OBJECTIVES: To estimate the budget impact implied by the introduction of RTX after failure of one or more TNF a therapies for the French health care system.   METHODS: A Markov model reproduced the course, over 4 years, of patients treated by either infliximab, etanercept, adalimumab or RTX, after failure of one or more TNF a therapies. Markov states were composed of the 4 TNF a strategies. The success rate for each strategy (ACR20) was based on a bayesian mixed treatment comparisons model.Health care consumptions were estimated from an observational study conducted retrospectively in 59 hospital rheumatology services (TC2 study). Three hypotheses were tested for the penetration of RTX into the market: RTX is not used (H1), RTX is used progressively (H2), RTX is the only treatment available (H3). As a sensitivity analysis, we applied coefficients to health care resources in third line. RESULTS: The RTX target population is expected to be 5700 patients. Under H1, direct total medical costs  are expected to be €47,180 000. Under H3 these costs reach €32,600,000, thus a decrease of  31% respectively compared to H1. Under H2, costs progressively decreased as RTX penetrates the market. The impact of multiplication coefficients is important. At the beginning of the simulation, the difference between the base-case analysis (H2) and the worst case sensitivity analysis (coeff2=20 %) concerning the total direct medical costs reached €3,160,000 (respectively €47,560,000 for H2 and €50,720,000 for H2-20%). However, while RTX progressively enters into the market, this difference decreases, attaining only €1,175,000 at the end of the simulation. CONCLUSIONS: RTX is expected to produce important savings if used after failure of one or more TNF a therapies. This is due to the administration rhythm of this product, generally every 6 to 12 months.