Because its charge was to identify those patients who would benefit from early initiation of treatment, the task force decided to use initiation of methotrexate (MTX) as the gold standard against which variables would be measured. If physicians had started patients on MTX, this meant that, in their clinical judgment, these patients were at risk for developing persistent or erosive RA. This approach avoided the circularity that would occur if the 1987 criteria were used as the gold standard.
Daniel Aletaha, MD, MSc, of the Medical University of Vienna, Austria, also a member of the task force, described the scientific challenges of the three-step data analysis of Phase One in his presentation entitled, “Developing New RA Classification Criteria: Learning from an Early Disease Cohort.” Investigators from nine centers gave the task force access to their early arthritis patient cohorts. With a total of 3,115 included patients, the cohorts yielded a large field of variables, with substantial variation. For instance, the average duration of symptoms at presentation ranged from 1.2 to 7 months, and the percentage of patients who were rheumatoid factor seropositive ranged from 13–40%. “This heterogeneity was one of the great strengths of the analysis because the [differences between the cohorts support] the generalizability of the results we generated,” Dr. Aletaha explained.
Step 1 of the data analysis began with what Dr. Aletaha called “a cloud of variables.” The team performed univariate regression analysis to look at each variable’s effect on MTX treatment at one year. Some of the variables, such as age, gender, and involvement of large joints, were eliminated as not predictive of persistent disease. Step 2 involved performing a principle component analysis to identify sets of variables that represented a similar theme. Each of these groupings, or factors, was given a name: metatarsalphalangeal (MTP) involvement, wrist involvement, hand–finger tenderness, acute phase response, proximal interphalangeal (PIP) involvement, and serology. The final multivariate regression analysis, Step 3, then yielded the independent associations of the key variables that would need to be considered in a final set of criteria.
After this three-step analysis, the Phase One effort had identified key variables and provided the relative weights of those variables to inform the final criteria. Among the findings:
- The importance of the hands as signified by swelling of the PIP, MCP, and wrist regions, with an independent contribution of tenderness;
- Symmetrical involvement did not have a significant effect over unilateral involvement; and
- Abnormal acute phase response, especially C-reactive protein values, had a considerable effect.
Phase Two: The Probability of RA
During Phase Two, the aim was to reach consensus on those patient factors that clinicians use to determine the patients’ probability of developing persist or erosive RA, according to task force member Gillian A. Hawker, MD, MSc, of the Women’s College Hospital, Toronto. In Phase Two, a 24-member expert panel was assembled, including task force leaders and other ACR and EULAR members, academicians and clinicians, and men and women. Each panel member submitted three to five patient-case scenarios based on real patients with undifferentiated inflammatory arthritis seen in their practices. They ranked their cases on a five-point Likert scale according to the probability that the patient would develop RA. From a total of 86 cases they received, the task force chose 30, which they simplified, anonymized, and disseminated to members of the panel. These panel members were then asked to rank all cases in descending order, from highest to lowest probability of RA, given the variables that had been identified in Phase One.
