The Real Implication of Incorporating PK Into Dose-Escalation Trials

• Activity Number: 351
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Cheng Zheng*+ and Lu-May Chiang and Pai-Hsi Huang
• Companies: Novartis and Novartis and Novartis
• Keywords: Bayesian hierarchical model ; dose-escalation ; pharmacokinetic ; Phase I

Abstract

Extensive research has been conducted using Bayesian techniques for modeling dose-toxicity relationship in oncology dose-escalation trials. Many have attempted to integrate pharmacokinetic (PK) information, such as drug exposure measured by area under the curve (AUC), into the model. However, given these proposals were rarely formally evaluated, the implication of such practice in real trials remains unknown. We are proposing a Bayesian hierarchical model considering both the PK exposure-toxicity curve and the relationship between dosage and PK. This model can be fitted via Gibbs sampling and is very flexible in addressing safety concerns. We have studied this model in details based on data from 5 past or ongoing Novartis Phase I trials with large number of enrolled patients and well-collected PK profiles. After comparing the new model with the standard Bayesian model for in-house dose-escalation trials, we propose conditions under which incorporating PK information is beneficial during dose-escalation in terms of both maximum tolerated dose (MTD) estimate accuracy and safety profile.