An observational study of response heterogeneity in children with attention deficit hyperactivity disorder following treatment switch to modified-release methylphenidate

For children with severe symptoms of attention deficit hyperactivity disorder (ADHD), methylphenidate (MPH) is one of the major treatment options. In randomized placebo-controlled trials, MPH has been shown to be effective and to be associated with, on average, large treatment effects [1, 2]. In addition to improving symptom control, MPH can have favourable effects on comorbid oppositional symptoms [3, 4] and social functioning [5, 6]. Side effects can occur and may require adaptation of the treatment plan [7]. Because of its efficacy and tolerability profile, MPH is recognized in treatment guidelines in Europe and the USA [8‐11].

The response achieved by drug treatment varies among children with ADHD; some children will experience adequate symptom control, whereas for others benefits will be minor [12]. There is currently little information on response prediction in ADHD, and detecting and explaining differential treatment effects should be considered an important part of the general treatment research agenda [13, 14].

Given the variation in responsiveness, it is reasonable to suggest that the population of patients treated with MPH is heterogeneous and may comprise distinct subgroups, which could be identified. For example, in therapy-naïve children receiving MPH for the first time, two subgroups might be hypothesized to exist: responders, showing good symptom reduction over time, and partial responders, who show few or no treatment effects. Subpopulations can be defined based on ‘observed’ variables that are directly measureable characteristics (e.g., comorbidity). For example, Ghuman et al. [15] found that comorbidities in preschool children with ADHD predicted response to MPH; the subgroup of children with three or more comorbid disorders did not respond to treatment compared with those with two or fewer comorbidities.

Subgroups may not always be identifiable by ‘a priori’-defined measureable characteristics and instead have to be inferred from the data [16, 17]. As in this case, group membership is not known beforehand, so these subgroups can be called ‘latent’ or ‘unobserved’ [18]. In clinical research, there is growing interest in the detection of unobserved subgroups associated with differential treatment effects [19]. However, in the treatment of ADHD, knowledge in this area remains very limited. In analyses by Sonuga-Barke et al. [20] of data from COMACS (Comparison of Methylphenidates in the Analog Classroom Setting) [21] and by Swanson et al. [22] of data from the MTA (Multimodal Treatment Study of Children with ADHD) [23], subgroups could be identified and subgroup membership could be linked to differential treatment effects. Owing to differences in methodology and study design, including varying observation periods of 12 hours (COMACS) and 3 years (MTA), direct comparisons of the results of these analyses are not meaningful. Nonetheless, the conclusion from both studies was that, for symptom change in children with ADHD, the overall population is heterogeneous and it is reasonable to look for subgroups with different treatment responses.

MPH is available in immediate-release and extended-release formulations [24]. For immediate-release preparations, the average duration of action is about 4 hours; extended-release formulations are designed to be effective for 8–12 hours [25, 26]. Several extended-release formulations, with different compositions and pharmacokinetic properties, are available [27, 28]. Meta-analytical comparisons point to similar effect sizes for both delivery forms of MPH [29]. As both immediate- and extended-release formulations have advantages and disadvantages, treatment guidelines often provide recommendations specifying which delivery format is appropriate given a patient’s individual circumstances [27].

Equasym XL® (Shire Pharmaceuticals Ireland Ltd) [30] is a modified-release MPH (MR-MPH) formulation comprising 30% immediate-release coated beads and 70% extended-release coated beads, delivering an initial rapid rise in plasma concentration and a slower rise over about 3 hours, with therapeutic plasma concentrations and efficacy in symptom reduction maintained for about 8 hours. The safety profile and efficacy of MR-MPH have been demonstrated in controlled clinical trials [31].

The OBSEER study (OBservation of Safety and Effectiveness of Equasym XL in Routine care), a 3-month, prospective, observational study involving children and adolescents with ADHD in Germany, was conducted under routine care conditions to gain insights into the effectiveness and safety of treatment with MR-MPH in everyday use. The study included children and adolescents diagnosed with ADHD for whom therapy with MR-MPH was already planned by the attending physician. Patients could have a treatment history with other MPH formulations, other drugs for ADHD or non-pharmacological interventions, or could be treatment naïve. In OBSEER, statistically significant treatment improvements in ADHD symptoms were seen from baseline to last visit [32]. The effectiveness of MR-MPH was rated better than prior therapy by both teachers and parents at all measured time points across the day, particularly late morning and early afternoon.

The aim of the present analysis was twofold. First, to identify subgroups with different treatment response trajectories, thereby extending previous work to attain a more complete picture of response heterogeneity during treatment [20, 22]. Second, to identify predictors of subgroups with differential responsiveness that might allow physicians to better select the appropriate treatment for an individual patient.

Treatment response to MPH is variable, supporting the contention that the population of children receiving treatment for ADHD is not homogeneous. This analysis sought to identify and determine the distinctive features of clusters of patients with different response trajectories to treatment with MR-MPH in parent and teacher ratings under real world conditions using data from the OBSEER study [32].

After switching medication, four response subgroups were detected: two with substantial treatment effects, both of which had more severe symptoms at study start, and two with little or no treatment effects, one of which had low initial symptom scores and one that had more severe symptoms at study start.

The subgroup with low initial symptom scores was the largest, comprising about half of all children. Here, only small treatment effects were observed, and for adequate interpretation of this finding, the nature of observational studies has to be considered. These children may already have reached good symptom control with the previous medication, leaving relatively little room for symptom improvement with transition to MR-MPH. In this study, the meaning of the term treatment response is therefore to understand the response following the change in treatment relative to previous medication effects and is no indication for absolute efficacy of MR-MPH. However, some symptom reduction was observed in this group. On a descriptive level, children in the low-start subgroup were older and had fewer conduct problems. Statistical tests confirmed that age and conduct problems were useful predictors to distinguish the low-start subgroup from the other groups, which concurs with previous research findings. Generally, it is accepted that ADHD symptoms decrease with age, although the finding of symptom reduction with age may reflect the developmental insensitivity of the DSM-IV, not the natural history of ADHD [48], and that children without comorbid conduct problems generally exhibit fewer ADHD symptoms than their counterparts with such comorbidities [49].

Approximately one quarter of the children fell into one of two subgroups with a strong response to MR-MPH. One feature distinguishing the two more responsive subgroups from the two more stable subgroups was the MPH dose at study start. On a descriptive level, children in the more-responsive groups were treated less intensively prior to study start; statistical analysis confirmed that lower MPH dose before medication switch was linked to a higher probability of falling into a more-responsive subgroup. Due to the fact that the premedication in the more-responsive subgroups was lower than that in the less-responsive subgroups, the relative increase in MPH dosage in these groups was the highest after the medication switch. These findings are in line with the assumption of a linear dose–response relationship between MPH treatment and ADHD symptoms, as has previously been demonstrated [50, 51]. However, more-responsive versus low-response subgroups could not be distinguished by the prescribed MR-MPH doses during the observation period.

One third of children were classified as ‘high-start, low-response’, with higher initial symptom scores and no or only modest treatment effects after switch to MR-MPH. These children represent a population that is difficult to treat. The average MPH dose was about 19 mg/day at study start and 26 mg/day at Visit 2 for this subgroup. For patients in this cluster, adjustment or change of medication or additional psychosocial interventions should be considered [52, 53].

Emotional symptoms were not useful for distinguishing between subgroups. Early work addressing the importance of anxiety on the effect of stimulant medication in ADHD indicated that behavioural effects are weaker in anxious children [54]. From this conclusion, for the current analysis, it would have been expected that more anxious children would be less likely to be in one of the more-responsive groups. However, more recent studies have not replicated these findings [55, 56], and Tannock [57] suggests that current evidence points to similar effects of stimulant medication in children with and without comorbid anxiety. The current analysis is in line with this evaluation.

Conduct problems proved to be useful for distinguishing subgroups with more-severe ADHD symptoms from the subgroup with less-severe symptoms at study start, but this variable was less suited to differentiating between subgroups with and without a good treatment response after switching medication. Connor et al. [3] suggest that the effects of pharmacotherapy are similar in children with ADHD, both with and without comorbid conduct problems, which corresponds with our findings.

In OBSEER, Döpfner et al. [58] reported that treatment-naïve as well as pretreated children demonstrated ADHD symptom reduction during the observational period of the study, and that the greatest reduction was seen in treatment-naïve patients. In the present analysis, these two groups were not explicitly distinguished, but it was found that less-intensively pretreated children were more likely to be members of one of the two response subgroups, which is in line with the previous analysis.

