A comparison of the measurement properties and estimation of minimal important differences of the EQ-5D and SF-6D utility measures in patients with systemic sclerosis
Kwakkenbos, L.; Fransen, J.; Vonk, M. C.; Becker, E. S.; Jeurissen, M.; van den Hoogen, F. H.; van den Ende, C. H.
Clinical & Experimental Rheumatology
OBJECTIVES: To compare measurement properties of the EQ-5D and SF-6D utility measures, to assess the association and agreement between these measures and to estimate minimal important differences (MID) in patients with systemic sclerosis (SSc).; METHODS: Both measures were assessed twice in an observational prospective design over a 12-month period (n=211). Spearman’s rank correlation between the EQ-5D and SF-6D was calculated at baseline. Agreement was assessed using Lin’s concordance coefficient (LCC) and a Bland-Altman plot. MIDs were estimated using three anchors; the global rating of change item (SF-36) and changes on the Health Assessment Questionnaire-Disability Index (HAQ-DI) of >0.14 and >0.22.; RESULTS: At baseline, the mean EQ-5D and SF-6D scores were 0.64 (SD=0.25) and 0.65 (SD=0.11), respectively. The correlation between EQ-D and SF-6D scores was r=0.74. Agreement was moderate (LCC=0.49), and the Bland-Altman plot showed a mean difference of 0.003 but wide limits of agreement (-0.38 to 0.39) and a structural bias for lower scores. The mean MID estimate for the EQ-5D was 0.08 for the improved subgroup, and -0.13 for the deteriorated subgroup. For the SF-6D, the MID estimate was 0.05 for the improved and -0.04 for the deteriorated subgroup.; CONCLUSIONS: Although there was a marked correlation between the measures, the moderate agreement implies that EQ-5D and SF-6D scores cannot be used interchangeably. The MID estimates we provided can be used to calculate sample sizes for clinical trials involving SSc patients, and in interpreting the relevance and importance of treatment effects.