In a study designed to establish the safety of a marketed drug, interim analyses performed to detect harm can protect trial participants and the wider public before the final analysis occurs. Monitoring for harm within a safety study is different from monitoring for benefit, so techniques commonly used in an efficacy study of an experimental drug may not apply. We propose potentially more suitable techniques in this setting, including a novel spending function and conditional power. These techniques have reasonable operating characteristics in a simulation. The appropriate technique to implement will depend on circumstances of specific to the individual safety study.