(1) First-line disease-modifying treatment for rheumatoid arthritis is based on "slow-acting" antirheumatic agents, generally methotrexate. Subsequent options include a TNF-alpha antagonist, followed by rituximab or possibly abatacept; (2) Tocilizumab, a monoclonal antibody, inhibits interleukin-6 receptors. It is licensed in the European Union for patients with rheumatoid arthritis in whom other drugs have failed; (3) Clinical evaluation includes 4 placebo-controlled trials of the methotrexate-tocilizumab combination, after failure of a slow-acting antirheumatic drug (3 trials) or failure of a slow-acting antirheumatic drug and a TNF-alpha antagonist (1 trial). An indirect comparison suggests that tocilizumab is no more effective than rituximab in patients with multiple treatment failure; (4) Tocilizumab, like TNF-alpha antagonists, is an immunosuppressant. It carries a risk of serious infections, haematological disorders (neutropenia, thrombocytopenia), gastrointestinal bleeding, hepatic disorders, and systemic and local reactions during the infusion; (5) the adverse effects of long-term tocilizumab therapy are unknown, particularly the risk of cancer; (6) Tocilizumab carries a risk of interactions with drugs that are metabolised by cytochrome P450 isoenzymes. Clinical consequences cannot be ruled out when co-administered drugs have a narrow therapeutic margin; (7) Tocilizumab is administered intravenously every 4 weeks, making it slightly more convenient that rituximab at the beginning of treatment; (8) In patients with rheumatoid arthritis and multiple treatment failure, it remains to be shown whether tocilizumab has a better risk-benefit balance than rituximab, a drug with which we have more experience. It is therefore better to continue to use rituximab, or possibly abatacept.