Extension of Bayesian Logistic Regression Model (BLRM) for Dose Timing Selection in Oncology Phase I Combination Studies

• Activity Number: 247 - Clinical Trial Design- 2
• Type: Contributed
• Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yiyun Zhang* and Nigel Yateman and Fang Xiang and Lan Yi and Kapildeb Sen and Beat Neuenschwander
• Companies: Novartis and Novartis and Novartis and Novartis and Novartis and Novartis
• Keywords: Bayesian logistic regression model (BLRM); Escalation with overdose control (EWOC); Clinical trial; Dose escalation; Meta-analytic-predictive (MAP); Oncology

Abstract

Adaptive Bayesian logistic regression model (BLRM) guided by the escalation with overdose control (EWOC) principle has been used to make dose escalation recommendations and estimate the recommended dose for expansion (Neuenschwander et al 2014).

We consider a study design where the optimal timing of treatment initiation is the primary interest: Treatment A is administered to the patient. The other supportive treatment B will be started at certain time after treatment A. The synergy between the 2 treatments is expected to improve the efficacy, but also potentially lead to additional toxicity compared with monotherapy with treatment A. A modified BLRM is proposed to model the additional toxicity due to treatment B by modeling the dose-timing vs toxicity in the BLRM framework. This approach is an extension of the current BLRM and can be easily adapted to other situations where the dose-toxicity relationship is different to that assumed in the classical BLRM. A meta-analytic-predictive (MAP) approach (Neuenschwander et al 2010) can be used to derive the prior distribution of the BLRM parameters.

The method will be illustrated in an oncology Phase 1 combination study.