Adrian Dunne
Department of Statistics, University College Dublin, Ireland.
In some clinical trials the pharmacodynamic response is measured on an ordinal categorical scale. This means that each measured response belongs in one of a number of categories which are ordered in some natural way. For example, mental health status may be categorised as impaired, moderate, mild or unimpaired. Conventional PK/PD modelling is not suitable for the description and analysis of such data. Models which take account of the measurement scale of the response are required. This presentation introduces the cumulative logit model for ordinal categorical response data with categorical and/or continuous explanatory variables. This modelling approach is incorporated into a population PK/PD modelling strategy using a hierarchical nonlinear model. The technique is applied to population PK/PD modelling of the results from a clinical trial of an analgesic agent in which the pharmacodynamic response was measured by means of pain relief scores over time in groups of patients randomly assigned to receive placebo or one of a number of doses of the analgesic following a pain inducing event.
Reference: PAGE 6 (1997) Abstr 590 [www.page-meeting.org/?abstract=590]
Poster: oral presentation