A Unified Bayesian Framework for Assessing Predictive Probability of Trial Success

• Activity Number: 153 - Contributed Poster Presentations: Biopharmaceutical Section
• Type: Contributed
• Date/Time: Monday, August 8, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Archie Sachdeva* and Ming Zhou and Ram Tiwari
• Companies: University of Florida and Bristol Myers Squibb and Bristol Myers Squibb
• Keywords: Bayesian analysis; Probability of trial success; Oncology; Surrogate endpoints; phase 3 go/no go decision

Abstract

Data-driven decision-making plays a key role in drug development. One important decision-making tool is the probability of trial Success (POS), which estimates the likelihood of a successful trial in the future. In this project, we aim to i) improve POS assessment based on information from a surrogate endpoint or historical data from multiple trials; ii) provide improved and bias-adjusted treatment effect estimate for phase 2 study results. First, we consider POS calculation based on Phase 2 data from the same endpoint. Second, we extend the method to incorporate a surrogate endpoint. Third, we use Bayesian meta-analytic approach to incorporate information from multiple historical studies to help improve treatment effect estimation and POS assessment. Furthermore, we outline a unified Bayesian framework for POS assessment, which can potentially incorporate multiple endpoints, multiple studies with multi-arm/dose, and multiple compounds. We illustrate POS assessment using a few examples from oncology and immunology studies. In addition, we also developed software packages for study teams to use.