Clinical Implications
• Meta-analyses are considered to provide a good level of evidence and are frequently used to ground treatment recommendation. It is possible, however, that important topics of clinical relevance for the treatment of attention-deficit/hyperactivity disorder (ADHD) in adults are not yet covered by meta-analyses. It is important to identify clinically relevant questions covered and not covered by meta-analyses to plan future clinical trials and meta-analyses.
• There are, to date, 8 meta-analyses computing effect size (ES) of different interventions for the treatment of adult ADHD.
• Stimulants are effective on a short-term basis with a medium to large ES. Bupropion is superior to placebo but less effective than stimulants. Pooled estimation for the ES of other drugs is not described in meta-analyses. No high quality meta-analyses for psychosocial interventions were identified.
• Many questions of crucial importance for the treatment of adults with ADHD are not covered by available meta-analyses.
Introduction
Attention-Deficit/Hyperactivity Disorder (ADHD) is one of the most common mental disorders in childhoodReference Polanczyk, de Lima, Horta, Biederman and Rohde1. Although childhood ADHD is one of the most studied psychiatric conditions worldwideReference Lopez-Munoz, Alamo, Quintero-Gutierrez and Garcia-Garcia2, far less attention has been given to adult ADHD, a concerning reality given the fact that at least half of those children affected will carry symptoms and associated functional deficit to adult lifeReference Faraone, Biederman and Mick4.
Although the prevalence of adult ADHD is not yet well established, there is some evidence that at least 2% of the adult population suffers from the disorderReference Simon, Czobor, Balint, Meszaros and Bitter5, Reference Kessler, Adler and Barkley6, Reference Faraone and Biederman7. A pooled estimation of studies from non-representative samples of the population found a prevalence of 2.5% (2.1–3.1)Reference Simon, Czobor, Balint, Meszaros and Bitter5; a study approaching a representative community based population but using indirect estimation found an even higher prevalence of around 4.4% in the USAReference Kessler, Adler and Barkley6 and a rate of 5.2%Reference Faraone and Biederman7 for cross-national data.
The number of adults affected contrasts markedly with the imbalance of available literature about children and adult ADHD; hundreds of clinical trials have already been conducted to test different interventions for children with ADHD, and at least 17 meta-analyses have been published on this topic.Reference Schachter, Pham, King, Langford and Moher8–Reference Faraone, Biederman, Spencer and Aleardi24 Nevertheless, empirically based information about the treatment of adults with ADHD is considerably less frequent in the literature. A systematic review of guidelines for the treatment of ADHD available until 2011 found 13 guidelines.Reference Seixas, Weiss and Muller25 Among those, only 1 was exclusively dedicated to the treatment of adults, but its quality was compromised by serious methodological problems.Reference Ebert, Krause and Roth-Sackenheim26 Three other guidelines address the treatment of adults and children,Reference Nutt, Fone and Asherson27–29 and all the others are dedicated to the treatment of only children with ADHD. Most guidelines suggest stimulants as first-line treatments for adults with ADHD, but recommendations are more vague and divergent for psychotherapy and other drugs. These kinds of recommendations contrast with those for children with ADHD, for whom psychosocial interventions are frequently recommended as first line treatments, either alone or in combination with pharmacotherapy.28, 30, Reference Pliszka31
Although meta-analyses have several important limitations, many treatment guidelines have used meta-analyses as the ultimate level of evidence in the past few years. This approach can lead to distorted recommendations based on which topics are covered by meta-analysis. Because the scientific literature about ADHD has grown exponentially in the past few decades,Reference Lopez-Munoz, Alamo, Quintero-Gutierrez and Garcia-Garcia2 the large number of studies available makes meta-analysis a suitable approach for the systematization of available data on this topic. It is possible, however, that important clinical questions have not been explored in meta-analyses. We systematically reviewed the literature, searching for meta-analyses about the treatment of adult ADHD to verify to what extent clinically relevant questions about the treatment of adults with the disorder can be answered by available meta-analyses. The identification of clinically relevant questions covered and not covered by meta-analyses can help to plan future clinical trials and meta-analyses in this field, and in addition can provide evidence-based information for treatment recommendations.
