The effectiveness of cognitive behavioral therapy (CBT) for childhood obsessive compulsive disorder (OCD) has been well established [1
]. However, average symptom reduction is limited and almost half of the patients still have considerable complaints after standard treatment [6
]. The way CBT is applied may not always be optimally effective [9
CBT for OCD intends to change behavior (compulsions) and cognitions (obsessions). This is mostly done by a combination of exposure plus response prevention (ERP), and cognitive therapy (CT) [1
]. Despite their common basis in the learning theory, distinct mechanisms leading to symptom reduction are assumed in ERP and in CT. From a behavioral perspective it is assumed that a behavioral change through exposure to feared situations is the primary process leading to essential corrective learning experiences. For this reason, behaviorists advocate ERP as the core of treatment [11
]. In the cognitive model, it is assumed that the optimal way to corrective learning is through explicitly reframing dysfunctional beliefs. Therefore, in CT cognitive restructuring is advocated as the core of treatment [14
In line with the cognitive model, a relation between dysfunctional beliefs and obsessive–compulsive (OC) symptoms has been found in several child studies [16
]. However, other studies yielded equivocal evidence for the cognitive model in childhood OCD [26
]. In two studies it has been tested if inflated responsibility beliefs affected OC symptoms in children based on an experimental design. These studies yielded mixed results [24
]. Taken together, despite a well-developed theoretical base for a key role of dysfunctional beliefs in (childhood) OCD, the evidence is equivocal.
Nevertheless, cognitive models have strongly influenced treatment for OCD. Almost all treatment packages for childhood OCD contain some type of cognitive interventions [1
]. Full cognitive treatment protocols are developed even for children [32
], and in combined treatment packages it is not unusual to start with a focus on cognitive interventions, followed by ERP. Efficacy of behavioral as well as cognitive oriented treatment protocols have been demonstrated [1
]. However, a favorable effect of ERP over CT has been reported [34
], and the addition of cognitive interventions to ERP did not always result in more effective treatment packages [9
]. Overall, critiques on cognitive models are arising [35
At present, the core of CBT is still unknown. Is explicitly changing cognitions the main mechanism through which symptom reduction is achieved, or would it be more effective to primarily focus on ERP?
A better understanding of the mechanisms of change in CBT may contribute to better treatment [1
]. Identification of these mechanisms can help to improve treatment because effective treatment components can be added or strengthened, and ineffective components can be removed [36
]. However, mechanisms of change are rarely studied in treatments for childhood OCD.
Studying change mechanisms in treatment requires a formal test of statistical mediation. Mediation refers to the process through which change occurs. A change in the proposed mediator variable should precede a change in outcome. Consequently, a repeated measurements design is needed to demonstrate mediation [37
]. Unfortunately, most treatment studies traditionally rely on pre-post test designs. An exception is a small pilot study [38
]. In this study it was examined if changes in dysfunctional cognitions were associated with treatment effect in CT for pediatric OCD. Six adolescents with OCD (12–17 years, M
= 14.3) reported responsibility beliefs and OCD symptoms at every treatment session. It was found that decreases in responsibility beliefs were associated with decreases in OCD symptoms, but the direction of this relation remained unclear, leaving the question of mediation unanswered [38
Repeated measurements designs have been used slightly more often in adult OCD studies, with equivocal results. Whereas the findings of some studies provided some support for the cognitive model [39
], findings of other studies did not [34
], or were inconclusive [45
]. Most studies are based on small samples, and differences across studies in design and statistical analyses make it hard to draw clear conclusions. A preliminary observation is that most studies, including those with the largest sample sizes and based on thorough statistical analyses [34
], did not support the cognitive mediation hypothesis.
