Exploring the minimal important difference in the treatment of paediatric obsessive-compulsive disorder using selective serotonin reuptake inhibitors

Background

The Children’s Yale-Brown Obsessive-Compulsive Scale (CYBOCS) is a clinician rated instrument developed to quantify severity of obsessive-compulsive disorder (OCD) in children and adolescents.1 Although the CYBOCS is the gold standard for outcome in clinical research and practice, the minimal required symptom reduction that constitutes the threshold for clinical improvement is unknown. One way of defining the ‘minimal important difference’ (MID) is by comparing the CYBOCS with the Clinical Global Impression-Improvement (CGI-I) scale, where a CGI of 3 (‘minimally improved’) is used as the MID. Equipercentile linking can be employed to establish corresponding outcomes between the CYBOCS and the CGI-I scale. The method has been used to link CGI with psychometric scales of major depressive disorder (MDD), anxiety, bipolar disorder and schizophrenia.2 In this study, we linked the CYBOCS to the CGI-I in order to find the MID for paediatric OCD.

Study selection and analysis

We used individual participant data from short-term double-blind, randomised, placebo-controlled trials with selective serotonin reuptake inhibitors (SSRIs) that were submitted to the Dutch Medicines Evaluation Board (MEB). All trials were approved by the original ethical commission and participants and caregivers signed informed consent. Sponsors shared anonymised data with the MEB under the restriction that we could not cite the original trial or release the name of the used compound. We included all available studies on children and adolescents that used both CGI-I and CYBOCS. Our primary outcome was the change in CYBOCS scores compared with baseline that was linked to a CGI-I of 3. As clinical relevance of symptom reduction might depend on baseline severity, we conducted a separate analysis for CGI-I and percentage CYBOCS reduction. Additionally, we performed a post hoc analysis calculating the MID for patients with moderate-to-severe OCD (CYBOCS ≥22) and with mild OCD (CYBOCS <22), using established severity benchmarks.3

We pooled included studies and used CGI-I and CYBOCS for weeks 2, 4, 6, 8, 10 and 12. For each week, we employed list-wise deletion to exclude participants with missing outcome data. We assessed the correlation of outcomes, predefining sufficient correlation as a Spearman’s rank correlation coefficient of 0.5 or above. If these conditions were met, we used equipercentile linking with log-linear smoothing, applying a bootstrap approach which allowed for generation of 95% CIs. We then presented a single average linking score, representing the collective outcomes across time points. Analyses were conducted using R V.4.2, employing the equate package.4 We preregistered our analysis plan on the Open Science Forum (doi https://doi.org/10.17605/OSF.IO/JEM7V).

Findings

Out of four available paediatric OCD trials, two used CGI-I along with CYBOCS measures (pooled n=309). 16 patients were not allocated to a treatment group and were excluded from analysis. We found a retention of 275 patients at week 10 (94%) after which one study stopped follow-up, leaving 155 patients at week 12. Baseline patient characteristics did not differ between the SSRI and placebo group. See table 1 for a basic overview of patient characteristics at baseline.

CGI-I and CYBOCS change scores were sufficiently correlated with a mean Spearman’s rank correlation of 0.70. Equipercentile linking scores were similar for each week (for visual inspection see figure 1). Results showed that a CGI-I score of 3 (‘minimally improved’) was linked to a 5.1-point CYBOCS change compared with baseline (95% CI 3.9 to 5.1) and a 19% change (see table 2). A CGI-I of 3 was linked to a 6.4-point CYBOCS reduction (8.2–4.6) in patients with moderate-to-severe OCD, and to a reduction of 3.3 (5.2–1.4) in patients with mild OCD. Omitting week 12, which comprised one study, did not alter our results.

Figure 1

Equipercentile linking of CGI-I and change in CYBOCS score, per week. The black line is the average linking score across all weeks. CGI-I, Clinical Global Impression-Improvement; CYBOCS, Children’s Yale-Brown Obsessive-Compulsive Scale.

Conclusions and clinical implications

This is the first study to examine the MID for children and adolescents with OCD. Using equipercentile linking, we identified that a reduction of 5.1 points on the CYBOCS is required for clinicians to observe minimal improvement. We further showed that the MID is larger in patients with a higher baseline symptom score.

The only other study using equipercentile linking to calculate an MID for psychiatric treatment in children and adolescents is a recently published manuscript on SSRIs for paediatric MDD.5 This study revealed an MID (14 points on the Children’s Depression Rating Scale-Revised) that was nearly double the observed treatment difference between active medication and placebo. In a recent meta-analysis assessing the efficacy of SSRIs compared with placebo in paediatric OCD, a small standardised mean difference was demonstrated, yet mean CYBOCS differences were not reported which impedes interpretability.6 For assessing clinical relevance, it would be valuable to determine whether the placebo-active treatment separation equals the MID in youth with OCD.

A shortcoming of our study was the inclusion of only two trials which were not specifically designed to examine our research question. While yielding an adequate sample size for linking analysis in a well-defined and relatively homogeneous trial population, this restricts generalisability. Finally, because we were unable to use patient-reported outcome measures, and because the CGI-I focusses on illness-specific symptoms, our MID is assessor-based and does not incorporate non-illness-specific factors such as quality of life.

Although we believe our findings help interpretation of results from previous paediatric clinical OCD trials and can guide future research, they also emphasise the need for replication in larger cohorts and across various treatments, including pharmacotherapy and psychotherapy. These studies should ideally also incorporate patient-reported outcomes to inform shared decision-making in paediatric OCD management.