Objective: To evaluate whether exercise treatment based on behavioural graded activity comprising booster sessions is a cost-effective treatment for patients with osteoarthritis of the hip and/or knee compared with usual care. Methods: An economic evaluation from a societal perspective was carried out alongside a randomised trial involving 200 patients with osteoarthritis of the hip and/or knee. Outcome measures were pain, physical functioning, self-perceived change and quality of life, assessed at baseline, 13, 39 and 65 weeks. Costs were measured using cost diaries for the entire follow-up period of 65 weeks. Cost and effect differences were estimated using multilevel analysis. Uncertainty around the cost-effectiveness ratios was estimated by bootstrapping and graphically represented on cost-effectiveness planes. Results: 97 patients received behavioural graded activity, and 103 patients received usual care. At 65 weeks, no differences were found between the two groups in improvement with respect to baseline on any of the outcome measures. The mean (95% confidence interval) difference in total costs between the groups was -€773 (-€2360 to €772) - that is, behavioural graded activity resulted in less cost but this difference was non-significant. As effect differences were small, a large incremental cost-effectiveness ratio of €51 385 per quality adjusted life year was found for graded activity versus usual care. Conclusions: This study provides no evidence that behavioural graded activity is either more effective or less costly than usual care. Yielding similar results to usual care, behavioural graded activity seems an acceptable method for treating patients with osteoarthritis of the hip and/or knee.