Evaluation of Multiple Imputation in a Vaccine Immunogenicity Trial

• Activity Number: 113
• Type: Contributed
• Date/Time: Monday, July 30, 2007 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Elizabeth Zell*+ and Michela Baccini and Constantine E. Frangakis and Fan Li and Fabrizia Mealli and Brian D. Plikaytis and Charles E. Rose and Donald B. Rubin
• Companies: Centers for Disease Control and Prevention and University of Florence and Johns Hopkins University and Harvard Medical School and University of Florence and Centers for Disease Control and Prevention and Centers for Disease Control and Prevention and Harvard University
• Keywords: Missing Data ; Clinical Trials ; Multiple Imputation

Abstract

In randomized vaccine trials that require repeated visits, it is inevitable that some people will miss some visits. Data should be collected despite these protocol violations, but it is difficult to obtain immunogenicity data (antibody levels) or reactogenicity data (side effects) for missed visits. Some sort of imputation is required if ITT analyses are to be done using all randomized subjects. Thus it is desirable to impute the missing data. Multiple imputation has been shown to be valid under a broad range of circumstances. CDC's current Anthrax Vaccine Trial presents challenges because of the large number of measurements per person (~2000) and the limited number of subjects in each treatment arm (~200). Here we present preliminary results from a large-scale evaluation of the validity of the multiple imputations in this trial, which were created using a new state of the art procedure.