Keywords: basket trial; Type III error; umbrella trial
We consider a simplified exploratory master protocol that intends to investigate multiple test drugs in the same tumor indication (i.e., an umbrella trial) or investigate the same test drug in multiple tumor indications (i.e., a basket trial).
The primary objective of an exploratory umbrella trial is to test whether ANY of the drugs is effective, whereas individual drug cohorts are analyzed independently. In contrast, the primary objective of an exploratory basket trial is to test whether the test drug is effective in ANY of the tumor indications, whereas data are subject to pooling. The distinction is less explored in literature and is often overlooked in practice. In this presentation, we illustrate it with the pruning-and-pooing method that was first proposed for designing confirmatory basket trials (Chen et al. 2016). The method is extended to address a high-level strategic question: how many cohorts should be considered in each master protocol under resource constraint?