Attention-deficit hyperactivity disorder (ADHD) (1) is a common
psychiatric disorder occurring in approximately 5% of school-aged
children, adolescents, and adults (1,2). It is a frequent reason
for referral to mental health services (3). ADHD is first evident
in the preschool years (4) and persists in as many as 70% of individuals
throughout childhood and adolescence and into adulthood (5,6). Affected
individuals are at risk for oppositional defiant disorder (ODD),
conduct disorder (CD), depression, anxiety, and learning disability
(LD) (7). They are also at risk for the development of delinquency
or criminality, and for school suspension, academic underachievement,
low self-esteem, and substance abuse (8). Treatments must be of
sufficient intensity to have an immediate impact on the core ADHD
symptoms, and they must be of adequate duration and intensity to
alter adverse outcomes, many of which do not develop for several
In 1997, the Agency for Healthcare Research and Quality charged
the McMaster Evidence-based Practice Centre with the conduct of
a comprehensive, systematic review of the literature on ADHD treatment
. Our first objective was to conduct a formal critical appraisal
of existing systematic reviews and metaanalyses. In addition to
other issues, we wished to determine whether existing reviews had
adequately evaluated the evidence on the effectiveness of extended
treatment for ADHD. The main conclusion from the appraisal of published
reviews (9) was that most existing systematic reviews and metaanalyses
had extensive flaws due to poor description of the methods used
to identify, select, assess, and synthesize information. Of the
13 reviews (3,1021), only 2 (10,13) had minimal flaws, but
these reviews were narrowly focused on treatment effects as measured
on the Continuous Performance Task (13) or on school-based interventions
(10). Only 1 review (15) specifically evaluated evidence for long-term
ADHD treatment, but that study had significant methodological limitations.
Consequently, the evidence for the effectiveness of long-term pharmacologic
and nonpharmacologic interventions for ADHD was not addressed by
previously published reviews.
This article systematically appraises the methods and results of
published studies in which ADHD treatment persisted for more than
12 weeks (Table
1, 22–35). Although ADHD treatment typically continues for many
years (36), we identified those studies in which treatment persisted
for a minimum of 12 weeks under randomized conditions. This criterion
was arbitrary but ensured that the review would identify the longest
studies from which to draw preliminary conclusions. This distinction
does not, of course, imply that 3 months of treatment defines optimal
extended therapy. The entire report is available at (http://www.ahcpr.gov:80/
clinic/index.html#evidence). We focused on evidence provided
by randomized controlled trials (RCTs) because they are the simplest
and most powerful research designs in which to evaluate the efficacy
and effectiveness of interventions.
Selection of Studies
Inclusion and Exclusion Criteria. We regarded as
potentially eligible full reports published in peer-reviewed journals,
in any language, that focused on the treatment of ADHD in humans.
We selected any reported RCT in which treatment persisted for a
minimum of 12 weeks under randomized conditions. Where studies included
subjects with conditions other than ADHD, they were selected only
if they provided separate analyses for patients with ADHD.
Search Strategy. Potentially eligible studies were
identified through a systematic search of Medline (from 1966), CINAHL
(from 1982), HEALTHStar (from 1975), Psycinfo (from 1984), and EMBASE
(from January 1, 1984, to November 30, 1997). (The search strategy
may be obtained from the principal author.) We also searched the
Cochrane Library (issue 4, 1997). Finally, we searched the reference
lists of any eligible article from these sources, as well as files
of research team members and partner organizations.
We developed and tested data extraction forms for this project
Two reviewers independently extracted data from each full report.
Differences were resolved by consensus and by referring to the information
in the original report. Any differences that the 2 reviewers could
not resolve were resolved by a third reviewer. The original reports
were not masked because masking is time-consuming and does not have
an important impact on the results of systematic reviews (37).
Data were extracted on each of the treatment arms, outcomes, and
tests used in each study. Information on outcomes was extracted
regardless of the instrument or test used within each study. Only
information gathered during the subjects participation in
the randomized groups was extracted: no data gathered after discontinuation
of therapy were sought.
The outcomes of interest, selected a priori by the panel of experts
(led by the AAP Task Force), were as follows:
Behaviour Symptoms: observed core symptoms of inattention,
hyperactivity, and impulsivity, and the results of a laboratory
test of attention.
Academic Performance: results of achievement tests
and grades, as well as verbal, reading, mathematics, and spelling
Social Behaviour, Conduct, and Oppositional Disturbance:
diagnosis of ODD or CD, degree of aggressiveness and social competence.
Internalizing Symptoms: depression, anxiety, crying,
sadness, global mood, and level of emotional well-being and self-esteem.
Assessment of Study Quality
We assessed the methodological quality of the studies in several
Quality Assessment Scale. The methodological quality
scale (38) produces a minimum score of 0 points and a maximum score
of 5 points. This scale assesses whether the study includes a description
of appropriate methods to generate the randomization sequence (2
points), double blinding (2 points), and a detailed account of withdrawals
and dropouts (1 point). The higher the score, the better the methodological
quality of the RCT. Studies with a score of less than or equal to
2 points have been shown to exaggerate estimates of intervention
effects by more than 30%, on average (39). Blinding helps prevent
ascertainment bias and protects the randomization sequence after
allocation, but cannot always be implemented, as was the case for
most trials in this review (40). Even when a trial was not double-blind,
a study could still be awarded 3 points if it included a description
of randomization and adherence.
Concealment. We determined whether the allocation
of individuals to the different study groups had been concealed
until after their consent was obtained. Studies in which allocation
was unclear or inadequately concealed have been shown to exaggerate
estimates of intervention effects, by more than 35%, on average
(39,40). Allocation concealment prevents selection bias, protects
the randomization sequence before the interventions are given to
study participants, and can always be implemented.
Relation with Industry. We also sought information
on any association between the investigators and the pharmaceutical
or a related industry. It has been shown that reports of trials
sponsored by pharmaceutical companies are more likely to favour
the experimental intervention than are trials not sponsored by pharmaceutical