Article type
Year
Abstract
Background: Outcome measures are critical to accurately estimate the treatment effects in clinical trials and observational studies. The ACR 20 is the standard measure to assess treatment effects in Rheumatoid arthritis (RA) clinical trials (RCTs). Because the ACR 20, which measures a 20% improvement in a number of different measures, is too sensitive it often fails to discriminate amongst disease modifying antirheumatic drugs (DMARDs). In an effort to increase stringency, trials use ACR at 50% and 70%. We investigated the behavior of these and other scoring systems to quantify treatment outcome in RA RCTs.
Methods: Three trials with a total of eight arms were analyzed from The Cooperative Systematic Studies for the Rheumatic Diseases( CSSRD). Using raw data for each patient entered we calculated ACR 20, 50 and 70 for each arm. We also calculated the average # of criteria met for the ACR 20, 50 and 70 and their respective areas under the curve.
Results: In each trial DMARD therapy was superior to placebo. The ACR 20 failed to discriminate amongst active therapies; however, the ACR 50 was too stringent and no patients in these trials satisfied the ACR 70. The average # of criteria satisfied discriminated best at the ACR 50 as did the area under the curve.
Conclusions: The average number of ACR criteria met at a 20, 50 or 70% level discriminates better than the traditional ACR criteria but more of the information is preserved by the area under the curve of the # of ACR criteria satisfied at each level because they preserve both the continuous nature of the data and its time dependence. Further study will be necessary to generate a sensitive, specific, time-dependent, responsive and domain preserving metric to use as the primary outcome measure in trials of agents for the treatment of rheumatoid arthritis.
Methods: Three trials with a total of eight arms were analyzed from The Cooperative Systematic Studies for the Rheumatic Diseases( CSSRD). Using raw data for each patient entered we calculated ACR 20, 50 and 70 for each arm. We also calculated the average # of criteria met for the ACR 20, 50 and 70 and their respective areas under the curve.
Results: In each trial DMARD therapy was superior to placebo. The ACR 20 failed to discriminate amongst active therapies; however, the ACR 50 was too stringent and no patients in these trials satisfied the ACR 70. The average # of criteria satisfied discriminated best at the ACR 50 as did the area under the curve.
Conclusions: The average number of ACR criteria met at a 20, 50 or 70% level discriminates better than the traditional ACR criteria but more of the information is preserved by the area under the curve of the # of ACR criteria satisfied at each level because they preserve both the continuous nature of the data and its time dependence. Further study will be necessary to generate a sensitive, specific, time-dependent, responsive and domain preserving metric to use as the primary outcome measure in trials of agents for the treatment of rheumatoid arthritis.