OBJECTIVES : To quantify patients’ preferences for multiple sclerosis (MS) treatment attributes and how they may be impacted by treatment experience.

METHODS : In a double-blind, active-controlled, randomized controlled clinical trial, OPTIMUM (NCT02425644), subjects completed a questionnaire at screening, baseline and follow-up (FU1; 15 days after end of treatment). The questionnaire aimed to elicit preferences for 7 MS treatment attributes (number of relapses in 2 years, years until MS worsens, disability due to disease progression, likelihood of liver failure within 10 years, likelihood of serious infection leading to disability within 10 years, likelihood of developing cancer within 10 years and time spent in hospital (for monitoring of the heart) using the Measuring Attractiveness by a Categorical Based Evaluation Technique (MACBETH). Dirichlet regression models were used to analyse observable preference heterogeneity.

RESULTS : In total, 293 (63% female, mean age 38 ± 8 years) subjects with relapsing MS completed the preference questionnaire. At all times, subjects most valued decreasing ‘disability due to disease progression’ (screening=0.218; baseline=0.220; FU1=0.222), followed by increasing ‘time until your MS disease gets worse’ (screening= 0.176; baseline=0.172; FU1=0.170) and decreasing ‘number of relapses in two years’ (screening= 0.165; baseline= 0.165; FU1= 0.164). Subjects in the ponesimod treatment arm cared significantly less about disability due to disease progression at FU1 (p-value=0.024) compared to those in the teriflunomide treatment arm. Patients who had experienced >1 relapses in the past 12 months cared significantly more about ‘disability due to disease progression’ at screening (p-value=0.041) and ‘time until MS gets worse’ at FU1 (p-value=0.029) compared to those who had ≤1 relapse in the past 12 months.

CONCLUSIONS : Trial subjects cared most about disability due to disease progression and their preferences were stable throughout the trial. MACBETH technique was able to capture treatment preferences accurately and Dirichlet regression was appropriate for quantifying observable preference heterogeneity.