Learn on sample-size re-estimation, conditional power and the promising-zone method.
In this webinar, we will play the part of a sponsor consulting a statistician with the goal of designing a randomized trial with sample-size re-estimation. We will discuss the concept of conditional power and the promising-zone method, and we will present a pragmatic view of some issues, including when to plan the sample-size re-estimation in the course of the accrual, how to select the amount of sample-size increase, and how to operate such trials in practice. We will also discuss the pros- and cons- of using this type of design in relation to the clinical setting in which the trial is designed and endpoints it uses. Moreover, we will discuss the regulators’ views and illustrate our discussion with practical examples.
KEY TAKEWAYS:
- Be exposed to the concepts of sample-size re-estimation, conditional power, and the promising-zone method
- Gain insights into the planning of sample-size re-estimation and the operationalization of such designs
- Understand the benefits and limitations of sample size re-estimation in relation to the clinical setting in which the study is conducted
It is often the case in drug development, particularly in oncology, that a randomized trial is launched with considerable uncertainty about the final sample size required to demonstrate a clinically meaningful and statistically significant improvement in results over the control treatment. Group-sequential designs have an established role in randomized trials and allow for safety monitoring as well as early decisions regarding efficacy and futility of the new treatment. In some trials, it may be interesting to foresee the possibility of adapting the final sample size if results at an interim analysis suggest the possibility of a positive trial, albeit with a less pronounced treatment effect than initially anticipated. Such adaptive designs that allow for sample-size re-estimation have gained traction in recent years. However, they pose several issues that require consideration and advanced statistical input.
SPEAKERS:
