Evaluating MRI-Confirmed Relapses as a Novel Primary Endpoint in Multiple Sclerosis Trials

Background

Clinically defined relapses are the traditional primary endpoint of randomized control trials (RCTs) in multiple sclerosis (MS), yet a substantial proportion lack new inflammatory lesions. Confirming relapses with brain and spinal cord MRI to distinguish relapses with active MRI (RAM) from acute clinical events with stable MRI (ACES) may provide a more sensitive primary outcome for future trials.

Objective

To estimate RAM and ACES rates in MS trials and evaluate the impact on statistical power of using RAM.

Methods

We used two approaches: an aggregated data (AD) approach, combining population-level data from RCTs with observational data from the French MS registry, and an individual patient data (IPD) approach from the PRIMUS platform. Trials were selected if they evaluated DMTs sufficiently represented in the OFSEP ancillary study or were available in PRIMUS. Eleven pivotal RCTs were included, evaluating natalizumab, cladribine, dimethyl fumarate, teriflunomide, fingolimod, or ocrelizumab; 7 were analyzed with AD only, 1 with IPD only, and 3 with both. For the AD approach, population-level characteristics were extracted from published reports; expected RAM probabilities were then derived from a RAM model fitted on OFSEP observational data, applied to each RCT arm population. For the IPD approach, clinically defined relapses were directly classified as RAM/ACES according to radiological activity on subsequent brain MRI. Main outcomes were the treatment effect on RAM and ACES rates, compared with the effect on clinically defined relapses.

Results

Across 11 RCTs, treatment effects were consistently equal or greater for RAM than for clinically defined relapses, with both AD and IPD approaches. No DMT significantly reduced ACES rates, which remained stable at approximately 0.08 events/year across arms. The IPD approach yielded systematically lower RAM probabilities than the AD approach.

Using RAM as the endpoint improved statistical power in most scenarios: e.g. a trial with annualized relapse rates of 0.15/year (active arm) vs 0.30 (control arm) requires one-third fewer participants.

Conclusion

Adopting RAM as the primary outcome could substantially enhance the power of future MS trials and better target the effect of treatment on inflammatory activity.