A systematic review and analysis of long-term outcomes in attention deficit hyperactivity disorder: effects of treatment and non-treatment

Based on the potential for long-term persistence of ADHD into adulthood and equivocal reports on treatment, we initiated this evidence-based systematic review to understand the long-term outcomes of ADHD with emphasis on a comprehensive synthesis of published data. Given the large differences in study design and measurements used, we decided that using a binary feature common to all the studies ('significantly different' and 'not significantly different') to define outcome results for LTOs would allow us to summarize all the included studies in a non-statistical fashion. Binary variables are used to simplify data in clinical trials for ADHD quite frequently (for example, Clinical Global Impression-Improvement (CGI-I), ADHD Rating Scale (ADHD-RS) responder analyses). 'Long term' was defined as 2 years or more and 'outcomes' were defined as life consequences, distinct from symptoms.

Our analysis sought to answer the research question: what are the long-term outcomes in participants with ADHD compared to baseline or controls and do long-term outcomes of ADHD improve with treatment (including pharmacological, non-pharmacological and multimodal)?

Our search method has been described in a previous publication [35]. To identify as many published studies as possible, 12 databases were searched: Academic Search Premier, CINAHL, Cochrane CRCT (including EMBASE), Criminal Justice Abstracts, ERIC, MEDLINE, Military & Government collection, NHS Economic Evaluation database, PsycARTICLES, PsycINFO, SocINDEX, and Teacher Reference Center. MEDLINE was searched using two different search engines. Duplicates were eliminated electronically and manually, yielding 5,467 studies.

Based primarily on title and abstract, these studies were reviewed manually and inclusion was agreed on by two researchers. This yielded 351 studies for inclusion in the analysis. A list of all studies included in the analysis has been published [35].

All disagreements between researchers on study inclusion were resolved by examining the full text of the study. Studies included participants who were diagnosed with ADHD or symptomatic presentation of ADHD as reported by the authors of each study. Only those studies in which ADHD was the primary disorder under study were included. Studies included both naturalistic examination of ADHD course (vs non-ADHD controls or start-of-study baseline) and/or treated ADHD (vs ADHD natural course, pretreatment baseline, or non-ADHD controls). Treatments included pharmacological, non-pharmacological, and/or multimodal treatment.

Data from each study was manually extracted from the full text of the study to a database, including: (1) study location, (2) study sample size, (3) study length, (4) participant age range, (5) study support, (6) diagnostic criteria, (7) study type (longitudinal, cross-sectional, prospective, retrospective), (8) outcome measures, (9) outcome results, (10) comparator type, (11) treatment type, and (12) treatment duration. Outcome results were measured by dichotomizing all results into a binary variable of either 'poorer' or 'similar' outcomes, and with respect to outcomes with treatment, either 'benefit' or 'no benefit' with treatment. Many studies (44%) reported more than one outcome result, thus the number of outcome results is more than the number of studies. While the outcomes may be the item of interest, the number of studies from which these outcomes are derived is also informative and so is also reported.

Outcome measures were compiled into nine major groups based on commonality of outcome characteristics (Figure 1). This grouping of outcomes has also been described in a prior publication [35]. These groups included: (1) non-medicinal drug use/addictive behavior (for example, use, abuse, and dependence on alcohol, cigarettes, marijuana, stimulants, or illicit drugs; age at first use; multiple substance use; gambling), (2) academic (for example, achievement test scores, grade point average, repeated grades, years of schooling, degrees earned), (3) antisocial behavior (for example, school expulsion, delinquency, self-reported crimes, arrests, detainment, incarceration, repeat convictions), (4) social function (for example, relationships, peer nomination scores, marital status, multiple divorces, activities, hobbies), (5) occupation (for example, employment, military service, job changes, occupation level, socioeconomic status), (6) self-esteem (for example, self-esteem scales, self-perception, suicide ideation, suicide attempts, suicide rate) (7) driving (for example, accidents, traffic violations, license suspensions, driving record), (8) services use (for example, justice system, emergency health care, financial assistance), and (9) obesity (body mass index (BMI), weight).

