Abstract
OBJECTIVE: To estimate the degree of redundancy among articular response variables used routinely in juvenile chronic arthritis (JCA) clinical trials. METHODS: We obtained data from a large, computerised databank holding information from multiple clinical trials in children with JCA to determine Pearson correlation coefficients for changes from baseline values in articular response variables. All 508 JCA patients who entered the analysis of efficacy of the original trials were included in the present study. An r value of 0.7 or greater was considered evidence of colinearity. RESULTS: When changes in joint counts of articular manifestations were compared with changes in the respective severity scores (for example swollen joint count versus swelling severity score), the r values were always > or = 0.7. Three articular response variables related to the assessment of pain are all highly correlated with each other. The number of joints with active arthritis was correlated with the number of joints with swelling and the swelling score. Change in the overall severity score was highly correlated (r > or = 0.7) with change in nearly all articular response variables. CONCLUSIONS: Many endpoints reported in JCA trials are redundant, especially counts and scores of the same articular measure. Eliminating redundant variables would provide the same information with less complexity, and decrease the probability of statistical error and ambiguous results.