S. Walker, J.C. Wakefield
Imperial College, London, UK
In an important paper Sheiner et al discussed clinical trials design for dose ranging studies. It is the aim of this paper to discuss estimation methods for such studies. We carry out a Bayesian analysis and compare this with some of the alternatives described in the literature and in particular the package NONMEM. We prefer to compare such analyses not by parameter estimates but by the selection of an initial dose under well defined criteria on the grounds that these are the objectives of any dose ranging study. We argue that these criteria are best based upon the predictive distribution, for a random patient from the population under study, of achieving at least a specified response with a specified probability. We show that the Bayesian approach advocated leads to reasonably accurate initial dose predictions based upon these criteria whereas NONMEM can seriously underestimate the dose required.
Reference: PAGE 3 () Abstr 851 [www.page-meeting.org/?abstract=851]
Poster: poster