Background/Purpose: There is growing emphasis on the cost-effectiveness of treating early arthritis (EA). As few studies directly record the utility measures needed for economic analyses, mapping is often used. Health Assessment Questionnaire (HAQ) is ‘converted’ into utility using regression. The use of such transformed data by regulatory bodies which determine drug availability raises concern as it involves mathematical transformation between measures which may not be clinically equivalent and with potentially variable interrelationships over time.

We aimed 1) To assess characteristics associated with SF-6D and EQ-5D utility measures in EA 2) To check whether these associations are stable over 3 years.

Methods: – Patients: included in the French nationwide cohort of EA ESPOIR (at least 2 swollen joints for less than 6 months and suspicion of RA).

               – Data available: SF-6D and EQ-5D utility measures were longitudinally assessed in 813 patients with EA (at baseline, 6 months, 1, 2 and 3 years). Bio-clinical variables and X-rays were also recorded.

               – Analysis:  The determinants of SF-6D and EQ-5D utility measures at each time-point were assessed by multivariate linear regressions in 618 EA patients followed over 3 years. Partial R-squares were used to assess the relative importance of variables to the variation in EQ-5D and SF-6D scores.

Results: At baseline, SF-6D was essentially determined by function, HAQ explaining 50.2% of the variance, whereas after 6 months, SF-6D was essentially determined by mental status (the Arthritis Impact Measurement Scale 2–Short Form) (55.8 to 57.6% of the variance) and the HAQ represented only 7.3 to 13.2%. At each time-point, EQ-5D was essentially determined by function, HAQ explaining 36.9 to 44.2% of the variance, except at 1 year, it was essentially determined by mental status, explaining 40.8% of the variance and the HAQ only 11% (figure).

Conclusion: The major impact of functional ability and mental status, and the variability of the utility determinants over time in addition of the bimodal distribution of the EQ-5D raise concerns about mapping to estimate utilities from clinical instruments. Evaluation of treatment cost-utility should not be based on utility data transformed from HAQ.