Selection of markers to predict tumour response or survival: description of a novel approach

Abstract

A major issue with current cancer therapy is patient selection to ensure maximum benefit with minimum toxicity. Current methodologies for the prediction of patient outcome at diagnosis are almost exclusively based on the size of the primary tumour and the degree of local (lymph node) invasion or distant metastasis. Biomarkers have not made a major impact as yet for any solid tumour [1]. However, the Human Genome Mapping Project and other initiatives are producing a large amount of new data with potential relevance for predicting response or survival. In fact, the number of potential biomarkers is in near exponential increase, bringing further problems in selecting which marker will be most important for each tumour type. As we move into the post-genome era we need practical methods to assess rapidly multiple markers to evaluate their potential as prognostic predictors. Guidelines have been proposed for assessing the utility of putative markers of prognosis from existing literature [2], but no approaches appear to be identified for objectively selecting markers with the greatest potential to provide useful information in terms of predicting prognosis, predicting response to therapy, or in guiding therapy selection.