Online Program Home
My Program

Abstract Details

Activity Number: 660
Type: Contributed
Date/Time: Thursday, August 4, 2016 : 8:30 AM to 10:20 AM
Sponsor: Biopharmaceutical Section
Abstract #321207 View Presentation
Title: Why the Likelihood Ratio Test Is Inappropriate to Test Dose-Response and Sharing of Information Among Antibody Therapies for Dose Selection
Author(s): Russell Reeve*
Companies: Quintiles
Keywords: Hill model ; Likelihood ratio test ; dose-response ; psoriasis
Abstract:

Suppose we are interested in estimating an appropriate dose to achieve a therapeutic level for an antibody NME for psoriasis. The question that arises: Can we use data on other compounds to improve the estimates of our NME? The dose-response curve is estimated using a Hill model. If the parameters of the Hill model can be shown to be shared among compounds, then using prior data would augment the data from the NME and improve the dose-response estimation. To do this, we need to test the hypothesis that the dose-response curve is non-flat. Since the null hypothesis is a subset of the parameter space of the alternative, one would expect that the likelihood ratio test (LRT) would be satisfactory. However, this is not the case. We describe why this is not the case, and find that the sampling distribution of the LRT can be expressed as a mixture of non-integral chi-squared random variables. We also show that the LRT does not behave even this well when applied to only summary data. We explore the distribution in this case as well. Finally, if we have time we will show that competitor dose-response curve provide useful information to help generate dose information on novel NMEs.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association