The main endpoint of a phase III trial in oncology must be relevant clinically. In practice, this endpoint is either the disease-free or the overall survival, which requires both long term follow-up and a large number of patients. In phase II trials, it is essential to proceed rapidly, and one therefore uses a surrogate endpoint. Surrogate endpoints are adequate for phase II trials. Their use for phase III trials should be restricted to the rare situations where their validity has been established for the therapeutic and a under study. This surrogate endpoint must be faster to obtain than the clinical endpoint it replaces. Ideally, it should capture all of the treatment effect on the main endpoint, i.e. there should be no effect of the treatment on survival, once the value of the surrogate endpoint is known.