Objective: To investigate the cost-effectiveness of five different tumour necrosis factor inhibitor tapering strategies in patients with rheumatoid arthritis (RA) and stable low disease activity, using a modelling design.Method: Using Markov models based on data from the DRESS and STRASS randomized controlled trials, and the Nijmegen RA cohort, five tapering strategies for etanercept and adalimumab were tested against continuation: 1, four-step tapering (DRESS strategy); 2, five-step tapering; 3, tapering without withdrawal; 4, use of a stricter flare criterion; and 5, use of a theoretical predictor for successful tapering. We also examined how well a biomarker should be able to predict in order for strategy 5 to become cost-effective compared to the other strategies.Results: All examined tapering strategies were cost saving (range: EUR 5128 to 7873) but yielded more short-lived flares compared to continuation. The change in utilities compared to continuation was minimal and not clinically relevant (range: -0.005 to 0.007 quality-adjusted life-years). Strategy 1 was cost-effective compared to all other strategies [highest incremental net monetary benefit (iNMB)]. However, there was a large overlap in credible intervals, especially between strategies 1 and 2. Scenario analyses showed that 50% reduction of drug prices would result in the highest iNMB for strategy 2. A biomarker only becomes cost-effective when it is inexpensive and has a sensitivity and specificity of at least 84%.Conclusion: Because our study showed a comparable iNMB for tapering in four or five steps (including discontinuation), we recommend a choice between these strategies, based on shared decision making.