Exact Group Sequential Designs for Phase I Oncology Trials

• Activity Number: 70
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Karen Messer*+
• Companies: Moores UCSD Cancer Center
• Keywords: Phase I trail ; group sequential ; exact test

Abstract

The standard Phase I oncology trial is designed to find the maximum tolerated dose of an agent in a setting where serious drug-related toxicity is expected. However, newer targeted anti-cancer agents may act without increasing the background rate of adverse events. In this setting, formal 'toxicity-evaluation' designs have been proposed, based the traditional 3+3 design. (Messer et al, 2010) These enable tests of hypothesis and estimates of toxicity rates using a simple 3+3 schema. For example, if 4 cohorts have received the proposed therapeutic dose without stopping in the 3+3 design, then a formal test of hypothesis will conclude that the toxicity rate is below 33% with 95% confidence. In this talk we compare these 3+3 toxicity-evaluation designs to exact group sequential tests, which are more complex but also more efficient. We show in many settings there is not much to be gained, but for low toxicity rates the added complexity may be worthwhile. We recommend use of the simpler designs in many settings.