Analytical Methodologies for Clinical Endpoint Data with Excessive Zero

• Activity Number: 471
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Nancy Liu*+ and Amarjot Kaur and Jing Li and Ziliang Li
• Companies: Merck and Merck Research Laboratories and Merck and Merck
• Keywords: rescue medication ; zero-inflated log-normal ; overall treatment effect

Abstract

Zero-inflated normal or log-normal models have been introduced in literature for analyzing continuous data with high frequency of zeros and when the non-zero part of data approximately follow normal or log-normal distributions. However, the focus of available methods has primarily been on the estimations on the parameters of each separate model of the mixed-distribution. This article will discuss the estimation of the overall treatment effect along with its inferences entailing confidence intervals and p-values based on the zero-inflated log-normal model. The inferences based on zero-inflated normal model can be derived similarly. The zero-inflated log-normal model and the analysis of variance method will be compared using simulations and the rescue medication use data from an illustrative example study.