Analyzing Incomplete Longitudinal Clinical Trial Data

• Activity Number: 157
• Type: Invited
• Date/Time: Monday, August 4, 2003 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Geert Molenberghs*+
• Companies: Limburgs University Centrum
• Keywords: missing data ; clinical trials ; LOCF ; imputation ; likelihood

Abstract

Using standard missing data taxonomy, largely due to Rubin and coworkers, and using simple algebraic derivations, it is argued that some simple but commonly used methods to handle incomplete longitudinal clinical trial data, such as complete case analyses and methods based on last observation carried forward, are poorly principled and restrictive. Given the availability of flexible software for analyzing longitudinal sequences of unequal length, it is argued there is not even a computational reason for not shifting to a likelihood-based ignorable analysis. Such analyses are valid under the much weaker assumption of MAR. While the occurrence of MNAR missingness cannot be ruled out, it is argued that such analyses are, themselves, surrounded with problems and therefore, rather than either forgetting about them or blindly shifting to them, their optimal place is within a sensitivity analysis. The concepts developed here are exemplified using data from three clinical trials, where it is shown that shifting the analysis method may have an impact on the conclusions of the study.