gms | German Medical Science

Flexible designs are an attractive option for clinical trials since they provide the possibility to react to unforeseen events during an ongoing trial while still controlling the type I error rate. Additionally, the application of an adaptive design may lead to a decrease in average sample size and, therefore, to ethical and economic benefits.

During the planning stage, adaptive designs are often formulated as classical group-sequential designs (i.e., based on a combination test), and it is intended to recalculate the sample size following conditional power considerations. However, the pre-specification of a group-sequential design that may actually never be realized is counterintuitive and may be inefficient. Instead, an adaptive design can be chosen to optimize a trial-specific objective criterion as, e.g., the expected sample size under constraints on the type I error rate, overall power, and conditional power. The resulting designs are an appealing choice in clinical practice due to their optimal performance characteristics.

The freely available R-package adoptr allows its users to compute optimal two-stage designs for a user-specific objective function under a user-specific set of constraints. Furthermore, the underlying data distribution can be defined by choosing an appropriate prior for the effect size of interest. Due to its flexibility, the software is an exciting option to design a clinical trial since the optimization problem can be formulated according to the specific needs of the respective trial.

In this talk, it is illustrated for various clinical trial examples how an optimal two-stage design can be derived by means of adoptr. In particular, the options regarding the data distribution of the endpoint and the possibility of extending the package's functionality by any user are presented. Furthermore, we discuss the issue of when an unplanned design adaptation may become necessary and how one should proceed in this case.

The authors declare that they have no competing interests.

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