Methods
Relevant publications were identified by searching the PubMed and Cochrane databases using the following search keywords: “ADHD” and [“pharmacotherapy” or “stimulants” or “antidepressant” or “atomoxetine” or “modafinil” or “alpha-2 agonists” or “psychosocial” or “psychotherapy” or “school based intervention” or “behavioral therapy” or “cognitive therapy” or “cognitive training” or “complementary therapy” or “alternative therapy”]. No limit for date was set. The matches were restricted to publication type, and only reviews were retrieved. This search resulted in 191 titles. Abstracts for those 191 titles were independently reviewed by two evaluators (authors FST and KMF). Articles were included in this systematic review if (1) they described the results of a meta-analysis for the treatment of ADHD, (2) at least 1 study included in the meta-analysis was conducted in the adult population (age older than 18 years), (3) the intervention tested was a drug or a psychosocial treatment, and (4) the article was available in the English language. If the authors were aware of any other studies fulfilling these criteria that were not retrieved using the search strategy, those were also included. Reviews of reference lists of all articles included were also performed.
Information of interest was extracted from the manuscript independently by 3 of the authors (FST, KMF, and TSM), and 1 of the authors (TSM) congregated all information that was collected. The following information was extracted: first author name and year of publication, number of studies included, total sample size, age of patients included, proportion of males, inclusions and exclusion criteria, interventions under comparison for the calculation of effect size (ES), duration of the trials included, sources and dates of search, overall quality of studies included, and main results in regard to ES for the main outcomes considered. When publication bias was identified (regularly using funnel plots), corrected ES was preferred. Evaluators were also instructed to make notes about other data they considered to be of clinical relevance or importance for the evaluation of the quality of the meta-analysis.
Results
We identified 8 studies that fulfilled inclusion criteria (see Tables 1–3 for an overview of the studies). Three studies computed pooled effect sizes for the comparison of drugs with placeboReference Faraone, Spencer, Aleardi, Pagano and Biederman32–Reference Verbeeck, Tuinier and Bekkering34 (Table 1), and 4 compared different drugsReference Peterson, McDonagh and Fu35–Reference Faraone and Glatt38 (Table 2). Only 1 study covered psychosocial interventionReference Linderkamp and Lauth39 (Table 3), but this study has methodological problems and should be carefully considered. The main findings for these studies are described in the following sections and in Tables 1–3.
Table 1 Meta-analyses of trials testing drugs in comparison to placebo
*This is stated in the methodologies, but there is no placebo group described in one of the studies described, the one testing lithium.
ADHD = attention-deficit/hyperactivity disorder; CGI = Clinical Global Improvement Scale; DSMII, DSMIII, DMSIIIR, DSMIV = Diagnostic and Statistical Manual of Mental Disorders, 2nd, 3rd, 3rd rev., or 4th edition, respectively; ES = effect size; MAS = mixed amphetamine salts; OR = odds ratio; RR = relative risk; vs = versus; SD = standard Deviation; SR = standard release; y = years.
Table 2 Meta-analyses of trials comparing different drugs for ADHD
AACAP = American Academy of Child and Adolescence Psychiatry; ADHD = attention-deficit/hyperactivity disorder; APA = American Psychiatric Association; CI = confidence interval; DSMII, DSMIII, DMSIIIR, DSMIV = Diagnostic and Statistical Manual of Mental Disorders, 2nd, 3rd, 3rd rev., or 4th edition, respectively; ES = effect size; LA-St = long-acting stimulants; MAS = mixed amphetamine salts; RR = relative risk; vs = versus; SA-St = short-acting stimulants; SR = standard release; y = years.
Table 3 Meta-analysis of trials comparing psychosocial interventions and pharmacotherapy for the treatment of ADHD
ADHD = attention-deficit/hyperactivity disorder; ES = effect size.
