Predicting long-term sustained disability progression in multiple sclerosis.

Introduction: Prevention of disability over the long-term is currently the main treatment goal in multiple sclerosis (MS). Randomized clinical trials evaluate short-term treatment effect on disability in the form of 3-6-month confirmed disability progression. Using global MSBase registry, this study establishes 6-month confirmed disability progression events as indicators of long-term disability worsening suitable for use in randomized clinical trials. Method(s): A Cox proportional hazards model in the development cohort identified associations between demographic and clinical variables at the time of disability progression and the probability of an event not being sustained in the future. The coefficients from this model were used to calculate a sustained progression score. Its association with the risk that an event will be sustained over timewas evaluated with a Cox model in the validation cohort. Result(s): 11,435 confirmed progression events identified in 6,902 patients constituted the development cohort. Patient characteristics at the time of progression associated with lower probability of subsequent improvement were age (hazard ratio (HR)=0.98), primary progressive (HR=0.37) and progressive-relapsing (HR=0.36) MS, expanded disability status scale score >=6 (HR=0.71) and its change from baseline (HR=0.67), number of affected functional system scores (HR=0.92) and pyramidal(HR=0.79) functional system score (p< 0.05). The strength of the association with pyramidal score decreased with time(HR=1.01). Occurrence of a relapse within previous month (HR=1.46) and worsening in sensory functional system score (HR=1.17) were associated with higher probability of improvement after progression. The associations were confirmed by two sensitivity analyses. The sustained progression score, ranged 0.39-4.79 in the validation cohort with 1,271 progression events, estimated a 53% lower chance of improvement with each unit increase in the score (HR=0.47, 95% confidence interval