Clinical trials in cancer often have several measures of outcome, such as recurrence, survival, and toxic effects of treatment, and it is rare that the same treatment will be better for all outcomes. Benefits and risks therefore need to be balanced against one another. Quantification of this trade-off is a particular issue in so-called equivalence and non-inferiority trials in which a very small loss in survival (affecting few patients) is frequently deemed acceptable for a very large gain in toxic effects (affecting many). For various reasons, views about this trade-off can change during a trial's progress. From our experience of a recent chemotherapy trial for advanced testicular cancer, we discuss the issues that investigators, statisticians, and data monitoring and ethics committees face when the factors that affect the trade-off change substantially during the course of a clinical trial. We show the value of structured questionnaires--completed by clinicians who participate in the trial--that elicit individual decision-making change-points (or ranges of equivalence) with respect to the primary outcome measure, and discuss their use in the design and monitoring of equivalence and non-inferiority trials.