A Correlated Logistic Regression Model for Population Pharmacodynamic Data

• Activity Number: 121
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 12:00 PM to 1:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): John Kwagyan*+ and Nnenna Kalu and Robert E. Taylor
• Companies: Howard University College of Medicine and Howard University College of Medicine and Howard University College of Medicine
• Keywords: pharmacodynamics ; correlated models ; alcohol sensitivity ; likelihood estimation

Abstract

Dichotomous type population pharmacodynamic data often arise in practice, especially in a clinical pharmacological setting. Data from subjects may consist of a multiple yes or no response to a defined stimulus at a given concentration and/or time. For example, in toxicity studies, the response will be whether the adverse effect occurs or not at a given concentration. In alcohol studies, the response will be whether the individual becomes intoxicated or stimulated at a given concentration or time. There is usually the tendency of certain individuals not responding to the stimulus at every concentration or time. In the study of acute tolerance to alcohol, greater than 50% of subjects reported no subjective intoxication effects at every assessment time. Consequently, in studying the effect of certain stimulus, data that is randomly sampled will lead to individuals that are largely devoid of the outcome of interest. The overall distribution of responsiveness in such data should appropriately be a mixture. We present an adaptation of the disposition model for correlated outcomes to dichotomous population pharmacodynamic data.