Safety data come from a number of sources. Randomized clinical trials tend to be relatively short, exclude patients with significant comorbidity, have limited numbers of subjects and are primarily powered for efficacy. The most useful post-marketing data come from large national registries, such as Britain's BSRBR, Sweden's ARTIS, Germany's RABBIT, France's DANBIO, Spain's BIODASER and North America's CORRONA. Among the most commonly used non-biologic DMARDs, MTX is associated with risks of hepatotoxicity and cytopenia, as well as pneumonitis, particularly during the first year of treatment. Regarding TNF inhibitors, there is an increased risk of infection (including serious infections) by bacterial pathogens, atypical fungi and opportunistic pathogens. When possible, pneumococcal and influenza vaccines should be given before initiation of treatment with any biologic DMARD. Screening for latent tuberculosis is recommended for all TNF inhibitors, and has been shown to reduce the risk of reactivation. Evidence from registries suggests that there is no increased risk of solid tumours with TNF inhibitor treatment; however, non-melanoma skin cancers are more common. Specific risks with other biologic DMARDs include gastrointestinal perforation with tocilizumab, progressive multifocal leucoencephalopathy with rituximab and pulmonary infections with abatacept. Overall, the safety of biologic and non-biologic DMARDs appears to be reasonable, particularly compared with the risks associated with the disease itself.