Missing Data Methodology: Past, Present, and Future Trends

• Activity Number: 391
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Cristina Labrador Sotto*+
• Companies: I-Biostat
• Keywords: Direct Likelihood ; Imputation ; Sensitivity Analysis

Abstract

Incompleteness in data arising from clinical trials often occurs, due largely to the longitudinal nature of such studies. In the past, simplistic approaches, e.g., complete case analysis and last observation carried forward, have been used, primarily for their straightforward implementation. Over the years, however, research on such methods has drawn attention to their less-than-ideal characteristics, at least under scenarios that are often inconsistent with clinical trial data. Thus, there has been a conscious shift towards more appropriate methodology in recent decades, with computing technology that enable implementation using standard software. In the future, methodology for missing data can only be expected to develop even further, providing the researcher with an ample repository of tools to address incompleteness. An overview of existing methods for missing data will be presented.