How Many Times Should We Assess an Efficacy Measurement?

• Activity Number: 34
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Bongin Yoo*+ and George Manos
• Companies: Bristol-Myers Squibb and Bristol-Myers Squibb
• Keywords: Linear mixed effects model ; longitudinal clinical trial ; missing data ; bias ; type I error ; type II error

Abstract

In a longitudinal clinical trial, the number and interval of efficacy measurement needs to be decided on a study-by-study basis. However, there has been little research showing the impact of limiting the number of repeated measures when a linear mixed effects model (LMM) is adopted. LMM's have been gaining a lot of popularity in the statistical community, even in the regulatory setting, due to their robustness and availability. The objective of this study is to investigate the impact of reduced number of assessments on LMM under different missing mechanisms with respect to any change on Type I and II errors and magnitude and/or direction of biases on treatment effects and their standard error estimates. Our simulations and real data analysis show that reducing the number of repeated measures in LMM approaches do not have much impact on bias of the treatment effects and preserving type I and II error rates under MCAR or balanced MAR. On the other hand, under MNAR and unbalanced MAR mechanism, reducing the number of repeated measures in LMM approaches results in substantial drawbacks such as noticeable inflation of type I errors and bias and fluctuation of type II errors.