A fundamental goal of oncology trials is to reduce overall mortality. But the time horizon from trial entry to death is often long, and in order to conduct a cost-effective and timely trial it is often necessary to rely on surrogate endpoints such as tumor shrinkage or progression-free survival to evaluate the efficacy of new treatments. In order for a surrogate endpoint to be appropriate for use in a trial, it must correlate with the primary endpoint of interest, overall survival. A commonly used measure of tumor shrinkage is the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines. But response proportion as defined by RECIST guidelines has not always been a reliable surrogate endpoint. This talk will begin by reviewing the foundational issues involved in the selection and use of surrogate markers in clinical trials. To illustrate the challenges in identifying a relevant surrogate marker, it will then describe the results from a study analyzing the strength of association between various measures of response and progression with overall survival in the context of vemurafenib treatment for melanoma.