Assessing missing data impact in a clinical trial prior to unblinding using a parametric bootstrap approach

• Activity Number: 223
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jianing Di*+
• Companies: Janssen Alzheimer Immunotherapy R&D, LLC
• Keywords: missing data ; type I error rate ; parametric bootstrap ; mixed model for repeated measures ; simulation

Abstract

The problem of missing data is frequently encountered in clinical studies. The potential impact of missing data ranges from estimation inefficiency to estimation bias/invalidity. In practice, to assess the robustness of the primary efficacy analysis method to missing data, sensitivity analyses are conducted after data unblinding. This paper discusses an alternative simulation-based framework that can be used to assess the missing data impact by applying the primary method to data generated with different characteristics. The proposed approach can be used prior to data unblinding to evaluate the missing data impact on any metrics or statistical methods. An example of using such framework to assess the type I error rate of the mixed model for repeated measures in a parallel-group study is used to illustrate the methodology.