A Likelihood-Based Model Adequacy Test of the Four-Parameter Logistic Model for Data with No Replicate Response

• Activity Number: 233
• Type: Contributed
• Date/Time: Monday, August 3, 2009 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Wijith P. Munasinghe*+ and Kurex Sidik
• Companies: Kansas State University and Pfizer Global Research & Development
• Keywords: IOS test ; Dose-Response models ; Model adequacy

Abstract

The four-parameter nonlinear logistic (4PL) model is widely used in dose-response modeling in drug discovery and development. Assessing model adequacy of fitted 4PL model is important for making reliable inference about the model parameters, for example, inference for the EC50/ED50 parameter based on fitted model. However, when replicate responses at dose levels are not available, assessing model adequacy is problematic. The application of the general purpose goodness-of-fit IOS test (Presnell and Boos, 2004, JASA) is explored for assessing the 4PL model adequacy. Empirical power of the IOS test is investigated, it shows reasonable power for the cases of homogeneous error variance, and it indicates useful results for the cases of heterogeneous error variance. The test procedure provides a sound methodology to detect outliers in a data set, which also can cause for model inadequacy.