Given the mixed evidence with regard to the potential role of dysfunctional beliefs in the treatment of childhood OCD, the aim of the present study was to further examine if changing dysfunctional cognitions is a mediator of treatment outcome in CBT for children and adolescents with OCD. It was examined whether changes in dysfunctional beliefs preceded changes in OCD severity, whether they were a consequence of changes in OCD severity, or whether this relation was bidirectional. Based on cognitive models it was our hypothesis that cognitive changes precede changes in OCD severity.
The aim of the present study was to examine if changing dysfunctional beliefs was a mediator of CBT for pediatric OCD. Fifty-eight children with OCD (8–18 years old) received sixteen weekly sessions of CBT consisting of ERP and CT. Dysfunctional beliefs and OCD severity were assessed pre-treatment, mid-treatment, post-treatment, and at 16-week follow-up. According to cognitive models, we expected cognitive changes to precede changes in OCD severity. Contrary to this hypothesis, results showed that changes in OCD severity statistically predicted changes in dysfunctional beliefs rather than the reverse. In other words, changes in severity explained changes in dysfunctional beliefs within time intervals. More specifically, changes in OCD severity explained 43% of the change in beliefs at mid-treatment, 36% between mid- and post-treatment, and 42% between post-treatment and follow-up. It is important to note that the present results do not allow for conclusions about causality, because we did not find a relation between cognitions and OCD severity over time (across assessment points). Therefore, we cannot determine whether a decrease in dysfunctional beliefs actually was an effect of a decrease in OCD severity. Nevertheless, the present findings cast doubt on the assumption of cognitive models suggesting that changing beliefs plays a key role in the treatment of OCD.
In line with the present results, several previous studies evaluating mediating mechanisms in psychological treatment for OCD were not supportive for cognitive models [34
]. However, this field of research is still in its infancy and is hampered by the limited amount of studies, the main focus on adult samples, and methodological limitations and differences across studies. There are some other issues that merit discussion.
Studying cognitive mediation is a challenging task. One of the biggest challenges is to measure actual thought processes. Cognitive processes may rely on conscious thoughts as well as on unconscious, automatic thoughts. Explicit measures, either standardized or idiographic, can only provide indications of beliefs that are accessible for personal introspection. In addition to explicit measures, implicit paradigms may be needed to shed light on the role of unconscious, automatic thoughts [64
]. Although still in its early stages, several cognitive bias paradigms have been developed for this purpose [66
]. It would be interesting to use such paradigms in future studies. It could be for example, that contrary to the explicit approach in cognitive therapy, ERP addresses implicit cognitive processes. Implicit paradigms may be more suitable to detect these mechanisms.
Another challenge is that mediation requires a relation between the mediating variable and the outcome variable over time [37
]. One could wonder if—even in case of established cognitive mediation—the mediating processes can be disentangled over time. One may assume that as soon as thought processes change, for example the patient does not overestimate the importance of an intrusion anymore, no raise in anxiety will appear, and consequently there may be no urge to perform compulsions. These processes may occur in acute response to each other, making it impossible to observe a temporal lag between a change in cognitive processes and in OC symptoms [67
]. Following this line of reasoning, it may be complicated to get a grip on these fast and dynamic processes.
A third challenge is that mediating processes may differ across individuals. This hypothesis is supported by a study in adult OCD patients on the role of dysfunctional beliefs in CBT [42
]. Indeed, there is some evidence that not all OCD patients experience more dysfunctional beliefs than non-clinical individuals [68
], and cognitive mediation may not be expected for these patients. In addition, our findings showed a wide range of OBQ-CV scores, indicating that there are substantial individual differences in dysfunctional beliefs, and consequently there may be differences in cognitive change processes during treatment.
