OBJECTIVES : Paroxysmal nocturnal hemoglobinuria (PNH) is a rare hematological disorder that causes hemolytic anemia and thrombosis. Ravulizumab, a new treatment administered every 8 weeks (q8w), has been shown to be clinically non-inferior to eculizumab administered every 2 weeks (q2w). The objectives of this study were to evaluate the UK general public’s preferences for less-frequent dosing and other treatment attributes for PNH and to estimate disutilities for use in cost-effectiveness analyses.

METHODS : A stated preference discrete choice experiment survey evaluated attributes including overall survival, treatment administration, burden of hemolysis, risk of meningitis, and need for blood transfusions. Attributes and defined levels were combined using a published orthogonal array into choice sets. The survey was administered to UK general public participants aged ≥18 years. The mixed logit model estimated strength of preference for the attributes and disutilties for each attribute. Marginal rates of substitution were estimated between survival and other attributes in order to estimate disutilities, weighted against average life expectancy.

RESULTS : Data from 385 participants were analyzed following logical consistency checks, reflecting the UK demographics. Analysis of the choice data indicated that maximizing life expectancy was the most important attribute. Regarding other attributes, participants preferred an infusion frequency of q8w compared with q2w (disutility = −0.057), were concerned about the risk of developing meningitis infection (disutility = −0.040) and severe hemolysis (disutility = −0.158), and concerned regarding the need for annual blood transfusions (disutility = −0.073).

CONCLUSIONS : This study estimated disutilities for events that would be difficult to measure in a clinical study. In this rare disease context, this study showed the value that the general public places on attributes of treatment for PNH. We believe that this is a useful method for estimating disutilities for economic modelling where EQ-5D data are not available.