Test of Acceptable Vaccine Immunogenicity: t-Test vs. Longitudinal Data Analysis Model

• Activity Number: 324
• Type: Contributed
• Date/Time: Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Xiaoming Li*+ and Frank Liu and Devan V. Mehrotra
• Companies: Merck Research Laboratories and Merck Research Laboratories and Merck Research Laboratories
• Keywords: Immunogenicity ; Vaccine ; Acceptability ; Longitudinal data analysis

Abstract

Immunogenicity studies often have a 2-fold goal: 1) to compare a study vaccine vs. a control in terms of geo. mean titer ratio post vaccination, and 2) to test the acceptability of the study vaccine as measured by geo. mean fold-rise from pre to post vaccination. For goal 1) the longitudinal data analysis (LDA) model by Liang and Zeger (2000) is often used, for 2) paired t-test is commonly used. However, there are a few potential drawbacks in the paired t-test: 1) subjects with any missing data are excluded, 2) if the missing data mechanism deviates from missing completely at random, the naive estimate of fold-rise may be biased, and 3) the paired t-test may not be able to adjust for a prognostic factor. We will investigate the potential gains by using the LDA model (based on data from only the study vaccine arm or both arms) for the acceptability test as compared to the paired t-test.