Finding the Right Patient Population for a Drug
*Richard Simon, National Cancer Institute 


The development of molecularly targeted treatments has stimulated the use of companion diagnostics for identifying the patients most likely to benefit from a new treatment and has enhanced the role of enrichment designs in which eligibility is restricted based on the diagnostic. Enrichment designs can provide substantial reductions in sample size compared to more traditional broad eligibility clinical trials, can result in larger treatment effects and have been used in recent therapeutic successes in oncology (1). In cases where a companion diagnostic with compelling biological relevance has not been developed by the start of a pivotal phase III trial, the sponsors and investigators have several design options including: (i) A traditional broad eligibility design with post-hoc subset analysis; (ii) An adaptive signature design with broad eligibility and cross-validated classifier development at the end of the trial (2); (iii) An adaptive enrichment design (3-7); (iv) Delaying initiation of the phase III trial to do more phase II studies. I will discuss some of the relative merits of these approaches. Traditional broad eligibility designs may result in intended use populations for which relatively few patients benefit. Attempting to discover and confirm on the same dataset requires special care to avoid biases and inflation of type I error and makes explicit the uncertainty about “who benefits” that is hidden in standard confirmatory trials.