gms | German Medical Science

Platform studies are considered to increase both, operational and statistical efficiency. Operational efficiency is achieved through a flexible study design which allows to test multiple interventions in possibly multiple sub-populations under one overarching master protocol. The design process of a platform study will typically be of higher complexity and longer in duration as compared to designing a single standard clinical study. Operational efficiencies of the platform approach increase with an increasing number of tested interventions in the study, due to decreasing required time for study start up, internal governance and regulatory interactions for each new intervention. Larger statistical efficiency as compared to standard clinical study designs is achieved by sharing of control data. Instead of running 4 studies with a 1:1 randomization, one study with 2:1:1:1:1 randomization would require less patients in total and on the control group to reach the same power for each individual active arm. Depending on the objectives of the platform study and the similarity of interventions under investigation, additional statistical efficiency can be generated by combining data from the study interventions. This may be of particular relevance in Phase 2 platform studies with the aim to investigate new mechanisms of action and to establish signs of activity for the tested interventions in the new mechanisms. Optimized contrast testing approaches (e.g. MCPMod) have been introduced as an efficient alternative to standard pairwise comparisons in multi-armed dose-finding studies. Generalizations of the contrast testing approach to allow efficient decision making in platform studies will be introduced in this presentation. Connections, benefits and limitations as compared to the simple sharing of control data and the more complex hierarchical modelling of treatment effects will be discussed.

The authors declare that they have no competing interests.

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