Impact of Missing Data on the Analysis of Crossover Data

• Activity Number: 625
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Anjela Tzontcheva+ and Mahtab Marker*
• Companies: Merck & Co., Inc. and Merck & Co., Inc.
• Keywords: cross-over ; missing data ; simulations ; Phase I studies

Abstract

Cross-over designs are often used in Phase I studies. Since treatment comparisons are made using only the within-subject variability, these designs require smaller sample sizes compared to parallel designs. Often due to dropouts observations from the later periods are missing. Due to small sample size, it is important to use all information available in analysis. Discarding incomplete data could lead to biased results. In our talk we assess the impact of missing data when you carry out standard analysis of cross-over data in SAS Proc MIXED. The setting we consider is a 3 treatment-3 period design. The primary goal is treatment comparisons. Through extensive simulation studies we estimate the bias, efficiency, error rates and power with-without incomplete data. Different scenarios considered are: balanced/unbalanced, different error variability, presence/absence of carry-over effects.