Designing Phase II MS Clinical Trials Without Control Arm

Introduction: Phase II trials in multiple sclerosis (MS) are increasingly challenged by ethical concerns about randomizing patients to a control arm, given the availability of several effective treatments. While alternative trial designs that omit control arms have been considered, these risk compromising methodological rigor. In parallel, efforts have been made to define new Phase II outcome measures, including radiological biomarkers like brain parenchymal fraction (BPF) in progressive MS and laboratory markers such as neurofilament light chain (NfL) in relapsing-remitting MS. Objectives/Aims: To propose a robust MS trial design that minimizes the need for control arms while accommodating emerging MS outcome measures. Methods: Through extensive simulations, we demonstrate the potential of a novel clinical trial design aimed at evaluating treatment effects on a biomarker indicative of disease progression. The design is a single-arm trial, wherein each participant starts as a control and transitions to the active intervention at a randomly assigned time. We use curve alignment, a statistical technique from functional data analysis, to reconstruct the timing of the intervention from noisy, repeated biomarker measurements. The detection of the treatment effect is then enabled by evaluating the correlation between the estimated and actual timing of the intervention. Our simulations explored a grid of parameters inspired by Expanded Disability Status Scale, BPF and NfL values reported in the literature, as well as both challenging and favorable scenarios. Results: Simulation results demonstrated that the correlation between estimated and actual intervention times reliably identifies treatment effects when present. Across scenarios, the method achieved a type I error consistently below 4% and a power above 80% for detecting treatment efficacy based on realistic assumptions of treatment effects and progression rates, as reported in the literature. Performance improved with larger treatment effects, larger sample sizes, denser and longer biomarker observations, and faster biomarker worsening. Additionally, compared to conventional control-arm trials, the proposed design enhances statistical power. Conclusion: The proposed design for Phase II trials demonstrated strong effectiveness in simulations for detecting intervention activity and could be practically implemented to reduce the need for a control arm.