Paper SP11 Incremental Methods of Imputation for Longitudinal Data with Informative

In longitudinal clinical trials, missing data are mostly related to dropouts. Some dropouts appear completely at random. The source for other dropouts is withdrawal from trials due to lack of efficacy. One of the approaches to comply with the intent-to-treat principle is the imputation of incomplete data. When the dropout process is related to the outcome process (informative dropouts), it creates tremendous challenges in analyzing such data. No commercial software currently considers the dropout mechanisms in dealing with informative or non-random dropout. Consequently, the results can be biased and misleading. This paper deals with the incremental methods of imputation applied to incomplete longitudinal data sets with informative dropouts. It is shown by simulation that the incremental methods are more precise than some other imputation methods (including the last observation carried forward method, the multiple imputation method, mixed models). The drop-out mechanism depends on individual previously and currently observed values of a secondary endpoint correlated with the primary endpoint