An Example of Simulation-Based Calculations of the Probability of Technical Success

• Activity Number: 386
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Committee on Applied Statisticians
• Author(s): Neal Thomas*+
• Companies: Pfizer Inc.
• Keywords: Bayes ; MCMC ; Dose Response ; Posterior predictive distribution

Abstract

Simulation-based Bayesian analyses of a dose ranging study with binary longitudinal data were used to make decisions about the doses to be included in a second study. MCMC methods implemented using OpenBUGS software were used to fit an Emax model with timing-varying parameters to binary longitudinal data. A second study was required to better estimate the dose response curve and locate the lower portion of the dosing range with insufficient efficacy. A limited number of lower doses were available for the second study without an extended delay to manufacture new lower doses, so a key question was whether to proceed with existing doses or delay. Analyses based on the MCMC model output were used to compute the probability that a dose from the existing drug supply is below the ED50. Posterior predictive distributions of the outcomes of a new study were also produced to assess the likelihood that the trial would produce sufficient differentiation between the lower and higher doses.