Objective. To investigate the utility of serum COMP level measurements as a predictor of future damage of the weight-bearing (large) joints in RA patients participating in intensive exercise. Methods. Data of the 281 completers of a 2-yr randomized controlled trial (Rheumatoid Arthritis Patients In Training; RAPIT) comparing the effects of usual care physical therapy with high-intensity weight-bearing exercises were analysed. The primary outcome variable was defined as the change in radiological joint damage (Larsen score) of the large joints. Potential predictors of outcome were defined: baseline and change in serum level of COMP after 3 months, baseline radiological damage of the large and small joints, number of months on glucocorticoids, change in disease activity and in physical capacity (aerobic fitness and muscle strength) after 2 yrs, and participation in the exercise group. Results. In cross-sectional evaluation of baseline data, we found strong association between the high serum COMP level and current damage of the large joints. Serum COMP level at baseline, however, was not associated with an increased rate of radiological joint damage after 2 yrs of follow-up. Furthermore, neither interaction between baseline COMP level and participation in exercises, nor change in COMP level after 3 months of exercising were associated with future damage of the large joints. Conclusion. Neither baseline serum COMP level nor its individual change after 3 months from start of intensive exercise predict longitudinal progression of damage of the large joints in this population.