Moving from correlative clinical science to predictive medicine

摘要: ED01-03 New technology and biological knowledge make it increasingly feasible to predict which patients require systemic therapy are most likely benefit or suffer severe adverse events from a specific reatment. Using genomic classifiers target treatment can greatly improve the therapeutic ratio of effects reduce societal medical costs. There are, however, many challenges in effectively co-developing new drugs with predictive classifiers.
 Much conventional wisdom about how develop utilize biomarker is flawed does not lead definitive evidence for well defined population. The data used classifier must be distinct test hypotheses effect subsets determined by classifier. Developmental studies exploratory, but on effectiveness claims based should that pre-specified (1). purpose phase III clinical trial drug utilizes determine whether provides population pre-defined refine demonstrate repeating development process independent results same classifier.
 My presentation will describe designs utilizing conjunction development. Simon Maitournam (2,3) evaluating efficiency enrichment selecting described. Randomized described Wang (4) (5-7) do restrict eligibility permit overall all randomized as one subset This design supports broad labeling indications cases where activity sufficient, strong prospectively more selective. adaptive Freidlin permits during initial then testing identified also discussed Jiang et al. identification best cut-point index. Reprints, interactive software planning targeted trials available at http://linus.nci.nih.gov/brb.
 References 1.Simon R. Roadmap developing validating therapeutically relevant classifiers. Journal Clinical Oncology 2005;23:7332-41.
 2. R Maitnourim A. Evaluating trials. Cancer Research 2004;10:6759-63; Supplement correction 2006;12:3229.
 3. A On Statistics Medicine 2005;24:329-39.
 4. R, SJ. Use signatures therapeutics Pharmacogenomics 2006;6:166-73.
 5. genomics design. (in press).
 6. Simon, Designs analysis plans pivotal companion diagnostics. Expert Opinion Medical Diagnostics 7. Lost translation: Problems pitfalls translating laboratory observations utility. European 8. B Adaptive signature design: An generating gene expression sensitive patients. 2005;11:7872-8.
 9. W, Varma S, Calculating confidence intervals prediction error microarray classification using resampling. Statistical Applications Genetics Molecular Biology :7(1)article 8, 2008.