OBJECTIVE: The revolution of early aggressive therapy in early arthritis (EA) has fueled the search for better approaches to establish cost-effectiveness. Our objective was to compare the EuroQol EQ-5D health outcome measure and the SF-6D and to investigate their relationship to clinical variables in a large prospective cohort of patients with EA. METHODS: The EQ-5D and SF-6D utility measures were longitudinally assessed in 813 patients with EA. Agreement and aspects of validity (construct validity, discrimination) were assessed. RESULTS: At baseline, mean values for EQ-5D were 0.52 ± 0.31 (range -0.59 to 1.0) and for SF-6D were 0.58 ± 0.11 (range 0.30 to 0.92), with a bimodal distribution for the EQ-5D. Agreement was low for patients with severe disability or active disease: the utility was systematically lower with EQ-5D. The intraclass correlation coefficient was 0.42 at baseline and increased to 0.53 at 6 months and 0.57 at 1 and 2 years. Correlations between the 2 utility scores and the Health Assessment Questionnaire were good, and remained similar and stable over 2 years (r = -0.70). Correlations with the Disease Activity Score for 28 joints and the physical component of the MOS 36-item Short-form Health Survey (SF-36) were moderate to good and stable. In contrast, correlation with the mental component of the SF-36 was better with the SF-6D, and the correlation with pain, weak at baseline, improved at 6 months and remained stable thereafter. The SF-6D was better able to discriminate patients with high disease activity. CONCLUSION: There was systematic disagreement between EQ-5D and SF-6D in EA, especially in patients with worse clinical outcomes. Using the 2 instruments could be appropriate to conduct sensitivity analyses of cost-utility ratios because the instruments measure utility with closely similar measured properties, but at different levels.