Before a drug can be marketed, an initial dose must be established. Sh
einer et al.(1) argue that a population approach leads to the most inf
ormed and rational decision making. We discuss the choice of an initia
l dose from both a predictive and estimative viewpoint. Our criteria a
re based upon evaluating the probabilities that a patient from the spe
cified population obtains a response that is at least of a specified s
ize. We demonstrate the approach using a simulation study and compare
estimation of population parameters and initial dose using Bayesian an
d likelihood-based methods. (C) 1997 by John Wiley & Sons, Ltd.