A flexible design for multiple armed screening trials

The goal of a screening trial is to examine several treatments with the intention of selecting one or more treatment(s) for further testing. Such a trial frequently occurs in the phase II setting, with the intention that the selected treatment be formally compared to a standard treatment in a phase III clinical trial. In this framework it is not essential that the very best treatment is definitely selected (such a decision would require a formal phase III trial), rather it is important that a substantially inferior treatment is not selected when a superior treatment exists. The level of evidence required to select a treatment based on this standard is lower than the evidence required to definitively order treatments. This paper introduces a flexible design for such trials, the flexibility coming from the use of a particular decision rule. The proposed rule states that if the observed difference in the success rates of the treatments is larger than some prespecified quantity d, then the treatment with the highest observed success rate is selected. However, if the observed difference is less than or equal to d, other factors may be considered in the selection. The goal of this design is to mirror clinical practice, where a treatment's success probability is often only one of many considerations in determining a treatment recommendation for a particular patient.