The number of outcome results in each outcome group can be seen in Figure 1. Drug use/addictive behavior was the most-studied outcome (160 outcome results), followed by academic (119 outcome results), antisocial behavior (104 outcome results), social function (98 outcome results), occupation (45 outcome results) self-esteem (44 outcome results), driving (30 outcome results), and services use (26 outcome results). Obesity was the least studied outcome (ten outcome results). Note that the total number of outcomes results (636) is greater than the total number of studies (351) because some studies reported more than a single outcome result.

Figure 2 shows the total number of studies published per year. There was a noticeable rise in studies of long-term outcomes of ADHD published worldwide between 1980 and 2008. The number of long-term outcome studies published at the peak in 2008 was 42 studies, dropping back to 28 in 2009 and 2010. The mean study length varied little by year, with a total mean of 9 years for which researchers collected data for each subject and range of 2 to 40 years for which researchers collected data for each subject. Data collection refers either to follow-up measures in the case of prospective studies or analysis of past records or reports in the case of retrospective studies.

The ages of the participants were examined by studies of specific outcomes (Figure 3A). Studies of children with a mid-range or mean age of 6 to 12 years measured services use, self-esteem, social function, academic outcomes, obesity, antisocial behavior, and drug use/addictive behavior, in that order of frequency. All nine outcome groups were measured in adults and adolescents. Within age categories (Figure 3B), social function and academic outcomes comprised the largest proportion of children outcomes (53%), while drug use/addictive behavior and antisocial behavior comprised the largest proportion of adult and adolescent outcomes (43 and 46%, respectively). A substantial proportion of outcomes in children and adolescents together were self-esteem and social function outcomes (28%).

Poorer outcomes were generally observed in untreated participants with ADHD (Figure 4). In all, 89 studies showed that people with untreated ADHD had outcomes not substantially different from controls (26% of outcome results), whereas 244 studies showed that untreated participants with ADHD experienced poorer long-term outcomes (74% of outcome results). Note that more outcomes were observed than studies because some studies reported more than one outcome. There were a few studies (6) that reported outcomes (6) for participants with untreated ADHD that were significantly better than non-ADHD controls. The derived or reported effect sizes were not large, and these few outcomes were included in the outcomes that were 'similar' to controls. No single outcome group was represented; the outcomes varied among drug use/addictive behavior, occupation, self-esteem, and social function outcomes.

Also found in the search were five studies that followed untreated participants with ADHD over 2 to 9 years and compared the long-term outcomes with the participants' status at baseline [42‐46]. Four of these studies reported a significant deterioration from baseline without treatment [42‐45]. Of these four, three reported a deterioration in academic outcomes (increased number of failing grades over 2 years and decline in math and reading scores over 9 years) [42‐44] and one [45] reported a worsening in tobacco use outcomes (levels of salivary cotinine measured daily rose in untreated participants with ADHD over 2 years. Cotinine is an alkaloid from tobacco and a metabolite from nicotine, used as a measure of the number of tobacco cigarettes smoked per day). One study followed untreated participants with ADHD for 8 years into adolescence and reported an improvement in social function compared with the participant's baseline measured at the beginning of the study, although function remained significantly poorer than non-ADHD control levels [46].

Many studies did not report effect size, and effect sizes varied among studies that did report it. For example, a study of driving outcomes reported a small effect size with a Cohen's d of 0.33 (P = 0.04) for the difference in the number of traffic accidents in the last 6 months for participants with ADHD (0.29 ± 0.73 accidents) compared with non-ADHD controls (0.15 ± 0.43 accidents), albeit this small effect size represented almost twice the rate of accidents for the ADHD group [47]. Another study of social function reported a large effect size with a Cohen's d of 1.03 (P < 0.001) for the difference in the parent-reported peer rejection scores for children with ADHD (0.45 ± 0.55) compared with non-ADHD controls (0.07 ± 0.23), while controlling for conduct disorder as a comorbidity [48]. It is possible in studies of smaller sample size that there may have been small effects that were not reported as differences, because statistical significance was not demonstrable due to the small sample size. For the purposes of the present analysis, to provide an overall comprehensive synthesis of reported study results, all results were analyzed as reported without additional interpretive changes on our part. Inclusion in this regard was limited by the report having passed through the peer-review process.

