Multiple Partial Imputation for Longitudinal Data with Missing Values in Clinical Trials

Biomedical research is plagued with problems of missing data, especially in clinical trials of medical and behavioral therapies adopting longitudinal design. After a comprehensive literature review on modeling incomplete longitudinal data based on the full-likelihood functions, this paper proposes a set of imputation-based strategies for implementing three advanced models for handling intermittent missing values and dropouts that are potentially nonignorable according to various criteria. In multiple partial imputation (MPI), intermittent missing values are first imputed several times. Then, each partially imputed data set is analyzed using selection, pattern-mixture, or shared-parameter models to deal with dropouts. If imputations are additionally created for dropouts, it is a 2-stage version of MPI. Depending on models used for making imputations, various strategies can offer a framework for analyzing the sensitivity of parameter estimation to the assumptions of the missingness mechanism. For illustration, both continuous and dichotomized binary data from a smoking cessation clinical trial are analyzed. Both likelihood-based and Markov Chain Monte Carlo (MCMC) based inferences are also described