A Bayesian Approach in Assessing Surrogacy of Clinical Trial Endpoints

• Activity Number: 473
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Manuela Buzoianu, PhD*+ and Shengyan Hong
• Companies: FDA and MedImmune
• Keywords: oncology clinical trial ; surrogate endpoint ; Bayesian analysis ; hazard ratio ; error-in-variable model ; Markov Chain Monte Carlo

Abstract

Using a surrogate endpoint in a clinical trial is common when it is difficult to use the definitive one. In particular, in oncology trials overall survival (OS), the gold standard for showing clinical benefit, requires large randomized studies with long follow-up and is potentially confounded by the effect of subsequent therapies to study drug. Progression-free survival (PFS) is an attractive endpoint because it requires shorter studies and is not influenced by subsequent lines of therapies. Whether and to what extent an improvement in PFS results in an improvement in OS remain to be evaluated depending on disease and treatment settings. A Bayesian meta-analysis approach is employed to assess trial-level surrogacy of PFS for OS, in particular to show how the treatment effect (hazard ratio) on PFS predicts the treatment effect (hazard ratio) on OS in first line advanced non-small cell lung cancer. Thus, a Bayesian error-in-variable model is built to account for the variability in the hazard ratio estimates from different clinical trials. In addition, posterior inference is made to evaluate the relationship between PFS hazard ratio and OS hazard ratio in different treatment settings.