Stimulant medications for the treatment of adults with ADHD
The available meta-analyses consistently showed stimulants to significantly decrease ADHD symptoms on a short-term basis when compared to placebo with a medium to high ES (see Tables 1 and 2). For the direct comparison with placebo, ESs were 0.9 for methylphenidate,Reference Faraone, Spencer, Aleardi, Pagano and Biederman32 0.73 for mixed amphetamine salts,Reference Castells, Ramos-Quiroga, Bosch, Nogueira and Casas36 0.6 for dexamphetamine,Reference Castells, Ramos-Quiroga, Bosch, Nogueira and Casas36 0.8 for lisdexamphetamine extracted from a single study,Reference Castells, Ramos-Quiroga, Bosch, Nogueira and Casas36 and 0.67 when stimulants were considered in general.Reference Meszaros, Czobor and Balint37 The use of higher doses was associated with larger effect size.Reference Faraone, Spencer, Aleardi, Pagano and Biederman32 Proportion of respondents was used as an outcome only for one meta-analysis, which showed short-acting stimulants to reach an ES of 4.32, while long-acting stimulants performed significantly lower with an ES of 1.35.Reference Peterson, McDonagh and Fu35 This difference regarding ES for long- and short-acting stimulants was also found in another meta-analysis by Faraone and Glatt,Reference Faraone and Glatt38 which found a nonsignificant statistical difference for the effect of short- versus long-acting stimulants (0.96 versus 0.73, respectively). Only one meta-analysis looked at the issue of adherence. Castells etal Reference Castells, Ramos-Quiroga, Bosch, Nogueira and Casas36 compared the pooled effect of different amphetamine derivatives and found all of them to significantly decrease ADHD symptoms. However, only mixed amphetamine salts (MAS) increased retention to treatment.
Nonstimulant medications for the treatment of adults with ADHD
Significantly fewer meta-analytic data exist with regarding the effect of nonstimulant drugs (see Table 2 for an overview). Two meta-analyses computed the ES of nonstimulants as a group. One found nonstimulants to have a significant lower ES of 0.39 than that of stimulants (ES for short-acting stimulants was 0.96 and for long-acting stimulants was 0.73Reference Faraone and Buitelaar11). In the other meta-analysis, the ES of nonstimulants was 0.59 versus 0.67 of stimulants, but between-group differences were not tested.Reference Meszaros, Czobor and Balint37 Data on the effect of individual drugs are scarce, and pooled estimations are only available for bupropion. Three different studies computed the pooled ES for bupropion. All of them used response rate as the main outcome; 2 of these meta-analyses were based on the same 5 trials, but used different statistical methods and outcomes and found different results. Maneeton etal Reference Maneeton, Maneeton, Srisurapanont and Martin33 found a pooled relative risk (RR) of 1.67, and Verbeeck etal Reference Verbeeck, Tuinier and Bekkering34 found an odds ratio (OR) of 2.42. In both cases, results favored bupropion over placebo. In Maneeton etal's meta-analysis, the number needed to treat found for bupropion was 4.6, and the discontinuation rate due to adverse events was not higher for bupropion than that of placebo.Reference Maneeton, Maneeton, Srisurapanont and Martin33 Peterson etal Reference Peterson, McDonagh and Fu35 used similar inclusion criteria for a systematic search of clinical trials but found only 3 of the 5 studies identified by the other 2 authors; for this study, pooled ES for bupropion was 1.87 for the comparison with placebo, and a RR of 2.24 was found favoring short-acting stimulants over bupropion. Some data about individual studies for other antidepressants were described in Verbeeck etal's meta-analysis (Table 2),Reference Verbeeck, Tuinier and Bekkering34 but because only one study per drug was available, pooled estimation was not possible. No pooled estimation was identified for atomoxetine, alpha-2-agonists, modafinil, and other antidepressants.
Psychosocial interventions for the treatment of adults with ADHD
Only one meta-analysis was identified that computed the pooled ES for psychosocial interventions; however, this meta-analysis has important methodological problems.Reference Linderkamp and Lauth39 The inclusion criteria for trials of psychosocial interventions are not clearly stated, but apparently noncontrolled trials were included, while for the estimation of the ES for pharmacological treatment, only controlled trials were admitted. The authors found an ES of 0.84 for cognitive behavioral therapy and 0.44 for pharmacotherapy, but the manuscript lacks a clear description about which outcome measures were considered for the ES estimation.
Discussion
We conducted a systematic review to identify meta-analyses about the treatment of adults with ADHD. We identified 8 meta-analyses that computed the ES of pharmacological and psychosocial interventions. Most data available were about stimulants; very little information about nonstimulants or psychosocial treatments was available.
