On the Calculation of Terminal Phase Half-life and Missing Not At Random

• Activity Number: 246
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Shu Zhang*+
• Companies: Eisai, Inc.
• Keywords: Half-life ; MNAR ; Sensitivity ; PMM ; Censor

Abstract

Terminal phase half-life (t1/2) is an important pharmacokinetic parameter characterizing a compound. Underestimating t1/2 can result in a shorter-than-necessary dosing interval, which in turn can cause undesired accumulation of the compound in the body. The population t1/2 is estimated by first estimating t1/2 for each subject, and then take the average of all the t1/2s to obtain the mean t1/2. However, in current practice, the estimated t1/2 for some subjects may be deemed unreliable based on certain "reliability criteria"; as a result, these t1/2s are discarded, rendering the t1/2 for those subjects a missing value. In this work, we examine the missingness mechanism and the overall impact of the missing values on the estimated parameters, and apply that concept to the case of estimating mean t1/2; we demonstrate that in some cases, excluding these "unreliable" t1/2s can result in the biasness of the mean t1/2. We propose some alternative methods to estimate the mean t1/2, in particular, use sensitivity analysis, Pattern Mixture Model (PMM) and censoring method. We implement our methods through simulation, and show the improvement of our methods over the "traditional" method.