Comparing Candidate General Surrogates of Protection

• Activity Number: 321
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Erin Gabriel*+ and M. Elizabeth Halloran
• Companies: Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
• Keywords: Surrogates ; Causal inference ; Meta-analysis ; Two-phase Sampling

Abstract

An immune measure that accurately predicts efficacy in a new setting is a general surrogate of protection (GSoP) (Gilbert et al. 2008, Statistics in Medicine). Evaluating a GSoP aims to estimate the causal association between the treatment effect on the GSoP and the treatment effect on the clinical endpoint among a set of similar trials, then use this information to estimate the clinical effect in a trial measuring only the GSoP. Most methods of evaluating a GSoP, such as Gail et al., 2000, Biostatistics, focus on esti- mating a point and interval for the clinical treatment effect in a new setting, rather than comparing candidate GSoPs. We propose using the absolute error statistic (Tian et al., 2007 Biometrika) to compare candidate GSoPs within a set of trials to determine the best candidate GSoP. We extend the concept of GSoP to consider functions of any set of post-randomization vari- ables. Our proposed methods also allow for differing levels of data collection, where previous methods assume access to all individual level data in each trial and do not address missing or two-phase sampled data. We use surrogates of vaccine protection to motivate our methods.