Adaptive Intra Patient Escalation Design for Phase I Trials in Oncology

• Activity Number: 187
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Laura Levette Fernandes*+ and Jeremy Michael George Taylor
• Companies: University of Michigan and University of Michigan
• Keywords: adaptive design ; dose ranging ; Markov model ; Bayesian analysis. ; oncology ; stochastic

Abstract

Phase I Dose escalation studies in oncology focus on finding the maximum tolerable dose (MTD) that gives an acceptable levels of toxicity. Often times the therapy is given in multiple cycles with breaks between the cycles. Usually the dose assignment for a patient does not change from one cycle to the next and the toxicity is determined from the first cycle only or from the whole regimen. An alternative design is to allow intra-patient variation in dose and to assess the toxicity separately in each cycle. In this scheme the next dose assigned to a patient would be escalated or deescalated based on past responses from this and other patients and driven by a statistical model. We propose a stochastic model for analyzing such data. The model is a 3 parameter Markov model in which the probability of toxicity at cycle j given no prior toxicity is a function of the cumulative dose, the current dose and the maximum of the previous doses administered to the patient. The data is then analyzed in a Bayesian way using Winbugs. Correct modeling of the data and proper estimation of the parameters enables projecting the outcomes of the trial when different doses are assigned at the next stage.