OBJECTIVES: The aim of the study was to evaluate by quantitative magnetic resonance imaging the effect of celecoxib 200 mg daily on cartilage volume loss over 12 months in knee osteoarthritis. METHODS: The primary outcome of this study was to evaluate cartilage volume loss in the medial compartment of the knee (femoral condyle and tibial plateau) assessed by quantitative magnetic resonance imaging on subjects receiving continuous treatment with celecoxib 200 mg daily for 12 months compared with a modelized historical control cohort, as expressed by the percentage loss from baseline. Safety of the medication was also assessed. Comparison of the observed volume loss to the expected loss was evaluated by a multivariate linear regression model based on a historical cohort. RESULTS: For the primary outcome, the 95% confidence intervals for the mean observed celecoxib cohort joint medial compartment cartilage volume loss (6.81% [6.01; 7.60]) and mean predicted loss (modelized historical cohort) (5.65% [5.10; 6.19]) overlap, indicating no significant difference and hence no effect of celecoxib on the medial compartment cartilage volume loss. Similar findings were demonstrated for the lateral compartment cartilage loss. The safety data reported several minor adverse events similar to those typically seen in a 1-year clinical trial. CONCLUSIONS: Although celecoxib was demonstrated to be safe for knee osteoarthritis at a 200 mg daily dose, it did not provide a protective effect on knee cartilage loss. Cohort modelization is an efficient and unbiased way to provide a comparator group for the assessment of novel treatments when classic head-to-head randomized controlled trials are not feasible.