A Simulation Study to Evaluate the Operating Characteristics of a Two-Stage Study to Develop and Validate a Panel of Biomarkers for Predicting Prostate Cancer Recurrence

• Activity Number: 179
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Joseph Koopmeiners*+ and Rachel Isaksson Vogel
• Companies: University of Minnesota and University of Minnesota
• Keywords: biomarkers ; predictive model ; two-stage design ; ROC curve

Abstract

We consider study design for a two-stage study to develop and validate a panel of biomarkers for predicting prostate cancer recurrence. In stage one, a predictive model for prostate cancer recurrence is developed using a set of candidate biomarkers. The study will be allowed to terminate for futility after stage one if initial estimates of prognostic accuracy are unacceptable. If initial estimates of prognostic accuracy are promising, the prognostic accuracy of the predictive model is evaluated using a set of independent samples in stage two. We present results from a simulation study to evaluate the effect of design parameters (the proportion of samples used in stage one and the cutoff for early termination) and marker parameters (number of markers truly associated with prostate cancer recurrence, correlation between markers, etc.) on the type-I error rate, power and expected sample size.