Safety meta-analyses could play an important role in capturing fragmented safety data from underpowered clinical studies. Their utility, however, is hampered by the same constraints surrounding clinical studies, including limited external validity, inconsistent classification and the underreporting of safety data. When long-term adverse effects are studied, like the risk of malignancy associated with the use of inhibitors of tumour necrosis factor (TNF) alpha, particular scrutiny should be exercised in deciding whether the individual trials included have allowed sufficient follow-up time for the safety endpoint to occur.