ABSTRACT
In terms of multiregional clinical trial (MRCT) for a common disease, conducting with a fixed design may be sufficient, but utilizing the available regional information (e.g., regional treatment effect and standard deviation) from historical trials or early stage data at the design stage can also be a key to success. The proposed optimal designs in Chapter 10 can handle this scenario. However, if no historical information is available for a disease, an adaptive design may be preferable by applying different reallocation
CONTENTS
14.1 Introduction .............................................................................................. 177 14.2 Challenges of Adaptive Design for MRCT ........................................... 178 14.3 Statistical Framework .............................................................................. 179
14.3.1 Test Statistics ............................................................................... 179 14.3.2 Conditional Power ...................................................................... 180 14.3.3 Conditional Assurance Probability ......................................... 180 14.3.4 Conditional Success Rate .......................................................... 182
14.4 Adaptive Strategies .................................................................................. 183 14.4.1 Step 1: Initial Sample Size Planning ........................................ 184 14.4.2 Step 2: Decision Making and Sample Size Adaptation
at Interim ..................................................................................... 184 14.4.3 Step 3: Final Decision ................................................................. 186
14.5 Comparisons between Proposed Adaptive Design and Classical Design for MRCT ..................................................................................... 187 14.5.1 Initial Sample Size Planning .................................................... 187 14.5.2 Overall Type I Error Rate Control Simulation: Design
and Results .................................................................................. 188 14.5.3 Power and Success Rate Performance Simulation:
Design and Results..................................................................... 189 14.6 Discussion and Summary ....................................................................... 190 References ............................................................................................................. 191
ratios to different regions and possibly by increasing the sample size after an interim analysis of the ongoing trial. Chow and Chang (2011) define an adaptive design of a clinical trial as a design that allows adaptations or modifications to some aspects of the trial after its initiation without undermining the validity and integrity of the trial. The commonly used adaptive designs were nicely introduced in Chang (2014) and Chow et al. (2011). The sample size planning for MRCT was also discussed in Hung et al. (2010), Quan et al. (2010), Chen et al. (2012b), and Uesaka (2009). Even though the topics of both MRCT and adaptive design are not new, there are only a very limited amount of adaptive designs proposed for MRCT. Luo et al. (2010) proposed an optimal adaptive design for MRCT in which an optional supplemental stage of the MRCT may be needed to provide additional data to address local regulation requirements. Chen et al. (2012a) proposed an adaptive strategy to identify an imbalanced factor or an effect modifier based on the blinded data and also provided a stratified analysis method to adjust for the imbalanced factor. In this chapter, we propose an unblinded region-level adaptive design to perform sample size recalculation and reallocation at interim based on the observed values of each region. Both the entire MRCT and each individual region are allowed to be stopped at interim for efficacy or futility, or to continue to the next stage with sample size reallocation and recalculation which are performed to ensure certain conditional overall power and conditional probability of success for each region of interest.
