Application of Markov Chains to Multiple Sclerosis Clinical Trial Data to Estimate Disease Trajectories

Background: Multiple Sclerosis (MS), an autoimmune disease affecting millions worldwide, is characterized by its variable course, in which some patients will experience a more benign disease course and others a more active one, with the latter leading to permanent neural damage and disability. Methods: This study uses a Markov Chain model to demonstrate the probability of movement across different states on the Expanded Disability Status Scale (EDSS) and attempted to define worsening, improvement, cycling, and stability of these different pathways. Most importantly we were interested in assessing the lack of impermanence of confirmed disability worsening and if it could be estimated from the Markov model. Results: The study identified only 8.1% were considered worsening, 5.6% consistent improving and 86% cyclers and less than 1% consistently stable. More importantly we also found that many (approximately 30%) of participants with confirmed disability worsening (CDW) regressed to stages that were not considered worsening, on subsequent visits after CDW. Conclusions: These finding are similar to what has been reported previously as predictors of worsening, and also for a lack of durability of CDW, but our results suggest that clinical trial endpoints may need to be modified to more accurately capture differences between the treatment and control groups. Further, this suggests that the rate of worsening in trials that use time to CDW are overestimating the extent of CDW. The trials remain valid since the regressing applies to both treatment and control groups, but that the results may be underestimating the treatment benefit due to misclassification.