Poor to modest agreement between rheumatoid arthritis response measures in clinical practice

Objective: To evaluate the agreement among several rheumatoid arthritis (RA) response measures in a clinical setting. Methods: 529 patients with RA were seen at 2 regular visits where the following response measures were determined: ACR-20, EULAR good or moderate (EULAR-GM), Simplified Disease Activity index moderate (SDAI-M), Clinical DAI moderate (CDAI-M), and Patient Reported Outcomes Index-M 20 (PRO-IM-20). Each measure was modified to include a "worse" response, i.e. the inverse of the respective guidelines for a positive improvement response. Introduced for comparison was the Real-time Assessment of Disease Activity in Rheumatoid Arthritis (RADARA), a response measure that registers improvement if the patient's tender and swollen joint counts and HAQ score all improve and worsening if all three increase. Contingency tables comparing the three responses (worse, no change, and improvement) along with Cohen's kappa were calculated. Results: The mean (SD) baseline characteristics of the patients included: age 66.5 (10.7) years, RA duration 12.9 (11.0) years, 91.3% male, 84.1% rheumatoid factor positive, and a Disease Activity Score-28 of 3.5 (1.3). The percentage of patients who improved/worsened were as follows: ACR-20 4.7/9.1, EULAR-GM 23.4/26.3, SDAI-M 16.1/20.6, CDAI-M 16.3/20.0, PRO-IM-20 22.5/34.4, and RADARA 7.0/11.5. Agreement (kappa) was poor to slight (≤ 0.4) between most of the response measures with the exception of RADARA/ACR-20 which showed substantial agreement (0.67) and SDAI/EULAR-GM and CDAI/EULAR-GM, which showed moderate agreement (0.54 and 0.52, respectively). Conclusion: RA response measures can be made more informative by the addition of a "worse" response, although even in this case the agreement in the clinic setting is primarily poor to moderate.