Using a Predictive Probability Design for Faster Decision-Making in an Early-Phase Oncology Trial

• Activity Number: 659
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Sharon Murray* and Grace Zhang and Allison Florance
• Companies: PAREXEL International and Merck and GlaxoSmithKline
• Keywords: oncology ; predictive probability ; adaptive design ; clinical trials

Abstract

In early-phase oncology trials, it is common to have two parts to the first-time-in-human (FTIH) trial. In the first part, a dose matching the targeted toxicity profile is identified. This is followed by an expansion arm to confirm the toxicity and examine efficacy before starting a Phase II trial. The expansion arm could consist of a single small cohort of subjects or could include review of two or more cohorts with start of the later cohort(s) based on observed efficacy, essentially treating the expansion arm as a small single-arm Phase II trial. At GlaxoSmithKline, we are using the Predictive Probability Design of Lee and Liu (Clinical Trials, 2008) to continually examine efficacy results during the expansion arm once a minimum number of subjects has been enrolled. This design is more likely to stop early for futility in the case of a non-efficacious drug than a two-cohort design and provides more information on efficacy than a small single cohort. This presentation will review our recent experiences using this design for several FTIH trials. Details will include the operating characteristics of the design, experiences working with clinical teams, and trial outcomes.