Analysis of a 2-by-2 Crossover Trial of Binary Data with Missing Values Using Multiple Imputation

• Activity Number: 413
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section for Statistical Programmers and Analysts
• Author(s): Junxiang Luo*+ and Timothy Costigan and Eileen Brown
• Companies: Eli Lilly and Company and Eli Lilly and Company and Eli Lilly and Company
• Keywords: missing data ; multiple imputation ; nonparametric ; binary ; crossover

Abstract

Prescott's test is one of powerful nonparametric-like tests for testing the null hypothesis of no treatment difference in a 2×2 cross-over trial without giving any estimate of the treatment effect size. Recently, Schouten proposed a new simple test by averaging the treatment differences over the sequences. This test gives the estimate of the treatment effect size and provides attractive testing power. But both of the two tests are implemented on subjects with completed observations by ignoring missing data. This paper tried to extend Prescott's test and Schouten's method to trials with missing data by using multiple imputations (MI). We proposed to impute unobserved values by using logistic regression method. Simulation studies were conducted to exam the efficiency of the tests with MI under different missing mechanisms.