Using Bayesian Methods to Incorporate Historical Information on Control Rate and Treatment Effect: An HIV-Paediatric Trial

• Activity Number: 140 - The Challenges and Advantages of Utilizing Bayesian Statistical Methodology in Extrapolation of Adult Use Data to Pediatric Study Designs and Evaluation
• Type: Invited
• Date/Time: Monday, July 31, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Clara Dominguez Islas* and Nicky Best and Rebecca Turner and Naomi Givens and Rimgaile Urbaityte and Sophie Barthel and Adrian Mander
• Companies: MRC Biostatistics Unit and GlaxoSmithKline and MRC Biostatistics Unit and GlaxoSmithKline and GlaxoSmithKline and GlaxoSmithKline and MRC Biostatistics Unit
• Keywords: Paediatric trials ; Bayesian analysis ; Extrapolation ; Historical controls ; Prior-data conflict

Abstract

Incorporating historical information in the design and analysis of a paediatric trial could reduce the number of subjects required. Our motivating example is a non-inferiority trial on an experimental treatment for the maintenance of viral load suppression in HIV-1 infected children. We see two main challenges: to combine and synthesize diverse sources of historical information and to incorporate the resulting information to the new trial in a robust way, allowing for potential conflict. In addition, we are interested on using historical information both on the rate of positive response in the control arm and on the treatment effect as estimated from an adult population. For this, we propose an extension of the meta-analytic-predictive priors approach proposed by Schmidli et al. (2014), in the form of a mixture of bivariate prior distributions. In this presentation we will discuss different ways of robustifying an informative empirical prior (by mixing it with less informative priors), to allow for potential conflict. We will show results from a simulation study that evaluate the operating characteristics of the proposed methods and compare it to other suggested approaches.