Abstract:
|
Missing data are frequently encountered in clinical trials across all phases and pose challenges for data analysis. Among Phase 1 studies, crossover studies with long treatment periods may be especially prone to dropouts and thus missing observations. Recent work suggests that the analysis method using the within-subject difference in period-specific baseline responses as a covariate could increase power for crossover studies (Mehrotra 2014). However, this method may potentially suffer even more from missing data issue, as the baseline measurements from both periods to be compared are required for analysis. Various approaches, including likelihood based methods and multiple imputation methods, have been proposed to handle missing data. However, few literatures were focused on missing data issue in crossover studies with relatively small sample sizes. In this presentation, we evaluate and compare the performance of several missing data analysis methods in crossover studies with baseline measurements, and relatively small sample sizes that are typically seen in Phase 1 studies. Simulation results are presented.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.