OBJECTIVES: The revolution of early aggressive treatments for early arthritis (EA) has fuelled the search for better approaches to establishing their cost-utility ratio. The authors aimed to compare the responsiveness of the EQ-5D and the SF-6D in a large prospective cohort of patients with EA. METHODS: EQ-5D and SF-6D utility measures were assessed in 813 patients with EA over 2 years. Responsiveness was analysed by the standardised response mean (SRM) and effect size between baseline and 6, 12 and 24 months for the entire sample and subgroups by disease evolution (increase or decrease in Disease Activity Score for 28 joints). Bootstrap methods were used to estimate 95% CI. RESULTS: The EQ-5D provided larger absolute mean change estimates with greater variance than the SF-6D, whatever the direction of change. At 12 months, the SF-6D was more sensitive to change with improved condition than the EQ-5D: SRM 0.83 (0.82 to 0.84) versus 0.57 (0.56 to 0.58). In contrast, the EQ-5D was more sensitive to change with deteriorated condition than the SF-6D: SRM -0.20 (-0.23 to -0.18) versus -0.11 (-0.14 to -0.08). Results were similar for 6 and 24 months. CONCLUSIONS: The SF-6D was more responsive than the EQ-5D with improved EA condition. Confidence in the relative cost-effectiveness of two treatments would be better with the SF-6D because of its smaller variance. The SF-6D provided more conservative cost-effectiveness ratios than the EQ-5D and may be more appropriate for trials of biological treatments for patients with EA.