To borrow or not to borrow: on non-concurrent control data in platform trials

Music City Center 

Platform trials are a type of clinical trial in which multiple treatments can be evaluated simultaneously using a single infrastructure. However, implementing control arms in platform trials can be challenging since new treatments are continually added and tested over time. Non-concurrent control (NCC) in platform trials refers to the use of a common control group for multiple experimental arms that are not evaluated at the same time. Although it is appealing to borrow NCC data to evaluate treatment effect for the benefits of faster accrual, flexibility and improved statistical efficiency, regulatory acceptance of NCC analysis results may be hindered by bias arising from lack of randomization. In this presentation, I will discuss the scenarios when borrowing NCC data in statistical analysis is appropriate, and the statistical methods that can be used to minimize statistical bias. In particular, when the outcome of interest is a time-to-event endpoint, I will introduce a novel approach to borrow the concurrent observation part of the NCC data by left truncation using a simple decision-making flowchart and with the use of restricted mean survival time for robust statistical inference.

platform trial

Non-concurrent control

Time-to-event

restricted mean survival time