The advancement of evidence-based medicine has had a major impact on the healthcare system and intervention research [59, 60]. High quality studies are required to demonstrate the efficacy of a particular treatment [61]. The quality of evidence from clinical studies varies depending on the study design and, in general, evidence from randomized, controlled trials is considered to be of high quality and that from observational studies is sometimes considered to be of low quality [62]. Nevertheless, both research designs have strengths and limitations, and instead of regarding them in a strictly hierarchical fashion, with randomized, controlled trials as the gold standard, a different perspective is to view them as complementary research designs [63, 64]. Randomized, controlled trials are high on internal validity but there may be a lack of generalizability of their findings. For example, in the MTA, the largest clinical trial in ADHD to date, children with irregular school visits were not eligible for participation and the generalizability of study findings for these children is unclear [23]. While the possibility for drawing causal inferences about the reasons for symptom reduction is limited in observational trials, they can provide valuable information regarding whether a treatment is likely to be effective under real world conditions, across different patient types and settings.

Because this was an observational study it has specific strengths and limitations. The pre-study medication history of the children evaluated in the study was varied, with the majority of children switching from other MPH formulations at a range of doses, and there was no baseline medication/dose stabilization phase. Furthermore, administration of MPH did not follow a standardized treatment protocol and treatment plans were individualized. Differences in symptomatology, previous medication and actual medication are therefore confounded and interpretation of treatment effects is complicated. Low internal validity is one of the major drawbacks of observational studies. However, this analysis identifies clusters of patients that can be observed in common medical practice. Information about a patient’s initial situation and ADHD treatment in real world settings, including treatment regimes and treatment responses, is valuable for clinical practice as well as research. For the practitioner, knowledge about the typical subgroups of patients in routine clinical practice may be useful for treatment individualization. Furthermore, such studies are useful to show the need for treatment optimization.

There were several limitations to this study. First, this was an observational study with no (placebo) control group and parents were not blinded to the study treatment or dose, which may have influenced parent-ratings of ADHD symptoms. The use of a placebo control group would have allowed us to disentangle actual perceived improvements from bias based on hope or expectations of the parents. Yet, teachers were not formally informed of the change of treatment, and although they could have been told by parents or children, most were not aware. Second, the results for the previously treated group can only be generalized to a population in which a switch to MR-MPH is planned due to suboptimal efficacy with prior medication. Third, in particular for teacher ratings of ADHD symptoms, there was a substantial proportion of missing data. For this analysis, an inclusive missing data handling strategy was used and all available data were taken into account. Even when the underlying assumptions of such strategies are not met, they are still considered to be an improvement over ad hoc procedures like listwise deletion [65]. Advantages of this study are the relatively long period of observation, the large sample size and the use of multiple informants.

Together with those by Sonuga-Barke et al. [20] and Swanson et al. [22], this is one of the few studies that has investigated the presence of latent subgroups in patients with ADHD with varying treatment responses after medication with MPH. All three studies indicate that there is heterogeneity in treatment response. The total population of children with ADHD – either previously treated or treatment naïve – receiving modified-release MPH under routine care conditions is best conceived as a composite of distinct subpopulations with varying treatment responses.

The current analysis revealed four subgroups after switching medication. Distinctive features of subgroups could be identified. Age, MPH dose at study start and conduct problems were particularly useful to discriminate between clusters of patients. Older children and children with few conduct problems were more likely to be members of a patient cluster with few symptoms at study start. Children with a low MPH dose in the beginning had a higher chance of being in the patient cluster with a strong treatment response. These results were in line with expectations. More research is needed to replicate these findings and to explore additional predictors to achieve a more complete picture of the differentiating characteristics of responsiveness subgroups. The current analyses should assist in identifying children likely to achieve a favourable treatment course with MPH and, additionally, those who are in need of alternative treatment options.

Christopher Hautmann is a clinical psychologist and psychotherapist. His main research interests involve externalizing disorders, treatment evaluation and the application of advanced statistical methods in clinical research. Aribert Rothenberger is Head and Chair of Child and Adolescent Psychiatry at University Medicine Göttingen, Germany. Manfred Döpfner is Professor in Child and Adolescent Psychiatry and Head of the School of Child and Adolescent Psychotherapy at the University Hospital, Cologne, Germany.