;1. www.ctu.dk/tsa.
62 Moher D, Liberati A, Tetzla J, Altman DG. Preferred reporting items for
systematic reviews and meta-analyses: the PRISMA Statement. Open
Med 2009;3:e123-e130.
63 Martinez-Raga J, Knecht C, Szerman N, Martinez MI. Risk of serious
cardiovascular problems with medications for attention-decit
hyperactivity disorder. CNS Drugs 2013;27:15-30.
64 Lundh A, Sismondo S, Lexchin J, Busuioc OA, Bero L. Industry
sponsorship and research outcome. Cochrane Database Syst Rev
2012;12:MR000033.
65 Schroll JB, Bero L, Gotzsche PC. Searching for unpublished data for
Cochrane reviews: cross sectional study. BMJ 2013;346:f2231.
66 Murray DW, Kollins SH, Hardy KK, etal. Parent versus teacher ratings
of attention-decit/hyperactivity disorder symptoms in the
preschoolers with Attention-Decit/Hyperactivity Disorder Treatment
Study (PATS). J Child Adolesc Psychopharmacol 2007;17:605-19.
67 Lavigne JV, Dulcan MK, LeBailly SA, Binns HJ. Can parent reports serve
as a proxy for teacher ratings in medication management of
attention-decit hyperactivity disorder? J Dev Behav Pediatri
2012;33:336-42.
68 Caye A, Machado JD, Rohde LA. Evaluating parental disagreement in
ADHD diagnosis: can we rely on a single report from home? J Atten
Disord 2013; published online 4 Oct.
69 Efstratopoulou M, Simons J, Janssen R. Concordance among physical
educators’, teachers’, and parents’ perceptions of attention problems
in children. J Atten Disord 2013;17:437-43.
70 Epstein T, Patsopoulos NA, Weiser M. Immediate-release
methylphenidate for attention decit hyperactivity disorder (ADHD) in
adults. Cochrane Database Syst Rev 2014;9:CD005041.
71 Storebø O, Gluud C. Criticism to “Immediate-release methylphenidate
for attention decit hyperactivity disorder (ADHD) in adults”. Email to:
T Epstein via Wiley Online Feedback form, 12 May 2015.
© BMJ Publishing Group Ltd 2015
