A large amount of medical research is never published and studies that are published are more likely to report favourable results. This blog explores how this ‘publication bias’ is a scientific and ethical problem that can lead to the benefits of treatments being overestimated, and harms being underestimated.