Treated ADHD versus untreated ADHD was compared in 48 studies with 76 outcomes (Figure 5). 'Untreated ADHD' comparators included both pretreatment baseline comparisons and comparison with an untreated group of participants with ADHD. Overall, treatment of ADHD resulted in favorable outcomes for most outcomes reported (55 of 76 outcome results; 72%). Three types of outcome results for which treatment was considered beneficial were: (1) improvement compared with participants with untreated ADHD (38% outcome results), (2) improvement compared with pretreatment baseline (22% outcome results), and (3) stabilization compared with pretreatment baseline (12% outcome results). Stabilization compared to pretreatment baseline was considered a benefit of treatment because it indicated that treatment may have alleviated the natural-course deterioration in outcomes over time that has been observed in separate study samples of untreated ADHD [42, 43, 45, 46]. None of the studies reporting stabilization compared with pretreatment baseline also included an untreated ADHD group, therefore this is an across-study sample comparison, subject to limitations. Three types of outcome results for which there was considered no benefit with treatment were: (1) no difference compared with participants with untreated ADHD (25% outcome results), (2) poorer outcomes compared with participants with untreated ADHD (1.5% outcomes results), and (3) poorer outcomes compared with pretreatment baseline (1.5% outcomes results). In outcomes for which there was no difference compared with participants with untreated ADHD, while there was no benefit to treatment, there was also no detriment either, such as increased incidence of substance use disorder or increased rate of suicide. No significantly increased incidence of substance abuse disorders or suicide rate compared with participants with untreated ADHD (or compared with pretreatment baseline either) was reported in any study included in this analysis. There was a single outcome following treatment that was worse for participants with untreated ADHD (increased experimentation with cocaine). A single outcome was reported to be worse than pretreatment baseline (greater percentage of study participants with decreased grade point average).

In 42 studies, the outcome results (n = 76) of participants with treated ADHD were compared with the outcomes of non-ADHD controls. Again, more outcomes were observed than studies because some studies reported more than one outcome. Most such studies did not show normalization with treatment. Only 18 outcomes in 16 studies were similar for participants with treated ADHD versus non-ADHD controls. A total of 58 outcomes (76% of outcomes) in 35 studies, poorer outcomes were observed for participants with treated ADHD relative to non-ADHD controls.

Benefit with treatment was analyzed according to specific outcome group for participants with treated versus untreated ADHD (Figure 6). Note that this analysis involved the same 48 studies and 76 outcomes in the analysis shown in Figure 5, with the same 3 conditions considered as beneficial with treatment and the same 3 considered to exhibit no benefit with treatment. For 100% of driving and obesity outcomes reported, treatment of ADHD was beneficial. For 90% of self-esteem outcomes, 83% of social function outcomes, 71% of academic outcomes, 67% of drug use/addictive behavior outcomes, 50% of antisocial behavior outcomes, 50% of services use outcomes and 33% of occupation outcomes, treatment was reported to be beneficial. In the case of services use, less use of services (for example, emergency room visits, financial assistance) was considered to be an improvement with treatment.

Four of the nine outcome groups exhibited a substantial percentage of no benefit with treatment. These four outcome groups were drug use/addictive behavior, antisocial behavior, services use, and occupation. A subanalysis of these four outcomes examined the post-hoc hypothesis that less aggressive/consistent treatment of ADHD in the rest of the world compared with Northern America (based in part on stricter diagnosis criteria for the International Classification of Diseases, tenth edition (ICD-10) versus the DSM-IV-TR [49, 50]) may account for the rate of reported treatment benefit observed in these four outcome groups in Figure 6. Thus, treatment outcome by region for this subgroup of outcomes was examined (Figure 7). For these 4 outcome groups, studies performed in Northern America were evenly split in reporting outcome results exhibiting treatment benefit versus no benefit (11 outcome results each, 50% each). In contrast, studies performed in countries in the rest of the world (all from Europe, in this case) reported a higher percentage of outcome results exhibiting treatment benefit (six of seven outcome results; 86%, four of which were improvement in drug use/addictive behavior outcomes) versus no benefit with treatment (one of seven outcome results; 14%) (Figure 7). This result clearly shows that reported results of treatment in the rest of the world do not underlie the higher percentage of outcome results exhibiting no benefit with treatment for these four outcomes. On the contrary, studies from the rest of the world reported a larger percentage of outcomes exhibiting treatment benefits.

