| The development of a diagnostic biomarker is a several phase process. Candidate markers are identified in marker discovery and their performance is evaluated over several stages of studies. The need to conserve scarce resources, specimens, for example, while fully evaluating the performance of a candidate marker motivates the use of group sequential study designs when evaluating the performance of a diagnostic biomarker. This dissertation considers issues related to group sequential diagnostic biomarker studies. We begin by developing thoeretical results that allow existing group sequential methodology to be applied in the diagnostic testing setting. Sequential empirical process theory is used to investigate the asymptotic properties of the sequential empirical ROC, PPV and NPV curves. Next, we consider estimation after a two-stage diagnostic biomarker study that allows early termination for futility. Estimators and confidence intervals that correct for the bias caused by the possibility of early termination are presented and used to estimate ROC(t) and PPV(u) after a two-stage diagnostic biomarker study that allows early termination for futility. Finally, we consider the design of the PCA3 and proPSA validation studies. The operating characteristics of various two-stage designs are investigated by simulation and the results are used to draw conclusions about the design of two-stage diagnostic biomarker studies that allow early termination for futility. |
