Harms caused by pharmaceutical treatments can be measured on absolute scales (for example, risk difference) or on relative scales (for example, relative risk), but there is no clear guidance on the situations for which one type of scale should be preferred over the other. The situation is analogous for the measurement of benefits; with respect to benefits we have previously shown that, when the clinical variable of interest is a time-to-event, the hazard ratio is preferred over absolute scales. In this presentation we extend that work to the evaluation of harms, and will consider the pros and cons of absolute and relative scales for safety endpoints, and for benefit:risk evaluation.