Biotechnology Healthcare - June 2008 - (Page 38) Multiplexing will enable diagnoses based on a more informative assessment of biomarker panels, providing better disease prognosis and more effective patient management. actually prognosticates and predicts what the developmental studies claimed. Ideally, validation studies use different patients and are conducted by researchers independent of those who performed the developmental studies. “You really shouldn’t use the same data to develop the classifier and then to evaluate it,” says Simon. “Everybody loves to do developmental studies, creating new things with potential medical applications. A validation study is not quite so sexy — collecting a lot more data so that you can validate something that someone else did. The biggest caveat is that it’s not ready for use with patients unless it’s been validated in a separate study.” Are patients sufficiently homogeneous to be therapeutically relevant? Simon explains: “With tumor tissue specimens, what’s often done is to use a department collection accumulated from patients with breast cancer over 10 or 20 years. However, these specimens usually are from a wide variety of patients who were treated in a wide variety of ways and whose cancer may have advanced to a wide variety of stages. Your new assay seems to be correlated with, say, disease-free survival. A lot of published oncology studies are of this type, and it’s very hard to figure out what to do with that. It’s important in doing these kinds of studies to have a therapeutically relevant question that can be answered and will help oncologists select treatments for patients. What oncologists really want to know is how to treat patients.” With Oncotype DX, the therapeutically relevant question is “Does this node-negative, ERpositive patient have a sufficiently good prognosis on tamoxifen alone that the potential advantage from chemotherapy is minimal?” The recurrence score provides the answer. Does the study provide information about assay reproducibility? “Genomic Health did the analytic validation of the Oncotype DX assay,” says Simon. “It’s really technical reproducibility kind of stuff, showing that you can take a patient’s tumor tissue and ship it across the country, and if you do the assay on different parts of the same tumor, you get the same answer, and if you do the assay twice in their lab, they get the same answer.” Focus on the importance of analytic validation was heightened after studies presented at the 2007 ASCO annual meeting reported inconsistent results from commercial HER2 assays of breast tumor tissue. HER2 tests identify responders to treatment with trastuzumab (Herceptin). ASCO addressed this issue in its guidelines by recommending that laboratories offering HER2 testing be annually accredited. Shak contrasts the HER2 testing troubles with the Oncotype DX assay. “We’ve made the investment, and continue to invest greatly, in standardizing our assay so that it can be run in a single lab where we get the same results again and again,” he says. “That’s so, so important.” EVIDENCE TO DEFINE QUALITY OF DATA When developing a diagnostic for cancer treatment planning, “We basically took a lot of the good principles used in how to develop drugs and get the evidence to justify the use of a new drug and applied those same principles,” says Shak. “We also listened to people like Richard Simon, who has outlined how to perform rigorous studies in the papers he’s published.” Daniel F. Hayes, MD, clinical director of the Breast Cancer Oncology Program at the University of Michigan Comprehensive Cancer Center, Ann Arbor, was cochair of the 2007 ASCO Tumor Markers Update Panel. Hayes, who recused himself from the deliberations regarding multigene assays such as Oncotype DX because of his past association with Genomic Health as a paid consultant and active scientific collaborator, readily acknowledges Simon’s influence. “A lot of what I have learned, I have learned from Richard,” says Hayes. A panel committee reviewed and analyzed data published since 1999 and evaluated 13 tumor markers, with some identified for multiple applications. The Multiparameter Gene Expression Analysis for Breast Cancer, which includes evaluations of the Breast Cancer Gene Expression Ratio; MammaPrint; and the Rotterdam Signature were new considerations for the 2007 guidelines. 38 BIOTECHNOLOGY HEALTHCARE · MAY/JUNE 2008
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.