The complete evaluation of a therapeutic intervention for many diseases such as AIDS and cancer requires lengthy and costly follow-ups of patients both for long-term benefit (e.g., survival) and late-occurring adverse events. In recent years, attention has been given to investigation of potential ``surrogate" endpoints for these ``true" endpoints. The potential benefits in terms of earlier results and less expensive studies are obvious. But there are significant risks as well. There are examples in which the use of surrogate endpoints produced dangerously misleading results. This ppresentation will explore some of the statistical issues in evaluating surrogate markers. Examples will focus primarily on cancer clinical trials, including prevention, screening, and therapeutic trials.

Back to Workshop Schedule