|
Abstract:
|
In cancer research, there is a growing need in finding biomarkers (e.g., molecular, genetic, imaging) that predict patients' prognosis or response to a specific treatment. Validation of the clinical utility of such biomarkers often require large prospective study. However, due to budget constraint and high cost of bioassays, investigators often have to select a subset from all registered subjects for biomarker assessment. Coupled with the losing of patients with biomarker measures, the correlation between the biomarker and the outcome is commonly moderate and the prevalence rates of positive outcome and positive biomarker are low in the study population. In these situations, investigators face a critical issue of how to select a fixed number of subjects that maximizes statistical power in detecting the correlation of the biomarker and the outcome.
|
