Objectives Meta-analysis of randomised controlled trials has shown that TNF-inhibitors are effective in RA. These trials however are undertaken in highly selective populations; under an experimental setting that may differ from that of clinical practice, and follow up rarely extend beyond 1 year. In addition all trials were sponsored by the manufacturer. We were asked to extend the review and meta-analysis of RCTs with a review of data from registries to evaluate efficacy and safety of TNF-inhibitors when used in clinical practice (real world). We focused particularly on the following questions: What is the efficacy of TNF-inhibitors when used outside clinical trials? What is the efficacy of TNF-inhibitors after long term use? What adverse events are reported in these studies? What is the risk of malignancies following long term use? What are the experiences concerning use of medication, treatment compliance and change of medication. Methods We searched Medline and Embase June 2006 by combining search terms for registries, cohort studies and databases with terms for TNF-inhibitors and rheumatoid arthritis (RA). We included publications from registries or databases on adalimumab, etanercept and infliximab for treatment of rheumatoid arthritis (RA), ankylosing spondylitis, psoriatic arthritis or juvenile idiopathic arthritis. Manufacturers were also invited to submit data. Outcomes considered were efficacy, safety and medication use. Results The search gave 290 hits, 64 references were retrieved and assessed in full text, and 23 publications finally included. These studies covered patients with RA, in addition we found one study on juvenile idiopathic arthritis. We did not identify relevant studies on ankylosing spondylitis or psoriatic arthritis. The summary of the results from the studies are as follows: Effectiveness: We included seven studies from registries and databases reporting clinical effects of TNF-inhibitors. In summary these studies showed that TNFinhibitors were effective also when used in clinical practice. The effect however appeared to be lower compared with RCTs. This could be explained by a more heterogeneous patient population. In addition patients in clinical practice often continued with existing medication, opposed to most clinical trials where patients often discontinued existing medications before enrolling. Although we aimed to 9 assess long-term effectiveness, few patients have been followed beyond 2-3 years of treatment.One study assessed patients with JIA, in this study treatment with TNF-inhibitor (etanercept) led to a significant reduction of the disease activity in most of the patients. Combination therapy: Two randomised controlled trials and data from registries evaluated the combination of TNF and MTX treatment. Treatment with TNFinhibitors and methotrexate (MTX) appeared more effective than treatment with TNFinhibitor alone in reducing the disease activity in patients with RA. Cancer: We included six publications that assessed cancer risk following TNFtreatment. A general comments to these studies is that patients have not been followed sufficiently long to allow for conclusions regarding cancer risk. Four studies analysed risk of lymphoma or leukaemia, with inconsistent results. Two studies analysed risk of solid cancer, with inconsistent results. Experiences from transplantation patients shows that cancer usually develops 10-15 years after immunosuppressive medication.Hence, these studies does not give any further information about the risk of developing cancer following treatment with TNF-inhibitors than reported in the randomized controlled trials. Infections: Treatment with TNF-inhibitors were associated with increase the risk of infections. In particular, the risk of reactivation of latent tuberculosis. However, routine screening and treatment of tuberculosis prior to TNF-treatment have reduced this risk considerably. • Compliance: Continuation of treatment with TNF-inhibitors (etanercept og infliximab) after one year was between 62-73 %. This number is lower than compared with RCTs . The reasons for ceasing TNF-treatment were in most cases adverse events or lack of effect. However, it was found that the compliance to TNF-inhibitors was higher then for traditional DMARDs. Conclusion In conclusion, results from clinical trials and registries show that TNF-inhibitors are effective, also when used on a broader patient population outside the setting of clinical trials. Treatment with TNF-inhibitors is associated with increased risk of infections, in particular tuberculosis. Included studies does not allow for conclusion regarding risk of cancer. Thus, the issue of long term safety is at time being incomplete, with a follow up of 2-3 years in most studies. A national registry for treatment with TNF-inhibitors (and other biologics) in Norway would be a very helpful tool to identify the effect and adverse events after long treatment with TNF-inhibitors.