Objective: To further the development of criteria for clinically important changes in outcomes seen in rheumatoid arthritis (RA) patients and trials.

Methods: Small group discussions and voting on specially designed profiles based on a 6 factor 2 level factorial design for changes seen in patients and trials. The purpose was to bring out the implicit opinions of participants on these issues, to complement the explicit opinions expressed in the OMERACT questionnaire. This took place at a conference of rheumatologists, methodologists, biostatisticians, regulatory, pharmaceutical and biotechnology industry personnel with an interest in therapies for patients with RA. Data from patients with RA and randomized clinical trials of second line drugs in patients with RA formed the basis to create 64 patient profiles and 64 trial profiles. The profiles contained information on changes in 6 measures: swollen joint count, tender joint count, pain, patient global assessment, physician global assessment and physical disability. The profiles were prepared on 4" x 5" cards and presented to the participants in packages of 64 in random order, in 2 different group sessions. Participants were assigned to 8 groups that contained a mix of all types of participants, with a majority of clinicians in each group. In the patient profile session, individual participants scored whether the profile represented important improvement. In the trial profile session, participants did likewise for important difference between the drugs. After structured discussion the group then voted: consensus was defined as agreement by at least 70% of the group. We decided that an important improvement or difference was present in the profiles on which at least 6 of the 8 groups had achieved consensus. The changes in the profiles showing important improvement or difference can be used to suggest minimum criteria for each of the 6 measures.

Results: The lower quartile of the change present in the patient profiles with important improvement varied from 17% (swollen joints) to 49% (disability); the median of these quartiles was 36%. The lower quartile of the differences present in the trial profiles with important differences varied from 13% (tender joints) to 26% (physician global); the median of these quartiles was 18%.

Conclusions: This approach has provided a beginning for less arbitrary definition of criteria for important change in patients and trials. It might be suggested that for patients, an improvement of at least 36% should be clinically important, while an active drug needs to be at least 18% better than placebo to be clinically important.