ABSTRACT
Personalized medicine aims to deliver the right drug to the right patient. In addition to discovering the right drug, nding the right dose, and identifying the right patient, it’s also desirable to shorten the time of drug development in order to bring the drug to the patient faster. Different clinical trial designs have been proposed for this purpose. One such design is a seamless phase II/III drop-the-losers (or pick-the-winner) design, which has the potential to terminate the inferior treatment groups (i.e., the “losers”) early if no efcacy is shown. It minimizes “white space” between phase II and
CONTENTS
5.1 Introduction ................................................................................................ 129 5.2 Motivations and Concepts ........................................................................ 131 5.3 Issues in Conventional One-Level Correlation Model ......................... 132 5.4 Two-Stage Winner Design with Proposed Model................................. 133
5.4.1 Two-Stage Winner Design ............................................................ 133 5.4.2 Two-Level Correlation Model for Two-Stage
Winner Design ............................................................................ 134 5.4.3 Test Statistic and Its Distribution................................................. 135 5.4.4 Type I Error Rate Control .............................................................. 137
5.5 Performance Evaluation of Two-Stage Winner Design under Proposed Model ............................................................................. 138 5.5.1 uX and uY Are Linearly Related .................................................... 138 5.5.2 uX and uY Are Not Linearly Related ............................................ 140
5.6 Parameter Estimation ................................................................................ 142 5.7 Discussion and Summary ........................................................................ 143
5.7.1 Distribution of Final Test Statistic ............................................... 145 5.7.2 R Code for Calculation of Stopping Boundaries ....................... 146 5.7.3 Derivation Details for σ rj2
.............................................................. 147 Acknowledgments .............................................................................................. 147 References ............................................................................................................. 147
phase III of the studies, and efciently uses all the patient data both in the learning and conrming phases.