Stimulants were consistently shown to decrease ADHD symptoms, and the ESs found were medium to high, independent of the stimulant drug considered. The highest ES was found for methylphenidate,Reference Faraone, Spencer, Aleardi, Pagano and Biederman32 but no direct comparison between methylphenidate and other drugs is available. Short-acting stimulants were superior to long-acting stimulants.Reference Peterson, McDonagh and Fu35, Reference Faraone and Glatt38 Apart from stimulants, the only drug to have a pooled estimation calculated for its ES was bupropion.Reference Maneeton, Maneeton, Srisurapanont and Martin33, Reference Peterson, McDonagh and Fu35 Bupropion was inferior to stimulants. When the ES for all nonstimulants as a group was compared to stimulants, nonstimulants performed significantly lower.Reference Meszaros, Czobor and Balint37, Reference Faraone and Glatt38 The only study to compute the pooled ES for psychosocial intervention has limitations, and no definitive conclusion can be drawn from its data.Reference Linderkamp and Lauth39
It must be highlighted that meta-analyses have several limitations. Most importantly, they rely on available clinical trials, and the methodology used by each author may vary markedly. For this reason, the ESs reported in this manuscript should not be interpreted as unequivocal, and, most importantly, they are not comparable between each other. Values of ES should be interpreted under the context of the methodology used in each specific meta-analysis. A good example was described in this review, in which two different meta-analyses computed the ES of bupropion based on the same 5 clinical trials and found different values.Reference Maneeton, Maneeton, Srisurapanont and Martin33, Reference Verbeeck, Tuinier and Bekkering34 The quality of studies included and the outcomes measures can also lead to distorted results. If the ES is computed based on a pre–post treatment comparison, it is more likely to overestimate the power of the intervention than ESs that are calculated based on the comparison with a passive control group. ES based in passive control groups, however, is likely to be overestimated when compared to that coming from studies that used active control groups. The dose of medication used in the trials included also has to be considered; when trials using low doses of medication are included, there will be an obvious tendency to report lower ESs.
The results presented here show a concerning lack of meta-analytic data to answer a number of important clinical questions.
First, the mean age range for almost all meta-analyses was between 35 and 45 years, and consequently no conclusions can be drawn with regarding the treatment response of older adults and the elderly population. Although there is limited information on the prevalence of ADHD among the elderly, a recent, well conducted epidemiological study from the Netherlands revealed a prevalence of 2.8% for the full-blown syndrome, but a higher prevalence of 4.2% when symptomatic cases not fulfilling criteria were considered.Reference Michielsen, Semeijn and Comijs40 These data suggest that a large population of sufferers is being disregarded in clinical trials with better methodology. It is important to note that these are the ones included in meta-analyses.
Second, although 2 meta-analyses provide the ES for nonstimulants as a group, pooled estimations of the ES for most drugs individually are not available. It is worth mentioning that there is no pooled estimation for atomoxetine, a drug that is recommended as first-line treatment in different guidelines and with a number of clinical trials already available.Reference Seixas, Weiss and Muller25 It is important to note that the computation of the ES for nonstimulants as a group is of reduced clinical relevance, since nonstimulants make reference to a very heterogeneous group of drugs without shared pharmacological properties, and consequently have very different effects with no expected similar efficacy.
Third, most meta-analyses have used mean scores from ADHD rating scales for the estimation of ES. The field has progressed in recent years, as researchers have become much more interested in functional outcomes than simply in reduction of symptom scores.Reference Weiss, Gadow and Wasdell41 It would be interesting to have more information about other important outcomes, such as academic and occupational functioning, interpersonal relationships, and reduction in accidents.
