gms | German Medical Science

Platform trials allow the evaluation of multiple experimental treatments under a single master protocol; they are designed to allow new treatments to be seamlessly added to the trial as they become available. They provide large potential gains in efficiency through 1) comparing arms against a common control group; 2) allowing an adaptive design to drop less promising experimental arms as outcome data is collected; 3) adding new treatments to an existing trial instead of starting a new trial from scratch. With these efficiencies comes additional complexities in the practical and statistical aspects of the trial.

In this talk I will discuss some recent statistical work that aims to improve the statistical robustness and efficiency of platform clinical trials.

The first area of work is error rate control. In an ongoing platform trial, it is challenging to control the overall error rate of the trial, due to the number of hypotheses to be tested not being known at the start of the trial. Recent work in statistical theory has proposed methods for ‘online control of error rates’. These methods allow control of the false discovery rate and family-wise error rate in settings where hypotheses are tested sequentially and it is unknown how many will be tested in the experiment. I will discuss how these methods perform when applied to platform clinical trial settings.

The second area of work is on when new treatment arms should be added in. Adding in arms is generally advantageous compared to starting a new trial; nevertheless, adding an arm may have negative consequences on the evaluation of existing treatments in the trial. I will discuss a decision-theoretic approach to considering when it is beneficial to add in a new arm.

I will finish by considering other areas of methodology research that are required to enable maximum utility of the platform trial approach.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.