Biomarkers and surrogate end points have great potential for use in clinical oncology, but their statistical validation presents major challenges, and few biomarkers have been robustly confirmed.