摘要：  

Clinical trials utilizing predictive biomarkers have become a topic of increasing research in the era of personalized medicine. In the talk we will focus on the stratified biomarker design where patients with the same biomarker status are randomly assigned to either an experimental arm or the standard of treatment. The primary interest of a stratified biomarker design is to investigate if patients respond differently to treatment based on their biomarker status. Despite the advancements in molecular assays, correctly identifying the biomarker status remains a challenging task. We analytically and numerically demonstrate the profound adverse effects of misclassified biomarker status on the estimates of treatment effect, biomarker effect, treatment-biomarker interaction, the corresponding confidence intervals, power of the tests, and required sample sizes. We further propose respective remedies that tackle the misclassified biomarkers in the design and analysis phase of clinical trials. Some challenging statistical issues will be discussed including optimal cutpoints, combination of multiple biomarkers.