Although the present study has several strengths such as a longitudinal design with a mid-treatment assessment, a representative sample of youth with OCD, and the use of a treatment protocol that has already been implemented in clinical practice, the study has some limitations too. First, the present sample size did not allow for adding extra variables to the models, and therefore we could not control for effects of possible moderating variables such as OCD subtype or developmental level. However, a previous study showed no effect of age on OBQ-CV score in a clinical OCD sample (mainly the same sample as the present study) [28
]. This finding makes it less likely that results would have been different when age was included. Furthermore, the present treatment protocol allowed for tailoring cognitive interventions to the individual to preclude the risk that these interventions would have been too difficult to understand for younger children, or too childish for older children. This way, we aimed to provide a treatment that suited all participants. Second, due to the time interval between assessment points we may have been unable to detect change processes that have occurred in-between assessment points. Third, reports of dysfunctional beliefs were solely based on a standardized questionnaire. The OBQ-CV was selected because adequate reliability and validity had been demonstrated in pediatric OCD and in a youth community sample [28
]. Furthermore, the OBQ-CV includes multiple dysfunctional belief domains assumed to be relevant in OCD, instead of a limited selection of domains [71
]. However, despite these strong features of the OBQ-CV, the incorporation of multiple belief domains also entails a disadvantage. Participants who frequently experience beliefs in one domain, may have a relatively low OBQ-CV total score despite the high frequency of particular beliefs. In these cases only small changes can be found, and even if dysfunctional beliefs change during treatment it would be hard to demonstrate mediation effects over time. We have considered to perform analyses using OBQ subscales instead of a general total scale. However, besides that this would have added extra parameters to the model, a study on psychometric properties of the OBQ-CV showed high correlations among subscales, and a single higher-order factor (OBQ total score) that explained the correlations between the subscales quite well [28
]. For these reasons, we did not differentiate between subscales. Another complicating factor of using explicit measures is that a certain level of insight in thought processes is required for a valid report of dysfunctional beliefs. This may be demanding for all individuals, and especially for children as meta-cognitive skills are not fully developed at this age. Finally, the main part of the sample was of a Western background. This may limit the generalizability of the results to samples of other cultural backgrounds.
Notwithstanding these limitations, some clinical implications and recommendations for future research can be derived from the present study. Based on the present results, we conclude that restructuring explicit cognitions may not be a necessary component in the treatment of OCD in children and adolescents. At least not for all patients with OCD. For future studies it would be interesting to shed more light on potential moderating variables, for example OCD subtype and children with and without obsessions (tic-related OCD), in combination with mediating variables. Furthermore, idiosyncratic and implicit measures could be combined with standardized, explicit measures of dysfunctional beliefs to examine the role of both explicit and implicit cognitive processes in CBT.
For therapists it would be interesting to know if treatment for pediatric OCD could be confined to solely ERP. The present results do not allow for such a conclusion. First of all, our results need to be replicated. Second, this conclusion would assume a single, one-to-one relationship between CT and a change in dysfunctional beliefs. This seems no realistic representation of matters. We cannot equate CT with cognitive change, and ERP with behavioral change. At present, the actual mechanisms of change of CT and ERP are inadequately understood. Different explicit and implicit processes may be active in CT and ERP, and there may be different ways to achieve treatment effect. Furthermore, in many cases elements of exposure are incorporated in CT in the form of behavioral experiments, and ERP is often complemented by cognitive interventions. The latter is explicitly prescribed in the inhibitory learning model of ERP where expectations about what may happen following exposure are challenged by the therapist [12
]. Challenging cognitions is not specifically prescribed in the habituation model of ERP, but this intervention can be used to support ERP [13
Future studies that aim to evaluate mediating (cognitive) mechanisms in treatment for OCD, should include more in-treatment assessment points. To gain more insight in change processes of dysfunctional beliefs and OC symptoms, these constructs could be measured each session or every day during treatment. Case-based time-series designs can then be used to closely follow changes over time within individuals [42
], before resource demanding, large randomized controlled trials are conducted. Although seldom used, case-based time series designs are recognized as fair methodological approaches for treatment studies [73
], and can provide information about what interventions work for whom. The latter may be especially important in the light of the large individual differences in treatment effects for pediatric OCD, and because mediating mechanisms may differ across patients.