Fourth, most trials included in the meta-analyses were short and lacked information about adherence. Adherence to treatment is a very relevant issue, because naturalistic studies have shown that persistence for drug treatment with stimulants is not the rule. More than 50% of patients quit stimulants after 90 days of treatment, and treatment maintenance for more than 180 days is unusual.Reference Marcus, Wan, Kemner and Olfson42–Reference Lawson, Johnsrud, Hodgkins, Sasane and Crismon44 Of particular interest is the issue of adherence to long- versus short-acting stimulants. Long-acting drugs are more expensive, and different guidelines suggest that their prescription would increase adherence due to their better once-a-day posology. Adherence has only been analyzed as an outcome for the ES estimation in one meta-analysis. This meta-analysis, however, did not compare long- to short-acting drugs, but instead, it compared different amphetamine derivatives (MAS, lisdexamphetamine, and dextroamphetamine).Reference Castells, Ramos-Quiroga, Bosch, Nogueira and Casas36 Also of note is the tendency for higher ESs for short- versus long-acting formulations (although this was not statistically significant in one of the meta-analyses). This finding is surprising, since short-acting and long-acting stimulants are pharmacologically identical, with the only difference being pharmacokinetic properties favoring long-acting stimulants. In the meta-analysis Faraone etal,Reference Faraone, Spencer, Aleardi, Pagano and Biederman32 however, there is some evidence that the superiority found for short-acting drugs may have been caused by publication bias.Reference Castells, Ramos-Quiroga, Bosch, Nogueira and Casas36 After correcting for study characteristics that were significantly biasing the studies’ results, differences between short- and long-acting stimulants disappeared.Reference Castells, Ramos-Quiroga, Bosch, Nogueira and Casas36
Finally, of major concern is the lack of meta-analytic data about psychosocial treatments. Psychosocial treatments are routinely prescribed for ADHD patients in an attempt to reduce psychosocial problems related to symptoms. For the pediatric population, there are at least 7 meta-analyses on this topic.Reference Klassen, Miller, Raina, Lee and Olsen16–Reference Storebo, Skoog and Damm22 One natural explanation for the lack of such adequate meta-analysis is the reduced number of trials. Although coming from a single trial, one encouraging finding in the field comes from a randomized clinical trial showing significantly reduced ADHD scores for adults using a combination of medication and CBT.Reference Safren, Sprich and Mimiaga45 However, it is important to bear in mind that Linderkamp and LauthReference Linderkamp and Lauth39 included both controlled and uncontrolled trials and identified only 12 studies in their meta-analysis. Nevertheless, the use of meta-analytic methods to evaluate the efficacy of psychosocial interventions is questionable because interventions in this field are markedly heterogeneous, making it difficult to interpret the results of pooled estimations.Reference Fabiano, Pelham and Coles20
Conclusion
Although the literature about child and adolescent ADHD is extensive, less data exist about adult ADHD. Very few meta-analyses have assessed pharmacological interventions for adults with ADHD, and we identified only one meta-analysis that evaluated psychosocial interventions. Results from meta-analyses suggest that stimulants are effective in decreasing ADHD symptoms on a short-term basis with a medium to large ES. Short-acting stimulants might be superior to long-acting stimulants, but no data on difference in adherence are available for the comparison of these two types of formulations. Bupropion is superior to placebo, but is less effective than stimulants. No pooled estimations are available for other drugs. No conclusions about the impact of psychosocial interventions can be drawn based on the meta-analytic data. There are no meta-analysis showing the effectiveness of some intervention whose effectiveness has been clearly demonstrated in clinical trials. This fact is concerning, and should be addressed in future research.
Disclosures
Tais S. Moriyama receives a CAPES PhD scholarship and has received continuous medical education support from Astra Zeneca, Eli-Lilly, and Janssen-Cilag. She has also served as a speaker for Ely-Lilly. Guilherme V. Polanczyk has served as a speaker and/or consultant to Eli-Lilly, Novartis, and Shire Pharmaceuticals; has developed educational material for Janssen-Cilag; and receives unrestricted research support from Novartis and from the National Council for Scientific and Technological Development (CNPq, Brazil). Fernanda S. Terzi and Kauy M. Faria have nothing to disclose. Luis A. Rohde was on the speakers’ bureau and/or acted as consultant for Eli-Lilly, Janssen-Cilag, Novartis, and Shire in the last 3 years. He also received travel awards (air tickets and hotel costs) from Novartis and Janssen-Cilag in 2010 for taking part in 2 child psychiatric meetings. The ADHD and Juvenile Bipolar Disorder Outpatient Programs chaired by LR received unrestricted educational and research support from the following pharmaceutical companies in the last 3 years: Abbott, Eli-Lilly, Janssen-Cilag, Novartis, and Shire.
