Bayesian dose-escalation models for combination phase I trials in oncology
View Presentation View Presentation
*Suman Sen, Novartis Pharmaceuticals Corporation  Stuart Bailey, Novartis Pharma  Beat Neuenschwander, Novartis Pharmaceuticals 

Keywords: Bayesian models, dose-escalation, oncology, historical data, prior

Within oncology there has been a paradigm shift moving from single agent therapies to combinations of drugs. In this setting there is no longer one MTD but a range of dose pairs. It is important to understand the impact of combining two or more compounds on patient safety, and to define the toxicity profile interaction. We review existing approaches and present an extension incorporating interaction on the odds-scale to address potential synergistic or antagonistic effects on safety. The focus is on practical and methodological issues, covering a range of phase Ib studies. We reflect on the use of single-agent data to support starting combinations, and discuss experiences with implementation of such a design in real studies.Critical aspects regarding the statistical model, appropriate specification of prior information, and monitoring of patient risk are discussed.