Further analysis of the results observed in Figure 7 showed that this result did not appear to be associated with types of treatment, comparator groups, or outcome groups of interest in Europe compared with Northern America [35]. A greater percentage, however, of these outcome results from the rest of the world were reported in retrospective studies of adults (3 of 4 studies; 75%) compared with those from Northern America (2 of 19 studies; 11%). The majority of these outcomes from Northern America were reported in prospective studies of various age groups (15 of 19 studies; 79%). Thus, differences in study design may underlie the result shown in Figure 7. The various treatment types mentioned in all the studies are listed in Additional File 2. Of the 130 studies that mention treatment, 120 (92%), 49 (38%), and 24 (18%) studies mentioned pharmacological treatment, non-pharmacological treatment, and/or multimodal treatment, respectively.

Several possible risks of bias and limitations need to be considered regarding the included studies. First, a publication and cultural bias could have resulted from including only studies that were published in English. In addition, the analysis excluded unpublished studies that might have been presented at conferences, for example. Also, our search relied on search engines for 'peer reviewed' status. Moreover, by strictly adhering to Cochrane systematic review guidelines and only including studies that were identified in our original electronic search, it is possible that some relevant studies may have been missed, introducing a search engine and literature database bias. This bias was reduced by extensive searching of 12 databases. Nonetheless, we are aware of four studies that would have met inclusion criteria, but were not identified by the search engines due to a technical limitation or inadvertent search string exclusion. Examination of these studies shows that the reported results are consistent with the overall results of the present analysis. The results of these four studies are summarized briefly here. One study reported poorer outcomes for participants with ADHD versus non-ADHD controls with regard to academic achievement, occupational adjustment, antisocial behavior, relationships, and substance use [65]. A second study found a high incidence of ADHD (65%) in 23 adolescents who attempted suicide [66]. A third study reported that by young adulthood, participants with ADHD were similar to non-ADHD controls in minor aspects of social and occupational outcomes (e.g., time socializing with friends and with hobbies), but had poorer outcomes in major aspects of these outcomes (e.g., had many more offspring and most were not living with them) [67]. A fourth study found that stimulant treatment in children with ADHD significantly improved reading scores and decreased grade retention [68]. The results of these studies are consistent with our overall finding that untreated ADHD is associated with poor long-term outcomes and that these outcomes improve with treatment.

Researcher bias could also be a possible source of bias in this analysis, however, this was reduced by having two researchers independently agree on the articles included and strict, simple inclusion criteria were established prior to searching.

Other sources of bias could include biases of individual research groups, which was eliminated by including only electronically identified studies (as mentioned above) and not selectively including the studies of specific groups and omitting others. As observed, different study designs may also lead to different conclusions and taking comparators into consideration is critical. We included studies of various designs, which may minimize such bias.

A further possible bias could arise from changes over time in diagnostic criteria or discrepancy between classification systems, specifically differences in the definition of hyperkinetic disorder (ICD-9 or ICD-10) versus ADHD (DSM-III/DSM-III-R/DSM-IV). Differences are less likely within classification systems [3]. One study found that 93% of children diagnosed with ADHD using DSM-III-R diagnosis also received a DSM-IV ADHD diagnosis, indicating good correspondence between classification systems [69]. Rediagnosis of the MTA sample by ICD criteria, however, resulted in only 25% of the DSM-IV-diagnosed MTA sample of combined-type ADHD qualifying as having hyperkinetic disorder or hyperkinetic conduct disorder by ICD-10 criteria [69].