Application of Multiple Imputation in Analysis of Data from Clinical Trials with Treatment Related Dropouts

Abstract

Many longitudinal clinical studies suffer from nonignorable dropouts. Ramakrishnan and Wang (2005) proposed a mixed effects analysis treating missing data as missing due to truncation (MDT), estimated the parameters under a multivariate truncated normal model, and suggested an adjustment to the degrees of freedom to account for the implicit estimation of the missing data. We propose a multiple imputation (MI) in conjunction with the MDT method as an alternative to accurately accommodate the uncertainty introduced by imputation. The data used in Ramakrishnan and Wang (2005) is considered for illustration. A comparison of various methods using a simulation study is presented.

Keywords:

• Missing due to truncation
• Multiple imputation
• Nonignorable missing

Mathematics Subject Classification:

• 6207
• 62F03