Two Examples on the Use of Simulations to Guide Model Selections in MMRM Analyses

• Activity Number: 652
• Type: Topic Contributed
• Date/Time: Thursday, August 4, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Peiling Yang*+ and Phillip Dinh
• Companies: U.S. Food and Drug Administration and U.S. Food and Drug Administration
• Keywords: Missing Data ; MMRM ; Simulation

Abstract

Mixed model for repeated measures (MMRM) offers an alternative approach for dealing with missing data in longitudinal studies. The performance of the method has been demonstrated to be superior to the traditional ANCOVA analysis with missing data imputed by the Last Observation Carried Forward (LOCF) method under various scenarios. (See, for example, Mallinckrodt, et al. (2001, 2004), Molenberghs, et al. (2004), Lane, (2008), Siddiqui, et al. (2009)). However, MMRM analyses often have extra levels of complexity in terms of model specifications. In this talk, simulations will be used to guide the model specifications in two examples: 1. how to deal with titration visits in MMRM analyses and 2. how to handle baseline scores in MMRM analyses. Our simulations suggest that one should include all data (including titration visits) and should include a treatment-by-baseline interaction in an MMRM analysis. Examples from real schizophrenia trials will be used to contrast different methods.