Methods: Data from a randomized trial were used for this secondary analysis. 180 patients were interviewed using instruments including the SF-36 at baseline, 6 weeks, and at subsequent 12-week intervals. The SF-6D algorithm was applied to the SF-36 data to calculate disutilities in 6 domains and an overall health utility score. Annual QALYs were calculated, using varying amounts of data between baseline and the 48-week interview. Population average models were used to assess the relationship between levels and changes in the SF-6D and respectively in the Birmingham Vasculitis Activity Score (BVAS), the Vasculitis Damage Index (VDI), and WG flares.
Results: In the model assessing the relationship between the clinical measures and level of health utility, VDI alone was a significant predictor, but BVAS was only significant when controlling for VDI. In the model assessing changes in utility, both changes in BVAS and VDI were significantly associated, but not the occurrence of flares between interviews. Initially, the role limitations dimension was associated with the least disutility among study subjects. Its associated disutility increased in magnitude over time while other disutilities either remained constant or decreased. Calculating QALYs using baseline, 24, and 48 week observations was associated with a 10% decrease in standard deviation in comparison with calculating QALYs based only on baseline and 48 week observations. Using additional observations to calculate QALYs was not associated with greater precision.
Conclusion: Health utility among WG patients is predicted by clinical measures. Vasculitis damage is consistently associated with the health utility level. However, given the relative constancy of VDI, its' changes were not as strongly associated with health utility changes as the changes in BVAS. Despite flares, possible changes in symptoms for patients with a high level of vasculitis activity, and effects of treatment, the intertemporal correlation of health utility measures was high. Adding additional observations did not improvement the precision of estimated QALYs except when changing from annual to semi-annual measurements.