Purpose of review: We critically review the methods used in microarray profiling, the reliability of published genomic signatures and the design of ongoing trials based on two of these signatures.
Recent findings: The main limitations of microarray prognostic signatures are known: instability of gene lists, overoptimistic performance indicators and inadequate validation. Two commercially available gene expression-based prognostic tests (MammaPrint and OncotypeDX) are currently being assessed in two large randomized clinical trials. The overall concordance between patient classification resulting from these tests is only slightly better than the concordance between MammaPrint and the Saint Gallen risk. In North America, TAILORx is specifically evaluating the usefulness of chemotherapy in intermediate-risk patients. In the case of MINDACT, ongoing in the European Union, clinical validation of the prognostic test will be considered successful if the proportion of failures among patients classified as having a good prognosis by the prognostic test is below a predefined level. Randomization will not address the validation of the gene expression-based prognostic test in either of these two trials.
Summary: Microarray-based prognostic tests are irremediably moving to the clinics, but their clinical utility might never be formally established.