Criteria for improvement in rheumatoid arthritis: Alternatives to the American College of Rheumatology 20

Objective. Appropriate outcome measures are critical to estimating treatment effects in clinical trials and observational studies. The American College of Rheumatology 20 (ACR 20) is the standard measure to assess treatment effects in rheumatoid arthritis (RA) randomized controlled clinical trials (RCT). Although the ACR 20, which measures at least a 20% improvement in a number of different measures, is a very useful dichotomous measure for identifying novel treatments with potential activity, it may fail to discriminate among active treatments. In an effort to increase stringency, trials have used the ACR at 50% and 70%. We investigated the behavior of these and other scoring systems to quantify treatment outcome in several RCT in RA. Methods. The Cooperating Clinics for the Systematic Study of Rheumatic Disease (CSSRD) database contained 2 trials with a total of 6 arms. Using raw data for each patient entered, we calculated ACR 20, 50, and 70 for each arm. We also calculated the average number of criteria met for the ACR 20, 50, and 70 and their respective areas under the curve over time. Results. The ACR 20 failed to discriminate among active therapies; however, the ACR 50 was too stringent, and only one patient in these trials satisfied the ACR 70. The average number of criteria satisfied at the 50% level at the final trial visit discriminated well, as did the area under the curve. Conclusion. The average number of ACR criteria met at a 20% or 50% level discriminated better than the traditional ACR criteria in these trials. More of the information is preserved by the area under the curve of the number of ACR criteria satisfied at each level because area preserves both the number of criteria and the time dependence. The area under the curve of the number of ACR criteria met is a discriminatory, specific, time-dependent, responsive, and domain-preserving metric to use as the primary outcome measure in trials of agents for the treatment of RA. These conclusions should be tested in additional data sets.