Improving pediatric multiple sclerosis interventional phase III study design: a meta-analysis

Abstract

Background: To support innovative trial designs in a regulatory setting for pediatric-onset multiple sclerosis (MS), the study aimed to perform a systematic literature review and meta-analysis of relapse rates with interferon β (IFN β), fingolimod, and natalizumab and thereby demonstrate potential benefits of Bayesian and non-inferiority designs in this population. Methods: We conducted a literature search in MEDLINE and EMBASE from inception until 17 June 2020 of all studies reporting annualized relapse rates (ARR) in IFN β-, fingolimod-, or natalizumab-treated patients with pediatric-onset relapsing–remitting MS. These interventions were chosen because the literature was mainly available for these treatments, and they are currently used for the treatment of pediatric MS. Two researchers independently extracted data and assessed study quality using the Cochrane Effective Practice and Organization of Care – Quality Assessment Tool. The meta-analysis estimates were obtained by Bayesian random effects model. Data were summarized as ARR point estimates and 95% credible intervals. Results: We found 19 articles, including 2 randomized controlled trials. The baseline ARR reported was between 1.4 and 3.7. The meta-analysis-based ARR was significantly higher in IFN β-treated patients (0.69, 95% credible interval: 0.51–0.91) versus fingolimod (0.11, 0.04–0.27) and natalizumab (0.17, 0.09–0.31). Based on the meta-analysis results, an appropriate non-inferiority margin versus fingolimod could be in the range of 2.29–2.67 and for natalizumab 1.72–2.29 on the ARR ratio scale. A Bayesian design, which uses historical information for a fingolimod or natalizumab control arm, could reduce the sample size of a new trial by 18 or 14 patients, respectively. Conclusion: This meta-analysis provides evidence that relapse rates are considerably higher with IFNs versus fingolimod or natalizumab. The results support the use of innovative Bayesian or non-inferiority designs to avoid exposing patients to less effective comparators in trials and bringing new medications to patients more efficiently.