Sometimes, a new candidate predictive biomarker is identified after the definitive trials have demonstrated benefit for the specific drug, making new prospective randomized clinical trials of that agent impractical. Simon et al. (2009) indicated that in some cases, it may be possible to use archived specimens collected in the past from appropriate, previously conducted therapeutic trials in a way that preserves the focus, control of type I error, and statistical power of properly designed fully prospective studies. Indeed, when there is substantial preliminary evidence that a new marker predicts benefit from a specific drug, it may sometimes be possible to assay the marker in archived specimens from randomized clinical trials that were conducted to evaluate the drug, as was done for KRAS mutations in colorectal cancer (Amado et al., 2008, 591; Karapetis et al., 2008, 671).

When suitable archived tissue is available, this prospective–retro-spective approach can facilitate and expedite delivery of valuable cancer diagnostics that may be of considerable benefit to patients. Simon et al. (2009) developed stringent conditions required for such a prospective–retrospective approach to have a degree of reliability that approaches that of a fully prospective clinical trial, and proposed a refinement to the previously published Level of Evidence (LOE) Scale that permits a more critical analysis of the quality of tumor marker studies using archived specimens.